We agree with Joel Ray that the ICD-9 coding system may not adequately ensure retrieval of all relevant cases. To ensure that we captured all relevant charts, we also reviewed other diagnostic categories such as myositis, gangrene, gas gangrene and erysipelas.1 We also reviewed all charts of children with group A β-hemolytic Streptococcus cultured from sterile sites during the time period of the study.
With respect to doing a sample size calculation, our goal was to collect all of the possible cases of necrotizing fasciitis at our institution over the last 16 years. Our problem was not with the sample size but with the power to discern major factors associated with necrotizing fasciitis, because of the small number of cases (8). For this reason, the reported results only suggested a trend and did not confirm it. Consequently we did not do a retrospective power calculation.
Gary Liss refers to a multivariate analysis and the fact that we were unable to obtain odds ratios from conditional logistic regression. To clarify, we did not do a multivariate analysis. Because of the small number of patients, odds ratios could not be obtained from a conditional logistic regression. Instead, we verified each estimate using a logit estimate of the odds ratio, adjusting for matching.
Reference
- 1.