In assessing the relationship between organizational characteristics of health care facilities and patient outcomes, standard questionnaires about the organizational components are often administered to health care facility staff and the results from the questionnaires are then related to the outcomes. The questionnaires used in these studies have often been developed and tested in populations somewhat different from the study populations. Hence, there may be considerable noise in the instruments as they relate to the study participants.
In this paper we consider such a situation when we look at the organizational characteristics of nine neonatal intensive care units (NICUs) in the Washington, DC area. The ultimate objective of the study was to look at the relationship between organizational characteristics and neonatal morbidity and mortality. However, in preparation for this an investigation was made of the organizational characteristics data generated from a standard instrument developed for adult intensive care units by Shortell. The goal of this preliminary investigation was to screen out items (questions) that did not discriminate among hospitals and to develop optimal linear composite scores of the discriminating items to be used in the morbidity and mortality analyses.
