|
Abstract
|
BACKGROUND: In hospital, falls are frequent adverse events. Certain drugs affect the fall risk, therefore studying prescriptions may reveal perilous combinations and support falls prevention. As neurologic diseases frequently increase fall risk, neurologic patients require special attention concerning fall prevention.
AIM: To analyse the performance of the electronic adverse drug reaction check programmes VERIKO(®) and SCHOLZ Datenbank(®) in identifying neurologic patients with a high drug-associated fall risk.
METHOD: Falls in the Department of Neurology in 2016 were matched to fall-free control patients of the same age, sex and principal diagnosis. Their estimated fall risk and other risk factors were compared using univariate and a multifactorial conditional logistic regression. Receiver operating characteristic curves visualised the performance of both programmes. R² for a model with and without software was calculated.
RESULTS: Eighty-seven matched pairs were analysed. In the univariate analyses, VERIKO risk estimations showed a significant correlation to fall events (OR=1.448, CI=1.061-1.975). Additionally, the number of comorbidities (OR=1.086, CI=1.013-1.164), the Hospital Frailty Risk Score (OR=1.085, CI=1.025-1.149), impaired balance (OR=3.6, CI=1.337-9.696), gait abnormality (OR=4.75, CI=1.616-13.962), presence of delirium (OR=3.4, CI=1.254-9.216) and previous falls (OR=8.0, CI=1.839-34.793) were related to high fall risk. Polypharmacy and the number of potentially inappropriate medications did not correlate with fall events. In the multivariate analysis, the Hospital Frailty Risk Score was associated to fall risk (OR=1.390, 95%-CI=1.049-1.842). Both programmes showed an area under the receiver operating characteristics curves < 0.6 and improved the model performance slightly (ΔR² ≤ 0.0006).
CONCLUSION: VERIKO risk estimations correlated significantly to fall events. Nevertheless, both programmes showed little accuracy in identifying drug-associated fall risk.
Language: en |