Abstract
Background: Acute lymphoblastic leukemia (ALL) is the most common pediatric cancer and requires intensive and prolonged chemotherapy. There are limited published data describing the number and characteristics of admissions to the hospital following initial diagnosis. This study aimed to describe the number and timing of admissions in children with ALL at hospitals across the United States. Admissions were categorized as expected or unexpected to provide a more comprehensive understanding of the hospitalization patterns for children with ALL.
Methods: A retrospective cohort study was designed using the Pediatric Health Information System (PHIS) database. PHIS, an administrative database that includes inpatient data from over 45 freestanding children's hospitals, contains demographic information, dates of admission and discharge, length of stay (LOS), discharge diagnosis and procedure codes, as well as billing encounters for medications. A cohort of pediatric patients with newly diagnosed ALL was previously identified using ICD-9 discharge diagnosis codes and chemotherapy exposure data. Patients were classified as high risk (HR) if billing data suggested receipt of an anthracycline within the first month of chemotherapy, otherwise patients were considered standard risk (SR). Admission patterns were explored by risk classification, age group, gender, race and ethnicity. Each admission was assessed for utilization of chemotherapy agents or parenteral antibiotics in the first two hospital days. Based on these resource utilization data admissions were hierarchically categorized as "chemotherapy related", "infection related", or "neither chemotherapy nor infection related". The former two categories were considered expected admissions while the third was considered unexpected. Each admission was also classified by the presence of billing codes for resources consistent with intensive care unit (ICU) level of care. Patient follow-up was censored at time of inpatient death or the projected end of treatment (28 months for females, 40 for males). Results: The majority of pediatric ALL patients (94.9% of HR, 87.5% of SR) had one or more admission after diagnosis. HR patients had a median of 7 admissions (range 0-43), with a 5-day median LOS (range 1-414) and median total of 55 hospitalized days (range 1-728). In contrast, SR patients had a median of 3 admissions (range 0-31), with a 4-day LOS (range 1-352) and 24 total hospitalized days (range 1-453). The distribution of admissions was skewed toward the first 6-months after diagnosis (Figure 1). Limited variability in number, duration, and timing of admissions was seen when stratifying by sex, race, or ethnicity.
Of all admissions following diagnosis admission, 26.5% were chemotherapy-related, 49.1% were infection-related, and 24.3% were unexpected admissions. HR patients had a median of 1 chemotherapy-related admission (range 0-23), 3 infection-related admissions (range 0-26), and 1 unexpected admission (range 0-22). In SR patients there were 0 (range 0-21), 2 (0-20), and 1 (0-22), respectively. Children with infant leukemia (diagnosed <1yo) had a median of 11 admissions (range 0-37), median LOS of 5 days (range 1-193) and median of 139 total hospitalized days (range 7-520). They had a median of 5 chemotherapy-related admissions (range 0-14), 4 infection-related admissions (range 0-19), and 2 unexpected admissions (range 0-12).
The need for ICU level of care within an admission was common; 18.7% of HR patients had at least 1 admission that required ICU care, 4.1% had 2 or more. In SR patients, 5.3% had at least 1 admission that required ICU care, 0.4% had 2 or more such admissions.
Conclusion: This study is the first step in benchmarking typical admission rates and characteristics after the index diagnosis admission. Patients who receive anthracyclines at diagnosis experience more frequent admissions early after diagnosis and admissions are more likely to be of higher acuity. Unexpected admissions were present in both groups, and account for nearly a quarter of all hospitalizations after diagnosis. It is imperative that we determine the etiology and risk factors associated with these unexpected admissions in order to identify potentially avoidable inpatient stays. Understanding predictive factors can inform the development of interventions to reduce avoidable admissions and to mitigate associated human and financial costs.
Fisher:Merck: Research Funding; Pfizer: Research Funding.
