Assessing the Estimands and Estimates of Hospitalization Rates in Health Economics and Clinical Medicine

ABSTRACT

Even though data on hospital admissions are widely used in health research, hospitalization‐related estimands measured using these data are not always clearly conceptualized. Consequently, estimators of these quantities can have unclear rationales and undesirable properties. We evaluate three “rate” estimators for measuring hospitalization‐related estimands. Using the Grossman human capital model, we motivate the importance of measuring healthy time. We show that an upper bound on healthy time can be calculated using lengths of hospital stay without assumptions about health status outside the hospital. We illustrate the empirical value of these bounds. Next, we find that an admission rate conventionally used in clinical research is a patient follow‐up time weighted average that lacks a clear basis for the weights. We propose an alternative estimator with more desirable properties and weaker assumptions. We assess its performance using a model of hospital admissions and death. Finally, we evaluate the Centers for Medicare and Medicaid Services (CMS) use of risk‐standardized readmission rates to penalize hospitals by showing that risk‐standardized rates can be sensitive to patient case mix, potentially leading to hospital rankings that do not reflect hospital quality. We propose treating hospital specific intercepts in the CMS risk‐standardization model as a measure of quality.