The effect of ACGME resident duty hour reforms on outcomes of physicians after completion of residency

Abstract

In 2003, work hours for physicians in training (residents) were mandatorily reduced to be no more than 80 hours per week, leading to the hotly debated but unexplored issue of whether physicians today are less well trained as a result of these work hour reforms. Using a unique database of nearly all hospitalizations in Florida during 2000–2009 which were linked to detailed information on the medical training history of the physician of record for each hospitalization, we studied whether hospital mortality and length of stay for internists varied according to the number of years an internist was exposed to the 2003 duty hour regulations during their own residency. Using a difference-in-difference analysis which compared trends in outcomes of junior physicians pre- and post-2003 to a control group of senior physicians who were not exposed to these reforms during their own residency, we found that the 2003 reforms did not affect the quality of physician training reflected by hospital mortality and length of stay.

INTRODUCTION

In 2003, amidst rising concerns about medical errors due to long resident work hours, the Accreditation Council for Graduate Medical Education (ACGME) implemented national duty hour regulations which established a maximum 80-hour work week and reduced shift lengths to no longer than 30 consecutive hours.¹ In 2011, shift lengths were further limited to a maximum of 16 consecutive hours for first year trainees and 28 hours for other trainees.²

Resident duty hour reform has become among the most hotly debated subjects in medical education.^3–5 Arguments in favor of duty hour restrictions cite evidence that reductions in resident fatigue lead to improvements in patient outcomes,^6–10 increased learning,⁸ and improvements in resident wellbeing and safety.^11–13 Arguments against duty hour restrictions rely on evidence that increased patient hand-offs and lower cumulative clinical experience of residents may lead to more frequent errors and worse patient outcomes,^14–16 as well as evidence that the cost of implementing duty hour rules may be high.¹⁷ Motivated by these competing hypotheses, an important series of studies analyzed the short-term impact of the 2003 ACGME duty hour reform on mortality of patients hospitalized at teaching-intensive hospitals.^18–22 These studies found either no change or a slight improvement in mortality in teaching hospitals in the two years after the 2003 duty hour reforms.

Prior studies of duty hour reform have entirely focused on the impact of work hour restrictions on the immediate outcomes of patients treated by resident physicians, rather than seeking to understand the different question of whether reductions in residency work hours impact the clinical outcomes of physicians entering unsupervised practice after completion of residency. Training competent physicians is the fundamental objective of graduate medical education and yet the effects of the 2003 ACGME reforms on this educational outcome are unknown. Reductions in residency work hours and the transition towards shift-work may lead physicians to have less overall clinical experience on which to base diagnostic and management decisions after entry into independent practice and may also adversely affect professional development.^23–26 Alternatively, reductions in resident fatigue and efforts by residency programs to improve resident education may lead to improved clinical skills after completion of residency. The preparedness of resident physicians for unsupervised practice after residency is an important dimension of duty hour reform that is heavily discussed but has not been quantitatively studied in terms of patient outcomes.

We studied the impact of the 2003 ACGME duty hour reforms on outcomes of physicians who completed residency after 2003. We analyzed a unique database of nearly all hospitalizations in Florida hospitals which were linked to detailed data on the medical training history of the attending physician of record for each hospitalization. Since the ACGME implemented further residency work schedule reforms in 2011, we analyzed Florida data from between 2000 and 2009.² Focusing on physicians who completed an internal medicine residency without further sub-specialization, we analyzed whether average hospital mortality and length of stay for internists varied according to the number of years an internist was exposed to the 2003 duty hour regulations during their own residency.

METHODS

Data sources

We identified admissions to Florida acute care hospitals - excluding state operated, federal, and Shriner’s hospitals - from 2000 to 2009 on the basis of de-identified discharge records collected by the Florida Agency for Health Care Administration (AHCA). Data included information on patient age, sex, race, and zip-code of residence; quarter and year of admission; International Classification of Diseases, Ninth Revision (ICD-9), codes for principal and secondary diagnoses and procedures as well as diagnosis related groups (DRGs); disposition (e.g., inpatient death); and length of stay. Data were exempt from human subjects review at Harvard Medical School.

A unique feature of the data was its inclusion of the Florida license number of the attending physician who discharged the patient from the hospital. We linked each discharge according to license number to a separate licensure database which included the calendar years and specific types of residency and fellowship completion of each physician. Linkage of each discharge to the year of residency completion of the physician of record allowed us to study the association of hospital outcomes with the number of years that a physician was exposed to the 2003 ACGME duty hour reforms during their residency.

Study sample

We identified all discharges in which the physician of record was an internist, defined by completion of an internal medicine residency without sub-specialization. We excluded physicians with sub-specialty training for two reasons. First, these physicians accumulate additional clinical experience compared to non-fellowship trained physicians prior to entering independent practice. Second, because our data extended to 2009, we were unable to include physicians who began residency after 2003 and completed 3 or more years of sub-specialty training. For this reason, we also did not study surgical specialties which generally require five or more years of residency, recognizing however that the impact of duty hour reductions on outcomes of surgeons after residency may be large given the association between cumulative surgical experience and patient outcomes. We also included only physicians with medical doctorates rather than osteopathic physicians, in order to study outcomes among as homogenous a sample of physicians as possible. Finally, although several physicians may be involved in the care of a single patient (e.g. overnight care may be provided by a hospitalist physician distinct from the attending of record), we attributed outcomes of each patient to the attending physician of record at discharge.

Main outcome measures

Our main outcome measures were in-hospital death and length of stay, adjusted for patient demographics, diagnosis-related groups, and co-morbidities. Although estimates of in-hospital death may be biased by differences in length of stay across physicians or hospitals,²⁷ our data did not include information on post-discharge mortality. Nevertheless, risk-adjusted in-hospital mortality rates and length of stay remain important indicators of the quality of care that inpatients receive.

Statistical analysis

We compared average hospital mortality and length of stay across three groups of physicians: those with zero exposure to duty hour reforms (i.e., those completing residency before 2003), those with partial exposure (i.e., those who completed residency in 2004 or 2005 and had 1 or 2 years of exposure), and those with three full years of exposure (Exhibit 1). Importantly, because of declining trends in hospital mortality and length of stay between 2000 and 2009, physicians with full exposure to duty hour reforms would be expected to have lower mortality and length of stay compared to earlier cohorts of physicians with zero or partial exposure, irrespective of any effect of duty hour exposure on subsequent clinical outcomes of physicians.

EXHIBIT 1 (figure).

Difference-in-difference approach to estimating the effect of duty hour reforms on outcomes of physicians after completion of residency

Notes: The effect of physician exposure to duty hour reforms on subsequent clinical outcomes after completion of residency was identified by comparison of three groups: physicians with zero years of exposure to duty hour reforms (those completing residency before 2003), partial exposure (i.e, 1 or 2 years), and 3 full years of exposure. Exposure to duty hour reforms was confounded with the year in which a physician began practice (i.e., a comparison of outcomes of physicians with 3 versus zero years of exposure to duty hour reforms would be confounded by any secular time trends in patient outcomes). Therefore, outcomes were compared in post-residency years by using a difference-in-difference analysis. The difference-in-difference model compared outcomes of fully and partially exposed residency cohorts to similarly experienced cohorts in earlier years, and compared this difference to the trend in outcomes between 2000 and 2009 among physicians with zero exposure to duty hour reforms (control group). For example, the effect of full exposure to duty hour reforms was identified by (1) first comparing hospitalization outcomes in 2006–2009 of physicians who began residency after 2003 to similarly experienced cohorts in earlier years and (2) comparing this difference to the difference in outcomes before and after duty hour reforms among physicians with zero years of exposure to duty hour reforms during their own residency (control group). Trends in this control group of ‘senior physicians’ captured changes in hospital care that affected outcomes of all physicians, irrespective of the physician’s exposure to duty hour reforms during their own residency.

We addressed this potential bias through a series of difference-in-difference study designs. First, for each year between 2000 and 2009, we estimated average inpatient mortality of physicians who completed residency in the year prior (i.e., junior physicians) and of physicians who completed residency 10 or more years prior to the hospitalization year (i.e., senior physicians). We then compared the difference in outcomes of junior physicians before and after duty hour reform to the difference in outcomes of senior physicians before and after reform (control group). Assuming that trends in hospital care impact physicians of all levels of experience equally, this approach uses senior physicians with 10 or more years of post-residency experience as a control group to account for these trends. Differences in hospital mortality among patients treated by newly trained physicians before and after duty hour reforms are therefore measured relative to differences before and after duty hour reforms for patients treated by more experienced physicians.

Second, we expanded this approach to estimate how outcomes varied with four levels of physician exposure to duty hour reforms during the physician’s own residency: 0, 1, 2, and 3 years of exposure. We estimated a discharge-level multivariate linear probability model in which the dependent variable was inpatient mortality and a multivariate linear regression model in which the dependent variable was length of stay (days). The main independent variable was years of physician exposure to duty hour reforms. Independent variables also included a set of calendar year indicators, indicator variables for year of physician residency completion, an indicator variable for whether a physician graduated from a top-25 ranked internal medicine residency program according to the 2013 U.S. News and World Report (occurred in 3.7% of all hospitalizations), hospital indicators (to account for time-invariant hospital characteristics such as teaching hospital status), and indicator variables for years elapsed between physician residency completion and hospitalization year (i.e., physician experience). It is important to account for time-invariant hospital factors such as teaching hospital status since teaching hospitals may experience different trends in hospital quality compared to non-teaching hospitals, which may bias our results if junior physicians were increasingly more likely to work in teaching hospitals than non-teaching hospitals after duty hour reform. We also accounted for the reputation of a physician’s residency program since prior research in obstetrics and gynecology has demonstrated improved outcomes associated with higher reputation of the residency program from which a physician graduated.²⁸ Differential trends in the rate of Florida physicians graduating from high reputation residency programs may bias our results if not accounted for. We adjusted for physician experience since physicians exposed to duty hour reforms by definition had fewer years of post-residency experience, which may influence outcomes. We also adjusted for patient age, sex, race, and indicator variables for diagnosis-related group and Charlson-Deyo co-morbidities. Finally, we adjusted for the number of full and part time hospitalists – internists specializing in hospital care – employed by a hospital in a given year, since newer residency graduates may be more likely to work as hospitalists after completion of residency. A growing propensity among recent residency graduates to work as hospitalists may confound our results if the hospitalist model of care itself leads to lower mortality and LOS, as some evidence suggests.²⁹

Our analytic approach is analogous to a basic difference-in-difference model which typically compares outcomes before and after an intervention for a single control and single treatment group. Instead, the single control group included patients whose physician completed residency in 2003 or earlier, while the multiple treatment groups included patients whose physician completed residency in 2004 (1 year of physician exposure to the 2003 duty hour reforms), 2005 (2 years of exposure), or 2006 or later (3 years of exposure). To describe the estimated effect of duty hour reforms, we reported adjusted inpatient mortality and length of stay according to years of physician exposure to duty hour reforms during the physician’s own residency.

In order to estimate whether physician exposure to duty hour reforms had a differential effect on outcomes of high-risk patients, we estimated this model separately for high-risk patients, defined as those with a total number of Charlson-Deyo co-morbidities exceeding the top quartile in entire sample.³⁰ Standard errors were clustered at the hospital.

Limitations

Our study had several limitations. First, our outcomes were limited to inpatient mortality and length of stay rather than more refined measures of the quality of inpatient care. Adherence to established process measures of quality, long-term mortality, and readmission rates are better proxies for quality of hospital care and may differ between physicians with varying exposure to the 2003 duty hour reforms. Similarly, the post-training effect of duty hour restrictions may be largest for uncommon or complicated disorders for which greater clinical experience prior to beginning independent practice is necessary. However, overall mortality is the first outcome that should be considered in evaluating the impact of duty hour reform on the post-residency outcomes of physicians and has formed the basis of prior work studying the contemporaneous impacts of duty hour reforms on the outcomes of patients treated by resident physicians.^18–22

Second, length of stay may itself not accurately reflect the quality of physician decision-making since it is influenced by many other factors including the availability of post-acute hospital beds, the quickness with which home health services can be set up, and patient insurance and socio-economic issues. Moreover, discharging patients from the hospital earlier is not always the correct clinical decision.

Third, early adherence to duty hours by residency programs was variable across programs, particularly with respect to the number of continuous hours a resident could work and adequate time off for recovery sleep.^31,32 Delays in adherence to duty hour regulations may bias us away from finding adverse effects associated with reduced training.

Fourth, our difference-in-difference approach assumed that junior and senior physicians were equally impacted by unobserved secular changes in the quality of hospital care that occurred over the study period. However, prior research suggests that more experienced physicians may sometimes deliver lower quality care,³³ which means that using trends in outcomes of senior physicians over time may not accurately reflect the influence of unobserved trends in hospital quality on the outcomes of junior physicians before and after duty hour reform.

Fifth, we did not consider the impact of duty hour reductions on aggregate measures of physician health care utilization such as inpatient costs and use of diagnostic tests, imaging, procedures, and referrals.

Sixth, our data was limited to hospitalizations, precluding an analysis of the effect of duty hour reforms on outpatient care quality. Because our data did not extend past 2009, we were also unable to study surgical specialties which generally require five or more years of residency training. The effect of reductions in duty hours on post-residency outcomes of surgeons may be particularly pronounced given the importance of cumulative surgical experience for patient outcomes from surgery.

Seventh, we could not explicitly account for growth in advanced care personnel such as nurse practioners, whose hiring by hospitals may offset reductions in the quality of training of physicians who completed residency in recent years. Our study addressed this issue to the extent that over time, junior and senior physicians were equally likely to use these personnel. However, if junior physicians were more likely to use these personnel after duty our reform, our results would be biased away from finding an adverse effect of duty hour reforms on the quality of training.

Eighth, our analysis did not explore the impact of the 2011 duty hour reforms, which further reduced shift lengths to 16 hours for interns.

Ninth, our sample was not nationally representative and we could not verify with actual medical records how frequently the attribution of a particular attending physician to a hospital discharge was accurate.

Finally, we could not account for differential trends in rates of palliative care and hospice use among older versus younger physicians, which may lead to spuriously lower inpatient mortality among recent residency graduates if those physicians more often refer patients to outpatient hospice.

RESULTS

Among 4,608,508 hospitalizations in which the physician of record was an internist, 382,932 (8.3%) involved physicians who completed residency after the 2003 duty hour reforms; 132,433 (2.9%) involved physicians who completed residency after 2006 and were therefore fully exposed to duty hour rules during their three-year residency (Exhibit 2). The percent of total hospitalizations attributable to physicians in later residency cohorts was lower since these physicians had fewer years of independent practice; for example, for physicians who completed residency in 2008, hospitalization data was only available for 2009, whereas physicians who completed residency in 1980 could have hospitalizations at any point in our study period.

EXHIBIT 2 (table).

Hospital mortality, length of stay, and characteristics of patients treated by internists during 2000–2009, according to physician’s year of residency completion

Year of residency completion of physician of record	No. of inpatients	No. of deaths	% Mortality	Average length of stay (days)	Average patient age (years)	Patient sex (% male)	Average number of Charlson- Deyo comorbidities
Prior to 2000	3,706,788	130,465	3.5	5.3	66.4	45.6	1.1
2001	190,009	5,149	2.7	5.1	63.6	47.0	1.1
2002	187,944	5,342	2.8	5.2	62.6	48.9	1.2
2003	140,835	3,693	2.6	5.2	62.5	47.4	1.1
2004	134,021	3,609	2.7	5.1	62.4	48.0	1.1
2005	116,478	2,993	2.6	5.0	62.9	47.9	1.2
2006	74,616	1,726	2.3	5.1	61.1	47.9	1.2
2007	38,413	853	2.2	4.8	60.8	47.5	1.2
2008	19,404	451	2.3	5.0	60.9	48.9	1.2
Full sample	4,608,508	154,281	3.4	5.3	65.7	46.0	1.1

Notes The number of inpatients treated by physicians completing residency in a given year declined as the year of residency completion became more recent because physicians completing residency in more recent years had fewer years of independent practice. For example, physicians completing residency in 2008 could only have hospitalizations attributed in 2009, whereas physicians completing residency in 1980 could have hospitalizations attributed at any point in our study period (2000–2009).

Source Authors’ analysis of Florida hospital and physician license data.

Unadjusted hospital mortality was lower among patients whose physician(s) completed residency after the 2003 reforms, while length of stay was similar (Exhibit 2). For example, there were 144,649 deaths among 4,225,576 hospitalizations in which the physician of record completed residency before 2003 (3.4%), compared to 9,632 deaths among 382,932 hospitalizations in which the physician of record completed residency after 2003 (2.5%), p<0.001. Across physicians in different residency cohorts, patients were similar in sex and the average number of Charlson-Deyo comorbidities per patient. Average patient age slightly declined with more recent residency cohorts.

Basic difference-in-difference results

Lower hospital mortality among internists completing residency after 2003 may be explained by a number of factors, other than a direct positive effect of duty hour reforms on the clinical outcomes of physicians after residency. These factors may include improving time trends in hospital care affecting all physicians and/or lower mortality among patients treated by physicians who are within a few years of residency completion (e.g. if average patient complexity rises with years of physician experience). Exhibit 3 uses a differences-in-difference approach to illustrate the importance of accounting for hospital mortality trends and differences in mortality across levels of physician experience when estimating the effect of the 2003 duty hour reforms on the clinical outcomes of physicians newly entering independent practice. In 2000, unadjusted hospital mortality among patients whose physician of record completed residency one year earlier (i.e., 1999) was 3.3%, while in 2009 the mortality rate was 2.5% (among patients whose physician of record completed residency in 2008), p<0.001. While mortality declined from 2000 to 2009 among patients treated by physicians who were within 1 year of residency, a similar decline occurred for patients whose physician completed residency 10 or more years prior to hospitalization (p=0.08 for difference in trends). Therefore, using outcomes of patients treated by physicians with 10 or more years of post-residency experience to adjust for overall trends in hospital care, there was no change in hospital mortality after the 2003 duty hour reforms among patients treated by internists in their first year of independent practice.

EXHIBIT 3 (figure).

Trends in hospital mortality of Florida inpatients whose physician of record was in their first year after residency completion

Source: Authors’ analysis of Florida hospital and physician license data.

Notes: Figure plots unadjusted hospital mortality in each year for two groups of inpatients: those whose physician of record was an internist who completed residency 1 year prior to the hospitalization year (newly independent internist) versus 10 or more years prior (senior internist).

Effect of varying exposure of physicians to duty hour reforms

Exhibit 4 (Panel A) presents the association between hospital mortality and length of stay with the number of years of physician exposure to the 2003 duty hour reforms during the physician’s own residency. Adjusted hospital mortality and length of stay did not clinically meaningfully vary with the number of years a physician was exposed to the 2003 duty hour reforms during their residency. For example, adjusted hospital mortality among patients whose physician completed an internal medicine residency in 2003 or earlier (zero exposure) was 3.29% compared to 3.17% among patients whose physician of record completed residency in 2006 or later (three years exposure), p=0.05. Similarly, adjusted length of stay among patients whose physician completed residency in 2003 or earlier was 5.3 days, identical to patients whose physician of record completed residency in 2006 or later, p=0.82.

EXHIBIT 4 (table).

Association of adjusted hospital mortality and length of stay with physician exposure to duty hour reforms during physician’s own residency

Exposure of patient's physician to 2003 duty hour reforms during physician's own residency	A. All patients		B. High-risk patients
Years	Adjusted hospital mortality % (95% CI)	Adjusted length of stay Days (95% CI)	Adjusted hospital mortality % (95% CI)	Adjusted length of stay Days (95% CI)
0	3.29 (3.28 – 3.30)	5.26 (5.25 – 5.27)	7.71 (7.69 – 7.73)	6.87 (6.86 – 6.88)
1	3.25 (3.21 – 3.29)	5.27 (5.24 – 5.30)	7.58 (7.47 – 7.69)	6.89 (6.84 – 6.94)
2	3.21 (3.13 – 3.29)	5.29 (5.21 – 5.37)	7.44 (7.20 – 7.68)	6.92 (6.81 – 7.03)
3	3.17 (3.04 – 3.30)	5.30 (5.18 – 5.42)	7.31 (6.93 – 7.68)	6.94 (6.77 – 7.11)

Notes Physicians completing residency in 2003 or earlier had zero years of exposure to duty hour reforms; 2004, 1 year of exposure; 2005, 2 years of exposure; 2006 or later, 3 years of exposure. Adjusted hospital mortality and length of stay were estimated from logistic and linear models, respectively, which adjusted for patient age, sex, race, hospitalization year, indicators for diagnosis-related group and Charlson-Deyo co-morbidities, years of residency exposure of physician of record to the 2003 duty hour reforms, year of physician residency completion, years of physician experience, and hospital indicators. Standard errors were clustered at the physician-level. High-risk medical inpatients were defined as those with a total number of Charlson-Deyo co-morbidities in the top quartile of all patients. Regressions were separately estimated for all patients and high-risk patients.

Source Authors’ analysis of Florida hospital and physician license data.

Outcomes among high-risk patients

Among high-risk patients, those treated by physicians who had greater exposure to the 2003 duty hour reforms during their residency had slightly lower adjusted hospital mortality (Exhibit 4, Panel B). For example, adjusted hospital mortality among high-risk patients whose physician completed residency in 2003 or earlier (zero exposure) was 7.71%, statistically greater (p=0.04) than 7.31% among high-risk patients whose physician of record completed residency in 2006 or later (three years exposure). Adjusted length of stay among high-risk patients was not statistically significantly associated with the number of years the physician of record was exposed to the 2003 duty hour reforms during their own residency.

DISCUSSION

The impact of resident duty hour reform on the clinical competency of physicians completing residency is among the most contested issues in medical education. We analyzed the associations between inpatient mortality and length of stay with the number of years during a physician’s own residency in which he or she was exposed to the 2003 ACGME duty hour regulations. We found that internists who were either partially or fully exposed to the reduced residency work hours of the 2003 reforms did not have statistically different rates of inpatient mortality or length of stay after completion of residency when compared to physicians who were not exposed to the duty hour regulations during their own residency. Moreover, although most concerns about the effect of duty hour reforms on physician training would predict poorer outcomes among high-risk patients – due to less cumulative clinical experience on which to base treatment decisions – we found that high-risk patients treated by physicians who had greater exposure to duty hour reforms during their residency had slightly lower hospital mortality. Our finding that hospital mortality trended downwards from 2000 to 2009 for all physicians also highlights the importance of accounting for this trend in comparing the outcomes of physicians recently completing residency before and after the 2003 reforms.

Although a substantial body of research has addressed whether the 2003 duty hour reductions were associated with changes in mortality attributable to care by resident physicians in teaching hospitals^18–22, no studies have explored whether reductions in residency work hours have impacted the clinical outcomes of physicians after residency completion. Because reductions in residency work hours and the transition towards shift-work may lower overall clinical experience prior to beginning independent practice, understanding the long-term consequences of duty hour reform is important. Concerns about the impact of reduced work hours on clinical training have also surfaced in Europe, where even more stringent work hour requirements for physicians in training are in place.³⁴ Although our study did not analyze outcomes in surgery, these concerns may be particularly valid there because potential reductions in operative volume and clinical care time during residency may lead to inadequate surgical preparation for independent practice.^35,36

Our findings are important because they suggest that the national reduction in residency work hours beginning in 2003 did not adversely impact two specific measures of the quality of hospital care provided by physicians who completed residency after 2003. Several factors may explain our findings. First, mortality and length of stay may be imperfect proxies for the clinical competency of physicians. These measures may ignore more nuanced measures of clinical competency such as the ability to diagnose and treat uncommon or complicated medical conditions and to care for a larger number of hospitalized inpatients in any given period of time. Second, although reductions in residency work hours and a transition to shift-work may have led to less cumulative clinical training on which to base independent practice, reductions in resident fatigue and efforts by residencies to improve resident education may have offset this effect. Several studies suggest, however, that resident education did not improve with duty hour reform.^12,37

Third, hospital care has changed in a myriad of ways since the implementation of the 2003 duty hour reforms. In particular, many hospitals now hire hospitalists to deliver in inpatient care. If hospitalist care is associated with lower mortality and length of stay, as some evidence suggests, a growing propensity among recent residency graduates to work as hospitalists may confound our results.^29,38,39 However, our analysis adjusted for the number of hospitalists employed by hospitals in our sample. Similarly, the 2003 duty hour reforms themselves placed new demands on inpatient care, leading many teaching hospitals to develop new inpatient care models that included growth in non-teaching hospitalist services and advanced practice personnel such as nurse practioners.¹⁷ Since teaching hospitals traditionally employ larger numbers of physicians recently completing residency, improvements over time in these hospitals may partly offset any deterioration in the quality of training of physicians who completed residency in recent years. Moreover, these same improvements over time in teaching hospitals may improve the quality of training of new attending physicians who completed residency in these hospitals. To address these potential concerns, our analysis adjusted for time-invariant hospital characteristics such as teaching hospital status. More broadly, however, an increasing focus on quality of care, increasing use of multi-disciplinary teams, computerized order entry systems, automated medication checks, and inpatient pharmacists may mitigate the long-term effect of less intensive residency training after the 2003 reforms.⁴⁰

In summary, our study suggests that the 2003 ACGME duty hour reforms did not adversely impact hospital mortality and length of stay of patients cared for by new attending physicians who were partly or fully exposed to reduced duty hours during their own residency. Further assessment of the impact of the 2003 and 2011 duty hour reforms on other aspects of physician quality, for particular patient sub-populations, and in national data is important.