Induction Mortality In Pediatric Acute Lymphoblastic Leukemia (ALL): a Retrospective Cohort Analysis From the Pediatric Health Systems Information (PHIS) Database, 1999–2009

Abstract 3239

We sought to define factors identifying children at increased risk for mortality during the induction phase of chemotherapy for ALL. We identified a cohort of 11,145 pediatric patients with newly diagnosed ALL using PHIS data from 1999–2009. The PHIS database contains de-identified records of all admissions from 42 children's hospitals across the United States, representing 85% of pediatric admissions to freestanding children's hospitals. Data are internally validated by PHIS, and participating hospitals must have an error rate of <2%. Patients were further validated by manual comparison of induction chemotherapy to published ALL induction standards. Information included in the database are patient demographics, ICD-9 codes, and resource utilization codes for pharmacy, radiology and other procedures. Our strategy to identify newly diagnosed ALL patients included all patients from birth to age 18.99 years with an ICD-9 code for ALL (204.xx) who received one of a series of 3- or 4-drug induction chemotherapy regimens based on standard practices for the era studied. Data were analyzed from the first 30 days after initial admission for ALL. Table 1 shows demographic characteristics, as well as clinical characteristics hypothesized to predict increased mortality. There were 172 patients who died during the study period (1.54%; Table 2). Age <1 year was associated with markedly increased mortality compared to ages 1–9.99 years; patients ages 10–18.99 also had increased mortality. Sex, race and Hispanic ethnicity were not predictive of mortality. We hypothesized that Down syndrome, anthracycline exposure and dexamethasone exposure would be risk factors; however, none of these exposures significantly predicted increased mortality compared to patients without these exposures. Certain interventions were strong predictors of mortality, such as intubation, use of pressors, intubation with concomitant pressors, extracorporeal membrane oxygenation (ECMO), and hemodialysis (Table 2). We tabulated the frequency of ICD-9 codes during the admission in which mortality was observed. The most commonly associated diagnoses were: respiratory failure, acidosis, aplastic anemia, pleural effusion, hemorrhagic stroke, renal failure, sepsis, coagulopathy, complications of a vascular device, hyponatremia, hypotension/shock, pneumonia, and seizure. We are in the process of further analyzing these diagnostic codes for prediction of mortality risk. In summary, we present the largest published dataset of induction mortality in pediatric ALL. Additionally, this is the first use of a nationally representative pediatric ALL dataset that includes patients irrespective of clinical trial enrollment. We are currently performing multivariate analyses using ICD-9 discharge, procedure and utilization codes to define risk factors for induction mortality. Using these analyses, we will develop a model that may allow practitioners to intervene against poor outcomes in high-risk patients. Finally, these data can provide national benchmarks for ALL induction mortality and complication rates that will be critically important in this era of increasing emphasis on patient safety.