Artikel
Using results from a mixed effects regression model analysis for a binary outcome in the prognosis of risks
Suche in Medline nach
Autoren
Veröffentlicht: | 6. September 2007 |
---|
Gliederung
Text
Introduction: Multiple logistic regression is a common technique to analyze medical data with binary outcome [Ref. 1]. In the case of repeated observations of the same sampling unit, simple logistic regression may not be valid and other procedures (e.g. GEE and mixed effects regression models) should be used to account for the correlation [Ref. 2]. Both procedures consider the correlation in a different manner, since the parameters have different interpretation as marginal model, or conditional model, respectively. The latter procedure allows the description of individual effects.
Material and Methods: In our study, newly referred outpatients with lung diseases (COPD, asthma, others) using four kinds of inhalers were included. The main objective of our work was an investigation of the influence of various factors (age, severity of disease, kind of inhaler, and training of the patient) on ineffective inhalation in these patients. Because of multiple observations in the patients (some of the patients used more than one inhaler), a mixed effects regression model for binary outcome was fitted to the data using proc glimmix in SAS 9.1 [Ref. 3]. The results from the model fit were used for the development of a risk prognosis model.
Results: We want to present the structure of the model and the most important results of the model fit. A main factor for the prognosis of ineffective inhalation is the training of the patient. Development, application and implications from the risk prognosis model will also be shown. The estimated risks for ineffective inhalation cover a wide variety.
Discussion: Mixed effects regression models for binary data are a useful tool to analyze correlated data. In extension of the “ordinary logistic regression model” [Ref. 4], results from the model fit could be used for predicting risks for specific patient profiles. A prediction for individual patients, included in the study is also possible.
References
- 1.
- Levy PS, Stolte K. Statistical methods in public health and epidemiology: a look at the recent past and projections for the next decade. Stat Methods Med Res. 2000;9(1):41-55.
- 2.
- Fitzmaurice GM, Laird NM, Ware JH. Applied Longitudinal Analysis. Hoboken, New Jersey: John Wiley & Sons, Inc.; 2004.
- 3.
- The GLIMMIX procedure. Nov. 2005. Available at http://support.sas.com/rnd/app/papers/glimmix.pdf.
- 4.
- Kleinbaum DG, Klein M, Prior ER. Logistic Regression: A Self-learning Text. New York/Berlin/Heidelberg: Springer; 2002.