|
|
Background Although the Accreditation Council for Graduate Medical Education (ACGME) limits the work hours of residents, concerns about fatigue persist. A new Institute of Medicine (IOM) report recommends, among other changes, improved adherence to the 2003 ACGME limits, naps during extended shifts, a 16-hour limit for shifts without naps, and reduced workloads.
Methods We used published data to estimate labor costs associated with transferring excess work from residents to substitute providers, and we examined the effects of our assumptions in sensitivity analyses. Next, using a probability model to represent labor costs as well as mortality and costs associated with preventable adverse events, we determined the net costs to major teaching hospitals and cost-effectiveness across a range of hypothetical changes in the rate of preventable adverse events.
Results Annual labor costs from implementing the IOM recommendations were estimated to be $1.6 billion (in 2006 U.S. dollars) across all ACGME-accredited programs ($1.1 billion to $2.5 billion in sensitivity analyses). From a 10% decrease to a 10% increase in preventable adverse events, net costs per admission ranged from $99 to $183 for major teaching hospitals and from $17 to $266 for society. With 2.5% to 11.3% decreases in preventable adverse events, costs to society per averted death ranged from $3.4 million to $0.
Conclusions Implementing the four IOM recommendations would be costly, and their effectiveness is unknown. If highly effective, they could prevent patient harm at reduced or no cost from the societal perspective. However, net costs to teaching hospitals would remain high.
This analysis extends a study commissioned to assist the deliberations of the IOM's Committee on Optimizing Graduate Medical Trainee (Resident) Hours and Work Schedules to Improve Patient Safety.21 Our objective was to estimate the labor costs associated with implementing four major IOM recommendations in ACGME-accredited residency programs: adherence to the 80-hour workweek, a 5-hour nap during extended shifts, a 16-hour limit for shifts without naps, and a reduced workload. Additional recommendations that appeared to be unlikely to generate substantial costs were not included. Shifting residents' work to substitute providers or an expanded population of residents would generate the labor costs.
Secondary objectives included estimating potential net costs to major teaching hospitals and cost-effectiveness to society — specifically, costs per averted death. Preventable adverse events generate costs during and after hospitalization22,23; therefore, if the changes decreased these events, labor costs would be offset by the reduction in costs related to preventable adverse events. Because handoffs of patient care would increase, increases in these costs would also be possible. We drew the framework for this study from an analysis of the 2003 reforms but considered the new recommendations, current working patterns of residents, and cost-effectiveness.24
Methods
Study Design
Our analysis involved two main steps. First, we estimated labor costs across 1206 hospitals with ACGME-accredited programs by considering the current working patterns of residents and the providers who are able to perform residents' work. Second, to estimate net costs and cost-effectiveness at major teaching hospitals (defined as members of the Council of Teaching Hospitals), we developed a probability model representing labor costs as well as mortality and costs associated with preventable adverse events.25 The model simulates hypothetical changes in preventable adverse events, ranging from a 10% increase to the percentage decrease that would exactly offset the labor costs. The reference year was 2006.
PubMed searches provided model variables. For working patterns and substitutes, we gave preference to studies that included national samples, multiple hospitals, multiple specialties, or specialties involving many residents.26 We searched the literature in PubMed for a combination of the terms "resident" and "duty hour" and "resident" and "work hour," with the limit "after July 1, 2003." For variables related to preventable adverse events, we selected studies that used medical-record review to identify various preventable events in U.S. hospitals; we searched for a combination of the terms "preventable adverse events" and "rate"; "cost" and "medical error"; "cost" and "preventable adverse events"; and "cost" and "medical injuries" and "adverse events" — all with the limit "after 1993." We also examined the first 20 articles that PubMed listed as being related to each study, references from the relevant articles, and publicly available data.24
Working Patterns
Variables in the model included current workweek, duration and frequency of extended shifts, and time on inpatient rotations (Table 1). Data sources comprised a national survey of first-postgraduate-year (PGY-1) residents in all specialties; three multihospital surveys of residents in general surgery, orthopedics, and pediatrics in multiple class years; and two surveys of residents in multiple specialties and class years at three major teaching hospitals. Respondents in the PGY-1 survey reported longer workweeks than did those in the other surveys (in which PGY-1 residents accounted for 35% or fewer of the respondents).12,13,15,16,28,29 One survey of residents in multiple specialties and multiple class years showed that workweeks exceeding the 80-hour limit were common in some specialties (among 22% of respondents); we therefore assumed that 22% of residents in the second postgraduate year (PGY-2) or beyond had workweeks similar to those of PGY-1 residents.13 This survey provided information on the duration of extended shifts for residents in PGY-2 or beyond.13 The general-surgery survey provided information on the frequency of extended shifts.15
|
Since the 2003 reforms, residency programs have often employed midlevel providers (physician assistants and nurse practitioners) to replace resident labor, but many residents still spend 9 to 24 hours per week on noneducational tasks that lower-level providers can perform.13,15,64,65,66,67,68,69,70,71,72,73 Programs have also shifted work among residents, but this can detract from educational opportunities.17 The IOM committee advocated greater involvement of attending physicians.
For each of the four recommendations, we assessed the transfer of current work beyond the proposed limit ("excess work") to substitute providers (Table 2). Attending physicians are logical substitutes for subspecialty residents; specialty residents are discussed below. We considered implementation of the recommendations serially to prevent double counting. For example, the workload would not need to be reduced during hours eliminated by adhering to the 80-hour workweek.
|
In the base-case scenario, midlevel providers and attending physicians assumed most of the excess work. A total of 3.9% of specialty residents working more than 100 hours transferred 12 hours to lower-level providers; the types of substitutes used were derived from an internal-medicine substitution model.74 We assumed that adherence to the 80-hour workweek would make adherence to the other limits of the 2003 reforms possible at no additional cost.
Protected Naps during Extended Shifts
Residents currently nap for 2 to 3 hours, but interruptions are frequent. Protected (i.e., uninterrupted) 5-hour naps would require 300 minutes of coverage (100 to 300 minutes in sensitivity analyses).14,15,75,76 Attending physicians may be the preferred substitutes because many patients are admitted during extended shifts and newly admitted patients often require a high level of care.
16-Hour Limit for Shifts without Naps
Residency programs can shorten extended shifts to 16 or fewer hours instead of providing for uninterrupted naps. Providing for naps when current shifts exceed 21 hours and shortening shifts when they last up to 21 hours limit the excess work to 5 hours per shift. The base-case scenario involved midlevel providers and attending physicians as substitutes.
Reduced Workload
The IOM report did not specify which residents would have reduced workloads or by how much the workloads would be reduced.20 In the base-case scenario, we assumed a 10% decrease for PGY-1 residents (5 to 20% in sensitivity analyses). We used midlevel providers as substitutes and assumed that costs would be the same as those for equivalent reductions in work hours because workload is not easily quantifiable.
Assumptions Relevant to All Four Recommendations
We assumed that individual substitutes would perform excess work transferred from multiple residents. Sensitivity analyses were performed to account for the following possibilities: the types of substitutes used may vary, rested substitutes may be more or less efficient than fatigued residents, providers' wages may be higher in urban areas than in rural areas, demand for midlevel providers and physicians could increase their wages, and some excess work may not be transferred to substitutes for several reasons.
Additional Residents
The Council on Graduate Medical Education recommends expanding the number of residency positions, which could reduce the work hours and workloads of residents.77 We determined the minimum overall increase in positions required, calculating excess work as described above. Costs per resident were based on stipends and benefits, average expenditures from public sources for graduate medical education (GME), excluding indirect payments, and average GME expenditures from all sources. Stipends remained stable after the 2003 reforms.39
Preventable Adverse Events
We based mortality and costs associated with preventable adverse events on one study that met our inclusion criteria. Thomas et al.22 identified preventable adverse events by examining medical records from 14,000 hospitalizations in Utah and Colorado, and they used consensus judgments to estimate disability and lifetime health care utilization related to these events. The authors projected the rate and number of preventable adverse events and attributable health care costs, lost wages, and the cost of hiring other persons to perform household duties in the two states in 1996 (discounting future costs at 2.75%).22 Rates of preventable adverse events did not vary according to hospital teaching status.43,78
Statistical Analysis
To derive labor costs, we calculated the total excess hours for each recommendation (number of excess hours per resident per week multiplied by numbers of affected weeks and residents) and then multiplied the result by substitutes' hourly wages plus benefits. When additional residents were considered, we divided total excess hours by the hours residents would work annually and multiplied the result by the annual per-resident costs. Sensitivity analyses (SA) (performed with the use of Stata software) examined how alterations in the variables and assumptions (sensitivity-analysis numbers SA1 through SA19) affected labor costs.
Estimating net costs and cost-effectiveness involved the following: calculating labor costs at major teaching hospitals; determining mean costs per preventable adverse event in 2006 (1996 costs multiplied by geographic and temporal adjustment factors); applying rates of preventable adverse events, mortality, and costs to admissions at major teaching hospitals; calculating changes in the incidence of preventable adverse events, mortality, and costs across changes in the rate of preventable adverse events; determining net costs to major teaching hospitals; and determining incremental cost-effectiveness ratios (performed with the use of Excel software). The teaching-hospital model included labor and inpatient costs related to preventable adverse events. The societal model included labor and all costs related to preventable adverse events.
Results
Annual labor costs under the base-case scenario were $1.6 billion (in 2006 U.S. dollars) (Table 3), a quarter of which pertained to improving adherence to the 80-hour workweek and half of which pertained to reducing fatigue related to extended shifts. Expanding residency positions appears to be less costly (approximately $500 million) when calculated on the basis of stipends and benefits but slightly more costly ($1.7 billion) based on per-resident GME expenditures from all sources. Costs per major teaching hospital would be $3.2 million for substitute providers and $990,000 to $3.5 million for additional residents (data not shown).
|
|
|
Discussion
Reducing resident physicians' work hours and workloads in accordance with four IOM recommendations would involve $1.6 billion of substitute labor; costs would range from $1.1 billion to $2.5 billion, depending on the current working patterns of residents, the wages and efficiency of substitute providers, and other considerations. With the use of additional residents instead of substitute providers, at least 8% more residency positions would be required, costing up to $1.7 billion annually.
Although the Council on Graduate Medical Education recommends expanding residency positions, the use of this approach to reduce residents' work raises three considerations.77 First, it will expand the long-term supply of physicians. Basing residency positions for each specialty on its residents' current work hours could alleviate shortages in some specialties but create surpluses in others. Pediatrics residents work long hours, for example, but the supply of pediatricians appears to be sufficient; therefore, expanding residency positions could create a surplus.12,79 Second, the additional graduates from U.S. or international medical schools might have weaker qualifications than current trainees. Third, admissions to teaching hospitals have been increasing; therefore, more residents may be needed in the future.58,59
With the use of either substitute providers or additional residents to reduce current residents' work, annual implementation costs would be high as compared with both GME expenditures and other interventions for patient safety. In 2006, direct Medicare GME payments were $2.3 billion and indirect Medicare GME payments were $5.9 billion.80 Public and private support totaled $18.7 billion in 2003.40 The three new IOM recommendations that we analyzed (i.e., protected naps, 16-hour shifts, and reduced workload) would cost more than the 2003 reforms may have cost ($1.2 billion vs. $0.7 billion to $1.1 billion).24 The cost per hospital ($3.2 million annually) is higher than the cost associated with computerized physician order entry ($3.3 million to $11.8 million over a period of 10 years) and medication bar-coding systems ($2.2 million over a period of 5 years).50,81,82 However, the cost per admission ($141) is lower than the cost associated with implementation of policies requiring one nurse for every four patients ($261).83
It remains unknown whether implementing the IOM's recommendations would reduce preventable adverse events. Although a randomized trial of shorter shifts suggests that a 25% decrease in serious errors might be plausible in medical intensive care units, few errors cause injuries, and effects could differ in other clinical settings.3 Also, the additional handoffs of patient care among residents could increase preventable adverse events.
If the number of preventable adverse events is reduced, will the investments of teaching hospitals in reducing residents' work hours and workloads lead to net savings?84,85,86 A return on investment has been shown for computerized physician order-entry systems, bar-coding systems, and having more registered nurses care for hospitalized patients.50,81,87 However, hospitals are likely to incur losses under the IOM-recommended changes. Averting preventable adverse events may, in addition to reducing the costs of inpatient medical care, avert malpractice claims, preserve the hospital's reputation, and improve patient throughput, but these latter effects on hospital profitability appear to be modest and unpredictable.84 For example, data on the association between negligent injuries and malpractice claims are equivocal, and hospitals bear only 22% of event costs, even when malpractice is considered.88,89,90,91,92,93
From the societal perspective, the cost-effectiveness of the recommendations depends on their success in reducing preventable adverse events and the overall effect of these events on health-related quality of life. Eighty percent of these events cause temporary disability.22 With regard to mortality, regulatory agencies have valued a "statistical life" at $1.7 million to $6.3 million ($2.0 million to $7.4 million in 2006 U.S. dollars).41,94 Although even a 2.5% reduction in preventable adverse events yields a cost-effectiveness ratio within this range ($3.4 billion), the life expectancy of patients who have these events appears to be much shorter than that assumed by the "statistical life" valuations. For example, one study showed that adjustment for the probability of living until hospital discharge if no preventable adverse event occurred reduced mortality rates associated with these events by a factor of four.49
The fact that the recommended changes may be cost-effective for society but cost-generating for teaching hospitals makes financing a complex issue. Possible strategies that teaching hospitals could use to manage the additional costs include reducing residents' salaries, increasing the workload of faculty physicians without increasing compensation, increasing charges to patients, allowing profitability to decrease, reducing clinical services, allowing quality of care to decline, improving efficiency, and securing subsidies — or not implementing the recommendations.
If the recommendations are implemented, oversight will be important because of the financial incentives for hospitals to minimize expenditures. Nonadherence to the 2003 limits can contribute to a program's loss of accreditation and Medicare GME support.95 The ACGME cited 8.8% of programs for nonadherence in 2006–2007, but this rarely affected accreditation.96 Yet researchers have documented widespread nonadherence, lack of attention to handoffs of patient care, and some erosion of educational opportunities.12,13,17,28,97 To address these issues, the IOM committee recommended greater oversight.20
The available data create uncertainty in our estimates. Work-hour data for PGY-1 residents are from 2004. For residents in PGY-2 and beyond, available studies have either focused on one specialty or on one or two hospitals; however, these studies were large, addressed major specialties, and were based at major teaching institutions. Few studies have indicated which substitutes can perform the tasks of residents, but midlevel providers and attending physicians are widely accepted as substitutes. The recommended changes could alter patterns of health care utilization measures such as length of stay, which would have unpredictable effects on costs and clinical outcomes. Additional IOM recommendations, including an additional day off per month, were omitted from our study, but they appear to be unlikely to involve substantial costs. The costs of training substitutes and managing residents' work hours have seldom been described.70 The omission of the costs of providing the additional day off, training substitutes, and managing residents suggests that $1.6 billion could be too low, whereas several sensitivity analyses suggest that this amount may be too high. Although we performed sensitivity analyses for various scenarios, we are unable to predict with certainty how teaching hospitals would choose to implement these recommendations.
Preventable adverse events have not been assessed in teaching hospitals recently. Thomas et al.22 may have overestimated the rate of these events.44,49,98 We excluded studies that assessed rates or costs of one type of event or of negligent adverse events, studies that were limited to specific populations or settings, those that used administrative data, and those conducted in other countries, but the estimates of Thomas et al.22 were not markedly different from estimates in the studies we excluded.23,44,45,46,47,49,99,100 Care costs more at teaching hospitals; therefore, preventable adverse events may also cost more.101,102 Payers sometimes reimburse the costs associated with these events; however, this decreases the likelihood of net savings to hospitals, and Medicare has stopped reimbursing hospitals for the costs associated with certain events.86,89 Reducing residents' motor-vehicle crashes and occupational injuries would provide additional benefits to society.
In conclusion, the IOM has recommended improved implementation of the 2003 reforms, a 5-hour nap during extended shifts, a 16-hour limit for shifts without naps, and reductions in residents' workloads. The labor costs associated with transferring all the excess work to alternative providers would be relatively high ($1.1 billion to $2.5 billion), and the effectiveness of doing so is unknown. If the changes were highly effective, they could prevent patient harm at reduced or no cost to society. However, financial incentives for teaching hospitals are not well aligned with the implementation of these reforms.
Supported by the Institute of Medicine, under contract to the Agency for Healthcare Research and Quality.
No potential conflict of interest relevant to this article was reported.
The views expressed in this article are those of the authors and do not necessarily represent the views of the Institute of Medicine or its committees and convening bodies.
We thank the Institute of Medicine committee members for their input and comments on the original analysis; Daniel Polsky, Ph.D., for his review of previous versions of the analysis; Mark Noah, M.D., Darryl Sue, M.D., and Soma Wali, M.D., for their general insights as directors of residency programs; and Krasimir Karamfilov for assistance with the preparation of an earlier draft of the manuscript.
Source Information
From the Division of General Internal Medicine and Health Services Research, David Geffen School of Medicine at UCLA, Los Angeles (T.K.N., J.J.E.); RAND, Santa Monica, CA (T.K.N., J.J.E.); Stanford University, Stanford, CA (J.B.); and the National Academy of Sciences, Institute of Medicine, Washington, DC (D.M.W., C.U.).
Address reprint requests to Dr. Nuckols at 911 Broxton Plaza, Los Angeles, CA 90095-1736, or at tnuckols{at}mednet.ucla.edu.
References
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||
Related Letters:
Residents' Work Hours
Schwenk T. L., Brenner M. J., Auer R., Aujesky D., Rodondi N., Johns M. M.E., Goldberg J. L., Sandroni S., Blanchard M. S., Meltzer D., Polonsky K. S.
Extract |
Full Text |
PDF
N Engl J Med 2009;
361:928-930, Aug 27, 2009.
Correspondence
This article has been cited by other articles:
HOME | SUBSCRIBE | SEARCH | CURRENT ISSUE | PAST ISSUES | COLLECTIONS | PRIVACY | TERMS OF USE | HELP | beta.nejm.org Comments and questions? Please contact us. The New England Journal of Medicine is owned, published, and copyrighted © 2010 Massachusetts Medical Society. All rights reserved. |
Previous
