Medical Record Level-Evaluation of Impact of Demographic and Socioeconomic Factors on Pediatric Neuro-Oncology Outcomes at Children’s Hospital Colorado

Purpose A medical record-level cohort study to investigate demographic and socioeconomic factors influencing treatment, timing of care, and survival outcomes in pediatric patients diagnosed with central nervous system (CNS) tumors. Methods Using electronic health records of patients at Children’s Hospital Colorado from 1986–2020, we identified 898 patients treated for CNS tumors. The primary outcomes of interest were 5-year survival, timing of diagnosis, and treatment. Multivariable logistic regression and Cox regression were used to identify covariates associated with our outcomes of interest. Results We found that age, race, tumor type, diagnosis year, and social concerns influenced receipt and timing of treatment. Age, race, patient rural vs. urban residence, and tumor impacted survival outcomes. Time to presentation and treatment were significantly different between White and minority patients. American Indian/Alaska Native and Black patients were less likely to receive chemo compared to White patients (OR 0.28, 0.93 p = 0.037, < 0.001). Patients with 3 + social concerns were more likely to survive after 5 years than children with no or unknown social concerns (OR 1.84, p = 0.011). However, with an adjusted hazards ratio, children with 2 social concerns were less likely to survive to 5 years than children with no or unknown concerns (OR 0.58, p = 0.066). Conclusions Demographic and socioeconomic factors influence timing of care and survival outcomes in pediatric patients with CNS tumors. Minority status, age, social factors, rural, and urban patients experience differences in care. This emphasizes the importance of considering these factors and addressing disparities to achieve equitable care.


Introduction
While great strides in healthcare in the last decades have substantially improved childhood cancer survival, disparities still exist in outcomes between different children based on demographic and socioeconomic factors [1,2].Previous cohort studies have identi ed disparities in the pediatric CNS cancer population Our primary outcome of interest was 5-year survival.Other outcomes of interest included receipt of treatment (surgery, chemotherapy, targeted chemotherapy, radiation, and clinical trial treatment), and days from symptom onset to rst medical care and diagnosis.Several variables were organized into categorical variables to be adjusted for in the analyses, shown in Table 1.Some covariates were excluded from the nal analysis due to lack of statistical or clinical signi cance, or to eliminate collinearity.The covariates included in our nal analyses were patient gender, age, ethnicity, race, urban residence, rural residence, year of diagnosis, tumor type, tumor grade, primary site, metastatic disease, tumor size, and number of social concerns (sum of insurance concerns, nancial concerns, and transportation concerns).Covariates were chosen based on having a univariate p-value < 0.1.

Statistical Analysis
Logistic regression was used to estimate the unadjusted and adjusted odds ratios for receipt of treatment.For our analysis of time from symptom onset to rst medical visit and diagnosis, we created a binary variable using the median number of days to care to use as a threshold in a binary logistic regression analysis to examine variables associated with more or less time to care.To estimate survival, Kaplan-Meier methods and the log-rank test were used to compare survival across various strata.We then used Cox proportional hazards models to estimate unadjusted and adjusted hazard ratios.All analyses were completed in SAS version 9.4 (SAS Inc., Cary NC) and evaluated as signi cant at p < 0.1, given the single-institution nature and relatively small size of the cohort.

Results
Our nal cohort included pediatric patients diagnosed with CNS tumors during our study period.Table 1 shows the cohort characteristics of patients by demographic, socioeconomic, and clinical variables.

Treatment
Treatment characteristics by demographic, socioeconomic, and clinical variables are shown in Table 2.The univariate treatment analysis showed that female patients were less likely to receive any chemotherapy (OR 0.74, p = 0.022) and cytotoxic chemotherapy (OR 0.68, p = 0.006) than male patients (Table 2).American Indian and Alaskan Native patients were more likely to receive radiation therapy than white patients (OR 2.35, p-value = 0.061).Asian and Paci c Islander patients were less likely to be enrolled in a clinical trial than white patients (OR 0.48, p-value = 0.095).Black children were also less likely than white to receive any type of chemotherapy (OR 0.93 p-value < 0.001).Children from outside the CHCO region were more likely to receive any type of chemotherapy when compared with children within 12.5 miles of CHCO (OR 2.61, p = 0.012).Rural patients were more likely to receive targeted chemotherapy when compared with urban patients (OR 1.41, p = 0.094).Patients with insurance concerns, nancial concerns and transport concerns were more likely than patients without such concerns to receive radiation treatment (OR 1.80, 2.60, 2.23 p-values 0.005, < 0.001, 0.005) and chemotherapy (Table 5).Patients with nancial concerns were also more likely to be enrolled in a clinical trial than patients without nancial concerns (OR 1.66, p-value = 0.002).Diagnosis year, tumor type, grade, metastatic disease status, and tumor size at diagnosis were all associated with differences in treatment, as expected.
Our next step was to perform a multivariable analysis comparing treatment across categories of patients (Table 3).In terms of rural vs. urban residence, patients with unknown status were less likely than rural patients to receive surgery, radiation therapy, general chemotherapy, or cytotoxic chemotherapy.
American Indian/Alaska Native patients were less likely than White patients to receive any chemotherapy or targeted chemotherapy (OR 0.28, 0.13 p = 0.037, 0.023).Patients with one social concern were more likely to receive radiation and targeted chemotherapy when compared to patients with no/unknown social concerns.Patients with three or more social concerns were more likely to receive any form of chemotherapy but were more likely to receive radiation or be enrolled in a clinical trial than patients with no/unknown social concerns.In general, patients with more aggressive diagnoses, such as high-grade glioma or medulloblastoma, were more likely to receive each treatment than patients diagnosed with less aggressive disease, such as low-grade glioma.Older children were more likely to receive surgery and radiation than younger age groups.1-4 year-old patients were the most likely to receive any form of chemotherapy and to be enrolled in a clinical trial.Year of diagnosis played a role in what treatment patients received, with patients diagnosed 2001-2010 more likely to receive surgery.Patients diagnosed before 2010 were more likely to receive cytotoxic chemotherapy and be enrolled in a clinical trial, but less likely to receive targeted chemotherapy.Metastatic disease status also affected which patients received different treatments, as expected.

Diagnostic Timing
We examined the timing from the onset of symptoms to rst medical care in our patients for whom we had diagnostic timing data.Due to known differences in timing of symptoms based on disease type, we ran separate analyses for children with low-grade (N = 237) and high-grade (N = 183) tumors.To compare the time from rst symptoms to care, the median number of days from symptom onset to rst medical care was used as the threshold; 54 days was the median time among low-grade tumors, and 31 days was the median for those with high-grade tumors.Among low-grade tumor patients, children 5-9 years old were more likely than 1-4-year-old children to receive care within 54 days of symptom onset (OR 2.96, p = 0.006, Table 4).Likewise, Black children with lowgrade tumors were more likely than White children to receive care within 54 days of symptom onset (OR 8.56, p = 0.047).Children with low-grade tumors and 3 or more social concerns were also more likely than children with no/unknown social concerns to receive care within 54 days of symptom onset (OR 2.62, p = 0.033).Among those with high-grade tumors, children 5-9 or 15 + years old were less likely to receive care within 31 days of symptom onset than children 1-4 years old (OR 0.28, 0.18, p = 0.004, < 0.001).Hispanic children were less likely than non-Hispanic children with a high-grade tumor to receive care within 31 days of symptom onset (OR 0.52, p = 0.046).Children diagnosed with a high grade tumor between 2001 and 2010 were more likely to get care within 31 days than children diagnosed between 2011-2020 (OR 2.02, p-value = 0.072).

Table 1
Cohort Characteristics.This table lists the number (n) and percentage (%) of patients matching each characteristic on char

Table 2
Univariate analysis of receipt of treatment.This table looks at the odds ratio with a 90% con dence interval and p-value of receiving surgery, radiation, enrollment in a clinical trial, any form of chemotherapy, cytotoxic chemotherapy, or targeted chemotherapy.Signi cant p-values are bolded.

Table 3
Multivariable analysis of receipt of treatment.This table lists the odds ratio with a 90% con dence interval and p-value of receiving surgery, radiation, enrollment in a clinical trial, any form of chemotherapy, cytotoxic chemotherapy, or targeted chemotherapy.Signi cant p-values are bolded.*Sum of insurance, nancial, and transportation concerns

Table 4
Univariate and multivariable analysis of time from onset of symptoms as noted in the patient's history to rst time receiving medical care.Signi cant p-values are bolded

Table 5
Univariate and multivariable analysis of time from the rst medical care for presenting symptoms to diagnosis of a CNS malignancy.Signi cant p-values are bolded odds ratio; ATRT = atypical teratoid rhabdoid tumors; OR = odds ratio; CI = con dence interval, *sum of insurance, nancial, and transportation concerns Table 6 Univariate and multivariable analysis of hazard ratios of survival for variables of interest.Signi cant p-values are bolded Overall survival-all patients (N = 858) aOR = adjusted odds ratio; ATRT = atypical teratoid rhabdoid tumors; OR = odds ratio; CI = con dence interval, *sum of insurance, nancial, and transportation concerns