The Effect of Alternative Case-Mix Adjustments on Mortality Differences Between Municipal and Voluntary Hospitals in New York City
Published In: HSR, Health Services Research, v. 29, no. 1, 1994, Apr. 1994, p. 95-112
Posted on RAND.org on December 31, 1993
Looks at how mortality rate differences in voluntary and municipal hospitals in New York City are affected by case-mix adjustment methods. More specifically, it addressed the question of whether hospital quality of care varies if the coded secondary diagnosis represents conditions that patients bring to the hospital or conditions that result from care received in the hospital. The sample included 10,000 random admissions for patients hospitalized with one of six conditions in New York City during 1984-1987. The article compares regression models that adjust for age and gender, principal diagnosis, secondary diagnoses that are unlikely to result from care received during the hospitalization (limited other diagnoses), all secondary diagnoses irrespective of whether they are due to care received (full other diagnoses), previous diagnoses, and other variables. For five of the six conditions there was higher mortality in the municipal hospitals when the limited other diagnoses algorithm was used, (i.e., when the model included secondary diagnoses that were more likely to be present prior to the admission than caused by the admission). However, when all diagnoses were entered into the model, only two of the six conditions showed significant mortality differences between the municipal and voluntary hospitals. This article demonstrates that judgments of quality based on administrative data are sensitive to how 'sickness at admission' is defined. It is relevant to current policy discussions about adding a digit to the diagnosis code to indicate whether a diagnosis was present before admission or was a result of the hospitalization.