OBJECTIVES: Estimation of cost-of-illness typically involves the analysis of skewed medical costs that include large outliers. Log transformations are frequently used to overcome these problems.  Linear regression models (OLS) are then applied to the transformed data.  The estimated model coefficients are retransformed back to the linear scale using the smearing approach. Implementing this approach in statistical packages requires customized programming. We propose an alternative to using log transformations: Generalized Linear Models (GLM) with a log link function.  We compare the performance of both models in estimating cost-of-illness. METHODS: We derived data from a large administrative database representing 143,593 discharges from 39 US hospitals from January 2004 to December 2005. We estimated total medical costs among hospitalized patients attributable to hyponatremia. Using a cross-validation approach, we compared the performance of two models: log transformed OLS with smearing and GLM with a log link function and a normal error distribution.  We used the Root Mean Squared Error (RMSE) and the Mean Absolute Error (MAE) to assess model performance.  Covariates in both models included patient age, gender, race, geographic region, Deyo-Charlson comorbidity index, primary diagnosis, teaching status of hospital, and admission source.  All analyses were contacted using SAS®. RESULTS:  The GLM with log-link and a normal error distribution had both the smallest RMSE (23,688) and MAE (11,304) compared to the log transformed OLS with smearing (24,057 and 11,392, respectively).  Furthermore, by using GLM, there was no need to compute a retransformation estimate, since the log link function relates the response mean to the original scale. CONCLUSIONS: In this cross-validation study, GLM outperformed OLS with smearing. GLM is easier to implement using SAS® with no need for retransformation estimates.  Because of its ease of use and statistical accuracy, GLM is a useful alternative to log-transformed OLS models with smearing, when estimating cost-of-illness.