OBJECTIVES: Monetary cost, positive values truncated at zero, violates normality assumption when used as the dependent variable in ordinary least squares (OLS) regressions. Log transformation of cost removes the skewness, but the resulting coefficients are not directly interpretable as raw dollars. Simply taking exponent of fitted regression coefficients causes retransformation bias. Duan's nonparametric smearing estimator factors into the mean of the anti-log of the residuals, thus correcting retransformation bias. The goal of this analysis is to apply Duan's smearing technique to retransform logged costs to evaluate the incremental cost of ChHD using claims data from an HMO. METHODS: A 13-item self-assessment questionnaire identifying ChHD and its severity was developed, validated, and mailed to 1,380 randomly selected members of a Massachusetts HMO. Average monthly costs for questionnaire respondents were calculated by the sum of approved and co-payment amounts from claims filed between 4/1/2001 – 12/31/2003 divided by months of observation. OLS regression of logarithm of cost was used, with covariates consisting of a ChHD dummy, and demographic and co-morbid factors. Using Duan's estimator, average monthly incremental cost of ChHD was calculated by multiplying percentage cost increase for ChHD from the OLS regression by predicted average monthly cost for Non-ChHD patients (which is the average cost after removing the effect of the ChHD dummy). RESULTS: 140 of 507 questionnaire respondents were identified as ChHD. Univariate comparison showed no statistical difference in monthly cost between the ChHD and Non-ChHD groups (Non-ChHD, $326.98±$29.52, ChHD $270.87±$23.59, p=0.1383). A skewness and kurtosis test rejected normality. However, multivariate analysis showed that ChHD patients had a statistically significant monthly cost increase of 25.2%(±2.5%) compared to Non-ChHD (p<0.001), amounting to an average monthly incremental cost of ChHD of $70.69(±$7.00) per patient. CONCLUSIONS: Duan's smearing estimator may be valid for inferring incremental cost in OLS regression models of logarithm of cost.