Payer and Race/Ethnicity Influence Length and Cost of Childhood Cancer Hospitalizations

Abstract

Background

Health disparities related to race, ethnicity, socioeconomic status and insurance status impact quality, access and health outcomes for children. Medicaid is a proxy for poverty and restricted access to health care. The goal of this study was to determine if there are discrepancies in the length and cost of hospitalizations between admissions covered by Medicaid or commercial insurance for pediatric patients with cancer.

Methods

Childhood cancer related admissions were identified from the 2012 Kids Inpatient Database (KID) using the International Classification of Diseases, Ninth revision. Length of hospitalization and cost of hospitalization were compared amongst hospitalizations paid for by Medicaid or commercial insurance. Total admission charges were converted to costs using cost-to-charge ratios and survey weighting methods were used for all analyses. Linear multiple regression models for both length of hospitalization and cost were developed to include patient level factors (race, sex, age, diagnosis, reason for admission).

Results

In 2012, there were 104,597 childhood cancer related admissions. Hospitalizations paid for by Medicaid were significantly longer than those paid for by commercial insurance. Hispanic ethnicity was associated with higher cost of hospitalization regardless of payer, and black race was associated with higher costs within the Medicaid population.

Conclusions

This analysis identifies differences in healthcare utilization for pediatric cancer related admissions paid for by Medicaid compared with commercial insurance. Prolonged hospitalizations and increased costs create burdens on children and their families, medical delivery systems and third-party payers. Further exploration into the causes of these disparities is warranted.

Background

Children with cancer are a population with high inpatient utilization, requiring admissions for chemotherapy, surgery, and infectious complications. Although childhood cancer (CC)-related admissions account for 5% of non-newborn pediatric hospitalizations, they account for 12% of non-newborn pediatric inpatient hospital costs.¹ In 2009, CC hospitalizations were 8 days longer with 70% higher costs per hospital day than pediatric hospitalizations for other conditions, comprising 1.9 billion dollars.¹ In adult patients with cancer, insurance-related health disparities are well documented, however, less is known in the pediatric oncology population.^2–5

After the creation of the Children’s Health Insurance Program in 1997 (www.medicaid.gov/chip), most children in the United States (U.S.) have access to health insurance. The Patient Protection and Affordable Care Act of 2010 extended Medicaid coverage to children whose family income is ≤133% of the federal poverty level (www.medicaid.gov/medicaid/eligibility), and in 2016, the National Center for Health Statistics estimated the rate of uninsured children at 5%.⁶ Nonetheless, health disparities related to race, ethnicity and socioeconomic status (SES) are pervasive and impact quality, access and health outcomes for children.^7–9 Medicaid status impacts length of hospitalization, cost and outcomes for several pediatric conditions including diabetes and asthma.^10–16 A growing body of evidence suggests medical complications and outcomes are impacted by SES, race, ethnicity and other sociodemographic factors including insurance coverage.¹⁷ Both the causes and impacts of these differences deserve further study in order to optimize the delivery of equitable care. The objective of this study was to determine whether coverage by Medicaid is associated with the length and cost of hospitalizations for CC patients.

Methods

Childhood cancer related admissions were identified from the 2012 Kids Inpatient Database (KID), a publication of the Healthcare Cost and Utilization Project from the Agency for Healthcare Research and Quality.¹⁸ It includes data from 80% of hospital discharges for U.S. pediatric patients. The 2012 database includes 3,195,782 unweighted de-identified admission records from 5,118 hospitals in 44 states.¹ Childhood cancer related admissions were identified using International Classification of Diseases ninth revision, clinical modification diagnosis (ICD-9-CM) codes. Admissions were included in our analysis if any of the 25 discharge diagnoses contained an ICD-9-CM code for cancer. Cancer diagnoses were categorized as leukemia (204.X-207.X), lymphoma (200.X-201.X), central nervous system (CNS) tumor (191.X, 192.X, 194.3, 194.4) or solid tumor ((170.X140.X, 150.X, 171.X–175.X, 179.X–189.X, 190.X, 193.X, 194.0–194.29, 194.5–194.9, 195.X and 209.0–209.3).¹⁹ Admissions with diagnoses of more than one category of cancer were categorized as “Dual”. Chronic leukemia and benign/premalignant neoplasms were excluded.

The KID categorizes race/ethnicity as White, African American/Black, Hispanic/Latino, American Indian/Alaska Native, Asian/Pacific Islander, Multiracial and Unknown. Age was categorized as <1 year, 1–4 years, 5–9 years, 10–14 years, and 15–19 years. Primary payers were grouped into Medicaid (includes Medicare), Commercial, or Other (includes self-pay, international and no charge). The Other category was dropped from analyses because of its heterogeneity to more directly compare Medicaid and Commercial payer groups. Reason for hospitalization was categorized as Chemotherapy, Procedure, Infections or Non-infectious toxicity.¹⁹ Chemotherapy admissions were defined as admissions where chemotherapy was delivered within the first 2 days of the admission. Admissions where patients had a cancer related procedure and did not get chemotherapy in the first 2 days of hospitalization were considered procedure admissions. Length of stay (LOS) and total admission charges are reported in KID. Total admission charges were converted to costs using cost-to-charge ratios and are presented in 2012 US dollars.²⁰ Total aggregate national cost was the product of the number of weighted admissions and the weighted mean cost per admission, and aggregate national bed days was the product of weighted admissions and the weighted mean length of stay. This de-identified dataset was considered exempt from human subjects review by the Institutional Review Board of Baylor College of Medicine.

Statistical analysis

Means and standard error of the means were calculated for continuous variables, and frequency counts and percentages were used to summarize categorical variables. Categorical variables were compared using χ² tests. Mean LOS and costs were compared using an adjusted Wald test. Linear multiple regression models for both LOS and cost were developed to include patient level factors (race, sex, age, diagnosis and reason for admission) and whether the admission occurred in a freestanding children’s hospital. Hospital location and teaching status were not included in the models as 95% of all CC admissions took place at urban teaching hospitals, with no difference between Medicaid and commercially covered admissions. For these models, total cost and LOS were log-transformed to adjust for the highly skewed nature of these measures. Results were considered statistically significant if p values were <0.05. All analyses were performed on weighted data using survey procedures to generate national estimates. All analyses were performed using STATA Version 13 (College Station, Texas).

Results

Demographics

Of the 104,597 weighted CC admissions in 2012, third party payer information was available for 104,395. Demographic information is presented in Table 1. Half of CC admissions were paid by commercial insurance, and 40% were paid by Medicaid. Nine percent of admissions were paid for by other means including self-pay, international, or no charge. These were dropped from the analysis to directly compare Medicaid with Commercially insured admissions. When comparing Medicaid to Commercially insured admissions, patients covered by Medicaid were significantly younger at an average of 8.8 years (95% CI 8.8–8.9) than those with commercial insurance, who were an average of 10 years old (95% CI 9.9–10.1). Race also differed amongst payers, with Blacks and Hispanics overrepresented in the Medicaid population (p<0.001). Medicaid admissions were more frequently associated with leukemia diagnoses (p<0.001) and more frequently involved infections or toxicity than commercially covered admissions (p<0.001).

TABLE 1.

Demographics of Childhood Cancer Admissions

	Total (%)	Medicaid 40.52% (%)	Commercial 50.63% (%)	P
Race				<0.001
White	50.8	39.2	72.1
Black	9.4	15.6	6.1
Hispanic	21.4	36.6	11.8
Asian/Pacific Islander	4.8	2.6	4.7
Native American	0.6	1	0.4
Other	5.3	5	5
Sex				0.3
Male	55.4	55.1	55.5
Female	44.6	44.9	44.5
Age (years)				<0.001
<1	3.2	3.9	2.8
1–4	25.4	29.5	22.3
5–9	22.1	21.8	22.1
10–14	21.8	21.0	22.8
15–19	27.6	23.8	30.1
Diagnosis				<0.001
Leukemia	36.5	39.5	34.5
Lymphoma	10.1	8.7	11.3
Brain Tumor	12.9	12.6	13.1
Solid Tumor	39.7	38.5	40.3
Dual	0.8	0.8	0.9
Reason for Admission				<0.001
Chemotherapy	45.5	44.4	46.3
Procedure	14.3	13.3	14.9
Infection	12.2	13.6	11.2
Toxicity	20.7	21.2	20.3
Other	7.4	7.5	7.3
Children’s Hospital				<0.001
No	55.5	55.4	56.7
Yes	45.5	44.6	43.3

Length of Stay

The mean LOS for all admissions was 6.9 days. Medicaid admissions were a mean of 1.1 days longer than commercial admissions (p<0.001). Medicaid admissions were significantly longer regardless of sex, age (except infants <1 year), diagnosis (except those with dual cancer diagnoses, Figure 1A), reason for hospitalization (Figure 1B) and whether the hospitalization was at a children’s hospital. Additionally, LOS was significantly longer for Medicaid patients of the three largest racial groups, White, Black and Hispanic (Supplemental Table 1).

Figure 1.

Graphic representation of mean LOS by diagnosis (A) and reason for admission (B), and mean total cost of admission by diagnosis (C) and reason for admission (D). * indicates statistical significance with p<0.05.

The largest difference in LOS between Medicaid and commercially covered admissions was in Procedure admissions. Although a higher proportion of Medicaid Procedure admissions were for minor procedures including central line placement (17.8% Medicaid vs 16.3% commercial p=0.05), bone marrow biopsy (17.7% Medicaid vs 15.3% commercial p=0.002) and lumbar puncture (13% Medicaid vs 11.6% commercial p=0.04), the Medicaid-covered admissions were 2.8 days longer than procedure admissions covered by commercial insurance (11.2 days vs 8.4 days, p<0.001).

Given the significant discrepancy in the length of Procedure-related admissions based on Medicaid status, these admissions were analyzed independently to identify drivers of longer length of stay. Procedure hospitalizations were analyzed for rates of post-procedure infections, non-infectious toxicities, and administration of post-procedure chemotherapy (>2 days after admission). Notably, Medicaid admissions had a significantly higher rate of post-procedure infection (47.1%) compared with commercial insurance (37.9%) p<0.001 (Supplemental Table 2). Medicaid admissions with post-procedure infections were on average 3.5 days (21%) longer than those with commercial insurance. Medicaid admissions also had a higher rate of non-infectious toxicities and were more likely to have chemotherapy administered post procedure within the same hospitalization (27.5% vs 20%, p<0.001). Post-procedure infections, toxicities and chemotherapy administration were all associated with significantly longer hospitalizations by 3.5–4.8 days for those with Medicaid compared with commercial insurance; however, these longer hospitalizations did not translate to significantly higher costs of admission.

Cost per admission

The mean total costs per admission were significantly higher when the admission was paid for by Medicaid compared with commercial insurance (Supplemental Table 3). The mean cost of Medicaid admissions was $23,464 compared to $21,850 for commercially insured hospitalizations. There was no difference in cost of admission by race between those who were insured commercially or by Medicaid, except for “Other” race. When comparing cost within the Medicaid population, admissions associated with Black and Hispanic race were 9.1% (p<0.001) and 22% (p<0.001) costlier than Whites, respectively. Within the commercially insured population, Hispanics had hospitalizations costing nearly 10% more than Whites.

Costs of hospitalization were highest for children aged 0–12 months with no significant difference between the commercially insured and Medicaid admissions. Costs of hospitalization were significantly higher for patients with Medicaid than commercial insurance in children 1–14 years old. When analyzed by diagnosis, costs were highest for admissions associated with dual cancer diagnoses. Amongst those with single cancer diagnoses, leukemia was the most expensive diagnosis, followed by CNS tumors. The only category that showed a significant difference in cost related to admission payer was solid tumors (Figure 1C). Admissions for chemotherapy, procedures, and non-infectious toxicities were all significantly more expensive if paid by Medicaid (Figure 1D).

Multi Regression Models

In the multi regression models, when controlling for race, sex, age, diagnosis, reason for admission and hospital type, payer still had a significant effect on the mean LOS, with commercially insured admissions being 6% shorter than Medicaid admissions. When controlled for the same factors, race also influenced LOS, with Black and Hispanic having 8% and 4% longer admissions than White race, respectively. (Table 2)

TABLE 2.

Multiple Regression Model Length of Stay

		Coefficient	95% CI	P
Payer	Medicaid	Ref
	Commercial	−.06	−0.08– −0.04	<0.001
Sex	Male	Ref
	Female	0.02	0.001–0.03	0.03
Race/Ethnicity	White	Ref
	Black	0.08	0.05–0.11	<0.001
	Hispanic	0.04	0.02–0.06	<0.001
	Asian/Pacific Islander	0.08	0.04–0.11	<0.001
	Native American	0.08	−0.001–0.17	0.06
	Other	0.07	0.03–0.11	<0.001
Age (years)	<1	Ref
	1–4	−0.18	−0.23– −0.13	<0.001
	5–9	−0.28	−0.34– −0.23	<0.001
	10–14	−0.21	−0.27– −0.17	<0.001
	15–19	−0.17	−0.22– −0.11	<0.001
Diagnosis	Leukemia	Ref
	Lymphoma	−0.19	−0.22– −0.16	<0.001
	Brain Tumor	−0.28	−0.31– −0.26	<0.001
	Solid Tumor	−0.34	−0.36– −0.32	<0.001
	Dual	0.12	0.02– 0.22	0.02
Reason for Admission	Chemotherapy	Ref
	Surgery	0.41	0.39–0.44	<0.001
	Infection	0.07	0.05–0.10	<0.001
	Toxicity	0.08	0.06–0.10	<0.001
Children’s Hospital	No	Ref
	Yes	0.04	0.02–0.05	<0.001

The multiple regression model demonstrated no significant difference in cost of admission related to payer when controlling for race, sex, age, diagnosis, reason for admission and admission to a children’s hospital. However, race did have a significant impact on cost of admission, with admissions associated with Black and Hispanic race being 7% and 10% more costly than those associated with White race, respectively.(Table 3)

TABLE 3.

Multiple Regression Model Total Cost of Admission

		Coefficient	95% CI	P
Payer	Medicaid	Ref
	Private	0.01	−0.01–0.03	0.3
Sex	Male	Ref
	Female	0.01	−0.002–0.03	0.09
Race	White	Ref
	Black	0.07	0.04–0.10	<0.001
	Hispanic	0.10	0.08–0.12	<0.001
	Asian/Pacific Islander	0.16	0.11–0.21	<0.001
	Native American	0.10	−0.001–0.19	0.05
	Other	0.15	0.11–0.19	<0.001
Age	<1	Ref
	1–4	−0.17	−0.22– −0.12	<0.001
	5–9	−0.22	−0.28– −0.16	<0.001
	10–14	−0.06	−0.11–0.002	0.06
	15–19	0.08	0.02–0.13	0.01
Diagnosis	Leukemia	Ref
	Lymphoma	−0.14	−0.17– −0.11	<0.001
	Brain Tumor	−0.12	−0.14– −0.08	<0.001
	Solid Tumor	−0.26	−0.28– −0.24	<0.001
	Dual	0.31	0.19–0.43	<0.001
Reason for Admission	Chemotherapy	Ref
	Surgery	0.92	0.90–0.95	<0.001
	Infection	−0.04	−0.07– −0.01	0.01
	Toxicity	−0.01	−0.03– 0.01	0.35
Children’s Hospital	No	Ref
	Yes	0.54	0.52–0.56	<0.001

The cost of admission was most disparate by diagnosis group (Supplemental Table 4). Differences in both LOS and cost were most striking in admissions associated with leukemia. Regardless of payer, leukemia admissions associated with black race were approximately 2 days longer than leukemia admissions associated with white race, and total costs of admission were close to $10,000 more. Hispanic leukemia admissions were also more expensive than White admissions although the LOS was similar. Admissions associated with CNS tumors were also 1 to 2 days longer for Hispanics and Blacks than Whites if they were paid by Medicaid.

Our analysis of the KID estimates that the total cost of CC admissions in 2012 was $2.4 billion. Medicaid accounted for $990 million, 41% of total costs and 40% of all admissions. Commercial insurance accounted for $1.2 billion. Overall, children with a cancer diagnosis occupied 711,000 bed days in 2012. Medicaid admissions, 40% of the admissions, accounted for 44% of these beds days while commercial admissions were 46% of total bed days.

Discussion

Childhood cancer hospitalizations are significantly longer when paid for by Medicaid compared with commercial insurance. On average, Medicaid admissions were 1.1 days longer than admissions paid for by commercial insurance, and subsets of admissions were up to 3 days longer for Medicaid patients. Our study is the first to demonstrate payer as an independent predictor of length of stay in the CC population. These findings are consistent with published literature demonstrating hospitalizations are longer and more expensive for patients with public payers compared to commercial insurance in a variety of pediatric and adult illnesses.^{10,11,13,15,16,21,22} Reasons for these differences in care are likely multifactorial, including access to similar resources, medical provider biases and social determinants of health.

Our data demonstrate that admissions associated with CC paid by Medicaid are different from those paid by commercially insurance (Table 1). Medicaid admissions were more commonly younger, racial minorities, associated with leukemia, and associated with infections and toxicity compared to admissions with commercial insurance. Costs and LOS were different in almost all descriptive categories. The greatest difference was for procedure admissions, with Medicaid patients remaining hospitalized nearly 3 days longer than those with commercial insurance. One reason for this discrepancy was related to post-procedure infections, which were significantly higher in Medicaid admissions. This is consistent with published literature documenting primary payer status was an independent predictor of postoperative mortality, morbidity and resource utilization for common pediatric surgical procedures.¹⁶ Further, post-surgical infection rates ²³ and length of stay ^24,25 are linked to obesity, a patient factor more common in children with low socioeconomic status.²⁶ While obesity can be coded for in KID, it is often under-recognized and may be under-documented in billing codes. Therefore, identifying and associating potential factors for post-procedural complications will need to be investigated in different data sets.

Medicaid patients also had a higher rate of receiving chemotherapy during the same hospitalization as the procedure for which they were admitted. While this increases the cost of that individual hospitalization, it is not clear what impact this practice has on overall health care spending. It is possible that individual patients with Medicaid have fewer hospitalizations overall because the next scheduled chemotherapy is coupled with a procedure and given prior to discharge. It is also possible that the next scheduled chemotherapy could have been given in an outpatient setting, which may be less costly. It is interesting that despite having 3 days longer length of stay for procedure admissions, the cost of these admissions for Medicaid patients was only $5,000 more on average. This suggests that, while some patients remain in the hospital after the procedure to receive expensive care such as treatment of infections and administration of chemotherapy, many remain in the hospital without such care, suggesting that social factors may be playing a role in the prolongation of admissions.

Perhaps most importantly, Medicaid status is an important marker of poverty because it provides insurance coverage for children with family incomes ≤133% of the federal poverty level. Poverty has been shown to have several consequences in the care of children with cancer. Adolescents with Hodgkin lymphoma, low SES and public or no insurance are at higher risk for presenting with advanced disease.²⁷ Hispanic children and those from low SES are more likely to have extraocular (advanced) retinoblastoma.²⁸ Poverty has also been shown to be independently related to poor outcomes in CC, with children afflicted by poverty experiencing higher rates of early relapse and lower overall survival.^29,30 Our study shows that these factors may be reflected in the cost and length of admissions for these patients.

Aside from poverty itself, factors associated with inferior outcomes for publicly insured adults include language barriers, level of education, low health literacy, inadequate nutrition and poor health maintenance.^31–33 While fewer studies have directly compared these factors in publicly insured children, it is reasonable to assume that similar factors apply to children with Medicaid. Several studies have documented increased LOS for children whose parents have limited English proficiency.^34,35 These children are known to have lower rates of private insurance and higher rates of both Medicaid and uninsured status.³⁶ Further, the parents of children eligible for Medicaid are at high risk of having limited health literacy skills.³⁷ A study of general medicine patients showed that low health literacy was associated with an 11% longer LOS in adults.³¹ Since children transitioning from the hospital to their homes are reliant on the health literacy of adult caregivers, it is reasonable to assume that parental limited health literacy may prolong pediatric hospitalizations.

Childhood cancer patients often have significant requirements for care at home. One study showed that a third of families of children with cancer have one parent quit or change jobs as a result of a cancer diagnosis.³⁸ For families living in poverty, loss of a job may be financially impossible. Furthermore, parental low health literacy may make caring for these children at home difficult for some families. In the absence of a robust social or medical in-home support system, patients may remain in the hospital for care that could otherwise be delivered in the outpatient setting, leading to longer hospitalizations. Home health care has been shown to be a cost-effective alternative to inpatient hospital care,^39,40 providing a wide range of services often required in patients with cancer including intravenous administration of antimicrobial agents and parenteral nutrition, enteral tube feeding and wound care. Access to these services has been shown to be inadequate throughout most of the U.S. for numerous reasons including inadequate payment and restrictive insurance and managed care policies.⁴¹ Although Medicaid provides a comprehensive home health benefit for children for services deemed medically necessary by the state, recipients still face barriers to adequate home health services. Due to inadequate reimbursement rates, many home health agencies do not accept Medicaid referrals.⁴² For the 45% of children in rural areas enrolled in Medicaid, there may be limited options for home health services leading to prolongation of hospitalization.^43,44

Implicit provider bias is increasingly recognized as impacting care. Several studies have implicated race, ethnicity and SES as sources of provider bias in patient care.⁴⁵ In this situation provider perception of parents’ ability to provide adequate care or return for emergency or planned follow up care may be a factor in prolonged hospitalizations for Medicaid patients.

Although LOS was significantly longer for Medicaid admissions compared with admissions paid for by commercial insurance, the cost of admissions was not significantly different in multiple regression analysis. While payer was not a significant factor in cost of admissions, race was. Several studies have demonstrated disparities in outcomes of CC related to race and ethnicity, with non-white children having inferior overall survival for a variety of cancers including acute leukemia, lymphoma, retinoblastoma and neuroblastoma.^29,46–49 These disparities in outcome likely play a role in racial differences in length and cost of hospitalization as children with relapsed or refractory disease often have longer hospitalizations and sometimes more costly treatments.

Our study has several limitations. KID is a billing database and diagnoses and procedure codes are reported in the form of ICD-9 codes which may be misclassified. KID data is reported at the admission level, thus recurring admissions from individual patients are not captured, potentially biasing the data toward the characteristics of patients with more frequent admissions. These data also do not allow us to look at the distinct components of an admission such as pharmaceuticals to determine the drivers of cost for particular admissions. However, the number of admissions and geographic sampling outweigh the limitations of the design of KID with respect to this analysis. In addition, this data set represents admissions taking place in 2012, seven years prior to this publication. This analysis was completed prior to the recent release of the 2016 KID in December of 2018, which includes ICD-10 codes rather than ICD-9. Changes in cost, quality and safety initiatives, public policy, and a greater awareness of health care disparities may lead to different results if this study were repeated with a more recent dataset. Children with cancer qualify for the Medicaid Medically Needy program in some states, which allows children from families with higher incomes than would normally qualify for Medicaid to become eligible after incurring significant medical expenses (https://www.medicaid.gov/medicaid/eligibility/index.html). While the composition of this specific Medicaid population limits the generalizability of these data to the entire pediatric Medicaid population, the results are relevant to other pediatric conditions requiring repeated hospitalizations and expensive care.

Overall, this analysis of admissions for CC identifies different characteristics and healthcare utilization for children with Medicaid compared to those with commercial insurance. It remains unclear from this dataset if these differences affect overall patient outcomes, but the significant differences in LOS likely impose differential effects on the families, medical delivery systems and Medicaid payer systems. This work highlights the need for further research and policy improvements for patients with CC living in poverty in order to understand and address drivers of health disparities.

Abbreviations

CC

Childhood cancer

KID

Kids Inpatient Database

ICD 9 CM

International Classification of Diseases ninth revision, clinical modification diagnosis

CNS

Central nervous system

LOS

Length of stay

SES

Socioeconomic status

U.S.

United States