This paper studies the performance of 45 Brazilian federal teaching hospitals, within five semesters as from 1998 by connecting and assessing technical efficiency scores and service indicators. The efficiency scores are obtained from data envelopment analysis (DBA) model. The service indicators used are the occupancy rate, the turnover rate and the average length of stay. The approach allows for regional and temporal analysis, besides bed size and clinical specialties (medicine and surgery) insights. By connecting efficiency scores and service indicators in a panel data regression model, some interesting relationships between these variables are achieved. We could find that fixed effects, specific to each hospital, are important, but increases in relative efficiency may be achieved trough decreases in average length of stay, what configures a delicate coordination problem.