A Colombian survey found intensive care mortality ratios were better in private vs. public hospitals Adriana P erez a,b, * , Rodolfo J. Dennis a,c , Martin A. Rondo ´n a , M. Alison Metcalfe d , Kathy M. Rowan e a Clinical Epidemiology and Biostatistics Unit, School of Medicine, Pontificia Universidad Javeriana, Bogota´, Colombia b Division of Biostatistics, School of Public Health, The University of Texas at Houston Health Science Center, 80 Fort Brown SPH, RAHC, Room number 200, Brownsville, TX 78520, USA c Departments of Medicine and Research, Fundacio´n Cardioinfantil, Bogota´, Colombia d Cancer and Public Health Unit, London School of Hygiene and Tropical Medicine, London, UK e Intensive Care National Audit & Research Centre, London, UK Accepted 24 June 2005 Abstract Background: Our main outcome was to identify organizational characteristics that help to evaluate the differences between the inten- sive care mortality ratios adjusted by APACHE II. We incorporated the variation associated with the ranking of institutions simulating its random effects under a binomial distribution. Methods: A nationwide survey on structure, technology, and staffing resources available in Colombian intensive care units during 1997–1998 was conducted. We collected data on admissions from 20 randomly selected adult medical and surgical intensive care units. Results: The mortality ratio from the 20 intensive care units ranged from 0.59 to 2.36; 80% of the intensive care units had a mortality ratio greater than 1. All four intensive care units with the lowest mortality ratio belonged to private institutions, while four of five insti- tutions with the highest mortality belonged to the public sector. Intensive care units in private institutions also had fewer number of beds, lower median length of stay, lower occupancy rates, higher education training for specialists and nurses and fewer emergency nonelective surgical procedures. Conclusion: We successfully accounted for intensive care mortality baseline differences and random effects variations. There were sub- stantial differences between intensive care units in institution type, bed availability, technology, staffing resources, and degree of training, which may have been associated with patient outcome. These results are of crucial importance to track, detect and assess future changes. Ó 2006 Elsevier Inc. All rights reserved. Keywords: Evaluation; Research; Colombia; APACHE; Ranking; League tables; Mortality 1. Introduction Amendments to the 1993 Social Security law in Colom- bia guaranteed basic health care coverage to the worker and his/her family. It was hoped that, by the year 2010, the whole of the population would have access to basic health care [1]. The increasing population in Colombia puts a bur- den on the ancillary services that have the potential to influ- ence dramatically not only the burden of disease but also the costs of health care. In this situation, it is crucial to be able to perform rigorous and unbiased evaluations of the health services provided through the health care system in the public and private sectors, as well as the costs of provision. To do this, we need measures capable of allow- ing rigorous evaluation of health care [2]. In the intensive care setting, it is well known that there are systematic differences in the provision of care, which are reflected in costs but not necessarily in quality [3]. Be- sides measuring processes in health care, it is equally im- portant to measure the outcome of care, as they may help identify inequities in health [4,5]. However, in nonrandom- ized comparisons in such a complex setting as intensive care, it is not easy to be sure that adjustment for confound- ing by case mix (e.g., age, acute severity, surgical status, di- agnosis/reason for admission) and random variation is sufficient [2,6]. In previous studies on patient outcomes in intensive care, it has become increasingly clear that such confound- ing could explain most of the observed variation in hospital mortality between intensive care units [7,8]. There may be, * Corresponding author. Tel.: 956-882-5160; fax: 956-882-5153. E-mail address: adriana.perez@uth.tmc.edu (A. P erez). 0895-4356/06/$ – see front matter Ó 2006 Elsevier Inc. All rights reserved. doi: 10.1016/j.jclinepi.2005.06.004 Journal of Clinical Epidemiology 59 (2006) 94–101