Identification problems are addressed by utilizing both propensity score matching and two-stage estimation approaches.
We controlled for a wide range of covariates using both propensity score-matched analysis and logistic regression analysis.
In general, both propensity score method and MCMC method are recommended for the imputation of continuous variable [ 21].
We used both propensity methods and an instrumental variable (IV) approach to account for imbalances in baseline patient risk.
Stratified models on deciles of both propensity scores were used to adjust for likelihood to use statins and fibrates during follow-up period.
In some of the studies, both adjusted results (OR, or RR, or hazard ratio (HR)) and count data (the numbers of survivors and nonsurvivors) were provided; or both propensity matched results and no propensity matched results were available.
Despite this, both propensity and individual matching were successful in creating balance between cases and matched controls on case mix factors, although balance was closer for individually matched control patients.
In conclusion, this simulation study revealed that, even in case of small study samples or low prevalence of treatment, both propensity score matching and inverse probability of treatment weighting can yield unbiased estimations of treatment effect.
In addition, to assess the potential role of CS on late mortality, we calculated propensity scores for having presented CS, using logistic regression analysis 1) in patients alive at hospital discharge and at 30 days, and 2) in patients who were alive at one year (c-statistic = 0.76 for both propensity scores).
To minimize the potential of confounding factors, each patient in the atypical LAT group was matched with up to three unique patients in the oral AP group based on both propensity scores, using the 2.5 percentiles of the propensity score distribution and exact matching factors.
