OBJECTIVES: The objective of this paper is to empirically compare inverse probability of treatment weights (IPTW) adjustment with propensity score matching method using the Thomson Reuter's MarketScan commercial claims database. METHODS: There are three treatment groups in this study: treatment with nasal corticosteroids, with nasal antihistamines and combination treatment. We applied both the IPTW method and the traditional matching by propensity scores. In the propensity score matching, we used nearest neighbor method to match the patients in the nasal corticosteroids group with those in the antihistamines group. A second matching was implemented between the nasal corticosteroids and combination treatment group. Alternatively, a generalized multinomial logit model was estimated to obtain the propensity scores. The conditional probabilities of receiving the particular type of treatment given the pre-treatment factors were used as the generalized propensity scores. A propensity score weight, also referred to as IPTW, was calculated for each case in each treatment group as the inverse of the generalized propensity score. Descriptive Statistics were obtained and difference between the groups tested after applying IPTW and propensity score matching. Logistics regressions of probability of any respiratory infection were estimated using both methods. We compared the point estimates and standard errors of treatment effect using two methods. RESULTS: Both IPTW and propensity score matching methods reduced the differences in the pre-treatment factors across treatment groups. In the multivariate regressions, patients treated with corticosteroids were found to have significantly lower probability to have any respiratory infection compared with those treated with intranasal antihistamine when using the propensity score matched sample. However, using the IPTW method, we found that the treatment effects had larger standard errors and were not significant. CONCLUSION: IPTW is less time-consuming compared with traditional matching procedure. However, the IPTW weighting method generates bigger standard errors compared with the traditional matching method.