OBJECTIVES: estimate the per capita income and the global attention expenditure for a person that is part of the contributive health subsystem, through an econometric model, using a clinical diagnosis group of patients, which permit the sufficiency resources evaluation. METHODS: Using an econometric model developed from information detailed health expenditure of 9.6 million members attended and unattended in four different insurers, he sought to build an explanation of expenditure information 2013, using three categories of variables: sociodemographic (age, sex), chronic (ten groups) and risk factors. To estimate the function a multiple regression was used to identify statistically significant variables that explain spending the best model, low goodness of fit and reporting criteria. Reliability tests on individual basis and robustness tests for each variable and the model generally applied. Furthermore, deviations between different insurers were identified. RESULTS: The final linear function includes 18 different variables, all significant at 10%, with an R2 of 7.03%. The estimation shows that men are spending on average lower than that of women. The differences in betas by age groups were not significant, but different in the groups with chronic diseases and comorbidity. Can be estimated average costs for patients with chronic diseases. When testing model fit on the same database, there is a goodness of fit of 99%, where health spending in real per capita was $ 447,375 versus estimated by the model equals $ 447,370. CONCLUSIONS: estimating a model that includes variables associated with the disease gives higher accuracy than when only demographic variables are used and allow a more equitable distribution of risk-based resources and could be a better alternative in defining the insurance premium for a country like Colombia.