Pediatric Hospital Volume and Induction Mortality in Pediatric Acute Lymphoblastic Leukemia (ALL)

Introduction:

Survival in pediatric acute lymphoblastic leukemia (ALL) has increased dramatically over the past 40 years from less than 10% to 85%. Despite these gains, induction mortality rates have been shown to vary between institutions. This variation may relate to hospital factors such as patient volume. A volume-outcome association describes the relationship between how frequently a service is provided by a hospital to the quality of care received at the patient level. In adult care, higher procedural and patient volume has been associated with improved outcomes. Few studies have investigated a volume-outcome relationship in pediatric leukemia. We hypothesized that an inverse relationship exists between a hospital’s pediatric volume and their pediatric ALL induction mortality rates.

Methods:

A retrospective cohort of new onset ALL patients ages 0 to <19 years was assembled using three years of consecutive data from the Pediatric Health Information System (PHIS) and Perspective Data Warehouse (Premier). PHIS is comprised of free-standing children’s hospitals while Premier represents community institutions caring for both adult and pediatric patients. The primary outcome was inpatient mortality in the first 60 days after initiation of induction chemotherapy. Hospital pediatric volume was defined as the average annual number of inpatient discharges for patients ages of 1–19-years. Hospitals were then grouped into tertiles as low, medium or high volume. Descriptive analysis represented by counts and percentages with 95% confidence intervals. Fisher’s exact test and the non-parametric tests for trend were applied to assess association between a hospital’s ALL mortality rate and their pediatric volume tertile using STATA version 13.

Results:

3456 patients with new onset ALL from 75 individual hospitals were included in our analysis. The induction mortality rate for the entire cohort was 0.87% (30/3456) with a range from 0.58% to 1.2% (Table 1). There was no significant inverse linear trend in mortality across the three pediatric volume categories (p= 0.218). Inpatient induction mortality was lowest in the low pediatric volume institutions, with resultant significant increased induction mortality rate when comparing medium volume institutions and high volume institutions to low volume institutions (Table 1). There was no difference in mortality rates between medium and high volume centers (p=0.468).

Conclusions:

Induction mortality in this pediatric ALL cohort is low and consistent with previously published literature. Contrary to our hypothesis, we observed a lower induction mortality rate in lowest volume institutions. While this finding may be real, it is likely that there are unmeasured confounders impacting the volume-outcome association in this ALL cohort. For instance, low pediatric volume centers may transfer sicker or more complex patients to medium or high volume centers due to lack of familiarity or comfort caring for a complicated patient with a rare diagnosis. The current dataset is limited and would require merger of alternate data sources to capture patient transfers from low to medium and high volume hospitals. Future analyses of these data will focus on adjusting for severity of illness at presentation among patients at the medium and high volume hospitals to see if an inverse association exists between volume and mortality.

*97.5% one sided confidence interval

**p value of Fisher’s exact test with reference as low pediatric volume tertile.