Hospital Variation in Cesarean Delivery: A Multilevel Analysis

Objectives: To assess the issue of hospital variations in Colombia and to contribute to the methodology on health care variations by using a model that clusters the variance between hospitals while accounting for individual-level reimbursement rates and objective health-status variables. Methods: We used data on all births (N = 11,954) taking place in a contributory-regimen insurer network in Colombia during 2007. A multilevel logistic regression model was used to account for the share of unexplained variance between hospitals. In addition, an alternative variance decomposition specification was further carried out to measure the proportion of such unexplained variance due to the region effect. Results: Hospitals account for 20% of the variation in performing cesarean sections, whereas region explains only one-third of such variance. Variables accounting for preferences on the demand side as well as reimbursement rates are found to predict the probability of performing cesarean sections. Conclusions: Hospital variations explain large variances within a single-payer's network. Because this insurer company is highly regarded in terms of performance and finance, these results might provide a lower bound for the scale of hospital variation in the Colombian health care market. Such lower bound provides guidance on the relevance of this issue for Colombia. Some factors such as demand-side preferences and physician reimbursement rates increase variations in health care even within a single-payer network. This is a source of inefficiencies, threatening the quality of health care and financial sustainability. The proposed methodology should be considered in further research on health care variations.