Hospital volume and 30-day mortality for three common medical conditions

Abstract

Background: The association between hospital volume and the death rate for patients who are hospitalized for acute myocardial infarction, heart failure, or pneumonia remains unclear. It is also not known whether a volume threshold for such an association exists.

Methods: We conducted cross-sectional analyses of data from Medicare administrative claims for all fee-for-service beneficiaries who were hospitalized between 2004 and 2006 in acute care hospitals in the United States for acute myocardial infarction, heart failure, or pneumonia. Using hierarchical logistic-regression models for each condition, we estimated the change in the odds of death within 30 days associated with an increase of 100 patients in the annual hospital volume. Analyses were adjusted for patients' risk factors and hospital characteristics. Bootstrapping procedures were used to estimate 95% confidence intervals to identify the condition-specific volume thresholds above which an increased volume was not associated with reduced mortality.

Results: There were 734,972 hospitalizations for acute myocardial infarction in 4128 hospitals, 1,324,287 for heart failure in 4679 hospitals, and 1,418,252 for pneumonia in 4673 hospitals. An increased hospital volume was associated with reduced 30-day mortality for all conditions (P<0.001 for all comparisons). For each condition, the association between volume and outcome was attenuated as the hospital's volume increased. For acute myocardial infarction, once the annual volume reached 610 patients (95% confidence interval [CI], 539 to 679), an increase in the hospital volume by 100 patients was no longer significantly associated with reduced odds of death. The volume threshold was 500 patients (95% CI, 433 to 566) for heart failure and 210 patients (95% CI, 142 to 284) for pneumonia.

Conclusions: Admission to higher-volume hospitals was associated with a reduction in mortality for acute myocardial infarction, heart failure, and pneumonia, although there was a volume threshold above which an increased condition-specific hospital volume was no longer significantly associated with reduced mortality.

Figure 1. Frequency Distribution of 30-Day Rates of Death, According to Medical Condition and Hospital Volume

The upper boundaries of the boxes represent the 75th percentile, the white horizontal line within each box represents the median or 50th percentile, and the lower boundaries of the boxes represent the 25th percentile. Individual data points represent outliers. The I bars represent the highest and lowest values within 1.5 times the interquartile range.

Figure 2. Relationship between Hospital Condition-Specific Volume and Risk-Adjusted Odds of Death from Any Cause at 30 Days, According to Medical Condition

The dashed line represents the fitted regression analysis for acute myocardial infarction, heart failure, and pneumonia, with each data point representing one hospital. The hospital volume was the annual condition-specific volume of patients who were treated at hospitals, averaged over the 3-year study period.

Figure 3. Predicted Effect of an Increase of 100 Patients in Annual Hospital Volume on the Adjusted Odds of Death from Any Cause at 30 Days and Volume Threshold, According to Medical Condition

For each condition, the association between the hospital volume and the patient outcome was attenuated as the hospital's volume increased. For patients with acute myocardial infarction, once the annual volume reached 610 patients, an increase of 100 patients in hospital volume was no longer significantly associated with reduced odds of death. The volume threshold for hospitalization was 500 patients for heart failure and 210 patients for pneumonia.