To measure healthcare accessibility, a previous study formulated a stochastic cost frontier model, and applied it to distances of communal centres from nearest emergency hospitals in Chile. Based on a larger set of variables, this study re-estimates the distance threshold, and tests alternative specifications and distribution assumptions over the period 2000-05. Complementing the analysis, ordered probit regressions help examine the cross-commune allocation of hospitals with varying complexity of medical services. Thirdly, excess distance estimates are used with other covariates in tobit models of health sector infrastructure investments. Socio-economic, demographic and geophysical conditions, along with non-hospital healthcare facilities, explain spatial inequality of hospitals. Excess distances are robust to different model specifications and distribution assumptions, and do not appear to be compensated by standards of services of the nearest hospitals or new healthcare infrastructure and upgrading investments in support to the communes concerned. Communes with perceived inadequacy in timing of medical attention turn out to often register large location inefficiency too. For the most critical communal cases, the excess distance burden amounts to nearly 80% of the respective, commune-specific 'best practice' hospital distance estimates.