Coderch Jordi, Sánchez-Pérez Inma, Ibern Pere, Carreras Marc, Pérez-Berruezo Xavier, Inoriza José M
Grup de Recerca en Serveis Sanitaris i Resultats en Salut (GReSSiReS), Palamós (Girona), España; Serveis de Salut Integrats Baix Empordà (SSIBE). Palamós, Girona, España.
Grup de Recerca en Serveis Sanitaris i Resultats en Salut (GReSSiReS), Palamós (Girona), España; Barcelona Graduate School of Economics, Barcelona, España; Centre de Recerca en Economia i Salut, Barcelona, España.
Gac Sanit. 2014 Jul-Aug;28(4):292-300. doi: 10.1016/j.gaceta.2014.03.003. Epub 2014 Apr 13.
To develop a predictive model for the risk of high consumption of healthcare resources, and assess the ability of the model to identify complex chronic patients.
A cross-sectional study was performed within a healthcare management organization by using individual data from 2 consecutive years (88,795 people). The dependent variable consisted of healthcare costs above the 95th percentile (P95), including all services provided by the organization and pharmaceutical consumption outside of the institution. The predictive variables were age, sex, morbidity-based on clinical risk groups (CRG)-and selected data from previous utilization (use of hospitalization, use of high-cost drugs in ambulatory care, pharmaceutical expenditure). A univariate descriptive analysis was performed. We constructed a logistic regression model with a 95% confidence level and analyzed sensitivity, specificity, positive predictive values (PPV), and the area under the ROC curve (AUC).
Individuals incurring costs >P95 accumulated 44% of total healthcare costs and were concentrated in ACRG3 (aggregated CRG level 3) categories related to multiple chronic diseases. All variables were statistically significant except for sex. The model had a sensitivity of 48.4% (CI: 46.9%-49.8%), specificity of 97.2% (CI: 97.0%-97.3%), PPV of 46.5% (CI: 45.0%-47.9%), and an AUC of 0.897 (CI: 0.892 to 0.902).
High consumption of healthcare resources is associated with complex chronic morbidity. A model based on age, morbidity, and prior utilization is able to predict high-cost risk and identify a target population requiring proactive care.
开发一种用于预测高医疗资源消耗风险的模型，并评估该模型识别复杂慢性病患者的能力。
在一个医疗管理组织内进行了一项横断面研究，使用连续两年的个体数据（88795人）。因变量包括第95百分位数（P95）以上的医疗费用，包括该组织提供的所有服务以及机构外的药品消费。预测变量包括年龄、性别、基于临床风险组（CRG）的发病率以及先前利用情况的选定数据（住院使用情况、门诊护理中高成本药物的使用情况、药品支出）。进行了单变量描述性分析。我们构建了一个置信水平为95%的逻辑回归模型，并分析了敏感性、特异性、阳性预测值（PPV）和ROC曲线下面积（AUC）。
费用超过P95的个体累积了44%的总医疗费用，并且集中在与多种慢性病相关的ACRG3（汇总CRG水平3）类别中。除性别外，所有变量均具有统计学意义。该模型的敏感性为48.4%（CI：46.9%-49.8%），特异性为97.2%（CI：97.0%-97.3%），PPV为46.5%（CI：45.0%-47.9%），AUC为0.897（CI：0.892至0.902）。
高医疗资源消耗与复杂的慢性病发病率相关。基于年龄、发病率和先前利用情况的模型能够预测高成本风险并识别需要积极护理的目标人群。
