Hospitalist Practice Models

Survey Instrument

A detailed description of the survey design, sampling strategy, data collection, and response rate calculations is described elsewhere.16 Portions of the 118‐item survey instrument assessed characteristics of the respondents' hospitalist group (12 items), details about their individual work patterns (12 items), and demographics (9 items). Work patterns were evaluated by the average number of clinical work days, consecutive days, hours per month, percentage of work assigned to night duty, and number of patient encounters. Average hours spent on nonclinical work, and the percentage of time allocated for clinical, administrative, teaching, and research activities were solicited. Additional items assessed specific clinical responsibilities, pretax earnings in FY2010, the availability of information technology capabilities, and the adequacy of available resources. Job and specialty satisfaction and 11 satisfaction domain measures were measured using validated scales.1726 Burnout symptoms were measured using a validated single‐item measure.26, 27

Sampling Strategy

We surveyed a national stratified sample of hospitalists in the US and Puerto Rico. We used the largest database of hospitalists (>24,000 names) currently available and maintained by the SHM as our sampling frame. We linked hospitalist employer information to hospital statistics from the American Hospital Association database28 to stratify the sample by number of hospital beds, geographic region, employment model, and specialty training, oversampling pediatric hospitalists due to small numbers. A respondent sample of about 700 hospitalists was calculated to be adequate to detect a 0.5 point difference in job satisfaction scores between subgroups assuming 90% power and alpha of 0.05. However, we sampled a total of 5389 addresses from the database to overcome the traditionally low physician response rates, duplicate sampling, bad addresses, and non‐hospitalists being included in the sampling frame. In addition, 2 multistate hospitalist companies (EmCare, In Compass Health) and 1 for‐profit hospital chain (HCA, Inc) financially sponsored this project with the stipulation that all of their hospitalist employees (n = 884) would be surveyed.

Data Collection

The healthcare consulting firm, Press Ganey, provided support with survey layout and administration following the modified Dillman method.29 Three rounds of coded surveys and solicitation letters from the investigators were mailed 2 weeks apart in November and December 2009. Because of low response rates to the mailed survey, an online survey was created using Survey Monkey and sent to 650 surveyees for whom e‐mail addresses were available, and administered at a kiosk for sample physicians during the SHM 2010 annual meeting.

Data Analysis

Nonresponse bias was measured by comparing characteristics between respondents of separate survey waves.30 We determined the validity of mailing addresses immediately following the survey period by mapping each address using Google, and if the address was a hospital, researching online whether or not the intended recipient was currently employed there. Practice characteristics were compared across 5 model categories distilled from the SHM & Medical Group Management Association survey: local hospitalist‐only group, multistate hospitalist group, multispecialty physician group, employer hospital, and university or medical school. Weighted proportions, means, and medians were calculated to account for oversampling of pediatric hospitalists. Differences in categorical measures were assessed using the chi‐square test and the design‐based F test for comparing weighted data. Weighted means (99% confidence intervals) and medians (interquartile ranges) were calculated. Because each parameter yielded a single outlier value across the 5 practice models, differences across weighted means were assessed using generalized linear models with the single outlier value chosen as the reference mean. Pair‐wise Wilcoxon rank sum test was used to compare median values. In these 4‐way comparisons of means and medians, significance was defined as P value of 0.0125 per Bonferroni correction. A single survey item solicited respondents to choose exactly 4 of 13 considerations most pertinent to job satisfaction. The proportion of respondents who scored 4 on a 5‐point Likert scale of the 11 satisfaction domains and 2 global measures of satisfaction, and burnout symptoms defined as 3 on a 5‐point single item measure were bar‐graphed. Chi‐square statistics were used to evaluate for differences across practice models. Statistical significance was defined by alpha less than 0.05, unless otherwise specified. All analyses were performed using STATA version 11.0 (College Station, TX). This study was approved by the Loyola University Institutional Review Board.

Survey data required cleaning prior to analysis. Missing gender information was imputed using the respondents' name. Responses to the item that asked to indicate the proportion of work dedicated to administrative responsibilities, clinical care, teaching, and research that did not add up to 100% were dropped. Two responses that indicated full‐time equivalent (FTE) of 0%, but whose respondents otherwise completed the survey implying they worked as clinical hospitalists, were replaced with values calculated from the given number of work hours relative to the median work hours in our sample. Out of range or implausible responses to the following items were dropped from analyses: the average number of billable encounters during a typical day or shift, number of shifts performing clinical activities during a typical month, pretax earnings, the year the respondent completed residency training, and the number of whole years practiced as a hospitalist. The proportion of selective item nonresponse was small and we did not, otherwise, impute missing data.

Response Rate

Of the 5389 originally sampled addresses, 1868 were undeliverable. Addresses were further excluded if they appeared in duplicate or were outdated. This yielded a total of 3105 eligible surveyees in the sample. As illustrated in Figure 1, 841 responded to the mailed survey and 5 responded to the Web‐based survey. After rejecting 67 non‐hospitalist respondents and 3 duplicate surveys, a total of 776 surveys were included in the final analysis. The adjusted response rate was 25.6% (776/3035). Members of SHM were more likely to return the survey than nonmembers. The adjusted response rate from hospitalists affiliated with the 3 sponsoring institutions was 6% (40/662). Because these respondents were more likely to be non‐members of SHM, we opted to analyze the responses from the sponsor hospitalists together with the sampled hospitalists. The demographics of the resulting pool of 816 respondents affiliated with over 650 unique hospitalist groups were representative of the original survey frame. We analyzed data from 794 of these who responded to the item indicating their hospitalist practice model. Demographic characteristics of responders and nonresponders to the practice model survey item were similar.

Figure 1

Characteristics of Hospitalists and Their Groups

Table 1 summarizes the characteristics of hospitalist respondents and their organizations by practice model. More (44%) respondents identified their practice model as directly employed by the hospital than other models, including multispecialty physician group (15%), multistate hospitalist group (14%), university or medical school (14%), local hospitalist group (12%), and other (2%). The median age of hospitalist respondents was 42 years, with 6.8 years of mean experience as a hospitalist. One third were women, 84% were married, and 46% had dependent children 6 years old or younger at home. Notably, hospitalists in multistate groups had fewer years of experience, and fewer hospitalists in local and multistate groups were married compared to hospitalists in other practice models.

Several differences in respondent group characteristics by practice model were found. Respondents in multistate hospitalist groups were more likely from the South and Midwest, while respondents from multispecialty groups were likely from the West. More multistate group practices were based in smaller hospitals, while academic hospitalists tended to practice in hospitals with 600 or more beds. Respondents employed by hospitals were more likely to practice at 1 hospital facility only, while local group practices were more likely to practice at 3 or more facilities. The median number of physicians in a hospitalist group was 11 (interquartile range [IQR] 6, 19). Local and multistate groups had fewer hospitalists compared to other models. Nonphysician providers were employed by nearly half of all hospitalist practices. Although almost all groups had access to some information technology, more academic hospitalists had access to electronic order entry, electronic physician notes, electronic clinical guidelines resources and communication technology, while local and multistate groups were least likely to have access to these resources.

Work Pattern Variations

Table 2 further details hospitalist work hours by practice model. The majority of hospitalists (78%) reported their position was full‐time (FTE 1.0), while 13% reported working less than full‐time (FTE <1.0). Only 5% of local group hospitalists worked part‐time, while 20% of multispecialty group hospitalists did. An additional 9% reported FTE >1.0, indicating their work hours exceeded the definition of a full‐time physician in their practice. Among full‐time hospitalists, local group members worked a greater number of shifts per month than employees of multispecialty groups, hospitals, and academic medical centers. Academic hospitalists reported higher numbers of consecutive clinical days worked on average, but fewer night shifts compared to hospitalists employed by multistate groups, multispecialty groups, and hospitals; fewer billable encounters than hospitalists in local and multistate groups; and more nonclinical work hours than hospitalists of any other practice model. Academic hospitalists also spent more time on teaching and research than other practice models. Hospitalists spent 11%‐18% of their time on administrative and committee responsibilities, with the least amount spent by hospitalists in multistate groups and the most in academic practice.

Table 3 tabulates other work pattern characteristics. Most hospitalists indicated that their current clinical work as hospitalists involved the general medical wards (100%), medical consultations (98%), and comanagement with specialists (92%). There were wide differences in participation in comanagement (100%, local groups vs 71%, academic), intensive care unit (ICU) responsibilities (94%, multistate groups vs 27%, academic), and nursing home care (30%, local groups vs 8%, academic). Among activities that are potentially not reimbursable, academic hospitalists were less likely to participate in coordination of patient transfers and code or rapid response teams, while multistate groups were least likely to participate in quality improvement activities. In total, 99% of hospitalists reported participating in at least 1 potentially nonreimbursable clinical activity.

Hospitalist compensation schemes were significantly different across the practice models. Salary‐only schemes were most common among academic hospitalists (47%), while 72% of multistate groups used performance incentives in addition to salary. More local groups used fee‐for‐service compensation than other models. Incentives differed by practice model, with more multistate groups having incentives based on patient satisfaction, while more multispecialty physician groups had incentives based on clinical processes and outcomes than other models. Finally, mean earnings for academic hospitalists were significantly lower than for hospitalists of other practice models. Local and multistate group hospitalists earned more than any other practice model (all P <0.001), and $60,000 more than the lowest compensated academic hospitalists.

Components of Job Satisfaction

Hospitalists' rankings of the most important factors for job satisfaction revealed differences across models (Figure 2). Overall, hospitalists were most likely to consider optimal workload and compensation as important factors for job satisfaction from a list of 13 considerations. Local groups and academics were least likely to rank optimal workload as a top factor, and local group hospitalists were more likely to rank optimal autonomy than those of other models. Academic hospitalists had less concern for substantial pay, and more concern for the variety of tasks they perform and recognition by leaders, than other hospitalists.

Figure 2

Job Satisfaction and Burnout Risk

Differences in the ratings of 4 of the 11 satisfaction and job characteristic domains were found across the practice models (Figure 3). Multispecialty group hospitalists were less satisfied with autonomy and their relationship with patients than other practice models, and along with multistate groups, reported the highest perceived workload. Organizational fairness was rated much higher by local group hospitalists than other practice models. Despite these differences in work patterns and satisfaction, there were no differences found in level of global job satisfaction, specialty satisfaction, or burnout across the practice models. Overall, 62% of respondents reported high job satisfaction (4 on a 1 to 5 scale), and 30% indicated burnout symptoms.

Figure 3

Survey Instrument

A detailed description of the survey design, sampling strategy, data collection, and response rate calculations is described elsewhere.16 Portions of the 118‐item survey instrument assessed characteristics of the respondents' hospitalist group (12 items), details about their individual work patterns (12 items), and demographics (9 items). Work patterns were evaluated by the average number of clinical work days, consecutive days, hours per month, percentage of work assigned to night duty, and number of patient encounters. Average hours spent on nonclinical work, and the percentage of time allocated for clinical, administrative, teaching, and research activities were solicited. Additional items assessed specific clinical responsibilities, pretax earnings in FY2010, the availability of information technology capabilities, and the adequacy of available resources. Job and specialty satisfaction and 11 satisfaction domain measures were measured using validated scales.1726 Burnout symptoms were measured using a validated single‐item measure.26, 27

Sampling Strategy

We surveyed a national stratified sample of hospitalists in the US and Puerto Rico. We used the largest database of hospitalists (>24,000 names) currently available and maintained by the SHM as our sampling frame. We linked hospitalist employer information to hospital statistics from the American Hospital Association database28 to stratify the sample by number of hospital beds, geographic region, employment model, and specialty training, oversampling pediatric hospitalists due to small numbers. A respondent sample of about 700 hospitalists was calculated to be adequate to detect a 0.5 point difference in job satisfaction scores between subgroups assuming 90% power and alpha of 0.05. However, we sampled a total of 5389 addresses from the database to overcome the traditionally low physician response rates, duplicate sampling, bad addresses, and non‐hospitalists being included in the sampling frame. In addition, 2 multistate hospitalist companies (EmCare, In Compass Health) and 1 for‐profit hospital chain (HCA, Inc) financially sponsored this project with the stipulation that all of their hospitalist employees (n = 884) would be surveyed.

Data Collection

The healthcare consulting firm, Press Ganey, provided support with survey layout and administration following the modified Dillman method.29 Three rounds of coded surveys and solicitation letters from the investigators were mailed 2 weeks apart in November and December 2009. Because of low response rates to the mailed survey, an online survey was created using Survey Monkey and sent to 650 surveyees for whom e‐mail addresses were available, and administered at a kiosk for sample physicians during the SHM 2010 annual meeting.

Data Analysis

Nonresponse bias was measured by comparing characteristics between respondents of separate survey waves.30 We determined the validity of mailing addresses immediately following the survey period by mapping each address using Google, and if the address was a hospital, researching online whether or not the intended recipient was currently employed there. Practice characteristics were compared across 5 model categories distilled from the SHM & Medical Group Management Association survey: local hospitalist‐only group, multistate hospitalist group, multispecialty physician group, employer hospital, and university or medical school. Weighted proportions, means, and medians were calculated to account for oversampling of pediatric hospitalists. Differences in categorical measures were assessed using the chi‐square test and the design‐based F test for comparing weighted data. Weighted means (99% confidence intervals) and medians (interquartile ranges) were calculated. Because each parameter yielded a single outlier value across the 5 practice models, differences across weighted means were assessed using generalized linear models with the single outlier value chosen as the reference mean. Pair‐wise Wilcoxon rank sum test was used to compare median values. In these 4‐way comparisons of means and medians, significance was defined as P value of 0.0125 per Bonferroni correction. A single survey item solicited respondents to choose exactly 4 of 13 considerations most pertinent to job satisfaction. The proportion of respondents who scored 4 on a 5‐point Likert scale of the 11 satisfaction domains and 2 global measures of satisfaction, and burnout symptoms defined as 3 on a 5‐point single item measure were bar‐graphed. Chi‐square statistics were used to evaluate for differences across practice models. Statistical significance was defined by alpha less than 0.05, unless otherwise specified. All analyses were performed using STATA version 11.0 (College Station, TX). This study was approved by the Loyola University Institutional Review Board.

Survey data required cleaning prior to analysis. Missing gender information was imputed using the respondents' name. Responses to the item that asked to indicate the proportion of work dedicated to administrative responsibilities, clinical care, teaching, and research that did not add up to 100% were dropped. Two responses that indicated full‐time equivalent (FTE) of 0%, but whose respondents otherwise completed the survey implying they worked as clinical hospitalists, were replaced with values calculated from the given number of work hours relative to the median work hours in our sample. Out of range or implausible responses to the following items were dropped from analyses: the average number of billable encounters during a typical day or shift, number of shifts performing clinical activities during a typical month, pretax earnings, the year the respondent completed residency training, and the number of whole years practiced as a hospitalist. The proportion of selective item nonresponse was small and we did not, otherwise, impute missing data.

Response Rate

Of the 5389 originally sampled addresses, 1868 were undeliverable. Addresses were further excluded if they appeared in duplicate or were outdated. This yielded a total of 3105 eligible surveyees in the sample. As illustrated in Figure 1, 841 responded to the mailed survey and 5 responded to the Web‐based survey. After rejecting 67 non‐hospitalist respondents and 3 duplicate surveys, a total of 776 surveys were included in the final analysis. The adjusted response rate was 25.6% (776/3035). Members of SHM were more likely to return the survey than nonmembers. The adjusted response rate from hospitalists affiliated with the 3 sponsoring institutions was 6% (40/662). Because these respondents were more likely to be non‐members of SHM, we opted to analyze the responses from the sponsor hospitalists together with the sampled hospitalists. The demographics of the resulting pool of 816 respondents affiliated with over 650 unique hospitalist groups were representative of the original survey frame. We analyzed data from 794 of these who responded to the item indicating their hospitalist practice model. Demographic characteristics of responders and nonresponders to the practice model survey item were similar.

Figure 1

Characteristics of Hospitalists and Their Groups

Table 1 summarizes the characteristics of hospitalist respondents and their organizations by practice model. More (44%) respondents identified their practice model as directly employed by the hospital than other models, including multispecialty physician group (15%), multistate hospitalist group (14%), university or medical school (14%), local hospitalist group (12%), and other (2%). The median age of hospitalist respondents was 42 years, with 6.8 years of mean experience as a hospitalist. One third were women, 84% were married, and 46% had dependent children 6 years old or younger at home. Notably, hospitalists in multistate groups had fewer years of experience, and fewer hospitalists in local and multistate groups were married compared to hospitalists in other practice models.

Several differences in respondent group characteristics by practice model were found. Respondents in multistate hospitalist groups were more likely from the South and Midwest, while respondents from multispecialty groups were likely from the West. More multistate group practices were based in smaller hospitals, while academic hospitalists tended to practice in hospitals with 600 or more beds. Respondents employed by hospitals were more likely to practice at 1 hospital facility only, while local group practices were more likely to practice at 3 or more facilities. The median number of physicians in a hospitalist group was 11 (interquartile range [IQR] 6, 19). Local and multistate groups had fewer hospitalists compared to other models. Nonphysician providers were employed by nearly half of all hospitalist practices. Although almost all groups had access to some information technology, more academic hospitalists had access to electronic order entry, electronic physician notes, electronic clinical guidelines resources and communication technology, while local and multistate groups were least likely to have access to these resources.

Work Pattern Variations

Table 2 further details hospitalist work hours by practice model. The majority of hospitalists (78%) reported their position was full‐time (FTE 1.0), while 13% reported working less than full‐time (FTE <1.0). Only 5% of local group hospitalists worked part‐time, while 20% of multispecialty group hospitalists did. An additional 9% reported FTE >1.0, indicating their work hours exceeded the definition of a full‐time physician in their practice. Among full‐time hospitalists, local group members worked a greater number of shifts per month than employees of multispecialty groups, hospitals, and academic medical centers. Academic hospitalists reported higher numbers of consecutive clinical days worked on average, but fewer night shifts compared to hospitalists employed by multistate groups, multispecialty groups, and hospitals; fewer billable encounters than hospitalists in local and multistate groups; and more nonclinical work hours than hospitalists of any other practice model. Academic hospitalists also spent more time on teaching and research than other practice models. Hospitalists spent 11%‐18% of their time on administrative and committee responsibilities, with the least amount spent by hospitalists in multistate groups and the most in academic practice.

Table 3 tabulates other work pattern characteristics. Most hospitalists indicated that their current clinical work as hospitalists involved the general medical wards (100%), medical consultations (98%), and comanagement with specialists (92%). There were wide differences in participation in comanagement (100%, local groups vs 71%, academic), intensive care unit (ICU) responsibilities (94%, multistate groups vs 27%, academic), and nursing home care (30%, local groups vs 8%, academic). Among activities that are potentially not reimbursable, academic hospitalists were less likely to participate in coordination of patient transfers and code or rapid response teams, while multistate groups were least likely to participate in quality improvement activities. In total, 99% of hospitalists reported participating in at least 1 potentially nonreimbursable clinical activity.

Hospitalist compensation schemes were significantly different across the practice models. Salary‐only schemes were most common among academic hospitalists (47%), while 72% of multistate groups used performance incentives in addition to salary. More local groups used fee‐for‐service compensation than other models. Incentives differed by practice model, with more multistate groups having incentives based on patient satisfaction, while more multispecialty physician groups had incentives based on clinical processes and outcomes than other models. Finally, mean earnings for academic hospitalists were significantly lower than for hospitalists of other practice models. Local and multistate group hospitalists earned more than any other practice model (all P <0.001), and $60,000 more than the lowest compensated academic hospitalists.

Components of Job Satisfaction

Hospitalists' rankings of the most important factors for job satisfaction revealed differences across models (Figure 2). Overall, hospitalists were most likely to consider optimal workload and compensation as important factors for job satisfaction from a list of 13 considerations. Local groups and academics were least likely to rank optimal workload as a top factor, and local group hospitalists were more likely to rank optimal autonomy than those of other models. Academic hospitalists had less concern for substantial pay, and more concern for the variety of tasks they perform and recognition by leaders, than other hospitalists.

Figure 2

Job Satisfaction and Burnout Risk

Differences in the ratings of 4 of the 11 satisfaction and job characteristic domains were found across the practice models (Figure 3). Multispecialty group hospitalists were less satisfied with autonomy and their relationship with patients than other practice models, and along with multistate groups, reported the highest perceived workload. Organizational fairness was rated much higher by local group hospitalists than other practice models. Despite these differences in work patterns and satisfaction, there were no differences found in level of global job satisfaction, specialty satisfaction, or burnout across the practice models. Overall, 62% of respondents reported high job satisfaction (4 on a 1 to 5 scale), and 30% indicated burnout symptoms.

Figure 3

Survey Instrument

A detailed description of the survey design, sampling strategy, data collection, and response rate calculations is described elsewhere.16 Portions of the 118‐item survey instrument assessed characteristics of the respondents' hospitalist group (12 items), details about their individual work patterns (12 items), and demographics (9 items). Work patterns were evaluated by the average number of clinical work days, consecutive days, hours per month, percentage of work assigned to night duty, and number of patient encounters. Average hours spent on nonclinical work, and the percentage of time allocated for clinical, administrative, teaching, and research activities were solicited. Additional items assessed specific clinical responsibilities, pretax earnings in FY2010, the availability of information technology capabilities, and the adequacy of available resources. Job and specialty satisfaction and 11 satisfaction domain measures were measured using validated scales.1726 Burnout symptoms were measured using a validated single‐item measure.26, 27

Sampling Strategy

We surveyed a national stratified sample of hospitalists in the US and Puerto Rico. We used the largest database of hospitalists (>24,000 names) currently available and maintained by the SHM as our sampling frame. We linked hospitalist employer information to hospital statistics from the American Hospital Association database28 to stratify the sample by number of hospital beds, geographic region, employment model, and specialty training, oversampling pediatric hospitalists due to small numbers. A respondent sample of about 700 hospitalists was calculated to be adequate to detect a 0.5 point difference in job satisfaction scores between subgroups assuming 90% power and alpha of 0.05. However, we sampled a total of 5389 addresses from the database to overcome the traditionally low physician response rates, duplicate sampling, bad addresses, and non‐hospitalists being included in the sampling frame. In addition, 2 multistate hospitalist companies (EmCare, In Compass Health) and 1 for‐profit hospital chain (HCA, Inc) financially sponsored this project with the stipulation that all of their hospitalist employees (n = 884) would be surveyed.

Data Collection

The healthcare consulting firm, Press Ganey, provided support with survey layout and administration following the modified Dillman method.29 Three rounds of coded surveys and solicitation letters from the investigators were mailed 2 weeks apart in November and December 2009. Because of low response rates to the mailed survey, an online survey was created using Survey Monkey and sent to 650 surveyees for whom e‐mail addresses were available, and administered at a kiosk for sample physicians during the SHM 2010 annual meeting.

Data Analysis

Nonresponse bias was measured by comparing characteristics between respondents of separate survey waves.30 We determined the validity of mailing addresses immediately following the survey period by mapping each address using Google, and if the address was a hospital, researching online whether or not the intended recipient was currently employed there. Practice characteristics were compared across 5 model categories distilled from the SHM & Medical Group Management Association survey: local hospitalist‐only group, multistate hospitalist group, multispecialty physician group, employer hospital, and university or medical school. Weighted proportions, means, and medians were calculated to account for oversampling of pediatric hospitalists. Differences in categorical measures were assessed using the chi‐square test and the design‐based F test for comparing weighted data. Weighted means (99% confidence intervals) and medians (interquartile ranges) were calculated. Because each parameter yielded a single outlier value across the 5 practice models, differences across weighted means were assessed using generalized linear models with the single outlier value chosen as the reference mean. Pair‐wise Wilcoxon rank sum test was used to compare median values. In these 4‐way comparisons of means and medians, significance was defined as P value of 0.0125 per Bonferroni correction. A single survey item solicited respondents to choose exactly 4 of 13 considerations most pertinent to job satisfaction. The proportion of respondents who scored 4 on a 5‐point Likert scale of the 11 satisfaction domains and 2 global measures of satisfaction, and burnout symptoms defined as 3 on a 5‐point single item measure were bar‐graphed. Chi‐square statistics were used to evaluate for differences across practice models. Statistical significance was defined by alpha less than 0.05, unless otherwise specified. All analyses were performed using STATA version 11.0 (College Station, TX). This study was approved by the Loyola University Institutional Review Board.

Survey data required cleaning prior to analysis. Missing gender information was imputed using the respondents' name. Responses to the item that asked to indicate the proportion of work dedicated to administrative responsibilities, clinical care, teaching, and research that did not add up to 100% were dropped. Two responses that indicated full‐time equivalent (FTE) of 0%, but whose respondents otherwise completed the survey implying they worked as clinical hospitalists, were replaced with values calculated from the given number of work hours relative to the median work hours in our sample. Out of range or implausible responses to the following items were dropped from analyses: the average number of billable encounters during a typical day or shift, number of shifts performing clinical activities during a typical month, pretax earnings, the year the respondent completed residency training, and the number of whole years practiced as a hospitalist. The proportion of selective item nonresponse was small and we did not, otherwise, impute missing data.

Response Rate

Of the 5389 originally sampled addresses, 1868 were undeliverable. Addresses were further excluded if they appeared in duplicate or were outdated. This yielded a total of 3105 eligible surveyees in the sample. As illustrated in Figure 1, 841 responded to the mailed survey and 5 responded to the Web‐based survey. After rejecting 67 non‐hospitalist respondents and 3 duplicate surveys, a total of 776 surveys were included in the final analysis. The adjusted response rate was 25.6% (776/3035). Members of SHM were more likely to return the survey than nonmembers. The adjusted response rate from hospitalists affiliated with the 3 sponsoring institutions was 6% (40/662). Because these respondents were more likely to be non‐members of SHM, we opted to analyze the responses from the sponsor hospitalists together with the sampled hospitalists. The demographics of the resulting pool of 816 respondents affiliated with over 650 unique hospitalist groups were representative of the original survey frame. We analyzed data from 794 of these who responded to the item indicating their hospitalist practice model. Demographic characteristics of responders and nonresponders to the practice model survey item were similar.

Figure 1

Characteristics of Hospitalists and Their Groups

Table 1 summarizes the characteristics of hospitalist respondents and their organizations by practice model. More (44%) respondents identified their practice model as directly employed by the hospital than other models, including multispecialty physician group (15%), multistate hospitalist group (14%), university or medical school (14%), local hospitalist group (12%), and other (2%). The median age of hospitalist respondents was 42 years, with 6.8 years of mean experience as a hospitalist. One third were women, 84% were married, and 46% had dependent children 6 years old or younger at home. Notably, hospitalists in multistate groups had fewer years of experience, and fewer hospitalists in local and multistate groups were married compared to hospitalists in other practice models.

Several differences in respondent group characteristics by practice model were found. Respondents in multistate hospitalist groups were more likely from the South and Midwest, while respondents from multispecialty groups were likely from the West. More multistate group practices were based in smaller hospitals, while academic hospitalists tended to practice in hospitals with 600 or more beds. Respondents employed by hospitals were more likely to practice at 1 hospital facility only, while local group practices were more likely to practice at 3 or more facilities. The median number of physicians in a hospitalist group was 11 (interquartile range [IQR] 6, 19). Local and multistate groups had fewer hospitalists compared to other models. Nonphysician providers were employed by nearly half of all hospitalist practices. Although almost all groups had access to some information technology, more academic hospitalists had access to electronic order entry, electronic physician notes, electronic clinical guidelines resources and communication technology, while local and multistate groups were least likely to have access to these resources.

Work Pattern Variations

Table 2 further details hospitalist work hours by practice model. The majority of hospitalists (78%) reported their position was full‐time (FTE 1.0), while 13% reported working less than full‐time (FTE <1.0). Only 5% of local group hospitalists worked part‐time, while 20% of multispecialty group hospitalists did. An additional 9% reported FTE >1.0, indicating their work hours exceeded the definition of a full‐time physician in their practice. Among full‐time hospitalists, local group members worked a greater number of shifts per month than employees of multispecialty groups, hospitals, and academic medical centers. Academic hospitalists reported higher numbers of consecutive clinical days worked on average, but fewer night shifts compared to hospitalists employed by multistate groups, multispecialty groups, and hospitals; fewer billable encounters than hospitalists in local and multistate groups; and more nonclinical work hours than hospitalists of any other practice model. Academic hospitalists also spent more time on teaching and research than other practice models. Hospitalists spent 11%‐18% of their time on administrative and committee responsibilities, with the least amount spent by hospitalists in multistate groups and the most in academic practice.

Table 3 tabulates other work pattern characteristics. Most hospitalists indicated that their current clinical work as hospitalists involved the general medical wards (100%), medical consultations (98%), and comanagement with specialists (92%). There were wide differences in participation in comanagement (100%, local groups vs 71%, academic), intensive care unit (ICU) responsibilities (94%, multistate groups vs 27%, academic), and nursing home care (30%, local groups vs 8%, academic). Among activities that are potentially not reimbursable, academic hospitalists were less likely to participate in coordination of patient transfers and code or rapid response teams, while multistate groups were least likely to participate in quality improvement activities. In total, 99% of hospitalists reported participating in at least 1 potentially nonreimbursable clinical activity.

Hospitalist compensation schemes were significantly different across the practice models. Salary‐only schemes were most common among academic hospitalists (47%), while 72% of multistate groups used performance incentives in addition to salary. More local groups used fee‐for‐service compensation than other models. Incentives differed by practice model, with more multistate groups having incentives based on patient satisfaction, while more multispecialty physician groups had incentives based on clinical processes and outcomes than other models. Finally, mean earnings for academic hospitalists were significantly lower than for hospitalists of other practice models. Local and multistate group hospitalists earned more than any other practice model (all P <0.001), and $60,000 more than the lowest compensated academic hospitalists.

Components of Job Satisfaction

Hospitalists' rankings of the most important factors for job satisfaction revealed differences across models (Figure 2). Overall, hospitalists were most likely to consider optimal workload and compensation as important factors for job satisfaction from a list of 13 considerations. Local groups and academics were least likely to rank optimal workload as a top factor, and local group hospitalists were more likely to rank optimal autonomy than those of other models. Academic hospitalists had less concern for substantial pay, and more concern for the variety of tasks they perform and recognition by leaders, than other hospitalists.

Figure 2

Job Satisfaction and Burnout Risk

Differences in the ratings of 4 of the 11 satisfaction and job characteristic domains were found across the practice models (Figure 3). Multispecialty group hospitalists were less satisfied with autonomy and their relationship with patients than other practice models, and along with multistate groups, reported the highest perceived workload. Organizational fairness was rated much higher by local group hospitalists than other practice models. Despite these differences in work patterns and satisfaction, there were no differences found in level of global job satisfaction, specialty satisfaction, or burnout across the practice models. Overall, 62% of respondents reported high job satisfaction (4 on a 1 to 5 scale), and 30% indicated burnout symptoms.

Figure 3