Physician performance improves with experience and training. PICU performance may also reflect the physician staff's training. We hypothesized that severity and diagnosis adjusted PICU mortality in PICUs staffied by critical care fellows would be better than those with only pediatric residents.

Methods: Consecutive admissions to 8 PICUs with critical care fellows and 8 without (all volunteer with intensivists) were studied. Severity (PRISM) and diagnosis adjusted mortality risks were computed using a predictor developed in an independent sample. Evaluations of PICU-level performance used Standardized Mortality Ratios (SMRs) and patient-level analyses used logistic regression analysis. Monitoring and therapeutic modalities were compared with severity-adjusted odds ratios.

There were 2744 admissions (145 deaths) to the 8 fellowship PICUs and 3006(150 deaths) admissions to the 8 nonfellowship units. Hospital characteristics were similar. The crude mortality rates were similar (fellowship: 5.28%; nonfellowship: 4.99%, p =.714). Fellowship PICUs were ranked better than PICUs without such programs (Wilcoxin's rank sum test, p=.020). The 6 best and the worst SMRs occurred PICUs with fellowship programs. Patient-level analyses indicated that outcome was significantly influence by fellowship status. Using two different analytic approaches, the odds of dying in a fellowship hospital versus a nonfellowship hospital were 0.592 (95% confidence interval.468 -.749, p<.0001) and 0.707 (95% confidence interval.523 -.955, p =.024). Fellowship PICUs did more (p<.05) invasive monitoring including intra-arterial and central venous pressure catheters, and more complex therapies such as mechanical ventilation.

Conclusion: In general, PICUs with critical care fellowship programs have better outcomes than those without such programs. However, the presence of a critical care training programs does not insure improved outcomes.