Surgical volume and center effects on early mortality after pediatric cardiac surgery: 25-year North American experience from a multi-institutional registry

Abstract

Mortality after pediatric cardiac surgery varies among centers. Previous research suggests that surgical volume is an important predictor of this variation. This report characterizes the relative contribution of patient factors, center surgical volume, and a volume-independent center effect on early postoperative mortality in a retrospective cohort study of North American centers in the Pediatric Cardiac Care Consortium (up to 500 cases/center/year). From 1982 to 2007, 49 centers reported 109,475 operations, 85,023 of which were analyzed using hierarchical multivariate logistic regression analysis. Patient characteristics varied significantly among the centers. The adjusted odds ratio (OR) for mortality decreased more than 10-fold during the study period (1982 vs. 2007: OR, 12.27, 95 % confidence interval [CI], 8.52-17.66; p < 0.0001). Surgical volume was associated inversely with odds of death (additional 100 cases/year: OR, 0.84; 95 % CI, 0.78-0.90; p < 0.0001). In the analysis of interactions, this effect was fairly consistent across age groups, risk categories (except the lowest), and time periods. However, a volume-independent center effect contributed substantially more to the risk model than did the volume. The Risk Adjusted Classification for Congenital Heart Surgery, version 1 (RACHS-1) risk category remains the strongest predictor of postoperative mortality through the 25-year study period. In conclusion, center-specific variation exists but is only partially explained by operative volume. Low-risk operations are safely performed at centers in all volume categories, whereas regionalization or other quality improvement strategies appear to be warranted for moderate- and high-risk operations. Potentially preventable mortality occurs at centers in all volume categories studied, so referral or regionalization strategies must target centers by observed outcomes rather than assume that volume predicts quality.

Fig. 1

STROBE-style flow diagram. Note that all operations in eligible years (asterisk) count toward institutional volume, whereas multivariate analysis was performed on a subset, as described in the text

Fig. 2

Annual volume and contribution of each center to the cohort. Centers are arranged (x-axis) by mean annual volume (bars, left y-axis) regardless of the number of years of participation. The percentage contribution of each center (thick and thin lines, right y-axis) accounts for length of participation

Fig. 3

Institutional surgical volumes in the PCCC. a Annual center activity (color) for all 57 PCCC centers (y-axis) over time (x-axis). The duration of participation varied among the individual centers (NA = center’s data incomplete or unavailable for that year) and eight centers were excluded. *High transfer rate, **outside North America, ***fewer than ten cases/year. b Statistical distribution of the 49 included centers’ volumes. The x-axis indicates the years and number of centers in each time period, and the y-axis indicates surgical volume annualized by time period. Boxes represent median and IQR (interquartile range, 25–75th percentiles); whiskers represent range within 1.5 × IQR; and circles represent values outside 1.5 × IQR

Fig. 4

Unadjusted mortality by time and risk category. The percentage of admissions ending in death each year is shown for the overall cohort and by risk category

Fig. 5

Mortality by time and risk category. Risk-adjusted mortality (x-axis, log scale) is shown as odds ratios (circles) and 95 % confidence intervals (whiskers). a Adjusted mortality over time (reference: year 2007). b Adjusted mortality across time periods (TPs) by risk category (reference: TP 5). c Adjusted mortality across risk categories by TP (reference: risk category 1)

Fig. 6

Individual center mortality by volume. The 49 analyzed centers are arranged by mean annual surgical volume (y-axis), with odds of mortality (x-axis, log scale) adjusted for a risk category, time period, age group, and sex, or for b risk category, time period, age group, sex, and volume. Symbols are as above, with centers having significantly increased or decreased odds of mortality highlighted respectively in red or green. For some centers, volume was adjusted up or down slightly to prevent overlap in the figure. This did not change the order of the centers or alter any center’s volume by more than two cases per year