Analysis of the fall‐related risk of pharmacotherapy in Czech hospitals: A case control study

Abstract

Rationale, aims, and objectives: Falls are among the major problems occurring in hospital setting, when drugs are viewed as important modifiable risk factor of falling.The aim was to analyse the effect of pharmacotherapy on the risk of falls in hospitalized patients.Methods: A multicentre prospective case-control study was conducted in 2017 retrieving data from four hospitals in South Bohemia, Czech Republic. An online database was constructed to collect patient and fall-related data. Each fall that occurred during hospitalization was assigned to appropriate controls (no fall during hospitalization) based on gender, age, length of hospitalization, and the number of drugs. Univariate and multivariate correlations were performed with a significance level of P < .05.Results: A total of 222 fall cases (107 males; median age, 81 y) and 1076 controls (516 males; median age, 80 y) were included. According to the first ATC level classification, drugs from groups S, N, and P were significantly associated with fall-related risk compared with controls (P < .05); further analysis of ATC levels showed that only psycholeptics (N05), antipsychotics (N05A), and tiapride were significantly associated with falls. Regression analysis revealed use of psycholeptics N05 (OR = 2.06; 95% CI, 1.56-2.76), or ophthalmologicals S01 (OR = 2.72; 95% CI, 1.37-5.41), as factors with the highest fall-related risk.Conclusions: Apart from the commonly considered fall-risk increasing drugs, other groups, such as ophthalmologicals, should also be considered; however, regarding clinical practice, it is difficult to evaluate the effects of individual drugs in the context of other risk factors of falls, due to the multifactorial nature of falls.

Risco de queda relacionado a medicamentos em hospitais: abordagem de aprendizado de máquina

“…Relaxantes musculares, quimioterápicos, insulina e medicamentos oftálmicos, identificados como fatores de risco em outros estudos, não estão incluídos. (25)(26)(27)(28) Quando os mesmos medicamentos utilizados para calcular o MFRS foram analisados nos quatro algoritmos de aprendizado de máquina, o que apresentou melhor desempenho foi o modelo desenvolvido por meio do algoritmo Naive Bayes. A área sob a curva ROC foi de 0,678 e o índice de Youden alcançado foi de 0,274, superando os respectivos escores de 0,603 e 0,218.…”