RESEARCH DOSSIER
Health Insurance Provider Network Adequacy and Access
Abstract. Provider network adequacy is the capacity of a health financing arrangement to make covered services realistically obtainable through providers who are clinically appropriate, contractually accessible, geographically reachable, available within clinically reasonable time, and capable of delivering acceptable quality. A nominal roster, geographic coverage threshold, or insurance card measures a different object from completed access. This monograph develops an international and source-critical analysis of provider network sufficiency as a problem in insurance contracting, health economics, spatial allocation, clinical capacity, consumer protection, regulatory implementation, and empirical identification. It distinguishes network composition from effective appointment access and evaluates the relationship among network breadth, bargaining, expenditure, clinical continuity, and unequal treatment opportunities. Its empirical discussion draws on peer-reviewed studies of network breadth, directory validity, secret-shopper access, and plan pricing, together with current U.S. Marketplace, Medicaid managed care, and Medicare Advantage rules, Dutch health insurers' duty of care, and international OECD and WHO indicators. The 2027 U.S. Marketplace rule is treated separately from earlier requirements, avoiding the presumption that regulatory policy was unchanged. Original analytical synthesis develops reproducible comparisons of travel-time eligibility, provider capacity, appointment probability, and multistage access attrition without claiming a new theory or an independently estimated causal effect. The analysis concludes that credible adequacy assessment requires a verified provider denominator, a member-specific feasible choice set, observed or independently tested appointment availability, clinical and distributive quality measures, and dated supervisory remedies. Evidence quality and legal applicability remain conditional on population, market, geography, and observation year.
Keywords: Health insurance; provider network adequacy; access to care; managed care; provider directories; secret shopper; geographic access; travel time; appointment wait time; mental health; specialist access; essential community providers; narrow networks; Medicare Advantage; Medicaid managed care; Marketplace qualified health plans; insurer contracting; quality oversight; health equity; regulatory compliance; health economics.
Introduction
Health insurance promises financial participation in medical care, but the promise is valuable only when an entitled individual can locate and receive the appropriate service. An insurance contract may list surgery as covered while the nearest contracted facility is several hours away; it may name a psychiatrist in the network while the office rejects new patients; or it may offer generous payment for treatment that requires prior authorization, referral, transport, or a specialist that the insured cannot secure. These are distinguishable failures along one care pathway. A scientific account of adequacy must identify where access is produced and where an ostensibly complete entitlement ceases to function.
The National Association of Insurance Commissioners (NAIC) defines network adequacy by reference to the health plan's capacity to deliver promised benefits through sufficient participating providers. Its network adequacy account documents the evolution of Model Act #74 and the practical significance of directories, specialty access, and limited-choice plans. This institutional framing is consistent with a larger health-system distinction: legal coverage, service availability, realized treatment, and protection against household expenditure constitute different measurable outcomes. The WHO–World Bank 2025 global monitoring report separates coverage of essential services from financial hardship, illustrating why an insured population statistic cannot measure provider availability by itself.
This research dossier asks three connected questions. First, how should network composition and geographic reach be defined when nominal provider counts differ from reachable and available clinical capacity? Second, which organizational and economic mechanisms convert plan participation into real appointments, specialist referrals, and continuity of treatment? Third, which quality and enforcement metrics can distinguish regulatory compliance on paper from observed care access across jurisdictions and population groups? The answers are grounded in existing law, research methods, and empirical studies. Originality consists in the comparative synthesis, identification of inferential limitations, and explicit measurement relationships developed here.
The research perimeter includes U.S. network-based coverage and comparable insurer-purchaser responsibilities abroad. The Netherlands provides a contrasting obligation to arrange or mediate access. European cross-border entitlements and OECD unmet-needs statistics illuminate portability and realized access without being reclassified as American-style contracted networks. Global WHO indicators establish the outer health-system context. Each national rule applies only to its specific program, product, benefit year, population, and enforcement authority.
The evidence hierarchy prioritizes current consolidated legal text and issuing-agency rule materials for binding requirements, original survey and research methods for measured outcomes, and systematic syntheses for interpretation across studies. A regulator's policy rationale is evidence about the regulation's stated purpose; it is not proof that the policy achieved its intended outcome. A cross-sectional association between network breadth and premiums is informative about product structure without identifying the isolated causal effect of one additional provider. Those distinctions govern the analysis throughout.
Foundations and Source-Critical History
Contractual and institutional meaning
A provider network is the set of clinicians, facilities, and service organizations linked to a plan under participation arrangements that determine recognition, reimbursement, referral pathways, and enrollee liabilities. The network may be assembled through direct contracts, delegated arrangements, integrated delivery systems, or specialist subcontractors. A provider may participate in one product offered by a carrier and be absent from another product bearing the same brand. Consequently, the defensible identifier is a plan–product–network–provider–location–specialty–time relation, rather than a bare carrier or physician name.
The choice of network changes the practical content of insurance. In the HealthCare.gov plan-type descriptions, an HMO generally restricts nonemergency care to participating clinicians, an EPO generally covers services in-network except emergencies, a PPO permits greater out-of-network choice at different member expense, and a POS arrangement often uses primary-care referrals. These are broad descriptions; actual benefits depend on policy language and applicable law. A network adequacy measure that ignores benefit design may rate two legally different products as if their membership in the same directory conveyed equal access.
A useful distinction is made among structural adequacy, operational adequacy, experienced adequacy, and clinical adequacy. Structural adequacy concerns the number, location, specialties, and physical infrastructure of participating providers. Operational adequacy asks whether the provider can be contacted, recognizes the plan, accepts eligible patients, and supplies appointments. Experienced adequacy concerns the obstacles, waiting, travel, administrative work, and financial payments actually encountered by enrollees. Clinical adequacy concerns fit between the available provider's qualifications, the patient's need, treatment continuity, and quality. These four dimensions are analytic distinctions used to interpret existing evidence, not a newly instituted framework or regulatory doctrine.
Regulatory evolution and different risk pools
The U.S. network adequacy movement reflects the growth of selective contracting and managed care. NAIC's Model Act number 74 was revised in 2015 to give greater attention to provider directories, network sufficiency, and consumer-facing protections. Model legislation supplies a template for state adoption; it is not automatically law in every state. State statutes, implementing regulations, licensed product classes, and enforcement practices must be separately verified before describing a model provision as an individual insurer's legal obligation.
The Affordable Care Act Marketplace regime is addressed in 45 CFR section 156.230, which requires qualifying network plans to include essential community providers and maintain sufficient types and numbers of clinicians, including mental health and substance use specialists, so that covered services are accessible without unreasonable delay. It also prescribes directory accessibility and continuity protections for certain provider terminations. Historically, federal rules introduced quantitative time–distance and appointment access oversight for parts of the federally facilitated Marketplace system. The 2025 Payment Notice summary extended quantitative time–distance requirements to State Exchanges for plan years beginning in 2026.
That history must be read with later amendments. The CMS 2027 Payment Notice final rule, issued May 15, 2026 and effective July 20, 2026, withdrew the requirement that State Exchanges and state-based exchanges on the federal platform impose quantitative time–distance requirements at least as stringent as federal-facilitated standards. It also provided a route for eligible FFE states to perform their own provider-access certification reviews under an Effective Provider Access Review Program. In FFE states where CMS retains review responsibility, the federal standards and related review methods continue as described by the plan-year 2027 CMS guidance. No inference about a particular state's certification practice follows without identifying its exchange and reviewer for the relevant plan year.
The Medicaid managed care regime is separately governed by 42 CFR section 438.68. State standards encompass primary care, obstetrics and gynecology, behavioral health, designated specialists, hospitals, pharmacies, pediatric dental care where covered, and certain long-term services and supports. The 2024 Medicaid managed care final rule imposed additional appointment-access and independent secret-shopper provisions, subject to staggered applicability dates. The official Medicaid implementation resources distinguish a rule's legal effective date from the contract rating periods in which individual reporting, survey, and remedy obligations first apply. Federal Medicare Advantage networks are governed by the distinct 42 CFR section 422.116, including county-specific time-and-distance standards and provider categories. Combining these programs into a single numerical national threshold would be legally and statistically mistaken.
In the Netherlands, the regulatory starting point is the insurer's duty to ensure appropriate access to covered care, known as zorgplicht. The Dutch supervisory explanation describes obligations concerning timely, geographically accessible, sufficient, and appropriate care and recognizes different legal consequences for benefit-in-kind versus reimbursement policies. Thus, the comparative question is not solely how many physicians appear on a contracted roster; it is also what an insurer must do when an individual cannot obtain insured care. These institutional differences explain why international evidence requires function-based comparison rather than unqualified transplantation of U.S. network terminology.
Economic evidence and the history of narrow-network evaluation
Selective contracting may reduce prices and guide patients toward less costly providers. It may also narrow choice, increase the consequences of directory errors, and intensify bargaining disputes. Dafny, Hendel, Marone, and Ody (2017), using early Marketplace plan data, reported an adjusted association in which plans with narrow physician and hospital networks had premiums around 16% lower than otherwise comparable broad-network plans; the original Health Affairs study also explained effects on subsidy benchmarks. That estimate is specific to its years, markets, model, and operational definition of breadth. It should not be presented as a permanent causal saving guaranteed by a narrow network.
Liebman and Panhans (2021) examined Colorado's nongroup market and decomposed narrower-network cost savings into lower utilization, negotiated price differences, and steering away from expensive hospitals. Their Health Economics decomposition attributed approximately 67%, 15%, and 18%, respectively, to those channels in the study's empirical setting. The percentages describe that model's accounting of observed differences, not universal policy constants. A subsequent systematic review of narrow networks by Mazurenko, Taylor, and Menachemi (2022) found cost reductions in many studies while emphasizing a more varied and limited evidence base for access and quality outcomes. The scientific task is therefore to examine both cost formation and whether needed care remains realistically obtainable.
Network Composition and Distance
The correct unit of network measurement
A provider directory usually starts with a count of entries. The analyst should instead specify an observation unit comprising a provider, practice location, clinical service, contractual product, eligibility conditions, and reference date. A multisite physician counted as three entries may represent one clinician with divided capacity. Conversely, a multispecialty team at one address may supply distinct appointment pathways. Neither practice-location counts nor unique clinician counts can be treated as full-time equivalent appointment capacity without further information.
For specialty s, geographic area g, member group m, and reference period t, define an observed eligible network set N(s,g,m,t) consisting of facilities or practitioners with verified plan recognition, relevant service scope, geographic reach, and evidence of accepting appropriate appointments. Its nominal counterpart D(s,g,t) consists of directory-listed entries. Usually N is a verified subset of a larger administrative file, but measurement error can produce missed network members as well as falsely included entries. The magnitude of either discrepancy depends on call protocols, data linkage, and time of observation.
The 2024 KFF study estimated that on average Marketplace enrollees had access to approximately 40% of doctors near their homes through their networks and that 23% were enrolled in plans including one quarter or fewer of local doctors. These are results of the study's provider-data linkage and 2021 reference data, not a 2026 census of actual appointment offers. The KFF network breadth estimates show both the meaning and the limits of relative breadth: 40% of many urban clinicians may yield more candidates than 70% of very few rural clinicians. Absolute supply, clinical mix, and member distribution matter alongside percentage participation.
A network should be represented as a bipartite set of member-demand nodes and provider-capacity nodes with contractual and feasible-travel edges. Each member node has clinically appropriate service demand. Each provider node has time-varying capacity and the ability to accept the relevant product. Edges may be removed by travel barriers, referral requirements, language mismatch, disability access, age restrictions, network-tier restrictions, and closed panels. This representation derives from standard network and allocation concepts and is used solely as an analytical description; it does not establish a new named theory.
Geographical reach and travel measurement
Distance is a screening metric for geographic feasibility. Straight-line distance, road-network distance, expected travel time, and public-transport travel time are separate quantities. Road topology and terrain can create large differences between Euclidean proximity and actual accessibility, and urban congestion can make a geographically close clinic slower to reach than a farther one. The CMS network-distance methodology describes geocoded locations and geographic routes that account for infrastructure and topographic constraints in Marketplace reviews. Analytical replication requires documenting geocoder precision, road network vintage, transport mode, travel-time assumption, and population-sampling frame.
Regulators sometimes stratify geographic standards by county type to acknowledge different underlying service capacities. The Medicare Advantage county-based access standards are an example: the published table specifies different maximum times and distances by specialty and county designation. Such thresholds are administrative test criteria. They do not themselves establish the clinical sufficiency of a network, particularly for subspecialties, urgent interventions, or patients whose travel options differ from the assumed mode.
A meaningful threshold-based population coverage rate is the weighted share of covered people for whom at least one relevant participating provider is within the stated travel limit. For many clinical services, the presence of only one nearby facility produces a fragile access arrangement: its closure, refusal of a new enrollee, loss of a contract, or temporary suspension can convert positive geographic coverage to no feasible option. Alternative-provider counts and second-nearest travel times reveal redundancy. Sensitivity tests should evaluate the access result after the loss of a large local provider or a specialty-specific service interruption.
Travel eligibility also varies by user. A wheelchair user requiring accessible equipment, a person without a vehicle, a family needing pediatric subspecialty services, and a patient requiring interpretation may have materially different feasible networks at one residential point. The Medicaid network standard factors explicitly mention language, reasonable accommodations, physical access, and technical arrangements among considerations for standards. A mean travel time calculated over all enrollees hides these subgroup constraints.
Network breadth, selective contracting, and bargaining
Network breadth is an economic design variable with at least two distinct margins: which providers are contracted and which contracted services are financially attractive to an enrollee. PPO tiers, differential coinsurance, referral restrictions, and facility-specific prior-authorization pathways affect effective choice even when contractual roster breadth is unchanged. A physician who is technically participating but assigned to a high-cost tier may be less usable for a financially constrained member than another clinician with similar clinical qualifications.
A basic decomposition of expenditures is total allowed spending = quantity of covered services × average allowed price, allowing for service mix. Selective contracting may change prices by altering provider bargaining power, quantities by steering, and the mix of episodes by selection and care management. The Colorado network cost study demonstrates these pathways. The tiered hospital network study by Prager (2020) found that a simplified set of observable patient prices could redirect demand toward lower-priced hospitals in its studied environment. These findings explain why a smaller network can sometimes coexist with cost containment. They do not substitute for an adequacy test under a given clinical service obligation.
Provider and insurer market power both influence negotiations. A dominant hospital may obtain higher contracted prices; a concentrated insurer may leverage network inclusion to bargain down reimbursement, potentially affecting provider participation and willingness to accept a plan's patients. An insurer can improve a premium measure while reducing service access if reimbursement changes outpace providers' capacity to absorb costs. Assessing that possibility requires provider participation histories, actual offered appointments, and service-level utilization rather than premium data alone.
Network design also produces selection incentives. Individuals with established specialists, complex chronic conditions, or predictable high utilization may place greater value on a broad network than healthier people. Differences in medical expenditure between narrow- and broad-network enrollees may therefore reflect self-selection, provider pricing, service substitution, or inhibited care. The comparative evidence synthesis finds outcome heterogeneity across settings; causal attribution requires designs that address sorting and competing mechanisms.
Essential community providers and geographic equity
Essential community provider (ECP) provisions recognize access for underserved populations in the Marketplace regulatory setting. The 2027 CMS ECP certification discussion retained the 35% general threshold and parallel federally qualified health center and family-planning thresholds described in its final rule account. This is a particular program-level certification measure, not a substitute for assessing whether underserved enrollees obtain care. A provider's ECP classification, its presence in a service area, and its ability to deliver an appointment are separate events.
Location-sensitive evaluation should report results for urban, suburban, rural, and remote populations alongside the relevant local provider supply. A high percentage of participating practitioners may coexist with absolute scarcity, while a lower percentage in a dense metropolitan setting may offer sufficient appointments. Public transport, service hours, referral prerequisites, and non-English communication capacity change the usable opportunity set. Haeder and Xu's 2024 consumer preferences survey found support for broad definitions of adequacy encompassing such accessibility features. Preferences from a U.S. opinion survey should inform what to measure, while actual service availability must be checked through operational evidence.
Appointment and Specialist Access
From a listing to a completed appointment
Access is a sequence of conditional steps. A patient must recognize a clinical need, identify an appropriate provider, find contact details, reach the office, confirm plan eligibility, satisfy clinical and administrative criteria, obtain an offered slot, and complete the encounter. Failure at any step can eliminate a provider from the patient's feasible options. The probability of completing a routine appointment is therefore influenced by each conditional stage; multiplication of observed transition probabilities is valid only when the probabilities are correctly defined on sequential conditional denominators.
The central measurement problem is that directory inclusion does not establish either availability or eligibility for an appointment. Haeder, Weimer, and Mukamel (2016) tested 743 primary-care providers in five California Marketplace pricing regions and reported that an appointment with the initially selected clinician could be scheduled in fewer than 30% of cases. Their secret-shopper access findings concerned a specific time, set of products, and sampling frame. It established a serious mismatch between nominal choice and practicable choice in that sample without establishing a universal national failure rate.
CMS's early Medicare Advantage provider directory review found that 45.1% of reviewed directory locations had inaccurate information. That administrative audit measured inaccuracies in a specified sample of plans and locations; the result is not the share of all physicians who were unreachable or the share of all Medicare Advantage appointments denied. The same caution applies whenever a publication interchanges the denominator of locations, unique providers, calls, listed services, or enrollees.
Burman, Haeder, and Xu (2023) examined California mental-health directory and appointment data covering 2018–2019 across multiple insurance markets. Their mental-health access investigation found substantial directory inaccuracies and constrained timely appointment opportunities, with differences across commercial, Marketplace, and Medicaid products. These are comparative associations from survey and administrative observations. They illustrate that ostensibly strong regulatory standards can coexist with practical obstacles, and that an evaluation must observe implementation rather than infer it from the text of the law.
Behavioral health and specialty bottlenecks
Mental health and substance use disorders create distinctive network adequacy questions because service types, patient needs, professional qualifications, acceptance rules, therapeutic continuity, and visit frequency vary. A psychologist, psychiatrist, counselor, inpatient psychiatric facility, and crisis service are not interchangeable slots. Primary-care access can be adequate while psychiatry or pediatric behavioral health services remain unobtainable. Specialist classification therefore needs clinical scope and referral feasibility, not merely broad professional categories.
Haeder and Zhu (2025) assessed 8,306 mental-health counselors reported in Pennsylvania ACA Marketplace regulatory filings for plan year 2024. Their filing-to-directory comparison found 19.9% of filed listings absent from directories and only 35.3% fully matched. Among 2,152 listings completely verified by calls, 65.2% had at least one inaccuracy and only 321, or 14.9%, yielded available appointments; mean offered wait among the relevant observed appointments was 33.2 days. These denominators differ. The 14.9% is a percentage of fully verified listings, not 14.9% of all enrollees obtaining care, and the 33.2 days is not a population-weighted national wait time.
The 2025 findings strengthen a source-critical conclusion: regulatory submission, public directory, reachable office, and appointment offer should be measured as separate linked tables. This enables identification of errors at each layer rather than attributing all unsuccessful calls to a generic category of network inadequacy. It also permits intervention-specific evaluation: correction of directories may help patients locate genuine providers, while an insufficient clinical workforce may require recruitment, reimbursement change, telehealth integration, or alternate capacity creation.
The Pennsylvania directory follow-up by Haeder and Zhu (2024) recontacted 5,170 previously inaccurate provider listings after approximately four to nine months. At follow-up, 19.0% had been removed, 44.8% remained inaccurate, 11.6% were verified accurate, and 24.6% could not be reached for verification. Because the cohort was selected for earlier inaccuracy, these percentages are not an estimate of the overall error rate among all listings. Their importance is temporal: inaccuracies can persist and correction performance differs by insurer. A one-time annual roster audit may therefore miss prolonged barriers between review dates.
Timeliness, urgency, and clinically appropriate priority
Appointment latency should be recorded from a standardized request date to the first clinically appropriate offered appointment, with distinctions between new and established patients, routine and urgent need, and in-person versus virtual encounters. Median wait, upper-tail wait, and the probability of meeting a specified threshold each answer a different question. A mean can be distorted by long tails; a threshold compliance rate can conceal delays just beyond the cutoff and extremely long delays within the noncompliant group.
Under Medicaid managed care rules, specified routine appointment standards include no longer than 10 business days for covered outpatient mental health and substance use services and 15 business days for covered primary care and obstetrics/gynecology, with other evidence-based state-selected service criteria. The detailed secret-shopper compliance mechanism includes an at-least-90% appointment availability measure for the specified standards. The Medicaid rule implementation schedule matters: some new independent surveying and remedy requirements begin with rating periods in 2028, while other reporting measures begin earlier. A study of 2025 operations cannot be classified retroactively as a failure to perform a 2028 survey obligation.
For Medicare Advantage, network access regulation and complementary plan requirements organize different provider and service standards. For qualified health plans, Marketplace network rules and the 2027 certification materials govern a different covered population. A single ten-day target would thus be an invented national measure. Where evidence lacks the relevant contractual class and regulatory year, the correct result is a descriptive wait-time distribution and an explicitly unclassified compliance determination.
Referral, continuity, telehealth, and financial access
A member can secure a primary-care appointment and still fail to reach the specialist specified by a referral. Network adequacy thus extends to the end-to-end pathway: referring clinician, authorized specialty, diagnostic site, procedure center, rehabilitation, laboratory, medication access, and follow-up. Coordinated care can reduce duplication and improve continuity, but poorly integrated referral rules can create additional queues. Measurement should trace referrals initiated, accepted, completed, and abandoned, specifying elapsed time between milestones and whether a clinical substitution was appropriate.
Continuity matters when network contracts terminate during treatment. Under 45 CFR continuity provisions, certain QHPs must give specified notices and allow qualifying active treatment to continue under defined circumstances following a provider termination without cause. The details of this federal rule cannot be assumed to govern every employer, Medicaid, or foreign insurance product. At the analytic level, continuity is a time-dependent property: a network adequate at enrollment may become inadequate after provider turnover, plan redesign, or relocation.
Telehealth can expand feasibility where travel or specialty shortage is binding, but effective remote access depends on the suitability of the service, technical connectivity, privacy, licensure, language, clinical follow-up, and coverage. A virtual appointment is not an equivalent substitute for surgery, physical examination, or certain diagnostic procedures. In Medicaid access considerations, telemedicine is included among technological features states may consider. Analysts should document when a telehealth offering satisfies a specific regulatory standard and when it is an additional pathway without legal equivalence.
Even a geographically and clinically adequate participating network may be financially inaccessible if deductibles, coinsurance, facility fees, or uncovered episodes prevent use. Out-of-network remedies may expand possibilities but also create uncertain household liability unless coverage rules protect the enrollee. The Marketplace plan-type distinctions demonstrate why identical provider lists can convey different effective financial options. Network adequacy research must therefore connect the physical choice set to coverage terms rather than equating any possible treatment with insured access.
Quality and Enforcement Metrics
Measuring quality without reducing it to roster size
Quality-sensitive adequacy asks whether available providers can deliver safe, effective, and appropriate care. A technically reachable but clinically unsuitable provider is not an adequate substitute for a required specialist. Quality indicators may include preventable complications, avoidable admissions, readmissions, guideline-concordant treatment, patient-reported experience, and continuity of chronic disease management. Each indicator requires risk adjustment, encounter definitions, attribution rules, observation periods, and careful interpretation because illness severity and selection into providers can confound raw comparisons.
An analytically defensible performance dashboard separates structural supply, verified access, clinical process, patient experience, and outcomes. Reporting these dimensions side by side protects against compensatory scoring in which excellent geographic coverage obscures high appointment refusal rates. It also prevents clinical outcome variation outside an insurer's plausible control from being mistaken for direct proof of network noncompliance. The question of legal compliance remains tied to the governing statute or contract, while quality outcomes can inform substantive assessment beyond the legal minimum.
The consumer perspective broadens this evaluation. Haeder and Xu (2024), in a U.S. sample of 4,008 adults, found support for definitions incorporating transport, culturally appropriate care, and language accessibility alongside conventional physician counts and distance. Their consumer network adequacy study measures expressed preferences; it does not establish observed clinical outcomes or a democratically enacted threshold. Yet it supports the substantive inclusion of characteristics omitted by simple spatial indices.
Directory integrity as an enforcement problem
Directory error is a compound problem of identification, governance, and incentives. Providers change affiliations and addresses, contract terms vary across products, practices close panels, and delegated administrators may not transmit updates promptly. A regulator seeking valid rosters should require identifiable timestamps, reliable product mapping, source provenance, and independent verification with a documented sampling frame. More frequent directory updates may reduce stale entries, but update frequency alone does not demonstrate that corrections are accurate.
The CMS Medicare Advantage audit and the directory persistence study show why data-quality problems are substantive consumer protection issues. A false listing can waste time, impose travel or communications costs, and delay treatment. It can also exaggerate a network's measured adequacy if regulators count nominal entries as functional providers. Audit findings should distinguish materially harmful errors, such as wrong network status or unavailable contact, from cosmetic inconsistencies that do not prevent access.
A strong audit uses independent callbacks or appointment-request testing for a representative sample, captures repeated attempts and nonresponse, and distinguishes an office that is closed from one that refuses the specific plan or cannot offer a timely appointment. It should not treat a voicemail as a completed verification. Analysts should preserve the original directory snapshot and contemporaneous contract roster so that later corrections do not erase the conditions under investigation.
Complaint data, enforcement, and remedies
Consumer complaints, appeals, grievances, denials, and calls to assistance services provide useful signals about access. They have selection and reporting biases: awareness of complaint channels and capacity to contest a problem differ across populations. An increase in complaints after outreach may reflect improved detection rather than deteriorating access. Such measures belong alongside observed appointment outcomes and should be disaggregated by plan, location, service, and severity.
In the United States, enforcement instruments include certification conditions, corrective action, network expansion, exceptional access arrangements, directory correction, public reporting, and contractual or financial remedies where authorized. The 2024 Medicaid rule account includes annual independent secret-shopper surveys, enrollee-experience monitoring, and remedy plans, with precise applicability determined by regulation and rating period. The Marketplace 2027 rule allocates provider-access review authority among federal and qualified state bodies. Enforcement should be attributed to the agency with actual jurisdiction rather than to an abstract national regulator.
The Dutch NZa wait-list mediation guidance supplies a contrasting remedy orientation. When a contracted provider cannot offer timely insured care, an insurer's duty may require helping the member obtain an alternative, including through a noncontracted provider under the specified circumstances. The NZa describes insurer obligations and reimbursement consequences dependent on the policy and mediation facts. The broader analytic significance is that an access obligation can be evaluated through concrete remedial action for an individual, not just through an annual roster score.
This comparison yields an interpretive conclusion: the maturity of supervision depends partly on whether evidence can trigger a defined operational remedy. Certification without monitoring risks validating a static snapshot. Monitoring without enforceable follow-up documents persistent deficiency. A remedial system should record the responsible institution, affected population, detected failure, corrective action, timetable, subsequent patient access, and consequences for nonperformance.
Equity, accessibility, and distributional adequacy
Provider shortages disproportionately constrain people with complex needs when their viable options are small. Assessments should identify adults and children separately; behavioral-health demand and supply; disability and mobility; language; age; rural residence; socioeconomic disadvantage; and medically underserved communities, subject to lawful and ethical data processing. Comparing subgroup access rates with a systemwide average establishes where aggregate compliance conceals risk.
The OECD unmet-care analysis reports that 3.4% of residents across 28 countries with comparable data experienced unmet medical needs because of cost, distance, or waiting in 2024, with 2.3% citing waiting times on average. The publication also reports substantial national and socioeconomic differences. These are survey-based cross-country outcomes, not U.S. insurer network adequacy pass rates. They demonstrate why timely access is measurable separately from insurance enrollment and why the meaning of a pooled national rate depends on its country sample and survey instrument.
An insurer may be unable to create clinicians in a severely underserved region, but it can still be assessed on contracting, referral management, reimbursement, transport assistance, telehealth when appropriate, and reasonable use of external capacity. The attribution of responsibilities must respect local law and the limits of the insurer's institutional role. At the level of policy evaluation, a good outcome measure should reveal whether the patient actually obtained clinically necessary care, independent of the administrative explanation for an initially inadequate network.
Data, Measurement, and Validation
Source architecture and reproducibility
A minimum reproducible dataset requires versioned provider-directory snapshots, contracted provider files, provider identity and location records, specialty taxonomy, service area, covered population, eligibility and benefit design, and timestamps. Stronger analysis integrates new-patient acceptance status, appointment offers, referral completions, claims or encounters, service quality, patient-reported barriers, and administrative complaints. The linkage key must identify the product and network, not just an insurer or a provider name.
Each data source has characteristic measurement failures. Directories may include erroneous or obsolete locations. Contracted-provider files may omit capacity restrictions. Claims record realized billing rather than unmet demand. Electronic scheduling may exclude patients unable to obtain an initial account or referral. Call surveys face refusal, nonresponse, script effects, and timing differences. Patient surveys face recall and nonresponse biases. Geographic files may geocode the practice headquarters instead of the clinic where appointments occur. An evidence ledger should record these vulnerabilities as properties of the measurement process.
The Pennsylvania filing validation study demonstrates one data-linkage failure: filings and public directories diverged before office verification even began. The California primary-care test demonstrates the subsequent conversion loss from listed providers to offered appointments. A scientifically robust audit traces both attrition stages and their uncertainties. A plan with a more accurate directory can display fewer listed providers after correction even while providing a more truthful description of access; unadjusted counts could mistakenly categorize that improvement as deterioration.
Formal measures and denominators
Let M denote members, P providers, S clinical service types, and t the observation date. Define I(m,p,s,t) as 1 when member m has a financially and clinically eligible path to provider p for service s at t, including applicable travel constraints, and 0 otherwise. Define A(m,p,s,t) as 1 when an appropriately timed appointment can actually be offered, and 0 otherwise. A simple structural coverage proportion is:
C_s(t) = [sum over m in M of w_m · 1{there exists p in P with I(m,p,s,t)=1}] / [sum over m in M of w_m].
An appointment-realizable proportion replaces I by the conjunction I·A. Here w_m are member or sampling weights, specified before estimation. If a survey measures only a probabilistic appointment offer, estimated A values must reflect call protocols and uncertainty. Neither proportion is interpretable without defining service s, acceptable travel, urgency, plan product, eligible member set, and observation period.
An illustrative access attrition calculation can be reproduced without empirical inference. Suppose a hypothetical file lists 100 clinician-location entries. Of these, 80 can be reached; among those, 60 verify active network participation; among those, 45 accept the relevant patient type; and among those, 30 offer an appointment within the prespecified time threshold. Then the end-to-end verified appointment yield is 30/100 = 30% of initial entries. The conditional stage yields are 80/100 = 80%, 60/80 = 75%, 45/60 = 75%, and 30/45 = 66.7%. Their product is 0.80 × 0.75 × 0.75 × (2/3) = 0.30. This is an arithmetical illustration, not a reported insurer performance estimate. It shows precisely why denominators must remain visible.
For travel, let T(m,p) be door-to-door journey time for the specified transport mode, and τ(s,g) be the applicable test threshold for service s in geography g. Reachability is R(m,p,s)=1 when T(m,p)≤τ(s,g), with required clinical and accommodation conditions recorded separately. A redundancy index may count the number of independently capable providers reachable by m, but the number is an input to vulnerability assessment rather than a validated outcome measure in its own right. Capacity allocation and competing member demand must also be considered.
For waiting times, define W_i as business days between a standardized request and the first clinically appropriate offered appointment. Report the number attempted, successfully reached, fully verified, and offered, as well as the share with W_i≤k. Stratify by urgency and whether the requested clinician or any appropriate in-network clinician was offered. Estimates should carry uncertainty intervals reflecting the design and clustering at practice, network, plan, and geographic levels. Observational point estimates without sampling uncertainty can convey unwarranted precision.
Sampling, validation, and uncertainty
Secret-shopper samples should be probability-based or transparently stratified over provider types, plan products, locales, and enrollment-weighted demand. Multiple attempts at defined hours can distinguish persistent inaccessibility from one unsuccessful call. Standardized scripts reduce interview variation but may obscure complexities in eligibility verification or communication accommodations. Scripts must be reviewed for legality and ethics; survey design should avoid occupying real patient appointments unnecessarily.
Noncontact requires explicit treatment. Coding every unanswered call as a refusal gives one bound; excluding noncontacts gives a potentially optimistic complete-case rate; treating them as unknown preserves uncertainty. If 200 listings are sampled, 120 verified, and 48 successfully offer an appointment, the verified-respondent rate is 40%, whereas the conservative initial-listing yield is 24%. The maximum overall yield under an assumption that all 80 unverified listings would have succeeded is 64%. These are hypothetical sensitivity bounds, not a confidence interval and not a causal estimate. The comparison conveys the importance of nonresponse.
Directory correction should be tested in longitudinal panels with stable provider identifiers and product definitions. Haeder and Zhu's longitudinal follow-up evidence illustrates both the feasibility of recontact and the selection limits of a cohort drawn from previously inaccurate listings. Before–after declines in listed errors may reflect removals rather than newly accessible appointments. A complete evaluation should distinguish corrected information, restored access, and withdrawn entries.
Causal inference and economic identification
Network breadth is endogenous to insurer negotiation, local supply, plan prices, patient demand, and risk selection. Regressing spending on network size alone cannot establish the effect of restriction because high-demand regions, attractive hospitals, and patient health may influence both variables. The Marketplace premium study and Colorado cost decomposition contribute valuable empirical evidence with study-specific designs and controls; their coefficients should be described at the level of inference supported by those designs.
Where a policy threshold changes exogenously or insurer networks alter their composition, causal research might exploit difference-in-differences, event studies, discontinuities, instrumental variables, or policy stagger, subject to credible identification assumptions. A difference-in-differences study requires plausible counterfactual trends and attention to anticipatory contracting, plan entry, population sorting, and spillovers. Provider mergers and shared clinic systems can transmit changes across nominal treatment and control networks. Time-varying physician supply may confound estimated effects attributed to oversight.
An informative causal evaluation tracks both unintended underuse and economically efficient substitution. A narrower network could reduce spending by obtaining lower contracted prices, redirecting care to equally capable providers, delaying needed care, or attracting a healthier membership. These mechanisms differ in welfare consequences. The narrow-network systematic review offers a documented synthesis; the tiered-network demand analysis provides evidence of steerage under simplified price signals. Observed reduced utilization should therefore be tested against care quality, urgency, health status, and episodes avoided or deferred.
International measurement comparability
Cross-country comparisons should distinguish insurance network rules from general public purchaser obligations, availability of health workers, financing entitlements, and total care capacity. The OECD 2025 unmet-needs indicator estimates self-reported barriers using a population survey design; WHO–World Bank global universal coverage indicators measure broader service coverage and financial hardship. Neither is a direct proxy for contracted network adequacy. A country with universal statutory entitlement may have no analogous commercial network roster, and a country with broad private network listings may still experience unmet care needs.
Differences in survey recall windows, exclusions, insured populations, service definitions, urbanization, and benefit design can reverse naive rankings. Comparative work should first harmonize the phenomenon—timely receipt of needed care—then report how each financing institution controls or fails to control the relevant service pathway. Legal compliance should remain jurisdiction-specific rather than transformed into an invented international benchmark.
International Synthesis and Scientific Conclusions
United States: differentiated program regulation
The United States illustrates the importance of identifying the specific regulated product. Marketplace QHPs have federal minimum access requirements and, depending on exchange type, review allocation, and plan year, quantitative certification procedures. Medicaid managed care imposes state-administered standards and federally structured reporting and appointment requirements. Medicare Advantage has its own contracted-network and geographic access regime. The applicable sources are Marketplace provider access regulation, Medicaid network standards, and Medicare Advantage standards. A claim that a single U.S. plan must satisfy all three programs' detailed requirements simultaneously would be erroneous.
A historically specific comparison matters in November 2026. The earlier State Exchange requirement and the later 2027 final rule revision point in different regulatory directions. The former expanded quantitative review, while the latter restored state discretion and created a pathway for certain federally facilitated states to perform certified reviews. Identifying the applicable rule's effective date and benefit year is necessary for any legal QA. A policy analysis that ignores amendment chronology would produce misleading statements even if every individual historical quotation were authentic.
U.S. empirical studies also demonstrate that plan rosters and directory filings can materially overstate access. The 2025 mental-health audit offers a particularly clear example of distinct denominators and multiple failure stages. Its clinical and geographic specificity limits extrapolation, but its research design shows why regulator-controlled datasets should be independently validated. The appropriate conclusion is a measurement requirement and an empirically grounded concern, not a universal estimate that all U.S. insurance networks are inadequate.
Netherlands: insurer duty and individual access remedy
The Netherlands provides a different institutional approach because its statutory basic health insurance is paired with a duty for insurers to ensure access to necessary covered care. The NZa duty-of-care explanation emphasizes timely, available, geographically accessible, and appropriate services. The NZa mediation guidance describes the insurer's obligations when waiting times exceed accepted norms and the practical role of alternative providers. These features make remedial performance part of the substantive inquiry.
The result is an important comparative distinction: a U.S. certification review often asks whether a contracted network satisfies a specified structural or appointment standard, while the Dutch duty-of-care process also evaluates what an insurer does for a person who lacks timely covered treatment. Neither system is reducible to a universal physician-per-thousand-population metric. Both require documented pathways from the enrolled individual to the care promised under the financing arrangement.
European portability and the wider health system
EU cross-border rules introduce another perspective. The European Commission's cross-border health care overview explains patient rights under Directive 2011/24/EU and associated reimbursement conditions. These rights can create additional lawful treatment options, but the entitlement's reimbursement prerequisites and applicable national arrangements differ from an insurer's domestic provider network. A specialist available abroad is not automatically an in-network specialist under a given private insurance product.
The OECD unmet-care comparison and WHO–World Bank universal coverage monitoring reveal that financial entitlement and effective service access must be evaluated together. WHO reported a 2023 global service coverage index of 71 and substantial remaining deficiencies, but those figures refer to a multidimensional health-service outcome, not a plan-level network audit. It would be a category error to convert the global index into a verdict on an individual insurer's network.
Integrated scientific findings
Several conclusions follow from the source-critical comparison.
First, a network is a dynamic service capability rather than a static list of names. Its effective boundary depends on contractual participation, service scope, geographic and functional reach, appointment availability, and patients' ability to navigate the care pathway. Administrative roster size is a useful input, but independent verification is necessary when that input carries a regulatory conclusion.
Second, adequacy and efficiency are linked through provider contracting and prices. Narrower networks may produce savings through bargaining, steerage, and member composition, as illustrated by the Colorado mechanisms study. A scientifically defensible efficiency claim must preserve the distinction between genuine lower resource cost and forgone necessary treatment. Premium comparisons, expenditure regressions, and quality indicators answer different questions and need to be analyzed together.
Third, regulatory requirements have different legal subjects and dates. The 2027 Marketplace rule, 2024 Medicaid final rule, Medicare Advantage regulation, and Dutch care obligation allocate responsibilities differently. The comparison supports institutional specificity as a methodological imperative. It does not authorize a single global compliance score.
Fourth, an independently observed appointment offer supplies evidence that a directory listing cannot provide. The early California field study, mental-health directory evaluation, and Pennsylvania filing audit indicate distinct measurement failures across settings. Their separate samples should be compared by study design and conditional denominators rather than pooled into a fictitious global error rate.
Fifth, quality and equity belong inside substantive adequacy analysis. Travel limits and specialist counts are meaningful only relative to patient needs, disability access, language, transport, clinical urgency, and capacity competition. The consumer preference research supports including these variables in the measurement agenda, while administrative and appointment records are needed to establish service outcomes.
Sixth, enforcement has a temporal dimension. A compliant network at certification can deteriorate because of changes in providers, patients, or contracts. A mature evaluation therefore uses repeated observation, timely directory correction, measured appointment performance, specific remedies, and read-back evidence that the remedy worked. It assigns responsibility according to law rather than to a generic organizational concept.
Outstanding empirical questions
The most consequential unanswered empirical problem is the gap between contracted network capacity and expected realized access for clinically defined groups. Reliable multi-plan panels linking actual appointment slots, clinician capacity, specialty, patient eligibility, and network status are uncommon. Where data exist, they often omit rejected requests or services not used because the patient anticipated failure.
Another question is whether stronger monitoring changes access or primarily changes the accuracy of reported rosters. Better disclosure can reduce consumer search costs without increasing treatment supply. An additional question is how insurers respond when geographic shortages make statutory thresholds structurally difficult to achieve. These questions warrant original data acquisition and preregistered or quasi-experimental study designs before asserting new causal effects.
The present evidence allows a strong analytical conclusion about what is necessary to evaluate provider network adequacy. It does not establish that any particular optimal network size exists for all settings, that a particular legal regime outperforms all others, or that an unobserved quality benefit follows from a given provider count. Scientific decisions must remain tied to their actual measurement and source basis.
Conclusion
Health insurance provider network adequacy is a measurable condition of effective entitlement. Its assessment begins with accurate provider identity and a legally relevant product definition, proceeds through reachable and clinically capable providers, and reaches its substantive endpoint in timely access to appropriate covered services. The relationship between these stages is sufficiently complex that neither nominal network breadth, low premiums, geographic thresholds, nor favorable utilization figures can independently establish adequacy.
The documentary record demonstrates three durable findings. First, network composition must be measured at the plan–provider–location–specialty–time level, with explicit member denominators and service geography. Second, appointment availability and patient experience supply indispensable evidence of access beyond the roster; research in U.S. markets has repeatedly identified discrepancies among filings, directories, and offered appointments. Third, quality oversight and enforceable remediation depend on jurisdiction, applicable year, and the ability to verify that corrected administrative information has led to better access.
The original contribution of this monograph is its source-grounded integration of economic mechanisms, clinical access stages, spatial measurement, regulatory revision, and quality assurance into a reproducible critical analysis. It identifies causal ambiguities, gives transparent measurement definitions, and preserves the distinctions among U.S. health insurance programs and other national arrangements. These results support future empirical research and responsible supervision without proposing a new authorial theory or converting analytical interpretation into a proprietary doctrine.
Appendix A: Reproducible Measurement Definitions
The following are methodological specifications for reproducible research, not regulations, validated clinical thresholds, or original estimates.
- Member denominator: individuals actually enrolled and entitled to the specified service during a defined interval; distinguish eligible persons, enrolled persons, and users of care.
- Provider denominator: unique product-specific clinical service opportunities, with alternative views by person, organization, site, specialty, clinical capacity, and contracted facility.
- Verified active-network rate: verified in-network service entries divided by attempted directory entries, with noncontact reported separately.
- Appointment yield: timely clinically appropriate appointment offers divided by all standardized attempts, and separately divided by fully verified contacts.
- Travel access: weighted proportion of members linked to at least one qualifying provider within defined routing and mobility criteria; report the second-nearest option and no-alternative share.
- Capacity: available appointment-equivalent service time adjusted for clinical suitability, not an unverified directory headcount.
- Continuity: documented completion of necessary care across provider contract changes, referrals, and network transitions.
- Quality: predeclared clinical and patient-experience measures with risk, case mix, and attribution limitations reported.
- Equity: subgroup differences expressed as absolute percentage-point gaps and relative ratios, each with sample sizes and uncertainty.
- Enforcement: verified failure event, legal authority, relevant remedy, deadline, corrective evidence, and longitudinal read-back.
For the hypothetical 100-entry access illustration, the four sequential conditional fractions equal 0.80, 0.75, 0.75, and 2/3. Their product equals 0.30. For the hypothetical noncontact sensitivity example with 48 offers out of 120 verified attempts from 200 listings, verified-case success is 48/120 = 40%, conservative all-attempt yield is 48/200 = 24%, and the logical upper bound assuming all 80 unknown cases succeed is (48+80)/200 = 64%. These calculations require no empirical parameter estimate; they are worked arithmetic intended to prevent denominator confusion.
Appendix B: Research Design and Evidence Appraisal
This dossier is a structured, international documentary and analytical synthesis based on openly reachable original statutory or supervisory materials, regulatory summaries, independently published empirical studies, and comparative health-system reporting. It is not a preregistered systematic review. The analysis did not obtain proprietary insurer rosters, individual claims, patient-level health outcomes, or live provider appointment inventory. A linked source therefore supports the specific claim for which it is cited, within that source's population and dates, rather than a broader generalization by association.
Legal statements are dated to the source as checked November 16, 2026. U.S. Marketplace rules are differentiated between the 2025 and 2027 Payment Notices; Medicaid managed care effective and applicability dates are distinct; Medicare Advantage requirements remain a separate regulatory category. Dutch supervisory descriptions are evidence of Dutch health insurer obligations and should not be applied to another jurisdiction. UHC and OECD measures are population-level context, not substitutes for network audit outcomes.
Original empirical studies use different samples, denominators, and statistical designs. KFF's 2024 report uses underlying 2021 provider data, while the 2025 Pennsylvania study reports 2024 Marketplace mental-health listings. Results are therefore not pooled as if contemporaneous national observations. The peer-reviewed review by Mazurenko, Taylor, and Menachemi supplies evidence synthesis through its own search period, which is not identical to this dossier's source-review window. All hypothetical numeric illustrations are explicitly labeled and arithmetically reproducible.
A future empirical extension should prespecify clinical services and target groups, obtain data-use permission where needed, protect health information, independently verify plans and providers, record call nonresponse, preregister inference where suitable, preserve source versions, and report both realized access and clinical consequences. The manuscript advances original comparative reasoning and source-critical interpretations; it asserts no independently discovered causal effect, new regulatory standard, or proprietary analytical doctrine.
References
National Association of Insurance Commissioners (NAIC) (2025). Network Adequacy. Insurance Topics; discussion of Health Benefit Plan Network Access and Adequacy Model Act #74. https://content.naic.org/insurance-topics/network-adequacy
United States, Code of Federal Regulations (2026). 45 CFR § 156.230, Provider Access Standards for Network Plans (consolidated text reviewed November 2026). https://www.law.cornell.edu/cfr/text/45/156.230
Centers for Medicare & Medicaid Services (CMS) (2026). HHS Notice of Benefit and Payment Parameters for 2027 Final Rule. Fact Sheet, May 15. https://www.cms.gov/newsroom/fact-sheets/hhs-notice-benefit-payment-parameters-2027-final-rule
Centers for Medicare & Medicaid Services (CMS) (2024). HHS Notice of Benefit and Payment Parameters for 2025 Final Rule. Fact Sheet. https://www.cms.gov/newsroom/fact-sheets/hhs-notice-benefit-and-payment-parameters-2025-final-rule
Centers for Medicare & Medicaid Services (CMS) (2026). Network Adequacy FAQs: Plan Year 2027 Time and Distance Methodology. QHP Certification. https://www.qhpcertification.cms.gov/QHP/faqs/Network-Adequacy-FAQs
United States, Code of Federal Regulations (2026). 42 CFR § 438.68, Medicaid Managed Care Network Adequacy Standards (consolidated text reviewed November 2026). https://www.law.cornell.edu/cfr/text/42/438.68
Centers for Medicare & Medicaid Services (CMS) (2024). Medicaid and Children's Health Insurance Program Managed Care Access, Finance, and Quality Final Rule (CMS-2439-F). Fact Sheet. https://www.cms.gov/newsroom/fact-sheets/medicaid-childrens-health-insurance-program-managed-care-access-finance-quality-final-rule-cms-2439
Centers for Medicare & Medicaid Services (CMS) (2024). Medicaid and CHIP Managed Care Final Rules: Implementation and Applicability Resources. https://www.medicaid.gov/medicaid/managed-care/guidance/medicaid-and-chip-managed-care-final-rules
United States, Code of Federal Regulations (2026). 42 CFR § 422.116, Medicare Advantage Network Adequacy (consolidated text reviewed November 2026). https://www.law.cornell.edu/cfr/text/42/422.116
Nederlandse Zorgautoriteit (NZa) (2026). Zorgplicht: Legal Duty to Ensure Access to Insured Care. https://www.nza.nl/onderwerpen/z/zorgplicht
Nederlandse Zorgautoriteit (NZa) (2026). De wachttijden bij gecontracteerde zorgaanbieder zijn te lang. Wat moet de zorgverzekeraar doen? Guidance on Wait-List Mediation. https://www.nza.nl/documenten/vraag-en-antwoord/zorgverzekeraars/de-wachttijden-bij-gecontracteerde-zorgaanbieder-zijn-te-lang.-wat-moet-de-zorgverzekeraar-doen
Organisation for Economic Co-operation and Development (OECD) (2025). Unmet Needs for Healthcare. Health at a Glance 2025. https://www.oecd.org/en/publications/health-at-a-glance-2025_8f9e3f98-en/full-report/unmet-needs-for-healthcare_c994e8b1.html
World Health Organization (WHO) and World Bank (2025). Tracking Universal Health Coverage: 2025 Global Monitoring Report. Report and Monitoring Materials. https://www.who.int/teams/health-financing-and-economics/global-monitoring-report/2025
European Commission (2026). Cross-Border Healthcare: Overview of Directive 2011/24/EU and Patient Rights. https://health.ec.europa.eu/cross-border-healthcare/overview_en
Rae, Matthew; Pollitz, Karen; Pestaina, Kaye; Long, Michelle; Lo, Justin; and Cox, Cynthia (2024). How Narrow or Broad Are ACA Marketplace Physician Networks? KFF, August 26. https://www.kff.org/private-insurance/how-narrow-or-broad-are-aca-marketplace-physician-networks/
Haeder, Simon F.; Weimer, David L.; and Mukamel, Dana B. (2016). Secret Shoppers Find Access to Providers and Network Accuracy Lacking for Those in Marketplace and Commercial Plans. Health Affairs 35(7):1160–1166. DOI:10.1377/hlthaff.2015.1554. https://pubmed.ncbi.nlm.nih.gov/27385229/
Burman, Abigail; Haeder, Simon F.; and Xu, Wendy Yi (2023). Provider Directory Inaccuracy and Timely Access for Mental Health Care. American Journal of Managed Care 29(2):96–102. DOI:10.37765/ajmc.2023.89318. https://www.ajmc.com/view/provider-directory-inaccuracy-and-timely-access-for-mental-health-care
Haeder, Simon F., and Zhu, Jane M. (2024). Inaccuracies in Provider Directories Persist for Long Periods of Time. Health Affairs Scholar 2(6):qxae079. DOI:10.1093/haschl/qxae079. https://academic.oup.com/healthaffairsscholar/article/2/6/qxae079/7687298
Haeder, Simon F., and Zhu, Jane M. (2025). ACA Network Regulatory Filings Are Inaccurate, Poorly Match Provider Directories. American Journal of Managed Care 31(9):e265–e269. DOI:10.37765/ajmc.2025.89791. https://www.ajmc.com/view/aca-network-regulatory-filings-are-inaccurate-poorly-match-provider-directories
Centers for Medicare & Medicaid Services (CMS) (2017). Online Provider Directory Review Report Final. January 13. https://www.hhs.gov/guidance/document/online-provider-directory-review-report-final
HealthCare.gov / U.S. Department of Health and Human Services (2026). Health Insurance Plan and Network Types: HMOs, PPOs, EPOs, and POS Plans. https://www.healthcare.gov/choose-a-plan/plan-types/
Haeder, Simon F., and Xu, Wendy Yi (2024). When Is a Network Adequate? Consumer Perspectives on Network Adequacy Definitions. American Journal of Managed Care 30(9):424–432. DOI:10.37765/ajmc.2024.89601. https://www.ajmc.com/view/when-is-a-network-adequate-consumer-perspectives-on-network-adequacy-definitions
Dafny, Leemore S.; Hendel, Igal; Marone, Victoria; and Ody, Christopher (2017). Narrow Networks on the Health Insurance Marketplaces: Prevalence, Pricing, and the Cost of Network Breadth. Health Affairs 36(9). DOI:10.1377/hlthaff.2016.1669. https://www.healthaffairs.org/doi/10.1377/hlthaff.2016.1669
Liebman, Eli, and Panhans, Matthew T. (2021). Why Do Narrow Network Plans Cost Less? Health Economics 30(10):2437–2451. DOI:10.1002/hec.4385. https://doi.org/10.1002/hec.4385
Mazurenko, Olena; Taylor, Heather L.; and Menachemi, Nir (2022). The Impact of Narrow and Tiered Networks on Costs, Access, Quality, and Patient Steering: A Systematic Review. Medical Care Research and Review 79(5):607–617. DOI:10.1177/10775587211055923. https://pmc.ncbi.nlm.nih.gov/articles/PMC9817087/
Prager, Elena (2020). Healthcare Demand under Simple Prices: Evidence from Tiered Hospital Networks. American Economic Journal: Applied Economics 12(4):196–223. DOI:10.1257/app.20180422. https://www.aeaweb.org/articles?id=10.1257%2Fapp.20180422