Abstract
Objective
Comorbidity has an important role in risk prediction and risk adjustment modelling in observational studies. However, it is unknown which comorbidity index is most accurate to predict mortality in hip fracture patients. We aimed to evaluate the prediction ability, including discrimination and calibration of Charlson comorbidity index (CCI), Elixhauser comorbidity index (ECI) and Rx-risk index for 30 day- and 1 year mortality in a population-based cohort of hip fracture surgery patients.
Methods
Using the Danish Multidisciplinary Hip Fracture Registry in the period 2014–2018, 31,443 patients were included. CCI and ECI were based on discharge diagnoses, while Rx-Risk index was based on pharmacy dispensings. We used logistic regression to assess discrimination of the different indices, individually and in combinations, by calculating c-statistics and the contrast in c-statistic to a base model including only age and gender with 95% confidence intervals (CI).
Results
The study cohort were primarily female (69%) and older than 85 years (42%). The 30-day mortality was 10.1% and the 1-year mortality was 26.6%. Age and gender alone had a good discrimination ability for 30-day and 1-year mortality (c-statistic=0.70, CI: 0.69–0.71 and c-statistic=0.68, CI: 0.67 −0.69, respectively). By adding indices individually to the base model, Rx-risk index had the best 30-day and 1-year mortality discrimination ability (c-statistic=0.73, CI: 0.72–0.74 and 0.71 CI: 0.71–0.72, respectively). By adding combination of indices to the base model, a combination of CCI and the Rx-risk index had a 30-day and 1-year mortality discrimination ability of c-statistic=0.74, CI: 0.73–0.75 and c-statistic=0.73, CI: 0.73–0.74, respectively. Calibration of indices was similar.
Conclusion
The highest discrimination ability was achieved by combining CCI and Rx-risk index in addition to age and gender. However, age and gender alone had a fair mortality discrimination ability.
Disclosure
The authors report no conflicts of interest in this work. The authors’ affiliation, Department of Clinical Epidemiology is, however, involved in studies with funding from various companies as research grants to (and administered by) Aarhus University.