Wayne Altman, MD, FAAFP

ABSTRACT

Background: The Altman Rule, a simple tool for consumers seeking to make healthier packaged food choices at the point of sale, applies to packaged carbohydrates. According to the Altman Rule, a food is a healthier option if it has at least 3 g of fiber per serving and the grams of fiber plus the grams of protein exceed the grams of sugar per serving. This study sought to evaluate whether the Altman Rule is a valid proxy for glycemic load (GL).

Methods: We compared the binary outcome of whether a food item meets the Altman Rule with the GL of all foods categorized as cereals, chips, crackers, and granola bars in the Nutrition Data System for Research Database (University of Minnesota, Version 2010). We examined the percentage of foods in low-, medium-, and high-GL categories that met the Altman Rule.

Results: There were 1235 foods (342 cereals, 305 chips, 379 crackers, and 209 granola bars) in this analysis. There was a significant relationship between the GL of foods and the Altman Rule (P < .001) in that most low-GL (68%), almost half of medium-GL (48%), and very few high-GL (7%) foods met the criteria of the rule.

Conclusions: The Altman Rule is a reasonable proxy for GL and can be a useful and accessible tool for consumers interested in buying healthier packaged carbohydrate foods.

Nutrition can be complicated for consumers interested in making healthier choices at the grocery store. Consumers may have difficulty identifying more nutritious options, especially when food labels are adorned with claims such as “Good Source of Fiber” or “Heart Healthy.”¹ In addition, when reading food labels, consumers may find it difficult to decipher which data to prioritize when carbohydrates, total sugars, added sugars, total dietary fiber, soluble fiber, and insoluble fiber are all listed.

The concept of glycemic load (GL) is an important consideration, especially for people with diabetes. GL approximates the blood sugar response to different foods. A food with a high GL is digested quickly, and its carbohydrates are taken into the bloodstream rapidly. This leads to a spike and subsequent drop in blood sugars, which can cause symptoms of hyperglycemia and hypoglycemia in a person with diabetes.^2,3 Despite its usefulness, GL may be too complicated for a consumer to understand, and it does not appear anywhere on the food label. Since GL is calculated using pooled blood sugar response from individuals after the ingestion of the particular food, estimation of the GL is not intuitable.⁴

Point-of-sale tools. People seeking to lose weight, control diabetes, improve dyslipidemia and/or blood pressure, and/or decrease their risk for heart disease may benefit from point-of-sale tools such as the Altman Rule, which simplifies and encourages the selection of more nutritious foods.¹ Other tools—such as Guiding Stars (https://guidingstars.com), NuVal (www.nuval.com), and different variations of traffic lights—have been created to help consumers make more informed and healthier food choices.^5-8 However, Guiding Stars and NuVal are based on complicated algorithms that are not entirely transparent and not accessible to the average consumer.^6,7 Evaluations of these nutrition tools indicate that consumers tend to underrate the healthiness of some foods, such as raw almonds and salmon, and overrate the healthiness of others, such as fruit punch and diet soda, when using traffic light systems.⁶ Furthermore, these nutrition tools are not available in many supermarkets. Previous research suggests that the use of point-of-sale nutrition apps decreases with the time and effort involved in using an app.⁹

Continue to: The Altman Rule

The Altman Rule was developed by a family physician (author WA) to provide a more accessible tool for people interested in choosing healthier prepackaged carbohydrate foods while shopping. Since the user does not need to have a smartphone, and they are not required to download or understand an app for each purchase, the Altman Rule may be more usable compared with more complicated alternatives.

The Altman Rule can be used with nutrition labels that feature serving information and calories in enlarged and bold type, in compliance with the most recent US Food and Drug Administration (FDA) guideline from 2016. Many foods with high fiber also have high amounts of sugar, so the criteria of the Altman Rule includes a 2-step process requiring (1) a minimum of 3 g of total dietary fiber per serving and (2) the sum of the grams of fiber plus the grams of protein per serving to be greater than the total grams of sugar (not grams of added sugar or grams of carbohydrate) per serving (FIGURE 1A). Unlike the relatively complicated formula related to GL, this 2-part rule can be applied in seconds while shopping (FIGURE 1B).

The rule is intended only to be used for packaged carbohydrate products, such as bread, muffins, bagels, pasta, rice, oatmeal, cereals, snack bars, chips, and crackers. It does not apply to whole foods, such as meat, dairy, fruits, or vegetables. These foods are excluded to prevent any consumer confusion related to the nutritional content of whole foods (eg, an apple may have more sugar than fiber and protein combined, but it is still a nutritious option).

Since the user does not need to have a smartphone, the Altman Rule may be more usable compared with more complicated alternatives.

This study aimed to determine if the Altman Rule is a reasonable proxy for the more complicated concept of GL. We calculated the relationship between the GL of commercially available packaged carbohydrate foods and whether those foods met the Altman Rule.

METHODS

The Altman Rule was tested by comparing the binary outcome of the rule (meets/does not meet) with data on all foods categorized as cereals, chips, crackers, and granola bars in the Nutrition Data System for Research (NDSR) Database (University of Minnesota, Version 2010).

Continue to: To account for differences...

To account for differences in serving size, we used the standard of 50 g for each product as 1 serving. We used 50 g (about 1.7 oz) to help compare the different foods and between foods within the same group. Additionally, 50 g is close to 1 serving for most foods in these groups; it is about the size of a typical granola bar, three-quarters to 2 cups of cereal, 10 to 12 crackers, and 15 to 25 chips. We determined the GL for each product by multiplying the number of available carbohydrates (total carbohydrate – dietary fiber) by the product’s glycemic index/100. In general, GL is categorized as low (≤ 10), medium (11-19), or high (≥ 20).

We applied the Altman Rule to categorize each product as meeting or not meeting the rule. We compared the proportion of foods meeting the Altman Rule, stratified by GL and by specific foods, and used chi-square to determine if differences were statistically significant. These data were collected and analyzed in the summer of 2019.

RESULTS

There were 1235 foods (342 breakfast cereals, 305 chips, 379 crackers, and 209 granola bars) used for this analysis. There is a significant relationship between the GL of foods and the Altman Rule in that most low-GL (68%), almost half of medium-GL (48%), and only a few high-GL foods (7%) met the rule (P < .001) (TABLE 1). There was also a significant relationship between “meeting the ­Altman Rule” and GL within each food type (P < .001) (TABLE 2).

The medium-GL foods were the second largest category of foods we calculated; thus we further broke them into binary categories of low-medium GL (values 11-14) and high-medium GL (values 15-19) to explore the results of the Altman Rule. About half of the foods in medium-GL category met the Altman Rule. About eighty-five percent of the foods with low-medium GL passed the Altman Rule, while only 39% of the foods with high-medium GL did.

Foods that met the rule were more likely to be low GL and foods that did not pass the rule were more likely high GL. Within the medium-GL category, foods that met the rule were more likely to be low-medium GL. 

Continue to: The findings within food categories...

The findings within food categories showed that very few cereals, chips, crackers, and granola bars were low GL. For every food category, except granola bars, far more low-GL foods met the Altman Rule than those that did not. At the same time, very few high-GL foods met the Altman Rule. The category with the most individual high-GL food items meeting the Altman Rule was cereal. This was also the subcategory with the largest percentage of high-GL food items meeting the Altman Rule. Thirty-nine cereals that were high GL met the rule, but more than 4 times as many high-GL cereals did not (n = 190).

DISCUSSION

Marketing and nutrition messaging create consumer confusion that makes it challenging to identify packaged food items that are more nutrient dense. The Altman Rule simplifies food choices that have become unnecessarily complex. Our findings suggest this 2-step rule is a reasonable proxy for the more complicated and less accessible GL for packaged carbohydrates, such as cereals, chips, crackers, and snack bars. Foods that meet the rule are likely low or low-medium GL and thus are foods that are likely to be healthier choices.

Our findings suggest this 2-step rule is a reasonable proxy for the more complicated and less accessible glycemic load for packaged carbohydrates.

Of note, only 9% of chips (n = 27) passed the Altman Rule, likely due to their low dietary fiber content, which was typical of chips. If a food item does not have at least 3 grams of total dietary fiber per serving, it does not pass the Altman Rule, regardless of how much protein or sugar is in the product. This may be considered a strength or a weakness of the Altman Rule. Few nutrition-dense foods are low in fiber, but some foods could be nutritious but do not meet the Altman Rule due to having < 3 g of fiber.

With the high prevalence of chronic diseases such as hypertension, diabetes, hyperlipidemia, and cardiovascular disease, it is essential to help consumers prevent chronic disease altogether or manage their chronic disease by providing tools to identify healthier food choices. The tool also has a place in clinical medicine for use by physicians and other health care professionals. Research shows that physicians find both time and lack of ­knowledge/resources to be a barrier to providing nutritional counseling to patients.¹⁰ Since the Altman Rule can be shared and explained with very little time and without extensive nutritional knowledge, it meets these needs.

Limitations

Glycemic load. We acknowledge that the Altman Rule is not foolproof and that assessing this rule based on GL has some limitations. GL is not a perfect or comprehensive way to measure the nutritional value of a food. For example, fruits such as watermelon and grapes are nutritionally dense. However, they contain high amounts of natural sugars—and as such, their GL is relatively high, which could lead a consumer to perceive them as unhealthy. Nevertheless, GL is both a useful and accepted tool and a reasonable way to assess the validity of the rule, specifically when assessing packaged carbohydrates. The simplicity of the Altman Rule and its relationship with GL makes it such that consumers are more likely to make a healthier food choice using it.⁹

Continue to: Specificity and sensitivity

Specificity and sensitivity. There are other limitations to the Altman Rule, given that a small number of high-GL foods meet the rule. For example, some granola bars had high dietary protein, which offset a high sugar content just enough to pass the rule despite a higher GL. As such, concluding that a snack bar is a healthier choice because it meets the Altman Rule when it has high amounts of sugar may not be appropriate. This limitation could be considered a lack of specificity (the rule includes food it ought not to include). Another limitation to consider would be a lack of sensitivity, given that only 68% of low-GL foods passed the Altman Rule. Since GL is associated with carbohydrate content, foods with a low carbohydrate count often have little to no fiber and thus would fall into the category of foods that did not meet the Altman Rule but had low GL. In this case, however, the low amount of fiber may render the Altman Rule a better indicator of a healthier food choice than the GL.

Hidden sugars. Foods with sugar alcohols and artificial sweeteners may be as deleterious as caloric alternatives while not being accounted for when reporting the grams of sugar per serving on the nutrition label.⁷ This may represent an exception to the Altman Rule, as foods that are not healthier choices may pass the rule because the sugar content on the nutrition label is, in a sense, artificially lowered. Future research may investigate the hypothesis that these foods are nutritionally inferior despite meeting the Altman Rule.

The sample. Our study also was limited to working only with foods that were included in the NDSR database up to 2010. This limitation is mitigated by the fact that the sample size was large (> 1000 packaged food items were included in our analyses). The study also could be limited by the food categories that were analyzed; food categories such as bread, rice, pasta, and bagels were not included.

The objective of this research was to investigate the relationship between GL and the Altman Rule, rather than to conduct an exhaustive analysis of the Altman Rule for every possible food category. Studying the relationship between the Altman Rule and GL in other categories of food is an objective for future research. The data so far support a relationship between these entities. The likelihood of the nutrition facts of foods changing without the GL changing (or vice versa) is very low. As such, the Altman Rule still seems to be a reasonable proxy of GL.

CONCLUSIONS

Research indicates that point-of-sale tools, such as Guiding Stars, NuVal, and other stoplight tools, can successfully alter consumers’ behaviors.⁹ These tools can be helpful but are not available in many supermarkets. Despite the limitations, the Altman Rule is a useful decision aid that is accessible to all consumers no matter where they live or shop and is easy to use and remember.

The Altman rule can be used in clinical practice by health care professionals, such as physicians, nurse practitioners, physician assistants, dietitians, and health coaches. It also has the potential to be used in commercial settings, such as grocery stores, to help consumers easily identify healthier convenience foods. This has public health implications, as the rule can both empower consumers and potentially incentivize food manufacturers to upgrade their products nutritionally.

Additional research would be useful to evaluate consumers’ preferences and perceptions about how user-friendly the Altman Rule is at the point of sale with packaged carbohydrate foods. This would help to further understand how the use of information on food packaging can motivate healthier decisions—thereby helping to alleviate the burden of chronic disease.

CORRESPONDENCE
Kimberly R. Dong, DrPH, MS, RDN, Tufts University School of Medicine, Department of Public Health and Community Medicine, 136 Harrison Avenue, MV Building, Boston, MA 02111; kimberly.dong@tufts.edu

ABSTRACT

Background: The Altman Rule, a simple tool for consumers seeking to make healthier packaged food choices at the point of sale, applies to packaged carbohydrates. According to the Altman Rule, a food is a healthier option if it has at least 3 g of fiber per serving and the grams of fiber plus the grams of protein exceed the grams of sugar per serving. This study sought to evaluate whether the Altman Rule is a valid proxy for glycemic load (GL).

Methods: We compared the binary outcome of whether a food item meets the Altman Rule with the GL of all foods categorized as cereals, chips, crackers, and granola bars in the Nutrition Data System for Research Database (University of Minnesota, Version 2010). We examined the percentage of foods in low-, medium-, and high-GL categories that met the Altman Rule.

Results: There were 1235 foods (342 cereals, 305 chips, 379 crackers, and 209 granola bars) in this analysis. There was a significant relationship between the GL of foods and the Altman Rule (P < .001) in that most low-GL (68%), almost half of medium-GL (48%), and very few high-GL (7%) foods met the criteria of the rule.

Conclusions: The Altman Rule is a reasonable proxy for GL and can be a useful and accessible tool for consumers interested in buying healthier packaged carbohydrate foods.

Nutrition can be complicated for consumers interested in making healthier choices at the grocery store. Consumers may have difficulty identifying more nutritious options, especially when food labels are adorned with claims such as “Good Source of Fiber” or “Heart Healthy.”¹ In addition, when reading food labels, consumers may find it difficult to decipher which data to prioritize when carbohydrates, total sugars, added sugars, total dietary fiber, soluble fiber, and insoluble fiber are all listed.

The concept of glycemic load (GL) is an important consideration, especially for people with diabetes. GL approximates the blood sugar response to different foods. A food with a high GL is digested quickly, and its carbohydrates are taken into the bloodstream rapidly. This leads to a spike and subsequent drop in blood sugars, which can cause symptoms of hyperglycemia and hypoglycemia in a person with diabetes.^2,3 Despite its usefulness, GL may be too complicated for a consumer to understand, and it does not appear anywhere on the food label. Since GL is calculated using pooled blood sugar response from individuals after the ingestion of the particular food, estimation of the GL is not intuitable.⁴

Point-of-sale tools. People seeking to lose weight, control diabetes, improve dyslipidemia and/or blood pressure, and/or decrease their risk for heart disease may benefit from point-of-sale tools such as the Altman Rule, which simplifies and encourages the selection of more nutritious foods.¹ Other tools—such as Guiding Stars (https://guidingstars.com), NuVal (www.nuval.com), and different variations of traffic lights—have been created to help consumers make more informed and healthier food choices.^5-8 However, Guiding Stars and NuVal are based on complicated algorithms that are not entirely transparent and not accessible to the average consumer.^6,7 Evaluations of these nutrition tools indicate that consumers tend to underrate the healthiness of some foods, such as raw almonds and salmon, and overrate the healthiness of others, such as fruit punch and diet soda, when using traffic light systems.⁶ Furthermore, these nutrition tools are not available in many supermarkets. Previous research suggests that the use of point-of-sale nutrition apps decreases with the time and effort involved in using an app.⁹

Continue to: The Altman Rule

The Altman Rule was developed by a family physician (author WA) to provide a more accessible tool for people interested in choosing healthier prepackaged carbohydrate foods while shopping. Since the user does not need to have a smartphone, and they are not required to download or understand an app for each purchase, the Altman Rule may be more usable compared with more complicated alternatives.

The Altman Rule can be used with nutrition labels that feature serving information and calories in enlarged and bold type, in compliance with the most recent US Food and Drug Administration (FDA) guideline from 2016. Many foods with high fiber also have high amounts of sugar, so the criteria of the Altman Rule includes a 2-step process requiring (1) a minimum of 3 g of total dietary fiber per serving and (2) the sum of the grams of fiber plus the grams of protein per serving to be greater than the total grams of sugar (not grams of added sugar or grams of carbohydrate) per serving (FIGURE 1A). Unlike the relatively complicated formula related to GL, this 2-part rule can be applied in seconds while shopping (FIGURE 1B).

The rule is intended only to be used for packaged carbohydrate products, such as bread, muffins, bagels, pasta, rice, oatmeal, cereals, snack bars, chips, and crackers. It does not apply to whole foods, such as meat, dairy, fruits, or vegetables. These foods are excluded to prevent any consumer confusion related to the nutritional content of whole foods (eg, an apple may have more sugar than fiber and protein combined, but it is still a nutritious option).

Since the user does not need to have a smartphone, the Altman Rule may be more usable compared with more complicated alternatives.

This study aimed to determine if the Altman Rule is a reasonable proxy for the more complicated concept of GL. We calculated the relationship between the GL of commercially available packaged carbohydrate foods and whether those foods met the Altman Rule.

METHODS

The Altman Rule was tested by comparing the binary outcome of the rule (meets/does not meet) with data on all foods categorized as cereals, chips, crackers, and granola bars in the Nutrition Data System for Research (NDSR) Database (University of Minnesota, Version 2010).

Continue to: To account for differences...

To account for differences in serving size, we used the standard of 50 g for each product as 1 serving. We used 50 g (about 1.7 oz) to help compare the different foods and between foods within the same group. Additionally, 50 g is close to 1 serving for most foods in these groups; it is about the size of a typical granola bar, three-quarters to 2 cups of cereal, 10 to 12 crackers, and 15 to 25 chips. We determined the GL for each product by multiplying the number of available carbohydrates (total carbohydrate – dietary fiber) by the product’s glycemic index/100. In general, GL is categorized as low (≤ 10), medium (11-19), or high (≥ 20).

We applied the Altman Rule to categorize each product as meeting or not meeting the rule. We compared the proportion of foods meeting the Altman Rule, stratified by GL and by specific foods, and used chi-square to determine if differences were statistically significant. These data were collected and analyzed in the summer of 2019.

RESULTS

There were 1235 foods (342 breakfast cereals, 305 chips, 379 crackers, and 209 granola bars) used for this analysis. There is a significant relationship between the GL of foods and the Altman Rule in that most low-GL (68%), almost half of medium-GL (48%), and only a few high-GL foods (7%) met the rule (P < .001) (TABLE 1). There was also a significant relationship between “meeting the ­Altman Rule” and GL within each food type (P < .001) (TABLE 2).

The medium-GL foods were the second largest category of foods we calculated; thus we further broke them into binary categories of low-medium GL (values 11-14) and high-medium GL (values 15-19) to explore the results of the Altman Rule. About half of the foods in medium-GL category met the Altman Rule. About eighty-five percent of the foods with low-medium GL passed the Altman Rule, while only 39% of the foods with high-medium GL did.

Foods that met the rule were more likely to be low GL and foods that did not pass the rule were more likely high GL. Within the medium-GL category, foods that met the rule were more likely to be low-medium GL. 

Continue to: The findings within food categories...

The findings within food categories showed that very few cereals, chips, crackers, and granola bars were low GL. For every food category, except granola bars, far more low-GL foods met the Altman Rule than those that did not. At the same time, very few high-GL foods met the Altman Rule. The category with the most individual high-GL food items meeting the Altman Rule was cereal. This was also the subcategory with the largest percentage of high-GL food items meeting the Altman Rule. Thirty-nine cereals that were high GL met the rule, but more than 4 times as many high-GL cereals did not (n = 190).

DISCUSSION

Marketing and nutrition messaging create consumer confusion that makes it challenging to identify packaged food items that are more nutrient dense. The Altman Rule simplifies food choices that have become unnecessarily complex. Our findings suggest this 2-step rule is a reasonable proxy for the more complicated and less accessible GL for packaged carbohydrates, such as cereals, chips, crackers, and snack bars. Foods that meet the rule are likely low or low-medium GL and thus are foods that are likely to be healthier choices.

Our findings suggest this 2-step rule is a reasonable proxy for the more complicated and less accessible glycemic load for packaged carbohydrates.

Of note, only 9% of chips (n = 27) passed the Altman Rule, likely due to their low dietary fiber content, which was typical of chips. If a food item does not have at least 3 grams of total dietary fiber per serving, it does not pass the Altman Rule, regardless of how much protein or sugar is in the product. This may be considered a strength or a weakness of the Altman Rule. Few nutrition-dense foods are low in fiber, but some foods could be nutritious but do not meet the Altman Rule due to having < 3 g of fiber.

With the high prevalence of chronic diseases such as hypertension, diabetes, hyperlipidemia, and cardiovascular disease, it is essential to help consumers prevent chronic disease altogether or manage their chronic disease by providing tools to identify healthier food choices. The tool also has a place in clinical medicine for use by physicians and other health care professionals. Research shows that physicians find both time and lack of ­knowledge/resources to be a barrier to providing nutritional counseling to patients.¹⁰ Since the Altman Rule can be shared and explained with very little time and without extensive nutritional knowledge, it meets these needs.

Limitations

Glycemic load. We acknowledge that the Altman Rule is not foolproof and that assessing this rule based on GL has some limitations. GL is not a perfect or comprehensive way to measure the nutritional value of a food. For example, fruits such as watermelon and grapes are nutritionally dense. However, they contain high amounts of natural sugars—and as such, their GL is relatively high, which could lead a consumer to perceive them as unhealthy. Nevertheless, GL is both a useful and accepted tool and a reasonable way to assess the validity of the rule, specifically when assessing packaged carbohydrates. The simplicity of the Altman Rule and its relationship with GL makes it such that consumers are more likely to make a healthier food choice using it.⁹

Continue to: Specificity and sensitivity

Specificity and sensitivity. There are other limitations to the Altman Rule, given that a small number of high-GL foods meet the rule. For example, some granola bars had high dietary protein, which offset a high sugar content just enough to pass the rule despite a higher GL. As such, concluding that a snack bar is a healthier choice because it meets the Altman Rule when it has high amounts of sugar may not be appropriate. This limitation could be considered a lack of specificity (the rule includes food it ought not to include). Another limitation to consider would be a lack of sensitivity, given that only 68% of low-GL foods passed the Altman Rule. Since GL is associated with carbohydrate content, foods with a low carbohydrate count often have little to no fiber and thus would fall into the category of foods that did not meet the Altman Rule but had low GL. In this case, however, the low amount of fiber may render the Altman Rule a better indicator of a healthier food choice than the GL.

Hidden sugars. Foods with sugar alcohols and artificial sweeteners may be as deleterious as caloric alternatives while not being accounted for when reporting the grams of sugar per serving on the nutrition label.⁷ This may represent an exception to the Altman Rule, as foods that are not healthier choices may pass the rule because the sugar content on the nutrition label is, in a sense, artificially lowered. Future research may investigate the hypothesis that these foods are nutritionally inferior despite meeting the Altman Rule.

The sample. Our study also was limited to working only with foods that were included in the NDSR database up to 2010. This limitation is mitigated by the fact that the sample size was large (> 1000 packaged food items were included in our analyses). The study also could be limited by the food categories that were analyzed; food categories such as bread, rice, pasta, and bagels were not included.

The objective of this research was to investigate the relationship between GL and the Altman Rule, rather than to conduct an exhaustive analysis of the Altman Rule for every possible food category. Studying the relationship between the Altman Rule and GL in other categories of food is an objective for future research. The data so far support a relationship between these entities. The likelihood of the nutrition facts of foods changing without the GL changing (or vice versa) is very low. As such, the Altman Rule still seems to be a reasonable proxy of GL.

CONCLUSIONS

Research indicates that point-of-sale tools, such as Guiding Stars, NuVal, and other stoplight tools, can successfully alter consumers’ behaviors.⁹ These tools can be helpful but are not available in many supermarkets. Despite the limitations, the Altman Rule is a useful decision aid that is accessible to all consumers no matter where they live or shop and is easy to use and remember.

The Altman rule can be used in clinical practice by health care professionals, such as physicians, nurse practitioners, physician assistants, dietitians, and health coaches. It also has the potential to be used in commercial settings, such as grocery stores, to help consumers easily identify healthier convenience foods. This has public health implications, as the rule can both empower consumers and potentially incentivize food manufacturers to upgrade their products nutritionally.

Additional research would be useful to evaluate consumers’ preferences and perceptions about how user-friendly the Altman Rule is at the point of sale with packaged carbohydrate foods. This would help to further understand how the use of information on food packaging can motivate healthier decisions—thereby helping to alleviate the burden of chronic disease.

CORRESPONDENCE
Kimberly R. Dong, DrPH, MS, RDN, Tufts University School of Medicine, Department of Public Health and Community Medicine, 136 Harrison Avenue, MV Building, Boston, MA 02111; kimberly.dong@tufts.edu

ABSTRACT

Background: The Altman Rule, a simple tool for consumers seeking to make healthier packaged food choices at the point of sale, applies to packaged carbohydrates. According to the Altman Rule, a food is a healthier option if it has at least 3 g of fiber per serving and the grams of fiber plus the grams of protein exceed the grams of sugar per serving. This study sought to evaluate whether the Altman Rule is a valid proxy for glycemic load (GL).

Methods: We compared the binary outcome of whether a food item meets the Altman Rule with the GL of all foods categorized as cereals, chips, crackers, and granola bars in the Nutrition Data System for Research Database (University of Minnesota, Version 2010). We examined the percentage of foods in low-, medium-, and high-GL categories that met the Altman Rule.

Results: There were 1235 foods (342 cereals, 305 chips, 379 crackers, and 209 granola bars) in this analysis. There was a significant relationship between the GL of foods and the Altman Rule (P < .001) in that most low-GL (68%), almost half of medium-GL (48%), and very few high-GL (7%) foods met the criteria of the rule.

Conclusions: The Altman Rule is a reasonable proxy for GL and can be a useful and accessible tool for consumers interested in buying healthier packaged carbohydrate foods.

Nutrition can be complicated for consumers interested in making healthier choices at the grocery store. Consumers may have difficulty identifying more nutritious options, especially when food labels are adorned with claims such as “Good Source of Fiber” or “Heart Healthy.”¹ In addition, when reading food labels, consumers may find it difficult to decipher which data to prioritize when carbohydrates, total sugars, added sugars, total dietary fiber, soluble fiber, and insoluble fiber are all listed.

The concept of glycemic load (GL) is an important consideration, especially for people with diabetes. GL approximates the blood sugar response to different foods. A food with a high GL is digested quickly, and its carbohydrates are taken into the bloodstream rapidly. This leads to a spike and subsequent drop in blood sugars, which can cause symptoms of hyperglycemia and hypoglycemia in a person with diabetes.^2,3 Despite its usefulness, GL may be too complicated for a consumer to understand, and it does not appear anywhere on the food label. Since GL is calculated using pooled blood sugar response from individuals after the ingestion of the particular food, estimation of the GL is not intuitable.⁴

Point-of-sale tools. People seeking to lose weight, control diabetes, improve dyslipidemia and/or blood pressure, and/or decrease their risk for heart disease may benefit from point-of-sale tools such as the Altman Rule, which simplifies and encourages the selection of more nutritious foods.¹ Other tools—such as Guiding Stars (https://guidingstars.com), NuVal (www.nuval.com), and different variations of traffic lights—have been created to help consumers make more informed and healthier food choices.^5-8 However, Guiding Stars and NuVal are based on complicated algorithms that are not entirely transparent and not accessible to the average consumer.^6,7 Evaluations of these nutrition tools indicate that consumers tend to underrate the healthiness of some foods, such as raw almonds and salmon, and overrate the healthiness of others, such as fruit punch and diet soda, when using traffic light systems.⁶ Furthermore, these nutrition tools are not available in many supermarkets. Previous research suggests that the use of point-of-sale nutrition apps decreases with the time and effort involved in using an app.⁹

Continue to: The Altman Rule

The Altman Rule was developed by a family physician (author WA) to provide a more accessible tool for people interested in choosing healthier prepackaged carbohydrate foods while shopping. Since the user does not need to have a smartphone, and they are not required to download or understand an app for each purchase, the Altman Rule may be more usable compared with more complicated alternatives.

The Altman Rule can be used with nutrition labels that feature serving information and calories in enlarged and bold type, in compliance with the most recent US Food and Drug Administration (FDA) guideline from 2016. Many foods with high fiber also have high amounts of sugar, so the criteria of the Altman Rule includes a 2-step process requiring (1) a minimum of 3 g of total dietary fiber per serving and (2) the sum of the grams of fiber plus the grams of protein per serving to be greater than the total grams of sugar (not grams of added sugar or grams of carbohydrate) per serving (FIGURE 1A). Unlike the relatively complicated formula related to GL, this 2-part rule can be applied in seconds while shopping (FIGURE 1B).

The rule is intended only to be used for packaged carbohydrate products, such as bread, muffins, bagels, pasta, rice, oatmeal, cereals, snack bars, chips, and crackers. It does not apply to whole foods, such as meat, dairy, fruits, or vegetables. These foods are excluded to prevent any consumer confusion related to the nutritional content of whole foods (eg, an apple may have more sugar than fiber and protein combined, but it is still a nutritious option).

Since the user does not need to have a smartphone, the Altman Rule may be more usable compared with more complicated alternatives.

This study aimed to determine if the Altman Rule is a reasonable proxy for the more complicated concept of GL. We calculated the relationship between the GL of commercially available packaged carbohydrate foods and whether those foods met the Altman Rule.

METHODS

The Altman Rule was tested by comparing the binary outcome of the rule (meets/does not meet) with data on all foods categorized as cereals, chips, crackers, and granola bars in the Nutrition Data System for Research (NDSR) Database (University of Minnesota, Version 2010).

Continue to: To account for differences...

To account for differences in serving size, we used the standard of 50 g for each product as 1 serving. We used 50 g (about 1.7 oz) to help compare the different foods and between foods within the same group. Additionally, 50 g is close to 1 serving for most foods in these groups; it is about the size of a typical granola bar, three-quarters to 2 cups of cereal, 10 to 12 crackers, and 15 to 25 chips. We determined the GL for each product by multiplying the number of available carbohydrates (total carbohydrate – dietary fiber) by the product’s glycemic index/100. In general, GL is categorized as low (≤ 10), medium (11-19), or high (≥ 20).

We applied the Altman Rule to categorize each product as meeting or not meeting the rule. We compared the proportion of foods meeting the Altman Rule, stratified by GL and by specific foods, and used chi-square to determine if differences were statistically significant. These data were collected and analyzed in the summer of 2019.

RESULTS

There were 1235 foods (342 breakfast cereals, 305 chips, 379 crackers, and 209 granola bars) used for this analysis. There is a significant relationship between the GL of foods and the Altman Rule in that most low-GL (68%), almost half of medium-GL (48%), and only a few high-GL foods (7%) met the rule (P < .001) (TABLE 1). There was also a significant relationship between “meeting the ­Altman Rule” and GL within each food type (P < .001) (TABLE 2).

The medium-GL foods were the second largest category of foods we calculated; thus we further broke them into binary categories of low-medium GL (values 11-14) and high-medium GL (values 15-19) to explore the results of the Altman Rule. About half of the foods in medium-GL category met the Altman Rule. About eighty-five percent of the foods with low-medium GL passed the Altman Rule, while only 39% of the foods with high-medium GL did.

Foods that met the rule were more likely to be low GL and foods that did not pass the rule were more likely high GL. Within the medium-GL category, foods that met the rule were more likely to be low-medium GL. 

Continue to: The findings within food categories...

The findings within food categories showed that very few cereals, chips, crackers, and granola bars were low GL. For every food category, except granola bars, far more low-GL foods met the Altman Rule than those that did not. At the same time, very few high-GL foods met the Altman Rule. The category with the most individual high-GL food items meeting the Altman Rule was cereal. This was also the subcategory with the largest percentage of high-GL food items meeting the Altman Rule. Thirty-nine cereals that were high GL met the rule, but more than 4 times as many high-GL cereals did not (n = 190).

DISCUSSION

Marketing and nutrition messaging create consumer confusion that makes it challenging to identify packaged food items that are more nutrient dense. The Altman Rule simplifies food choices that have become unnecessarily complex. Our findings suggest this 2-step rule is a reasonable proxy for the more complicated and less accessible GL for packaged carbohydrates, such as cereals, chips, crackers, and snack bars. Foods that meet the rule are likely low or low-medium GL and thus are foods that are likely to be healthier choices.

Our findings suggest this 2-step rule is a reasonable proxy for the more complicated and less accessible glycemic load for packaged carbohydrates.

Of note, only 9% of chips (n = 27) passed the Altman Rule, likely due to their low dietary fiber content, which was typical of chips. If a food item does not have at least 3 grams of total dietary fiber per serving, it does not pass the Altman Rule, regardless of how much protein or sugar is in the product. This may be considered a strength or a weakness of the Altman Rule. Few nutrition-dense foods are low in fiber, but some foods could be nutritious but do not meet the Altman Rule due to having < 3 g of fiber.

With the high prevalence of chronic diseases such as hypertension, diabetes, hyperlipidemia, and cardiovascular disease, it is essential to help consumers prevent chronic disease altogether or manage their chronic disease by providing tools to identify healthier food choices. The tool also has a place in clinical medicine for use by physicians and other health care professionals. Research shows that physicians find both time and lack of ­knowledge/resources to be a barrier to providing nutritional counseling to patients.¹⁰ Since the Altman Rule can be shared and explained with very little time and without extensive nutritional knowledge, it meets these needs.

Limitations

Glycemic load. We acknowledge that the Altman Rule is not foolproof and that assessing this rule based on GL has some limitations. GL is not a perfect or comprehensive way to measure the nutritional value of a food. For example, fruits such as watermelon and grapes are nutritionally dense. However, they contain high amounts of natural sugars—and as such, their GL is relatively high, which could lead a consumer to perceive them as unhealthy. Nevertheless, GL is both a useful and accepted tool and a reasonable way to assess the validity of the rule, specifically when assessing packaged carbohydrates. The simplicity of the Altman Rule and its relationship with GL makes it such that consumers are more likely to make a healthier food choice using it.⁹

Continue to: Specificity and sensitivity

Specificity and sensitivity. There are other limitations to the Altman Rule, given that a small number of high-GL foods meet the rule. For example, some granola bars had high dietary protein, which offset a high sugar content just enough to pass the rule despite a higher GL. As such, concluding that a snack bar is a healthier choice because it meets the Altman Rule when it has high amounts of sugar may not be appropriate. This limitation could be considered a lack of specificity (the rule includes food it ought not to include). Another limitation to consider would be a lack of sensitivity, given that only 68% of low-GL foods passed the Altman Rule. Since GL is associated with carbohydrate content, foods with a low carbohydrate count often have little to no fiber and thus would fall into the category of foods that did not meet the Altman Rule but had low GL. In this case, however, the low amount of fiber may render the Altman Rule a better indicator of a healthier food choice than the GL.

Hidden sugars. Foods with sugar alcohols and artificial sweeteners may be as deleterious as caloric alternatives while not being accounted for when reporting the grams of sugar per serving on the nutrition label.⁷ This may represent an exception to the Altman Rule, as foods that are not healthier choices may pass the rule because the sugar content on the nutrition label is, in a sense, artificially lowered. Future research may investigate the hypothesis that these foods are nutritionally inferior despite meeting the Altman Rule.

The sample. Our study also was limited to working only with foods that were included in the NDSR database up to 2010. This limitation is mitigated by the fact that the sample size was large (> 1000 packaged food items were included in our analyses). The study also could be limited by the food categories that were analyzed; food categories such as bread, rice, pasta, and bagels were not included.

The objective of this research was to investigate the relationship between GL and the Altman Rule, rather than to conduct an exhaustive analysis of the Altman Rule for every possible food category. Studying the relationship between the Altman Rule and GL in other categories of food is an objective for future research. The data so far support a relationship between these entities. The likelihood of the nutrition facts of foods changing without the GL changing (or vice versa) is very low. As such, the Altman Rule still seems to be a reasonable proxy of GL.

CONCLUSIONS

Research indicates that point-of-sale tools, such as Guiding Stars, NuVal, and other stoplight tools, can successfully alter consumers’ behaviors.⁹ These tools can be helpful but are not available in many supermarkets. Despite the limitations, the Altman Rule is a useful decision aid that is accessible to all consumers no matter where they live or shop and is easy to use and remember.

The Altman rule can be used in clinical practice by health care professionals, such as physicians, nurse practitioners, physician assistants, dietitians, and health coaches. It also has the potential to be used in commercial settings, such as grocery stores, to help consumers easily identify healthier convenience foods. This has public health implications, as the rule can both empower consumers and potentially incentivize food manufacturers to upgrade their products nutritionally.

Additional research would be useful to evaluate consumers’ preferences and perceptions about how user-friendly the Altman Rule is at the point of sale with packaged carbohydrate foods. This would help to further understand how the use of information on food packaging can motivate healthier decisions—thereby helping to alleviate the burden of chronic disease.

CORRESPONDENCE
Kimberly R. Dong, DrPH, MS, RDN, Tufts University School of Medicine, Department of Public Health and Community Medicine, 136 Harrison Avenue, MV Building, Boston, MA 02111; kimberly.dong@tufts.edu

ABSTRACT

Purpose Medical scribes are known to increase revenue by increasing visits to a medical practice. We examined whether medical scribes are associated with markers of financial benefit independent of increased visits.

Methods We conducted a pre- and post-­observational study with a control group, examining changes in the percentage of visits (1) coded as level of service 4 or 5, (2) with at least 1 hierarchical condition category code billed, and (3) at which orders for 3 pay-for-performance quality measures (screening for breast, cervical, and colon cancer) were placed, if due. We looked at changes in outcomes among scribed providers and compared them to nonscribed providers. We used generalized estimating equations with robust standard errors to account for repeated measures and the hierarchical nature of the data, controlling for patient demographics.

Results We examined 41,371 visits to 17 scribed providers and 230,297 visits to 78 nonscribed providers. In adjusted analyses, and compared to nonscribed providers, scribes were associated with an increase of:

• Frutiger LT Std9.2 percentage points in level-of-­service 4 or 5 billing (Frutiger LT StdP < .001)
• 3.6 percentage points in hierarchical condition category coding (Frutiger LT StdP < .001)
• 4.0 percentage points in breast cancer screening orders (Frutiger LT StdP = .01)
• 4.9 percentage points in colon cancer screening orders (Frutiger LT StdPFrutiger LT Std = .04).

Conclusions This study suggests that scribes are associated with financial benefit in addition to increased visit volume. Primary care practices should consider the financial benefit of scribes independent of their ability to add patient volume.

Increasingly, medical scribes are used in ambulatory care settings across the United States.¹ Scribes are trained personnel who accompany providers during visits to provide documentation support and assist with other administrative tasks. They are associated with reduced documentation time for providers^2-6 and improved provider satisfaction,^7-11 without detriment to^4-16 (or with possible improvement in^17-20) patient satisfaction in ambulatory care settings. At the same time, concerns remain that using scribes might inhibit patient communication, harm clinical reasoning, reduce the effectiveness of clinical decision-support tools, and simply serve as a work-around to fixing inefficiencies in the electronic medical record (EMR).^21-23

A driving force for the increased use of medical scribes is the expectation that they reduce the cost of providing care. Cost-­efficiency is typically described as resulting from a reduction in physician time per patient seen, which allows increased patient volume and, in turn, drives increased physician productivity.^{2,8,10,18,24-27} Whether scribes result in cost savings remains unclear; some papers suggest that scribes are cost efficient in ambulatory care,^2,10,18 while others have been unable to identify cost savings, particularly in primary care.⁴

One reason why scribes might not be associated with cost savings is that their financial benefit might be undercounted. Studies that focus on increased volume miss the opportunity to capture financial benefits conferred through mechanisms that are independent of seeing more patients:

• Scribes might help providers address and document more, and more complex, medical problems, allowing higher level-of-service (LOS) billing. For example, a provider chooses a lower LOS because they have insufficiently documented a visit to support a higher LOS; by assisting with documentation, the scribe might allow the provider to choose a higher LOS.
• Scribes might prompt a provider to use decision-support tools for risk coding (using appropriate medical codes to capture the patient’s level of medical complexity), thereby increasing reimbursement.
• Scribes might extend the time available during the visit for the provider to address pay-for-performance quality measures, such as cancer screening.

Continue to: Making visits count, not counting visits

Making visits count, not counting visits. In this study, we examined whether medical scribes in primary care are associated with improved markers of revenue that are independent of seeing more patients. Specifically, we examined whether scribes are associated with increased LOS coding, risk coding, and orders for pay-for-performance measures for all primary care visits and for nonpreventive primary care visits.

Methods

Design

This observational study compared the change in outcomes before implementation of scribes and during implementation of scribes, between scribed providers and nonscribed providers. We compared visits during the year prior to the implementation of scribes (July 2017–June 2018) with the year during their implementation (July 2018–2019).

The Cambridge Health Alliance Institutional Review Board considered this study exempt from review.

Setting

This study was conducted at a safety-net community academic health system that uses an EMR developed by Epic Systems [Verona, WI]. This EMR includes decision-support tools that prompt providers when pay-for-performance quality measures are due and when hierarchical condition category (HCC) codes—ie, specific diagnoses used by Medicare and other payers to reimburse providers for the complexity of their patients—might apply to the visit.

Primary care practices should consider the financial benefit of scribes, independent of their ability to add patient volume.

These EMR decision-support tools use algorithms that draw on age, gender, diagnoses that were billed previously or are on the problem list, laboratory findings, and prior imaging. They alert physicians when a patient is due for pay-for-performance quality measures, such as cancer screenings, and when HCC codes might be applicable.

Continue to: During the study period...

During the study period, the EMR ­decision-support tool for HCC coding underwent several changes designed to improve HCC coding. In addition, systematic changes to primary care visits took place, leading to an increase in the number of patients seen and screenings required.

Outcomes

We examined 2 categories of outcomes that confer financial benefit to many institutions: billing measures and pay-for-performance measures.

Billing measures included the percentage of visits (1) coded as LOS 4 or 5 and (2) with at least 1 HCC code billed (among those for which the decision-support tool identified at least 1 potential HCC code).

Pay-for-performance measures. We examined whether any of 3 pay-for-­performance quality measures were addressed during the visit, selecting 3 that are commonly addressed by primary care providers (PCPs) and that require PCPs to sign an order for screening during a primary care visit: breast cancer (mammography order), cervical cancer (Papanicolaou smear order), and colon cancer (an order for fecal occult blood testing or colonoscopy).

Intervention

Scribes were employees of Cambridge Health Alliance who had recently graduated from college and were interested in a career as a health care professional. Scribes received 3 days of training on how to function effectively in their role; 1 day of training in EMR functionality; and 2 hours of training on decision-support tools for pay-for-performance quality measures and risk coding. Scribes continued learning on the job through feedback from supervising PCPs. Scribes documented patient encounters, recording histories and findings on the physical exam and transcribing discussion of treatment plans and the PCP’s instructions to patients.

Continue to: The 14 scribes worked with 17 physician...

The 14 scribes worked with 17 physician and nurse practitioner PCPs beginning in July 2018. Participation by PCPs was voluntary; they received no compensation for participating in the scribe program. PCPs were not required to see additional patients to participate. PCPs who chose to work with a scribe were similar to those who declined a scribe, as regards gender, race, type of provider (MD or NP), tenure at the institution, and percentage of time in clinical work (see Table W-1).

The control group comprised providers who elected not to work with a scribe but who worked in the same clinics as the intervention providers.

Scribes might extend the time available during the visit for the provider to address pay-forperformance quality measures, such as cancer screening.

Scribes were assigned to a PCP based on availability during the PCP’s scheduled hours and worked with 1 PCP throughout the intervention (except for 1 PCP who worked with 2 scribes). All PCPs worked with their scribe(s) part time; on average, 49% of intervention PCPs’ visits were scribed.

Inclusion and exclusion criteria

Because the first year at an institution is a learning period for PCPs, we excluded those who worked at the institution for < 1 year before the start of the scribe program (n = 12). Based on the extensive clinical experience of 1 PCP (WA) with scribes, we excluded the first 200 visits or 6 weeks (whichever occurred first) with a scribe among all scribed providers, to account for an initial learning period (n = 2202, of 15,372 scribed visits [14%]). We also excluded 2 providers who left during the pre-intervention period or were in the intervention period for < 1 month.

To ensure that we captured visits to providers with clinically significant exposure to scribes, we required scribed providers to have ≥ 20% of their visits scribed during the intervention period. To minimize the potential for contamination, we excluded nonscribed visits to scribed providers during the intervention period (n = 2211), because such nonscribed visits were largely due to visits outside the scribe’s scheduled time.

Continue to: Analysis

We compared demographic characteristics for patients and providers using the chi-square test for categorical variables and the t test for continuous variables. We compared the change in outcomes from before implementation of scribes to during implementation of scribes among scribed providers, compared to nonscribed providers, using generalized estimating equations with robust standard errors to account for repeated measures (ie, multiple visits by the same patients) and the hierarchical nature of the data (ie, patients nested within providers). We then recalculated these estimates, controlling for patient demographics (age, gender, race, and ethnicity). We repeated these analyses for patients presenting for nonpreventive visits.

Results

Visit characteristics

We examined 271,768 visits, including 41,371 visits to 17 scribed providers and 230,397 visits to 78 nonscribed providers (Table 1). Patients were most likely to be female, > 21 years of age, have English as their language of care, and be non-White. Most visits were by established patients and were nonpreventive.

We noted no clinically significant differences in characteristics between visits with scribed providers and visits with nonscribed providers, and over time. Patient complexity measures, including care management enrollment and hospital admissions, were also similar between groups, and over time.

Billing measures

HCC coding. In 28.6% of visits, the decision-support tool identified at least 1 potential HCC code. Among these, the percentage of visits with at least 1 HCC code billed increased by 10.1 percentage points (from 3.9% before implementation of scribes to 14.0%) among scribed providers, compared to increasing by 6.5 percentage points (from 2.9% before implementation to 9.3%) among nonscribed providers (TABLE 2). Scribes were therefore associated with an additional 3.6 percentage-point increase in visits with at least 1 HCC code billed (P < .0001)—a difference that remained significant after adjusting for patient demographics (P < .0001).

LOS coding. Scribed providers increased the number of visits billed as LOS 4 or 5 by 9.6 percentage points (from 47.3% before implementation to 56.8%); during the same period, nonscribed providers increased the number of visits billed as LOS 4 or 5 by 1.3 percentage points (from 46.5% before implementation to 47.8%) (TABLE 2). Scribes were therefore associated with an additional 8.3 percentage points in LOS 4 or 5 billing (P < .001) (TABLE 2). This difference remained significant after adjusting for patient demographics (P < .001).

Continue to: Pay-for-performance quality measures

Breast cancer screening. Scribed providers increased the number of visits at which breast cancer screening was ordered by 2.7 percentage points (from 17.3% before implementation of scribes to 20.0%); during the same period, the number of visits at which breast cancer screening was ordered by nonscribed providers decreased by 1.9 percentage points (from 19.5% to 17.6%). Scribes were therefore associated with an increase of 4.6 percentage points in breast cancer screening orders, compared to nonscribed providers (P < .003) (TABLE 2). That difference remained significant after adjusting for patient demographics (P = .01).

Colon cancer screening. Similarly, scribed providers increased the number of visits at which colon cancer screening was ordered by 1.2 percentage points (from 19.2% before implementation of scribes to 20.3%); during the same period, the number of visits at which colon cancer screening was ordered by nonscribed providers decreased by 2.7 percentage points (from 18.5% to 15.9%) (P = .112). After adjusting for patient demographics, scribes were associated with an increase of 4.9 percentage points in colon cancer screening orders, compared to nonscribed providers (P = .044) (TABLE 2).

Cervical cancer screening. The rate at which cervical cancer screening was ordered did not change among scribed providers and decreased (by 2.5 percentage points) among nonscribed providers—a difference that was not statistically significant (P = .26).

Nonpreventive visits. Our findings overall did not change in analyses focused on nonpreventive visits, in which scribes were associated with an increase of 8.2 percentage points in LOS 4 or 5 billing (P < .001); an increase of 3.1 percentage points in HCC coding (P < .001); and an increase of 3.2 percentage points in breast cancer screening orders (P = .03) (TABLE 3). Although scribes were associated with an increase of 1.5 percentage points in cervical cancer screening orders and an increase of 3.1 percentage points in colon cancer screening orders, these increases did not reach statistical significance.

Discussion

We found that implementation of scribes is associated with (1) an increase in LOS coding and risk coding and (2) a higher frequency of addressing 2 of 3 pay-for-performance quality measures in primary care. In adjusted analyses in our study, and compared to nonscribed providers, scribes were associated with an additional 9.2 percentage points in LOS 4 or 5 billing; 3.6 percentage points in HCC coding; 4.0 percentage points in breast cancer screening orders; and 4.9 percentage points in colon cancer screening orders. Cervical cancer screening orders followed a similar pattern, with an increase of 2.3 percentage points in the adjusted screening order rate among scribed providers, compared to nonscribed providers, during implementation of scribes—although the increase was not significant. These findings did not change in analyses focused on nonpreventive visits.

Continue to: Our findings are consistent

Our findings are consistent with those of earlier studies. Prior examinations in ambulatory specialties found that scribes increased HCC coding,⁴ LOS billing,²⁴ and pay-for-performance metrics.¹⁸ The only study to examine these areas in primary care found that scribes were associated with increased pay-for-performance measure documentation,²⁰ a change that is necessary but insufficient to realize increased pay-for-­performance revenue. Therefore, our study confirms, for the first time, that PCPs can better address pay-for-performance measures, LOS billing, and HCC coding when working with a scribe in primary care.

Providers who used a scribe during the intervention period of our study increased the number of visits billed as LOS 4 or 5 by 9.6 percentage points.

Demands on primary care visits are increasing.²⁸ Physicians are required to provide more documentation; there is greater emphasis on PCPs meeting pay-for-­performance measures; and there are more data in the EMR to review. In this context, addressing pay-for-performance measures and gaps in risk coding is likely to be increasingly challenging. Our study suggests that scribes might provide a mechanism to increase risk coding, LOS billing, and pay-for-performance measures, despite increased demands on primary care visits.

Increase in LOS billing. In the settings in which we work, a fee-for-service LOS 4 primary care visit generates, on average, $20 to $75 more in revenue than an LOS 3 visit. Using an average of $50 additional revenue for LOS 4 billing, we estimate that a full-time scribe is associated with roughly $7,000 in additional revenue annually. We arrived at this estimate using an average of 1500 visits at LOS ≤ 3 for every PCP full-time equivalent. A 9.2 percentage–point increase in LOS 4 billing would lead to roughly 140 additional LOS 4 visits, with each visit generating an additional $50 in revenue.

This analysis does not account for increased revenue associated with increased pay for HCC coding identified in our study.

Our study also suggests that scribes might provide a mechanism to increase risk coding and payfor-performance, despite increased demands on primary care visits.

Furthermore, in our conservative assumption, the entire increase in LOS billing was from level 3 to level 4; in fact, a small percentage of that increase would be from level 2 and another small percentage would be to level 5—both of which would generate additional revenue. Our assumption therefore underestimates the full financial value associated with scribes in the absence of increased patient volume. Nonetheless, the assumption suggests that increases in LOS billing offset a substantial percentage of a scribe’s salary.

Continue to: Limitations of this study

Limitations of this study. Our study should be interpreted in the context of several limitations:

• The study was conducted at 1 institution. Our findings might not be generalizable beyond this setting.
• The study measures the impact of scribes when providers work with scribes part time. Because providers who utilize a scribe for all, or nearly all, their visits are likely to use a scribe more efficiently, our study might underestimate the full impact of a scribe.
• In some settings, team members such as medical assistants are trained to assist with documentation and other responsibilities (such as closing care gaps) in addition to other patient care responsibilities.^29-32 The extent to which our findings transfer to other models is unclear; studies comparing the impact of other models (which might provide even stronger outcomes) to the impact of medical scribes would be an interesting area for further research.
• In addition to the variability of models, there is likely variability in the quality and interactions of medical scribes, which might impact outcomes. We did not examine the qualities of scribes that led to outcomes in this study.
• We examined the impact of scribes on quality measure–ordering behaviors of providers, not on whether quality measures actually improved. Because scribes are associated with more face-to-face time with patients,²⁷ they might allow for increased attention being paid by physicians to barriers to pay-for-performance measures (eg, patient education). This could increase the likelihood that patients complete a multitude of screenings, and thus improve adherence and follow-up. However, the impact of scribes on quality measures is a topic for future study.

Value beyond volume. Any limitations notwithstanding, our study suggests that scribes are associated with financial benefit in addition to the benefit of increased volume. Primary care practices should therefore consider the financial benefit of scribes independent of their ability to add patient volume. By recognizing this additive value, primary care practices might more fully capture the benefit of scribes, which might then allow practices to employ scribes with less demand to increase volume. This added support without increased volume would, in turn, likely reduce provider burnout (and the costly associated turnover) and increase patient satisfaction, leading to a synergistic financial benefit.

CORRESPONDENCE
Wayne Altman, MD, FAAFP, Tufts University School of Medicine, 200 Harrison Avenue, Boston, MA 02111; wayne. altman@tufts.edu

ABSTRACT

Purpose Medical scribes are known to increase revenue by increasing visits to a medical practice. We examined whether medical scribes are associated with markers of financial benefit independent of increased visits.

Methods We conducted a pre- and post-­observational study with a control group, examining changes in the percentage of visits (1) coded as level of service 4 or 5, (2) with at least 1 hierarchical condition category code billed, and (3) at which orders for 3 pay-for-performance quality measures (screening for breast, cervical, and colon cancer) were placed, if due. We looked at changes in outcomes among scribed providers and compared them to nonscribed providers. We used generalized estimating equations with robust standard errors to account for repeated measures and the hierarchical nature of the data, controlling for patient demographics.

Results We examined 41,371 visits to 17 scribed providers and 230,297 visits to 78 nonscribed providers. In adjusted analyses, and compared to nonscribed providers, scribes were associated with an increase of:

• Frutiger LT Std9.2 percentage points in level-of-­service 4 or 5 billing (Frutiger LT StdP < .001)
• 3.6 percentage points in hierarchical condition category coding (Frutiger LT StdP < .001)
• 4.0 percentage points in breast cancer screening orders (Frutiger LT StdP = .01)
• 4.9 percentage points in colon cancer screening orders (Frutiger LT StdPFrutiger LT Std = .04).

Conclusions This study suggests that scribes are associated with financial benefit in addition to increased visit volume. Primary care practices should consider the financial benefit of scribes independent of their ability to add patient volume.

Increasingly, medical scribes are used in ambulatory care settings across the United States.¹ Scribes are trained personnel who accompany providers during visits to provide documentation support and assist with other administrative tasks. They are associated with reduced documentation time for providers^2-6 and improved provider satisfaction,^7-11 without detriment to^4-16 (or with possible improvement in^17-20) patient satisfaction in ambulatory care settings. At the same time, concerns remain that using scribes might inhibit patient communication, harm clinical reasoning, reduce the effectiveness of clinical decision-support tools, and simply serve as a work-around to fixing inefficiencies in the electronic medical record (EMR).^21-23

A driving force for the increased use of medical scribes is the expectation that they reduce the cost of providing care. Cost-­efficiency is typically described as resulting from a reduction in physician time per patient seen, which allows increased patient volume and, in turn, drives increased physician productivity.^{2,8,10,18,24-27} Whether scribes result in cost savings remains unclear; some papers suggest that scribes are cost efficient in ambulatory care,^2,10,18 while others have been unable to identify cost savings, particularly in primary care.⁴

One reason why scribes might not be associated with cost savings is that their financial benefit might be undercounted. Studies that focus on increased volume miss the opportunity to capture financial benefits conferred through mechanisms that are independent of seeing more patients:

• Scribes might help providers address and document more, and more complex, medical problems, allowing higher level-of-service (LOS) billing. For example, a provider chooses a lower LOS because they have insufficiently documented a visit to support a higher LOS; by assisting with documentation, the scribe might allow the provider to choose a higher LOS.
• Scribes might prompt a provider to use decision-support tools for risk coding (using appropriate medical codes to capture the patient’s level of medical complexity), thereby increasing reimbursement.
• Scribes might extend the time available during the visit for the provider to address pay-for-performance quality measures, such as cancer screening.

Continue to: Making visits count, not counting visits

Making visits count, not counting visits. In this study, we examined whether medical scribes in primary care are associated with improved markers of revenue that are independent of seeing more patients. Specifically, we examined whether scribes are associated with increased LOS coding, risk coding, and orders for pay-for-performance measures for all primary care visits and for nonpreventive primary care visits.

Methods

Design

This observational study compared the change in outcomes before implementation of scribes and during implementation of scribes, between scribed providers and nonscribed providers. We compared visits during the year prior to the implementation of scribes (July 2017–June 2018) with the year during their implementation (July 2018–2019).

The Cambridge Health Alliance Institutional Review Board considered this study exempt from review.

Setting

This study was conducted at a safety-net community academic health system that uses an EMR developed by Epic Systems [Verona, WI]. This EMR includes decision-support tools that prompt providers when pay-for-performance quality measures are due and when hierarchical condition category (HCC) codes—ie, specific diagnoses used by Medicare and other payers to reimburse providers for the complexity of their patients—might apply to the visit.

Primary care practices should consider the financial benefit of scribes, independent of their ability to add patient volume.

These EMR decision-support tools use algorithms that draw on age, gender, diagnoses that were billed previously or are on the problem list, laboratory findings, and prior imaging. They alert physicians when a patient is due for pay-for-performance quality measures, such as cancer screenings, and when HCC codes might be applicable.

Continue to: During the study period...

During the study period, the EMR ­decision-support tool for HCC coding underwent several changes designed to improve HCC coding. In addition, systematic changes to primary care visits took place, leading to an increase in the number of patients seen and screenings required.

Outcomes

We examined 2 categories of outcomes that confer financial benefit to many institutions: billing measures and pay-for-performance measures.

Billing measures included the percentage of visits (1) coded as LOS 4 or 5 and (2) with at least 1 HCC code billed (among those for which the decision-support tool identified at least 1 potential HCC code).

Pay-for-performance measures. We examined whether any of 3 pay-for-­performance quality measures were addressed during the visit, selecting 3 that are commonly addressed by primary care providers (PCPs) and that require PCPs to sign an order for screening during a primary care visit: breast cancer (mammography order), cervical cancer (Papanicolaou smear order), and colon cancer (an order for fecal occult blood testing or colonoscopy).

Intervention

Scribes were employees of Cambridge Health Alliance who had recently graduated from college and were interested in a career as a health care professional. Scribes received 3 days of training on how to function effectively in their role; 1 day of training in EMR functionality; and 2 hours of training on decision-support tools for pay-for-performance quality measures and risk coding. Scribes continued learning on the job through feedback from supervising PCPs. Scribes documented patient encounters, recording histories and findings on the physical exam and transcribing discussion of treatment plans and the PCP’s instructions to patients.

Continue to: The 14 scribes worked with 17 physician...

The 14 scribes worked with 17 physician and nurse practitioner PCPs beginning in July 2018. Participation by PCPs was voluntary; they received no compensation for participating in the scribe program. PCPs were not required to see additional patients to participate. PCPs who chose to work with a scribe were similar to those who declined a scribe, as regards gender, race, type of provider (MD or NP), tenure at the institution, and percentage of time in clinical work (see Table W-1).

The control group comprised providers who elected not to work with a scribe but who worked in the same clinics as the intervention providers.

Scribes might extend the time available during the visit for the provider to address pay-forperformance quality measures, such as cancer screening.

Scribes were assigned to a PCP based on availability during the PCP’s scheduled hours and worked with 1 PCP throughout the intervention (except for 1 PCP who worked with 2 scribes). All PCPs worked with their scribe(s) part time; on average, 49% of intervention PCPs’ visits were scribed.

Inclusion and exclusion criteria

Because the first year at an institution is a learning period for PCPs, we excluded those who worked at the institution for < 1 year before the start of the scribe program (n = 12). Based on the extensive clinical experience of 1 PCP (WA) with scribes, we excluded the first 200 visits or 6 weeks (whichever occurred first) with a scribe among all scribed providers, to account for an initial learning period (n = 2202, of 15,372 scribed visits [14%]). We also excluded 2 providers who left during the pre-intervention period or were in the intervention period for < 1 month.

To ensure that we captured visits to providers with clinically significant exposure to scribes, we required scribed providers to have ≥ 20% of their visits scribed during the intervention period. To minimize the potential for contamination, we excluded nonscribed visits to scribed providers during the intervention period (n = 2211), because such nonscribed visits were largely due to visits outside the scribe’s scheduled time.

Continue to: Analysis

We compared demographic characteristics for patients and providers using the chi-square test for categorical variables and the t test for continuous variables. We compared the change in outcomes from before implementation of scribes to during implementation of scribes among scribed providers, compared to nonscribed providers, using generalized estimating equations with robust standard errors to account for repeated measures (ie, multiple visits by the same patients) and the hierarchical nature of the data (ie, patients nested within providers). We then recalculated these estimates, controlling for patient demographics (age, gender, race, and ethnicity). We repeated these analyses for patients presenting for nonpreventive visits.

Results

Visit characteristics

We examined 271,768 visits, including 41,371 visits to 17 scribed providers and 230,397 visits to 78 nonscribed providers (Table 1). Patients were most likely to be female, > 21 years of age, have English as their language of care, and be non-White. Most visits were by established patients and were nonpreventive.

We noted no clinically significant differences in characteristics between visits with scribed providers and visits with nonscribed providers, and over time. Patient complexity measures, including care management enrollment and hospital admissions, were also similar between groups, and over time.

Billing measures

HCC coding. In 28.6% of visits, the decision-support tool identified at least 1 potential HCC code. Among these, the percentage of visits with at least 1 HCC code billed increased by 10.1 percentage points (from 3.9% before implementation of scribes to 14.0%) among scribed providers, compared to increasing by 6.5 percentage points (from 2.9% before implementation to 9.3%) among nonscribed providers (TABLE 2). Scribes were therefore associated with an additional 3.6 percentage-point increase in visits with at least 1 HCC code billed (P < .0001)—a difference that remained significant after adjusting for patient demographics (P < .0001).

LOS coding. Scribed providers increased the number of visits billed as LOS 4 or 5 by 9.6 percentage points (from 47.3% before implementation to 56.8%); during the same period, nonscribed providers increased the number of visits billed as LOS 4 or 5 by 1.3 percentage points (from 46.5% before implementation to 47.8%) (TABLE 2). Scribes were therefore associated with an additional 8.3 percentage points in LOS 4 or 5 billing (P < .001) (TABLE 2). This difference remained significant after adjusting for patient demographics (P < .001).

Continue to: Pay-for-performance quality measures

Breast cancer screening. Scribed providers increased the number of visits at which breast cancer screening was ordered by 2.7 percentage points (from 17.3% before implementation of scribes to 20.0%); during the same period, the number of visits at which breast cancer screening was ordered by nonscribed providers decreased by 1.9 percentage points (from 19.5% to 17.6%). Scribes were therefore associated with an increase of 4.6 percentage points in breast cancer screening orders, compared to nonscribed providers (P < .003) (TABLE 2). That difference remained significant after adjusting for patient demographics (P = .01).

Colon cancer screening. Similarly, scribed providers increased the number of visits at which colon cancer screening was ordered by 1.2 percentage points (from 19.2% before implementation of scribes to 20.3%); during the same period, the number of visits at which colon cancer screening was ordered by nonscribed providers decreased by 2.7 percentage points (from 18.5% to 15.9%) (P = .112). After adjusting for patient demographics, scribes were associated with an increase of 4.9 percentage points in colon cancer screening orders, compared to nonscribed providers (P = .044) (TABLE 2).

Cervical cancer screening. The rate at which cervical cancer screening was ordered did not change among scribed providers and decreased (by 2.5 percentage points) among nonscribed providers—a difference that was not statistically significant (P = .26).

Nonpreventive visits. Our findings overall did not change in analyses focused on nonpreventive visits, in which scribes were associated with an increase of 8.2 percentage points in LOS 4 or 5 billing (P < .001); an increase of 3.1 percentage points in HCC coding (P < .001); and an increase of 3.2 percentage points in breast cancer screening orders (P = .03) (TABLE 3). Although scribes were associated with an increase of 1.5 percentage points in cervical cancer screening orders and an increase of 3.1 percentage points in colon cancer screening orders, these increases did not reach statistical significance.

Discussion

We found that implementation of scribes is associated with (1) an increase in LOS coding and risk coding and (2) a higher frequency of addressing 2 of 3 pay-for-performance quality measures in primary care. In adjusted analyses in our study, and compared to nonscribed providers, scribes were associated with an additional 9.2 percentage points in LOS 4 or 5 billing; 3.6 percentage points in HCC coding; 4.0 percentage points in breast cancer screening orders; and 4.9 percentage points in colon cancer screening orders. Cervical cancer screening orders followed a similar pattern, with an increase of 2.3 percentage points in the adjusted screening order rate among scribed providers, compared to nonscribed providers, during implementation of scribes—although the increase was not significant. These findings did not change in analyses focused on nonpreventive visits.

Continue to: Our findings are consistent

Our findings are consistent with those of earlier studies. Prior examinations in ambulatory specialties found that scribes increased HCC coding,⁴ LOS billing,²⁴ and pay-for-performance metrics.¹⁸ The only study to examine these areas in primary care found that scribes were associated with increased pay-for-performance measure documentation,²⁰ a change that is necessary but insufficient to realize increased pay-for-­performance revenue. Therefore, our study confirms, for the first time, that PCPs can better address pay-for-performance measures, LOS billing, and HCC coding when working with a scribe in primary care.

Providers who used a scribe during the intervention period of our study increased the number of visits billed as LOS 4 or 5 by 9.6 percentage points.

Demands on primary care visits are increasing.²⁸ Physicians are required to provide more documentation; there is greater emphasis on PCPs meeting pay-for-­performance measures; and there are more data in the EMR to review. In this context, addressing pay-for-performance measures and gaps in risk coding is likely to be increasingly challenging. Our study suggests that scribes might provide a mechanism to increase risk coding, LOS billing, and pay-for-performance measures, despite increased demands on primary care visits.

Increase in LOS billing. In the settings in which we work, a fee-for-service LOS 4 primary care visit generates, on average, $20 to $75 more in revenue than an LOS 3 visit. Using an average of $50 additional revenue for LOS 4 billing, we estimate that a full-time scribe is associated with roughly $7,000 in additional revenue annually. We arrived at this estimate using an average of 1500 visits at LOS ≤ 3 for every PCP full-time equivalent. A 9.2 percentage–point increase in LOS 4 billing would lead to roughly 140 additional LOS 4 visits, with each visit generating an additional $50 in revenue.

This analysis does not account for increased revenue associated with increased pay for HCC coding identified in our study.

Our study also suggests that scribes might provide a mechanism to increase risk coding and payfor-performance, despite increased demands on primary care visits.

Furthermore, in our conservative assumption, the entire increase in LOS billing was from level 3 to level 4; in fact, a small percentage of that increase would be from level 2 and another small percentage would be to level 5—both of which would generate additional revenue. Our assumption therefore underestimates the full financial value associated with scribes in the absence of increased patient volume. Nonetheless, the assumption suggests that increases in LOS billing offset a substantial percentage of a scribe’s salary.

Continue to: Limitations of this study

Limitations of this study. Our study should be interpreted in the context of several limitations:

• The study was conducted at 1 institution. Our findings might not be generalizable beyond this setting.
• The study measures the impact of scribes when providers work with scribes part time. Because providers who utilize a scribe for all, or nearly all, their visits are likely to use a scribe more efficiently, our study might underestimate the full impact of a scribe.
• In some settings, team members such as medical assistants are trained to assist with documentation and other responsibilities (such as closing care gaps) in addition to other patient care responsibilities.^29-32 The extent to which our findings transfer to other models is unclear; studies comparing the impact of other models (which might provide even stronger outcomes) to the impact of medical scribes would be an interesting area for further research.
• In addition to the variability of models, there is likely variability in the quality and interactions of medical scribes, which might impact outcomes. We did not examine the qualities of scribes that led to outcomes in this study.
• We examined the impact of scribes on quality measure–ordering behaviors of providers, not on whether quality measures actually improved. Because scribes are associated with more face-to-face time with patients,²⁷ they might allow for increased attention being paid by physicians to barriers to pay-for-performance measures (eg, patient education). This could increase the likelihood that patients complete a multitude of screenings, and thus improve adherence and follow-up. However, the impact of scribes on quality measures is a topic for future study.

Value beyond volume. Any limitations notwithstanding, our study suggests that scribes are associated with financial benefit in addition to the benefit of increased volume. Primary care practices should therefore consider the financial benefit of scribes independent of their ability to add patient volume. By recognizing this additive value, primary care practices might more fully capture the benefit of scribes, which might then allow practices to employ scribes with less demand to increase volume. This added support without increased volume would, in turn, likely reduce provider burnout (and the costly associated turnover) and increase patient satisfaction, leading to a synergistic financial benefit.

CORRESPONDENCE
Wayne Altman, MD, FAAFP, Tufts University School of Medicine, 200 Harrison Avenue, Boston, MA 02111; wayne. altman@tufts.edu

ABSTRACT

Purpose Medical scribes are known to increase revenue by increasing visits to a medical practice. We examined whether medical scribes are associated with markers of financial benefit independent of increased visits.

Methods We conducted a pre- and post-­observational study with a control group, examining changes in the percentage of visits (1) coded as level of service 4 or 5, (2) with at least 1 hierarchical condition category code billed, and (3) at which orders for 3 pay-for-performance quality measures (screening for breast, cervical, and colon cancer) were placed, if due. We looked at changes in outcomes among scribed providers and compared them to nonscribed providers. We used generalized estimating equations with robust standard errors to account for repeated measures and the hierarchical nature of the data, controlling for patient demographics.

Results We examined 41,371 visits to 17 scribed providers and 230,297 visits to 78 nonscribed providers. In adjusted analyses, and compared to nonscribed providers, scribes were associated with an increase of:

• Frutiger LT Std9.2 percentage points in level-of-­service 4 or 5 billing (Frutiger LT StdP < .001)
• 3.6 percentage points in hierarchical condition category coding (Frutiger LT StdP < .001)
• 4.0 percentage points in breast cancer screening orders (Frutiger LT StdP = .01)
• 4.9 percentage points in colon cancer screening orders (Frutiger LT StdPFrutiger LT Std = .04).

Conclusions This study suggests that scribes are associated with financial benefit in addition to increased visit volume. Primary care practices should consider the financial benefit of scribes independent of their ability to add patient volume.

Increasingly, medical scribes are used in ambulatory care settings across the United States.¹ Scribes are trained personnel who accompany providers during visits to provide documentation support and assist with other administrative tasks. They are associated with reduced documentation time for providers^2-6 and improved provider satisfaction,^7-11 without detriment to^4-16 (or with possible improvement in^17-20) patient satisfaction in ambulatory care settings. At the same time, concerns remain that using scribes might inhibit patient communication, harm clinical reasoning, reduce the effectiveness of clinical decision-support tools, and simply serve as a work-around to fixing inefficiencies in the electronic medical record (EMR).^21-23

A driving force for the increased use of medical scribes is the expectation that they reduce the cost of providing care. Cost-­efficiency is typically described as resulting from a reduction in physician time per patient seen, which allows increased patient volume and, in turn, drives increased physician productivity.^{2,8,10,18,24-27} Whether scribes result in cost savings remains unclear; some papers suggest that scribes are cost efficient in ambulatory care,^2,10,18 while others have been unable to identify cost savings, particularly in primary care.⁴

One reason why scribes might not be associated with cost savings is that their financial benefit might be undercounted. Studies that focus on increased volume miss the opportunity to capture financial benefits conferred through mechanisms that are independent of seeing more patients:

• Scribes might help providers address and document more, and more complex, medical problems, allowing higher level-of-service (LOS) billing. For example, a provider chooses a lower LOS because they have insufficiently documented a visit to support a higher LOS; by assisting with documentation, the scribe might allow the provider to choose a higher LOS.
• Scribes might prompt a provider to use decision-support tools for risk coding (using appropriate medical codes to capture the patient’s level of medical complexity), thereby increasing reimbursement.
• Scribes might extend the time available during the visit for the provider to address pay-for-performance quality measures, such as cancer screening.

Continue to: Making visits count, not counting visits

Making visits count, not counting visits. In this study, we examined whether medical scribes in primary care are associated with improved markers of revenue that are independent of seeing more patients. Specifically, we examined whether scribes are associated with increased LOS coding, risk coding, and orders for pay-for-performance measures for all primary care visits and for nonpreventive primary care visits.

Methods

Design

This observational study compared the change in outcomes before implementation of scribes and during implementation of scribes, between scribed providers and nonscribed providers. We compared visits during the year prior to the implementation of scribes (July 2017–June 2018) with the year during their implementation (July 2018–2019).

The Cambridge Health Alliance Institutional Review Board considered this study exempt from review.

Setting

This study was conducted at a safety-net community academic health system that uses an EMR developed by Epic Systems [Verona, WI]. This EMR includes decision-support tools that prompt providers when pay-for-performance quality measures are due and when hierarchical condition category (HCC) codes—ie, specific diagnoses used by Medicare and other payers to reimburse providers for the complexity of their patients—might apply to the visit.

Primary care practices should consider the financial benefit of scribes, independent of their ability to add patient volume.

These EMR decision-support tools use algorithms that draw on age, gender, diagnoses that were billed previously or are on the problem list, laboratory findings, and prior imaging. They alert physicians when a patient is due for pay-for-performance quality measures, such as cancer screenings, and when HCC codes might be applicable.

Continue to: During the study period...

During the study period, the EMR ­decision-support tool for HCC coding underwent several changes designed to improve HCC coding. In addition, systematic changes to primary care visits took place, leading to an increase in the number of patients seen and screenings required.

Outcomes

We examined 2 categories of outcomes that confer financial benefit to many institutions: billing measures and pay-for-performance measures.

Billing measures included the percentage of visits (1) coded as LOS 4 or 5 and (2) with at least 1 HCC code billed (among those for which the decision-support tool identified at least 1 potential HCC code).

Pay-for-performance measures. We examined whether any of 3 pay-for-­performance quality measures were addressed during the visit, selecting 3 that are commonly addressed by primary care providers (PCPs) and that require PCPs to sign an order for screening during a primary care visit: breast cancer (mammography order), cervical cancer (Papanicolaou smear order), and colon cancer (an order for fecal occult blood testing or colonoscopy).

Intervention

Scribes were employees of Cambridge Health Alliance who had recently graduated from college and were interested in a career as a health care professional. Scribes received 3 days of training on how to function effectively in their role; 1 day of training in EMR functionality; and 2 hours of training on decision-support tools for pay-for-performance quality measures and risk coding. Scribes continued learning on the job through feedback from supervising PCPs. Scribes documented patient encounters, recording histories and findings on the physical exam and transcribing discussion of treatment plans and the PCP’s instructions to patients.

Continue to: The 14 scribes worked with 17 physician...

The 14 scribes worked with 17 physician and nurse practitioner PCPs beginning in July 2018. Participation by PCPs was voluntary; they received no compensation for participating in the scribe program. PCPs were not required to see additional patients to participate. PCPs who chose to work with a scribe were similar to those who declined a scribe, as regards gender, race, type of provider (MD or NP), tenure at the institution, and percentage of time in clinical work (see Table W-1).

The control group comprised providers who elected not to work with a scribe but who worked in the same clinics as the intervention providers.

Scribes might extend the time available during the visit for the provider to address pay-forperformance quality measures, such as cancer screening.

Scribes were assigned to a PCP based on availability during the PCP’s scheduled hours and worked with 1 PCP throughout the intervention (except for 1 PCP who worked with 2 scribes). All PCPs worked with their scribe(s) part time; on average, 49% of intervention PCPs’ visits were scribed.

Inclusion and exclusion criteria

Because the first year at an institution is a learning period for PCPs, we excluded those who worked at the institution for < 1 year before the start of the scribe program (n = 12). Based on the extensive clinical experience of 1 PCP (WA) with scribes, we excluded the first 200 visits or 6 weeks (whichever occurred first) with a scribe among all scribed providers, to account for an initial learning period (n = 2202, of 15,372 scribed visits [14%]). We also excluded 2 providers who left during the pre-intervention period or were in the intervention period for < 1 month.

To ensure that we captured visits to providers with clinically significant exposure to scribes, we required scribed providers to have ≥ 20% of their visits scribed during the intervention period. To minimize the potential for contamination, we excluded nonscribed visits to scribed providers during the intervention period (n = 2211), because such nonscribed visits were largely due to visits outside the scribe’s scheduled time.

Continue to: Analysis

We compared demographic characteristics for patients and providers using the chi-square test for categorical variables and the t test for continuous variables. We compared the change in outcomes from before implementation of scribes to during implementation of scribes among scribed providers, compared to nonscribed providers, using generalized estimating equations with robust standard errors to account for repeated measures (ie, multiple visits by the same patients) and the hierarchical nature of the data (ie, patients nested within providers). We then recalculated these estimates, controlling for patient demographics (age, gender, race, and ethnicity). We repeated these analyses for patients presenting for nonpreventive visits.

Results

Visit characteristics

We examined 271,768 visits, including 41,371 visits to 17 scribed providers and 230,397 visits to 78 nonscribed providers (Table 1). Patients were most likely to be female, > 21 years of age, have English as their language of care, and be non-White. Most visits were by established patients and were nonpreventive.

We noted no clinically significant differences in characteristics between visits with scribed providers and visits with nonscribed providers, and over time. Patient complexity measures, including care management enrollment and hospital admissions, were also similar between groups, and over time.

Billing measures

HCC coding. In 28.6% of visits, the decision-support tool identified at least 1 potential HCC code. Among these, the percentage of visits with at least 1 HCC code billed increased by 10.1 percentage points (from 3.9% before implementation of scribes to 14.0%) among scribed providers, compared to increasing by 6.5 percentage points (from 2.9% before implementation to 9.3%) among nonscribed providers (TABLE 2). Scribes were therefore associated with an additional 3.6 percentage-point increase in visits with at least 1 HCC code billed (P < .0001)—a difference that remained significant after adjusting for patient demographics (P < .0001).

LOS coding. Scribed providers increased the number of visits billed as LOS 4 or 5 by 9.6 percentage points (from 47.3% before implementation to 56.8%); during the same period, nonscribed providers increased the number of visits billed as LOS 4 or 5 by 1.3 percentage points (from 46.5% before implementation to 47.8%) (TABLE 2). Scribes were therefore associated with an additional 8.3 percentage points in LOS 4 or 5 billing (P < .001) (TABLE 2). This difference remained significant after adjusting for patient demographics (P < .001).

Continue to: Pay-for-performance quality measures

Breast cancer screening. Scribed providers increased the number of visits at which breast cancer screening was ordered by 2.7 percentage points (from 17.3% before implementation of scribes to 20.0%); during the same period, the number of visits at which breast cancer screening was ordered by nonscribed providers decreased by 1.9 percentage points (from 19.5% to 17.6%). Scribes were therefore associated with an increase of 4.6 percentage points in breast cancer screening orders, compared to nonscribed providers (P < .003) (TABLE 2). That difference remained significant after adjusting for patient demographics (P = .01).

Colon cancer screening. Similarly, scribed providers increased the number of visits at which colon cancer screening was ordered by 1.2 percentage points (from 19.2% before implementation of scribes to 20.3%); during the same period, the number of visits at which colon cancer screening was ordered by nonscribed providers decreased by 2.7 percentage points (from 18.5% to 15.9%) (P = .112). After adjusting for patient demographics, scribes were associated with an increase of 4.9 percentage points in colon cancer screening orders, compared to nonscribed providers (P = .044) (TABLE 2).

Cervical cancer screening. The rate at which cervical cancer screening was ordered did not change among scribed providers and decreased (by 2.5 percentage points) among nonscribed providers—a difference that was not statistically significant (P = .26).

Nonpreventive visits. Our findings overall did not change in analyses focused on nonpreventive visits, in which scribes were associated with an increase of 8.2 percentage points in LOS 4 or 5 billing (P < .001); an increase of 3.1 percentage points in HCC coding (P < .001); and an increase of 3.2 percentage points in breast cancer screening orders (P = .03) (TABLE 3). Although scribes were associated with an increase of 1.5 percentage points in cervical cancer screening orders and an increase of 3.1 percentage points in colon cancer screening orders, these increases did not reach statistical significance.

Discussion

We found that implementation of scribes is associated with (1) an increase in LOS coding and risk coding and (2) a higher frequency of addressing 2 of 3 pay-for-performance quality measures in primary care. In adjusted analyses in our study, and compared to nonscribed providers, scribes were associated with an additional 9.2 percentage points in LOS 4 or 5 billing; 3.6 percentage points in HCC coding; 4.0 percentage points in breast cancer screening orders; and 4.9 percentage points in colon cancer screening orders. Cervical cancer screening orders followed a similar pattern, with an increase of 2.3 percentage points in the adjusted screening order rate among scribed providers, compared to nonscribed providers, during implementation of scribes—although the increase was not significant. These findings did not change in analyses focused on nonpreventive visits.

Continue to: Our findings are consistent

Our findings are consistent with those of earlier studies. Prior examinations in ambulatory specialties found that scribes increased HCC coding,⁴ LOS billing,²⁴ and pay-for-performance metrics.¹⁸ The only study to examine these areas in primary care found that scribes were associated with increased pay-for-performance measure documentation,²⁰ a change that is necessary but insufficient to realize increased pay-for-­performance revenue. Therefore, our study confirms, for the first time, that PCPs can better address pay-for-performance measures, LOS billing, and HCC coding when working with a scribe in primary care.

Providers who used a scribe during the intervention period of our study increased the number of visits billed as LOS 4 or 5 by 9.6 percentage points.

Demands on primary care visits are increasing.²⁸ Physicians are required to provide more documentation; there is greater emphasis on PCPs meeting pay-for-­performance measures; and there are more data in the EMR to review. In this context, addressing pay-for-performance measures and gaps in risk coding is likely to be increasingly challenging. Our study suggests that scribes might provide a mechanism to increase risk coding, LOS billing, and pay-for-performance measures, despite increased demands on primary care visits.

Increase in LOS billing. In the settings in which we work, a fee-for-service LOS 4 primary care visit generates, on average, $20 to $75 more in revenue than an LOS 3 visit. Using an average of $50 additional revenue for LOS 4 billing, we estimate that a full-time scribe is associated with roughly $7,000 in additional revenue annually. We arrived at this estimate using an average of 1500 visits at LOS ≤ 3 for every PCP full-time equivalent. A 9.2 percentage–point increase in LOS 4 billing would lead to roughly 140 additional LOS 4 visits, with each visit generating an additional $50 in revenue.

This analysis does not account for increased revenue associated with increased pay for HCC coding identified in our study.

Our study also suggests that scribes might provide a mechanism to increase risk coding and payfor-performance, despite increased demands on primary care visits.

Furthermore, in our conservative assumption, the entire increase in LOS billing was from level 3 to level 4; in fact, a small percentage of that increase would be from level 2 and another small percentage would be to level 5—both of which would generate additional revenue. Our assumption therefore underestimates the full financial value associated with scribes in the absence of increased patient volume. Nonetheless, the assumption suggests that increases in LOS billing offset a substantial percentage of a scribe’s salary.

Continue to: Limitations of this study

Limitations of this study. Our study should be interpreted in the context of several limitations:

• The study was conducted at 1 institution. Our findings might not be generalizable beyond this setting.
• The study measures the impact of scribes when providers work with scribes part time. Because providers who utilize a scribe for all, or nearly all, their visits are likely to use a scribe more efficiently, our study might underestimate the full impact of a scribe.
• In some settings, team members such as medical assistants are trained to assist with documentation and other responsibilities (such as closing care gaps) in addition to other patient care responsibilities.^29-32 The extent to which our findings transfer to other models is unclear; studies comparing the impact of other models (which might provide even stronger outcomes) to the impact of medical scribes would be an interesting area for further research.
• In addition to the variability of models, there is likely variability in the quality and interactions of medical scribes, which might impact outcomes. We did not examine the qualities of scribes that led to outcomes in this study.
• We examined the impact of scribes on quality measure–ordering behaviors of providers, not on whether quality measures actually improved. Because scribes are associated with more face-to-face time with patients,²⁷ they might allow for increased attention being paid by physicians to barriers to pay-for-performance measures (eg, patient education). This could increase the likelihood that patients complete a multitude of screenings, and thus improve adherence and follow-up. However, the impact of scribes on quality measures is a topic for future study.

Value beyond volume. Any limitations notwithstanding, our study suggests that scribes are associated with financial benefit in addition to the benefit of increased volume. Primary care practices should therefore consider the financial benefit of scribes independent of their ability to add patient volume. By recognizing this additive value, primary care practices might more fully capture the benefit of scribes, which might then allow practices to employ scribes with less demand to increase volume. This added support without increased volume would, in turn, likely reduce provider burnout (and the costly associated turnover) and increase patient satisfaction, leading to a synergistic financial benefit.

CORRESPONDENCE
Wayne Altman, MD, FAAFP, Tufts University School of Medicine, 200 Harrison Avenue, Boston, MA 02111; wayne. altman@tufts.edu

ABSTRACT

Purpose To avoid disruption of administrative and clinical workflow in an increasingly complex system of health information technology, health care systems and providers have started using medical scribes. The purpose of this study was to investigate the impact of medical scribes on patient satisfaction, physician satisfaction, and quality measure documentation in a family medicine office.

Methods We reviewed 1000 electronic health records for documentation of specified quality measures in the family medicine setting, before and after the use of medical scribes. We surveyed 150 patients on attitude, comfort, and acceptance of medical scribes during their visit. Five physicians shared their perceptions related to productivity, efficiency, and overall job satisfaction on working with medical scribes.

Results Documentation of 4 quality measures improved with the use of scribes, demonstrating statistical significance: fall risk assessment (odds ratio [OR] = 5.5; P = .02), follow-up tobacco screen (OR = 6.4; P = .01), follow-up body mass index plan (OR = 6.2; P < .01), and follow-up blood pressure plan (OR = 39.6; P < .01). Patients reported comfort with scribes in the examination room (96%, n = 144), a more focused health care provider (76%, n = 113), increased efficiency (74%, n = 109), and a higher degree of satisfaction with the office visit (61%, n = 90). Physicians believed they were providing better care and developing better relationships with patients while spending less time documenting and experiencing less stress.

Conclusions Use of medical scribes in a primary care setting was associated with higher patient and physician satisfaction. Patients felt comfortable with a medical scribe in the room, attested to their professionalism, and understood their purpose during the visit. The use of medical scribes in this primary care setting improved documentation of 4 quality measures.

[polldaddy:10339849]

The widespread implementation and adoption of electronic health records (EHRs) continues to increase, primarily motivated by federal incentives through the Centers for Medicare and Medicaid Services to positively impact patient care. Physician use of the EHR in the exam room has the potential to affect the patient-physician relationship, patient satisfaction, physician satisfaction, physician productivity, and physician reimbursement. In the United States, the Health Information Technology for Economic and Clinical Health Act of 2009 established incentive programs to promote meaningful use of EHRs in primary care.¹ Integrating EHRs into physician practice, adoption of meaningful use, and the increasing challenge of pay-for-performance quality measures have generated additional hours of administrative work for health care providers. These intrusions on routine clinical care, while hypothesized to improve care, have diminished physician satisfaction, increased stress, and contributed to physician burnout.²

The expanded role of clinicians incentivized to capture metrics for value-based care introduces an unprecedented level of multitasking required at the point of care. In a clinical setting, multitasking undermines the core clinical activities of observation, communication, problem solving, and, ultimately, the development of trusting relationships.^3,4 EHR documentation creates a barrier to patient engagement and may contribute to patients feeling isolated when unable to view data being entered.^5,6

Potential benefits of scribes. One means of increasing physician satisfaction and productivity may be the integration of medical scribes into health care systems. Medical scribes do not operate independently but are able to document activities or receive dictation critical for patient management—eg, recording patient histories, documenting physical examination findings and procedures, and following up on lab reports.⁷

Continue to: In a 2015 systematic review...

In a 2015 systematic review, Shultz and Holmstrom found that medical scribes in specialty settings may improve clinician satisfaction, productivity, time-related efficiency, revenue, and patient-clinician interactions.⁸ The use of scribes in one study increased the number of patients seen and time saved by emergency physicians, thereby increasing physician productivity.⁹ Studies have also shown that physicians were more satisfied during scribe engagement, related to increased time spent with patients, decreased work-related stress, and increased overall workplace satisfaction.^10-12

Sixty-one percent of patients were more satisfied with their office visit with a scribe present.

Studies on the use of medical scribes have mainly focused on physician satisfaction and productivity; however, the data on patient satisfaction are limited. Data about the use of the medical scribe in the primary care setting are also limited. The aim of our research was threefold. We wanted to evaluate the effects of using a medical scribe on: (1) patient satisfaction, (2) documentation of primary care pay-for-performance quality measures, and (3) physicians’ perceptions of the use of scribes in the primary care setting.

METHODS

Data collection

This study was conducted at Family Practice Group in Arlington, Massachusetts, where 5 part-time physicians and 3 full-time physician assistants see approximately 400 patients each week. The representative patient population is approximately 80% privately insured, 10% Medicaid, and 10% Medicare. The EHR system is eClinicalWorks.

The scribes were undergraduate college students who were interested in careers as health care professionals. They had no scribe training or experience working in a medical office. These scribes underwent 4 hours of training in EHR functionality, pay-for-performance quality measures, and risk coding (using appropriate medical codes that capture the patient’s level of medical complexity). The Independent Physician Association affiliated with Family Practice Group provided this training at no cost to the practice. The 3 scribes worked full-time with the 5 part-time physicians in the study. Scribes were not required to have had a medical background prior to entering the program.

After the aforementioned training, scribes began working full-time with physicians during patient visits and continued learning on the job through feedback from supervising physicians. Scribes documented the patient encounters, recording medical and social histories and physical exam findings, and transcribing discussions of treatment plans and physicians’ instructions to patients.

Continue to: We reviewed patient EHRs...

We reviewed patient EHRs of 5 family physicians over 2 time periods: the 3 months prior to having a medical scribe and the 3 months after beginning to work with a medical scribe. Chart data extraction occurred from 4/11/13 to 8/28/14. We reviewed 1000 patient EHRs—100 EHRs each for the 5 participating physicians before and after scribe use. Selected EHRs ran chronologically from the start of each 3-month period. Reviewing EHRs at 3 months after the onset of the medical scribe program allowed time for the scribes to be fully integrated into the practice and confident in their job responsibilities. Chart review was performed by an office administrator who was blinded as to whether documentation had been done with or without a scribe present during the visit.

Eight quality measures were evaluated in chart review. These measures were drawn from the Healthcare Effectiveness Data and Information Set (HEDIS), a tool used to measure performance in medical care and service.

We surveyed 30 patients of each of the 5 providers, yielding a total of 150 survey responses. A medical assistant gave surveys to patients in the exam room following each office visit, to be completed anonymously and privately. Patients were told that surveys would take less than 2 minutes to complete. Office visits included episodic visits, physical exams, and chronic disease management.

Physicians believed they were saving, on average, 1.5 hours each day with the use of a scribe.

After the trial period, we surveyed participating physicians regarding medical scribe assistance with documentation. We also asked the physicians 3 open-ended questions regarding their experiences with their medical scribe.

This study was reviewed and approved (IRB Approval #11424) by the Tufts Health Science Campus Institutional Review Board.

Continue to: Data analysis

During chart review, we assessed the rate at which documentation was completed for 8 quality outcome measures commonly used in the primary care setting (TABLE 1), before and after the introduction of medical scribes. These quality measures and pertinent descriptors are listed in TABLE 2.¹³ Presence or absence of documentation on all quality measures was noted for all applicable patients.

One hundred fifty patients were surveyed immediately after their office visit on their perceptions of medical scribes, including their attitude toward, comfort with, and acceptance of medical scribes (TABLE 3). Five participating physicians were surveyed to assess their perceptions related to productivity and job satisfaction with the use of medical scribes (TABLE 4), and regarding time saved and additional patients seen. Those who collected and analyzed the data from the surveys were blinded to patient and physician identifiers.

Statistical analysis

Using chi-squared tests, we compared the number of positive documentations for the 8 outcome measures before and after the use of medical scribes. Two-sided P values < .05 were considered statistically significant. All statistical analyses were performed with the use of STATA version 9 (StataCorp LP. College Station, Tex).

Physician survey data were calculated on a Likert scale, with a score of 1 corresponding to “strongly disagree,” 2 “disagree,” 3 “neither agree nor disagree,” 4 “agree,” and 5 “strongly agree.” Using the 5 answers generated from the 5 physicians, we calculated the mean for each question.

RESULTS

The use of scribes demonstrated a statistically significant improvement in the documentation of 4 (out of 8) pay-for-performance measures (TABLE 1): fall risk assessment (odds ratio [OR] = 5.5, P = .02), follow-up tobacco screen (OR = 6.4; P = .01), follow-up body mass index (BMI) plan (OR = 6.2; P < .01), and follow-up blood pressure plan (OR = 39.6; P < .01). Sample sizes of each quality measure vary as there were differing numbers of applicable patients for each quality measure within the overall 1000 charts.

Continue to: We established at the beginning...

We established at the beginning of the study a target of obtaining surveys from 30 patients of each of the 5 physicians (total of 150). Response rates for surveys were 100% for both the 150 patients and the 5 physicians. No patients declined to complete the survey, although some did not answer every question.

Patients generally had positive experiences with medical scribes (TABLE 3). The majority of patients (96%, n = 144) felt comfortable with the scribe in the room during the visit with their provider. Patients felt that the provider focused on them “a little to a lot more” (75.8%, n = 113) and thought their visit was more efficient (73.6%, n = 109) as a result of the scribe being present vs not being present. Most patients were more satisfied with their office visit with the scribe being present (60.8%, n = 90).

Physicians felt that working with a medical scribe helped them connect with their patients, made patients feel that their physician was more attentive to them, contributed to better patient care, decreased the time they spent documenting in EHR, and contributed to faster work flow (TABLE 4). The physicians also believed they had saved a mean of 1.5 hours each day with the use of a medical scribe, and that they did not have to change their schedule in any way to accommodate additional patients as a result of having a scribe.

DISCUSSION

Documentation of fall risk assessment, follow-up tobacco screening, follow-up BMI plan, and follow-up blood pressure plan all demonstrated statistically significant increases with the use of medical scribes compared with practice before scribes. Follow-up depression screen and transition of care management had relatively high ORs (3.2 and 8, respectively), but did not yield statistically significant values, in part due to small sample sizes as the number of patients who were hospitalized and the number of patients who screened positive for depression were relatively small out of the total group of 1000 patients. The use of scribes had little effect on depression screen and tobacco screen. This is likely due to the fact that there were already effective office systems in place at the practice that alerted medical assistants to complete these screens for each appropriate patient.

We found that the use of medical scribes in a primary care setting was associated with both higher patient and physician satisfaction. Although the 5 physicians in this study chose not to see additional patients when using a medical scribe, they believed they were saving, on average, 1.5 hours of time each day with the use of a scribe. All 5 physicians reported that medical scribes enabled them to provide better patient care and to help patients feel as though they had more of the physician’s attention. Patient respondents attested to their provider focusing on them more during the visit. According to patient surveys, 40.4% of respondents felt that physicians addressed their concerns more thoroughly during the visit, while the remainder of patients did not.

Continue to: Some concerns...

Some concerns of introducing medical scribes into a health care system include possible patient discomfort with a third party being present during the visit and the cost of employing medical scribes. In this study, the vast majority of patients (96%) felt comfortable with a scribe in the room. Future research could compare patient discomfort due to the presence of a medical scribe with patient discomfort due to a physician using a computer during the visit.

Limitations of this study include the small sample size of both physicians and patients; a lack of validated measures for calculating productivity, time/efficiency, and overall satisfaction; and short time periods leading up to and following the introduction of medical scribes. In addition, EHRs of patients were chosen sequentially and not randomly, which could be a confounder. Participating physicians were aware of being studied; therefore, documentation could have been affected by the Hawthorne effect. The study also was limited to one family medicine site. Although improved documentation of primary care pay-for-performance quality measures was reported, wide confidence intervals and small patient numbers hindered generalizability of findings.

Documentation of 4 out of 8 pay-forperformance measures showed statistically significant improvement with the use of scribes.

Additional studies are needed with a robust analytic plan sufficient to demonstrate baseline provider familiarity with EHRs, accuracy of medical scribe documentation, and improved documentation of pay-for-performance quality measures. Additional investigation regarding the variable competency of different medical scribes could be useful in measuring the effects of the scribe on a variety of outcomes related to both the physician and patient.

It is possible that the improved documentation yielded by the use of medical scribes could generate billing codes that reimburse physicians at a higher level (eg, a higher ratio of 99214 to 99213), leading to increased pay. Future research could aim to quantify this source of increased revenue. Furthermore, investigations could aim to quantify the revenue that medical scribes generate via improved quality measure pay-for-performance documentation.

CORRESPONDENCE
Jessica Platt, MD, 195 Canal Street, Malden, MA 02148; jpplatt@challiance.org.

ABSTRACT

Purpose To avoid disruption of administrative and clinical workflow in an increasingly complex system of health information technology, health care systems and providers have started using medical scribes. The purpose of this study was to investigate the impact of medical scribes on patient satisfaction, physician satisfaction, and quality measure documentation in a family medicine office.

Methods We reviewed 1000 electronic health records for documentation of specified quality measures in the family medicine setting, before and after the use of medical scribes. We surveyed 150 patients on attitude, comfort, and acceptance of medical scribes during their visit. Five physicians shared their perceptions related to productivity, efficiency, and overall job satisfaction on working with medical scribes.

Results Documentation of 4 quality measures improved with the use of scribes, demonstrating statistical significance: fall risk assessment (odds ratio [OR] = 5.5; P = .02), follow-up tobacco screen (OR = 6.4; P = .01), follow-up body mass index plan (OR = 6.2; P < .01), and follow-up blood pressure plan (OR = 39.6; P < .01). Patients reported comfort with scribes in the examination room (96%, n = 144), a more focused health care provider (76%, n = 113), increased efficiency (74%, n = 109), and a higher degree of satisfaction with the office visit (61%, n = 90). Physicians believed they were providing better care and developing better relationships with patients while spending less time documenting and experiencing less stress.

Conclusions Use of medical scribes in a primary care setting was associated with higher patient and physician satisfaction. Patients felt comfortable with a medical scribe in the room, attested to their professionalism, and understood their purpose during the visit. The use of medical scribes in this primary care setting improved documentation of 4 quality measures.

[polldaddy:10339849]

The widespread implementation and adoption of electronic health records (EHRs) continues to increase, primarily motivated by federal incentives through the Centers for Medicare and Medicaid Services to positively impact patient care. Physician use of the EHR in the exam room has the potential to affect the patient-physician relationship, patient satisfaction, physician satisfaction, physician productivity, and physician reimbursement. In the United States, the Health Information Technology for Economic and Clinical Health Act of 2009 established incentive programs to promote meaningful use of EHRs in primary care.¹ Integrating EHRs into physician practice, adoption of meaningful use, and the increasing challenge of pay-for-performance quality measures have generated additional hours of administrative work for health care providers. These intrusions on routine clinical care, while hypothesized to improve care, have diminished physician satisfaction, increased stress, and contributed to physician burnout.²

The expanded role of clinicians incentivized to capture metrics for value-based care introduces an unprecedented level of multitasking required at the point of care. In a clinical setting, multitasking undermines the core clinical activities of observation, communication, problem solving, and, ultimately, the development of trusting relationships.^3,4 EHR documentation creates a barrier to patient engagement and may contribute to patients feeling isolated when unable to view data being entered.^5,6

Potential benefits of scribes. One means of increasing physician satisfaction and productivity may be the integration of medical scribes into health care systems. Medical scribes do not operate independently but are able to document activities or receive dictation critical for patient management—eg, recording patient histories, documenting physical examination findings and procedures, and following up on lab reports.⁷

Continue to: In a 2015 systematic review...

In a 2015 systematic review, Shultz and Holmstrom found that medical scribes in specialty settings may improve clinician satisfaction, productivity, time-related efficiency, revenue, and patient-clinician interactions.⁸ The use of scribes in one study increased the number of patients seen and time saved by emergency physicians, thereby increasing physician productivity.⁹ Studies have also shown that physicians were more satisfied during scribe engagement, related to increased time spent with patients, decreased work-related stress, and increased overall workplace satisfaction.^10-12

Sixty-one percent of patients were more satisfied with their office visit with a scribe present.

Studies on the use of medical scribes have mainly focused on physician satisfaction and productivity; however, the data on patient satisfaction are limited. Data about the use of the medical scribe in the primary care setting are also limited. The aim of our research was threefold. We wanted to evaluate the effects of using a medical scribe on: (1) patient satisfaction, (2) documentation of primary care pay-for-performance quality measures, and (3) physicians’ perceptions of the use of scribes in the primary care setting.

METHODS

Data collection

This study was conducted at Family Practice Group in Arlington, Massachusetts, where 5 part-time physicians and 3 full-time physician assistants see approximately 400 patients each week. The representative patient population is approximately 80% privately insured, 10% Medicaid, and 10% Medicare. The EHR system is eClinicalWorks.

The scribes were undergraduate college students who were interested in careers as health care professionals. They had no scribe training or experience working in a medical office. These scribes underwent 4 hours of training in EHR functionality, pay-for-performance quality measures, and risk coding (using appropriate medical codes that capture the patient’s level of medical complexity). The Independent Physician Association affiliated with Family Practice Group provided this training at no cost to the practice. The 3 scribes worked full-time with the 5 part-time physicians in the study. Scribes were not required to have had a medical background prior to entering the program.

After the aforementioned training, scribes began working full-time with physicians during patient visits and continued learning on the job through feedback from supervising physicians. Scribes documented the patient encounters, recording medical and social histories and physical exam findings, and transcribing discussions of treatment plans and physicians’ instructions to patients.

Continue to: We reviewed patient EHRs...

We reviewed patient EHRs of 5 family physicians over 2 time periods: the 3 months prior to having a medical scribe and the 3 months after beginning to work with a medical scribe. Chart data extraction occurred from 4/11/13 to 8/28/14. We reviewed 1000 patient EHRs—100 EHRs each for the 5 participating physicians before and after scribe use. Selected EHRs ran chronologically from the start of each 3-month period. Reviewing EHRs at 3 months after the onset of the medical scribe program allowed time for the scribes to be fully integrated into the practice and confident in their job responsibilities. Chart review was performed by an office administrator who was blinded as to whether documentation had been done with or without a scribe present during the visit.

Eight quality measures were evaluated in chart review. These measures were drawn from the Healthcare Effectiveness Data and Information Set (HEDIS), a tool used to measure performance in medical care and service.

We surveyed 30 patients of each of the 5 providers, yielding a total of 150 survey responses. A medical assistant gave surveys to patients in the exam room following each office visit, to be completed anonymously and privately. Patients were told that surveys would take less than 2 minutes to complete. Office visits included episodic visits, physical exams, and chronic disease management.

Physicians believed they were saving, on average, 1.5 hours each day with the use of a scribe.

After the trial period, we surveyed participating physicians regarding medical scribe assistance with documentation. We also asked the physicians 3 open-ended questions regarding their experiences with their medical scribe.

This study was reviewed and approved (IRB Approval #11424) by the Tufts Health Science Campus Institutional Review Board.

Continue to: Data analysis

During chart review, we assessed the rate at which documentation was completed for 8 quality outcome measures commonly used in the primary care setting (TABLE 1), before and after the introduction of medical scribes. These quality measures and pertinent descriptors are listed in TABLE 2.¹³ Presence or absence of documentation on all quality measures was noted for all applicable patients.

One hundred fifty patients were surveyed immediately after their office visit on their perceptions of medical scribes, including their attitude toward, comfort with, and acceptance of medical scribes (TABLE 3). Five participating physicians were surveyed to assess their perceptions related to productivity and job satisfaction with the use of medical scribes (TABLE 4), and regarding time saved and additional patients seen. Those who collected and analyzed the data from the surveys were blinded to patient and physician identifiers.

Statistical analysis

Using chi-squared tests, we compared the number of positive documentations for the 8 outcome measures before and after the use of medical scribes. Two-sided P values < .05 were considered statistically significant. All statistical analyses were performed with the use of STATA version 9 (StataCorp LP. College Station, Tex).

Physician survey data were calculated on a Likert scale, with a score of 1 corresponding to “strongly disagree,” 2 “disagree,” 3 “neither agree nor disagree,” 4 “agree,” and 5 “strongly agree.” Using the 5 answers generated from the 5 physicians, we calculated the mean for each question.

RESULTS

The use of scribes demonstrated a statistically significant improvement in the documentation of 4 (out of 8) pay-for-performance measures (TABLE 1): fall risk assessment (odds ratio [OR] = 5.5, P = .02), follow-up tobacco screen (OR = 6.4; P = .01), follow-up body mass index (BMI) plan (OR = 6.2; P < .01), and follow-up blood pressure plan (OR = 39.6; P < .01). Sample sizes of each quality measure vary as there were differing numbers of applicable patients for each quality measure within the overall 1000 charts.

Continue to: We established at the beginning...

We established at the beginning of the study a target of obtaining surveys from 30 patients of each of the 5 physicians (total of 150). Response rates for surveys were 100% for both the 150 patients and the 5 physicians. No patients declined to complete the survey, although some did not answer every question.

Patients generally had positive experiences with medical scribes (TABLE 3). The majority of patients (96%, n = 144) felt comfortable with the scribe in the room during the visit with their provider. Patients felt that the provider focused on them “a little to a lot more” (75.8%, n = 113) and thought their visit was more efficient (73.6%, n = 109) as a result of the scribe being present vs not being present. Most patients were more satisfied with their office visit with the scribe being present (60.8%, n = 90).

Physicians felt that working with a medical scribe helped them connect with their patients, made patients feel that their physician was more attentive to them, contributed to better patient care, decreased the time they spent documenting in EHR, and contributed to faster work flow (TABLE 4). The physicians also believed they had saved a mean of 1.5 hours each day with the use of a medical scribe, and that they did not have to change their schedule in any way to accommodate additional patients as a result of having a scribe.

DISCUSSION

Documentation of fall risk assessment, follow-up tobacco screening, follow-up BMI plan, and follow-up blood pressure plan all demonstrated statistically significant increases with the use of medical scribes compared with practice before scribes. Follow-up depression screen and transition of care management had relatively high ORs (3.2 and 8, respectively), but did not yield statistically significant values, in part due to small sample sizes as the number of patients who were hospitalized and the number of patients who screened positive for depression were relatively small out of the total group of 1000 patients. The use of scribes had little effect on depression screen and tobacco screen. This is likely due to the fact that there were already effective office systems in place at the practice that alerted medical assistants to complete these screens for each appropriate patient.

We found that the use of medical scribes in a primary care setting was associated with both higher patient and physician satisfaction. Although the 5 physicians in this study chose not to see additional patients when using a medical scribe, they believed they were saving, on average, 1.5 hours of time each day with the use of a scribe. All 5 physicians reported that medical scribes enabled them to provide better patient care and to help patients feel as though they had more of the physician’s attention. Patient respondents attested to their provider focusing on them more during the visit. According to patient surveys, 40.4% of respondents felt that physicians addressed their concerns more thoroughly during the visit, while the remainder of patients did not.

Continue to: Some concerns...

Some concerns of introducing medical scribes into a health care system include possible patient discomfort with a third party being present during the visit and the cost of employing medical scribes. In this study, the vast majority of patients (96%) felt comfortable with a scribe in the room. Future research could compare patient discomfort due to the presence of a medical scribe with patient discomfort due to a physician using a computer during the visit.

Limitations of this study include the small sample size of both physicians and patients; a lack of validated measures for calculating productivity, time/efficiency, and overall satisfaction; and short time periods leading up to and following the introduction of medical scribes. In addition, EHRs of patients were chosen sequentially and not randomly, which could be a confounder. Participating physicians were aware of being studied; therefore, documentation could have been affected by the Hawthorne effect. The study also was limited to one family medicine site. Although improved documentation of primary care pay-for-performance quality measures was reported, wide confidence intervals and small patient numbers hindered generalizability of findings.

Documentation of 4 out of 8 pay-forperformance measures showed statistically significant improvement with the use of scribes.

Additional studies are needed with a robust analytic plan sufficient to demonstrate baseline provider familiarity with EHRs, accuracy of medical scribe documentation, and improved documentation of pay-for-performance quality measures. Additional investigation regarding the variable competency of different medical scribes could be useful in measuring the effects of the scribe on a variety of outcomes related to both the physician and patient.

It is possible that the improved documentation yielded by the use of medical scribes could generate billing codes that reimburse physicians at a higher level (eg, a higher ratio of 99214 to 99213), leading to increased pay. Future research could aim to quantify this source of increased revenue. Furthermore, investigations could aim to quantify the revenue that medical scribes generate via improved quality measure pay-for-performance documentation.

CORRESPONDENCE
Jessica Platt, MD, 195 Canal Street, Malden, MA 02148; jpplatt@challiance.org.

ABSTRACT

Purpose To avoid disruption of administrative and clinical workflow in an increasingly complex system of health information technology, health care systems and providers have started using medical scribes. The purpose of this study was to investigate the impact of medical scribes on patient satisfaction, physician satisfaction, and quality measure documentation in a family medicine office.

Methods We reviewed 1000 electronic health records for documentation of specified quality measures in the family medicine setting, before and after the use of medical scribes. We surveyed 150 patients on attitude, comfort, and acceptance of medical scribes during their visit. Five physicians shared their perceptions related to productivity, efficiency, and overall job satisfaction on working with medical scribes.

Results Documentation of 4 quality measures improved with the use of scribes, demonstrating statistical significance: fall risk assessment (odds ratio [OR] = 5.5; P = .02), follow-up tobacco screen (OR = 6.4; P = .01), follow-up body mass index plan (OR = 6.2; P < .01), and follow-up blood pressure plan (OR = 39.6; P < .01). Patients reported comfort with scribes in the examination room (96%, n = 144), a more focused health care provider (76%, n = 113), increased efficiency (74%, n = 109), and a higher degree of satisfaction with the office visit (61%, n = 90). Physicians believed they were providing better care and developing better relationships with patients while spending less time documenting and experiencing less stress.

Conclusions Use of medical scribes in a primary care setting was associated with higher patient and physician satisfaction. Patients felt comfortable with a medical scribe in the room, attested to their professionalism, and understood their purpose during the visit. The use of medical scribes in this primary care setting improved documentation of 4 quality measures.

[polldaddy:10339849]

The widespread implementation and adoption of electronic health records (EHRs) continues to increase, primarily motivated by federal incentives through the Centers for Medicare and Medicaid Services to positively impact patient care. Physician use of the EHR in the exam room has the potential to affect the patient-physician relationship, patient satisfaction, physician satisfaction, physician productivity, and physician reimbursement. In the United States, the Health Information Technology for Economic and Clinical Health Act of 2009 established incentive programs to promote meaningful use of EHRs in primary care.¹ Integrating EHRs into physician practice, adoption of meaningful use, and the increasing challenge of pay-for-performance quality measures have generated additional hours of administrative work for health care providers. These intrusions on routine clinical care, while hypothesized to improve care, have diminished physician satisfaction, increased stress, and contributed to physician burnout.²

The expanded role of clinicians incentivized to capture metrics for value-based care introduces an unprecedented level of multitasking required at the point of care. In a clinical setting, multitasking undermines the core clinical activities of observation, communication, problem solving, and, ultimately, the development of trusting relationships.^3,4 EHR documentation creates a barrier to patient engagement and may contribute to patients feeling isolated when unable to view data being entered.^5,6

Potential benefits of scribes. One means of increasing physician satisfaction and productivity may be the integration of medical scribes into health care systems. Medical scribes do not operate independently but are able to document activities or receive dictation critical for patient management—eg, recording patient histories, documenting physical examination findings and procedures, and following up on lab reports.⁷

Continue to: In a 2015 systematic review...

In a 2015 systematic review, Shultz and Holmstrom found that medical scribes in specialty settings may improve clinician satisfaction, productivity, time-related efficiency, revenue, and patient-clinician interactions.⁸ The use of scribes in one study increased the number of patients seen and time saved by emergency physicians, thereby increasing physician productivity.⁹ Studies have also shown that physicians were more satisfied during scribe engagement, related to increased time spent with patients, decreased work-related stress, and increased overall workplace satisfaction.^10-12

Sixty-one percent of patients were more satisfied with their office visit with a scribe present.

Studies on the use of medical scribes have mainly focused on physician satisfaction and productivity; however, the data on patient satisfaction are limited. Data about the use of the medical scribe in the primary care setting are also limited. The aim of our research was threefold. We wanted to evaluate the effects of using a medical scribe on: (1) patient satisfaction, (2) documentation of primary care pay-for-performance quality measures, and (3) physicians’ perceptions of the use of scribes in the primary care setting.

METHODS

Data collection

This study was conducted at Family Practice Group in Arlington, Massachusetts, where 5 part-time physicians and 3 full-time physician assistants see approximately 400 patients each week. The representative patient population is approximately 80% privately insured, 10% Medicaid, and 10% Medicare. The EHR system is eClinicalWorks.

The scribes were undergraduate college students who were interested in careers as health care professionals. They had no scribe training or experience working in a medical office. These scribes underwent 4 hours of training in EHR functionality, pay-for-performance quality measures, and risk coding (using appropriate medical codes that capture the patient’s level of medical complexity). The Independent Physician Association affiliated with Family Practice Group provided this training at no cost to the practice. The 3 scribes worked full-time with the 5 part-time physicians in the study. Scribes were not required to have had a medical background prior to entering the program.

After the aforementioned training, scribes began working full-time with physicians during patient visits and continued learning on the job through feedback from supervising physicians. Scribes documented the patient encounters, recording medical and social histories and physical exam findings, and transcribing discussions of treatment plans and physicians’ instructions to patients.

Continue to: We reviewed patient EHRs...

We reviewed patient EHRs of 5 family physicians over 2 time periods: the 3 months prior to having a medical scribe and the 3 months after beginning to work with a medical scribe. Chart data extraction occurred from 4/11/13 to 8/28/14. We reviewed 1000 patient EHRs—100 EHRs each for the 5 participating physicians before and after scribe use. Selected EHRs ran chronologically from the start of each 3-month period. Reviewing EHRs at 3 months after the onset of the medical scribe program allowed time for the scribes to be fully integrated into the practice and confident in their job responsibilities. Chart review was performed by an office administrator who was blinded as to whether documentation had been done with or without a scribe present during the visit.

Eight quality measures were evaluated in chart review. These measures were drawn from the Healthcare Effectiveness Data and Information Set (HEDIS), a tool used to measure performance in medical care and service.

We surveyed 30 patients of each of the 5 providers, yielding a total of 150 survey responses. A medical assistant gave surveys to patients in the exam room following each office visit, to be completed anonymously and privately. Patients were told that surveys would take less than 2 minutes to complete. Office visits included episodic visits, physical exams, and chronic disease management.

Physicians believed they were saving, on average, 1.5 hours each day with the use of a scribe.

After the trial period, we surveyed participating physicians regarding medical scribe assistance with documentation. We also asked the physicians 3 open-ended questions regarding their experiences with their medical scribe.

This study was reviewed and approved (IRB Approval #11424) by the Tufts Health Science Campus Institutional Review Board.

Continue to: Data analysis

During chart review, we assessed the rate at which documentation was completed for 8 quality outcome measures commonly used in the primary care setting (TABLE 1), before and after the introduction of medical scribes. These quality measures and pertinent descriptors are listed in TABLE 2.¹³ Presence or absence of documentation on all quality measures was noted for all applicable patients.

One hundred fifty patients were surveyed immediately after their office visit on their perceptions of medical scribes, including their attitude toward, comfort with, and acceptance of medical scribes (TABLE 3). Five participating physicians were surveyed to assess their perceptions related to productivity and job satisfaction with the use of medical scribes (TABLE 4), and regarding time saved and additional patients seen. Those who collected and analyzed the data from the surveys were blinded to patient and physician identifiers.

Statistical analysis

Using chi-squared tests, we compared the number of positive documentations for the 8 outcome measures before and after the use of medical scribes. Two-sided P values < .05 were considered statistically significant. All statistical analyses were performed with the use of STATA version 9 (StataCorp LP. College Station, Tex).

Physician survey data were calculated on a Likert scale, with a score of 1 corresponding to “strongly disagree,” 2 “disagree,” 3 “neither agree nor disagree,” 4 “agree,” and 5 “strongly agree.” Using the 5 answers generated from the 5 physicians, we calculated the mean for each question.

RESULTS

The use of scribes demonstrated a statistically significant improvement in the documentation of 4 (out of 8) pay-for-performance measures (TABLE 1): fall risk assessment (odds ratio [OR] = 5.5, P = .02), follow-up tobacco screen (OR = 6.4; P = .01), follow-up body mass index (BMI) plan (OR = 6.2; P < .01), and follow-up blood pressure plan (OR = 39.6; P < .01). Sample sizes of each quality measure vary as there were differing numbers of applicable patients for each quality measure within the overall 1000 charts.

Continue to: We established at the beginning...

We established at the beginning of the study a target of obtaining surveys from 30 patients of each of the 5 physicians (total of 150). Response rates for surveys were 100% for both the 150 patients and the 5 physicians. No patients declined to complete the survey, although some did not answer every question.

Patients generally had positive experiences with medical scribes (TABLE 3). The majority of patients (96%, n = 144) felt comfortable with the scribe in the room during the visit with their provider. Patients felt that the provider focused on them “a little to a lot more” (75.8%, n = 113) and thought their visit was more efficient (73.6%, n = 109) as a result of the scribe being present vs not being present. Most patients were more satisfied with their office visit with the scribe being present (60.8%, n = 90).

Physicians felt that working with a medical scribe helped them connect with their patients, made patients feel that their physician was more attentive to them, contributed to better patient care, decreased the time they spent documenting in EHR, and contributed to faster work flow (TABLE 4). The physicians also believed they had saved a mean of 1.5 hours each day with the use of a medical scribe, and that they did not have to change their schedule in any way to accommodate additional patients as a result of having a scribe.

DISCUSSION

Documentation of fall risk assessment, follow-up tobacco screening, follow-up BMI plan, and follow-up blood pressure plan all demonstrated statistically significant increases with the use of medical scribes compared with practice before scribes. Follow-up depression screen and transition of care management had relatively high ORs (3.2 and 8, respectively), but did not yield statistically significant values, in part due to small sample sizes as the number of patients who were hospitalized and the number of patients who screened positive for depression were relatively small out of the total group of 1000 patients. The use of scribes had little effect on depression screen and tobacco screen. This is likely due to the fact that there were already effective office systems in place at the practice that alerted medical assistants to complete these screens for each appropriate patient.

We found that the use of medical scribes in a primary care setting was associated with both higher patient and physician satisfaction. Although the 5 physicians in this study chose not to see additional patients when using a medical scribe, they believed they were saving, on average, 1.5 hours of time each day with the use of a scribe. All 5 physicians reported that medical scribes enabled them to provide better patient care and to help patients feel as though they had more of the physician’s attention. Patient respondents attested to their provider focusing on them more during the visit. According to patient surveys, 40.4% of respondents felt that physicians addressed their concerns more thoroughly during the visit, while the remainder of patients did not.

Continue to: Some concerns...

Some concerns of introducing medical scribes into a health care system include possible patient discomfort with a third party being present during the visit and the cost of employing medical scribes. In this study, the vast majority of patients (96%) felt comfortable with a scribe in the room. Future research could compare patient discomfort due to the presence of a medical scribe with patient discomfort due to a physician using a computer during the visit.

Limitations of this study include the small sample size of both physicians and patients; a lack of validated measures for calculating productivity, time/efficiency, and overall satisfaction; and short time periods leading up to and following the introduction of medical scribes. In addition, EHRs of patients were chosen sequentially and not randomly, which could be a confounder. Participating physicians were aware of being studied; therefore, documentation could have been affected by the Hawthorne effect. The study also was limited to one family medicine site. Although improved documentation of primary care pay-for-performance quality measures was reported, wide confidence intervals and small patient numbers hindered generalizability of findings.

Documentation of 4 out of 8 pay-forperformance measures showed statistically significant improvement with the use of scribes.

Additional studies are needed with a robust analytic plan sufficient to demonstrate baseline provider familiarity with EHRs, accuracy of medical scribe documentation, and improved documentation of pay-for-performance quality measures. Additional investigation regarding the variable competency of different medical scribes could be useful in measuring the effects of the scribe on a variety of outcomes related to both the physician and patient.

It is possible that the improved documentation yielded by the use of medical scribes could generate billing codes that reimburse physicians at a higher level (eg, a higher ratio of 99214 to 99213), leading to increased pay. Future research could aim to quantify this source of increased revenue. Furthermore, investigations could aim to quantify the revenue that medical scribes generate via improved quality measure pay-for-performance documentation.

CORRESPONDENCE
Jessica Platt, MD, 195 Canal Street, Malden, MA 02148; jpplatt@challiance.org.