Auyeung, Tung Wai (TW);Kng, Carolyn Poey Lyn (CPL);Chan, Tak Yeung (TY);Hui, Elsie (E);Leung, Chi Shing (CS);Luk, James Ka Hay (JKH);Sha, Kwok Yiu (KY);Yu, Teresa Kim Kum (TKK);

Author information

J Frailty Aging.2025 Mar 07;14(2):100021.doi:10.1016/j.tjfa.2025.100021

Abstract

BACKGROUND: Electronic health record (EHR) has been in place in many parts of the world. This fits in very well to the frailty index calculation proposed by Rockwood and thus a frailty index can potentially be generated automatically from an EHR database. Therefore, the Hong Kong Hospital Authority (HA) attempted to develop an electronic frailty index (HK eFI), by employing thirty-eight health variables from her own EHR database.

METHODS: Five cohorts of patients aged 60 years or above ever attended any services provided by the Hong Kong HA in the year 2015, 2016, 2017, 2018 and 2019, were included. The HK eFI trajectory with ageing, generated by the five cohorts, were compared to the one described by Rockwood's group. Following the UK eFI method, 4 levels of frailty were categorized, and they were examined whether they were related to mortality, readmission rate and hospitalization patient days.

RESULTS: Each successive cohort consisted of increasing number of patients aged 60 years or above. (2015, 1.14 million; 2016, 1.19 million; 2017,1.25 million; 2018, 1.31 million; 2019, 1.38 million). The gradients of the five trajectories ranged from 0.035 to 0.037, representing an increase in FI approximately 3.6 % annually. The intercept of the curves converged at 0.1, representing the FI at age 60 years was 0.1. Compared to the fit group, the adjusted hazard ratios of mortality of the mild, moderate and severe frail group were 1.77, 3.31 and 6.65 respectively and they were all statistically higher than the fit group. (p < 0.005) Likewise, there was a stepwise increase in readmission rate and hospital patient days utilization with increasing frailty levels.

CONCLUSION: It is feasible to develop an eFI and a biological age trajectory from a large EHR database with local adaptation.