Risk-Adjusted Variability in Duration of Mechanical Ventilation Across a Network of Community Hospitals

Background: Benchmarking quality of care for mechanically ventilated patient might help identify best practices and catalyze improvements in care across institutions. Ventilator-associated pneumonia (VAP) is the most commonly proposed metric for comparisons but the complexity and subjectivity of VAP measurement make it a poor choice for benchmarking.  Mean duration of mechanical ventilation might be a more suitable metric since it is easy to measure, objective, and responsive to specific care improvement initiatives such as sedative management and ventilator weaning protocols.  Little is known, however, about variability in risk-adjusted mean durations of mechanical ventilation across similar institutions.

Objective: To compare risk-adjusted mean durations of mechanical ventilation across a network of community hospitals.

Methods: We gathered clinical and administrative data on consecutive mechanically ventilated patients from 49 U.S. community hospitals participating in the Institute for Health Metrics Practice Improvement Network between January 1, 2007 and July 1, 2010.  Ventilator and hospital lengths of stay were calculated for each patient, averaged for each hospital, and compared.  We then developed a random effects Poisson regression model to compare risk-adjusted differences in the relative duration of mechanical ventilation for each hospital.  Model covariates included both patient factors (age, sex, time from admission to intubation, medical versus surgical status, Charlson index, and expected length of stay as derived from each patient’s diagnosis-related group code) and hospital factors (total number of beds, critical care beds, discharges, and patient-days per year). 

Results: Data were available on 12,363 episodes of mechanical ventilation.  After excluding patients under age 18, those with ventilator lengths of stay under 1 day, and patients transferred to a second institution while still on a ventilator, 9,010 episodes remained.  The unadjusted mean duration of mechanical ventilation per hospital varied from 2.6 to 14.2 days.  The adjusted mean duration of mechanical ventilation per hospital varied from 4.1 to 16.7 days.  In the random effects model, lengths of mechanical ventilation were significantly different from the population mean for the 8 hospitals with the lowest adjusted mean durations of mechanical ventilation and the 8 hospitals with the highest adjusted mean durations of mechanical ventilation.

Conclusions: Risk-adjusted mean durations of mechanical ventilation varied four-fold across a network of community hospitals.  In-depth analyses of outlier hospitals’ populations and practices of care are merited to try to discern whether residual differences are due to persistent case mix confounding or informative differences in patient care.