A population-based study suggested that adolescent and young adults with acute lymphoblastic leukemia (ALL) have superior survival when treated at specialized cancer centers (SCC; versus community hospitals). However, the association of location of initial care with early mortality (death within 60 days) has not yet been evaluated in children and adolescents and young adults with ALL. Therefore, we examined the impact of location of initial care, sociodemographic factors and complications on early mortality in children (0-18 years) and adolescent and young adults (AYAs; 19-39 years) with ALL.


Using the California Cancer Registry linked to a hospitalization database in California, we identified children and AYAs with first primary diagnosis of ALL who received inpatient treatment within 60-days of diagnosis between 1991 and 2014. SCCs were defined as Children's Oncology Group or National Cancer Institute (NCI)-designated cancer centers for children and NCI-designated cancer center for AYAs. Children and AYAs were classified as receiving all or part/none of their care at a SCC within 60-days of diagnosis. Mulitple logistic regression was used to identify factors associated with receiving all initial care at a SCC and to build a propensity score separately for children and AYAs. Inverse probability-weighted, multivariable logistic regression models estimated the associations between location of care, sociodemographic characteristics, clinical factors and complications with early mortality. Results are presented as odds ratios (ORs) and 95% confidence intervals (CI).


Of the 6,531 patients with ALL in our study, the overall early mortality rate for was 2.1% in children and 6.3% in AYAs. The majority of children received all of their care at an SCC (n=4752, 85.4%) compared to only 36.1% of AYAs (n=1779). Rates of early mortality varied significantly by location of care and age (children: all care at an SCC 1.9% vs part/none 3.2%, p=0.04; AYAs: all care at an SCC 3.9% vs part/none 7.7%, p=0.002). Among children, factors associated with receiving all care at a SCC were younger age, Hispanic race/ethnicity (vs non-Hispanic white), more recent diagnosis and having public/no insurance (vs private). Factors associated with receiving all care at a SCC among AYAs were younger age, non-Hispanic white race/ethnicity (vs Hispanic), public/no insurance, and living in a lower socioeconomic status (SES) neighborhood. In multivariable, inverse probability weighted models, receiving all care at a SCC (versus part/no care) was associated with significantly lower early mortality in AYAs (OR 0.44, CI 0.27-0.74), but not in children (OR 0.61, CI 0.34-1.08). Higher early mortality in children was associated with complications of bleeding (OR 5.73, CI 2.76-11.90), sepsis (OR 2.02, CI 1.01-4.04), renal (OR 4.1, CI 2.17-7.76) and respiratory failure (OR 5.48, CI 2.66-11.29). In AYAs, higher early mortality was associated with older age (>29), having a comorbidity (OR 1.83, CI 1.08-3.10) and complications of liver failure (OR 8.03, CI 2.41-26.76), renal failure (OR 2.45, CI 1.33-4.52) and respiratory failure (OR 11.89, CI 5.87-24.09). Early mortality was not associated with race/ethnicity, neighborhood SES or health insurance for either age group. Rates of complications were similar by location of care in children and AYAs except for thrombosis in children (all: 0.40% vs part/none: 1.00% p=0.03) and renal failure in AYAs (all: 14.3% vs part/none: 10.7% p=0.03).


Our results suggest that treatment at a SCC is associated with a significantly lower early mortality for AYA patients after adjusting for sociodemographic and clinical factors. We observed a borderline difference in mortality by location of care in children, likely because, in contrast to AYAs, the vast majority of children are seen at SCCs. The reduction in mortality at SCCs may be due to access to clinical trials and other specialized services available at these types of institutions.


Wun: Janssen: Other: Study steering committee and research support (site PI); Pfizer: Other: Study steering committee and research support (site PI).

Author notes


Asterisk with author names denotes non-ASH members.