Editor’s Note: A commentary by B.M. Wong and K. Imrie appears on pages 1209–1211.
First in 2003 then again in 2011, the Accreditation Council for Graduate Medical Education reduced the number of working hours for resident physicians.1 Although studies have demonstrated little change in patient outcomes after these shifts in duty hours regulations, concerns remain regarding the unintended consequences of these restrictions related to resident education and clinical experiences.2
One such unintended consequence of shorter resident duty hours that has received little attention is faculty workload. Given that resident physicians are less available, inpatient attending physicians may face increasing workloads with greater responsibility of patient care, which could in turn decrease their time for teaching.3 Although research on attending physicians’ workload and teaching is limited, previous outpatient studies demonstrated that attending physicians’ self-perceptions of workload are associated with increased fatigue and decreased satisfaction.4 An emergency department study demonstrated that attending physicians’ ability to provide comprehensive care decreased with an increased workload.5
Although data on attending physicians are limited, several studies confirm a negative association between increased resident physicians’ workload, measured by patient census, and educational and patient care outcomes. One study reported an increase in patient mortality with more on-call admissions to a resident service.6 Another study demonstrated that resident physicians perceived that less learning had occurred as patient volume increased.7
Understanding the association between inpatient attending physicians’ workload and teaching is critical to preserving the learning experience in residency training. With this study, we aimed to test the association between inpatient attending physicians’ self-reported workload and their perceptions of time for teaching before and after the 2003 resident duty hours regulations and to explore this relationship at different points in the academic year.
From 2001 to 2008, we surveyed all teaching attending physicians from a single institution at the end of their inpatient medicine rotations.8 Questions included five-item Likert-type scales to assess respondents’ satisfaction and agreement with statements focused on domains related to workload, teaching, and burnout. Questions assessed their satisfaction with work on the general medical wards, experience as attending physicians, and amount of perceived stress. Other questions measured time for teaching in two ways. First, we defined satisfaction with time for teaching as a response of agree or strongly agree to the statement “I had enough time for teaching.” Additionally, on an open-ended question, respondents reported the number of hours per week they spent doing didactic teaching. We dichotomized the responses to this question as above or below the median number of hours of teaching per week.
To measure workload, we used a conceptual framework for perceived workload initially developed by the National Aeronautics and Space Administration (NASA) and later adapted for physicians.4,9 This framework consists of six domains believed to define the experience of workload for an individual: mental demand, physical demand, temporal demand, effort, performance, and frustration level. We used three questions verbatim from the NASA framework (mental demand, frustration level, and temporal demand). For the domains of performance and effort, the questions on our survey had very similar meanings to those from the NASA framework but were not verbatim. For example, to measure performance, we asked how satisfied respondents were with their attending physician experience on the wards, whereas NASA asked, “How satisfied were you with your performance in accomplishing these goals?”9 To construct the workload index following the NASA method, we reverse coded survey items related to performance and effort to measure whether difficulty in these areas was positively associated with increased workload. To measure perceived workload, we created a composite workload score (range 6–30) from the sum of the four items (mental demand, physical demand, temporal demand, and frustration level) and the two reverse-coded items (performance and effort).
We also asked respondents to answer questions about demographics (i.e., gender) and practice characteristics (i.e., specialty), and we recorded the month and year in which they completed the survey. We used physician billing records to calculate respondents’ volume of patients admitted and the number of days each spent on service for the rotation. Because attending physicians’ time on service is typically either two weeks or four weeks, we dichotomized the number of days on service as 16 days or less and greater than 26 days.
We calculated descriptive statistics to summarize the distribution of respondents’ composite workload scores as well as demographics related to gender and specialty. We grouped the data in two categories based on year: pre implementation of the 2003 duty hours regulations and post implementation of the 2003 duty hours regulations. We used multivariate logistic regression, controlling for clustering by attending physicians and adjusting for time trends during the academic year, to ascertain the relationship between centered workload composite score, implementation of the 2003 duty hours regulations, and time for teaching (self-reported time for teaching, time spent conducting more than the median number of hours of didactic teaching per week). We divided the academic year into early (July–September), middle (October–February), and late (March–June). We used the early academic year period (July–September) as the reference group for our analysis.8 To account for the possibility that effect of workload varied by academic year period, we constructed interaction terms—we used the interaction term workload composite*early academic year as the reference group.
To analyze the validity of the composite workload score, we calculated the crude, unadjusted odds ratio (OR) for each of the components with regard to self-reported time for teaching and time spent conducting more than three hours of didactic teaching per week. Additionally, we examined the correlation between perceived workload and actual patient volume. Furthermore, we used a t test to assess the relationship between time on service and composite workload score.
We conducted sensitivity analyses to assess the robustness of our results. We also conducted a separate analysis on the composite workload score, the main predictor variable. We originally included the composite workload score as a continuous variable in the final model. We recoded this score as a nonlinear variable, in both tertiles and quartiles, and repeated the analysis to test whether the main and seasonal effects remained similar. We considered the presence of secular trends on perceived workload and time for teaching. In lieu of the duty hours variable, we placed individual year variables (2001–2007), using year 2008 as the reference category, into the final model. Because only one attending physician completed two weeks of service (instead of four weeks) pre implementation of the 2003 duty hours regulations, we repeated the analysis by dropping the duty hours variable and using length of service instead. We constructed similar models to test the association between patient volume (as opposed to composite workload score) and time for teaching. We also examined the correlation between time for teaching and other metrics, such as overall burnout. Lastly, we examined the association between composite workload score and burnout using the model specifications of implementation of the 2003 duty hours regulations and academic year period, clustered by attending physician. We performed all statistical tests using Intercooled Stata 11.0 (College Station, Texas).
We distributed surveys to all 738 attending physicians on the inpatient general medicine ward service from 2001 to 2008. We included the responses of 482 (65%) teaching attending physicians in our analysis. Although the response rate varied by month and year, the differences were not statistically significant (χ2 = 61.05, P = .909). Of 458 respondents, 211 (46%) were female and 388 (85%) identified internal medicine as their primary specialty (see Table 1). Among subspecialties, general internal medicine was most common (202; 44%), followed by geriatrics (73; 16%) and hospitalists (66; 14%) (see Table 1).
The composite workload scores ranged from 6 to 30, with a median score of 15 and an interquartile range of 14 to 17 (see Figure 1). We found a weak positive correlation (r = 0.25) with actual patient volume. Of attending physicians, 198 (43%) reported that they either agreed or strongly agreed that they had enough time for teaching. Overall, attending physicians conducted a median of three hours per week of didactic teaching, with an interquartile range of two to four hours. Furthermore, 282 (65%) spent more than the median time teaching on a weekly basis.
In our multivariate analyses, the odds of an attending physician reporting enough time for teaching were 21% lower for each point increase in composite workload score (OR = 0.79 [95% confidence interval (CI) 0.69–0.91]; P = .001). This score was substantially lower after the implementation of the 2003 duty hours regulations (OR = 0.38 [95% CI 0.19–0.76]; P = .006). There was no interaction between the implementation of the 2003 duty hours regulations and composite workload score. However, the odds of reporting enough time for teaching were lower in the middle (OR = 0.61 [95% CI 0.37–0.99]; P = .04) and late (OR = 0.69 [95% CI 0.42–1.15]; P = .161) periods of the academic year when compared with the early period. Interactions between composite workload score and period of the academic year were significant. Increased attending physicians’ workload during the middle (OR = 0.78 [95% CI 0.64–0.96]; P = .02) and late (OR = 0.80 [95% CI 0.63–1.02]; P = .08) periods of the academic year was associated with even less time for teaching than during the early period of the academic year.
For a hypothetical attending physician with a low composite workload score (in the 25th percentile) during the early academic year prior to the implementation of the 2003 duty hours regulations, we calculated the predicted probability of that attending physician reporting enough time for teaching as 80%. When compared with a hypothetical attending physician with a high composite workload score (in the 75th percentile) during the same academic year period and year, the predicted probability of him or her reporting enough time for teaching decreased to 57%. During the middle and late academic year periods (prior to the implementation of the 2003 duty hours regulations), the predicted probability of an attending physician with a high composite workload score reporting enough time for teaching was less than 50%, while the predicted probability of an attending physician with a low composite workload score was over 60%.
After the implementation of the 2003 duty hours regulations, the predicted probabilities of both hypothetical attending physicians (high and low composite workload score) to report enough time for teaching decreased. The predicted probability of an attending physician with a low composite workload score reporting enough time for teaching was about 67% versus 42% for an attending physician with a high composite workload score. Compared with the early academic year period, the middle and late academic year periods (post implementation of the 2003 duty hours regulations) are characterized by lower perceived time for teaching and are more sensitive to higher perceived workloads. Among the high composite workload score group, the predicted probability of an attending physician reporting enough time for teaching dropped to roughly 38%, although the predicted probability of an attending physician with a low composite workload score reporting enough time for teaching was just over 50% (see Figure 2).
When we examined whether an attending physician spent more than three hours per week conducting didactic teaching, the results were similar. For each point increase in composite workload score, the odds of teaching more than three hours per week decreased by 18% (OR = 0.82 [95% CI 0.75–0.90]; P < .001). Similarly, post implementation of the 2003 duty hours regulations, teaching more than three hours per week decreased by 64% (OR = 0.36 [95% CI 0.17–0.76]; P = .007) (see Table 2). Furthermore, teaching declined by 52% in the middle academic year period (OR = 0.48 [95% CI 0.29–0.83]; P = .008). There were no significant interactions between these variables.
Each component of the workload score contributed to the overall composite score. When we analyzed workload domains with respect to self-reported time for teaching, the crude, unadjusted OR was 0.33 (95% CI 0.25–0.45; P < .0001). Similarly, the crude OR for stress (frustration domain) was 0.63 (95% CI 0.50–0.80; P < .0001), for excessive patient volume (physical demand domain) was 0.51 (95% CI 0.40–0.65; P < .0001), and for ability to set the pace of work (temporal demand domain) was 0.43 (95% CI 0.34–0.54; P < .0001). All of these values represent the negative association of workload scores to self-reported time for teaching. Overall satisfaction (performance domain) had a crude OR of 2.32 (95% CI 1.70–3.02; P < .0001) and being up to date with inpatient medicine (effort domain) of 1.43 (95% CI 1.14–1.79; P = .002) (see Table 3). Each component of the composite workload score contributed equally to the overall score when we assessed the effects of workload on self-reported time for teaching (see Table 3).
Actual patient volume was not a significant predictor of time for teaching (OR = 0.99 [95% CI 0.98–1.01]; P = .86) when we controlled for duty hours, academic year period, and time on service. Self-reported faculty burnout and time for teaching were weakly negatively correlated (r = −0.13), whereas self-reported overall satisfaction and time for teaching were weakly positively correlated (r = 0.31). Controlling for all other covariates, the odds of reporting symptoms of burnout increased 1.17 times for each point increase in composite workload score (95% CI 1.00–1.37; P = .04).
When we included individual year variables into our final model in lieu of the duty hours variable, the effect of the composite workload score on attending physicians’ reported time for teaching remained similar (OR = 0.78 [95% CI 0.68–0.89]; P = .001). Additionally, controlling for attending physicians’ length of service (two weeks versus four weeks) did not change the relationship between the composite workload score and attending physicians’ reported time for teaching (OR = 0.78 [95% CI 0.69–0.89]; P < .001), indicating that the amount of time attending physicians spent on service did not reflect their self-perceived workload. Although a t test of the relationship between the composite workload score and time on service was statistically significant, the findings were not clinically relevant [composite workload score with two weeks of service = 15.5; composite workload score with four weeks of service = 16.1; P = .02]. We did note a dose–response effect when we examined the effect of the composite workload score, as a nonlinear variable, on attending physicians’ reported time for teaching.
We found that attending physicians’ greater self-perceived workload was associated with decreased time for teaching. We also found a significant workload–academic year period interaction, such that attending physicians’ greater perceived workload was associated with even less time for teaching during the winter compared with the summer. Moreover, time for teaching was substantially reduced after the implementation of the 2003 resident duty hours regulations. With residents working even fewer hours after the implementation of additional duty hours regulations in 2011, efforts to preserve teaching on inpatient ward rotations are critical, given the potential increases in attending physicians’ workload.
Attending physicians’ self-perceived workload was an important predictor of time for teaching, whereas actual patient census was not. This finding suggests that factors other than patient census contribute to the overall perception of workload, such as patient acuity. For example, having to conduct a family meeting to determine goals of care for a complex patient could represent such a circumstance. Attending physicians also have other responsibilities that contribute to their overall workload but that fall outside the domain of direct medical care and are not related to patient volume. Moreover, resident absenteeism due to scheduling concerns may contribute to feelings of disjointedness among members of the medical team and lead to decreased satisfaction. Subsequently, attending physicians could translate this feeling into reporting less time for teaching or an increased workload. Regardless, future studies examining attending physicians’ workload must measure more than patient volume.10
The significance of the correlation between academic year period and attending physicians’ self-reported time for teaching raises interesting questions about the etiology of this association. The winter season likely affects workload more than the other seasons because hospitals tend to experience a greater acuity and higher patient census during this time, though this is not reflected in our billing data claims.11 Alternatively, major emphasis is put on teaching during the month of July to appropriately orient resident physicians to their new roles. Once this provisional period has passed, attending physicians must address any previously unfinished administrative duties, which transiently increases their perceptions of their workload. The association we found between decreased time for teaching and the winter should prompt attending physicians to pay more attention to teaching during the winter and spring.
In addition, after the implementation of the 2003 duty hours regulations, attending physicians began assuming roles traditionally held by resident physicians.3,8 This transition of additional duties could lead to a stronger association between the increase in self-perceived workload and the decrease in time for teaching. This shift becomes more salient after the implementation of the 2011 duty hours regulations, when the further reduction in resident physicians’ work hours may shift additional duties to attending physicians. These added responsibilities might portend a stronger association between attending physicians’ self-perceived workload and decreased time for teaching. Reducing attending physicians’ workload is an important step in improving the learning experience of resident physicians and medical students.
Given these concerns, the academic medicine community must consider novel solutions to address attending physicians’ workload and to maintain the necessary time for teaching. To date, efforts to reduce their workload have focused on the number of patients attending physicians see. A dedicated teaching–attending physician at Brigham and Women’s Hospital has led to resident physicians’ increased satisfaction, more time spent on learning and teaching activities, and improved discharge summaries.12,13 Additionally, the Aliki Initiative at Johns Hopkins Bayview reduced patient census to bring a patient-centered care approach to inpatient teaching and increased the satisfaction of both physicians and patients as well as care outcomes.14 Researchers found similar effects after the patient census for resident teams was reduced at Mayo Clinic.15 Finally, our findings show that factors other than patient census contribute to self-perceived workload, and they demonstrate a need to continue to develop novel solutions to maintain teaching throughout the year.
There are several limitations to our study. First, we did not use the original NASA task load index instrument; instead, we retrospectively mapped our data to its conceptual domains. Although this method of workload evaluation has been used in the published medical literature, it has not been validated.4 Second, because we collected our data at a single institution, generalizability is limited. Third, self-reported workload could be subject to recall bias. Fourth, although we surveyed attending physicians on aspects of teaching other than didactics, responses yielded extreme outliers, and we excluded them from our analyses. Fifth, the composite workload score does not account for faculty members’ academic productivity, which may translate to less reported time for teaching. Finally, although we collected our data across an eight-year period, we did not capture the most recent trends in residency training. Our methods and the corresponding results, however, may have implications for the present. For example, although we did not examine the direct effects of the 2011 duty hours regulations, we did conclude that we must continue to examine the effects of attending physicians’ workload, especially with further restrictions in duty hours.
Attending physicians’ greater self-perceived workload was associated with decreased time for teaching and with even less time for teaching during the winter. Moreover, time for teaching was substantially lower after the implementation of resident physician duty hours regulations in 2003. With residents working even fewer hours after the 2011 duty hours regulations, efforts to preserve teaching on inpatient ward rotations are critical, given the potential increases in attending physicians’ workload.
Acknowledgments: The authors would like to thank Kimberly Beiting for her administrative support and the internal medicine attending physicians who participated in the survey process.
1. Philibert I, Friedmann P, Williams WTACGME Work Group on Resident Duty Hours. Accreditation Council for Graduate Medical Education. . New requirements for resident duty hours. JAMA. 2002;288:1112–1114
2. Arora VM, Georgitis E, Siddique J, et al. Association of workload of on-call medical interns with on-call sleep duration, shift duration, and participation in educational activities. JAMA. 2008;300:1146–1153
3. Reed DA, Levine RB, Miller RG, et al. Effect of residency duty-hour limits: Views of key clinical faculty. Arch Intern Med. 2007;167:1487–1492
4. Bertram DA, Hershey CO, Opila DA, Quirin O. A measure of physician mental work load in internal medicine ambulatory care clinics. Med Care. 1990;28:458–467
5. Wrenn K, Lorenzen B, Jones I, Zhou C, Aronsky D. Factors affecting stress in emergency medicine residents while working in the ED. Am J Emerg Med. 2010;28:897–902
6. Ong M, Bostrom A, Vidyarthi A, McCulloch C, Auerbach A. House staff team workload and organization effects on patient outcomes in an academic general internal medicine inpatient service. Arch Intern Med. 2007;167:47–52
7. Haney EM, Nicolaidis C, Hunter A, Chan BK, Cooney TG, Bowen JL. Relationship between resident workload and self-perceived learning on inpatient medicine wards: A longitudinal study. BMC Med Educ. 2006;6:35
8. Arora V, Meltzer D. Effect of ACGME duty hours on attending physician teaching and satisfaction. Arch Intern Med. 2008;168:1226–1228
9. Leedal JM, Smith AF. Methodological approaches to anaesthetists’ workload in the operating theatre. Br J Anaesth. 2005;94:702–709
10. Tomasko JM, Pauli EM, Kunselman AR, Haluck RS. Sleep deprivation increases cognitive workload during simulated surgical tasks. Am J Surg. 2012;203:37–43
11. Garfield M, Ridley S, Kong A, Burns A, Blunt M, Gunning K. Seasonal variation in admission rates to intensive care units. Anaesthesia. 2001;56:1136–1140
12. McMahon GT, Katz JT, Thorndike ME, Levy BD, Loscalzo J. Evaluation of a redesign initiative in an internal-medicine residency. N Engl J Med. 2010;362:1304–1311
13. Coit MH, Katz JT, McMahon GT. The effect of workload reduction on the quality of residents’ discharge summaries. J Gen Intern Med. 2011;26:28–32
14. Ratanawongsa N, Federowicz MA, Christmas C, et al. Effects of a focused patient-centered care curriculum on the experiences of internal medicine residents and their patients. J Gen Intern Med. 2012;27:473–477
15. Thanarajasingam U, McDonald FS, Halvorsen AJ, et al. Service census caps and unit-based admissions: Resident workload, conference attendance, duty hour compliance, and patient safety. Mayo Clin Proc. 2012;87:320–327