Using generalized estimating equations to analyze longitudinal data in nursing research.

Although use of GEE has become widespread for analyzing longitudinal data in many fields, much research in nursing still uses only the traditional statistical methods. This article illustrates use of GEE, using a data set describing symptoms of women with gynecological cancers, with eight data collection times across a period of 6 months. Three research questions are analyzed: (a) Did number of symptoms in women with gynecological cancer change within 6 months after surgery? (b) Did individual symptom (e.g., pain) in women with gynecological cancers change within 6 months after surgery? (c) Were psychosocial and disease variables (age, cancer site, new diagnosis or recurrence, treatment, emotional distress, and use of symptom management toolkit) associated with individual symptoms (e.g., fatigue) in women with gynecological cancers? For each research question, SAS code and explanation of SAS instructions are described in detail. Missing data and time intervals for longitudinal studies are also discussed.