Screening Tools Used by Clinical Pharmacists to Identify Elderly Patients at Risk of Drug-Related Problems on Hospital Admission: A Systematic Review

Abstract

In recent years, a number of studies have examined tools to identify elderly patients who are at increased risk of drug-related problems (DRPs). There has been interest in developing tools to prioritise patients for clinical pharmacist (CP) review. This systematic review (SR) aimed to identify published primary research in this area and critically evaluate the quality of prediction tools to identify elderly patients at increased risk of DRPs and/or likely to need CP intervention. The PubMed, EMBASE, OVID HMIC, Cochrane Library, PsychInfo, CINAHL PLUS, Web of Science and ProQuest databases were searched. Keeping up to date with research and citations, the reference lists of included articles were also searched to identify relevant studies. The studies involved the development, utilisation and/or validation of a prediction tool. The protocol for this SR, CRD42019115673, was registered on PROSPERO. Data were extracted and systematically assessed for quality by considering the four key stages involved in accurate risk prediction models—development, validation, impact and implementation—and following the Checklist for the critical Appraisal and data extraction for systematic Reviews of prediction Modelling Studies (CHARMS). Nineteen studies met the inclusion criteria. Variations in study design, participant characteristics and outcomes made meta-analysis unsuitable. The tools varied in complexity. Most studies reported the sensitivity, specificity and/or discriminatory ability of the tool. Only four studies included external validation of the tool(s), namely of the BADRI model and the GerontoNet ADR Risk Score. The BADRI score demonstrated acceptable goodness of fit and good discrimination performance, whilst the GerontoNet ADR Risk Score showed poor reliability in external validation. None of the models met the four key stages required to create a quality risk prediction model. Further research is needed to either refine the tools developed to date or develop new ones that have good performance and have been externally validated before considering the potential impact and implementation of such tools.