METHODS: We carried out a retrospective cohort study with administrative health databases from Colombian population affiliated in the contributory health insurance scheme. We used an operative definition to select the cohort with diabetes. Incremental cost and cost ratio of diabetes were estimated using an inverse probability weighting of treatment approach to find the causal effect of having the disease. Weights were calculated by a propensity score method using a Random Forest model. The flexibility of this machine learning algorithm allows to have a better specification and bias reduction. Additionally, we reported incremental costs and cost ratios with confidence intervals using bootstrapping and analyzed costs by age groups and complications associated with diabetes.
RESULTS: The estimated prevalence of diabetes was 2834 per 100 000 cases, in 2018. The group with diabetes was comprised 634 015 people and the control group 1 524 808. The calculated annual direct medical cost was $860, for which the incremental cost was $493 and the cost ratio 2.34. The incremental annual cost for some type of complication ranges from $1239 to $2043, renal complication being the most expensive. Incremental cost by age groups ranges from $347 to $878, being higher in younger people.
CONCLUSIONS: Although the cost of diabetes in Colombia ranges among the global averages and is similar to other Latin-American countries, a greater incremental cost was found in patients with renal, circulatory, and neurologic complications.
方法:我们使用来自缴费型健康保险计划所属哥伦比亚人口的行政健康数据库进行了一项回顾性队列研究。我们使用手术定义来选择糖尿病队列。使用治疗方法的逆概率加权估计糖尿病的增量成本和成本比,以发现患有该疾病的因果效应。使用随机森林模型通过倾向评分方法计算权重。这种机器学习算法的灵活性允许具有更好的规范和偏差减少。此外,我们使用自举法报告了增量成本和成本比以及置信区间,并按年龄组和与糖尿病相关的并发症分析了成本.
结果:估计糖尿病患病率为每10万例中2834例,在2018年。糖尿病组包括634015人,对照组为1524808人。计算得出的年度直接医疗费用为860美元,其中增量费用为493美元,费用比率为2.34。某些类型的并发症每年增加的费用从1239美元到2043美元不等,肾脏并发症是最昂贵的。按年龄组划分的增量成本从347美元到878美元不等,年轻人更高。
结论:尽管哥伦比亚的糖尿病费用在全球平均水平中不等,与其他拉丁美洲国家相似,在肾脏患者中发现了更大的增量成本,循环,和神经系统并发症。