A simulation study was conducted to calculate power in a generalized estimating equation (GEE) model for a proposed study of the effect of an intervention, designed to reduce lower back injuries which occur among nursing personnel working in nursing homes. Nursing homes were randomly placed in either an intervention or control group and all employees within the nursing home were treated alike. Data from injury statistics suggest that the baseline injury risk for each home can be reasonably modeled with a data distribution. It was assumed that the risk for any particular nurse in any of the nursing homes follows a Bernouli probability distribution expressed as a logic function of fixed covariates, which have values of odds ratios determined from earlier studies. One of the major concerns that was revealed during the study was that the probability of type- I error for the GEE models was significantly higher than 0.05, regardless of the size of the intracluster correlation. The authors found better estimates of type-I error from the robust estimates, although the robust estimates were also significantly higher than 0.05. The effect of ignoring intracluster correlation on power was investigated using GEE and logistic regression. The findings demonstrated that ignoring this intracluster effect in the analysis can cause overestimates of power as well as increased probabilities of type-I error, by as much as 20%. The authors also demonstrated that there is no evidence of intracluster correlation in the data, using the more complicated GEE model.