Examining the effect of Medicaid expansion on early detection of head and neck cancer of the oral cavity and pharynx by HPV‐type and generosity of dental benefits

Abstract Background Over a decade of evidence supports the claim that increased access to insurance through Medicaid expansions improves early detection of cancer. Yet, evidence linking Medicaid expansions to early detection of head and neck cancers (HNC) of the oral cavity and pharynx, specifically, may be limited by the lack of attention to Human Papillomavirus (HPV) etiology, generosity of dental coverage, and valid inference analyzing state cancer registry data. Aims This study reexamined the effect of Medicaid expansion on early detection of HPV+/− HNC in states offering extensive dental benefits. Materials and Methods Specialized data from the Surveillance, Epidemiology, and End Results (SEER) program was analyzed to account for, previously unmeasurable, differential detection patterns of HNCs associated with HPV. Then, to identify the effect of increasing Medicaid eligibility on staging patterns in states offering extensive benefits amidst potentially non‐common trends between states, a “Triple Differences” design identifies the differential effect of Medicaid Expansion (with dental coverage) on HPV‐negative HNCs relative to the change in HPV‐positive HNCs. For valid inference analyzing a small number of state clusters (12) in cancer registry data, each regression model applies a Wild Cluster Bootstrap. Results Expanding Medicaid eligibility was found to be associated with a decrease in the proportion of distant‐stage diagnoses of HPV(−) HNCs, but only among states which increased Medicaid dental generosity at the time of Medicaid expansion. Conclusions These results suggest that adding extensive Medicaid dental benefits was the primary mechanism impacting HNC detection. This study highlights the potential positive spillover effects of policies which increase access to public dental coverage for low‐income adults, while also showing the limitation of access to dental services for improving early detection of HPV+ HNCs.


| INTRODUCTION
This paper reexamines the potential impact of new access to affordable Medicaid insurance coverage on staging outcomes for head and neck cancers of the oral cavity and pharynx (HNC). Specifically, this study investigates the Affordable Care Act's (ACA) Medicaid expansions, a policy which has been rigorously studied for nearly a decade. 1 Studies have generally found that by increasing access to physicians, Medicaid expansions improved physician-based cancer screenings and early-stage cancer diagnoses of cancers with systematic screening protocols. 2 However, the limited research on HNC has not resulted in clear consensus, or sufficiently explained the potential mechanisms linking access to Medicaid with improved staging outcomes. [3][4][5][6] Comprising over 90% of all head and neck cancers, HNCs of the oral cavity and pharynx kill more than 10 000 adults each year. [7][8][9] Compared to localized-stage HNCs, a distant-stage diagnosis can lower the probability of 5-year survival by 40%-70%. 9 Late detection also raises healthcare expenditures and exacerbates financial toxicity.
The annual treatment cost-per-person for late-stage HNCs is $10 000 higher than the standard of care for early-stage HNCs. 10,11 In 2019, less than 30% of HNCs were diagnosed at localized stages. 8,9 Policies that increase the proportion of HNCs detected early could dramatically improve outcomes for this rare, but deadly disease.
Two recent studies found mixed evidence that Medicaid Expansion improved HNC detection. The first study analyzed hospital-based data to find that stage I-II diagnoses of nonoropharyngeal HNC increased in Medicaid Expansion states relative to non-Expansion states. 12 The second study analyzed populationbased data to find that Medicaid Expansion was associated with higher rates of early-stage diagnoses, but only in low-incidence populations (i.e., young adults, females). 13 Both studies used traditional Difference-in-Differences designs, but neither study empirically examined their analytical models for bias. More critically, these prior studies relied on the fundamental assumption that, upon gaining access to Medicaid coverage, low-income adults at risk of developing HNC were widely screened by physicians, largely ignoring the role of dentists in state Medicaid dental systems.
Dental services are a critical component of HNC detection. In the United States, dental professionals are among the only healthcare provider recommended to systematically conduct visual oral cancer examinations for adult patients. [14][15][16][17] This is likely why dentists conduct the overwhelming majority or HNC screenings. 18 Unfortunately for efforts to improve population oral health, financial barriers prohibit adequate utilization of dental services. 19,20 Medicaid has served as a critical access point for low-income adults, but barriers persist. Adult Medicaid dental benefits remain optional and, therefore highly volatile. The consequences of this variation in access to Medicaid dental services continues to be of great concern for policymakers and investigators.
If the ACA's Medicaid expansions were to improve HNC detection, the most likely pathway would be through increased access to Medicaid dental services and dentist-based HNC screenings.
However, HNCs comprise a heterogenous group of malignant tumors. 8,21 The efficacy of discriminatory screening varies not just by tumor site, but by Human Papillomavirus (HPV) etiology. 22 HNCs caused by HPV, which now account for the bulk of all HNCs, are less likely to be detected early by visual examination. 8,12,23 Researchers and clinicians continue to develop and test novel approaches for detecting HPV+ HNCs at earlier stages. [24][25][26][27] However, current HPVassociated oropharyngeal cancer screening protocols are currently not justified in the population. 28 HNCs not associated with HPV, rather, can be screened for and identified by a visual examination during a comprehensive oral exam; an exam which is most often performed by a dental professional. 18 39 Although, California and Washington added extensive dental benefits in 2014. 40,41 All other states did not change their Medicaid dental benefits during this time.
The final strategy tests if adding Medicaid dental coverage or increasing Medicaid eligibility is the critical mechanism impacting HNC outcomes. Here, the two states which added extensive Medicaid dental benefits in 2014 are considered the treated group and all other states as controls. To estimate the association between adding dental benefits and HNC outcomes, this study tests for differential changes between groups before and after 2014. Then, to test if increasing Medicaid eligibility (without dental benefits) was associated with HNC outcomes, this study tests for differential changes before and after 2012 (when both CA and WA began implementing Medicaid expansion) for California and Washington compared to all other states; this test excludes the years 2014-2016. 42

| Variables
The primary outcomes of interest are derived from the SEER Combined Summary Stage variable, which categorizes the presentation of each tumor as localized, regional, and distant. 43 The first outcome is a binary variable indicating if the HNC patient was diagnosed at a localized stage. The second outcome is a binary variable indicating if the HNC patient was diagnosed at a distant stage. Medicaid coverage serves as a secondary outcome. 44 To account for temporal trends and unobserved heterogeneity, all models control for year and state fixed-effects. The model also accounts for seasonal variation by including month dummy variables. All models control for patient age, race/ethnicity, sex, marital status, tumor site, and metro status.

| Empirical strategy
As a linear probability regression model, a Difference-in-Differences design identifies the association between Medicaid expansion and HNC outcomes. The initial specification aims, in part, to replicate prior evidence of Medicaid expansion's association with the probability of an early-stage HNC diagnosis. Stratification analyses test for differences by HPV status and the timing of extensive Medicaid dental coverage. The final specification constructs a Triple Differences Model, which identifies the association between Medicaid expansion with dental benefits and HNC outcomes for HPV(À) HNCs, relative to the association between Medicaid expansion and HNC outcomes for HPV + HNCs. This third difference accounts for non-common trends between expansion and non-expansion states, as long as those noncommon trends are consistent between HPV+ and HPV(À) HNCs.
For all models, an event-history study tests for pre-expansion differential trends in HNC outcomes between expanding and non-expanding states. See Appendix A1 for more details on these identification assumption and pre-trend tests.

| Statistical analysis and inference
Given that the treatment is assigned at the state-level, theory and common practice suggest that inference should utilize standard errors clustered at the state-level to account for within-state correlation.
However, because of the limited number of states participating in SEER, there are insufficient clusters for valid, robust inference. To mediate this threat to inference, this study implements robust Wild Bootstrapping procedures following each model. 45 Using Webb resampling weights to further ensure robustness against type 1 error due to small number of clusters, 999 replicate bootstraps with standard errors clustered at the state-level construct a 95% confidence interval set. 45 Statistical significance is set at alpha = .05. All analyses were conducted using STATA v. 17.   Table 1 for sample descriptive statistics.

| Empirical results
In the full sample, Medicaid expansion was associated with a 2.46 percentage point increase in the probability of Medicaid coverage (CI = [À0.0164, 0.1018]; p = .0821), but had a near-zero and statistically insignificant association with localized stage diagnoses (

| Expansion in states adding dental coverage in 2014
Similar to the estimates in the states always covering generous Medicaid dental benefits, in states adding generous dental coverage at the time of expansion there also appears to be a significant and positive association between expansion and Medicaid cover-  Estimates for localized diagnoses for HPV+ HNCs in these states may be threatened by differential pre-expansion trends (Supplemental Table 2). For this reason, the DDD estimates for HPV(À) may not be valid (Supplemental Table 5). However, there is no evidence from the DD or DDD pre-trend tests to suggest differential trends in distant-stage diagnoses for HPV(À) HNCs prior to expansion in these states (Supplemental Tables 3 and 6).

| Adding dental coverage in 2014
Adding extensive Medicaid dental benefits was not associated with changing Medicaid coverage (Tables 2 and 3). However, adding T A B L E 2 Medicaid expansion's association with HNC outcomes (DD).  This estimate corresponds to a 16.3% relative increase in the probability of localized-stage diagnoses relative to 2013 rates. The DDDestimates yield similar results in terms of magnitude and inference, but the DDD estimate for localized diagnoses may be threatened by significant pre-trends (Supplemental Table 8). The DD and DDD estimates for the distant-stage diagnoses are supported by the lack of significant pre-trend test statistics (Supplemental Tables 7 and 8).

None of the estimates for the association between expanding
Medicaid in 2012 (without dental benefits) and HNC staging were even marginally statistically significant. Similarly, this study only found an association between expanding Medicaid with specific subpopulations. 12,13 Perhaps the subgroups of these prior studies were, in fact, proxies for HPV(À) HNC. Note the Sineshaw study includes HPV-stratification as a supplemental aim, and this study affirms their hospital-based results at the populationlevel. 12 Viewed together, three studies now suggest that Medicaid expansion had some positive impact on earlier HNC detection, but only for certain adults. This work argues that those certain adults were patients with HPV(À) HNCs who gained access to extensive Medicaid dental services. Until future research extends the design using more years and more states, these results should only be generalized to California and Washington state adults, among which the vast majority of the sample were male, non-Hispanic White, and metro county residents.
To further contextualize these results, it is important to assess the validity of this study's inference. Here, the discussion extends beyond studies investigating Medicaid expansion and HNC, but all quasi-experimental, state-based policy research analyzing cancer registry data. In most of the cancer research, including the HNC studies motivating this work, estimating valid standard errors receives little detail or consideration.
More concerning was the extensive work evaluating Medicaid expansion T A B L E 3 Medicaid expansion's association with HPV(À) HNC outcomes (DDD). As of January 2023, all but 11 states have expanded Medicaid eligibility through the ACA. 46 According to the latest MACPAC report, 14 states do not cover any preventative dental services to the adult population. 39 Most states today operate their Medicaid dental program via a managed care system, each implementing different strategies to control oral health care and expenditures. 47 There is a well-established, causal relationship between access to generous Medicaid dental benefits and higher utilization of dental services. 48,49 Undoubtedly, by preventing low-income adults from accessing affordable health insurance, the persistence of non-expansion states likely contributes to health inequities. 50 This current study further illuminates the potential inequities in HNC detection that result from heterogenous and volatile access to Medicaid dental services.

AUTHOR CONTRIBUTIONS
The author confirms sole responsibility for the following: study conception and design, data collection, analysis and interpretation of results, and manuscript preparation.

FUNDING INFORMATION
National Institute of Dental and Craniofacial Research NIDCR 1F31DE032250-01.

CONFLICT OF INTEREST STATEMENT
The authors have stated explicitly that there are no conflicts of interest in connection with this article.

DATA AVAILABILITY STATEMENT
Data sharing is restricted by third-party. Investigators can access publicly available SEER HNC-HPV datafile at NCI. 30

ETHICS STATEMENT
This retrospective analysis uses deidentified, secondary data and does not meet the definition of human subjects research (45 CFR 46.101 (b)(4)).