Oral cancer screening prevalence in low-income adults before and after the ACA

Abstract

Background

Detecting oral cancer early is associated with higher probability of survival, reduced treatment costs, and improved quality of life. Unfortunately, less than 30% of oral cancers are detected early. Recent health insurance expansions from the Affordable Care Act (ACA) could improve outcomes by increasing access to screening. However, due to the differences in screening practices by physicians and dentists, the impact of expanded access to insurance on oral cancer screenings remains unknown.

Methods

Self-reported oral cancer screening data were obtained from The National Health and Nutrition Examination Survey (NHANES) for years 2011–2017. NHANES questionnaires ask respondents if they have received an oral cancer screen from a physician or dentist in the past year. Along with adjusting for demographic characteristics, this study accounts for unobserved heterogeneity by comparing “Differences-in-Differences” estimates of low-income adults (<200% FPL) with high-income adults, before and after the ACA (2014), for adults most exposed (< age 65) to insurance expansion.

Results

Before and after the ACA, low-income adults had the lowest prevalence of oral cancer screenings. However, relative to high-income adults, the ACA was associated with a 5–6%-point increase in oral cancer screenings for low-income adults under age 65, but only for screenings performed by dentists.

Conclusions

Overall, oral cancer screening rates have been declining across the population, but the ACA may have slowed the decline in low-income adults. Understanding why oral cancer screenings are declining could inform cancer control policies. Research evaluating the impact of access to oral cancer screenings remains warranted.

Introduction

With nearly 55,000 adults diagnosed in 2020, oral cancer is becoming one of the most common cancers in the United States [1–2]. Since 2000, the oral cancer incidence rate has grown by 2% each year [3–4]. 11,000 adults are expected to die from oral cancer in 2022.

Stage at diagnosis is a critical prognostic factor [5–9]. Compared to oral cancers diagnosed at a local stage, a distant stage oral cancer diagnosis can lower the probability of five-year survival dramatically [8–9]. For example, the probability of five-year survival for three common oral cancer sites (floor of the mouth, tongue, and lip) are 76%, 82%, and 94% if diagnosed at an early-stage [9]. However, for those same common oral cancer sites, the probability of five-year survival drops to 20%, 40%, and 32% for patients diagnosed at late stages [9]. Unfortunately, late-stage oral cancers represent the bulk of oral cancers diagnosed in the United States, as less than 30% of all oral cancers are diagnosed early [1,9].

In addition to lower probability of survival, an oral cancer diagnosed at a late stage is also more difficult and expensive to treat. Previous research has estimated that the annual cost per patient for late-stage oral cancer treatment can exceed $10,000 more than early-stage oral cancer treatment [10]. These excess costs not only increase aggregate healthcare expenditures but raise out-of-pocket costs for patients newly diagnosed with late-stage oral cancer [11]. Finally, late-stage oral cancer diagnoses have been found to be associated with lower overall quality of life, inability to perform normal Activities of Daily Living, and increased risk of suicide and self-harm [12–15].

Theory and empirical evidence suggest that expanding insurance coverage will improve cancer detection, generally, as increasing access to healthcare services will lead to more opportunities to screen for and diagnose cancers early [16–20]. Yet, evidence linking health insurance to early oral cancer detection, specifically, has been limited. Two recent studies have investigated whether the Affordable Care Act’s (ACA) Medicaid Expansion was associated with more early-stage oral cancer diagnoses [21–22]. One hospital-based study found that early-stage diagnoses of non-oropharyngeal head and neck carcinomas increased, relative to non-Expansion states, after the ACA’s Medicaid Expansion [21]. A second, population-based cancer registry study found that Medicaid Expansion was associated with a relative increase in early-stage oral cancer diagnoses for lip cancers [22].

While each study used different cancer datasets, both empirical studies used quasi-experimental approaches and arrived at a consistent conclusion: by increasing access to care, the ACA’s Medicaid Expansion likely increased the proportion of oral cancers diagnosed early (at least, relative to oral cancers diagnosed in non-Expansion states). While not explicitly stated, both studies appear to be relying on an assumption that after gaining access to the healthcare system through the ACA, patients were screened for oral cancer. However, despite the low survival and quality-of-life associated with late-stage oral cancer, and because of the limited evidence for efficacy, the three primary authorities do not recommend that physicians perform population or risk-stratified oral cancer screening [23–25]. So, while the ACA increased access to, and utilization of physician-based healthcare services, physicians may not be routinely performing oral cancer screenings, leaving the potential effect of the ACA on physician-based oral cancer screenings largely unknown.

Unlike physicians, however, dentists have established oral cancer screening as standard of care during routine examinations and consequently perform over 75% of all oral cancer screenings [26–28]. Not only are dentists most likely to perform an oral cancer screening, but most early-stage oral cancer diagnoses can be traced back to dental visits [29]. Unfortunately, low-income adults forgo necessary dental care due to cost more than any other healthcare service [30]. Strong evidence suggests that dental service visits among low-income adults increased after the ACA’s Medicaid Expansion, albeit only in states offering comprehensive adult Medicaid dental coverage [31–34]. Aside from expanding Medicaid, however, few provisions in the ACA directly addressed gaps in access to dental services, potentially limiting the ACA’s effect on dentist-based oral cancer screenings.

To improve outcomes, lower income adults will need access to oral cancer detection services (i.e., screening), which the current U.S. healthcare system may not be adequately providing. The ACA expanded access to many preventive cancer screening services, and ultimately decreased the proportion of late-stage diagnoses for many common cancers [35,46]. However, we cannot yet ascertain that these findings extend to oral cancer. The ACA may have improved physician-based oral cancer screenings, given the focus on increasing access to physician-based services, however, physicians are not recommended to routinely perform oral cancer screenings [23–25, 35,26]. Meanwhile, the ACA’s Medicaid Expansion improved access to dental services, the service most likely to include a routine oral cancer screening, but the ACA’s effect on dental services was limited and focused to certain populations in select states [26–29,31–34]. For low-income adults residing in states which did not expand Medicaid, or which do not cover comprehensive adult Medicaid dental benefits, the ACA may have done little to improve access to dentist-based oral cancer screenings.

Existing research has yet to systematically examine the potential mechanisms linking the ACA to oral cancer detection. Failing to consider the mechanisms linking access to new health insurance and early-stage oral cancer diagnoses can hinder the dissemination of evidence-based policy. This study aims to illuminate the potential mechanisms by investigating how the ACA may have impacted oral cancer screening rates among low-income adults. Knowing whether the ACA altered physician or dentist-based oral cancer screening trends can inform future healthcare reforms needed to reduce the burden of late-stage oral cancer across the country and inform future studies evaluating the efficacy of access to oral cancer screening on survival, quality of life, and mortality.

Materials and Methods

Data

Self-reported oral cancer screenings were obtained from The National Health and Nutrition Examination Survey (NHANES) [27]. The NHANES is a population-based survey. NHANES survey design incorporates complex sampling and probability weighting procedures to represent the health status of the U.S. population. For years 2011, 2013, 2015, and 2017, NHANES included an oral cancer screening question on the in-person survey. Respondents were asked if they “have ever had an exam for oral cancer in which the doctor or dentist pulls on their tongue, sometimes with gauze wrapped around it, and feels under the tongue and inside the cheeks?”. Respondents were then asked when they completed their most recent oral cancer screening and what type of provider performed the oral cancer screening.

Outcome

These self-reported oral cancer screening responses were used to develop a set of binary outcome variables, the first of which indicates if the respondent received an oral cancer screen in the past year. A second set of binary variables indicates if the respondent received an oral cancer screen in the past year from a specific provider (I.e., physician, dentist, nurse practioner, dental hygienist/assistant). Alternative specifications grouped together providers by setting (physicians grouped with nurse practioners, dentists with dental hygienists/assistants) and by level of provider (physician grouped with dentists, nurse practioners with dental hygienists/assistants).

Covariates

NHANES also includes self-reported demographic data. All analyses adjust for age, race/ethnicity, gender, education level, marital status. These covariates are expected to be exogenous to ACA healthcare reform and possibly associated with oral cancer screening patterns.

Exposure

To identify how the ACA may have impacted oral cancer screening rates for low-income adults, this study leverages the variation in exposure to new access to healthcare services after the ACA. The ACA increased access to affordable health insurance by expanding Medicaid and implementing ACA Marketplace subsidies [37,38]. However, only adults under 200% FPL were eligible for these new health insurance options [37, 38]. Further, adults under age 65 were more likely to take up the new ACA insurance options than adults at or above at 65, as the latter group was already eligible for Medicare. Based on the implementation of the ACA and empirical evidence, we expected that access to healthcare, through affordable insurance, would not significantly change for older and higher income adults, whereas lower-income adults under age 65 would experience the greatest benefit (and uptake of new insurance) from the ACA [39,40].

Statistical Analysis

Using differential exposure to the ACA’s impact, this study estimates the relative change in oral cancer screening rates for low-income adults by comparing “Differences-in-Differences” estimates of average oral cancer screening rates for low-income adults (<200% FPL) with high-income adults, before/after the ACA (2014), for adults more exposed (< age 65) and less unexposed (>=age 65) to insurance expansion. Along with adjusting for exogenous demographics, this approach accounts for unobserved heterogeneity by differencing away secular trends in oral cancer screening across the entire population, but also accounts for group-specific time-invariant factors which may be influencing differential oral cancer screening behavior [41–46]. All analyses incorporated the NHANES probability sampling weights. Each linear regression estimated standard errors robust to heteroskedasticity. Data extraction, cleaning, and analysis were conducted in STATA v. 17 [47].

Placebo Tests

The association, or average marginal effect estimate, between access to healthcare services after the ACA and oral cancer screening can be interpreted as a causal effect under the “parallel trends” assumption [45–48]. The “parallel trends” assumption states that in the absence of ACA, the group-specific oral cancer screening trends would not have changed [48,49]. While this assumption does not require equal levels of oral cancer screening prior to the ACA, the assumption cannot be completely tested since the post-ACA counterfactual cannot be observed. However, pre-ACA trends can be used to assess the validity of our analytical strategy and plausibility of the “parallel trends” assumption. Here, we implemented a placebo test to determine if group-specific oral cancer screening trends were parallel prior to the ACA’s implementation [45,46]. Coefficients estimated to be significantly different than zero would indicate pre-treatment differential trends and cast doubt on our causal interpretation.

Alternative Specifications

The primary sample is restricted to adults younger than 65, and specifies adults under 200% FPL as the “treated” group given that the ACA’s subsidies were most likely to benefit that population. To ensure that our results are not specific to the 200% FPL threshold, each analysis was recalculated using a range of FPL cutoff points from 60%−350% FPL. We expect the results to be consistent, but not meaningfully different, at and around the 200% FPL cutoff and expect null and insignificant results further away from the 200% FPL cutoff. We then include a set of alternate insurance outcomes to ensure our sample satisfies first order effects from the ACA insurance expansions. We also tested if the ACA was associated with changes in two-year oral cancer screening rates and reporting ever having had an oral cancer screen.

Subgroups

The subgroup analyses aimed to assess the generalizability of the ACA’s association on oral cancer screening rates. We conducted a series of subgroup analyses by factors exogenous to the ACA: race/ethnicity, gender, education level, marital status, and whether children were living in the home. We then test if the marginal effect estimates of the ACA on oral cancer screening differed within each group. Despite potential endogeneity threats, as an exploratory analysis we examine differences in the ACA’s association with oral cancer screening among adults based on smoking history and Human Papillomavirus (HPV) status.

Results

Before and after the ACA, low-income adults under age 65 had the lowest rate of oral cancer screenings (Table 1). However, the Differences-in-Differences estimates reveal that, relative to high-income adults, the ACA was associated with a 6.2-percentage point increase (p < 0.001) in oral cancer screenings for low-income adults under age 65 (Table 2). To our surprise, oral cancer screening rates have been declining across the population since 2011 (Figure 1–2).

Table 1 -.

Summary Statistics (Proportions %) of Outcome and Control Variables

	2011 & 2013				2015 & 2017
	Age >= 65, >= 200% FPL	Age >= 65, < 200% FPL	Age < 65, >= 200% FPL	Age < 65, < 200% FPL	Age >= 65, >= 200% FPL	Age >= 65, < 200% FPL	Age < 65, >= 200% FPL	Age < 65, < 200% FPL
Outcome Variables
Recent Oral Cancer Screen	43.4	17.5	29.2	8.3	34.4	7.4	19.7	5.6
By Physician	5.4	5.1	2.7	1.7	1.0	0.2	0.4	0.2
By Dentist	30.9	9.2	22.5	5.1	26.0	6.1	15.5	4.1
By Physician/Nurse Practioner	5.5	5.2	2.8	1.8	1.0	0.3	0.4	0.2
By Dentist/Dental Hygienist	37.4	12.1	26.4	6.5	33.3	7.0	19.1	5.0
By Nurse Practioner	0.0	0.2	0.1	0.0	0.0	0.1	0.1	0.0
By Dental Hygienist	6.5	2.9	3.9	1.4	7.4	0.8	3.6	0.9
By Physician/Dentist	36.4	14.3	25.2	6.8	27.0	6.3	15.9	4.3
By Nurse Practioner/Dental Hygienist	6.5	3.0	4.0	1.4	7.4	0.9	3.6	1.0
Control Variables
Male	49.3	38.7	49.3	47.0	47.5	40.8	49.8	45.0
non-Hispanic White	83.9	65.1	73.0	52.8	82.5	61.3	69.9	48.8
Hispanic	4.9	15.0	11.1	25.1	5.8	18.5	13.1	28.0
non-Hispanic Black	6.9	14.5	9.5	18.0	7.2	13.8	10.2	17.4
No High School Degree	10.4	34.8	8.2	31.8	8.5	29.8	7.2	27.5
High School Degree Only	20.0	26.9	17.0	28.0	22.0	30.9	19.1	30.4
At Least Some College	32.4	26.2	29.5	30.1	28.7	28.7	31.0	31.8
Married	72.9	45.0	69.9	45.0	68.4	42.0	69.1	43.0
No Children Living in Home	94.4	86.5	57.4	43.1	93.2	84.4	55.2	43.1

Table 2 -.

The Association Between the ACA and a Recent Oral Cancer Screening (Any Provider)

Coefficients	Difference-in-Differences			Triple-Differences
	Full Sample	Adults >= 65	Adults <65	Full Sample
POST	−0.086***	−0.091**	−0.087***	−0.090**
	(0.013)	(0.028)	(0.014)	(0.028)
FPL_UNDER200	−0.122***	−0.169***	−0.110***	−0.180***
	(0.012)	(0.028)	(0.013)	(0.027)
FPL_UNDER200*POST	0.048**	0.002	0.062***	−0.002
	(0.016)	(0.036)	(0.018)	(0.036)
AGE_30–64				−0.188***
				(0.056)
AGE_30–64*POST				0.004
				(0.032)
FPL_UNDER200*AGE_30–64				0.074*
				(0.030)
FPL_UNDER200*AGE_30–64*POST				0.063
				(0.041)
N	16639	4139	12500	16639

Figure 1.

Year-by-Year Oral Cancer Screening

Figure 2.

Screening by Physician Dentist

When investigating by type of provider, the ACA had a minimal and insignificant association with oral cancer screenings performed by a physician and/or a nurse practioner (Table 3; Supplemental Table 1). Conversely, the ACA was significantly associated with a 5.5-percentage point relative increase (p < 0.001) in oral cancer screenings performed by a dentist. The estimate for any dental professional was also positive and significant.

Table 3 -.

The Association Between the ACA and a Recent Oral Cancer Screening (By Type of Provider)

	Physician				Dentist
	Difference-in-Differences			Triple-Differences	Difference-in-Differences			Triple-Differences
Coefficients	Full Sample	Adults >= 65	Adults <65	Full Sample	Full Sample	Adults >= 65	Adults <65	Full Sample
POST	−0.027***	−0.044***	−0.023***	−0.044***	−0.059***	−0.052	−0.062***	−0.052
	(0.004)	(0.009)	(0.004)	(0.009)	(0.012)	(0.027)	(0.013)	(0.027)
FPL_UNDER200	−0.006	−0.001	−0.008	−0.002	−0.108***	−0.149***	−0.097***	−0.156***
	(0.004)	(0.013)	(0.004)	(0.013)	(0.010)	(0.025)	(0.011)	(0.024)
FPL_UNDER200*POST	0.005	-0.004	0.007	−0.004	0.049***	0.031	0.055***	0.029
	(0.005)	(0.013)	(0.005)	(0.013)	(0.014)	(0.033)	(0.016)	(0.033)
AGE_30–64				−0.039*				−0.153**
				(0.018)				(0.052)
AGE_30–64*POST				0.021*				−0.010
				(0.010)				(0.030)
FPL_UNDER200*AGE_30–64				−0.005				0.060*
				(0.013)				(0.026)
FPL_UNDER200*AGE_30–64*POST				0.011				0.026
				(0.014)				(0.036)
N	16639	4139	12500	16639	16639	4139	12500	16639

Placebo Tests

Each of the Triple Differences estimates were similar in magnitude and direction as the DD estimates, yet none of the Triple Differences estimates were significant. Investigating the DD estimates for older adults and each placebo test helps illuminate the inconsistency. First, consistent with our prediction, we did not observe any association between the ACA and oral cancer screenings by any provider for adults at or above age 65. Further, our placebo tests did not reveal any pre-ACA differential trends in oral cancer screenings performed by any provider (Table 4). However, while we find that the ACA was not associated with physician-based oral cancer screenings in older adults, the placebo test in this older age group indicates that physician-based oral cancer screening trends were not comparable prior to the ACA. We also observe the possibility of differential trends for dentist-based oral cancer screenings prior to the ACA for older adults (Table 5). These placebo tests suggest that the identifying assumption of “parallel trends” between high and low-income adults at or above age 65 is violated. Conversely, there are no pre-ACA differential trends for any provider, physician, or dentist in the under age 65 group, strengthening the validity of our primary estimates and supporting the traditional DD model as the most appropriate strategy for our study.

Table 4 -.

Placebo Tests – Comparing Pre-ACA Oral Cancer Screening Trends

	Difference-in-Differences			Triple-Differences
Coefficients	Full Sample	Adults >= 65	Adults <65	Full Sample
PLACEBO_POST	0.001	0.016	−0.003	0.014
	(0.018)	(0.040)	(0.020)	(0.040)
FPL_UNDER200	−0.101***	−0.145***	−0.092***	−0.157***
	(0.018)	(0.043)	(0.019)	(0.043)
FPL_UNDER200*PLACEBO_POST	−0.010	−0.014	−0.005	−0.016
	(0.022)	(0.054)	(0.024)	(0.055)
AGE_30–64				−0.106
				(0.069)
AGE_30–64*POST				−0.017
				(0.044)
FPL_UNDER200*AGE_30–64				0.069
				(0.046)
FPL_UNDER200*AGE_30–64* PLACEBO_POST				0.012
				(0.060)
N	8377	1983	6394	8377

Table 5 -.

Placebo Tests – Comparing Pre-ACA Oral Cancer Screening Trends (By Type of Provider)

Coefficients	Physician				Dentist
	Difference-in-Differences			Triple-Differences	Difference-in-Differences			Triple-Differences
	Full Sample	Adults >= 65	Adults <65	Full Sample	Full Sample	Adults >= 65	Adults <65	Full Sample
PLACEBO_POST	−0.007	−0.022	−0.004	−0.022	0.012	0.034	0.006	0.034
	(0.007)	(0.018)	(0.007)	(0.018)	(0.017)	(0.038)	(0.018)	(0.038)
FPL_UNDER200	−0.015*	−0.043*	−0.009	−0.041*	−0.088***	−0.096*	−0.086***	−0.108**
	(0.006)	(0.017)	(0.006)	(0.017)	(0.015)	(0.039)	(0.016)	(0.037)
FPL_UNDER200*PLACEBO_POST	0.020*	0.071**	0.008	0.069**	-0.010	-0.060	0.005	−0.063
	(0.009)	(0.025)	(0.009)	(0.025)	(0.020)	(0.048)	(0.022)	(0.048)
AGE_30–64				−0.044				−0.041
				(0.032)				(0.060)
AGE_30–64*POST				0.018				−0.027
				(0.019)				(0.042)
FPL_UNDER200*AGE_30–64				0.031				0.026
				(0.018)				(0.040)
FPL_UNDER200*AGE_30–64* PLACEBO_POST				-0.061*				0.068
				(0.026)				(0.053)
N	8377	1983	6394	8377	8377	1983	6394	8377

Alternative Specifications

Our results do not appear to be sensitive to the FPL-threshold used to construct the low-income “treatment” group. Supplemental Figure 1 visually shows the estimated association between the ACA and any recent oral cancer screen for adults under age 65, at varying levels of FPL-thresholds to determine “low-income” status. The estimates peak at predictable thresholds (130% and 200%), critically relevant FPL thresholds for ACA insurance subsidies. Further, the estimates maintain (p < 0.05) significance from 90%−260% FPL and (p < 0.001) significance from 130%−200% FPL. These results not only justify our primary specification for setting the FPL-threshold at 200% based on access to ACA subsidies, but also casts doubt that our results are sensitive to how we define low-income adults.

Supplemental Table 3 reports alternative outcomes, which appear to be consistent with the extensive research evaluating the ACA’s effect on insurance coverage [39,40]. We found that uninsurance rates declined in low-income adults after the ACA and that Medicaid coverage increased. We also found positive associations between the ACA and an oral cancer screen in the past two years, but no association with ever having had an oral cancer screen.

Subgroups

For most demographic categories, we failed to reject the null hypothesis that the ACA’s estimated association with oral cancer screening differed by subgroups. The one exception was race/ethnicity, as the association for non-Hispanic White adults was positive and significant, and significantly higher than the association for all other racial/ethnic groups (which were all smaller in magnitude and insignificant). See Supplemental Tables 4-5 for the subgroup results.

Discussion

Our study evaluated how the ACA impacted oral cancer screening for low-income adults under age 65. Prior to the ACA, this population had lower oral cancer screening rates than older and higher income adults. We expected the ACA to improve screening rates for low-income adults under age 65, as they were most likely to benefit from expanded access to insurance. We found that the ACA was associated with higher oral cancer screenings for low-income adults under age 65, and that this association was only found for screenings performed by dentists.

However, it is important to understand that the estimates for low-income adults are relative to high-income adults. We do not observe an overall increase in oral cancer screenings for low-income adults after the ACA. In fact, oral cancer screening rates in low-income adults declined over the study period. This decline, however, was less than higher income and older adults. Was this relatively smaller decline caused by the ACA? It is plausible that the ACA, by increasing access to care, slowed the declining trends. But it is possible that the slower decline was due to lower baseline screening rates. We also cannot rule out that the declining rates were caused by another event which may have changed preventative screening guidelines or scope of practice policies (such as the 2013 update from the U.S. Preventative Task Force) [23]. Or rather, perhaps the declining rates we observe from 2011–2017 are just part of a larger trend over the last two decades where the incidence of oral cancers with screening protocols are declining relative to oral cancers caused by Human Papillomavirus [1–4]. Understanding the downward trend in oral cancer screenings remains critical for the future of oral cancer control efforts.

Still, our results help affirm the prior evidence that the ACA may have increased early-stage oral cancer diagnoses [21,22]. By only investigating Medicaid Expansion, however, the prior studies may underestimate the ACA’s total effect on improving early detection as evident from our results that the ACA was associated with a relative increase in screenings for adults up to 200% FPL. Additionally, the average effect estimates of these prior studies may be subject to unmeasured state-level heterogeneity driven by the availability of Medicaid dental benefits, given our finding that the ACA was positively associated with oral cancer screenings performed by dentists. Finally, our finding that oral cancer screenings were declining across the population matches the evidence presented from the two Medicaid Expansion studies [21,22]. The estimates from both studies show that, relative to non-Expansion states, more oral cancers were diagnosed early in states that expanded Medicaid under the ACA.. Consistent with our results, both prior studies found that early-stage oral cancers were declining across the population, but the decline was relatively smaller in states that expanded Medicaid. While quite different in terms of data and design, our study and the two prior ACA oral cancer evaluation studies share a consistent finding: that the ACA was associated with increased screening and earlier diagnoses for low-income adults with or at risk of developing oral cancer; but these better outcomes are relative as outcomes are worsening outcomes across the population over time. Access to care is critical, but to improve outcomes we must implement evidence-based policies and programs that affirmatively increase early detection.

Limitations

Readers should be aware that our aim was not to investigate a single component of the ACA, but rather the ACA as a whole (i.e., marketplace subsidies, Medicaid Expansion, pre-existing condition exclusions, etc.). NHANES data availability and survey design limit our ability to identify any change by state-level exposure such as expanding Medicaid dental coverage through the ACA. NHANES data does not provide publicly available state-level geographic indicators. Moreover, the NHANES sample is only selected and weighted to approximate the U.S. population, not each state’s population. Analyzing state-level NHANES data are more costly and potentially less valid than NHANES studies analyzing the total U.S. population. Given the drawbacks of NHANES state-level research, we focus on the broader potential impact of the ACA on oral cancer screening for all low-income adults in the United States.

While NHANES is the only population-based survey to include oral cancer screening questions, the size and structure of the NHANES data introduced many limitations. The sample sizes are smaller than other population-based studies, which limits the power to detect effects. The power issues are most obvious in our subgroup analyses, which are considered supplementary aims. Further, the oral cancer screening module in NHANES is also only conducted biannually. More problematic are the inherent issues with self-reported survey responses, especially for a clinical procedure not typically provided and not well understood by the respondent. While NHANES qualifies the question regarding oral cancer screening by describing the procedure, the potential for recall bias may be higher than more common healthcare services and screenings. Yet, we have no evidence to expect that this recall bias would differ over time or by age and income groups. Finally, the identification strategy relies on assumptions which can be difficult to justify [44–45]. We followed best practices to approximate a counter-factual by leveraging differential exposures to the ACA. We then conducted sensitivity checks, placebo tests, and alternative outcome tests to help validate our approach. Still, the bridge of causality is difficult to cross. We found a statistically significant association between the ACA and increased oral cancer screenings for low-income adult under age 65. However, this effect estimate is purely driven by the declining oral cancer screening rates in all populations, with a smaller decline in low-income adults under age 65. Our design can adjust for such secular trends but cannot explain them.

Conclusion

Overall, oral cancer screenings across the population have been declining since 2011. The ACA may have slowed the decline for low-income adults under age 65, as oral cancer screenings for low-income adults increased relative to high-income adults after the ACA. The relative increase was only observed for oral cancer screenings performed by dentists, signaling that access to dental services may be the mechanism linking the ACA with improved early-stage oral cancer diagnoses. Policies aiming to increase the proportion of oral cancers detected early should strive to expand access to dental services and the scope of practice for performing oral cancer screenings. Whether the changing patterns in oral cancer screening impact early-detection, survival, and mortality remains an open question.