Health Literacy and Medication Use

Population

The PILL‐CVD study protocol and eligibility criteria has been previously published.9 Briefly, patients were eligible if they were at least 18 years old and admitted with acute coronary syndrome or acute decompensated heart failure. Patients were excluded if they: were too ill to complete an interview; were not oriented to person, place, or time; had a corrected visual acuity worse than 20/200; had impaired hearing; could not communicate in English or Spanish; were not responsible for managing their own medications; had no phone; were unlikely to be discharged to home; were in police custody; or had been previously enrolled in the study. For the present analysis, we also excluded any patient who was not on at least 1 prescription medication prior to admission. Saline nasal spray, saline eye drops, herbal products, nutritional supplements, vitamins, and over the counter (OTC) lotions and creams were not counted as prescription medications. Oral medications available both OTC and by prescription (eg, aspirin, nonsteroidal anti‐inflammatory drugs, and acid reflux medications) were counted as prescription medications.

Measures

At enrollment, which was usually within 24 to 48 hours of admission, participants completed the short form of the Test of Functional Health Literacy in Adults (s‐TOFHLA) in English or Spanish,10 the Mini‐Cog test of cognition,11 and the Medication Understanding Questionnaire (MUQ), as well as demographic information. The number of prescription medications prior to hospital admission was abstracted from the best available reference listthat documented by the treating physicians in the electronic health record (EHR). The EHR at each site was a home‐grown system and included both inpatient and outpatient records, which facilitated physicians' documentation of the medication list.

The s‐TOFHLA consists of 2 short reading‐comprehension passages. Scores on the s‐TOFHLA range from 0 to 36, and can be categorized as inadequate (0‐16), marginal (17‐22), or adequate (23‐36) health literacy.10 The Mini‐Cog includes 3‐item recall and clock‐drawing tests. It provides a brief measure of cognitive function and performs well among patients with limited literacy or educational attainment.11 Scores range from 0 to 5, with a score <3 indicating possible dementia.

The MUQ was administered verbally and assessed patients' understanding of their own preadmission medication regimen. It was developed for this study, based on published measures of medication understanding.2, 12 To administer the MUQ, research assistants (RAs) accessed the patient's preadmission medication list from the EHR and used a random number table to select up to 5 prescription medications from the list. If the patient was taking 5 or fewer medications, all of their medications were selected. Saline nasal spray, saline eye drops, herbal products, nutritional supplements, vitamins, and OTC lotions and creams were excluded from testing. The RA provided the brand and generic name of each medication, and then asked the patient for the drug's purpose, strength per unit (eg, 20 mg tablet), number of units taken at a time (eg, 2 tablets), and dosing frequency (eg, twice a day). For drugs prescribed on an as‐needed basis, the RA asked patients for the maximum allowable dose and frequency. Patients were instructed to not refer to a medication list or bottles when responding. The RA documented the patient's responses on the MUQ, along with the dosing information from the EHR for each selected medication.

One clinical pharmacist (MM) scored all MUQ forms by applying a set of scoring rules. Each medication score could range from 0 to 3. The components of the score included indication (1 point), strength (0.5 point), units (0.5 point), and frequency (1 point). The patient's overall MUQ score was an average of the MUQ scores for each tested medication.

Statistical Analysis

We summarized patient characteristics, number of preadmission medications, and MUQ scores using median and interquartile range (IRQ) for continuous variables, and frequencies and proportions for categorical variables. We conducted proportional odds logistic regression (ordinal regression) to estimate the effect of s‐TOFHLA score, other patient characteristics, and number of medications on MUQ scores.13

Important covariates were selected a priori based on clinical significance. These included age (continuous), gender, patient self‐reported race (white, black, other nonwhite), Mini‐Cog score (continuous), primary language (English or Spanish), years of education (continuous), number of preadmission medications (continuous), income (ordinal categories), insurance type (categorical), and study site. Covariates with missing data (household income, health literacy, and years of education) were imputed using multiple imputation techniques.14 The relationship between number of preadmission medications and MUQ scores was found to be nonlinear, and it was modeled using restricted cubic splines.14 We also fit models which treated health literacy and cognition as categorical variables. Results are reported as odds ratios (OR) with 95% confidence intervals (CI). Wald tests were used to test for the statistical significance of predictor variables. Two‐sided P values less than 0.05 were considered statistically significant. All analyses were performed using statistical language R (R Foundation, available at: http://www.r‐project.org).

Baseline Characteristics

Among the 862 patients enrolled in PILL‐CVD, 790 (91.7 %) had at least 1 preadmission medication and were included in this analysis (Table 1). Forty‐seven percent were admitted to VUH (N = 373) and 53% to BWH (N = 417). The median age was 61 (interquartile range [IQR] 52, 71), 77% were white, and 57% were male. Inadequate or marginal health literacy was identified among 11% and 9% of patients, respectively. The median number of preadmission medications was 8 (IQR 5, 11). Patients excluded from the analysis for not having preadmission medications were similar to included patients, except they were more likely to be male (76% vs 57%) and less likely to have health insurance (23% self‐pay vs 4%). (Data available upon request.)

MUQ Scores

The MUQ was administered in approximately 5 minutes. The median MUQ score was 2.5 (IQR 2.2, 2.8) (Table 2); 16.3% of patients scored less than 2. Subjects typically achieved high scores for the domains of indication, units, and frequency, while scores on the strength domain were lower (median = 0.2 [IQR 0.1, 0.4], maximum possible = 0.5).

Predictors of Medication Understanding

Unadjusted relationships of health literacy, cognition, and number of medications with medication understanding are shown in Figure 1 (panels A, B, and C, respectively). The figure demonstrates a linear relationship with both health literacy (Figure 1A) and cognition (Figure 1B), and a nonlinear relationship between number of preadmission medications and MUQ score (Figure 1C).

Figure 1

Adjusted relationships using imputed data for missing covariates are shown in Figure 2. Lower health literacy, cognitive impairment, male gender, and black race were independently associated with lower understanding of preadmission medications. Each 1 point increase in s‐TOFHLA or Mini‐Cog score led to an increase in medication understanding (OR = 1.04; 95% CI, 1.02 to 1.06; P = 0.0001; and OR = 1.24; 95% CI, 1.1 to 1.4; P = 0.001; respectively). Patients with marginal or inadequate health literacy had lower odds of understanding their regimen (OR = 0.53; 95% CI, 0.34 to 0.84; and OR = 0.49; 95% CI, 0.31 to 0.78, respectively) compared to those with adequate health literacy. Impaired cognitive function (Mini‐Cog score <3, indicating dementia) was also associated with lower odds of medication understanding (OR = 0.57; 95% CI, 0.38 to 0.86) compared to those with no cognitive impairment. An increase in the number of preadmission medications (up to 10) was also strongly associated with lower MUQ scores. For each increase by 1 medication, there was a significant decrease in medication understanding, up to 10 medications, beyond which understanding did not significantly decrease further. Patients on 6 medications were about half as likely to understand their medication regimen as patients on only 1 medication (OR = 0.52; 95% CI, 0.36 to 0.75). For patients on 11 medications, the odds of medication understanding were 24% lower than for patients on 6 medications (OR = 0.76; 95% CI, 0.65 to 0.89). Patients' age, years of schooling, and household income were not independently associated with medication understanding. Results were similar using data without multiple imputation.

Figure 2

Examples of Misunderstanding of Common Medications

Table 3 provides examples of incorrect patient responses for several commonly prescribed medications or drug classes, including aspirin, digoxin, nitroglycerin, and HMG‐CoA reductase inhibitors (statins). For aspirin, many patients were not aware of the strength. For digoxin, several participants reported splitting a higher‐strength pill to obtain the prescribed dose, which should not be done given the imprecision of splitting and narrow therapeutic index of this drug. Patients prescribed nitroglycerin sublingual tablets were commonly unable to report the correct dosing and frequency for angina treatment. Medications for cholesterol were often reported as being taken in the morning; this was scored strictly as a frequency error if the medication timing in the EHR was listed as evening or bedtime. We also identified many patients with poor understanding of opioid analgesics, particularly regarding their dosing and frequency.

Population

The PILL‐CVD study protocol and eligibility criteria has been previously published.9 Briefly, patients were eligible if they were at least 18 years old and admitted with acute coronary syndrome or acute decompensated heart failure. Patients were excluded if they: were too ill to complete an interview; were not oriented to person, place, or time; had a corrected visual acuity worse than 20/200; had impaired hearing; could not communicate in English or Spanish; were not responsible for managing their own medications; had no phone; were unlikely to be discharged to home; were in police custody; or had been previously enrolled in the study. For the present analysis, we also excluded any patient who was not on at least 1 prescription medication prior to admission. Saline nasal spray, saline eye drops, herbal products, nutritional supplements, vitamins, and over the counter (OTC) lotions and creams were not counted as prescription medications. Oral medications available both OTC and by prescription (eg, aspirin, nonsteroidal anti‐inflammatory drugs, and acid reflux medications) were counted as prescription medications.

Measures

At enrollment, which was usually within 24 to 48 hours of admission, participants completed the short form of the Test of Functional Health Literacy in Adults (s‐TOFHLA) in English or Spanish,10 the Mini‐Cog test of cognition,11 and the Medication Understanding Questionnaire (MUQ), as well as demographic information. The number of prescription medications prior to hospital admission was abstracted from the best available reference listthat documented by the treating physicians in the electronic health record (EHR). The EHR at each site was a home‐grown system and included both inpatient and outpatient records, which facilitated physicians' documentation of the medication list.

The s‐TOFHLA consists of 2 short reading‐comprehension passages. Scores on the s‐TOFHLA range from 0 to 36, and can be categorized as inadequate (0‐16), marginal (17‐22), or adequate (23‐36) health literacy.10 The Mini‐Cog includes 3‐item recall and clock‐drawing tests. It provides a brief measure of cognitive function and performs well among patients with limited literacy or educational attainment.11 Scores range from 0 to 5, with a score <3 indicating possible dementia.

The MUQ was administered verbally and assessed patients' understanding of their own preadmission medication regimen. It was developed for this study, based on published measures of medication understanding.2, 12 To administer the MUQ, research assistants (RAs) accessed the patient's preadmission medication list from the EHR and used a random number table to select up to 5 prescription medications from the list. If the patient was taking 5 or fewer medications, all of their medications were selected. Saline nasal spray, saline eye drops, herbal products, nutritional supplements, vitamins, and OTC lotions and creams were excluded from testing. The RA provided the brand and generic name of each medication, and then asked the patient for the drug's purpose, strength per unit (eg, 20 mg tablet), number of units taken at a time (eg, 2 tablets), and dosing frequency (eg, twice a day). For drugs prescribed on an as‐needed basis, the RA asked patients for the maximum allowable dose and frequency. Patients were instructed to not refer to a medication list or bottles when responding. The RA documented the patient's responses on the MUQ, along with the dosing information from the EHR for each selected medication.

One clinical pharmacist (MM) scored all MUQ forms by applying a set of scoring rules. Each medication score could range from 0 to 3. The components of the score included indication (1 point), strength (0.5 point), units (0.5 point), and frequency (1 point). The patient's overall MUQ score was an average of the MUQ scores for each tested medication.

Statistical Analysis

We summarized patient characteristics, number of preadmission medications, and MUQ scores using median and interquartile range (IRQ) for continuous variables, and frequencies and proportions for categorical variables. We conducted proportional odds logistic regression (ordinal regression) to estimate the effect of s‐TOFHLA score, other patient characteristics, and number of medications on MUQ scores.13

Important covariates were selected a priori based on clinical significance. These included age (continuous), gender, patient self‐reported race (white, black, other nonwhite), Mini‐Cog score (continuous), primary language (English or Spanish), years of education (continuous), number of preadmission medications (continuous), income (ordinal categories), insurance type (categorical), and study site. Covariates with missing data (household income, health literacy, and years of education) were imputed using multiple imputation techniques.14 The relationship between number of preadmission medications and MUQ scores was found to be nonlinear, and it was modeled using restricted cubic splines.14 We also fit models which treated health literacy and cognition as categorical variables. Results are reported as odds ratios (OR) with 95% confidence intervals (CI). Wald tests were used to test for the statistical significance of predictor variables. Two‐sided P values less than 0.05 were considered statistically significant. All analyses were performed using statistical language R (R Foundation, available at: http://www.r‐project.org).

Baseline Characteristics

Among the 862 patients enrolled in PILL‐CVD, 790 (91.7 %) had at least 1 preadmission medication and were included in this analysis (Table 1). Forty‐seven percent were admitted to VUH (N = 373) and 53% to BWH (N = 417). The median age was 61 (interquartile range [IQR] 52, 71), 77% were white, and 57% were male. Inadequate or marginal health literacy was identified among 11% and 9% of patients, respectively. The median number of preadmission medications was 8 (IQR 5, 11). Patients excluded from the analysis for not having preadmission medications were similar to included patients, except they were more likely to be male (76% vs 57%) and less likely to have health insurance (23% self‐pay vs 4%). (Data available upon request.)

MUQ Scores

The MUQ was administered in approximately 5 minutes. The median MUQ score was 2.5 (IQR 2.2, 2.8) (Table 2); 16.3% of patients scored less than 2. Subjects typically achieved high scores for the domains of indication, units, and frequency, while scores on the strength domain were lower (median = 0.2 [IQR 0.1, 0.4], maximum possible = 0.5).

Predictors of Medication Understanding

Unadjusted relationships of health literacy, cognition, and number of medications with medication understanding are shown in Figure 1 (panels A, B, and C, respectively). The figure demonstrates a linear relationship with both health literacy (Figure 1A) and cognition (Figure 1B), and a nonlinear relationship between number of preadmission medications and MUQ score (Figure 1C).

Figure 1

Adjusted relationships using imputed data for missing covariates are shown in Figure 2. Lower health literacy, cognitive impairment, male gender, and black race were independently associated with lower understanding of preadmission medications. Each 1 point increase in s‐TOFHLA or Mini‐Cog score led to an increase in medication understanding (OR = 1.04; 95% CI, 1.02 to 1.06; P = 0.0001; and OR = 1.24; 95% CI, 1.1 to 1.4; P = 0.001; respectively). Patients with marginal or inadequate health literacy had lower odds of understanding their regimen (OR = 0.53; 95% CI, 0.34 to 0.84; and OR = 0.49; 95% CI, 0.31 to 0.78, respectively) compared to those with adequate health literacy. Impaired cognitive function (Mini‐Cog score <3, indicating dementia) was also associated with lower odds of medication understanding (OR = 0.57; 95% CI, 0.38 to 0.86) compared to those with no cognitive impairment. An increase in the number of preadmission medications (up to 10) was also strongly associated with lower MUQ scores. For each increase by 1 medication, there was a significant decrease in medication understanding, up to 10 medications, beyond which understanding did not significantly decrease further. Patients on 6 medications were about half as likely to understand their medication regimen as patients on only 1 medication (OR = 0.52; 95% CI, 0.36 to 0.75). For patients on 11 medications, the odds of medication understanding were 24% lower than for patients on 6 medications (OR = 0.76; 95% CI, 0.65 to 0.89). Patients' age, years of schooling, and household income were not independently associated with medication understanding. Results were similar using data without multiple imputation.

Figure 2

Examples of Misunderstanding of Common Medications

Table 3 provides examples of incorrect patient responses for several commonly prescribed medications or drug classes, including aspirin, digoxin, nitroglycerin, and HMG‐CoA reductase inhibitors (statins). For aspirin, many patients were not aware of the strength. For digoxin, several participants reported splitting a higher‐strength pill to obtain the prescribed dose, which should not be done given the imprecision of splitting and narrow therapeutic index of this drug. Patients prescribed nitroglycerin sublingual tablets were commonly unable to report the correct dosing and frequency for angina treatment. Medications for cholesterol were often reported as being taken in the morning; this was scored strictly as a frequency error if the medication timing in the EHR was listed as evening or bedtime. We also identified many patients with poor understanding of opioid analgesics, particularly regarding their dosing and frequency.

Population

The PILL‐CVD study protocol and eligibility criteria has been previously published.9 Briefly, patients were eligible if they were at least 18 years old and admitted with acute coronary syndrome or acute decompensated heart failure. Patients were excluded if they: were too ill to complete an interview; were not oriented to person, place, or time; had a corrected visual acuity worse than 20/200; had impaired hearing; could not communicate in English or Spanish; were not responsible for managing their own medications; had no phone; were unlikely to be discharged to home; were in police custody; or had been previously enrolled in the study. For the present analysis, we also excluded any patient who was not on at least 1 prescription medication prior to admission. Saline nasal spray, saline eye drops, herbal products, nutritional supplements, vitamins, and over the counter (OTC) lotions and creams were not counted as prescription medications. Oral medications available both OTC and by prescription (eg, aspirin, nonsteroidal anti‐inflammatory drugs, and acid reflux medications) were counted as prescription medications.

Measures

At enrollment, which was usually within 24 to 48 hours of admission, participants completed the short form of the Test of Functional Health Literacy in Adults (s‐TOFHLA) in English or Spanish,10 the Mini‐Cog test of cognition,11 and the Medication Understanding Questionnaire (MUQ), as well as demographic information. The number of prescription medications prior to hospital admission was abstracted from the best available reference listthat documented by the treating physicians in the electronic health record (EHR). The EHR at each site was a home‐grown system and included both inpatient and outpatient records, which facilitated physicians' documentation of the medication list.

The s‐TOFHLA consists of 2 short reading‐comprehension passages. Scores on the s‐TOFHLA range from 0 to 36, and can be categorized as inadequate (0‐16), marginal (17‐22), or adequate (23‐36) health literacy.10 The Mini‐Cog includes 3‐item recall and clock‐drawing tests. It provides a brief measure of cognitive function and performs well among patients with limited literacy or educational attainment.11 Scores range from 0 to 5, with a score <3 indicating possible dementia.

The MUQ was administered verbally and assessed patients' understanding of their own preadmission medication regimen. It was developed for this study, based on published measures of medication understanding.2, 12 To administer the MUQ, research assistants (RAs) accessed the patient's preadmission medication list from the EHR and used a random number table to select up to 5 prescription medications from the list. If the patient was taking 5 or fewer medications, all of their medications were selected. Saline nasal spray, saline eye drops, herbal products, nutritional supplements, vitamins, and OTC lotions and creams were excluded from testing. The RA provided the brand and generic name of each medication, and then asked the patient for the drug's purpose, strength per unit (eg, 20 mg tablet), number of units taken at a time (eg, 2 tablets), and dosing frequency (eg, twice a day). For drugs prescribed on an as‐needed basis, the RA asked patients for the maximum allowable dose and frequency. Patients were instructed to not refer to a medication list or bottles when responding. The RA documented the patient's responses on the MUQ, along with the dosing information from the EHR for each selected medication.

One clinical pharmacist (MM) scored all MUQ forms by applying a set of scoring rules. Each medication score could range from 0 to 3. The components of the score included indication (1 point), strength (0.5 point), units (0.5 point), and frequency (1 point). The patient's overall MUQ score was an average of the MUQ scores for each tested medication.

Statistical Analysis

We summarized patient characteristics, number of preadmission medications, and MUQ scores using median and interquartile range (IRQ) for continuous variables, and frequencies and proportions for categorical variables. We conducted proportional odds logistic regression (ordinal regression) to estimate the effect of s‐TOFHLA score, other patient characteristics, and number of medications on MUQ scores.13

Important covariates were selected a priori based on clinical significance. These included age (continuous), gender, patient self‐reported race (white, black, other nonwhite), Mini‐Cog score (continuous), primary language (English or Spanish), years of education (continuous), number of preadmission medications (continuous), income (ordinal categories), insurance type (categorical), and study site. Covariates with missing data (household income, health literacy, and years of education) were imputed using multiple imputation techniques.14 The relationship between number of preadmission medications and MUQ scores was found to be nonlinear, and it was modeled using restricted cubic splines.14 We also fit models which treated health literacy and cognition as categorical variables. Results are reported as odds ratios (OR) with 95% confidence intervals (CI). Wald tests were used to test for the statistical significance of predictor variables. Two‐sided P values less than 0.05 were considered statistically significant. All analyses were performed using statistical language R (R Foundation, available at: http://www.r‐project.org).

Baseline Characteristics

Among the 862 patients enrolled in PILL‐CVD, 790 (91.7 %) had at least 1 preadmission medication and were included in this analysis (Table 1). Forty‐seven percent were admitted to VUH (N = 373) and 53% to BWH (N = 417). The median age was 61 (interquartile range [IQR] 52, 71), 77% were white, and 57% were male. Inadequate or marginal health literacy was identified among 11% and 9% of patients, respectively. The median number of preadmission medications was 8 (IQR 5, 11). Patients excluded from the analysis for not having preadmission medications were similar to included patients, except they were more likely to be male (76% vs 57%) and less likely to have health insurance (23% self‐pay vs 4%). (Data available upon request.)

MUQ Scores

The MUQ was administered in approximately 5 minutes. The median MUQ score was 2.5 (IQR 2.2, 2.8) (Table 2); 16.3% of patients scored less than 2. Subjects typically achieved high scores for the domains of indication, units, and frequency, while scores on the strength domain were lower (median = 0.2 [IQR 0.1, 0.4], maximum possible = 0.5).

Predictors of Medication Understanding

Unadjusted relationships of health literacy, cognition, and number of medications with medication understanding are shown in Figure 1 (panels A, B, and C, respectively). The figure demonstrates a linear relationship with both health literacy (Figure 1A) and cognition (Figure 1B), and a nonlinear relationship between number of preadmission medications and MUQ score (Figure 1C).

Figure 1

Adjusted relationships using imputed data for missing covariates are shown in Figure 2. Lower health literacy, cognitive impairment, male gender, and black race were independently associated with lower understanding of preadmission medications. Each 1 point increase in s‐TOFHLA or Mini‐Cog score led to an increase in medication understanding (OR = 1.04; 95% CI, 1.02 to 1.06; P = 0.0001; and OR = 1.24; 95% CI, 1.1 to 1.4; P = 0.001; respectively). Patients with marginal or inadequate health literacy had lower odds of understanding their regimen (OR = 0.53; 95% CI, 0.34 to 0.84; and OR = 0.49; 95% CI, 0.31 to 0.78, respectively) compared to those with adequate health literacy. Impaired cognitive function (Mini‐Cog score <3, indicating dementia) was also associated with lower odds of medication understanding (OR = 0.57; 95% CI, 0.38 to 0.86) compared to those with no cognitive impairment. An increase in the number of preadmission medications (up to 10) was also strongly associated with lower MUQ scores. For each increase by 1 medication, there was a significant decrease in medication understanding, up to 10 medications, beyond which understanding did not significantly decrease further. Patients on 6 medications were about half as likely to understand their medication regimen as patients on only 1 medication (OR = 0.52; 95% CI, 0.36 to 0.75). For patients on 11 medications, the odds of medication understanding were 24% lower than for patients on 6 medications (OR = 0.76; 95% CI, 0.65 to 0.89). Patients' age, years of schooling, and household income were not independently associated with medication understanding. Results were similar using data without multiple imputation.

Figure 2

Examples of Misunderstanding of Common Medications

Table 3 provides examples of incorrect patient responses for several commonly prescribed medications or drug classes, including aspirin, digoxin, nitroglycerin, and HMG‐CoA reductase inhibitors (statins). For aspirin, many patients were not aware of the strength. For digoxin, several participants reported splitting a higher‐strength pill to obtain the prescribed dose, which should not be done given the imprecision of splitting and narrow therapeutic index of this drug. Patients prescribed nitroglycerin sublingual tablets were commonly unable to report the correct dosing and frequency for angina treatment. Medications for cholesterol were often reported as being taken in the morning; this was scored strictly as a frequency error if the medication timing in the EHR was listed as evening or bedtime. We also identified many patients with poor understanding of opioid analgesics, particularly regarding their dosing and frequency.