Does Identifying Frailty from ICD-10 Coded Data on Hospital Admission Improve Prediction of Adverse Outcomes in Older Surgical Patients

Main Article Content

Lara Harvey
Barbara Toson
Ian Harris
Robert Gandy
Jacqueline Close


As the population ages, increasing numbers of older adults are undergoing surgery. Outcomes for older people are known to be worse than younger people following surgical procedures, and identifying which patients stand to benefit from surgery can be challenging.

Frailty is recognised as a major contributor to poor outcomes, however assessing frailty clinically is time-consuming and not routinely undertaken. Using data available from electronic medical records can potentially provide the opportunity to routinely screen for frailty electronically at time of admission.

Objectives and Approach
This population-based external validation study aimed to: 1. assess the performance of the Hospital Frailty Risk Score (HFRS) in the prediction of adverse outcomes (mortality, prolonged length of stay (LOS) and 28-day readmission), 2. to determine optimal age-groups and lookback periods and 3. compare HFRS performance against the Charlson Comorbidity Index (CCI).

Hospital and death data for individuals (n=487,197) aged >50 years admitted under a surgical specialty to all public/private hospitals in NSW, Australia, 2013-2017 were linked. Logistic regression models were tested for each outcome of interest. Area under receiving operator curve (AUC) and Akaike information criterion (AIC) were assessed for each model.

For prediction of 30-day, all models performed better than age and sex alone; however adjusting for CCI (AUC 0.76) provided marginally better prediction than adjusting for HFRS (AUC 0.75). Models consistently performed better in the younger age-group (50-65), providing excellent discrimination (AUC 0.82). In contrast, all models had poor ability to predict prolonged-LOS (AUC range 0.62- 0.63) or readmission (AUC range 0.62-0.65). Using a 5-year lookback period did not improve model discrimination over using a 2-year period.

Conclusion / Implications
Adjusting for frailty using the HFRS did not improve prediction of 30-mortality over that achieved by the CCI. Neither HFRS nor CCI were useful for predicting prolonged-LOS r 28-day unplanned readmission.

Article Details

How to Cite
Harvey, L., Toson, B., Harris, I., Gandy, R. and Close, J. (2020) “Does Identifying Frailty from ICD-10 Coded Data on Hospital Admission Improve Prediction of Adverse Outcomes in Older Surgical Patients”, International Journal of Population Data Science, 5(5). doi: 10.23889/ijpds.v5i5.1503.

Most read articles by the same author(s)