基于数据挖掘的抗菌药物使用强度研究

Data mining-based study on the antibiotics use density

目的探讨时间序列模型在抗菌药物使用强度预测中的应用。方法基于2012年1月至2019年12月的抗菌药物使用强度数据构建简单季节模型,并对2020年数据进行预测和评价。预测发现2020年抗菌药物使用强度偏高,采取措施整改。为预测2022年5-12月数据,应用2012年1月至2022年4月数据构建简单季节模型并评价。将干预前、干预后第1阶段、干预后第2阶段抗菌药物使用强度进行比较。结果基于2012年1月至2019年12月数据构建简单季节模型并将数据回代检验,结果显示,抗菌药物使用强度有90个月实测值均在95%置信区间之内,模拟效果良好。以此模型预测2020年抗菌药物使用强度,结果显示,其值高于48.37,11个月实际值均在预测值的95%置信区间内,相对误差最大为28.27%,相对误差最小为0.59%,采取相关措施整改。为预测2022年5-12月抗菌药物使用强度,采用2012年1月至2022年4月数据构建简单季节模型并将数据回代检验,结果显示,抗菌药物使用强度有117个月实测值均在95%置信区间之内,模拟效果良好。以此模型预测发现8个月抗菌药物使用强度预测值均高于40.67,并且实际值均在预测值的95%置信区间之内,相对误差最大为43.26%,相对误差最小为0.07%,并再次干预。抗菌药物使用强度(每100人每天消耗的抗菌药物DDDs)从干预前的50.70,降至第1阶段45.58、第2阶段38.20。结论简单季节模型能预测抗菌药物使用强度趋势,可以为精细化合理用药管控提供决策支持,提高合理用药水平,优化医疗资源配置。

Objective To explore the application of time series models in the prediction of antibiotics use density(AUD).Methods A simple seasonal model was constructed based on AUD data from January 2012 to December 2019,and data of 2020 was predicted and evaluated.The prediction found that the AUD in 2020 was high,and measures were taken to rectify the situation.A simple seasonal model was constructed and evaluated using the data from January 2012 to April 2022 to predict the data from May to December 2022.The AUD before the intervention,at the first phase and the second phase after intervention were compared.Results A simple seasonal model was constructed based on data from January 2012 to December 2019 and the data were backgenerated and tested,and the results showed that the actual values of AUD of 90 months were within the 95%confidence interval,which showed a good simulation effect.Prediction of AUD in 2020 with this model found that its value was higher than 48.37,and the actual values of 11 months were within the 95%confidence interval of the predicted values,with a maximum relative error of 28.27%and a minimum relative error of 0.59%,and relevant measures were taken to rectify the situation.A simple seasonal model was constructed using the data from January 2012 to April 2022 and the data were backgenerated and tested to predict the AUD from May to December 2022,and the results showed that the actual values of AUD of 117 months were within the 95%confidence interval,which showed good simulation results.Prediction using this model found that the predicted value of AUD of 8 months was higher than 40.67 and the actual values were all within the 95%confidence interval of the predicted values,with a maximum relative error of 43.26%and a minimum relative error of 0.07%,and the intervention was given again.The AUD(DDDs per 100 inpatents per day)decreased from 50.70(before the intervention)to 45.58(at the first phase)and 38.20(at the second phase).Conclusion The simple seasonal model can predict the trend of AUD,which can provide decision support for refining rational drug use control,improve rational drug use and optimize medical resource allocation.