Development and validation of a clinical score to identify hospitalised patients at high risk of drug-related problems.

Abstract

Background: Drug-related problems (DRPs) are a major health concern, with half being preventable and potentially resolvable through the application of pharmaceutical care (PC). However, performing PC to all hospitalised patients is unfeasible due to staff shortages coupled with an increasing number of patients. Hence, a risk score for identifying patients at high risk of DRPs is needed. This study aimed to develop and validate a DRP risk score for hospitalised patients.

Method: A prospective cohort study was conducted in a tertiary hospital in Northern Thailand. Adult patients (≥ 18 years) admitted to medical wards were included. DRPs were identified by clinical pharmacists specialising in internal medicine. Multivariable logistic regression analysis was used to construct a risk score. The score was validated using bootstrapping, and three risk groups were created based on both probability and severity. Score performance was assessed with the area under the receiver operating characteristic curve (AUROC), calibration plot, sensitivity, and specificity.

Results: Among 1350 eligible admissions, 155 (11.48%) experienced at least one clinically preventable DRP. The DRP risk score included 6 predictors, namely age ≥ 65 years, chronic cardiac disease, number of drugs used prior to admission, parenteral administration (excluding parenteral nutrition), drugs with special instructions, and drugs with a high potential for drug-drug interactions. The AUROC was 0.709 (95% CI 0.672, 0.751), with good calibration (calibration slope of 0.928, intercept 0.004). Patients with a score < 4 were classified as low risk, while score ≥ 8 indicated high risk. A score of 4 yielded a sensitivity of 93.55% and a specificity of 34.48%, whereas a score of 8 demonstrated a sensitivity of 43.87% and a specificity of 83.01%.

Conclusions: The DRP risk score has the potential to identify patients at risk of DRPs. External validation is needed to enhance its generalisability. Integration into automated systems may support timely pharmacist interventions.