OBJECTIVE: The first objective was to identify factors associated with prolonged hospital stay (PHS: hospitalized >42 weeks postmenstrual age) in extremely premature (EP: born less than or equal to 28 weeks gestation) infants. The second objective was to identify a PHS best-performing benchmark center. METHODS: This study was a retrospective cohort analysis of infants born < or =28 weeks gestation and admitted to one of 12 tertiary centers between January 1998 and October 2001. Risk-adjusted odds of PHS, defined as hospitalization beyond 42 weeks postmenstrual age, and the competing outcome, mortality, were assessed using logistic regression models. RESULTS: Among 3892 EP survivors who had complete data for multivariable analysis, 685 (18%) had PHS. Variables contributing to PHS included chronic lung disease (oxygen use at discharge home or 36 week postmenstrual age) (OR 6.75; 95% CI: 5.04 to 9.03), necrotizing enterocolitis requiring surgery (OR 13.83; 95% CI: 8.05 to 23.76), and >two episodes of late-onset sepsis (OR 2.39; 95% CI: 1.66 to 3.44). Centers' risk-adjusted PHS odds differed from the reference center, which had the lowest incidence of PHS and mortality (overall P-value <0.0001). Mortality contributed to PHS, but in an opposite direction compared to other factors. Centers with lowest PHS odds were among those with highest mortality. CONCLUSIONS: These findings suggest that reduction of CLD, surgical NEC, and late onset sepsis could reduce PHS in EP infants. Risk adjusted odds of PHS and mortality are both crucial for selecting a PHS best-performing center.