Research Report
Characteristics of Internal Medicine Physicians and Their Practices That Have Differential Impacts on Their Maintenance of Certification Rebecca S. Lipner, PhD, and Bradley G. Brossman, PhD
Abstract Purpose One way to ensure quality of health care in the United States is through maintenance of certification (MOC). In this study, the authors explored whether participation in the internal medicine MOC program varies by physician-level characteristics, professional activities, and the size and location of the practice. They also sought to determine which component of MOC was incomplete for physicians who participated but did not complete the program. Method The authors used a theoretical realist approach to understand whether
C
oncerns over the quality of health care in the United States continue to grow. One effort to address these concerns is to increase physicians’ accountability to the public1 through board certification and maintenance of certification (MOC), a nongovernmental voluntary process led by the medical profession.2 The 24 member boards of the American Board of Medical Specialties (ABMS) have adopted an MOC program to ensure that practicing physicians provide quality patient care by being up-to-date in their knowledge and clinical skills, and committed to professionalism, including interpersonal communication, lifelong learning, and quality improvement. Dr. Lipner is senior vice president, Department of Evaluation, Research, and Development, American Board of Internal Medicine, Philadelphia, Pennsylvania. Dr. Brossman is a psychometrician, Department of Evaluation, Research, and Development, American Board of Internal Medicine, Philadelphia, Pennsylvania. Correspondence should be addressed to Dr. Brossman, 510 Walnut St., Suite 1700, Philadelphia, PA 19106-3699; telephone: (215) 399-4249; e-mail: [email protected]. Acad Med. 2015;90:82–87. First published online August 5, 2014 doi: 10.1097/ACM.0000000000000445
82
participation in the American Board of Internal Medicine MOC program varies according to physician and practice characteristics. The data came from a study sample that consisted of all physicians whose original certification was granted in internal medicine from 1990 through 1999; the study was conducted in 2013. Chi-square tests of independence and a multinomial logistic regression were conducted to determine which physician-level characteristics, professional activities, and practice characteristics were significantly associated with MOC participation.
Results Results showed that physicians who completed MOC tended to have higher certification exam scores; were younger; were U.S. medical graduates; practiced as subspecialists and in the Midwest; spent more time in patient care, teaching, or administration; worked in nonsolo practices; or were employed in counties with less than 20% of persons in poverty.
Beginning with physicians who initially certified in 1990, the American Board of Internal Medicine (ABIM), a member board of ABMS, changed its board certification policy from time-unlimited to time-limited. Under this new policy, all general internists and subspecialists must complete the MOC program every 10 years to maintain their internal medicine and/or subspecialty certification.
success in MOC and performance in practice,6–8 and several other studies have demonstrated a positive relationship between initial certification and performance in practice.9 The realist approach encourages researchers to consider how the program may work differently for different physicians and contexts. In a study of family physicians, researchers found that although the MOC participation rate was high, those who practice in underserved areas, are solo practitioners, or graduated from international medical schools were more likely to have lapsed certification.10
Because the MOC program is a complex intervention that can produce different outcomes under different contexts, we use the theoretical realist approach to evaluation.3 As in any program evaluation, we seek to understand whether the program is working the way it is supposed to and whether it is practical and valid. A national survey found that internists who maintain certification said they did so largely for professional reasons and to keep current with changes in patient care, whereas those who chose not to maintain certification noted that it took too much time, was too expensive, and was not required for their employment.4 Similar findings have been shown for general pediatric physicians.5 Evidence of program validity has been demonstrated through a positive relationship between
Conclusions As certifying boards evaluate their programs, they need to continuously improve their features to assure the public that physicians maintaining certification are providing high-quality patient care.
To that end, we set out to understand whether participation in the internal medicine MOC program varies according to physicians’ characteristics (e.g., general internist or subspecialist, age, gender, medical school origin, country of birth), their professional activities (patient care, teaching, research, administration), the size of their practice (solo or nonsolo), and the practice’s geographic characteristics (e.g., census region and poverty areas). A secondary purpose was to determine which component of MOC was not completed for physicians who participated but did not complete the MOC program.
Academic Medicine, Vol. 90, No. 1 / January 2015
Research Report
Method
MOC program The basic goal of MOC is to ensure the public that physicians are maintaining competence throughout their careers. MOC is based on the six Accreditation Council for Graduate Medical Education and ABMS competency frameworks that attempt to measure many aspects of good patient care. Its structure has four components: evidence of good professional behavior, including good interpersonal and communication skills; self-assessment of medical knowledge; clinical problem solving; and application of quality improvement principles in practice. The goal, structure, and evidence of the ABMS MOC program are described more thoroughly elsewhere.11,12 Data and study sample The study sample consisted of all physicians whose original certification was granted in internal medicine between 1990 and 1999. The physicians in this time frame had certificates that were valid for only 10 years (i.e., time-limited). By restricting analyses to this population of physicians, it was possible to accurately classify physicians as those who completed MOC (MOC Participating and Completed), those who participated but did not yet complete MOC (MOC Participating but Not Completed), and those not participating in MOC (MOC Not Participating). These physicians were expected to have completed MOC by 2009 at the latest in order to not have a gap in their certification status. Physicians known to be deceased, retired, not in good standing, or 75 years or older were excluded from any subsequent analyses. In 2013, four primary data files were merged for the analyses: the ABIM physician-level file, the American Medical Association (AMA) Physician Masterfile, the Area Resource File (ARF), and the Health Resources and Services Administration (HRSA) Health Profession Shortage Areas and Medically Underserved Areas file. The ABIM physician-level dataset was first merged with the AMA Masterfile to obtain office zip codes. The resulting file was subsequently merged with the ARF via the state and county Federal Information Processing Standard code and with the HRSA file via the 2010 zip code tabulation areas (ZCTA) corresponding
Academic Medicine, Vol. 90, No. 1 / January 2015
to each physician’s office zip code. Zip codes and ZCTA codes are similar, with the differentiation that ZCTA codes are typically equal to the most frequently occurring zip code within a given census block. Our rationale for including the variables we used in this study was their previous inclusion in other similar study models predicting performance.10,13 The ABIM physician dataset included physicians’ characteristics (e.g., age, gender, birth country, medical school, and certification status) as well as initial first-attempt general internal medicine certification scores scaled to be comparable across administration years. Also included were responses to the ABIM Practice Characteristics Survey, given to all physicians at the time they enroll in MOC and updated every 18 months thereafter. In this study, we used the most recent responses. The ARF is a county-level health resource information file that contains 6,000 geographic and healthrelated variables, and the HRSA file is a ZCTA-level file that contains variables related to primary care Health Profession Shortage Areas (HPSA) and Medically Underserved Areas and Populations (MUA/P). We used the HRSA file from January 25, 2013 (HPSA and MUA/P designations may change daily). Specific geographic variables were selected from the ARF and the HRSA file for the analyses. Statistical analyses The statistical analyses consisted of descriptive statistics, chi-square tests of independence, and a multinomial logistic regression model. The chi-squares tested the association between MOC status (MOC Participating and Completed, MOC Participating but Not Completed, and MOC Not Participating) and categorical variables one at a time. Given that it was possible to identify clusters of correlated and conceptually related independent variables (e.g., HPSA, percentage of persons below poverty), we also used a multinomial logistic regression model to test the association between MOC status and both categorical and continuous variables after controlling for the effects of the other potentially confounding variables. Statistical analyses were conducted using SAS version 9.3 (SAS Institute, Cary, North Carolina). Our study did not require institutional
review board approval; when physicians enroll in the MOC program, they enter into a business associate agreement that allows the ABIM to use their data—only at an aggregate level—for research purposes.14 Physicians were not given incentives for this research study. All physicians have the opportunity to opt out of being included in research studies without any negative consequences. Results
MOC participation rates The dataset comprised 54,967 physicians. Table 1 displays the frequency and percentage of the 1990–1999 cohorts by their MOC status. Overall, 48,795 physicians (89%) participated in MOC. Of those, 43,038 (78%) completed their first 10-year cycle of the MOC process on time, and 5,757 (11%) participated but did not complete the process on time. The remaining 6,172 (11%) chose not to participate in MOC. Of the 5,757 physicians who partici pated but did not complete MOC (Table 2), 1,523 (26%) failed the secure examination, 3,534 (61%) never completed it, and 700 (12%) passed the exam but did not meet other MOC program requirements (e.g., had an invalid medical license, failed to complete the MOC self-assessment components). It should be noted that physicians who failed or did not complete the secure exam were included in those categories, even if they additionally did not meet the other MOC program requirements. Chi-square tests Chi-square tests of independence assessed the association between various categorical variables and MOC status. The frequency and percentage of physicians within each combination of MOC status and category, along with
Table 1 Maintenance of Certification (MOC) Participation Status From a 2013 Study for All 54,967 Internists Originally Certified From 1990 Through 1999 MOC status Participating Completed
No. (%) 48,795 (89) 43,038 (78)
Not completed
5,757 (11)
Not participating
6,172 (11)
83
Research Report
Table 2 Reasons From a 2013 Study That 5,757 Internists (Originally Certified From 1990 Through 1999) Who Enrolled in Maintenance of Certification (MOC) Did Not Complete MOC Reason Failed exam Did not take exam
No. (%) 1,523 (26) 3,534 (61)
Other
700 (12)
corresponding chi-square P values for each variable, are displayed in Table 3. Overall, most of the categorical variables under consideration revealed a significant relationship with MOC status. Physicians who participated and completed MOC tended to have a subspecialty; be younger (between the ages of 35 and 55); graduate from a U.S. medical school; spend more time in patient care, teaching, or administration; work in a group practice; not work in a HPSA; work in an MUA/P; practice in the Midwest; and be employed in counties with less than 20% of persons in poverty. Physicians who did not participate in MOC tended to be general internists; be older (between the ages of 65 and 75); graduate from international medical schools; live in an HPSA; or not work in an MUA/P. Multinomial logistic regression Because chi-square analyses tested the relationship between MOC status and each categorical variable one at a time, we also estimated a multinomial logistic regression model to test the relationship between MOC status and each of these variables after controlling for all other variables in the model. We also added the physicians’ original internal medicine certification exam scores to the model as a predictor of MOC status, and treated age as a continuous variable. Results for the regression model are displayed in Table 4. Estimated parameters include odds ratios and 95% confidence limits for each effect, as well as global P values. Because of model overspecification, we estimated parameters for N − 1 categories for each categorical variable (N represents the total number of categories); the final category served as the reference group for model estimation purposes. Similarly, we estimated effects for two of the three
84
MOC statuses: MOC Participating and Completed, and MOC Participating but Not Completed. The MOC Not Participating status served as the reference group. Overall, most of the variables in the model revealed significant relationships with MOC status. Physicians who participated and completed MOC tended to have higher internal medicine certification exam scores; practice in a subspecialty field; work in a nonsolo practice; spend more time in patient care and on administrative activities; and practice in the Midwest. Furthermore, the multinomial logistic regression results showed that women tended to have higher MOC completion rates than did men, which was not the case for the chi-square tests. This is attributable to the logistic regression procedure controlling for possible confounding variables, such as age. Along similar lines, whereas HPSA, MUA/P, and poverty level all showed significant chi-square associations, these variables were not significant in the regression model, presumably because of overlap between these variables. The 5,757 physicians who participated but did not complete MOC tended to have lower initial internal medicine certification exam scores than both physicians who completed MOC and physicians who did not participate in MOC. Of those, 3,534 (61%) never attempted the MOC exam; they tended to work in solo practice. Finally, physicians who did not participate in MOC tended to have lower certification exam scores than physicians who participated and completed MOC, but higher scores than physicians who participated but did not complete MOC. These physicians also tended to be general internists and work in a solo practice. Discussion
In this study, we examined the relationship between physician and practice characteristics and MOC status. Overall, the majority of the 54,967 internists with time-limited certificates (48,795; 89%) are participating in MOC for either their subspecialty or general internal medicine certificate (or both). Those who do not participate in MOC are more likely to be general internists, be older (between the ages of 65 and 75),
practice in solo practice, be international medical graduates, be males (after controlling for age and other relevant characteristics), have lower certification scores, and not practice in the Midwest. We know that those in solo practice tend to have lower certification exam scores12 and conjecture that they may be more reticent to expose themselves to potential failure, especially if MOC is not needed for employment. Poverty level of the practice area was not significant after controlling for similar variables, but the trend is in the expected direction, such that those not participating in MOC are more likely to practice in areas with a higher poverty level. We did not find differences between metropolitan versus nonmetropolitan areas because of the relatively small number of internists, once certified, practicing in completely rural areas; most of the internists classified as nonmetropolitan did not practice in completely rural areas and were therefore similar to the metropolitan internists. Not surprisingly, we found that participation in MOC in the Midwest region is higher than in other parts of the country. This finding could be related to the high quality and lower cost of patient care in this region, and perhaps maintaining one’s certification is most desirable.15 Interestingly, we found that those who participate but do not complete MOC have lower certification scores than those who never participate. This may explain why the most common reason for not completing MOC is because the physician failed or did not take the recertification exam (Table 2). The finding that more general internists than subspecialists do not participate in MOC is not surprising; a previous study showed that the rate of attrition from general internal medicine to other medical fields could be as high as 21%, and therefore those who change fields would no longer maintain their internal medicine certification.4 There are several limitations in this study. First, to get a clear picture of participation rates, we had to limit our findings to specific cohorts. Second, we were not able to match all physicians to the AMA database and had to rely on ABIM addresses for geographic location. Some were home zip codes and may differ from the office zip code, so an accurate classification of geographic characteristics may not be possible for these cases. However, through sensitivity analysis
Academic Medicine, Vol. 90, No. 1 / January 2015
Research Report
Table 3 Chi-Square Tests of Independence to Show Associations With Maintenance of Certification Status for 54,967 Internists Originally Certified From 1990 Through 1999 in a 2013 Study Participated and completed, no. (%)
Participated but did not complete, no. (%)
Did not participate, no. (%)
Specialty General internist
24,198 (73.56)
3,878 (11.79)
4,819 (14.65)
Subspecialist
18,840 (85.36)
1,879 (8.51)
1,353 (6.13)
35–44
5,796 (79.81)
645 (8.88)
821 (11.31)
45–54
29,999 (80.28)
3,455 (9.25)
3,913 (10.47)
55–64
6,633 (71.64)
1,446 (15.62)
1,180 (12.74)
65–74
610 (56.53)
211 (19.56)
258 (23.91)
Factor
Characteristics of Internal Medicine Physicians and Their Practices That Have Differential Impacts on Their Maintenance of Certification Rebecca S. Lipner, PhD, and Bradley G. Brossman, PhD
Abstract Purpose One way to ensure quality of health care in the United States is through maintenance of certification (MOC). In this study, the authors explored whether participation in the internal medicine MOC program varies by physician-level characteristics, professional activities, and the size and location of the practice. They also sought to determine which component of MOC was incomplete for physicians who participated but did not complete the program. Method The authors used a theoretical realist approach to understand whether
C
oncerns over the quality of health care in the United States continue to grow. One effort to address these concerns is to increase physicians’ accountability to the public1 through board certification and maintenance of certification (MOC), a nongovernmental voluntary process led by the medical profession.2 The 24 member boards of the American Board of Medical Specialties (ABMS) have adopted an MOC program to ensure that practicing physicians provide quality patient care by being up-to-date in their knowledge and clinical skills, and committed to professionalism, including interpersonal communication, lifelong learning, and quality improvement. Dr. Lipner is senior vice president, Department of Evaluation, Research, and Development, American Board of Internal Medicine, Philadelphia, Pennsylvania. Dr. Brossman is a psychometrician, Department of Evaluation, Research, and Development, American Board of Internal Medicine, Philadelphia, Pennsylvania. Correspondence should be addressed to Dr. Brossman, 510 Walnut St., Suite 1700, Philadelphia, PA 19106-3699; telephone: (215) 399-4249; e-mail: [email protected]. Acad Med. 2015;90:82–87. First published online August 5, 2014 doi: 10.1097/ACM.0000000000000445
82
participation in the American Board of Internal Medicine MOC program varies according to physician and practice characteristics. The data came from a study sample that consisted of all physicians whose original certification was granted in internal medicine from 1990 through 1999; the study was conducted in 2013. Chi-square tests of independence and a multinomial logistic regression were conducted to determine which physician-level characteristics, professional activities, and practice characteristics were significantly associated with MOC participation.
Results Results showed that physicians who completed MOC tended to have higher certification exam scores; were younger; were U.S. medical graduates; practiced as subspecialists and in the Midwest; spent more time in patient care, teaching, or administration; worked in nonsolo practices; or were employed in counties with less than 20% of persons in poverty.
Beginning with physicians who initially certified in 1990, the American Board of Internal Medicine (ABIM), a member board of ABMS, changed its board certification policy from time-unlimited to time-limited. Under this new policy, all general internists and subspecialists must complete the MOC program every 10 years to maintain their internal medicine and/or subspecialty certification.
success in MOC and performance in practice,6–8 and several other studies have demonstrated a positive relationship between initial certification and performance in practice.9 The realist approach encourages researchers to consider how the program may work differently for different physicians and contexts. In a study of family physicians, researchers found that although the MOC participation rate was high, those who practice in underserved areas, are solo practitioners, or graduated from international medical schools were more likely to have lapsed certification.10
Because the MOC program is a complex intervention that can produce different outcomes under different contexts, we use the theoretical realist approach to evaluation.3 As in any program evaluation, we seek to understand whether the program is working the way it is supposed to and whether it is practical and valid. A national survey found that internists who maintain certification said they did so largely for professional reasons and to keep current with changes in patient care, whereas those who chose not to maintain certification noted that it took too much time, was too expensive, and was not required for their employment.4 Similar findings have been shown for general pediatric physicians.5 Evidence of program validity has been demonstrated through a positive relationship between
Conclusions As certifying boards evaluate their programs, they need to continuously improve their features to assure the public that physicians maintaining certification are providing high-quality patient care.
To that end, we set out to understand whether participation in the internal medicine MOC program varies according to physicians’ characteristics (e.g., general internist or subspecialist, age, gender, medical school origin, country of birth), their professional activities (patient care, teaching, research, administration), the size of their practice (solo or nonsolo), and the practice’s geographic characteristics (e.g., census region and poverty areas). A secondary purpose was to determine which component of MOC was not completed for physicians who participated but did not complete the MOC program.
Academic Medicine, Vol. 90, No. 1 / January 2015
Research Report
Method
MOC program The basic goal of MOC is to ensure the public that physicians are maintaining competence throughout their careers. MOC is based on the six Accreditation Council for Graduate Medical Education and ABMS competency frameworks that attempt to measure many aspects of good patient care. Its structure has four components: evidence of good professional behavior, including good interpersonal and communication skills; self-assessment of medical knowledge; clinical problem solving; and application of quality improvement principles in practice. The goal, structure, and evidence of the ABMS MOC program are described more thoroughly elsewhere.11,12 Data and study sample The study sample consisted of all physicians whose original certification was granted in internal medicine between 1990 and 1999. The physicians in this time frame had certificates that were valid for only 10 years (i.e., time-limited). By restricting analyses to this population of physicians, it was possible to accurately classify physicians as those who completed MOC (MOC Participating and Completed), those who participated but did not yet complete MOC (MOC Participating but Not Completed), and those not participating in MOC (MOC Not Participating). These physicians were expected to have completed MOC by 2009 at the latest in order to not have a gap in their certification status. Physicians known to be deceased, retired, not in good standing, or 75 years or older were excluded from any subsequent analyses. In 2013, four primary data files were merged for the analyses: the ABIM physician-level file, the American Medical Association (AMA) Physician Masterfile, the Area Resource File (ARF), and the Health Resources and Services Administration (HRSA) Health Profession Shortage Areas and Medically Underserved Areas file. The ABIM physician-level dataset was first merged with the AMA Masterfile to obtain office zip codes. The resulting file was subsequently merged with the ARF via the state and county Federal Information Processing Standard code and with the HRSA file via the 2010 zip code tabulation areas (ZCTA) corresponding
Academic Medicine, Vol. 90, No. 1 / January 2015
to each physician’s office zip code. Zip codes and ZCTA codes are similar, with the differentiation that ZCTA codes are typically equal to the most frequently occurring zip code within a given census block. Our rationale for including the variables we used in this study was their previous inclusion in other similar study models predicting performance.10,13 The ABIM physician dataset included physicians’ characteristics (e.g., age, gender, birth country, medical school, and certification status) as well as initial first-attempt general internal medicine certification scores scaled to be comparable across administration years. Also included were responses to the ABIM Practice Characteristics Survey, given to all physicians at the time they enroll in MOC and updated every 18 months thereafter. In this study, we used the most recent responses. The ARF is a county-level health resource information file that contains 6,000 geographic and healthrelated variables, and the HRSA file is a ZCTA-level file that contains variables related to primary care Health Profession Shortage Areas (HPSA) and Medically Underserved Areas and Populations (MUA/P). We used the HRSA file from January 25, 2013 (HPSA and MUA/P designations may change daily). Specific geographic variables were selected from the ARF and the HRSA file for the analyses. Statistical analyses The statistical analyses consisted of descriptive statistics, chi-square tests of independence, and a multinomial logistic regression model. The chi-squares tested the association between MOC status (MOC Participating and Completed, MOC Participating but Not Completed, and MOC Not Participating) and categorical variables one at a time. Given that it was possible to identify clusters of correlated and conceptually related independent variables (e.g., HPSA, percentage of persons below poverty), we also used a multinomial logistic regression model to test the association between MOC status and both categorical and continuous variables after controlling for the effects of the other potentially confounding variables. Statistical analyses were conducted using SAS version 9.3 (SAS Institute, Cary, North Carolina). Our study did not require institutional
review board approval; when physicians enroll in the MOC program, they enter into a business associate agreement that allows the ABIM to use their data—only at an aggregate level—for research purposes.14 Physicians were not given incentives for this research study. All physicians have the opportunity to opt out of being included in research studies without any negative consequences. Results
MOC participation rates The dataset comprised 54,967 physicians. Table 1 displays the frequency and percentage of the 1990–1999 cohorts by their MOC status. Overall, 48,795 physicians (89%) participated in MOC. Of those, 43,038 (78%) completed their first 10-year cycle of the MOC process on time, and 5,757 (11%) participated but did not complete the process on time. The remaining 6,172 (11%) chose not to participate in MOC. Of the 5,757 physicians who partici pated but did not complete MOC (Table 2), 1,523 (26%) failed the secure examination, 3,534 (61%) never completed it, and 700 (12%) passed the exam but did not meet other MOC program requirements (e.g., had an invalid medical license, failed to complete the MOC self-assessment components). It should be noted that physicians who failed or did not complete the secure exam were included in those categories, even if they additionally did not meet the other MOC program requirements. Chi-square tests Chi-square tests of independence assessed the association between various categorical variables and MOC status. The frequency and percentage of physicians within each combination of MOC status and category, along with
Table 1 Maintenance of Certification (MOC) Participation Status From a 2013 Study for All 54,967 Internists Originally Certified From 1990 Through 1999 MOC status Participating Completed
No. (%) 48,795 (89) 43,038 (78)
Not completed
5,757 (11)
Not participating
6,172 (11)
83
Research Report
Table 2 Reasons From a 2013 Study That 5,757 Internists (Originally Certified From 1990 Through 1999) Who Enrolled in Maintenance of Certification (MOC) Did Not Complete MOC Reason Failed exam Did not take exam
No. (%) 1,523 (26) 3,534 (61)
Other
700 (12)
corresponding chi-square P values for each variable, are displayed in Table 3. Overall, most of the categorical variables under consideration revealed a significant relationship with MOC status. Physicians who participated and completed MOC tended to have a subspecialty; be younger (between the ages of 35 and 55); graduate from a U.S. medical school; spend more time in patient care, teaching, or administration; work in a group practice; not work in a HPSA; work in an MUA/P; practice in the Midwest; and be employed in counties with less than 20% of persons in poverty. Physicians who did not participate in MOC tended to be general internists; be older (between the ages of 65 and 75); graduate from international medical schools; live in an HPSA; or not work in an MUA/P. Multinomial logistic regression Because chi-square analyses tested the relationship between MOC status and each categorical variable one at a time, we also estimated a multinomial logistic regression model to test the relationship between MOC status and each of these variables after controlling for all other variables in the model. We also added the physicians’ original internal medicine certification exam scores to the model as a predictor of MOC status, and treated age as a continuous variable. Results for the regression model are displayed in Table 4. Estimated parameters include odds ratios and 95% confidence limits for each effect, as well as global P values. Because of model overspecification, we estimated parameters for N − 1 categories for each categorical variable (N represents the total number of categories); the final category served as the reference group for model estimation purposes. Similarly, we estimated effects for two of the three
84
MOC statuses: MOC Participating and Completed, and MOC Participating but Not Completed. The MOC Not Participating status served as the reference group. Overall, most of the variables in the model revealed significant relationships with MOC status. Physicians who participated and completed MOC tended to have higher internal medicine certification exam scores; practice in a subspecialty field; work in a nonsolo practice; spend more time in patient care and on administrative activities; and practice in the Midwest. Furthermore, the multinomial logistic regression results showed that women tended to have higher MOC completion rates than did men, which was not the case for the chi-square tests. This is attributable to the logistic regression procedure controlling for possible confounding variables, such as age. Along similar lines, whereas HPSA, MUA/P, and poverty level all showed significant chi-square associations, these variables were not significant in the regression model, presumably because of overlap between these variables. The 5,757 physicians who participated but did not complete MOC tended to have lower initial internal medicine certification exam scores than both physicians who completed MOC and physicians who did not participate in MOC. Of those, 3,534 (61%) never attempted the MOC exam; they tended to work in solo practice. Finally, physicians who did not participate in MOC tended to have lower certification exam scores than physicians who participated and completed MOC, but higher scores than physicians who participated but did not complete MOC. These physicians also tended to be general internists and work in a solo practice. Discussion
In this study, we examined the relationship between physician and practice characteristics and MOC status. Overall, the majority of the 54,967 internists with time-limited certificates (48,795; 89%) are participating in MOC for either their subspecialty or general internal medicine certificate (or both). Those who do not participate in MOC are more likely to be general internists, be older (between the ages of 65 and 75),
practice in solo practice, be international medical graduates, be males (after controlling for age and other relevant characteristics), have lower certification scores, and not practice in the Midwest. We know that those in solo practice tend to have lower certification exam scores12 and conjecture that they may be more reticent to expose themselves to potential failure, especially if MOC is not needed for employment. Poverty level of the practice area was not significant after controlling for similar variables, but the trend is in the expected direction, such that those not participating in MOC are more likely to practice in areas with a higher poverty level. We did not find differences between metropolitan versus nonmetropolitan areas because of the relatively small number of internists, once certified, practicing in completely rural areas; most of the internists classified as nonmetropolitan did not practice in completely rural areas and were therefore similar to the metropolitan internists. Not surprisingly, we found that participation in MOC in the Midwest region is higher than in other parts of the country. This finding could be related to the high quality and lower cost of patient care in this region, and perhaps maintaining one’s certification is most desirable.15 Interestingly, we found that those who participate but do not complete MOC have lower certification scores than those who never participate. This may explain why the most common reason for not completing MOC is because the physician failed or did not take the recertification exam (Table 2). The finding that more general internists than subspecialists do not participate in MOC is not surprising; a previous study showed that the rate of attrition from general internal medicine to other medical fields could be as high as 21%, and therefore those who change fields would no longer maintain their internal medicine certification.4 There are several limitations in this study. First, to get a clear picture of participation rates, we had to limit our findings to specific cohorts. Second, we were not able to match all physicians to the AMA database and had to rely on ABIM addresses for geographic location. Some were home zip codes and may differ from the office zip code, so an accurate classification of geographic characteristics may not be possible for these cases. However, through sensitivity analysis
Academic Medicine, Vol. 90, No. 1 / January 2015
Research Report
Table 3 Chi-Square Tests of Independence to Show Associations With Maintenance of Certification Status for 54,967 Internists Originally Certified From 1990 Through 1999 in a 2013 Study Participated and completed, no. (%)
Participated but did not complete, no. (%)
Did not participate, no. (%)
Specialty General internist
24,198 (73.56)
3,878 (11.79)
4,819 (14.65)
Subspecialist
18,840 (85.36)
1,879 (8.51)
1,353 (6.13)
35–44
5,796 (79.81)
645 (8.88)
821 (11.31)
45–54
29,999 (80.28)
3,455 (9.25)
3,913 (10.47)
55–64
6,633 (71.64)
1,446 (15.62)
1,180 (12.74)
65–74
610 (56.53)
211 (19.56)
258 (23.91)
Factor
