# Reliability and validity of the Swedish indicator ‘Drugs that should be avoided in older people’—an appraisal of a set of potentially inappropriate medications

## TL;DR

This study evaluates a Swedish indicator for identifying inappropriate medications in older people, finding it reliable but not valid for assessing treatment adequacy.

## Contribution

The study provides a novel appraisal of the Swedish PIM indicator's reliability and validity in older patients.

## Key findings

- The indicator reliably detects potentially inappropriate medications (PIMs) with high inter-rater agreement (kappa: 0.89).
- The indicator does not effectively predict inadequate drug treatment (AUC: 0.56).
- Only 4% of patients had PIMs requiring medical action after consensus evaluation.

## Abstract

To analyse the reliability and validity of the Swedish indicator ‘Drugs that should be avoided in older people’.

From a previous study that included consecutive primary care patients ≥ 65 years of age, all patients ≥ 75 years of age were analysed. Two physicians independently screened their medication lists and medical records, applying the Swedish indicator which includes potentially inappropriate medications (PIMs): long-acting benzodiazepines, drugs with anticholinergic action, tramadol, propiomazine, codeine, and glibenclamide. The clinical relevance of identified PIMs was independently assessed. Thereafter, the physicians determined in consensus whether some medical action related to the drug treatment was medically justified and prioritised before the next regular visit. If so, the drug treatment was considered inadequate, and if not, adequate.

A total of 1,146 drugs were assessed in 149 patients (75‒99 years, 62% female, 0‒20 drugs per patient). In 29 (19%) patients, at least one physician identified ≥ 1 PIM according to the indicator at issue; 24 (16%) patients were concordantly identified with ≥ 1 such PIM (kappa: 0.89). Of 26 PIMs concordantly identified, the physicians concordantly assessed four as clinically relevant and 12 as not clinically relevant (kappa: 0.17). After the consensus discussion, six (4%) patients had ≥ 1 PIM according to the studied indicator that merited action. Using the area under the receiver operating characteristic (ROC) curve, the indicator did not outperform chance in identifying inadequate drug treatment: 0.56 (95% confidence interval: 0.46 to 0.66).

The Swedish indicator has strong reliability regarding PIM detection but does not validly reflect the adequacy of drug treatment.

The online version contains supplementary material available at 10.1007/s00228-024-03700-x.