Factors Associated With High Inpatient Utilization Among Children With Newly Diagnosed Cancer Free

Data were sourced from the Florida 2016 to 2019 State Inpatient Database (SID). The Florida SID is a limited deidentified dataset and, thus, exempt from review by an institutional review board. Inpatient admissions for children aged 0 to 21 years with a cancer diagnosis were identified (n = 2652 patients). Cancer diagnoses were identified by the Clinical Classification Software Refined algorithm.⁶  The first encounter date coded with cancer was used to approximate the date of diagnosis. Patients were followed for 1 year (n = 11 067 admissions). Admissions were divided into planned admissions for diagnosis and treatment and unplanned for all others. Children experiencing a high utilization (HU) of inpatient services were defined as those in the top decile of cumulative length of stay (cLOS) for unplanned admissions.

Multivariable logistic regression was used to identify factors associated with HU adjusted for sex, race, ethnicity, age, payer, cancer type, median income quartile by zip code, and metro/non-metro county of residence. Covariates were obtained from the initial SID admission record. Results were reported as adjusted odds ratios (aOR) with 95% confidence intervals (CI). The average cLOS, cumulative admissions, and number of admissions by month were plotted over time and stratified by patient utilization.

Descriptive statistics of the cohort are shown in Table 1. Figure 1A reveals characteristics associated with HU from multivariable regression. Patients with HU had higher odds of being Black (aOR 1.64 [95% CI 1.12–2.39]) than non-Hispanic White. Those with HU had lower odds of having Hodgkins’s lymphoma (aOR 0.24 [95% CI 0.10–0.50]), leukemia (aOR 0.56 [95% CI 0.38–0.82]), and non-central nervous system (CNS) solid tumors (aOR 0.55 [95% CI 0.38–0.80]) versus CNS tumors. Patients with HU represented 10% of our cohort but accounted for 24.9% of total admission days and 45.2% of unplanned admission days.

Examining admission patterns, the average cLOS of unplanned admissions steeply increased among patients with HU but remained nearly unchanged for those with non-HU (Fig 1B). The average cumulative number of unplanned admissions for patients with HU continually increased over time, whereas unplanned admissions for those with non-HU tapered off (Fig 1C). When admissions were averaged by month, all patients showed the most unplanned admissions occurring in the first 2 months (Fig 1D). However, patients with HU averaged more unplanned admissions at all time points. Importantly, planned admissions were also longer and more frequent in those with HU (Fig 1B–D).

We identified a group of newly diagnosed pediatric oncology patients with dramatically more and longer hospitalizations. These patients were more likely to be Black and have CNS tumors. Although hospitalizations for children with CNS tumors are likely related to their grim prognosis and need for complex surgical resection,⁷  excess hospitalizations among Black children is disconcerting for structural racism, resulting in a lack of outpatient support, toxic stress, material hardship, delays in presentation to care, and discrimination.^5,8 

Our results must be considered with the following limitations. First, a reliance on billing codes to identify our study population and outcome may lead to underreporting the true number of patients diagnosed with cancer, as well as properly classifying unplanned admissions. Second, using the first reported hospital admission with a cancer diagnosis is only an approximation of the cancer date of diagnosis. Third, we were unable to control for potential confounders, such as disease severity at diagnosis and cancer subtype and staging, that influence care and treatment plans.

Despite these limitations, our findings generate critical hypotheses as to whether Black children are differentially burdened by inpatient admissions. Hospitalizations increase the risk for medical complications such as nosocomial infections and thromboembolism,⁹  as well as nonmedical complications, such as loss of employment, divorce/separation, and psychological distress.^2,4  Future efforts are needed to determine the underlying cause of differential utilization and provide targeted support to improve quality care.