HLTH5623 Module 2 crude and adjusted rates paper example

The Better Hospital on Paper: Crude and Risk-Adjusted Heart Failure Mortality at a Community Hospital and a Tertiary Center

Student Name

American College of Education

HLTH5623: Epidemiology and Public Health for Healthcare Administrators

Module 2 Assignment

Instructor Name

May 8, 2028

The Crude Comparison

A fictional health system created for this paper owns a 150-bed community hospital and a 500-bed tertiary center 30 miles apart. Its quality dashboard reports 30-day mortality for patients admitted with heart failure. Last year, the community hospital admitted 500 heart failure patients, of whom 49 died within 30 days, a crude mortality rate of 9.8 percent. The tertiary center admitted 1,000 heart failure patients, of whom 126 died, a crude rate of 12.6 percent. The system's board asked why patients were more likely to die at its flagship hospital and whether cardiology leadership there needed to change.

A crude rate is the number of events divided by the total population at risk, with no adjustment for differences in the characteristics of the people in each group. Celentano and Szklo (2019) explain that crude rates are appropriate for describing the actual burden of an outcome in a population but can be misleading for comparing populations that differ in factors strongly related to the outcome, such as age or severity of illness. Crude rates answer the question of what happened; they do not answer the question of which hospital performed better.

Why the Patients Differ

The two hospitals do not admit the same patients. The tertiary center receives transfers from the community hospital and from six other hospitals in the region, mostly patients too sick to be managed locally: those in cardiogenic shock, those needing mechanical circulatory support and those with advanced kidney disease. Its heart failure patients are older on average, more often have multiple chronic conditions and more often arrive with low blood pressure or elevated troponin. The community hospital keeps patients with less severe disease and transfers the sickest.

That pattern means the crude comparison is confounded by case mix. A patient admitted to the tertiary center was, on average, more likely to die before any care was given, simply because of who that patient was. Comparing crude rates attributes that difference in risk to the hospitals rather than to the patients. Iezzoni (2013) describes risk adjustment as the process of accounting for patient characteristics that affect outcomes independently of the care received, so that differences in outcomes can be more fairly attributed to differences in quality.

Indirect Standardization, Step by Step

Indirect standardization compares the number of deaths each hospital actually observed with the number it would be expected to have if its patients experienced the average risk of death for patients like them. The expected number comes from a risk model, a regression built on a large reference population that estimates each patient's probability of death from characteristics recorded at admission: age, sex, prior heart failure admissions, comorbidities such as kidney disease and diabetes and, where available, clinical measures. Adding up the predicted probabilities for all of a hospital's patients gives its expected deaths.

Applied to the two hospitals, the risk model predicts 44.0 deaths among the community hospital's 500 patients, an expected rate of 8.8 percent, and 138.0 deaths among the tertiary center's 1,000 patients, an expected rate of 13.8 percent. The observed-to-expected ratio is the observed deaths divided by the expected deaths. For the community hospital, it is 49 divided by 44.0, or 1.11: about 11 percent more deaths than expected. For the tertiary center, it is 126 divided by 138.0, or 0.91: about 9 percent fewer deaths than expected. Multiplying each ratio by the reference population's overall rate, 11.5 percent, gives a risk-standardized mortality rate of 12.8 percent for the community hospital and 10.5 percent for the tertiary center. Once patient risk is taken into account, the ranking reverses: the hospital that looked worse was treating sicker patients and doing better than expected with them.

How Public Reporting Does It

The method used for public reporting of hospital heart failure mortality in the United States follows the same logic with more refinement. Krumholz et al. (2006) developed an administrative claims model for profiling hospital performance on 30-day heart failure mortality that uses hierarchical logistic regression, which accounts for the clustering of patients within hospitals, and they reported that it performed well compared with a model based on medical record data. The measure compares each hospital's predicted deaths, reflecting its own performance, with the deaths expected for a hospital of average performance treating the same patients.

One feature of the hierarchical method matters for the community hospital. Hospitals with few patients have unstable rates, and the model pulls their estimates toward the overall average, a property called shrinkage. With 500 cases a year the community hospital is large enough for a stable estimate, but a smaller hospital with 50 heart failure admissions would see its apparent excess mortality substantially moderated by the model.

Limits of Adjustment

Risk adjustment is only as good as the variables in the model. Claims-based models capture diagnoses and prior use but not measures of severity at arrival, such as blood pressure or laboratory values, and they can be influenced by coding: a hospital that documents comorbidities more thoroughly will appear to have sicker patients and a more favorable observed-to-expected ratio. Adjustment also cannot account for differences that are legitimately part of quality, such as a hospital's decision to transfer patients quickly. The community hospital's ratio of 1.11 is therefore a signal to investigate, not a verdict. The review should examine whether deaths occurred among patients who might have been transferred earlier, whether guideline-directed medications were started before discharge and whether documentation of comorbidities was complete.

Explaining the Result to the Board

The board asked a simple question and deserves an answer it can repeat. A useful way to explain the result is with a comparison the members already understand: judging hospitals by crude death rates is like judging surgeons by how many of their patients die without asking how sick the patients were before surgery. The tertiary center takes the region's sickest heart failure patients, and on average those patients would be expected to die at a rate of 13.8 percent; fewer of them died than expected. The community hospital's patients would be expected to die at 8.8 percent; slightly more died than expected.

The explanation should also be honest about uncertainty. With 49 observed deaths, the community hospital's ratio of 1.11 could reflect chance as well as care, and a confidence interval around the ratio would likely include 1.0. The appropriate message for the board is therefore not that the community hospital performs poorly but that its result warrants a structured review, while the tertiary center's result is reassuring rather than a reason for change in its leadership.

Conclusion

The crude heart failure mortality rates suggested that the tertiary center was the weaker hospital. Indirect standardization, comparing observed deaths with deaths expected for each hospital's patients, reversed that conclusion: the tertiary center had fewer deaths than expected and the community hospital more. Crude rates remain useful for describing burden, and adjusted rates are necessary for comparing performance, but both depend on the quality of the data behind them. The system's dashboard should report crude rates, expected rates and observed-to-expected ratios side by side, so that the board asks the right question of the right hospital.

References

Celentano, D. D., & Szklo, M. (2019). Gordis epidemiology (6th ed.). Elsevier.

Iezzoni, L. I. (Ed.). (2013). Risk adjustment for measuring health care outcomes (4th ed.). Health Administration Press.

Krumholz, H. M., Wang, Y., Mattera, J. A., Wang, Y., Han, L. F., Ingber, M. J., Roman, S., & Normand, S.-L. T. (2006). An administrative claims model suitable for profiling hospital performance based on 30-day mortality rates among patients with heart failure. Circulation, 113(13), 1693-1701. https://doi.org/10.1161/CIRCULATIONAHA.105.611194