Health insurance continuity and mortality in children, adolescents, and young adults with blood cancer

Abstract

Background

Many uninsured patients do not receive Medicaid coverage until a cancer diagnosis, potentially delaying access to care for early cancer detection and treatment. We examined the association of Medicaid enrollment timing and patterns with survival among children, adolescents, and young adults with diagnosed blood cancers, where disease onset can be acute and early detection is critical.

Methods

We identified 28 750 children, adolescents, and young adults (birth to 39 years of age) with newly diagnosed blood cancers from the 2006-2013 Surveillance, Epidemiology, and End Results program–Medicaid data. Enrollment patterns included continuous Medicaid enrollment (preceding through diagnosis), newly gained Medicaid coverage (at or shortly after diagnosis), other noncontinuous Medicaid enrollment, and private/other insurance. We assessed cumulative incidence of death from diagnosis, censoring at last follow-up, 5 years after diagnosis, or December 2018, whichever occurred first. Multivariable survival models estimated the association of insurance enrollment patterns with risk of death.

Results

One-fourth (26.1%) of the cohort was insured by Medicaid; of these patients, 41.1% had continuous Medicaid enrollment, 34.9% had newly gained Medicaid, and 24.0% had other or noncontinuous enrollment. The cumulative incidence of all-cause death 5 year after diagnosis was highest in patients with newly gained Medicaid (30.2%, 95% confidence interval [CI] = 28.4% to 31.9%), followed by other noncontinuous enrollment (23.2%, 95% CI = 21.3% to 25.2%), continuous Medicaid enrollment (20.5%, 95% CI = 19.1% to 21.9%), and private/other insurance (11.2%, 95% CI = 10.7% to 11.7%). In multivariable models, newly gained Medicaid was associated with a higher risk of all-cause death (hazard ratio = 1.39, 95% CI = 1.27 to 1.53) and cancer-specific death (hazard ratio = 1.50, 95% CI = 1.35 to 1.68) compared with continuous Medicaid.

Conclusions

Continuous Medicaid coverage is associated with survival benefits among pediatric, adolescent, and young adult patients with diagnosed blood cancers; however, fewer than half of Medicaid-insured patients have continuous coverage before diagnosis.

In the United States, leukemia is the most common pediatric cancer, and lymphoma is the third-leading cancer in adolescents and young adults (1,2). Although the overall survival rates of pediatric, adolescent, and young adult cancers have increased (3,4), disparities in survival have been well documented for uninsured children, adolescents, and young adults with blood cancers (5,6). Yet, the timing and continuity of insurance enrollment and their association with survival are poorly understood for this young, vulnerable population.

Medicaid is the single largest insurer of children, covering more than 35.1 million children as of March 2023 and 1 in 3 children with newly diagnosed cancer in 2014 (7,8). Medicaid is also the primary insurance program for low-income adults and adults with disabilities (9,10). Importantly, many eligible individuals do not gain Medicaid coverage until the point of or shortly after a cancer diagnosis (ie, newly gained Medicaid) (11,12). Individuals who have gained Medicaid can experience gaps in enrollment and become uninsured, with or without subsequent reenrollment in Medicaid (ie, noncontinuous enrollment) (13). Such enrollment gaps, known as coverage churn, can span half a year or longer and occur repeatedly (14). Approximately 1 in 4 Medicaid beneficiaries experience noncontinuous enrollment in a given year, particularly nondisabled young adults and postpartum women (15), but this issue is understudied in pediatric, adolescent, and young adult oncology populations. Compared with children, adolescents, and young adults maintaining continuous Medicaid coverage, individuals who become enrolled in Medicaid only following a cancer diagnosis or otherwise experience noncontinuous Medicaid may lack access to health care that facilitates early cancer detection, present with more severe or acute disease, delay treatment until later stages, and so face a higher risk of the treatment-related complications and subacute morbidity that underly mortality (12,16,17).

Historically, evaluation of Medicaid coverage has focused largely on whether patients did or did not have Medicaid at the time of diagnosis. To date, there is a dearth of research examining the timing and continuity of Medicaid enrollment in children, adolescents, and young adults with blood cancers, for which there are no screening guidelines and disease onset can be acute, even fatal. The adolescent and young adult population has also historically exhibited the highest prevalence of uninsurance or underinsurance (18,19). Of the few studies examining the association of Medicaid enrollment continuity with cancer outcomes, most have focused on adult-onset solid tumors in single states (20). We contribute to this literature by examining the associations between insurance enrollment patterns and survival in pediatric, adolescent, and young adult patients with newly diagnosed blood cancers in a large, population-based cohort from 12 states.

Methods

Data

We used data from the Surveillance, Epidemiology, and End Results (SEER) program, linked to the Medicaid administrative enrollment files (SEER-Medicaid data) (21,22). The methods of SEER-Medicaid linkage have been described in detail elsewhere (23-25). In brief, the National Cancer Institute (NCI) and the Centers for Medicare & Medicaid Services linked data from 14 SEER cancer registries to the Personal Summary File of the Medicaid Analytic eXtract data (26). SEER contains all incident cancer cases captured during 2006-2013 in the 14 population-based registries across 12 geographically diverse states for which linked Medicaid data were available across the study period. These registries provide information about some sociodemographic factors, cancer diagnosis, insurance status at diagnosis, and vital status (month, year, and cause of death). The Medicaid files comprise records for all SEER patients enrolled in Medicaid for at least 1 day in a given year. The linkage was performed using a deterministic matching process based on patient-level identifiers such as Social Security number, date of birth, and sex (23,24).

The Medicaid files were used to trace monthly Medicaid enrollment status during the 12 months preceding, the month of, and 2 months following cancer diagnosis (ie, 15-month assessment window) (Supplementary Figure 1, available online) (23). We used this assessment window because continuity of coverage in the 12 months before diagnosis is important for timely access to care, early cancer detection and diagnosis, and timely treatment initiation (12). Further, patients who gain Medicaid coverage only at diagnosis are likely applying for and receiving Medicaid within the first 2 months following diagnosis (27,28).

The Emory University institutional review board approved this study.

Study sample

We identified all patients from birth to 39 years of age who had received a first primary diagnosis of leukemia or lymphoma between 2006 and 2013. We excluded patients with missing data on diagnosis or death. We further narrowed to patients whose diagnoses occurred between January 1, 2007, and October 31, 2013, to ensure complete observation during the 15-month assessment window. Our analytic sample consisted of 28 750 patients, including 7510 (26.1%) who were linked to the Medicaid files and 21 240 (73.9%) who were not (Supplementary Figure 2, available online).

Insurance enrollment pattern categorization

Patients linked to the Medicaid files were categorized into the following mutually exclusive groups: 1) continuous Medicaid enrollment preceding and through diagnosis (hereafter, “continuous Medicaid”—ie, patients enrolled in Medicaid for 12-13 months during the year preceding and through the month of diagnosis), 2) newly gained Medicaid at or shortly after diagnosis (hereafter, “newly gained Medicaid”—ie, patients who were not enrolled in Medicaid until the month before, the month of, or within 2 months following diagnosis), and 3) other noncontinuous Medicaid enrollment (patients enrolled in Medicaid at some point during the 15-month assessment window but not classified into the prior categories) (12,24).

Patients not linked to the Medicaid files were categorized into the following mutually exclusive groups using information from SEER registries on insurance type recorded at diagnosis: private/other insurance (military, US Department of Veterans Affairs, TRICARE, Medicare, or Indian Health Service or Public Health Service), unknown insurance status, or no insurance. In a supplementary analysis, we explored registry-reported insurance types at diagnosis among the aforementioned Medicaid groups.

Outcomes

Vital status was extracted from SEER registries on patients who had died on or before December 31, 2018, and used to identify all-cause and cancer-specific death within 5 years after diagnosis. Cancer-specific death refers to deaths attributable to the originally diagnosed cancer (29).

Covariates

Our analysis included sociodemographic (race and ethnicity, sex, age at first cancer diagnosis, rural-urban residence, neighborhood socioeconomic status [SES]) and cancer-related characteristics (year of diagnosis, primary hematologic cancer category). Race and ethnicity, as social constructs, are linked to insurance enrollment and survival outcomes; thus, they were included as confounding factors in multivariable models. Neighborhood SES was assessed by the Yost index, a multidimensional SES measure capturing the educational level, income, housing status, and employment rate in the US Census tract of patient residence (30).

Statistical analysis

Sample characteristics were described overall and by enrollment patterns. We applied the Kaplan-Meier method to model the cumulative incidence of death from diagnosis, with censoring defined as no death event by the last recorded follow-up, 5 years after diagnosis, or December 31, 2018—whichever occurred first. We compared the cumulative incidence curves by enrollment patterns using log-rank tests.

We used multivariable regressions to assess the association of insurance enrollment patterns and risk of death, adjusting for sociodemographic factors and year of diagnosis, with standard errors clustered at the state level. The Cox proportional hazards models assessed risk of all-cause death. The Fine-Gray subdistribution hazards model assessed risk of cancer-specific death, accounting for all other causes of death as a competing risk. Models were estimated for the full cohort, and then restricted to the Medicaid-insured groups only. To examine associations of enrollment patterns with survival by demographic and disease characteristics, subgroup analyses were conducted by age group, race and ethnicity, and cancer type. Adjusted hazard ratios (HRs) of death and associated 95% confidence intervals (CIs) were reported. Analyses were conducted using SAS, version 9.4, statistical software (SAS Institute, Inc, Cary, NC). Statistical significance was assessed at the .05 level using 2-sided tests.

Results

Within our cohort (N = 28 750), more than half were male (56.4%), non-Hispanic White patients (51.5%) and living in metropolitan areas (91.3%) (Table 1). The most prevalent blood cancer diagnoses were non-Hodgkin lymphoma (30.2%), Hodgkin lymphoma (27.7%), and lymphocytic leukemia (24.9%).

Table 1.

Sample characteristics of children, adolescents, and young adults with newly diagnosed blood cancer^a

Characteristic	Total, No. N = 28 750	Medicaid-insured groups			Private/other insurance (registry reported), % n = 16 204	Uninsured/unknown insurance (registry reported), % n = 5036
		Continuous Medicaid, % n = 3089	Newly gained Medicaid, %b n = 2618	Other noncontinuous Medicaid enrollment, % n = 1803
Sex
Male	16 229	9.3	10.5	5.3	55.9	19.0
Female	12 521	12.6	7.3	7.5	56.9	15.7
Race or ethnicity
Hispanic	7709	14.6	9.7	7.8	36.1	31.8
Non-Hispanic Black	3448	19.1	13.0	9.5	40.8	17.6
Non-Hispanic otherc or unknown race or ethnicity	2791	6.6	6.9	4.3	65.3	16.9
Non-Hispanic White	14 802	7.6	8.3	5.1	68.9	10.2
Age at first primary diagnosis, y
Birth to 5	4009	16.4	5.5	6.8	46.5	24.8
6-14	3925	16.6	6.5	5.9	50.2	20.8
15-25	7870	10.6	11.0	7.6	55.8	15.0
26-39	12 946	7.3	9.9	5.4	61.6	15.8
Rurality of residence
Metropolitan areas	26 255	10.1	8.6	5.9	57.5	17.9
Nonmetropolitan (including rural) areas	2495	17.3	14.4	10.3	44.1	14.0
Neighborhood SES quintile
1 (highest SES)	7092	3.5	5.1	2.4	79.4	9.6
2	6019	7.3	7.4	5.0	65.0	15.3
3	5049	10.6	9.8	7.0	53.5	19.2
4	4899	14.0	12.0	8.6	43.1	22.2
5 (lowest SES)	5137	22.1	13.5	10.5	28.9	25.1
Unknown	554	7.8	5.8	4.5	65.3	16.6
Blood cancer type
Hodgkin lymphoma	7975	8.6	6.5	6.3	64.8	13.9
Non-Hodgkin lymphoma	8671	8.6	8.8	5.9	59.7	17.0
Lymphocytic leukemia	7160	15.0	9.4	6.6	47.0	22.0
Myeloid and monocytic leukemia	4617	11.6	13.7	6.4	50.6	17.6
Other leukemia	327	15.3	11.6	7.0	46.5	19.6
Year of cancer diagnosis
2007	4194	10.3	8.6	6.4	58.9	15.9
2008	4270	10.0	9.6	6.2	58.7	15.5
2009	4263	11.0	9.9	5.9	56.2	17.1
2010	4169	11.6	9.9	6.3	54.0	18.2
2011	4199	12.1	9.2	5.8	55.2	17.7
2012	4199	10.7	8.8	6.0	55.4	19.1
2013	3456	9.4	7.4	7.6	56.0	19.6
SEER state
California	12 518	10.6	8.3	5.5	55.7	19.9
Connecticut	1221	12.9	6.7	5.9	62.5	12.0
Georgia	3226	7.7	10.6	6.5	51.9	23.4
Hawaii	422	11.1	8.5	4.7	56.2	19.4
Iowa	965	11.3	10.9	8.0	63.3	6.5
Kentucky	1339	12.6	12.7	9.0	52.1	13.6
Louisiana	1427	19.5	17.5	9.2	40.4	13.5
Michigan	1335	11.2	9.5	6.7	59.1	13.5
New Jersey	3111	7.8	5.6	4.3	65.9	16.5
New Mexico	605	23.3	8.8	13.7	39.2	15.0
Utah	1050	7.4	10.2	5.8	61.9	14.7
Washington	1531	9.4	8.8	7.6	61.7	12.6

^aAuthors’ analysis of the 2006-2013 SEER-Medicaid data. SEER = Surveillance, Epidemiology, and End Results; SES = socioeconomic status.

^bRow percentages were reported.

^cThe non-Hispanic other group includes Asian, Native American or Alaska Native, and Native Hawaiian or other Pacific Islander patients.

More than half the cohort had private/other insurance at diagnosis (56.4%), and 17.5% were classified as unknown insurance or uninsured at diagnosis (Supplementary Figures 2 and 3, available online). More than one-fourth (26.1%) had any (≥1 month) Medicaid enrollment during the assessment window; of these, 41.1% had continuous Medicaid, 34.9% had newly gained Medicaid, and 24.0% experienced other noncontinuous Medicaid enrollment. Notably, among the Medicaid-insured patients, 34.0% were recorded as having private/other insurance at diagnosis in SEER registries (Supplementary Table 1, available online).

Comparing insurance enrollment patterns by sample characteristics, the proportions with continuous Medicaid and newly gained Medicaid were higher in non-Hispanic Black and Hispanic (vs non-Hispanic White) patients, residents of rural and nonmetropolitan (vs metropolitan) areas, and neighborhoods with lower SES (Table 1). Among adolescent and young adult patients aged 15 to 39 years, the proportion with continuous Medicaid (7.3%-10.6% vs 16.4%-16.6%) was lower, whereas the proportion with newly gained Medicaid (9.9%-11.0% vs 5.5%-6.5%) was higher, compared with pediatric patients 14 years of age or younger.

Overall associations between enrollment patterns and risk of death

The cumulative incidence of all-cause death 5 years after diagnosis was highest in patients having newly gained Medicaid (30.2%, 95% CI = 28.4% to 31.9%), followed by other noncontinuous Medicaid enrollment (23.2%, 95% CI = 21.3% to 25.2%), and then continuous Medicaid (20.5%, 95% CI = 19.1% to 21.9%) (Figure 1, A). The cumulative incidence of all-cause death remained lowest for patients with private/other insurance, at 11.2% (95% CI = 10.7% to 11.7%) 5 years after diagnosis. Patterns were similar for cancer-specific deaths (Figure 1, B). Notably, of all deaths (n = 4408) that occurred 5 years after diagnosis, the vast majority were cancer-specific deaths (n = 3766).

Figure 1.

Cumulative incidence curves for death, by insurance continuity patterns. A) Cumulative incidence of all-cause death, by insurance continuity pattern. B) Cumulative incidence of cancer-specific death, by insurance continuity pattern. Authors’ analysis of the 2006-2013 Surveillance-Epidemiology, and End Results program–Medicaid data.

Multivariable models assessing all-cause death showed that compared with patients with continuous Medicaid, the hazard ratio (HR) for all-cause death was 1.39 (95% CI = 1.27 to 1.53, P < .001) for patients with newly gained Medicaid and 1.11 (95% CI = 1.06 to 1.16, P < .001) for patients with other noncontinuous Medicaid enrollment, whereas the hazard ratio was 0.54 (95% CI = 0.48 to 0.62, P < .001) for patients with private/other insurance and 0.76 (95% CI = 0.64 to 0.91, P = .002) for patients classified as having unknown insurance or being uninsured (Table 2). These differences persisted in multivariable models examining cancer-specific death. The hazard ratio of cancer-specific death was 1.50 (95% CI = 1.35 to 1.68, P < .001) for patients with newly gained Medicaid, 1.17 (95% CI = 1.03 to 1.33, P = .02) for other noncontinuous Medicaid enrollment, 0.58 (95% CI = 0.52 to 0.64, P < .001) for private/other insurance, and 0.80 (95% CI = 0.72 to 0.90, P < .001) for unknown insurance or uninsured compared with continuous Medicaid enrollees.

Table 2.

Overall associations between insurance continuity patterns and risk of death among children, adolescents, and young adults with newly diagnosed blood cancer^a

	All-cause death			Cancer-specific death
Characteristic	HR	95% CI	P	HR	95% CI	P
Insurance coverage continuity
Continuous Medicaid	(Referent)			(Referent)
Newly gained Medicaid	1.39	1.27 to 1.53	<.001	1.50	1.35 to 1.68	<.001
Other noncontinuous Medicaid enrollment	1.11	1.06 to 1.16	<.001	1.17	1.03 to 1.33	.02
Private/other insurance (registry reported)	0.54	0.48 to 0.62	<.001	0.58	0.52 to 0.64	<.001
Uninsured/unknown insurance (registry reported)	0.76	0.64 to 0.91	.002	0.80	0.72 to 0.90	<.001
Sex
Male	(Referent)			(Referent)
Female	0.78	0.75 to 0.80	<.001	0.80	0.76 to 0.86	<.001
Race or ethnicity
Hispanic	1.41	1.30 to 1.53	<.001	1.47	1.35 to 1.59	<.001
Non-Hispanic Black	1.44	1.31 to 1.58	<.001	1.49	1.36 to 1.64	<.001
Non-Hispanic other or unknown	1.24	1.10 to 1.40	<.001	1.29	1.16 to 1.45	<.001
Non-Hispanic White	(Referent)			(Referent)
Age at first primary diagnosis, y
Birth to 5	(Referent)			(Referent)
6-14	1.15	1.04 to 1.26	.004	1.13	0.98 to 1.30	.08
15-25	1.61	1.45 to 1.79	<.001	1.62	1.44 to 1.82	<.001
26-39	2.09	1.87 to 2.33	<.001	2.03	1.81 to 2.27	<.001
Rurality of residence
Metropolitan areas	(Referent)			(Referent)
Nonmetropolitan (including rural) areasb	0.94	0.84 to 1.06	.30	0.91	0.81 to 1.02	.10
Neighborhood SES quintile
1 (highest SES)	(Referent)			(Referent)
2	1.17	1.11 to 1.23	<.001	1.21	1.09 to 1.34	<.001
3	1.20	1.07 to 1.35	.002	1.23	1.11 to 1.37	<.001
4	1.28	1.18 to 1.39	<.001	1.25	1.12 to 1.40	<.001
5 (lowest SES)	1.43	1.32 to 1.54	<.001	1.44	1.29 to 1.60	<.001
Unknown SES	1.17	0.88 to 1.56	.27	1.18	0.93 to 1.50	.18
Year of cancer diagnosis
2007	(Referent)			(Referent)
2008	1.04	0.97 to 1.11	.27	1.02	0.91 to 1.13	.75
2009	1.00	0.92 to 1.10	.98	0.97	0.87 to 1.08	.52
2010	0.90	0.83 to 0.98	.01	0.89	0.79 to 0.99	.03
2011	0.93	0.85 to 1.02	.10	0.87	0.78 to 0.97	.01
2012	0.81	0.72 to 0.91	<.001	0.71	0.63 to 0.80	<.001
2013	0.88	0.78 to 0.99	.04	0.79	0.70 to 0.89	<.001

^aAuthors’ analysis of the 2006-2013 Surveillance, Epidemiology, and End Results program–Medicaid data (N = 28 750 patients). HR = hazard ratio; CI = confidence interval; SES = socioeconomic status.

^bIncluded 7 patients with unknown information about rurality.

Subgroup analyses by demographics and disease types among all patients

Statistically significant associations between enrollment patterns and risk of cancer-specific death were observed in the majority of demographic and disease subgroups. Notably, associations were larger in magnitude among young adults aged 26 to 39 years than in pediatric patients from birth to 14 years of age (newly gained [vs continuous] Medicaid: HR = 1.51, 95% CI = 1.28 to 1.78, P < .001 vs HR = 1.30, 95% CI = 0.99 to 1.72, P = .06) (Table 3). Moreover, associations were most notable among non-Hispanic Black patients (newly gained [vs continuous] Medicaid: HR = 1.75, 95% CI = 1.39 to 2.20, P < .001) across racial and ethnic groups and among patients with lymphoma (newly gained [vs continuous] Medicaid: HR = 1.52, 95% CI = 1.26 to 1.83, P < .001) across disease groups (Supplementary Table 2, available online). Similar patterns were observed in subgroup analyses estimating all-cause death (Table 3; Supplementary Table 2, available online).

Table 3.

Subgroup analyses between insurance continuity patterns and risk of death among children, adolescents, and young adults with newly diagnosed blood cancer, by age group^a

	All-cause death			Cancer-specific death
Characteristic	HR	95% CI	P	HR	95% CI	P
Among patients birth to 14 y of age (n = 7934)
Insurance coverage continuity
Continuous Medicaid	(Referent)			(Referent)
Newly gained Medicaid	1.28	0.99 to 1.66	.057	1.30	0.99 to 1.72	.06
Other noncontinuous Medicaid enrollment	1.28	1.09 to 1.51	.003	1.16	0.87 to 1.53	.31
Private/other insurance (registry reported)	0.79	0.65 to 0.96	.02	0.75	0.61 to 0.92	.006
Uninsured/unknown insurance (registry reported)	0.79	0.59 to 1.07	.12	0.78	0.63 to 0.96	.02
Among patients aged 15-25 y (n = 7870)
Insurance coverage continuity
Continuous Medicaid	(Referent)			(Referent)
Newly gained Medicaid	1.26	1.06 to 1.49	.001	1.31	1.07 to 1.60	.008
Other noncontinuous Medicaid enrollment	1.06	0.94 to 1.20	.34	1.10	0.88 to 1.38	.42
Private/other insurance (registry reported)	0.49	0.41 to 0.59	<.001	0.52	0.43 to 0.63	<.001
Uninsured/unknown insurance (registry reported)	0.79	0.65 to 0.96	.02	0.82	0.67 to 1.00	.05
Among patients aged 26-39 y (n = 12 946)
Insurance coverage continuity
Continuous Medicaid	(Referent)			(Referent)
Newly gained Medicaid	1.32	1.20 to 1.44	<.001	1.51	1.28 to 1.78	<.001
Other noncontinuous Medicaid enrollment	1.02	0.93 to 1.12	.64	1.17	0.96 to 1.42	.12
Private/other insurance (registry reported)	0.47	0.41 to 0.53	<.001	0.52	0.45 to 0.61	<.001
Uninsured/unknown insurance (registry reported)	0.68	0.57 to 0.80	<.001	0.75	0.63 to 0.89	.001

^aAuthors’ analysis of the 2006-2013 Surveillance, Epidemiology, and End Results program–Medicaid data. All regression models also controlled for sex, race and ethnicity, year of diagnosis, rurality of residence, and neighborhood socioeconomic status, with standard errors clustered at the state level. HR = hazard ratio; CI = confidence interval.

Analyses among Medicaid-insured patients

Findings were consistent when limiting the analysis to Medicaid-insured patients. Patients with newly gained Medicaid (HR of all-cause death = 1.30, 95% CI = 1.16 to 1.46, P < .001; HR of cancer-specific death = 1.41, 95% CI = 1.26 to 1.58, P < .001) and patients with other noncontinuous Medicaid enrollment (HR of all-cause death = 1.08, 95% CI = 1.03 to 1.13, P = .001; HR of cancer-specific death = 1.14, 95% CI = 1.00 to 1.30, P = .05) faced higher risks of death than did continuous Medicaid enrollees (Table 4).

Table 4.

Associations between insurance continuity patterns and risk of death among Medicaid-insured children, adolescents, and young adults with newly diagnosed blood cancer and by age group^a

	All-cause death			Cancer-specific death
Characteristic	HR	95% CI	P	HR	95% CI	P
Among all Medicaid-insured patients (n = 7510)
Insurance coverage continuity
Continuous Medicaid	(Referent)			(Referent)
Newly gained Medicaid	1.30	1.16 to 1.46	<.001	1.41	1.26 to 1.58	<.001
Other noncontinuous Medicaid enrollment	1.08	1.03 to 1.13	.001	1.14	1.00 to 1.30	.05
Among Medicaid-insured patients aged from birth to 14 y (n = 2290)
Insurance coverage continuity
Continuous Medicaid	(Referent)			(Referent)
Newly gained Medicaid	1.27	0.98 to 1.63	.069	1.29	0.97 to 1.71	.077
Other noncontinuous Medicaid enrollment	1.29	1.08 to 1.56	.006	1.18	0.89 to 1.56	.26
Among Medicaid-insured patients aged 15-25 y (n = 2297)
Insurance coverage continuity
Continuous Medicaid	(Referent)			(Referent)
Newly gained Medicaid	1.26	1.06 to 1.49	.007	1.31	1.07 to 1.60	.01
Other noncontinuous Medicaid enrollment	1.06	0.93 to 1.20	.37	1.09	0.87 to 1.37	.45
Among Medicaid-insured patients aged 26-39 y (n = 2923)
Insurance coverage continuity
Continuous Medicaid	(Referent)			(Referent)
Newly gained Medicaid	1.27	1.13 to 1.43	<.001	1.46	1.24 to 1.72	<.001
Other noncontinuous Medicaid enrollment	1.01	0.93 to 1.10	.84	1.16	0.95 to 1.41	.15

^aAuthors’ analysis of the 2006-2013 Surveillance, Epidemiology, and End Results program–Medicaid data. All regression models also controlled for sex, race and ethnicity, year of diagnosis, rurality of residence, and neighborhood socioeconomic status, with standard errors clustered at the state level. HR = adjusted hazard ratio; CI = confidence interval.

Discussion

In this multistate, population-based cohort study, we found significant associations between lacking continuous insurance and inferior survival among Medicaid-insured children, adolescents, and young adults with newly diagnosed leukemia or lymphoma. Compared with patients continuously enrolled in Medicaid, patients with newly gained Medicaid at or shortly after diagnosis faced a 26%-39% higher risk of all-cause death and a 29%-51% higher risk of cancer-specific death. This association persisted across demographic and disease subgroups. Patients with other noncontinuous Medicaid enrollment (vs continuous Medicaid) faced an 8%-29% higher risk of all-cause death and a 14%-17% higher risk of cancer-specific death. Despite such an elevated risk, more than half (58.9%) of Medicaid-insured children, adolescents, and young adults with blood cancers lacked continuous coverage preceding their blood cancer diagnosis.

These findings suggest that the timing and continuity of insurance enrollment play a pivotal role in driving survival outcomes in pediatric, adolescent, and young adult blood cancers, consistent with prior research (12,17). Notably, most previous studies assessed insurance status at only 1 time point (largely at the time of pediatric, adolescent, or young adult cancer diagnosis) (31-34) or were conducted in single states for adult-onset solid tumors and with relatively small sample sizes (17,24,28,35-37). In our sample, 2556 patients recorded in SEER registries as having private/other insurance at diagnosis were linked to Medicaid administrative data (Supplementary Table 1, available online), suggesting a potential for “Medicaid undercount” (discrepancies between registry-based estimates of Medicaid enrollment and those reported in state or federal administrative data), as noted in prior research based solely on registry-reported payor information (38). Our study advances this field by providing the first multistate, objective estimate of Medicaid enrollment patterns before, during, and shortly after blood cancer diagnosis. In addition, we captured the entire age spectrum of children, adolescents, and young adults, as defined by the NCI (39).

The observed survival benefits of continuous Medicaid enrollment likely reflect a period when continuous coverage of care is critical for early cancer detection, prompt diagnosis, and timely treatment initiation to promote optimal blood cancer outcomes. Unlike some adult-onset solid tumor cancers, currently there are no screening guidelines for blood cancers among children, adolescents, and young adults. Response largely depends on primary care professionals detecting early warning signs and early-onset symptoms during regular health checks, thus increasing the likelihood of early-stage cancer diagnosis (40). Patients with (vs without) continuous coverage are more likely to have stable access to primary care and early referral to specialists (eg, oncologists) and to receive timely testing for a cancer diagnosis (12,41). In contrast, a delay in access can result in an initial presentation with higher disease severity/acuity or more advanced stage and lead to treatment delays, potentially driving poorer prognosis.

For patients who newly gained Medicaid coverage at the point of or after diagnosis, the additional time needed to navigate the insurance system, including applying for and waiting to be processed and accepted by Medicaid for those who meet eligibility criteria, could prevent them from timely initiation of lifesaving therapies (37). Prior research has shown that delayed treatment initiation contributes to worse prognosis in adult cancers and increased risk of mortality for some pediatric cancers (42,43). Although our assessment focused in the periods before, during, and shortly after diagnosis, patients lacking continuous insurance in this period may also be more likely to experience insurance gaps or discontinuation over the course of treatment, after treatment, and during survivorship, which would further exacerbate the risk for inferior survival (17,44). Maintaining continuous insurance after treatment enables patients to receive regular follow-up care, including guideline-concordant surveillance and screening essential for early detection of cancer recurrence, initiation of salvage therapy, and management of chronic health conditions and subsequent malignancies associated with cancer treatment (45-47). More research is needed to understand the trajectory and continuity of insurance coverage after treatment through long-term survivorship in this high-risk, high-need population.

The association between lacking continuous Medicaid and poorer survival was strongest among young adults aged 26 to 39 years; unfortunately, this age group was the least likely to have continuous Medicaid coverage. This disparity may be attributed to the unique vulnerability of life stage in young adults to uninsurance or underinsurance. Previous adolescent and young adult–focused literature has shown that such vulnerability is often the result of transitions from pediatric care largely managed by parents to adult care managed by the patients themselves as they age out of parental insurance plans and face difficulties navigating the complex Medicaid insurance processes during young adulthood (48). This vulnerability could in turn affect adolescents and young adults’ access to care, treatment adherence, financial hardship, and ultimately prognosis (49).

Millions of individuals relying on Medicaid have lost coverage since March 2023, when the Medicaid Continuous Enrollment Protection (which required state Medicaid programs to keep their beneficiaries continuously enrolled during the COVID-19 pandemic) expired (50,51). Our findings from prepandemic data foreshadow the potential consequences of current coverage gaps and discontinuations, highlighting the urgent need for multilevel interventions tailored to meet the insurance needs of both the pediatric and the adolescent and young adult age groups. For example, state Medicaid programs could mitigate enrollment gaps and discontinuations by providing 12-month (or 24-month) continuous eligibility for children as well as adolescents and young adults (9). In addition, nonexpansion states could consider adopting the Patient Protection and Affordable Care Act Medicaid expansion and reducing system barriers to enrollment, which may mitigate coverage disruptions and particularly benefit young adults (52). States could also implement strategies for faster processing of new applications and simplified eligibility recertification procedures, thus reducing administrative delays that impede timely access to care for eligible children, adolescents, and young adults (53). Beyond state-level policy changes, establishing community-based or institution-based insurance navigation programs would help eligible young patients to effectively navigate the Medicaid system, maintain continuous coverage, and receive the primary and specialist care they need.

Limitations

Several study limitations are noted. First, we had no data to identify insurance continuity patterns for patients with private/other insurance or uninsured patients. The “uninsured” group, defined as patients with no insurance at the point of diagnosis in our study, may have had private/other insurance before or after diagnosis; thus, the observed association between being uninsured and survival may be biased. A prior survey-based cohort study reported increased mortality risk in working-aged adults with disruptions in either private or public coverage (19). Future research may use alternative data sources (eg, registry and all-payer claims data linkage in some states) to examine insurance continuity patterns before and after a cancer diagnosis among privately insured children, adolescents, and young adults.

Second, although the current SEER registries cover approximately 48% of the US population (3,54), our analysis was limited to the 12 states participating in the SEER-Medicaid data linkage effort, which may limit the generalizability of our results. Nevertheless, the SEER-Medicaid data constitute the only available linkage of multistate population-based registries with insurance data on children, adolescents, and young adults with cancer, providing the largest sample to date for studying the relationship between insurance enrollment dynamics and cancer survival in pediatric, adolescent, and young adult populations. Third, registry-reported insurance status at diagnosis may undercount Medicaid coverage (55,56); to address this issue, our study used administrative enrollment data linked to SEER registries. Without healthcare claims or encounter data, we did not have information about patients’ healthcare utilization, such as treatment for preexisting medical conditions. Moreover, our study followed patients up to 5 years after diagnosis to assess survival, which may be relatively short (57,58).

Finally, the survival differences observed between patients with continuous Medicaid and patients with private/other (largely Medicare and military/Department of Veterans Affairs) insurance may be explained by factors beyond insurance itself because Medicaid eligibility is based on having low household income or disability status. These factors include concurrent individual-level health-related social needs that negatively affect patients’ ability to seek care even with insurance: transportation challenges, household material hardship, insufficient social support, and low health literacy; difficulty accessing specialists who accept Medicaid insurance; inability to navigate complex care; and a higher burden of comorbidities, poorer overall health before diagnosis, and risky behaviors after cancer treatment (11,59). Such factors, often more prevalent among socioeconomically disadvantaged populations relying on Medicaid (11), were not accounted for in our analysis because these measures were unavailable. With the introduction of the International Statistical Classification of Diseases, Tenth Revision, codes for health-related social needs, such as those related to literacy and economic circumstances (60), future studies could use these codes to measure health-related social needs. This change would allow for an empirically well-grounded comparison of survival between Medicaid enrollees and patients with private/other insurance, accounting for health-related social needs–related differences.

We provide the first evidence that having continuous Medicaid coverage is associated with survival benefits for pediatric, adolescent, and young adult patients with diagnosed blood cancers; however, only 2 in 5 Medicaid-insured patients had continuous Medicaid coverage in the year preceding cancer diagnosis. These findings suggest that when examining the modifiable social drivers of blood cancer survival, it is crucial to differentiate patients with continuous insurance coverage from patients who newly gained coverage because of a cancer diagnosis. More work is needed to investigate the mechanisms through which insurance enrollment patterns affect survival of pediatric, adolescent, and young adult blood cancers. Important factors include disease severity at initial diagnosis; the quality of care during the active treatment phase; and access to appropriate resources during posttreatment survivorship, including supportive services, palliative care, and end-of-life care. Future work should also comprehensively evaluate the effects of state Medicaid policies, such as Medicaid expansion, parameters of waiver programs, and providing 12-month continuous eligibility (9), on coverage continuity for this young, vulnerable population.

Data availability

The authors cannot make the data publicly available because of the restrictions of the SEER-Medicaid Data Use Agreement. Researchers can visit https://healthcaredelivery.cancer.gov/seermedicaid/obtain/ for information about requesting and accessing the SEER-Medicaid data.

Author contributions

Xu Ji, PhD (Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Resources; Software; Validation; Visualization; Writing—original draft; Writing—review & editing); Xinyue Zhang, MPA (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Software; Validation; Visualization; Writing—review & editing); K. Robin Yabroff, PhD (Conceptualization; Investigation; Methodology; Resources; Validation; Visualization; Writing—review & editing); Wendy Stock, MD (Conceptualization; Investigation; Methodology; Resources; Validation; Visualization; Writing—review & editing); Patricia Cornwell, LCSW (Conceptualization; Investigation; Methodology; Resources; Validation; Writing—review & editing); Shasha Bai, PhD (Conceptualization; Funding acquisition; Investigation; Methodology; Validation; Visualization; Writing—review & editing); Ann C. Mertens, PhD (Conceptualization; Funding acquisition; Investigation; Methodology; Resources; Supervision; Validation; Writing—review & editing); Joseph Lipscomb, PhD (Conceptualization; Investigation; Methodology; Resources; Supervision; Validation; Visualization; Writing—review & editing); Sharon M. Castellino, MD, MSc (Conceptualization; Funding acquisition; Investigation; Methodology; Project administration; Resources; Supervision; Validation; Visualization; Writing—review & editing).

Funding

This work has been supported by grant No. HSR9015-23 from the Leukemia & Lymphoma Society’s Equity in Access Research Program (Ji [multiple principal investigator], Bai, Mertens, and Castellino [multiple principal investigator]) and by grant K01MD018637 (Ji [principal investigator], Mertens, and Castellino) from the National Institute on Minority Health and Health Disparities of the National Institutes of Health. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Leukemia & Lymphoma Society or the National Institutes of Health.

Conflicts of interest

K.R.Y. has served on the Flatiron Health Equity Advisory Board and a National Comprehensive Cancer Network working group; all honoraria are donated to her employer, the American Cancer Society. S.M.C. has served on the advisory board of Seagen Inc and Bristol Meyers Squibb. The authors have no other conflicts of interest to disclose.