OBJECTIVES: Propensity score matching (PSM) is a method used to address selection bias in observational studies and to date has been used mainly in survival analysis. This study has used PSM when determining antenatal costs for alternative referral modes (via telemedicine (TM) or direct referral) for obtaining specialist advice for pregnant women at risk of a fetal cardiac anomaly. METHODS: Three district hospitals (DH) in south-east England were offered the use of a fetal cardiology telemedicine service. Two hospitals (DH1 and DH3) continued to refer all patients directly to London, while the third hospital (DH4) used both referral modes. A logistic regression model was fitted to women in DH4 in order to predict what proportions of women in DH1 and DH3 would have been seen via TM, had that service been taken up. PSM was then used to match "TM" cases to "direct referral" cases. A total cost per woman was obtained for all antenatal resource use incurred during the study and the costs were adjusted for those predicted as TM cases. Finally, the costs for the predicted modes and matched modes were compared. RESULTS: The logistic regression model predicted 153 women be assessed via TM and the remainder (n=84) as direct referrals. Mean antenatal cost per patient for the TM group was slightly higher (stg606), although not significantly so, compared with the direct referral group (stg561). After applying PSM, 66 TM cases were matched to 55 direct cases. The magnitudes of the PSM costs (TM group, stg671, and direct group, stg551) were similar to the logistic regression model. Comparisons of cost results by referral mode and analytical method were not statistically significant. CONCLUSIONS: PSM indicated that the cost results from the logistic regression model were reliable. We conclude that there is a role for PSM in economic evaluation.