A comparison of the application of STOPP/START to patients' drug lists with and without clinical information

Abstract

Background: A European screening tool (STOPP/START) has been formulated to identify the prescribing of potentially inappropriate medicines (PIMs) and potential prescribing omissions (PPOs). Pharmacists working in community pharmacies could use STOPP/START as a guide to conducting medication use reviews; however, community pharmacists do not routinely have access to patients' clinical records.

Objective: To compare the PIM and PPO detection rates from application of the STOPP/START criteria to patients' medication details alone with the detection rates from application of STOPP/START to information on patients' medications combined with clinical information.

Setting: Community Pharmacy.

Method: Three pharmacists applied STOPP/START to 250 patient medication lists, containing information regarding dose, frequency and duration of treatment. The PIMs and PPOs identified by each pharmacist were compared with those identified by consensus agreement of two other pharmacists, who applied STOPP/START criteria using patients' full clinical records.

Main outcome measure: The main outcome measures were: (1) PIM and PPO detection rates among pharmacists with access to patients' clinical information compared to PIM and PPO detection rates among pharmacists using patients' medication information only, and (2) the levels of agreement (calculated using Cohen's kappa statistic (k)) for the three most commonly identified PIMs and PPOs.

Results: Pharmacists with access to patients' clinical records identified significantly fewer PIMs than pharmacists without (p = 0.002). The three most commonly identified PIMs were benzodiazepines, proton pump inhibitors and duplicate drug classes, with kappa (k) statistic agreement ranges of 0.87-0.97, 0.60-0.68 and 0.39-0.85 respectively. PPOs were identified more often (p < 0.001) when clinical information was considered. The three most commonly identified PPOs were: bisphosphonates, β2-agonists and anti-platelets, with kappa (k) statistic agreement ranges of 0.89-1.0, 0.50-0.80 and 0.5-1.0 respectively.

Conclusions: PIM detection is likely to be overestimated using STOPP and PPO detection underestimated using START when STOPP/START is used in isolation of clinical information. Agreement for a selected number of criteria for which clinical information is not required is good, suggesting that some criteria may be reliably deployed without clinical information during a medicines use review. However, for STOPP/START criteria to be deployed more effectively by pharmacists, access to the full clinical record is recommended.