Establishing reliability and validity of the Critical Care Family Satisfaction Survey

Objective: To develop and validate the Critical Care Family Satisfaction Su rvey as a proxy for patient satisfaction. Design: Instrument validation study. Setting and Time Frame: The Medical Intensive Care, Shock Trauma, Acute Cor onary Care, Central Nervous System, Surgical Intensive Care, and Special Ca re units of Lehigh Valley Hospital (Allentown, PA), for the period December 1997 through September 1998. Patients/Participants: One family member for each of 237 critical care pati ents. Intervention(s): Content and construct validity were examined on 37 items a nd 6 constructs thought to measure family satisfaction with the quality of critical care in hospitals. Initially, 14 items and 1 construct were remove d from the questionnaire based on this analysis. It was then administered t o 237 family members. Measurements and Main Results: Factor analysis and confirmatory factor anal ysis using path models were performed. Internal consistency using Pearson c orrelations and Cronbach's alpha, and discriminant validation were also cal culated. Factor analysis yielded a single eigenvalue >1 (3.712), whereas co nfirmatory factor analysis led to the final instrument being reduced to 20 items and 5 subscale constructs. One subscale ("Comfort") performed poorly, indicating the possible need for a four-factor model. Subsequently, intern al consistency assessed by Cronbach's alpha was 0.9101 for the five-factor model and 0.9327 for the four-factor model. Subscale correlations were no l ower than 0.750 for the five-factor model and 0.856 for the four-factor mod el. Conclusions: This study provides support that the Critical Care Family Sati sfaction Survey-which yields five subscales, "Assurance," "information," "P roximity," "Support," and "Comfort"-is reliable and valid. Using five const ructs rather than four is recommended because of the following: a) the inte rnal consistency loss of 0.0226 for the "Comfort" subscale is not enough to warrant its removal, b) a four-factor questionnaire can be administered an d totaled independently of this subscale, c) the need for the fifth constru ct is indicated by this study's results, and d) including the extra data ma y allow for more detailed analysis.