Demystifying Performance Measures for Hospitalists: Attribution Challenges

A hospital leader assigns the hospital medicine division chief to reduce the division’s length of stay (LOS). The division chief chooses LOS as a measure and focuses on individuals’ LOS by attributing each hospitalization’s LOS to the discharging hospitalist. The division chief reviews a seven-day hospitalization (with an expected LOS of five days). Three hospitalists saw the patient during their hospitalization: hospitalist A saw the patient for the first two days; hospitalist B saw the patient for the next four days, during which the patient was medically cleared and awaiting skilled nursing facility placement on day five; and hospitalist C saw the patient for the last day and discharge. Hospitalist C is concerned that this attribution method is unfair and that they are being punished for discharging this patient by chance with a longer-than-expected LOS. Which hospitalist is responsible for the extra two days over the expected LOS?

Why Attribution Matters

Attribution is the foundation on which valid, actionable performance measures are built. Problems with attribution have been a persistent challenge throughout the “Demystifying Performance Measures for Hospitalists” series.^1,2 As seen in the case, difficulty arose when trying to tie a hospitalization outcome to a single hospitalist, where the problem is not which performance measure is selected but rather the decision to set the attribution level. Therefore, before choosing any performance measure, leaders must first determine the attribution level.

Two Key Decisions: Attribution Level and Measure Type

For successful performance improvement, leaders must choose both the attribution level and the measure type (see Figure 1). Attribution level can be individual or group, while measures can be process- and outcome-based. Specifically, process-based measures focus on tasks performed by a single person during hospitalization (e.g., discharge orders placed before noon). In contrast, outcome-based measures focus on outcomes that occur on or after discharge (e.g., LOS).

The Attribution Challenge in Hospital Medicine

Attribution is especially problematic in hospital medicine, given care fragmentation and the interdependent nature of work with hospital processes. In contrast to primary care, where one physician is assigned to one patient, the case had three physicians during a single hospitalization, a common occurrence in practice.^3,4 Therefore, the relationship between patient outcomes and individuals in hospital medicine is less straightforward than in primary care.

In hospital medicine, group attribution is often more defensible given the fragmented care structure, as it avoids assigning outcomes to single hospitalists where multiple hospitalists share responsibility. Individual attribution is difficult to determine, especially with outcome-based measures, because a single hospitalist rarely controls the entire course of a hospitalization. As illustrated in the case, attributing outcomes to the discharging hospitalist can be problematic because the hospitalist may not have been responsible for most of the care during the hospitalization. When attribution and measurement level are misaligned, they can lead to both system-level and individual consequences.

System-Level Consequences of Getting Attribution Wrong: Failed Improvement

Poor attribution decisions can cause initiatives to fail at the system level, wasting resources and preventing them from achieving the ultimate goal. Performance measures with group attribution and outcome-based measures offer the greatest potential for systemic improvement, given the nature of hospital medicine’s work structure. However, while this combination is helpful for performance tracking, it makes intervention design more difficult.

Paradoxically, despite the granularity, individual attribution with either measure type can also lead to system-level failure. Inaccurate individual attribution could lead to improvement efforts with the wrong underlying assumptions. In one study, different attribution formulas led to an 88-percentile change in one individual’s LOS ranking, meaning that their rank reflected the attribution method rather than underlying performance.⁵ Any intervention based on inaccurate assumptions risks wasting resources without any improvement.

For outcome-based measures, regardless of attribution level, not understanding the underlying process drivers risks targeting improvement efforts at the wrong things. LOS, for example, as an outcome-based measure, is influenced by factors such as weekend services, placement delays, and season, with higher LOS in winter for heart failure patients.^6-8 Therefore, efforts to improve LOS may be unproductive because the underlying process driver is seasonal variation.

For process-based measures, regardless of the attribution level, there must be a statistical link between the process-based measure and the goal, as such measures are inherently proxies. Otherwise, any effort to improve the process-based measure will not lead to improvement. As reviewed in a previous article in the series, while discharge before noon (DBN), a process-based measure for LOS, is a reasonable choice that aligns with individual attribution, there is limited scientific support that earlier discharge timing reduces LOS.^9,10 On the other hand, there are alternative process-based measures statistically associated with LOS, as detailed previously in this series, including day-weighted LOS or individual billing patterns, which offer more scientifically validated measures of LOS. Even with well-designed performance measures optimized at the systemic level, individual-level consequences can still occur.^1,5,11

Individual-Level Consequences of Getting Attribution Wrong: Burnout and Disengagement

Group attribution combined with outcome-based measures seems like an attractive choice for improvement efforts in hospital medicine. However, it creates challenges at the individual level, as incorrect attribution or measure type erodes professional vitality. Professional vitality depends on fairness and autonomy, both of which are threatened by mismatched attribution when individuals feel like their efforts are not reflected in the outcome.¹²

Group attribution, combined with outcome-based measures, while more accurate, can limit the individual’s feeling of control over the outcome, as seen in hospital-acquired infections, patient satisfaction, and LOS.^1,2,13 When using LOS as an outcome-based measure with group attribution, if one hospitalist improves their performance only to see a colleague not put forth the same effort, it violates that hospitalist’s sense of fairness over the improvement process, increasing burnout.

Individual attribution with outcome-based measures, conversely, risks unfairness when circumstances beyond the individual’s control affect the outcome. If the individual does not have a direct impact on the measure, there is a risk of loss of autonomy, as the individual may feel they lack the resources to complete the objective.¹²

Mirroring the systemic consequences, using process-based measures with limited scientific links to the desired outcome, regardless of attribution type, risks burnout by adding more tasks without any link to patient outcomes. If hospitalists put a great deal of effort into improving DBN but, in the end, there is no meaningful change in LOS, that increases burnout. Additionally, poor performance measures increase the risk of incentivizing gaming, where meeting the performance measure becomes the goal as opposed to improving patient care.¹⁴

Conclusion

Attribution decisions shape whether performance measures are fair, meaningful, and ultimately successful. As illustrated in this article, hospital medicine’s fragmented care structure makes the complexity and consequences of attribution uniquely high. While there is no perfect combination of attribution level and measure type, the opening case highlights these trade-offs in practice when selecting both a measure and an attribution strategy. The first step for leaders is to be aware of the limitations and consequences of the combination they have chosen, and to determine whether it is consistent with their goals. In the next article in this series, we explore the next step, with practical approaches hospitalist groups can use to make attribution more meaningful.

The authors are members of SHM’s Performance Measurement and Reporting Committee, which created this series to explore quality measures common in hospital medicine. 

Dr. Roseman

Dr. Carmichael

Dr. Sinha

Dr. Iyer

Dr. Roseman is a med-peds academic hospitalist and assistant professor of medicine and pediatrics at UMass Chan Medical School–Baystate in Springfield, Mass. Dr. Carmichael is a hospitalist, researcher, and assistant professor of hospital medicine at Intermountain Health, a research fellow in the Intermountain Healthcare Delivery Institute in Murray, Utah, and an adjunct assistant professor of medicine at the University of Utah in Salt Lake City. Dr. Sinha is a hospitalist at Beth Israel Deaconess Medical Center and affiliated community hospitals in Plymouth, Mass. Dr. Iyer is a hospitalist in internal medicine at UCI Health and a clinical professor in medicine at the University of California, Irvine, both in Orange, Calif.

References

1. Abbasi S, et al. Demystifying performance measures: length of stay. The Hospitalist website. https://www.the-hospitalist.org/hospitalist/article/37452/quality-improvement/demystifying-performance-measures/. Published August 1, 2024. Accessed July 7, 2026.
2. Miller T, et al. Demystifying performance measures for hospitalists: HCAHPS. The Hospitalist website. https://www.the-hospitalist.org/hospitalist/article/37005/business-of-medicine/demystifying-performance-measures-for-hospitalists-hcahps/. Published May 1, 2024. Accessed July 7, 2026.
3. Epstein K, et al. The impact of fragmentation of hospitalist care on length of stay. J Hosp Med. 2010;5(6):335-338. doi:10.1002/jhm.675.
4. Goodwin JS, et al. Variation among hospitals in the continuity of care for older hospitalized patients: a cross-sectional cohort study. BMC Health Serv Res. 2021;21(1):552. doi:10.1186/s12913-021-06584-0.
5. Herzke CA, et al. A method for attributing patient-level metrics to rotating providers in an inpatient setting. J Hosp Med. 2018;13(7):470-475. doi:10.12788/jhm.2897.
6. Shyu M, et al. Analysing Monday discharges to identify lost opportunities for weekend discharge. Intern Med J. 2023;53(4):625-628. doi:10.1111/imj.16062.
7. Toh HJ, et al. Factors associated with prolonged length of stay in older patients. Singapore Med J. 2017;58(3):134-138. doi:10.11622/smedj.2016158.
8. Akintoye E, et al. Seasonal variation in hospitalization outcomes in patients admitted for heart failure in the United States. Clin Cardiol. 2017;40(11):1105-1111. doi:10.1002/clc.22784.
9. Bartlett C, et al. Demystifying performance measures for hospitalists: discharge before noon. The Hospitalist website. https://www.the-hospitalist.org/hospitalist/article/35197/clinical-guidelines/demystifying-performance-measures-for-hospitalists-dcbn/. Published August 1, 2023. Accessed July 7, 2026.
10. Burden M, et al. Discharge in the a.m.: a randomized controlled trial of physician rounding styles to improve hospital throughput and length of stay. J Hosp Med. 2023;18(4):302-315. doi:10.1002/jhm.13060.
11. Pierce L, et al. Individualized average length of stay: a timelier, provider-level LOS metric. J Hosp Med. 2024;19(6):539-541. doi:10.1002/jhm.13339.
12. National Academy of Medicine Committee on Systems Approaches to Improve Patient Care by Supporting Clinician Well-Being. Taking Action Against Clinician Burnout: A Systems Approach to Professional Well-Being. Washington, DC: National Academies Press (US); 2019. doi:10.17226/25521.
13. Bruti C, et al. Demystifying performance measures for hospitalists: CAUTI and CLABSI. The Hospitalist website. https://www.the-hospitalist.org/hospitalist/article/40144/quality-improvement/demystifying-performance-measures-for-hospitalists-cauti-and-clabsi/. Published December 1, 2025. Accessed July 7, 2026.
14. Tenbensel T, et al. Gaming New Zealand’s emergency department target: how and why did it vary over time and between organisations? Int J Health Policy Manag. 2020;9(4):152-162. doi:10.15171/ijhpm.2019.98.