Investigation of the Appropriate Threshold for Warning Dosage and Development of a Predictive Logistic Regression Model to Detect Dose- Error of Prednisolone Tablets
- VernacularTitle:プレドニゾロン錠の誤用量検出を目的とした処方 1 日量の最適閾値の検討および予測ロジスティック回帰モデルの構築
- Author:
Hiroyasu SATO
1
;
Yoshinobu KIMURA
2
;
Masahiro OHBA
3
;
Yoshiaki ARA
4
;
Susumu WAKABAYASHI
5
;
Hiroko NOMURA
6
;
Hiroaki WATANABE
7
Author Information
- Keywords: steroids; prednisolone; medication error; prediction model; dose audit
- From:Japanese Journal of Drug Informatics 2023;25(3):157-163
- CountryJapan
- Language:Japanese
- Abstract: Objective: The wrong dose of high-risk drugs such as oral steroids is a serious issue that needs to be addressed. This study aims to determine the appropriate upper tolerable dose threshold and to develop a multi-variable logistic regression model to detect dose-errors in oral prednisolone tablets.Methods: Data on Prednisolone prescriptions were obtained from a single center. Out of the data collected, positive cases consisted of cases where dose-related modifications were made. A univariate logistic regression model was developed with the current daily dose. In the model, the Youden Index was used to determine the upper tolerable dose threshold. The investigation was done to determine whether the performance of the multivariate model was improved by adding clinical department and previous prescription information as variables.Results: Univariate models (AUC: 0.645) with only current daily doses and estimated optimal thresholds of 6 mg/day or 11 mg/day, respectively were determined to be appropriate. Including variables improved the performance of the predictive model; the best performing model (AUC: 0.840) was derived when the following variables were entered: “current daily dose,” “current prescription days,” “clinical department,” “daily dose of the previous prescription,” and “prescription days of the previous prescription”.Conclusion: A single upper tolerance limit is insufficient to determine dose adequacy for prednisolone tablets owing to their broad clinical dose range. Itmay be possible to develop a high-performance dose audit support model by adding information.
