Testable Hypotheses and Research Designs
The eight ideas below extend the chapter’s empirical foundation with falsifiable predictions and proposed policy experiments. These are not evidence — the evidence is presented above. These are the roadmap for testing whether healthcare reform actually works.
1 The Diagnostic Desert: Geographic Access as a Self-Reinforcing Trap
For a disease where travel triggers PEM, specialist access is governed by a radius constraint — and that radius is shorter than for any other chronic disease because PEM makes travel an active harm, not just a cost or inconvenience. This creates a self-reinforcing diagnostic desert: (a) patients in deserts cannot be diagnosed by specialists → fewer diagnoses recorded in desert regions → the region appears low-prevalence to health planners → no justification for deploying specialists; (b) patients in deserts who are diagnosed are self-selected — they have the health literacy, financial resources, and milder severity needed to navigate the gap — so the “diagnosed-in-desert” cohort is systematically healthier than the undiagnosed desert population, further suppressing regional prevalence estimates.
This mechanism is consistent with the 10× variation in ME/CFS diagnosis rates across English ICBs (Samms and Ponting 2025), but has not been tested directly. It predicts that PEM-adjusted travel impedance (distance × PEM severity) predicts diagnosis probability better than raw distance, and that this PEM-modulation is specific to ME/CFS — fibromyalgia, which lacks PEM as a travel barrier, should show weaker geographic clustering. (Certainty: 0.45 — geographic variation is documented; PEM-as-travel-barrier is mechanistically plausible; the self-reinforcing trap mechanism is novel and untested. Falsifiable prediction: in the UK, PEM-adjusted distance from the nearest ME/CFS specialist clinic inversely correlates with diagnosis rate at the ICB level, and this correlation is stronger for ME/CFS than for fibromyalgia. Evidence source: Samms 2025 administrative data; inference target: healthcare-system design.)
Consequence: If the diagnostic desert hypothesis is correct, opening a specialist clinic in an underserved region should produce a diagnostic surge — not because disease prevalence was lower there, but because 10×-underdiagnosed patients suddenly have access. Severity applicability: most relevant for severe/very severe patients whose PEM radius is short enough to make specialist access functionally impossible. (Origin: brainstorm)
2 Implementation Gap Mechanisms
All clinical guidelines take time to reach practice — the evidence-to-practice literature estimates 17 years on average. For contested illnesses, the cascade duration may be systematically longer because: (a) physicians have pre-existing beliefs that the guideline contradicts (belief revision requires more exposure than belief formation); (b) contested illnesses lack specialty champions — for a new oncology guideline, oncologists advocate for uptake at tumor boards; ME/CFS has no equivalent specialty presence in most hospitals; (c) the guideline’s main recommendation is negative (don’t recommend GET) — negative recommendations are harder to audit than positive ones. These mechanisms predict that at 5 years post-NICE 2021, the proportion of UK primary care physicians who correctly identify that GET is no longer recommended will be under 50% — exceeding the typical guideline uptake lag. The German G-BA statutory directive (regulatory force, not just advisory) provides a natural experiment: if binding guidelines produce faster cascade duration, German PEM-contraindication awareness should exceed UK levels. (Certainty: 0.35 — general implementation science supports the prediction; contested-illness amplification is novel; cascade duration for NICE 2021 is currently unobserved. Falsifiable prediction: German physicians show higher PEM-concern awareness than UK physicians at matched time-from-publication.)
Consequence: If the contested-disease cascade duration is 17+ years, every adult physician practicing today will retire before their training catches up to the NICE 2021 standard. Guideline publication without implementation is a publishing event, not a practice change. Severity applicability: all — the cascade duration affects management across all severity levels. (Origin: brainstorm)
The Bontempo 2025 invalidation model identifies healthcare system avoidance as a consequence of invalidation. The chapter documents that PEM prevents clinic participation. These two barriers interact synergistically: a patient who has been invalidated is reluctant to seek care, and a patient with PEM cannot easily access care even when motivated. The result is a bootstrap trap: the patient is too traumatized to trust doctors AND too fatigued to shop for a new one. Even when good care becomes available (a new specialist opens, a local GP reads NICE 2021), the patient cannot re-enter the system because past invalidation produces anticipatory dread, PEM makes provider-searching physiologically prohibitive, and each failed search costs unrecoverable energy.
This predicts a perverse effect: healthcare reform investments may primarily benefit patients with milder disease and shorter invalidation histories, while the most harmed patients — those with the longest invalidation histories and most severe PEM — remain outside the system. The trap would be refuted if a cohort study showed that healthcare avoidance does not interact with PEM severity to predict time-to-specialist-contact. (Certainty: 0.40 — both mechanisms individually documented (Bontempo, Bontempo, and Duberstein 2025); the synergy is logically implied but not directly observed; the perverse-effect prediction is the riskiest component. Falsifiable prediction: when a new specialist clinic opens, the first 50 patients to access it will have lower baseline invalidation scores and milder PEM than the regional ME/CFS population.)
Consequence: Building better clinics may not be enough — patients who have been burned by years of invalidation may need active outreach, not just better clinic availability. Severity applicability: most critical for severe/very severe patients with long disease duration, whose accumulated invalidation trauma and PEM severity create the strongest trap. (Origin: brainstorm)
3 Policy Experiment Designs
Five years after NICE NG206 (October 2021) explicitly withdrew GET and recognized PEM, no audit measures whether UK primary care practice changed. The UK’s centralized healthcare data infrastructure (CPRD covers ~20 million patients) makes this question answerable with existing administrative data. The design is an interrupted time series: pre-NICE (April 2018–September 2021) vs post-NICE (January 2022–December 2024), with primary outcomes including new ME/CFS diagnoses per 100,000 patient-years, proportion of ME/CFS-coded patients referred to physiotherapy (GET proxy), and diagnostic delay. Fibromyalgia and rheumatoid arthritis serve as control conditions — if referral/exercise patterns change for ME/CFS but not for controls, the change is guideline-attributable. Feasibility certainty: 0.75 — data exists, method standard, Samms 2025 establishes precedence (Samms and Ponting 2025). If no structural break is detected, the chapter’s implementation-gap concern is confirmed.
Consequence: This study is cheap, fast, and could have been done in 2022. The fact that no one has done it is itself evidence of how little health-services research attention ME/CFS receives. (Origin: brainstorm)
Geographic proximity to a specialist clinic can serve as an instrumental variable: patients who happen to live near a clinic use it more, and distance should affect outcomes only through clinic use. In ME/CFS, PEM-adjusted distance refines the instrument — the impedance is not raw miles but miles × PEM cost. Two-stage least squares would estimate: (first stage) distance predicts specialist clinic use; (second stage) specialist clinic use predicts diagnostic delay, functional status, and PEM-related emergency department visits. MS serves as a comparator (known specialist benefit, no PEM travel barrier). Feasibility certainty: 0.55 — method standard in health economics; UK data exists; US claims data are richer but poorer on functional outcomes. If the null hypothesis supported (Do Specialist Clinics Improve Outcomes?), specialist clinics do not measurably improve outcomes after controlling for selection. (Origin: brainstorm)
Consequence: This study would tell us whether specialist clinics do what advocates claim — improve outcomes — or whether their apparent benefit reflects the fact that wealthier, milder, more health-literate patients are the ones who reach them. The answer determines whether funding specialist clinics is evidence-based medicine or equity-worsening favoritism. Severity applicability: the quasi-experimental design automatically accounts for selection bias by severity, since PEM severity is one of the factors that selects patients into or out of near-clinic residence.
The chapter’s upstream-failure model implies that fixing education fixes outcomes. This needs testing. A cluster-randomized trial (n=20–30 primary care practices, 1:1) would compare a structured 2-hour ME/CFS education module (diagnostic criteria, PEM recognition, guideline awareness, EHR decision-support tool) to usual care, with new ME/CFS diagnoses per 10,000 registered patients over 18 months as the primary outcome. The pilot is powered to detect a doubling of diagnosis rate. If education increases clinician knowledge but not diagnosis rates, the barrier is structural (time pressure, referral-pathway absence), not educational — and the chapter’s upstream-failure model needs revision. Feasibility certainty: 0.50 — method straightforward but 18-month follow-up may be too short for a disease where diagnostic delay averages 5+ years.
Consequence: We spend decades arguing that doctors need ME/CFS education — but has anyone tested whether teaching them changes what they do? This pilot costs less than a single year of a specialist clinic budget and would answer the question that anchors the chapter’s first section. (Origin: brainstorm)
Standard healthcare delivery is PEM-generating: travel, waiting room overload, cognitive strain, physical exam. A randomized trial (n=200 newly diagnosed patients, two arms with identical clinical content) would test whether PEM-avoiding delivery (telehealth, asynchronous messaging, smartphone actigraphy, care coordinator for referrals, energy-conservation pre-visit instructions) produces better outcomes than standard in-person clinic care. Primary outcome: SF-36 physical function at 12 months; secondary: PEM episode frequency. The null (delivery channel doesn’t matter) is plausible; the alternative (PEM-avoiding delivery is superior on functional outcomes) would transform ME/CFS care delivery design. Severe/housebound patients are hypothesized to benefit most. Feasibility certainty: 0.55 — telemedicine infrastructure exists; patient blinding to delivery channel is impossible, requiring actigraphy as an objective check on subjective endpoints.
Consequence: If we design healthcare visits around what ME/CFS does to patients — instead of around clinic-schedule convenience — do patients get better? The answer determines whether PEM-aware delivery is a patient preference or a clinical effectiveness lever. (Origin: brainstorm)