Mortality for hip replacement surgery: risk adjustment comparisons

Background The lengthening of life expectancy and improvements in surgical techniques have resulted in a worldwide growing trend in number of hip replacement (HR), accompanied by a decline in fatal outcomes. Short term adjusted mortality for HR is acknowledged indicator of orthopedic surgery quality, but few studies have investigated the ability of several risk adjustment (RA) models to predict outcomes. Our study aims to: 1) evaluate the in-hospital and 30-day mortality in hospitalized patients for HR; 2) compare the performances of two RA algorithms. Methods We studied the hospital discharge records (HD) of a retrospective cohort of hospitalized patients undergoing HR surgery from 2000 to 2005 at hospitals of Tuscan Region in Italy. The outcomes considered in the study were in-hospital mortality and 30-Day mortality after surgery. Two RA tools were adopted to predict the outcomes: All-Patient Refined Diagnosis Related Groups (APR-DRG) system, based on information of HD considered, and Elixhauser Index (EI), referring to the admitting diagnosis of the past three years. Logistic regression models were applied for the analysis of the performance of the two models. C statistic (C) was used to define their discriminative ability. Results The number of HD studied was 26.277, 70% women and 85% patients 65+. In-hospital and 30-Day crude mortality were 1,3 and 3%. Female gender was found to be a significant (p < 0,05) protective factor using the APR-DRG (OR 0,64 for in-hospital and 0,51 for 30-Day mortality) and for EI (OR 0,55 for in- hospital and 0,48 for 30-Day mortality). Among EI comorbid- ities, heart failure and liver diseases were associated with in- hospital mortality (OR 9,29 and 5,60; p < 0,05); at 30 days, the association is reduced (OR 6,36 and 3,26; p < 0,05). Increasing Age and APR-DRG class of risk were predictive for all the outcomes. Discriminative ability of EI was reasonable both for in-hospital and for 30-Day mortality (C 0,79 and 0,68), while it was good for APR-DRG (C 0,86 and 0,82). Conclusions Our study found that gender, age, EI comorbidities and APR- DRG risk of death are predictive factors of outcomes. Although APR-DRG has shown a slightly better performance in predicting mortality, it has an economic cost, while EI is freely available. Key messages Female, age, risk classes determined with APR-DRG/EI are predictive variables of in-hospital/30-Day mortality. RA tools are useful because allow a standardized comparison of Hospital Routine Data