Identifying Data Elements to Measure Frailty in a Dutch Nationwide Electronic Medical Record Database for Use in Postmarketing Safety Evaluation: An Exploratory Study

Introduction The role of frailty in postmarketing drug safety is increasingly acknowledged. Few European electronic medical records (EMRs) have been used to explore frailty in observational drug safety research. Objective The aim of this study was to identify data elements, beyond multimorbidity and polypharmacy, that could potentially contribute to measuring frailty among older adults in the Dutch nationwide Integrated Primary Care Information (IPCI) database. Methods Persons aged between 65 and 90 years in the IPCI database were identifed from 2008 to 2013. Clinical nondisease, non-drug measurements that could potentially contribute to measuring frailty were identifed and selected if they were recorded in>0.005% of patients and could be included in at least one of three defnitions of frailty: the frailty phenotype model, the cumulative defcit model, and direct evaluations of frailty through standardized frailty scores. The frequency of these measures was calculated. Results Overall, 314,191 (17% of the source population) elderly persons were identifed. Of these, 7948 (2.53%) had one or more of 12 clinical measurements identifed that could potentially contribute to measuring frailty, such as clinical evaluations of cognition, mobility, and cachexia, as well as direct measures of frailty, such as the Groningen Frailty Index. Three of fve measurements required for the frailty phenotype were identifed in<0.5% of the population: cachexia, reduced walking speed, and reduced physical activity; weakness and fatigue were not identifed. The measurements outlined above may be appropriate for the cumulative defcit defnition of frailty, provided that at least 30 defcits, including comorbidities and drug utilization, are evaluated in total. The most commonly recorded item identifed that could potentially be used in a cumulative frailty model was the Mini-Mental State Examination score (N=2850; 0.91%); the only recorded direct measurement of frailty was the Groningen Frailty Index (N=2382; 0.76%). Conclusion Non-disease, non-drug clinical data that could potentially contribute to a frailty model was not commonly recorded in the IPCI; less than 3% of a cohort of elderly persons had these data recorded, suggesting that the use of these data in postmarketing drug safety evaluation may be limited.

• View at Publisher