Japanese medical expenditures have increased rapidly in recent years and are predicted to continue rising. Indeed, it remains uncertain whether the current Japanese medical system can be sustained. In this paper, we a...Japanese medical expenditures have increased rapidly in recent years and are predicted to continue rising. Indeed, it remains uncertain whether the current Japanese medical system can be sustained. In this paper, we analyzed the medical expenditures for educational hospitalization of patients with type 2 diabetes. A dataset of 6173 patients from 36 hospitals was used in the analysis. The sample period was from July 2008 to March 2012. We analyzed the medical expenditures in two ways. First, we analyzed the length of hospital stay (LOS). Next, we analyzed the daily expenditure per patient. The Box-Cox transformation model was used in the first analysis and the ordinary least squares method in the second. Comorbidities and complications prolonged LOS and increased daily expenditures. The LOS was significantly different among hospitals. On the other hand, the differences in daily expenditures among hospitals were relatively small, such that LOS was the main determinant of medical expenditures. Previous studies suggested that LOS could be shortened without degradation of medical quality. Moreover, LOS could be controlled by introducing proper critical paths and improving and standardizing educational programs. Hence, it would be possible to control the medical expenditures for this disease.展开更多
目的基于机器学习构建和验证肝移植术后患者重症监护室(intensive care unit,ICU)住院时间延长(prolonged ICU stay,pLOS-ICU)的预测模型。方法回顾性纳入2013年4月至2024年4月的新疆医科大学第一附属医院接受肝移植并收治ICU的成年患...目的基于机器学习构建和验证肝移植术后患者重症监护室(intensive care unit,ICU)住院时间延长(prolonged ICU stay,pLOS-ICU)的预测模型。方法回顾性纳入2013年4月至2024年4月的新疆医科大学第一附属医院接受肝移植并收治ICU的成年患者。基于10种机器学习方法建立预测模型,通过比较受试者工作特征曲线下面积(area under the curve,AUC)、临床决策曲线等评估模型预测能力,使用可解释机器学习算法(shapley additive explanations,SHAP)对变量进行解释。结果本研究纳入230例患者,分为训练集162例,内部验证集68例。在训练集中对术前人口学特征、基础疾病、术中情况、术后早期实验室指标及支持治疗等共20个变量进行筛选,最终选取术后8 h中性粒细胞占比、术后8 h血红蛋白、术后8 h白蛋白、术后8 h非结合胆红素、术后24 h丙氨酸氨基转移酶(alanine aminotransferase,ALT)、术后是否使用多巴胺及术后是否使用肠外营养支持7个变量进行模型构建,其中Logistic模型为最佳模型(训练集AUC=0.797,内部验证集AUC=0.819)。采用SHAP值评估特征重要性,其中早期肝功能指标、营养指标和全身炎症反应在pLOS-ICU预测中占据核心地位。结论术后早期肝功能指标、炎症状态、循环与营养支持及基础疾病是肝移植患者PLOS-ICU的关键预测因素,SHAP值分析有助于个体化风险评估与早期干预决策。展开更多
文摘Japanese medical expenditures have increased rapidly in recent years and are predicted to continue rising. Indeed, it remains uncertain whether the current Japanese medical system can be sustained. In this paper, we analyzed the medical expenditures for educational hospitalization of patients with type 2 diabetes. A dataset of 6173 patients from 36 hospitals was used in the analysis. The sample period was from July 2008 to March 2012. We analyzed the medical expenditures in two ways. First, we analyzed the length of hospital stay (LOS). Next, we analyzed the daily expenditure per patient. The Box-Cox transformation model was used in the first analysis and the ordinary least squares method in the second. Comorbidities and complications prolonged LOS and increased daily expenditures. The LOS was significantly different among hospitals. On the other hand, the differences in daily expenditures among hospitals were relatively small, such that LOS was the main determinant of medical expenditures. Previous studies suggested that LOS could be shortened without degradation of medical quality. Moreover, LOS could be controlled by introducing proper critical paths and improving and standardizing educational programs. Hence, it would be possible to control the medical expenditures for this disease.
