Assessing causal effects of observational data requires attaining covariate balance. Unbalanced intervention and control sizes can result in outlier weights, with individuals or subgroups of patients having an oversized impact on the outcome. Identification and correction of outlier weights is often advised, but few guidelines are available. We assessed the impact of assessing and adjusting outlier weights on cost estimates for a national commercial dataset. Analysis involves cost trends, calculated as cost/risk score. Bootstrap sampling with replacement was performed on the baseline data and with weights iteratively trimmed at the 99-95th percentile. Outcomes were the width of the 95% confidence interval (variance) and the point estimate (bias) for baseline and trimmed values. Sensitivity analysis focused on subjects in geographic proximity to the patient in the population with the highest weight. The original population included more than five million patients with at least one preventative care visit to a primary care provider in the previous two years. The subset consisted of 56,786 patients across five contiguous geographic areas. For the full dataset, the maximum analytic weight was 625 (.02% of total weights) with a 99th percentile weight of 6.6. The bootstrap variance was .025; trimming at the 99th percentile resulted in similar percentage changes in bias and variance (22 vs 20% respectively). Additional trimming at the 98th to 95th percentile resulted in continued shifts in bias with minimal improvement in variance. For the subset analysis, the maximum analytic weight was 79 (.06% of total weights), and a 99th percentile weight of 6.4. Subset trimming results were consistent with those of the full sample. Trimming a large population dataset at the 99^th percentile resulted in moderate improvement in variance, with a similar shift in bias. Trimming beyond this point resulted in increases in bias with minimal improvement in variance.