The Shadow Price of Uncertainty: Consequences of Unpredictable Insurance Coverage for Access, Care, and Financial Security

Abstract

Policy Points.

• Health insurance reform in the United States has fostered enrollment to promote access to care and reduce financial insecurity. However, enrollees’ inability to reliably predict what insurance will cover (a.k.a. “coverage uncertainty”) impedes these goals, often as much as being uninsured.

• The Patient Protection and Affordable Care Act initially expanded enrollment and reduced coverage uncertainty. After the mid‐2010s, trends in coverage uncertainty plateaued, and it now impedes access to care for four times as many households as lack of health insurance.

• A variety of policies can moderate coverage uncertainty, but other popular reform strategies exacerbate it instead. Our findings suggest that structural reforms represent the most promising remedial strategies, particularly those that can enhance support for households negotiating coverage denials with insurers.

Context

Health insurance reform in the United States has focused on expanding enrollment, a goal inhibited by complex insurance provisions. Research documents this complexity and shows how it increases consumers’ challenges in anticipating needs and making informed choices, potentially deterring policy purchases. Little is known about how coverage uncertainty impacts those who have insurance.

Methods

Drawing on a multiwave survey with nationally representative data, we assessed consumer experiences and expectations in 2009, 2014, and 2021. Respondents identified (a) worries about the reliability of health insurance coverage, and (b) experiences of insurance not covering major medical expenses. Respondents also reported on three health care–related experiences—whether they delayed access to needed care, had been unable to effectively care for chronic health conditions, or felt anxious about future medical expenses. We estimated regressions associating metrics of coverage uncertainty with the three health care–related outcomes, controlling for socioeconomic status and other household characteristics.

Findings

Of American households, 32% reported intense worry about coverage reliability in 2009. This declined to 27% in 2014, then rebounded to 31% in 2021. Experiences of coverage shortfalls followed a similar pattern, declining from 27% to 17%, then rising back to 21%. Coverage uncertainty has statistically significant associations with all three outcomes, with access being the most sensitive to low‐level uncertainty. By 2021, coverage uncertainty deterred timely access in care for one in five American households, five times as many as among the uninsured.

Conclusions

Coverage uncertainty has become the predominant barrier to timely access. It also disrupts care for chronic conditions and exacerbates anxiety over medical expenses. These harms can be reduced. However, several popular health care reform strategies instead exacerbate coverage uncertainty. We explicate these overlooked cross‐policy connections and identify alternative strategies that could moderate the impact of coverage uncertainty in the United States

In december of 2024, a health insurance executive died on the streets of New York. Motives for Brian Thompson's murder remain murky but seem linked to UnitedHealthcare's frequent denials of coverage. ¹ The public's reaction—starkly conveyed through social media—was stunningly clear: an outburst of self‐righteous anger against insurers’ coverage determination practices, deemed to put patients’ lives at risk, each and every day. ² Whether UnitedHealthcare was in fact a pathological extreme or was emblematic of corporations putting “profits over people,” these interlocking tragedies can arguably be traced to a false step in US health policy a decade ago.

When the Affordable Care Act (ACA) became law in 2010, it embodied two distinct strategies for promoting access to care and financial security. The first involved expanding the number of Americans enrolled in health insurance by broadening Medicaid eligibility and creating subsidized exchanges for purchasing individual policies. The second pathway was intended to make coverage more reliable by prohibiting practices that induced unexpected coverage denials (e.g., exclusions for preexisting conditions) and by assisting consumers in negotiating coverage disputes with insurance companies. ³

Implementation skewed the balance among these strategies in part because of their relative newsworthiness. ⁴ Journalists (and their readers) could more readily track successes or failures expanding enrollment than enhancing coverage reliability. To get a “win” from media pundits, the Obama administration needed to increased insurance enrollment, and so they did. The number of uninsured Americans fell from 47 million in 2010 to 30 million by 2015. However, these impressive reductions came at a price. Resources originally intended for consumer assistance programs (CAPs) to help with coverage disputes were instead allocated to navigator programs promoting Medicaid and marketplace enrollment. ⁵ Between 2010 and 2012, a total of $60 million in federal funds went to CAPs, none after 2012. Beginning in 2013, federal funds for navigators totaled $275 million annually, with equal funding from the states. ⁶

That price carried a hidden cost. Many new enrollees struggled to make sense of their health insurance coverage. Although assistance with coverage issues was explicitly not within navigator programs remit, three‐quarters reported helping with the postenrollment problems, one in five indicating that this assistance was among their top three program activities. ⁵ Postenrollment cases were evenly split between coverage denials and coverage misunderstanding.

Confusion and denials were not unique to the newly insured. The Commonwealth Fund's cross‐national surveys revealed that in 2013, before insurance exchanges went operational, Americans were far more likely than citizens in other high‐income nations to report that health insurance had unexpectedly denied payment for a procedure or paid for less of a medical bill than expected (Table 1). Most strikingly, Americans were twice as likely to report such problems as those living in other countries that rely heavily on private health insurance markets, such as Switzerland or the Netherlands.

Table 1.

Unexpected Out‐of‐Pocket Medical Expenses as Markers of Coverage Uncertainty

Country of Residence	2013, % 7	2020, % 8
Countries with national health services
United Kingdom	3	3
Sweden	3	4
Norway	3	2
Countries with single‐payer insurance
Canada	14	17
Australia	15	17
France	17	15
New Zealand	6	4
Countries with multiple insurers
Netherlands	13	9
Switzerland	16	17
Germany	14	13
United States	28	34

The percentage of respondents who reported “insurance denied payment” or they “did not pay as much as expected” of the medical costs that they had incurred during the previous year was evaluated.

A variety of structural factors give rise to these cross‐national differences, including higher rates of enrollment turnover in the United States and less standardization of coverage provisions than in the European countries. ⁹ None of those structural factors changed markedly between 2013 and 2020, yet these markers for coverage misunderstanding increased in the United States while remaining level in other nations. By 2023, Americans’ coverage misunderstandings had become even more common, ¹⁰ contributing to rising medical debt ¹¹ and capturing media attention. ¹²

We refer here to the aggregated effects of experienced and anticipated coverage shortfalls as the impact of “coverage uncertainty” on American health care. Recent surveys reveal that in any given year, 55%‐60% of Americans experience some form of shortfalls in their insurance coverage. ¹⁰ Coverage uncertainty stems from a variety of sources, including open‐ended cost‐sharing requirements that accumulate to unexpectedly large out‐of‐pocket expenses, out‐of‐network charges for providers thought to have been in‐network, and denials of coverage under prior authorization requirements. Well over half of these coverage shortfalls involve insurers denying claims submitted for covered services rather than cost‐sharing provisions. ¹⁰

Drawing on a unique set of surveys fielded three times between 2009 and 2021, we estimate the scope of coverage uncertainty and its impact on access to medical care, care for chronic conditions, and financial insecurity among insured Americans. We show that the aggregate impact of coverage uncertainty on these outcomes affects a far larger share of the American public than the impact of not having insurance at all does. Coverage uncertainty also exacerbates health (care) inequities. To borrow a term from operations research, ¹³ it is this full array of harms and policy impingements that represents the true “shadow price” of coverage uncertainty in the United States—that is, the unmeasured but strikingly negative impact of health insurance coverage on which people cannot rely.

A Framework for Understanding and Assessing Coverage Uncertainty

To appropriately assess this impact, we must explicate how households form expectations regarding insurance coverage, then consider how expectations shape attitudes and behaviors.

The Origins of Coverage Uncertainty

Coverage uncertainty has remained a minor theme in health policy in part because it is challenging to assess. Most fundamentally, enrollees in health insurance plans are unaware of what they do not know. This characterization is not a tautology. Health insurance constitutes a contractual obligation to pay for the costs of future testing, preventive care, and treatments deemed medically necessary. ¹⁴ However, until individuals seek out (or learn from others’ seeking) specific tests, treatments, and services, it is difficult for them to anticipate which expenses are covered, let alone fully understand the criteria or procedural requirements in coverage determinations.

These procedural barriers are more common than consumers might initially expect. Roughly a quarter of all health insurance claims in the United States are initially denied, with considerable variation across insurers and geographic regions. ¹⁵ Denials may involve trivial paperwork errors (e.g., wrong billing codes), disputes over the medical necessity of a test or procedure, use of providers not in a plan's preferred network, or enrollees reaching utilization limits on certain services (most often, mental health or rehabilitation). ¹⁰

These coverage gaps or shortfalls typically come as a surprise to enrollees. ¹⁰ , ¹⁶ , ^{17 ,} Even when insurers strive for transparency, it's simply impossible to coherently describe coverage prospectively given all the discretionary decisions that are contingent on clinical circumstances. ¹² Limitations in coverage become apparent only when different medical needs emerge and patients (and their families) work through this process. Consequently—albeit somewhat ironically—the more experience enrollees have with a health plan, the less confidence they espouse in coverage. ¹⁰ , ¹⁸ New enrollees or those who have remained healthy throughout their enrollment consistently have more positive expectations out of ignorance.

Not all healthy enrollees, however, remain as optimistic about coverage. Some learn about coverage limitations through others’ experiences. ¹⁹ Roughly a quarter of all Americans who experience problems with their health insurance ask a friend or family member for help. ⁶ This generates a social network of learning. Others learn about coverage limitations through mass media²⁰—including evocative descriptions of insurers advertising for new employees to work as “denial nurses”¹⁴—and still others through social media. ²¹ , ²²

In terms of shaping subsequent choices and behaviors, the most consequential experiences related to coverage uncertainty likely evolve in one's own household. These experiences will outweigh those conveyed through social networks or media for several reasons. Unlike in countries with single‐payer insurance, the terms, conditions, and administrative practices shaping coverage in the United States vary greatly among insurers. ¹⁴ , ²³ Consequently, reports of unreliable coverage for others may be discounted when thinking about one's own coverage, unless experiences involve family sharing that coverage. Second, the extent to which coverage uncertainty undermines access or financial security will depend on enrollees’ capacity to appeal coverage denials. Appeals are often successful (30%‐40% of all cases) ¹⁰ , ¹⁵ but are rarely pursued. Fewer than one in 500 coverage denials induce a formal appeal. ⁶ As a result, the contestability of coverage determinations is even less understood by the public. ²⁴

Assessing the Impact of Coverage Uncertainty

Each household's assessment of coverage uncertainty therefore emerges from a complex process of experiential learning and expectations regarding future diagnostic needs, treatments, or screening procedures. For those with more substantial medical needs, these reminders about coverage uncertainty come frequently. Recent survey data reveal that 18% of American households reported that their insurance did “not pay for care you received, that you thought was covered” in the past 12 months. ⁸ Another 27% indicated that insurance paid “less than you expected for a bill you received from a doctor, hospital, or lab.” ¹⁰

The impact on expectations and subsequent care‐seeking behaviors will depend on households’ capacity to buffer these unexpected expenses. In roughly a third of the cases, the added costs are modest (less than $200). However, in 22% of the cases, out‐of‐pocket expenditures exceeded $1,000. ¹⁰ Larger out‐of‐pocket expenses can in themselves deter subsequent health care use, especially if they're associated with accumulating medical debt, which can erode credit ratings and lead to subsequent denials of care by health care providers. ¹¹ Anticipated medical expenses are likely to loom largest for those with chronic conditions requiring regular treatment or clinical monitoring because they face everaccumulating medical expenses. ²⁰

Households with greater financial resources will be more capable of buffering unexpected medical costs. Households that are financially well off can draw on some combination of income, savings, and other forms of fungible wealth. ²⁵ For households with limited incomes, dealing with medically induced financial shocks more often involves deferring payment of other bills and/or borrowing from family and friends. ²⁰

Uncertainty regarding future uncovered expenses can deter access or limit treatment even in the absence of past substantial out‐of‐pocket medical expenses. If households are sufficiently risk averse, the threat of future large expenses may in itself deter health care use, especially in households confused about their coverage limitations. Past research suggests that 30%‐40% of American households find health insurance hard to comprehend, heightening the perceived threat of unexpected medical expenses. ²⁶

Research on risk perceptions also suggests that emotional heuristics (e.g., using the intensity of anxiety or other negative feelings to gauge the magnitude of uncertain threats) can substantially shape behavior. These heuristics have been shown to be particularly powerful in health‐related settings. ²⁷ , ²⁸ Emotional heuristics are most impactful when decision makers have limited time and energy to contemplate uncertain prospects, ²⁹ which are circumstances that are most common in economically disadvantaged households that face many competing demands for their attention. ³⁰

Summing up, the literature suggests that coverage uncertainty and its impact on subsequent attitudes and behaviors are best understood as an interaction of multiple sources of experience, mediated by multiple drivers of expectations. In the analysis that follows, we apply these insights to measure coverage uncertainty in American households and its potential impact. More specifically, we track how coverage uncertainty has changed since the enactment and implementation of the ACA and identify the extent to which it continues to deter access, undermine care, and threaten American's ongoing financial security.

More specifically, in this paper, we test the following four hypotheses:

• Hypothesis 1: Coverage uncertainty reflects the interaction of people's experiences of past coverage shocks and their expectations regarding the risks of similar shortfalls in the future.

• Hypothesis 2: Coverage uncertainty, once it exceeds a threshold of prospective risk (again, reflecting the interaction of past experiences and future expectations), is associated with barriers to the use of needed health care.

• Hypothesis 3: Coverage uncertainty that exceeds thresholds of prospective risk catalyzes multiple emotional reactions, including anxiety and anger induced by future medical expenses.

• Hypothesis 4: Coverage uncertainty is so sufficiently common that it deters needed health care and induces problematic emotional responses among far more Americans who have health insurance than does the absence of insurance in other households.

Methods

This analysis aggregated three surveys of the American public, the first from spring of 2009, the second winter of 2014, and the third from spring of 2021. Data were thus collected under varied macroeconomic conditions. The spring of 2009 was the heart of the Great Recession; when the survey was fielded, unemployment nationwide was 8.7% and rising. By contrast, by 2014, the national unemployment rate had declined to 5.6% and was trending downward. Although the third survey was fielded in the economic shadow of COVID‐19–era lockdowns (when national unemployment peaked at 14.8%), as the survey was in the field, unemployment had already fallen to 6.0% with a continuing downward trajectory.

The first of these surveys was dubbed the Survey of Economic Risk Perceptions and Insecurity (SERPI). It was designed to provide a comprehensive assessment of Americans’ expectations and experiences with economic risks across 15 different domains. ²⁵ The survey also assessed respondents’ experiences in these same 15 domains, as well as a matching set of experiences reported by their social network (e.g., “members of your extended family or close friends not living with you”). Additional questions assessed households’ capacity to buffer economic shocks. This survey was replicated in the winter of 2014 with a smaller sample. The revised version incorporated some additional domains of economic risk. It also added additional measures of respondent's health status not part of the original survey.

The final wave analyzed here, collected a year after the emergence of the COVID‐19 pandemic, was designated the Survey of Health and Economic Risk Perceptions and Insecurity (SHERPI) because it incorporated a wide variety of pandemic‐related risk perceptions and experiences. As with the earlier surveys, the SHERPI collected expectations across 15 domains of economic risk, as well as the experiences of respondents and their social networks in the same 15 core domains. As with the two earlier waves, the survey incorporated measures of respondents’ health status and their households’ ability to buffer economic shocks.

Measures Used in the Analysis

Dependent Variables

Our primary measure to assess the impact of coverage uncertainty on access was drawn from one used extensively in the literature: whether the respondent had been deterred during the previous 12 months from using needed medical care because of concern about costs. More specifically, respondents were asked if they had done either of the following in the past 12 months: (a) “not gone to the doctor because of the cost,” and (b) “avoided getting medical care because you didn't know if your insurance would pay for it.” These questions were asked on all three waves of the survey.

Two additional outcome variables were available on only a single wave of the survey. These expand the array of consequences potentially associated with coverage uncertainty. However, unlike deterred access, trends over time in these associations cannot be assessed.

The first supplementary outcome relates to the household's sense of financial security in the face of potential future medical expenditures. The 2009 version of the survey included a question about whether respondents were feeling anxious about a variety of impending economic risks, including “facing major out‐of‐pocket medical expenses as a result of a serious illness or injury in your immediate family.” Responses were on a five‐point scale, ranging from “not at all anxious” to “extremely anxious.”

The second supplementary outcome was asked only in the final version of the survey. Respondents who reported that someone in their household was dealing with a chronic health problem were asked to assess their capacity to care for that chronic condition. More specifically, respondents to the SHERPI were asked whether during the previous 12 months they “had difficulty caring for a chronic medical condition” for anyone in their household.

Explanatory Variables

The questions used to identify expectational and experiential measures related to coverage uncertainty are at the core of these analyses. Expectations were assessed based on the intensity of worry reported on each survey, measured on a four‐point scale from “not at all worried” to “very worried.” Past research suggests that worry provides a useful metric for expectations related to uncertain future prospects because those who report that they are worried are conveying that it is both on their mind (how frequently they are thinking about the outcome) and that it carries a predominantly negative valence. ²⁵ , ³¹ Because subsequent attitudes and behaviors may respond to either the cognitive or emotional aspects of uncertainty, this duality of “worry” makes it a particularly useful explanatory variable. ³²

Two specific worries are most relevant to our analysis. The first relates most directly to coverage uncertainty, asking about whether the respondent is worried about “getting seriously ill and not being able to figure out what your insurance will pay for.” The second asks about consequences of coverage shortfalls, inquiring about worries related to “having a serious illness or injury in your immediate family that creates major out‐of‐pocket medical expenses.”

Our experiential measures have a comparable dual focus on both coverage and medical expenses. Respondents were asked whether over the previous 12 months they (a) “had problems getting your insurance to pay for major medical expenses,” or (b) “had major out‐of‐pocket medical expenses as the results of a serious illness or injury to you or your immediate family.” To assess learning through social networks, respondents were asked about these same outcomes for “members of your extended family or close friends not living with you.”

Because these surveys were designed to assess the scope and consequences of perceived economic insecurity, they incorporated multiple measures of household economic resources. In addition to household income, these included measures of (a) access to liquid assets (how long the household could meet basic needs if its income was discontinued), (b) equity (the current dollar value of stockholdings), and (c) potential to borrow from one's social network (again, measured in terms of dollar value). This final measure is particularly salient for economically disadvantaged households in which borrowing from family and friends often provides the primary buffer against financial shocks or other unexpected needs. ³³

In order to benchmark the impact of coverage uncertainty on our three outcome variables, we make use of several additional variables. The key benchmarking comparison involves households that were uninsured during the previous year. To control for variation in the reliability of coverage across sources of insurance, respondents were asked to identify if they were covered under public insurance programs (Medicare, Medicaid, or other government‐funded programs) or private insurance (employer based, individually purchased, or—beginning with the 2014 wave of the survey—purchased through an ACA marketplace).

Additional variables are incorporated into regression models to control for other factors that researchers have identified as affecting the magnitude of perceived risks or how people respond to those perceptions. ³⁴ , ³⁵ These include the following:

• Gender: Social psychologists have long recognized that men perceive uncertain prospects to be less risky than do women when objective probabilities for both are comparable. ³⁶ , ³⁷

• Self‐reported race and ethnicity: A similar pattern is evident within each gender: minority men and women perceive risks of threatening events to be higher than do White men and women, even when the actual prevalence is comparable. ³⁷ , ³⁸ , ³⁹

• Health status: As noted above, those with chronic or serious injuries or illnesses will be more likely to have encountered the coverage limits or procedural challenges than will healthy respondents. ¹⁵ , ⁴⁰

• Educational attainment: Respondents with additional education are likely to be more aware of coverage limitations and more likely to anticipate their impact on out‐of‐pocket expenses. ⁴¹ However, they may also be more adept at overcoming coverage limitation by effectively appealing coverage denials. This makes the predicted net impact of education on either risk perceptions or behavioral responses ambiguous.

• Generalized risk aversion: Respondents who are more risk averse than average are likely to have stronger attitudinal and behavioral responses to coverage uncertainty that they have experienced or expect in the future. ¹⁹ , ⁴⁰

Health status was assessed on the 2014 and 2021 surveys by two questions, one asking respondents about any “serious of life‐threatening illness or injuries” during the past 12 months, the second asking about any “chronic conditions that required regular medical care or monitoring” during that same time period. Those questions were not available on the 2009 survey; an alternative metric was substituted based on respondents’ reports of whether the respondent or someone in their household had “lost more than a month from work due to a serious illness or injury.” Generalized risk aversion is measured using the multistage choice question used on the Health and Retirement Survey to assess risk aversion in job choice.

We incorporated one additional variable into the models predicting deterred access using data from the 2021 survey (SHERPI). During the first year of the pandemic, many Americans avoided medical settings to reduce their risk of exposure. ⁴² Through spring 2021 (as vaccines were just rolling out), there was substantial regional variation in the severity of COVID‐19 and thus the extent of health care avoidance. ⁴³ We leveraged these regional differences to control for COVID‐19's impact using the state‐level frequency with which respondents on the monthly Census Pulse Survey reported avoiding medical care because of COVID‐related risks, even if they had some pressing symptoms.

Data Sources

Data for the 2009 and 2014 waves of the survey were collected via the standing internet panel known as the Knowledge Panel, original developed by Knowledge Networks and later operated under the auspices of Growth from Knowledge (GfK; currently owned by the Ipsos corporation). The 2009 data were collected as wave 16 of the 2008–2009 Longitudinal Survey fielded as part of the American National Election Study. A total of 2,612 panelists responded to this wave of the survey. The SERPI was fielded again in 2014 as a freestanding survey with a smaller sample size of 1,020. The SHERPI was fielded in spring 2021 through the AmeriSpeak internet panel maintained by the National Opinion Research Organization. This third wave collected responses from 2,908 panelists. Appendix 1 reports on the sociodemographic characteristics of these samples and demonstrates that all three waves are generally representative of the American public under the different macroeconomic circumstances in these 3 years.

Analytic Methods

Given the multifaceted origins of coverage uncertainty, our analysis of both its prevalence and impact made use of some unconventional analytic methods. Following the logic of our earlier conceptual discussion, we operationalized metrics of coverage uncertainty reflecting the intersection of expectations (e.g., worries about coverage unreliability) and experiences (e.g., past encounters with coverage shortfalls). We anticipated that households that have the most intense worries combined with the most extensive exposure to previous coverage shortfalls would report the most pronounced changes in behavior (e.g., deterred access) and attitudes (e.g., anxiety).

To assess the association between varying levels of coverage uncertainty and our health care–related outcomes, we estimated regression models that included the full array of interaction terms between worries and shortfalls, controlling for the household's capacity to buffer financial shocks and the extent of risk aversion among respondents. For ease of interpretation, we estimated these effects with linear probability models. (In Appendix 2, we show that findings remain consistent with alternative specifications, including logistic regression models.)

In the fourth and final stage of analysis, we “unpacked” the interactions of expectations and experience to assess whether it was coverage uncertainty itself that was associated with changes in health care attitudes and behavioral rather than the impact of the medical expenses that co‐occurred as households became aware of the limitations of the insurance coverage. We pursued this question in two ways. Because our primary models incorporated a full set of interactions between worries and experience, we could identify the distinctive association of reported worries about coverage reliability among households that reported neither experiencing a coverage denial nor unexpected medical expenses.

This offered perhaps the most straightforward test of the distinctive impact of uncertainty. However, it most likely represented a lower‐bound estimate of that impact because experiencing coverage shortfalls may also reveal previously overlooked circumstances associated with coverage unreliability. To take into account these distinctive opportunities for learning, one must deploy an alternative modeling approach.

This involved estimating a two‐stage regression model. In this alternative regression specification, we used coverage shortfalls that occurred in respondents’ social networks to instrument for variation in worries about coverage reliability, controlling for shortfalls that occurred in the respondents’ own household. Because learning through social networks was not accompanied by any additional medical costs, medical debt, or other financial harms that might in itself deter access or exacerbate anxieties, this instrumental variables approach made it possible to distinguish pure expectational effects of worry on our outcome variables.

Findings

The Prevalence of Coverage Uncertainty and Its Sequelae Among Insured Americans

Prevalence of Coverage Uncertainty

Our three waves of survey data allowed us to track the prevalence of both expectations (worries) and experiences (coverage shortfalls; out‐of‐pocket medical spending) over a 12‐year period that was associated with two distinct economic downturns. As Table 2 reveals, changing economic conditions are associated with variation in perceived worries about coverage. However, even under the most benign macroeconomic circumstances, more than a quarter of the insured American public reported high levels of worry about the reliability of their insurance coverage.

Table 2.

Prevalence of Worries Reported About Not Knowing Which Costs Health Insurance Will Cover and Coverage Shortfalls Involving Major Medical Expenses in the Previous 12 Months

Characteristic	Spring 2009, %	Winter 2014, %	Spring 2021, %
Level of worry
Not at all worried	33.3	36.7	31.1
Slightly worried	34.4	36.9	37.6
Fairly worried	18.0	15.4	19.2
Very worried	14.3	11.8	12.1
Total very or fairly worried	32.3	27.2	31.3
Nature of exposure
No coverage shortfalls	73.3	82.9	79.1
Only in own household	4.5	3.8	3.2
Only in own household	14.6	10.2	12.1
Own and other households	7.6	3.1	5.6
Respondents with any shortfall exposure	26.8	17.1	20.9

Data of insured respondents were collected from 2009 to 2021.

Economic circumstances were not the only change over these 12 years. In 2009, prior to the enactment of the ACA, more than 12% of the public (Table 2, row 2 plus row 4) reported experiencing a shortfall in their own coverage that left their household exposed to substantial medical costs. A total of 22% (Table 2, row 3 plus row 4) reported similar experiences in their social networks.

As the regulatory reforms associated with the ACA were implemented, there was a striking decline in prevalence of coverage gaps over the next 5 years (Table 2). Shortfalls in an individual's own coverage dropped under 7% by 2014. Shortfalls in social networks declined to just over 13%.

However, by the end of the decade, this benign trend had reversed. To be sure, the 2021 prevalence of shortfalls remained lower than the prevalence in 2009, suggesting that the ACA reforms had some persisting benefits. This later resurgence of coverage problems, consistent with changes reported on the Commonwealth Fund's surveys (Table 1), suggests the regulatory impact on coverage reliability could not be fully sustained.

Because we anticipated that the impact of coverage uncertainty would reflect the interaction of expectations and experiences, our final step in assessing prevalence combined our measures of worry with reported experiences of coverage shortfalls. To illustrate, we focused on the 2021 data, but similar patterns held for all three waves of the survey.

In Table 3, respondents are stratified into 16 categories based on combinations of experience and expectations. More than a quarter (27.7%) were carefree regarding coverage uncertainty, reporting neither worry about what insurance would cover nor any coverage shortfalls in the previous 12 months. Another 32% expressed slight worries about coverage reliability but, again, no recently experienced coverage shortfalls. That totals to about 60% of the public reporting relatively benign perceptions and experiences regarding their health insurance.

Table 3.

Percentage of American Households Reporting Worry About Health Insurance Coverage and/or Experiences of Coverage Shortfalls in Own Household During the Past 12 Months

Reported Level of Worry About Health Insurance Coverage	No Shortfalls, %	Cost Only Shortfalls, %	Coverage Only Shortfalls, %	Cost and Coverage Shortfalls, %
Not at all worried	27.7	2.6	0.3	0.4
Slightly worried	31.7	4.4	0.8	0.7
Fairly worried	13.9	2.5	1.6	1.3
Very worried	6.9	2.4	0.7	2.1

Data of the insured population were from spring 2021.

However, these benign expectations and experiences with insurance are, in part, a byproduct of most Americans being relatively healthy in any given year and therefore not fully testing the limits of their insurance coverage. Among respondents who had experienced neither a chronic nor serious illness in the past 12 months (not shown in Table 3), more than two‐thirds expressed little or no worry about their coverage reliability. For those who experienced a serious illness, those with few worries about coverage fell to 53%; for those with both chronic and serious illnesses, confidence in coverage declined to 46%.

About 40% of Americans expressed greater concerns about coverage reliability. A total of 20% reported being fairly or very worried about coverage, albeit with no recently experienced shortfalls. The other 20% reported one or more coverage shortfalls, divided evenly between those who expressed worry about coverage and those who did not. To what extent do these subsets of the American public experience sufficiently intense uncertainty that it affects their behaviors or expectations in potentially consequential ways?

Association of Attitudes and Behaviors With Reported Coverage Uncertainty

As noted above, one valued outcome—timely access to medical care—was collected on all three waves of the survey, allowing us to track changes over the 12 years between the first and final wave of the survey. This offers a perspective on secular trends in the impact of coverage uncertainty.

Figure 1 illustrates these trends over time, setting the stage for our multivariate models. In 2009, 7.8% of Americans had no health insurance and reported deferring medical care because of concerns about costs. (This was about half the uninsured population in that year.) In that same year, 17.7% of Americans who had health insurance reported deferring medical care because they were uncertain that their coverage would pay for that care. Another 9.7% of the insured reported that they had deferred medical treatment out of concern regarding potential medical costs. Because these two groups somewhat overlapped, a total of 24.5% of Americans had health insurance but were nonetheless deterred from seeking needed medical care because of concerns about future medical expenses or unreliable insurance coverage.

Figure 1.

Delayed Access to Medical Care Associated with Lack of Insurance and Coverage Uncertainty: Proportion of American Public Reporting Delayed Care Seeking, 2009–2021 [Colour figure can be viewed at wileyonlinelibrary.com]

Over the subsequent 12 years, the proportion of Americans who were uninsured declined as provisions in the ACA were rolled out, expanding Medicaid enrollment and private insurance purchased through state‐level marketplaces. This reduced the proportion of Americans who reported deferred medical care when they were without health insurance to under 5% in 2014 and 4.2% in 2021, as pandemic‐era policies allowed Medicaid recipients to retain enrollment without recertification. ⁴⁴

The impact of the ACA's insurance regulations and consumer assistance was also evident through 2014, as the proportion of households deferring access because of doubts about coverage or concerns about medical costs fell to under 20%. However, the resurgence of worry about coverage reliability and the growing prevalence of coverage shortfalls evident in Tables 1 and 2 were reflected in a modest increase in the deterrent effects of coverage uncertainty on access to care by 2021. In the final wave of the survey in 2021, almost five times as many Americans reported deferring medical care despite having health insurance, compared with those who deferred care because they were uninsured.

Sources of Coverage Uncertainty and Their Associations with Valued Outcomes

We sought to estimate the association of different intensities of coverage uncertainty with deferred access, difficulties caring for chronic conditions, and high levels of anxiety about future medical expenses. Overall, among insured Americans, 15.2% reported deferring access to care because of cost, 16.2% reported difficulties caring for a chronic condition in their household, and 13.9% reported being very or extremely anxious about future medical expenses.

We hypothesized above that uncertainty‐related barriers to timely and effective medical care will reflect a combination of experiences (coverage shocks) and expectations (worries). To test this proposition and to identify threshold effects, we estimated a set of regression models. The predicates of coverage uncertainty were measured by incorporating all the combinations of expectations and experience introduced in Table 3, omitting those who expressed no worry and reported neither coverage shortfalls nor large out‐of‐pocket medical expenses as the control group. The findings relevant to coverage uncertainty are reported in Table 4, the full regression models from which these results were extracted are presented in Appendix 3.

Table 4.

Estimated Association of Worries and Coverage Shocks with Health Care–Related Outcomes: Insured Americans

	Behaviors and Expectations Related to Medical Care
Sources of Coverage Uncertainty	Avoided Medical Care Because of Cost/Coverage Concerns, Spring 2021 a	Difficulty Caring for Chronic Illness, Spring 2021 b	Anxiety About Medical Costs, Spring 2009 c
Mean value among insured reporting no worries or coverage shocks	0.05	0.08	0.13
Not worried
Cost shocks only	0.06	0.17	−0.05
Coverage shocks only	0.22	0.06	0.10
Cost and coverage shocks	0.63	0.16	0.12
Slightly worried
No shocks	0.04	0.01	0.00
Cost shocks only	0.17	0.15	0.07
Coverage shocks only	0.37	0.04	0.04
Cost and coverage shocks	0.51	0.17	0.01
Fairly worried
No shocks	0.16	0.02	0.08
Cost shocks only	0.22	0.19	0.23
Coverage shocks only	0.49	0.14	0.07
Cost and coverage shocks	0.55	0.25	0.29
Very worried
No shocks	0.17	0.07	0.30
Cost shocks only	0.42	0.36	0.50
Coverage shocks only	0.45	0.27	0.28
Cost and coverage shocks	0.68	0.51	0.49

Values were compared with no worries or shocks. All regressions were controlled for household income, savings, value of stock holdings, ability to borrow from social network, whether the respondent had a serious illness or a chronic condition in past year, source of insurance (Medicare, Medicaid, other public, individually purchased policy), gender, race and ethnicity, educational attainment, and risk aversion. Regressions using data from the 2021 Survey of Health and Economic Risk Perceptions and Insecurity also controlled for state‐level differences in reported avoidance of medical care out of concern for COVID‐19–related risks. Bolded results are statistically significant at a p <0.05 level.

^a During the past 12 months, respondents did not seek needed medical care because of concern about potential medical expenses or perceived unreliability of insurance coverage.

^b During the past 12 months, respondents had difficulty caring for a chronic condition among some member of their household.

^c During the past 12 months, the respondents felt very or extremely anxious when thinking about potential medical expenses.

Consider first the association between the predicates of coverage uncertainty and reports of deferred access to medical care (left‐hand column of results). Exhibit E reported these associations for the 2021 survey responses, but the pattern of findings reported there was replicated by models estimated using the 2014 and 2009 waves of the survey. The overall pattern for deferred access was clear: experiences and expectations interacted such that the strongest predictors of deferred access involved (a) combinations of coverage and cost shocks, or (b) either of these shocks occurred in households in which concern about reliable coverage was already pronounced (e.g., fairly or very worried).

All told, as noted above, these deterrent effects on access were evident in just under a quarter of households in the United States, or about 27% of the insured population. There appeared to be a threshold effect with respect to worry—in the absence of recent coverage or medical cost shocks, worry about coverage uncertainty predicted deferred access only for respondents who reported being fairly or very worried. Coverage shocks had a stronger association with deterred access than cost shocks did.

Benchmarking the magnitude of these associations to deterred access reported by uninsured households helped to put these findings in perspective. To do this, we reestimated the regression models, this time including the uninsured in the analytic sample and comparing their deferred access with that reported by households with employer‐based insurance, again controlling statistically for socioeconomic status, demographics, and risk aversion. These estimates allowed us to compare the elevated rates of deferred access for uninsured households with increases in deferred access among insured households with varied intensity of coverage uncertainty.

Compared with households with employer‐based coverage and controlling for other household attributes, those without health insurance in 2021 were 36% more likely to report that they deferred access to needed medical care (Table 5, top row). This decrease in access among the uninsured was exceeded in magnitude for eight of the 15 subgroups that reported coverage uncertainty. Those for whom coverage uncertainty was associated with larger declines in access totaled 10% of the American public in 2021, roughly 12% of all insured households.

Table 5.

Estimated Association of Source of Insurance for Health Care–Related Outcomes: All Households

	Behaviors and Expectations Related to Medical Care
Sources of Coverage	Avoided Medical Care Because of Cost/Coverage Concerns, Spring 2021 a	Difficulty Caring for Chronic Illness, Spring 2021 b	Anxiety About Medical Costs, Spring 2009 c
Mean value among insured reporting no worries or coverage shocks	0.05	0.08	0.13
Uninsured	0.36	0.12	0.13
Private insurance
Individually purchased	0.05	−0.01	0.05
Purchased in ACA marketplace	0.12	0.05	NA
Public insurance
Medicare	−0.08	0.02	−0.03
Medicaid	−0.03	0.04	−0.01
Other government programs	−0.04	0.04	0.06

Values were compared with employer‐sponsored coverage. All regressions were controlled for household income, savings, value of stock holdings, ability to borrow from social network, whether the respondent had a serious illness or a chronic condition in past year, source of insurance (Medicare, Medicaid, other public, individually purchased policy), gender, race and ethnicity, educational attainment, and risk aversion. Regressions using data from the 2021 Survey of Health and Economic Risk Perceptions and Insecurity also controlled for state‐level differences in reported avoidance of medical care out of concern for COVID‐19–related risks. Bolded results are statistically significant at a p <0.05 level. ACA, Affordable Care Act; NA, not applicable.

^a During the past 12 months, the respondents did not seek needed medical care because of concern about potential medical expenses or perceived unreliability of insurance coverage.

^b During the past 12 months, the respondents had difficulty caring for a chronic condition among some member of their household.

^c During the past 12 months, the respondents felt very or extremely anxious when thinking about potential medical expenses.

A comparable pattern emerged for problems reported in caring for chronic conditions (middle column of results in Table 4) as well as reported high levels of anxiety about future medical expenses (right‐hand column in Table 4). For each of these outcomes, the associated probability of a harmful outcome was progressively higher for those who experienced a combination of more intense worry about coverage unreliability and one or more of the experienced shocks. Indeed, experience of coverage shortfalls predicted heightened anxiety only for those who reported being fairly or very worried about future coverage. Care for chronic conditions appeared at risk of disruption even when prevailing levels of worry were less intense but only when the past coverage shortfalls involved experiences with large out‐of‐pocket medical expenses. Anxieties also appeared to be more sensitive to cost shocks than to coverage shortfalls, a distinct contrast with the findings for deterred access to care.

Again, it was useful to benchmark our estimates against comparable associations for uninsured households. Not having health insurance was associated with a 12% increase in the proportion of households reporting problems caring for chronic illness (Table 5, middle column), compared with households with employer‐based insurance. That was matched or exceeded by nine subgroups of insured households reporting coverage uncertainty. For this outcome, in contrast to deferred access to care, worry about coverage in the absence of experienced shortfalls had smaller associations with difficulties caring for chronic conditions. However, for worried households that had experienced shocks, the association with worsened care for chronic illness was two to four times larger than that reported by uninsured households.

In 2009, respondents from households without insurance reported elevated levels of anxiety about future medical expenses, a 13% higher prevalence of intense anxiety compared with households with employer‐based insurance. This increase was exceeded by half a dozen subgroups of the insured reporting coverage uncertainty. These were, again, concentrated among households that were fairly or very worried about not understanding their health insurance coverage. Additionally, again, these associations were two to four times larger than the estimated association for the uninsured. The subsets of the insured reporting larger increases in anxiety than those having no health insurance at all totaled 15% of all insured households.

Some readers may be surprised that some combinations of expectations (worries) and experiences (coverage shocks) among insured respondents are estimated to produce a larger deterrent effect on the use of medical care or induce more anxiety about out‐of‐pocket medical expenses than not having any health insurance at all. After all, some coverage, even if unreliable, would seem to offer greater financial security than no insurance coverage at all. However, this interpretation fails to account for the impact of negative feelings when the coverage shortfall comes as a surprise. When anticipated economic security is unexpectedly supplanted by doubts about the future, this can induce a persisting negative emotional penumbra related to both health insurance and medical spending. That combination can pose a powerful deterrent to future health care use as well as a lasting sense of heightened anxiety.

How Is Coverage Uncertainty Associated with Problematic Outcomes?

The findings reported above suggest that coverage uncertainty is associated with reduced access to care, reductions in effective care for chronic conditions, and heightened anxieties, at least for those with the most intensive worry about coverage. Precisely because people learn about the limitations of their health insurance when they find themselves with uncovered expenses, it is easy to conflate the impact of uncertainty with the financial impact of unpaid medical bills. ¹⁰ , ¹¹ Using our unique data set, we can unpack these associations in two ways.

Households Experiencing No Coverage Shortfalls

Concerning the interactions presented in Table 4, from these results, we know that respondents who report themselves to be fairly or very worried about unreliable coverage report significantly more deterred access, problems with chronic conditions, and anxiety about medical costs. These patterns hold (albeit at lesser magnitude) for respondents who experienced no shocks and or experienced only coverage shocks. For delays in access, for example, pooling together subgroups suggests that 23% of the public experienced delays associated with coverage uncertainty, even though they had experienced no major recent out‐of‐pocket medical expenses.

Two‐Stage Regression Models

Our instrumental variable regressions used shortfalls experienced in social networks to parse out the associations related to coverage uncertainty per se. We started by reestimating the models predicting our three outcomes (deterred access to care, problems managing chronic conditions, and anxiety about medical expenses) using a simplified specification that replaces all the interaction terms from Table 4 with three independent variables: (a) intensity of worry about not being able to anticipate coverage, (b) having experienced in one's own household insurance not paying for an expected major expense, and (c) having experienced in one's household substantial out‐of‐pocket medical costs. Our focus for this next stage of the analyses was on the independent association of worry with our outcome variables, controlling for the two experienced coverage shortfalls.

The leftmost column in Table 6 presents the predicted coefficient on the worry variable from a single‐stage model, controlling for shocks and for the same set of additional explanatory variables as the models presented in Tables 4 and 5. Not surprisingly, given our prior findings, worry about coverage shows a statistically significant association with all three outcome variables—the largest with anxiety, the smallest with care for chronic conditions.

Table 6.

Estimated Association of Worries about Coverage with Health Care–Related Outcomes, Controlling for Coverage Shocks: Insured Americans

	Single‐Stage Linear Probability Model		Two‐Stage Least Squares, Worry Instrumented by Social Network Shocks
Outcome Measures	Coef.	SE	Coef.	SE	F statistic
Avoided medical care because of cost or coverage concerns a	0.068	0.008	0.475	0.083	18.70
Difficulty caring for chronic illness b	0.034	0.007	0.442	0.080	18.73
Anxiety about future medical expenses c	0.093	0.008	0.106	0.053	15.53

All regressions were controlled for experience of coverage shortfall or major out‐of‐pocket medical expenses in the respondent's household during the past 12 months, household income, savings, value of stock holdings, ability to borrow from social network, whether the respondent had a serious illness or a chronic condition in the past year, source of insurance (Medicare, Medicaid, other public, individually purchased policy), gender, race and ethnicity, educational attainment, and risk aversion. Regressions relying on data from the 2021 Survey of Health and Economic Risk Perceptions and Insecurity were also controlled for state‐level differences in reported avoidance of medical care out of concern for COVID‐19–related risks. Bolded results are statistically significant at a p <0.05 level. Coef., coefficient; SE, standard error.

^a During the past 12 months, the respondents did not seek needed medical care because of concern about potential medical expenses or perceived unreliability of insurance coverage.

^b During the past 12 months, the respondents had difficulty caring for a chronic condition among some member of their household.

^c During the past 12 months, the respondents felt very or extremely anxious when thinking about potential medical expenses.

We then estimated three matching two‐stage models using the shortfall experiences from respondents’ social networks to instrument for learning through others’ experiences with health insurance. We constructed measures of three network shocks. Respondents were asked whether members of their extended family or close friends had, in the previous 12 months, experienced the following: (a) major out‐of‐pocket medical expenses as a result of serious illness or injury in their immediate family, or (b) problems getting their insurance to pay for major medical expenses. We constructed dichotomous variables for three mutually exclusive subgroups: those with (a) only major out‐of‐pocket expenses, (b) only coverage problems, and (c) both out‐of‐pocket costs and coverage shortfalls. Those with no shocks were the comparison group.

As the second column in Table 6 reveals, once we instrumented for social network shocks in this manner, we obtained positive and statistically significant point estimates of the association of the portion of coverage uncertainty that is predicted by social network experiences on avoidance of medical care (row 1), difficulty caring for chronic conditions (row 2), and anxiety about medical costs (row 3). Note how these two‐stage estimates led to a different ordering of how sensitive these outcomes were to coverage uncertainty. For the single‐stage model, anxiety about medical expenses demonstrated the strongest associations, difficulty caring for chronic conditions being the weakest (though still statistically significant). By contrast, the two‐stage estimates suggested that anxiety was the least strongly related to worry about coverage reliability, with the other two outcomes both being more sensitive and equally responsive to coverage uncertainty. For all three outcomes, the two‐stage estimates were statistically significant, controlling for respondents’ own household coverage and medical cost shocks.

Discussion

Our findings suggest that coverage uncertainty is associated with deterred access, less efficacious care for chronic conditions, and heightened anxiety about medical expenses. For many insured households, these harmful sequelae are comparable or larger in magnitude than being uninsured. Moreover, coverage uncertainty affects many more American households than those affected by going without health insurance.

These persisting harms associated with limited coverage reliability do not diminish the other accomplishments of the ACA, which expanded insurance enrollment and initially reduced coverage uncertainty. However, the latter benefits were not fully sustained, eroding between 2014 and 2021. Among households with employer‐based insurance—the gold standard of coverage in the United States—coverage shortfalls were reported to almost double, growing from 5.6% in 2014 to 9.1% in 2021.

By 2021, one in five American households reported that coverage uncertainty had deterred access to care in the previous year. Although expanding and stabilizing insurance enrollment remains an important ongoing task, ⁴⁵ we believe that there is a strong case for returning the focus of health policy to also enhancing the reliability of coverage. The Biden administration moved in this direction, proposing to eliminate “junk insurance” from the market. ⁴⁵

Though well‐intended, the Obama administration already discovered the political perils of this approach. “Junk” insurance is cheap insurance—precisely because it pays out so little in benefits. As long as the enrolled households remain healthy, the unreliability of coverage remains invisible to them. Taking such policies off the market therefore largely engenders anger from those affected, seemingly depriving these purchasers of affordable policies that they have yet to discover was never really a good deal because it was never reliable coverage.

Moreover, the harms of coverage uncertainty have never been primarily a consequence of a few “bad apple” policies or insurance companies. This is evident from the cross‐national comparisons in Table 1. Consumers in countries with single‐payer insurance or more tightly regulated insurance markets still report substantial coverage uncertainty. The prevalence of coverage uncertainty is higher in the United States, elevated by the indirect impact of policies that are sensible in their own right but inevitably erode coverage reliability.

Policies That Indirectly Foster Coverage Uncertainty

Two policy approaches illustrate these spillover effects. The first involves value‐based purchasing. ⁴⁶ , ⁴⁷ In particular, the versions of that reform that adjust coverage to either encourage the use of treatment options that have higher average effectiveness or that increase consumer cost sharing to discourage use of lower‐value procedures. ⁴⁸ Favored by many policy advocates in the United States, public support has been more checkered. ⁴⁹ Although the origins of these public doubts have never been fully explored, one likely explanation is that value‐based coverage makes it difficult for patients and their families to reliably anticipate which treatments will be fully covered, a consideration that would most erode support among those with chronic conditions that portend a lifetime of future treatment choices. ⁵⁰

A second set of policies, more enthusiastically received by the American public, promotes consumer choice among health insurers. Policies promoting choice in employer‐based insurance date back to the 1980s. ⁵¹ More recently, advocates have shifted to encouraging choice within public programs such as Medicare and Medicaid. ⁵² Given Americans’ long‐standing distrust of their health insurers, the popularity of choice‐promoting policies is not surprising. However, choice is often extolled with little recognition that the act of switching among insurers exacerbates coverage uncertainty. Each time a household switches, those who are newly enrolled must learn each insurer's distinctive administrative protocols to obtain prior authorization for coverage or to effectively deal with coverage disputes. ¹⁴ , ¹⁵ , ⁵³ The more frequent the switching, the less complete or sustainable that learning will be.

Both value‐based coverage and choice‐promoting policies have compelling rationales—to a point. Nonetheless, if the shadow price of coverage uncertainty remains ill‐defined and largely ignored, ⁵⁴ policymakers are certain to pursue both these strategies in ways that fail to attend to their full consequences. Given the documented extent to which coverage uncertainty deters access, undermines care for chronic conditions, and exacerbates anxiety, the price being paid may be quite high, even if it is one not widely recognized.

Addressing the Impact of Coverage Uncertainty

Continuing to overlook the consequences of coverage uncertainty most harms minority populations who have been historically burdened by the impact of uneven insurance enrollment. ⁵⁵ , ⁵⁶ Because Black and Hispanic households have less accumulated wealth to buffer financial shocks, even if all Americans faced equally (un)reliable insurance coverage, one would expect these households to report elevated levels of worry about coverage uncertainty, ⁵⁷ and they do (Table 7). However, uneven financial reserves are not the full story because these same respondents also report disparities in experienced coverage shortfalls. The origins of these differences remain unclear because racial/ethnic disparities appear for each major source of insurance: employer based, individually purchased, Medicaid, and Medicare (see partial results in Appendix 4).

Table 7.

Racial/Ethnic Disparities for Coverage Uncertainty, Spring 2021: Among All Insured Americans

Self‐Identified Race/Ethnicity	Heightened Worry About Coverage, % a	Experienced Coverage Shortfall in Own Household, % b
White, non‐Hispanic	29.4	7.2
Black, non‐Hispanic	33.2	10.5
Hispanic (all races)	47.7	13.7

^a The percentage of respondents reporting they are fairly or very worried about not knowing if their health insurance will pay for a major medical expense was evaluated.

^b The percentage of respondents reporting they had “problems getting their insurance to pay for major medical expenses” during the past 12 months was evaluated.

To address these persisting inequities, policymakers must address the origins of coverage uncertainty itself. Three policy approaches have emerged, varying in effectiveness, for addressing reliability of health insurance coverage. The first involves reducing complexity by standardizing certain aspects of coverage. Although it is a mainstay reform in other countries relying on private insurance markets, the only notable American application involved federal reforms in 1990 to supplementary Medicare (a.k.a. “Medigap”) policies. Coverage standardization was effective at enhancing consumer understanding in this context ⁵⁸ but was designed to promote better informed purchasing, not reliable anticipation of subsequent coverage. Comparable approaches would have limited relevance for reducing uncertainty related to the most common sources of coverage denials documented in contemporary surveys, which have less to do with the scope of coverage than in its administration, particularly the determinations of medically necessary care. ⁶ , ¹⁰ , ¹³

A second approach strives to better prepare consumers to deal with coverage complexity. These interventions were motivated by evidence that many Americans have trouble comprehending the basic workings and lexicon of health insurance, motivating calls for remedial instruction to promote “health insurance literacy.” ⁵⁹ Research has established that people with a more limited understanding of insurance terminology used health services less and experienced heightened anxieties about medical costs. ⁴⁰ , ⁶⁰ However, these studies failed to control for the correlation between insurance literacy and socioeconomic status, making it impossible to discern whether limited insurance literacy or respondents’ economic vulnerabilities were the real source of reduced access or heightened anxieties.

If a lack of comprehension were the root cause of coverage uncertainty, one would expect to see a strong correlation between reported levels of coverage uncertainty and respondents’ educational attainment because past studies document a close correlation between education and familiarity with insurance terminology. ⁵⁸ However, the regression models reported above, which better control for socioeconomic status than previous research on health literacy had, suggest that there is at best a very modest educational gradient related to either perceived worry or experienced coverage shortfalls (Table 8). Respondents with graduate school degrees report just as much worry about coverage uncertainty as those who had never completed high school. It thus seems unlikely that reducing knowledge deficits offers much leverage for effectively addressing reliability of coverage.

Table 8.

Relationship of Educational Attainment to Reported Metrics for Coverage Uncertainty: Spring 2021

Educational Attainment	Prevalence in Sample Making Sense of Complexity, %	Coverage Shortfalls in the Past Year, % a , b	Fairly or Very Worried About Coverage, % b
Graduate school degree	13	9.9	28
College graduate	19	8.5	26
Completed some college/Assoc. degree	46	11.1	27
High school graduate	19	11.6	30
Did not complete high school	3.3	12.5	29

Regressions were controlled for household income, savings, ability to borrow from social network, whether the respondent had a serious illness or a chronic condition in past year, the respondent's risk aversion, and source of insurance coverage. Bolded values are statistically significant different from the sample mean at a p <0.05 level. Assoc., associate.

^a Whether the respondent reported having had “problems getting their insurance to pay for major medical expenses” during the past 12 months was evaluated.

^b Values are regression‐adjusted means.

The third strategy, by contrast, focuses on more structural reforms. This brings us back to the provisions of the ACA as enacted in 2010 or, perhaps more accurately, to the legislation's original name: the Patient Protection and Affordable Care Act (PPACA). As the insurance‐focused interventions related to the first part of the name rolled out during the initial years of implementation, ⁶¹ the proportion of American households reporting experienced coverage shortfalls in the prior year fell from 12.1% in 2009 to 6.9% in 2014 (Table 2). Although this reduction was not sustained over time, that is not surprising because the reforms protecting consumers lost their federal funding after 2012. Once involving 35 state‐operated programs assisting families with coverage disputes, extant programs have declined to about a dozen, now handling only about 3% of all coverage disputes nationwide. ⁶²

Engagement with external assistance can increase the proportion of coverage denials that are appealed and enhance the odds that coverage disputes are resolved in consumers’ favor, reducing coverage uncertainty for those involved in the appeals. ²⁴ More frequent and successful appeals also have the potential to foster insurer practices that are more predictable for all Americans by increasing the cost to insurers who deploy widespread coverage denials. ¹⁰ Whether a more lasting federal commitment to consumer assistance can be sustained is less clear because regulatory interventions often struggle to generate sustained public awareness and legitimacy. ⁶² However, recent surveys suggest that up to 75% of Americans would take advantage of a CAP if it were available to them, ⁶ making this policy option perhaps the most promising avenue for addressing coverage reliability.

Could a revival of aspirations for nationwide consumer assistance substantially reduce the pernicious effects of coverage uncertainty? We cannot fully test this proposition here, but we can offer a potentially helpful starting point. To do this, we compared the prevalence of coverage shocks reported by respondents living in jurisdictions with active CAPs like those envisioned in the PPACA (District of Columbia and ten states: California, Connecticut, Massachusetts, Maryland, Maine, Mississippi, North Carolina, New Mexico, New York, and Vermont) to reports from respondents in other states, controlling for other respondent characteristics. Of the 2021 sample, 19% lived in jurisdictions with active consumer assistance. We assess the prevalence of coverage shortfalls for major medical procedures, the focus of consumer assistance in the PPACA (Figure 2).

Figure 2.

Association of Consumer Assistance Programs With Reported Coverage Shortfalls: Fall 2021 [Colour figure can be viewed at wileyonlinelibrary.com]

Those living in states with active consumer assistance reported less frequent coverage shortfalls in their own households (7.1% vs. 9.1%, controlling for other household attributes) and fewer coverage shortfalls reported by friends and family in their social networks (11.5% vs. 16.2). The latter difference is statistically significant at a 0.01 confidence level; the former difference is not. Although one cannot readily generalize from these findings because the states with remaining CAPs differ in multiple ways from other jurisdictions, these findings suggest that expanding CAPs to address coverage uncertainty may prove a promising strategy.

All this brings us back to the streets of New York City in December 2024 and the firestorm of anger conveyed through social media regarding the claims review processes at UnitedHealthcare. To explore the linkage between coverage uncertainty and these angry reactions, we return to the 2009 survey wave, in which we assessed respondents’ emotional responses to out‐of‐pocket medical expenses. At that time, just over 11% of respondents reported being “very” or “extremely” angry about these medical expenses. Using this cutoff to define angry respondents, we can identify the associations with worry about coverage reliability and experienced coverage shocks (Figure 3).

Figure 3.

Association of Coverage Uncertainty with Heightened Anger* About Medical Expenses: Spring 2009 [Colour figure can be viewed at wileyonlinelibrary.com]

Findings from regression models that controlled for source of health insurance, household income, household savings, capacity to borrow from social network, value of investments health status, gender, education, race and ethnicity, and risk aversion. OOP, out‐of‐pocket.

*The percentage of respondents reporting feeling very or extremely angry about out‐of‐pocket medical expenses was evaluated.

Our starting point for comparison comes from respondents with employer‐based coverage who had experienced neither worry nor coverage shocks. Among these respondents, in 2009, only 3% reported being angry about out‐of‐pocket medical expenses. By contrast, for those very worried about coverage reliability, prevailing anger jumped to 25%, almost as common as the anger reported by the uninsured (27%). Recent coverage or cost shocks added another 11%‐12% to the count of the enraged. In short, coverage uncertainty is strongly associated with—possibly a primary driver of—the angry coda that accompanied Brian Thompson's death. When it emerged in social media in late 2024, this widespread anger was seen as a media‐worthy surprise, which in of itself is deeply troubling. If it persists, such anger seems likely to further erode the public's increasingly frail trust in American health care.

Methodological Limitations and Questions for Additional Research

The findings in this paper and the implications derived from them need to be considered in light of some methodological limitations that also point the way toward promising future research. First, our measures of both worry and experienced coverage shortfalls are based on episodes involving major medical expenses. However, if one compares the prevalence of coverage events reported on the Commonwealth Fund surveys (Table 1) with those reported on our surveys (Table 2), it becomes clear that there are many coverage events that do not involve major expenditures. How much these smaller expenses—especially if they accumulate over time for households with members who have chronic health problems—might matter for subsequent consumer expectations and behavior should be more carefully studied.

A second set of unanswered questions involves the origins of racial and ethnic differences in the prevalence of experienced coverage shortfalls. One possible explanation might involve differential access to clinicians willing to advocate in coverage disputes for their patients. If patients with particular health needs, ethnic or racial backgrounds, or other identifiable attributes are less able to obtain effective clinician support, this would exacerbate unreliable coverage and the impact of coverage uncertainty. ⁵² How best to address remains unclear and certainly would benefit from additional study. ⁸ , ²³

Third, as with most research that examines issues that have received relatively little prior study, our exploration of coverage uncertainty involves some measurement choices that could have been operationalized in other ways. For example, when assessing the impact of coverage uncertainty on access to care, we combined together respondents who had reported delaying access based on concerns about costs with those concerned about coverage. This is consistent with our discussion about how people learn from past experiences with insurance. Readers might wonder how our results would have differed had our access analyses separated these two sources of delay. To assess how this might have altered our results, we reestimated the models, separating these two outcomes (Appendix 2). These findings suggest that however one parses access, coverage uncertainty is associated with substantial reductions in access to care.

Fourth, our measures of both coverage uncertainty and health care–related outcomes are based on respondents’ self‐reports. This approach has advantages when people's perceptions of coverage reliability loom large in their behavioral or attitudinal responses. There may also be some identifiable aspects of plan practices and review processes that affect the reliability of uncertain coverage. These could be measured in other ways. Recently, data are becoming available on denial and appeal rates across health plans, at least for selected markets. ⁶³ How much access consumers have to this information and how they might interpret it remains unclear. Future studies of coverage uncertainty might combine information on enrollee perceptions with such metrics of health insurer practices.

Finally, our examination of value‐based cost‐sharing reforms makes clear that even with transparency of health insurance practices, coverage will never be entirely predictable because the extent of cost sharing will always depend on the particular health needs and treatments that clinicians prescribe. Can effective value‐based strategies be pursued that do not disrupt patients’ capacity to anticipate when their insurance coverage can be relied on? Addressing low‐value care in ways that safeguard against increased coverage uncertainty would potentially enhance public support for value‐based purchasing interventions.

Conclusion

As cross‐national comparisons reveal, payment for some medical expenses will remain unpredictable under any financing system. That said, when it comes to unexpected out‐of‐pocket costs for consumers, the United States is clearly an outlier—in ways that cost Americans dearly in more than just financial terms. With the United States being the one nation where the prevalence of unexpected coverage shortfalls has increased over the past decade, this concern has become more pressing and more politically salient over time.

In short, Americans have a clear and growing problem with coverage uncertainty, albeit one that often goes unrecognized because its prevalence is rarely measured and its consequences are challenging to discern. At a time when Americans are accumulating staggering amounts of medical debt, ⁶⁴ it is easy to attribute impaired access and inequities in timely treatment to these experienced expenses when in fact it may well be the unpredictability of future spending that is the primary source of pernicious outcomes. When growing numbers of Americans report being angry about medical costs, it is easy to presume that this reflects the burdens of unpaid medical bills when in fact their reactions may have more to do with a lack of support in dealing with coverage reviews and denials.

The saving grace, to the extent that one exists, is that the levels of coverage uncertainty that now prevail in the United States can certainly be reduced. Residents of other countries that rely heavily on competing private insurers report far less frequent experiences with coverage denials. When the Obama administration's reforms were still known as the PPACA, the multiple provisions targeting insurance practices reduced the prevalence of coverage shortfalls almost in half by 2014, compared with levels prevailing in 2009. By returning to these structural reforms, built into the PPACA but since largely abandoned, coverage shortfalls can almost assuredly be reduced. There are still a dozen states with operative CAPs. Future national policy can learn from and build on this infrastructure, assuring that all Americans can rely on these supportive policies when navigating their insurers’ coverage practices. No amount of external assistance will ever make health insurance coverage completely reliable. However, these forms of support can nonetheless moderate the price of persisting uncertainty.

Conflict of Interest Disclosures

No disclosures were reported.

Appendix 1.

Benchmarking Survey Samples Against Matching CPS Data: Spring 2021, Winter 2014, Spring 2009

	2021		2014		2009
Sociodemographic Attributes	Survey*	March 2020 CPS	Survey*	March 2015 CPS	Survey*	March 2009 CPS
Household Income
Less than $35,00	15.1%	16.0%	17.9%	22.9%	13.7%	26.4%
$35,000 ‐ $59,999	23.3%	20.7%	22.5%	19.9%	28.3%	21.4%
$60,000 ‐$99,999	22.7%	22.5%	32.4%	23.9%	37.7%	24.4%
$100,000 and above	41.4%	40.7%	27.3%	29.5%	20.3%	24.3%
Respondent Age
18 – 34	28.9%	29.3%	29.1%	29.9%	24.9%	30.3%
35 – 49	24.5%	24.2%	24.6%	24.6%	31.7%	27.3%
50 – 64	24.9%	24.5%	26.7%	25.1%	25.9%	23.9%
65 and above	21.7%	22.0%	19.5%	19.0%	18.5%	16.7%
Race/Ethnicity
White, Non‐Hispanic	63.2%	62.5%	65.5%	65.7%	78.0%	66.4%
Black, Non‐Hispanic	11.0%	12.1%	11.6%	13.0%	11.5%	12.7%
Hispanic	16.4%	16.8%	15.2%	15.6%	4.8%	15.9%
Educational Status
High School or Less	36.7%	37.9%	29.7%	37.9%	41.2%	54.4%
Some College/Assoc Degree	28.0%	27.1%	28.7%	34.5%	30.2%	26.4%
Completed College	15.5%	22.2%	17.1%	17.6%	19.2%	12.4%
Completed Graduate School	17.9%	12.8%	13.1%	10.1%	9.4%	6.8%
Household Ownership
Owner Occupied	66.5%		71.3%	‐	77.4%	‐
Renter Occupied	31.2%	‐	25.7%	‐	14.4%	‐
Occupied w/o payment	2.3%	‐	3.1%	‐	8.2%	‐
Marital Status
Married	52.6%	51.8%	53.6%	49.2%	65.3%	51.0%
Single	47.4%	48.2%	46.4%	50.8%	34.7%	49.0%
Sex
Male	48.7%	48.5%	48.2%	48.6%	47.4%	48.7%
Female	51.3%	51.5%	51.8%	51.4%	52.7%	51.3%

*Weighted distributions

Appendix 2.

Estimated Association of Worries and Coverage Shortfalls on Deterred Medical Access: Alternative Specifications of Models and Dependent Variables

		Alternative Specifications
Sources of Coverage Uncertainty	Avoid Medical Care (Cost or Coverage)	Avoid Medical Care (Cost Only)	Avoid Medical Care (Coverage Only)	Logistic Regression
	Baseline: Exhibit E			(Odds Ratios)
	Mean Value of Outcomes
		0.20	0.15	0.16
Interactions: Worries + Shocks
[compared with no worries or shocks]
Not Worried: Costs Only	0.06	0.05	0.03	2.0
Not Worried: Coverage Only	0.22	0.24	0.18	8.7
Not Worried: Costs + Coverage	0.63	0.43	0.46	13.6
Slightly Worried: No Shocks	0.04	0.01	0.03	1.3
Slightly Worried: Costs Only	0.17	0.12	0.19	3.8
Slightly Worried: Coverage Only	0.37	0.24	0.37	6.0
Slightly Worried: Cost + Coverage	0.51	0.41	0.50	15.5
Fairly Worried: No Shocks	0.16	0.10	0.14	3.2
Fairly Worried: Costs Only	0.22	0.15	0.24	4.7
Fairly Worried: Coverage Only	0.49	0.39	0.47	15.3
Fairly Worried: Cost + Coverage	0.55	0.55	0.58	21.0
Very Worried: No Shocks	0.17	0.08	0.18	2.6
Very Worried: Costs Only	0.42	0.41	0.35	14.1
Very Worried: Coverage Only	0.45	0.21	0.51	4.1
Very Worried: Cost + Coverage	0.68	0.62	0.70	31.5
Number of Observations	2,478	2,478	2,478	2,478

# Didn't seek needed medical care because of concern about medical expenses or coverage

Note: Regressions control for household income, savings, value of stock holdings, ability to borrow from social network, whether respondent had a serious illness or a chronic condition in past year requiring regular treatment or monitoring, gender, race and ethnicity, risk aversion.

Bolded results are statistically significant at a p <0.05 level

Appendix 3.

Complete Models Incorporating Interaction of Expectations and Experience: Linear Probability Models

	Association with Access, Treatment and Anxiety
Sources of Coverage Uncertainty	Avoided Medical Care Due to Cost or Coverage*		Difficulty Caring for Chronic Illness#		Anxiety about Medical Costs##
	Coef.	SE	Coef.	SE	Coef.	SE
Not Worried: Costs Only	0.06	0.06	0.17	0.06	−0.05	0.05
Not Worried: Coverage Only	0.22	0.09	0.06	0.08	0.10	0.08
Not Worried: Costs + Coverage	0.63	0.12	0.16	0.11	0.12	0.10
Slightly Worried: No Shocks	0.04	0.02	0.01	0.02	0.01	0.02
Slightly Worried: Costs Only	0.17	0.04	0.15	0.04	0.07	0.04
Slightly Worried: Coverage Only	0.37	0.07	0.04	0.06	0.04	0.06
Slightly Worried: Cost + Coverage	0.51	0.06	0.17	0.06	0.01	0.05
Fairly Worried: No Shocks	0.16	0.02	0.02	0.02	0.08	0.02
Fairly Worried: Costs Only	0.22	0.05	0.19	0.05	0.23	0.05
Fairly Worried: Coverage Only	0.49	0.07	0.14	0.06	0.07	0.05
Fairly Worried: Cost + Coverage	0.55	0.07	0.25	0.06	0.29	0.06
Very Worried: No Shocks	0.17	0.03	0.07	0.03	0.30	0.03
Very Worried: Costs Only	0.42	0.06	0.36	0.05	0.50	0.06
Very Worried: Coverage Only	0.45	0.07	0.27	0.06	0.28	0.06
Very Worried: Cost + Coverage	0.68	0.05	0.51	0.05	0.49	0.05
[Employer‐based Omitted]
Medicare	−0.04	0.02	0.05	0.02	−0.04	0.02
Medicaid	−0.01	0.03	0.06	0.02	−0.04	0.04
Other Public	−0.01	0.04	0.06	0.03	0.06	0.04
Individually Purchase	0.03	0.03	−0.02	0.03	0.02	0.02
Qualified Health Plan	0.11	0.04	0.04	0.03	NA	NA
Household Income	−0.01	0.01	−0.01	0.01	−0.01	0.01
Financial Reserves (Savings)	−0.04	0.01	−0.03	0.01	−0.02	0.01
Informal Borrowing Potential	−0.01	0.01	−0.01	0.01	−0.01	0.01
Value of Equity Holdings	−0.01	0.01	−0.01	0.01	−0.01	0.01
Seriously Ill in Past 12 months	−0.01	0.02	0.07	0.02	−0.02	0.02
Chronically Ill in Past 12 months	−0.01	0.02	0.14	0.02	NA	NA
Female	0.03	0.01	0.02	0.01	0.07	0.01
Educational Attainment	0.01	0.01	0.01	0.01	0.01	0.01
Black	−0.07	0.03	−0.05	0.02	−0.08	0.03
Latino	−0.03	0.03	−0.04	0.02	0.02	0.03
Risk averse	0.01	0.01	0.01	0.01	−0.01	0.01
COVID avoidance	−0.04	0.02	0.01	0.01	NA
R‐squared; N	0.25; n = 2,478		0.20; n = 2,488		0.17; n = 2,103

Appendix 4.

Racial/Ethnic Disparities in Experienced Coverage Shortfalls,* Spring 2021: Comparisons Across Sources of Insurance

Self‐Identified Race/Ethnicity	Employer‐Based Insurance	Individually Purchased Policy
White, non‐Hispanic	7.4%	11.8%
Black, non‐Hispanic	11.7%	26.7%
Hispanic (all races)	14.8%	21.4%

*Percent reporting respondent had “problems getting their insurance to pay for major medical expenses” during the past 12 months