This paper develops an initial estimation of the health cost-effectiveness threshold in the Colombian health system, from the perspective of the supply-side. This approach defines a theoretical and empirical framework, explicit and consistent, that can be replicated in different health settings. In the supply-side approach, the threshold is defined as the opportunity cost of allocating resources to a new technology, in terms of the health benefits displaced or forgone as a consequence of their unavailability to finance other alternatives that compete for the same budget; and its estimate is based on the quantification of the causal effect of the spending on health benefits. To carry out the research, an unbalanced panel-type econometric model with three dimensions was developed: 1) the Colombians insurers Entidades Administradoras de Planes de Beneficios (EAPB); 2) different groups of health conditions, based on the International Shortlist for Hospital Morbidity Tabulation of the World Health Organization; and 3) a temporary period from 2012 to 2016. The health outcome used were the Life-Years Gained (LYG), calculated using life tables, illness adjusted life expectancy and survival curves methodologies. The expected endogeneity was tried to be resolved through an instrumental variables approach. The estimation was carried out using two-stage ordinary last squares. To run the model, a novel database was built from multiple institutional sources available in the country from the period 2012-2016: Estudios de Suficiencia, Estadísticas Vitales, Registro Individual de Prestación de Servicios, Base de Datos Única de Afiliados, and the financial information of the Superintendencia de Salud. Due to the availability of the information, the estimate was delimited to the universe of the contributory regime and the technologies included in the Plan de Beneficios con cargo a la Unidad de Pago por Capitación (PBS-UPC). As instrumental variables, we explore the use of the financial reports of the EAPB each year, and the per capita expenditure of new technologies included in the PBS-UPC between 2012 and 2016 for each EAPB, each health conditions groups, and each year. The results of the econometric model, without adjusting for the instrumental variables, calculate a coefficient between the expenditure and the LYG of 0.0933 not statistically significant, which implies a cost-effectiveness threshold of $8,772,177 per LYG. Taking into account the instrumental variable of financial reports, the estimated coefficient was 5,521 statistically significant at a 95% confidence level, and a threshold of $157,775 per LYG. On the other hand, the estimation with the instrumental variable of the per capita expenditure of new technologies included in the PBS-UPC between 2012 and 2016, did not yield logical and statistically significant results. Taking into account this scenarios, it is considered that the best initial estimation of the cost-effectiveness threshold in the Colombian health system is $8,772,177 per LYG. To our knowledge, this is the first empirical estimate of the threshold per YLG from the supply-side approach for Colombia, Latin America, and for any low-middle income country. The main limitations of this paper are detailed below. First, the estimate is limited to the universe of the contributory regime and the technologies included in the PBS-UPC, so a future research agenda, if the information sources allow it, would be the inclusion of these two universes within the proposed model, Second, other measures beside the LYG are recommended in the literature, such as the Quality-Adjusted Life Years (QALY) and the Disability-Adjusted Life Years (DALY), which could be taken into account to estimate thresholds for different health outcomes. Third, a limitation of this paper is related to the impossibility in identifying and quantifying a good instrument that corrects the expected endogeneity between spending and health benefits. Future research could modify the estimation of the threshold if an appropriate instrument with the sufficient information is found.
