TriMod™: Trio Overlap Correction for Multimorbidity Population Estimation
This paper is part of the CYNAERA Long COVID Library, a growing collection of research advancing how infection-associated chronic conditions are understood, measured, and addressed.
By Cynthia Adinig
Population burden estimates become increasingly unreliable when related chronic conditions are counted as though they occur independently. This problem extends across infection-associated chronic conditions (IACCs), autoimmune disease, autonomic disorders, cardiometabolic disease, neurological conditions, and other areas where multimorbidity is common. Large population studies and systematic reviews have repeatedly demonstrated that chronic diseases cluster within individuals, that those clusters are often nonrandom, and that estimated multimorbidity varies substantially depending on the conditions included, case definitions, source population, and ascertainment methodology (Barnett et al., 2012; Fortin et al., 2012; Prados-Torres et al., 2014; Willadsen et al., 2016; Ho et al., 2022).
TriMod™ is CYNAERA’s trio overlap correction module for estimating the unduplicated population represented across three overlapping conditions. Rather than summing three prevalence estimates and treating the result as the number of unique people affected, TriMod incorporates pairwise and triple overlap to distinguish diagnostic burden from unique population burden. The underlying set arithmetic is based on the established inclusion-exclusion principle. TriMod’s proprietary value lies in the epidemiological implementation surrounding that calculation, including condition-trio selection, overlap evidence evaluation, denominator harmonization, ascertainment adjustment, uncertainty construction, structural validation, and downstream deployment.
TriMod extends CYNAERA’s existing corrected prevalence architecture. US-CCUC™ addresses systematic undercounting and overlap across infection associated chronic condition populations, while the Diagnostic Multiplier™ models the portion of disease prevalence hidden by incomplete diagnostic capture. CYNAERA’s corrected ME/CFS prevalence modeling further demonstrated that correcting individual disease populations without accounting for their intersection with Long COVID could recreate overcounting at the aggregate level (Adinig, US-CCUC, 2026; Adinig, Diagnostic Multiplier, 2026; Adinig, ME/CFS Prevalence, 2026). TriMod isolates that overlap problem as a reusable analytical module that can operate inside CYNAERA systems or independently across other disease ecosystems.
The Population Counting Problem
Most combined disease burden calculations begin with a simple operation: Condition A population + Condition B population + Condition C population. That calculation accurately represents the combined number of disease memberships only when the analyst intends to count each diagnosis separately. It does not necessarily represent the number of unique people affected.
The distinction matters because multimorbidity is not a marginal phenomenon. A population-level analysis of more than 1.7 million people in Scotland found multimorbidity across age groups and demonstrated particularly high complexity among socially deprived populations (Barnett et al., 2012). International reviews have subsequently shown that multimorbidity prevalence estimates can vary dramatically depending on the number and selection of conditions examined, definitions used, population age, and setting (Fortin et al., 2012; Willadsen et al., 2016; Ho et al., 2022). These differences are not simply statistical noise. They reflect a more fundamental problem: diseases do not distribute themselves across human populations as independent categories.
The problem becomes obvious when the same individual carries several related diagnoses. A person with Long COVID, ME/CFS, and POTS may appear in the prevalence estimate for each condition. If all three national estimates are added together without adjustment, that individual is counted three times. Conversely, collapsing that individual into only one disease category would erase clinically meaningful diagnostic complexity. TriMod preserves both realities. A person may represent one unique patient while simultaneously contributing several legitimate disease memberships.
Disease Overlap Is Structured, Not Random
Disease combinations frequently occur in recognizable patterns rather than through random coincidence. A systematic review of associative multimorbidity identified recurring combinations of cardiovascular, metabolic, musculoskeletal, psychiatric, respiratory, and other conditions across different study populations, although methodological differences influenced the exact patterns identified (Prados-Torres et al., 2014). Research specifically examining disease triads has similarly shown that three-condition combinations can reveal clinically meaningful multimorbidity structures that are not visible when diseases are examined only one at a time (Schäfer et al., 2014).
More recent computational research has expanded this problem to much larger populations. An analysis of electronic health records from more than ten million people with multimorbidity in England identified disease clusters using both co-occurrence and temporal-sequence approaches, demonstrating that the organization of disease populations depends not only on which diagnoses occur together but also on the order in which they develop (Kuan et al., 2024). These findings reinforce the need to treat overlap as an explicit component of population estimation rather than assuming that individual disease prevalence values can simply be added together.
TriMod therefore treats overlap as a measurable population structure. The central analytical question is not merely whether conditions can coexist, but how frequently the same people occupy multiple disease populations and how confidently that intersection can be estimated.
The TriMod™ Framework
TriMod answers a specific population question: How many unique individuals are represented across three overlapping condition populations?
Let P(A), P(B), and P(C) represent the prevalence of three conditions. Let P(A∩B), P(A∩C), and P(B∩C) represent the corresponding pairwise overlaps, while P(A∩B∩C) represents the population belonging to all three condition groups.
The TriMod-adjusted burden is:
Pₜ = P(A) + P(B) + P(C) − P(A∩B) − P(A∩C) − P(B∩C) + P(A∩B∩C)
The pairwise intersections are subtracted because those individuals were counted more than once in the original prevalence sum. The triple intersection is then added back because individuals occupying all three populations were removed too many times during pairwise correction.
The mathematical identity underlying this calculation is the established inclusion-exclusion principle. TriMod does not claim ownership of inclusion-exclusion mathematics. The analytical difficulty lies in determining what values appropriately belong in the overlap terms, whether those values describe the target population, and how uncertainty or missing evidence should alter the final estimate.
Knowing the equation alone does not establish whether a POTS rate observed in a tertiary autonomic clinic can be generalized to the national Long COVID population. It does not establish whether ME/CFS cohorts using different diagnostic criteria describe comparable populations, whether self-reported and clinically confirmed diagnoses should receive equal evidentiary weight, or whether administrative claims adequately capture conditions known to experience long diagnostic delays. TriMod is designed around these implementation problems.
TriMod Within the CYNAERA Prevalence Architecture
TriMod emerged from a recurring challenge within CYNAERA prevalence modeling. US-CCUC™ was developed to estimate disease populations that remain incompletely visible to conventional surveillance and to avoid treating corrected IACC estimates as completely independent populations when substantial overlap exists (Adinig, US-CCUC, 2026). The framework recognizes that correcting Long COVID, ME/CFS, POTS, Lyme-associated chronic illness, and related populations separately can create a second-order problem if corrected estimates are subsequently added together without removing shared patients.
The Diagnostic Multiplier™ addresses an earlier point in the same analytical chain. It estimates disease populations that remain hidden because of incomplete diagnostic capture, including factors such as clinician recognition, referral dependency, diagnostic tool limitations, healthcare access, subgroup underrecognition, and structural barriers (Adinig, Diagnostic Multiplier, 2026). TriMod asks a different question. Once those condition populations have been estimated, how many times have the same people been represented across them?
CYNAERA’s corrected ME/CFS prevalence analysis illustrates why this distinction matters. Post-COVID ME/CFS cannot simply be added wholesale to Long COVID prevalence because a large portion of that population is already represented within Long COVID estimates. The corrected ME/CFS model therefore distinguishes condition-specific classification burden from the number of additional unique individuals that should enter a broader IACC population estimate (Adinig, ME/CFS Prevalence, 2026).
The resulting architecture can be represented as:
Observed population → diagnostic correction → condition prevalence → overlap correction → unique population estimate
TriMod occupies the overlap-correction layer.
IACCs as a High-Overlap Test Case
Infection-associated chronic conditions provide an especially useful demonstration of the problem because diagnostic overlap is common and often biologically plausible. Long COVID, ME/CFS, POTS and other dysautonomias, fibromyalgia, mast cell-associated disorders, small fiber neuropathy, autoimmune conditions, and connective tissue disorders can coexist within the same people. Post-infectious disease research also demonstrates that overlapping chronic syndromes are not unique to SARS-CoV-2, with persistent disease reported following numerous viral and bacterial infections (Choutka et al., 2022; Adinig, PCT Framework, 2026).
The RECOVER-Adult study provides one of the clearest contemporary examples. Among participants infected with SARS-CoV-2, post-COVID ME/CFS occurred substantially more frequently than among uninfected participants, and 88.7% of participants meeting criteria for post-COVID ME/CFS also met the RECOVER definition for Long COVID (Vernon et al., 2025). The result demonstrates why separately estimating Long COVID and post-COVID ME/CFS and then simply adding the populations would create substantial duplication.
POTS and ME/CFS provide another example. Autonomic research has long identified clinically meaningful overlap between postural tachycardia and chronic fatigue syndromes, while contemporary Long COVID research has expanded recognition of autonomic dysfunction after infection (Okamoto et al., 2012; Davis et al., 2023). These relationships make Long COVID, ME/CFS, and POTS an intuitive demonstration trio, but they also illustrate an important TriMod principle: evidence of overlap does not automatically establish the nationally representative size of that overlap.
A specialty-clinic cohort, RECOVER cohort, patient survey, electronic health record analysis, and national surveillance dataset may all reveal valid information about the same disease relationship while producing different percentages because they observe different populations. TriMod therefore does not treat published overlap percentages as interchangeable.
Evidence Hierarchy and Overlap Construction
The most important TriMod decision occurs before the calculation. The analyst must determine which evidence should populate the intersection terms. Direct population-based estimates with compatible definitions and denominators provide the strongest inputs when available. Large prospective cohorts can provide strong phenotyping but may contain recruitment effects. Electronic health records provide substantial scale while depending on diagnostic recognition and coding. Claims databases can quantify healthcare-recognized disease but may undercount conditions with long diagnostic delays. Specialty clinics provide rich clinical characterization but commonly overrepresent severe or complex patients. Patient-reported cohorts can capture populations missed by conventional care while introducing different forms of selection and classification uncertainty.
This is not a theoretical concern. Multimorbidity prevalence research has repeatedly shown that changing disease definitions, the number of conditions included, the source dataset, or the characteristics of the population can substantially change the resulting estimate (Fortin et al., 2012; Willadsen et al., 2016; Ho et al., 2022). Ho and colleagues' review of 193 international studies found extreme heterogeneity in reported multimorbidity prevalence, with age and the number of included conditions among the major contributors to variation (Ho et al., 2022).
TriMod therefore evaluates overlap evidence across multiple dimensions, including case-definition compatibility, population representativeness, diagnostic versus symptom-based ascertainment, source type, temporal alignment, age structure, geography, referral bias, disease severity, underdiagnosis risk, and denominator compatibility. The objective is not to force heterogeneous evidence into a single number. It is to determine which evidence can reasonably inform the target overlap and how much confidence should be placed in it.
TriMod Evidence Modes
TriMod can operate under several evidence environments.
Direct Overlap Mode is used when sufficiently compatible evidence exists for the relevant pairwise and triple intersections.
Mixed Evidence Mode applies when some disease relationships are well characterized while others remain uncertain.
Proxy-Informed Mode permits clinically adjacent or partially representative evidence to establish defensible bounds when direct population estimates are unavailable.
Scenario Mode generates conservative, moderate, and higher-overlap structures so that downstream decisions can be tested against alternative plausible assumptions. This approach allows uncertainty to remain visible. Emerging chronic illnesses rarely offer the perfectly harmonized national datasets required for precise intersection estimates. In those circumstances, a bounded population range can be more scientifically useful than a precise number constructed from uncertain assumptions. TriMod therefore treats range construction as an analytical feature rather than a methodological weakness.
Structural Validation
A proposed overlap structure must satisfy mathematical constraints before it can be accepted. A pairwise intersection cannot exceed the smaller of the two populations it joins. A triple intersection cannot exceed any pairwise intersection containing it. The final unique population cannot exceed the naïve sum of all three disease populations and cannot be smaller than the largest individual population.
These constraints prevent impossible combinations, but mathematical validity alone is insufficient. Epidemiological compatibility also matters. A calculation may be numerically correct while combining prevalence values from incompatible age groups, geographic populations, disease definitions, or time periods.
Methodological research on multimorbidity patterns demonstrates why this additional validation is necessary. Prados-Torres and colleagues found substantial methodological heterogeneity among studies identifying associative multimorbidity patterns (Prados-Torres et al., 2014). Schäfer and colleagues further demonstrated the usefulness of disease clusters and triads for representing multimorbidity while also illustrating the analytical complexity that emerges as disease combinations expand (Schäfer et al., 2014). TriMod therefore functions as an overlap-integrity layer, not simply an arithmetic calculator.
TriMod™ Applied to US-CCUC™ Prevalence
CYNAERA’s 2026 US-CCUC™ model estimates approximately 48.5–64.6 million U.S. adults with Long COVID, 18–26 million with ME/CFS, and 20–28 million with dysautonomia. The same model estimates that approximately 75–90 million Americans live with at least one IACC, including 25–35 million people experiencing multiple overlapping conditions, demonstrating why individually corrected disease estimates cannot simply be added together (Adinig, US-CCUC, 2026; Adinig, ME/CFS Prevalence, 2026).
Using the midpoint of the corrected prevalence ranges produces central planning populations of approximately 56.55 million for Long COVID, 22 million for ME/CFS, and 24 million for dysautonomia. Without any overlap correction, these three conditions alone would produce:
56.55M + 22M + 24M = 102.55M disease memberships
That figure does not mean 102.55 million unique people have one of these three conditions. US-CCUC™ identifies substantial co-occurrence across major IACC pairs, with documented overlap rates frequently ranging from approximately 40% to 80% depending on the conditions and population studied (Adinig, US-CCUC, 2026). A substantial portion of the ME/CFS and dysautonomia populations is therefore already represented within the Long COVID population, while additional patients occupy both the ME/CFS and dysautonomia populations.
TriMod corrects this structure using:
Pₜ = 102.55M − P(LC∩ME/CFS) − P(LC∩Dysautonomia) − P(ME/CFS∩Dysautonomia) + P(LC∩ME/CFS∩Dysautonomia)
The three pairwise intersections remove duplicated patient counts, while the final three-way intersection restores patients who would otherwise be removed too many times. The resulting value represents the number of unique individuals, while the original 102.55 million represents cumulative disease memberships.
TriMod™ 2026 US-CCUC Input Table
Condition / Population Measure | 2026 US-CCUC Range | Central Planning Estimate | Role in TriMod |
Long COVID | 48.5–64.6M | 56.55M | Base condition population |
ME/CFS | 18–26M | 22M | Overlapping condition population |
Dysautonomia | 20–28M | 24M | Overlapping condition population |
Naïve Combined Total | 86.5–118.6M | 102.55M | Disease memberships before overlap correction |
Americans with ≥1 IACC | 75–90M | 82.5M | US-CCUC unique-population benchmark across all ten IACCs |
Americans with Multiple IACCs | 25–35M | 30M | US-CCUC multimorbidity benchmark |
This formulation preserves the full diagnostic burden represented by the three conditions while preventing the same individuals from being counted repeatedly when the analytical objective is to estimate unique population burden.
Example Pair and Trio Applications
TriMod can be applied at both the two-condition and three-condition level. CYNAERA’s 2026 US-CCUC™ model estimates 48.5–64.6 million adults with Long COVID, 18–26 million with ME/CFS, and 20–28 million with dysautonomia, while documenting substantial co-occurrence across major IACC pairs. The examples below demonstrate how those updated prevalence estimates can be translated into overlap-aware planning scenarios rather than treated as mutually exclusive populations (Adinig, US-CCUC, 2026; Adinig, ME/CFS Prevalence, 2026).
Population Combination | Individual Prevalence Inputs | Naïve Disease Memberships | TriMod Interpretation |
Long COVID + ME/CFS | 48.5–64.6M + 18–26M | 66.5–90.6M | A substantial portion of ME/CFS is already represented within Long COVID, so the unique population is materially lower than the summed total. |
Long COVID + Dysautonomia | 48.5–64.6M + 20–28M | 68.5–92.6M | Dysautonomia frequently occurs within the Long COVID population, requiring pairwise overlap correction before estimating unique individuals. |
ME/CFS + Dysautonomia | 18–26M + 20–28M | 38–54M | Clinical overlap between neuroimmune and autonomic illness means these populations cannot be assumed to be independent. |
Long COVID + ME/CFS + Dysautonomia | 48.5–64.6M + 18–26M + 20–28M | 86.5–118.6M | TriMod removes the three pairwise intersections and restores the three-way intersection to estimate the unique population represented across all three conditions. |
At the midpoint of the current US-CCUC™ ranges, the Long COVID, ME/CFS, and dysautonomia trio contains approximately 102.55 million disease memberships before overlap correction:
56.55M + 22M + 24M = 102.55M
The trio is then corrected as:
Pₜ = 102.55M − P(LC∩ME/CFS) − P(LC∩Dysautonomia) − P(ME/CFS∩Dysautonomia) + P(LC∩ME/CFS∩Dysautonomia)
Pairwise applications use the simpler two-condition form:
P(A∪B) = P(A) + P(B) − P(A∩B)
For example, the midpoint Long COVID and ME/CFS populations total 78.55 million disease memberships before correction, while Long COVID and dysautonomia total 80.55 million, and ME/CFS and dysautonomia total 46 million. Each of those totals represents diagnostic memberships rather than automatically representing an equivalent number of unique people.
This distinction becomes increasingly important as conditions are combined. A pair requires correction for one intersection. A trio requires correction for three pairwise intersections and restoration of the population occupying all three conditions. TriMod therefore allows researchers, policymakers, health systems, and market analysts to move from simple disease summation toward a population structure that reflects real-world multimorbidity.
Diagnostic Burden and Unique Population Burden
The distinction between diagnostic burden and unique population burden is central to TriMod. One person living with Long COVID, ME/CFS, and POTS represents one human being but three legitimate disease memberships. Both measurements can be useful. Diagnostic burden may be more appropriate when estimating specialty care needs, medications, diagnostic testing, clinical workload, disease-specific research demand, or treatment utilization. Unique population burden may be more appropriate when estimating how many people are affected, workforce participation, household impact, disability populations, geographic distribution, coalition reach, insurance populations, or addressable markets.
The problem arises when one quantity is mistaken for the other.
A coalition representing three diseases might legitimately represent 40 million disease memberships but only 29 million unique individuals. A health system serving one million multimorbid patients might need considerably more than one million specialty encounters. A pharmaceutical company may face several diagnostic opportunities within the same patient while having a smaller unique addressable population than naïve disease summation suggests.
TriMod allows those quantities to remain analytically separate.
Beyond IACCs
Although TriMod emerged from CYNAERA’s IACC prevalence work, its architecture is condition-agnostic. Multimorbidity is documented across cardiovascular, metabolic, respiratory, musculoskeletal, neurological, psychiatric, renal, oncologic, and autoimmune disease. Barnett and colleagues demonstrated the scale of this problem using records from more than 1.7 million people, showing that multimorbidity affects large sections of the population and occurs earlier in more socioeconomically deprived communities (Barnett et al., 2012).
Disease clustering research further demonstrates that multimorbidity can form recognizable structures rather than simple random accumulation (Prados-Torres et al., 2014; Schäfer et al., 2014). Large-scale English primary care data have now extended disease-clustering analysis to more than ten million people and more than 200 conditions (Kuan et al., 2024). TriMod can therefore be applied to autoimmune multimorbidity, cardiovascular-metabolic-renal populations, neurological conditions, mental and physical health overlap, oncology comorbidities, maternal health, disability populations, insurance risk modeling, pharmaceutical market sizing, clinical trial recruitment, and other areas where three diagnostic populations substantially intersect.
CYNAERA’s broader prevalence work already demonstrates that correction architecture can move across disease categories. The Diagnostic Multiplier™ is explicitly condition-agnostic, while CYNAERA prevalence modeling has been applied to autoimmune disease, military populations, Long COVID, ME/CFS, Lyme-associated illness, and other disease environments (Adinig, Diagnostic Multiplier, 2026; Adinig, Lupus Prevalence, 2026; Adinig, US-CCUC Military, 2026).
Health Access Hurdles
Observed overlap is also shaped by who receives diagnoses in the first place. Diagnostic visibility is not evenly distributed across disease populations or demographic groups. Access to specialty care, clinician recognition, referral requirements, healthcare coverage, geographic availability, sex and gender bias, racial disparities, and socioeconomic barriers can all influence whether a disease intersection becomes visible in a dataset (Adinig, Diagnostic Multiplier, 2026).
TriMod therefore distinguishes observed overlap from plausible underlying overlap when the evidence supports doing so. Two conditions may appear to intersect less frequently in a population not because the biological relationship is weaker, but because one of the conditions is systematically underdiagnosed. This creates an important interface between TriMod and the Diagnostic Multiplier™. Diagnostic Multiplier can estimate how much disease may be missing from observed prevalence, while TriMod evaluates how those corrected disease populations intersect. In health equity applications, the combination can help identify situations where conventional datasets understate both the size and complexity of affected populations.
Temporal Overlap and Disease Evolution
Disease overlap is not necessarily static. A patient may first enter one diagnostic population and later acquire a second or third diagnosis as disease evolves, symptoms emerge, diagnostic criteria are met, or healthcare access improves. This is particularly relevant in infection-associated chronic illness, where chronic disease states may develop over months or years following an initiating infection or other biological trigger (Adinig, PCT Framework, 2026; Choutka et al., 2022).
CYNAERA’s Primary Chronic Trigger Framework models the identification of potential ignition events and subsequent progression into chronic disease, while Stage Zero™ addresses measurable physiological instability that may exist before formal diagnostic thresholds are reached (Adinig, PCT Framework, 2026; Adinig, Stage Zero, 2026). These models suggest that the relationship between prevalence populations can change across time even when the underlying patient cohort remains the same. A person may initially be represented only within Long COVID prevalence, subsequently meet criteria for POTS, and later meet criteria for ME/CFS. Cross-sectional measurement at each stage would produce a different TriMod structure.
Large-scale computational research increasingly supports incorporating time into multimorbidity analysis. Kuan and colleagues identified disease clusters using not only co-occurrence but also disease-development sequence, demonstrating that temporal order contains information that cross-sectional overlap alone cannot capture (Kuan et al., 2024). TriMod can therefore be recalibrated across disease duration or calendar periods when evidence supports time-dependent overlap estimates.
TriMod as AI Infrastructure
The overlap problem becomes increasingly important as artificial intelligence and automated analytics are used to synthesize epidemiological information. An AI system can retrieve several individually valid prevalence estimates and still generate a materially invalid population total if it assumes those populations are additive. For example, an automated system might retrieve a Long COVID prevalence estimate, an ME/CFS estimate, and a POTS estimate from three credible sources. Each source may be accurate within its own methodology. The combined number can nevertheless be wrong as an estimate of unique people because the same patients occupy multiple populations.
TriMod introduces an explicit reasoning layer between prevalence retrieval and downstream calculation:
Condition prevalence → evidence compatibility → overlap structure → unique population estimate → downstream model
This architecture can support automated policy analysis, economic burden modeling, payer analytics, clinical trial recruitment, pharmaceutical market sizing, research prioritization, and health-system forecasting. The purpose is not merely to improve epidemiological arithmetic. It is to prevent duplicated population assumptions from being scaled automatically across decision systems.
Applications
TriMod can support national burden estimates, policy briefs, healthcare planning, economic modeling, funding justification, coalition strategy, market sizing, health equity analysis, clinical trial planning, payer modeling, research portfolio design, AI population intelligence, and other applications where multiple condition populations intersect. For policymakers, overlap correction can prevent aggregate disease estimates from overstating the number of unique constituents while preserving the greater service requirements associated with multimorbidity. For health systems, TriMod can separate patient population size from diagnostic and specialty demand. For research portfolios, it can reveal when nominally separate disease programs serve substantially overlapping populations. For advocacy coalitions, it can calculate collective reach without artificially inflating the number of individuals represented.
The market application is particularly important. A company may identify 12 million people in one diagnostic market, 8 million in another, and 5 million in a third and report a 25-million-person addressable market. If millions of people occupy two or three of those categories, 25 million may represent diagnostic opportunities rather than unique customers. TriMod allows those market concepts to be separated.
Modular Architecture
TriMod is intentionally designed around three-condition structures. Pairwise analysis is useful but cannot fully characterize situations where one person simultaneously occupies three populations. Conversely, higher-order inclusion-exclusion models become increasingly complex as more conditions are added because the number of possible intersections grows rapidly.
Disease trios provide a practical middle layer. They are complex enough to capture meaningful multimorbidity while remaining interpretable to researchers, policymakers, health systems, advocates, and commercial users. The existing multimorbidity literature has independently demonstrated the analytical usefulness of disease triads for representing population disease patterns (Schäfer et al., 2014).
TriMod units can also be linked when broader multimorbidity ecosystems must be modeled. A larger disease environment can be decomposed into clinically meaningful trios, analyzed separately, and incorporated into higher-order CYNAERA systems where necessary.
TriMod currently functions as a Core Analytical Module embedded within US-CCUC™ and PULSE™. Within US-CCUC™, TriMod can prevent corrected prevalence estimates from recreating overcounting when multiple related chronic conditions are aggregated. Within PULSE™, it can support population and market intelligence where several diagnostic categories correspond to overlapping human populations.
Licensing and Deployment
TriMod can be deployed through multiple licensing structures.
A TriMod Core License can provide the analytical architecture to organizations with their own validated prevalence and overlap datasets.
Disease Calibration Packs can provide CYNAERA-developed overlap structures, evidence hierarchies, parameter ranges, and validation rules for specific condition trios.
Research Licenses can support academic, nonprofit, and collaborative research use.
Government and Institutional Licenses can support federal agencies, state governments, health systems, public health organizations, and payers.
Embedded Technology Licensing can allow TriMod to function inside AI tools, population-health systems, epidemiological platforms, market intelligence products, and commercial analytics environments.
Enterprise and API Licensing can support repeated automated analysis across multiple disease combinations, geographic populations, datasets, or market scenarios. Separating the core architecture from disease-specific calibration creates an important licensing distinction. Inclusion-exclusion mathematics is established and publicly available.
The defensible value of TriMod lies in the evidence architecture surrounding it: disease-trio construction, intersection estimation, denominator harmonization, diagnostic-capture interaction, uncertainty construction, structural validation, recalibration logic, and domain-specific implementation.
A calibrated Long COVID–ME/CFS–POTS TriMod package, for example, is analytically different from a cardiovascular-kidney-diabetes package even though both use the same underlying set arithmetic. The evidence, diagnostic relationships, population assumptions, and validation constraints differ. That creates a modular licensing model in which the analytical engine and the calibrated evidence layers can be deployed separately.
Limitations
TriMod does not diagnose individuals, determine disease causation, or establish that overlapping conditions share identical pathophysiology. It does not assume that a disease association observed in a specialty cohort represents a nationally generalizable overlap rate. It also cannot eliminate bias already present within the base prevalence estimates. The reliability of TriMod therefore depends on the quality of its inputs. Conditions with substantial diagnostic delay, inconsistent criteria, weak surveillance, or rapidly evolving definitions may require wider uncertainty intervals. Disease intersections may also change over time as people acquire diagnoses or move between disease states.
TriMod is designed to expose these limitations rather than conceal them. Where evidence supports only a range, the appropriate output is a range. Where an intersection is based on proxy evidence, it should be identified as modeled rather than directly observed. Where available evidence is insufficient to support a credible correction, the model should preserve that uncertainty rather than manufacture precision.
Conclusion
Modern disease populations do not exist in clean, mutually exclusive diagnostic silos. Decades of multimorbidity research demonstrate that chronic illnesses cluster within individuals, that those relationships are frequently nonrandom, and that estimates can change substantially according to disease selection, case definition, data source, study population, and analytical method (Barnett et al., 2012; Fortin et al., 2012; Prados-Torres et al., 2014; Willadsen et al., 2016; Ho et al., 2022).
IACCs make the consequences of that problem particularly visible. Long COVID, ME/CFS, POTS, and other chronic conditions can occupy substantially overlapping patient populations, making naïve prevalence summation inappropriate when the desired output is the number of unique people affected (Vernon et al., 2025; Adinig, ME/CFS Prevalence, 2026; Adinig, US-CCUC, 2026). TriMod™ isolates that problem into a reusable analytical module. It preserves the distinction between the number of diagnoses represented within a population and the number of unique people carrying those diagnoses. The set mathematics required to remove duplicate counts is established. Determining which populations truly overlap, which evidence is transferable, how diagnostic invisibility affects observed intersections, how uncertainty should be bounded, how disease relationships change over time, and how corrected populations should move into downstream systems is considerably harder.That is the analytical layer TriMod was built to solve.
Appendix A. Recalibrated IACC Pair and Trio Estimates
The following estimates provide a 2026 planning application of TriMod™ across commonly overlapping infection-associated chronic conditions. CYNAERA’s updated US-CCUC™ methodology estimates that approximately 75–90 million Americans live with at least one IACC, including 25–35 million experiencing multiple overlapping conditions. Corrected condition estimates include Long COVID at 48.5–64.6 million adults, ME/CFS at 18–26 million adults, dysautonomia at 20–28 million adults, MCAS at 20–28 million adults, hypermobile Ehlers-Danlos syndrome at 12–18 million adults, fibromyalgia at 13–18 million adults, small fiber neuropathy at 6.5–8.5 million adults, Sjögren’s syndrome at 7–10 million adults, chronic Lyme disease and post-treatment Lyme disease syndrome at 5–7 million adults, and PANS/PANDAS at 2–4 million children (Adinig, US-CCUC, 2026). CYNAERA’s POTS-specific prevalence model provides a separate planning range of approximately 14–18 million, with a central estimate of 16.5 million (Adinig, US-CUCC POTS, 2025).
The estimates below are TriMod modeled planning estimates rather than directly observed national comorbidity prevalence. They are constructed from CYNAERA’s current condition-specific prevalence estimates, published evidence of substantial IACC co-occurrence, and structural constraints that prevent pairwise or three-way intersections from exceeding the populations that contain them. US-CCUC™ identifies co-occurrence rates of approximately 40–80% across major IACC condition pairs, providing a broad empirical range for modeling high-overlap disease populations when harmonized national intersection data are unavailable (Adinig, US-CCUC, 2026).
TriMod applies two fundamental structural rules. First, a pairwise intersection cannot exceed the prevalence of the smaller condition in that pair. Second, a three-way intersection cannot exceed any of the three pairwise intersections that contain it. These constraints prevent mathematically impossible overlap structures while allowing uncertainty to remain explicit.
Common Disease Pairings and 2026 TriMod™ Estimates
Condition Pairing | 2026 TriMod™ Modeled Overlap |
Long COVID + ME/CFS | ~17.6 million |
Long COVID + Dysautonomia | ~14.5 million |
Long COVID + POTS | ~12.4 million |
Long COVID + MCAS | ~9.4 million |
ME/CFS + Dysautonomia | ~15.4 million |
ME/CFS + MCAS | ~14.3 million |
ME/CFS + POTS | ~9.8 million |
POTS + MCAS | ~5.7 million |
POTS + hEDS | ~12.0 million |
MCAS + hEDS | ~12.0 million |
SFN + POTS | ~6.0 million |
Sjögren’s Syndrome + MCAS | ~6.8 million |
Chronic Lyme / PTLDS + ME/CFS | ~4.8 million |
Chronic Lyme / PTLDS + MCAS | ~4.1 million |
PANS/PANDAS + MCAS | ~2.0 million* |
These pairings illustrate why condition-specific prevalence estimates cannot be interpreted as mutually exclusive populations. Long COVID and ME/CFS, for example, have central prevalence estimates of approximately 56.55 million and 22 million, respectively. Simply adding them produces 78.55 million disease memberships, but a modeled intersection of approximately 17.6 million indicates that a substantial number of people are represented within both disease populations.
The same distinction applies to autonomic illness. Long COVID and dysautonomia produce approximately 80.55 million disease memberships at their central prevalence estimates, yet TriMod models approximately 14.5 million individuals as occupying both populations. Long COVID and POTS similarly produce approximately 73.05 million disease memberships using the 56.55-million Long COVID midpoint and 16.5-million POTS point estimate, while the modeled shared population is approximately 12.4 million.
High overlap is also visible outside Long COVID. ME/CFS and dysautonomia are modeled at approximately 15.4 million shared individuals, reflecting the substantial autonomic component documented across ME/CFS populations. ME/CFS and MCAS are modeled at approximately 14.3 million, while POTS and hEDS are modeled at approximately 12 million. These estimates should not be interpreted as proof that one condition causes another. They represent population intersections useful for estimating unique burden across clinically overlapping disease categories.
Common Disease Trios and 2026 TriMod™ Estimates
Three-way overlap requires additional correction because the same person can simultaneously occupy all three pairwise intersections. Every trio estimate must therefore remain smaller than or equal to the smallest pairwise intersection composing that trio.
Condition Trio | 2026 TriMod™ Modeled Overlap |
Long COVID + ME/CFS + Dysautonomia | ~14.0 million |
Long COVID + ME/CFS + POTS | ~9.0 million |
Long COVID + POTS + MCAS | ~5.2 million |
ME/CFS + POTS + MCAS | ~5.0 million |
POTS + MCAS + hEDS | ~5.0 million |
ME/CFS + POTS + SFN | ~5.0 million |
Chronic Lyme / PTLDS + ME/CFS + MCAS | ~3.4 million |
The largest modeled trio is Long COVID + ME/CFS + dysautonomia at approximately 14 million people. This estimate remains below the modeled Long COVID–ME/CFS intersection of 17.6 million, Long COVID–dysautonomia intersection of 14.5 million, and ME/CFS–dysautonomia intersection of 15.4 million. The three-way population is therefore structurally compatible with each of its constituent pairs.
The Long COVID + ME/CFS + POTS trio is modeled at approximately 9 million people. That value remains below the Long COVID–ME/CFS intersection of 17.6 million, Long COVID–POTS intersection of 12.4 million, and ME/CFS–POTS intersection of 9.8 million. This represents a particularly important TriMod use case because an individual living with all three conditions is simultaneously represented within three pairwise disease combinations and three individual prevalence estimates.
Likewise, the Long COVID + POTS + MCAS population is modeled at approximately 5.2 million people, constrained principally by the approximately 5.7-million POTS–MCAS intersection. The ME/CFS + POTS + MCAS and POTS + MCAS + hEDS trios are each modeled at approximately 5 million people, remaining below all relevant pairwise boundaries.
The chronic Lyme / PTLDS–ME/CFS–MCAS population is modeled more conservatively at approximately 3.4 million, reflecting both the smaller corrected chronic Lyme/PTLDS population and the requirement that the three-way intersection remain below the estimated Lyme–ME/CFS and Lyme–MCAS pairwise populations.
Illustrative TriMod™ Trio Calculation
The Long COVID–ME/CFS–POTS trio demonstrates why both pairwise and three-way corrections are necessary. Using central planning values:
Long COVID = 56.55MME/CFS = 22MPOTS = 16.5M
The naïve combined total is:
56.55M + 22M + 16.5M = 95.05M disease memberships
TriMod then incorporates the modeled pairwise intersections:
Long COVID ∩ ME/CFS = 17.6M
Long COVID ∩ POTS = 12.4M
ME/CFS ∩ POTS = 9.8M
and the modeled three-way intersection:
Long COVID ∩ ME/CFS ∩ POTS = 9.0M
The corrected calculation is:
95.05M − 17.6M − 12.4M − 9.8M + 9.0M = 64.25M
Under this TriMod planning scenario, approximately 95.05 million disease memberships correspond to approximately 64.25 million unique individuals across the Long COVID, ME/CFS, and POTS populations. The difference of approximately 30.8 million counts does not represent invalid diagnoses. It represents repeated population membership created when the same individuals appear across two or three disease categories. A person living with all three conditions remains one individual for unique population estimation while retaining three legitimate diagnoses for clinical, research, service-utilization, and economic planning.
Illustrative TriMod™ Pair Calculations
TriMod can also be used for two-condition estimates. For a pair, the calculation simplifies to:
P(A∪B) = P(A) + P(B) − P(A∩B)
For Long COVID and ME/CFS:
56.55M + 22M − 17.6M = 60.95M unique individuals
The two conditions therefore represent approximately 78.55 million disease memberships but 60.95 million unique people under the modeled overlap scenario.
For Long COVID and POTS:
56.55M + 16.5M − 12.4M = 60.65M unique individuals
This corresponds to approximately 73.05 million disease memberships but 60.65 million unique people.
For ME/CFS and dysautonomia:
22M + 24M − 15.4M = 30.6M unique individuals
The two conditions therefore represent approximately 46 million disease memberships but 30.6 million unique people. These examples demonstrate how the apparent size of a population can change materially depending on whether the analytical unit is the diagnosis or the person.
Interpretation and Evidence Status
The estimates in this appendix should be interpreted as 2026 TriMod planning estimates, not as direct nationally measured comorbidity prevalence. National surveillance systems do not currently provide harmonized estimates for every IACC pair and trio using consistent diagnostic criteria, age groups, sampling methods, and ascertainment standards. TriMod therefore combines corrected condition prevalence with evidence-informed overlap assumptions and explicit structural constraints. The confidence associated with individual estimates will vary. Relationships such as Long COVID–ME/CFS, Long COVID–dysautonomia, ME/CFS–POTS, and POTS–hEDS have comparatively strong clinical and epidemiological support for substantial overlap, while other combinations have less nationally representative evidence. TriMod is designed so that modeled intersections can be replaced with directly measured values as stronger registry, cohort, claims, EHR, or surveillance data become available.
POTS is treated separately from the broader dysautonomia category in this appendix. CYNAERA’s POTS-specific model estimates approximately 14–18 million Americans with POTS, with a central estimate of 16.5 million, while the June 2026 US-CCUC™ model estimates approximately 20–28 million adults with dysautonomia overall. POTS should therefore be understood as a major subset of the dysautonomia population rather than an additional population that can be added independently to it.
* PANS/PANDAS + MCAS: The US-CCUC™ PANS/PANDAS prevalence estimate applies to children, while the current US-CCUC™ MCAS estimate is reported for adults. The approximately 2-million overlap value should therefore be considered a provisional planning estimate rather than an age-harmonized national intersection.
Gulf War Illness combinations are not assigned updated numerical values in this appendix because a current 2026 CYNAERA prevalence denominator was not included in the June 2026 US-CCUC™ model. They can be added once an updated prevalence basis is established.
Overall, these estimates illustrate the central function of TriMod™. Multimorbidity creates two simultaneous realities: the healthcare system must manage every legitimate diagnosis, while population planning must avoid repeatedly counting the same people. TriMod preserves both by distinguishing disease memberships from unique individuals and ensuring that pairwise and three-way overlap estimates remain mathematically and epidemiologically coherent.
CYNAERA Framework Papers
This paper draws on a defined subset of CYNAERA Institute white papers that establish the methodological and analytical foundations of CYNAERA’s frameworks. These publications provide deeper context on prevalence reconstruction, remission, combination therapies and biomarker approaches. Our Long COVID Library, ME/CFS Library, Lyme Library, Autoimmune Library and CRISPR Remission Library are also in depth resources.
Author’s Note:
All insights, frameworks, and recommendations in this written material reflect the author's independent analysis and synthesis. References to researchers, clinicians, and advocacy organizations acknowledge their contributions to the field but do not imply endorsement of the specific frameworks, conclusions, or policy models proposed herein. This information is not medical guidance.
Patent-Pending Systems
Bioadaptive Systems Therapeutics™ (BST) and affiliated CYNAERA frameworks are protected under U.S. Provisional Patent Application No. 63/909,951. CYNAERA is built as modular intelligence infrastructure designed for licensing, integration, and strategic deployment across health, research, public sector, and enterprise environments.
Licensing and Integration
CYNAERA supports licensing of individual modules, bundled systems, and broader architecture layers. Current applications include research modernization, trial stabilization, diagnostic innovation, environmental forecasting, and population level modeling for complex chronic conditions. Basic licensing is available through CYNAERA Market, with additional pathways for pilot programs, institutional partnerships, and enterprise integration.
About the Author
Cynthia Adinig is the founder of CYNAERA, a modular intelligence infrastructure company that transforms fragmented real world data into predictive insight across healthcare, climate, and public sector risk environments. Her work sits at the intersection of AI infrastructure, federal policy, and complex health system modeling, with a focus on helping institutions detect hidden costs, anticipate service demand, and strengthen planning in high uncertainty environments.
Cynthia has contributed to federal health and data modernization efforts spanning HHS, NIH, CDC, FDA, AHRQ, and NASEM, and has worked with congressional offices including Senator Tim Kaine, Senator Ed Markey, Representative Don Beyer, and Representative Jack Bergman on legislative initiatives related to chronic illness surveillance, healthcare access, and data infrastructure. In 2025, she was appointed to advise the U.S. Department of Health and Human Services and has testified before Congress on healthcare data gaps and system level risk.
She is a PCORI Merit Reviewer, currently advises Selin Lab at UMass Chan, and has co-authored research with Harlan Krumholz, MD, Akiko Iwasaki, PhD, and David Putrino, PhD, including through Yale’s LISTEN Study. She also advised Amy Proal, PhD’s research group at Mount Sinai through its CoRE advisory board and has worked with Dr. Peter Rowe of Johns Hopkins on national education and outreach focused on post-viral and autonomic illness. Her CRISPR Remission™ abstract was presented at CRISPRMED26 and she has authored a Milken Institute essay on artificial intelligence and healthcare.
Cynthia has been covered by outlets including TIME, Bloomberg, Fortune, and USA Today for her policy, advocacy, and public health work. Her perspective on complex chronic conditions is also informed by lived experience, which sharpened her commitment to reforming how chronic illness is understood, studied, and treated. She also advocates for domestic violence prevention and patient safety, bringing a trauma informed lens to her research, systems design, and policy work. Based in Northern Virginia, she brings more than a decade of experience in strategy, narrative design, and systems thinking to the development of cross sector intelligence infrastructure designed to reduce uncertainty, improve resilience, and support institutional decision making at scale.
References
Adinig, C. (2026). ME/CFS Prevalence in the Long COVID Era: A Corrected U.S. Burden Estimate Using Infection-Associated Chronic Condition Modeling. SSRN. https://doi.org/10.2139/ssrn.6835159
Adinig, C. (2026). The Diagnostic Multiplier™: A Framework for Adjusting Disease Prevalence Based on Real-World Diagnostic Capture. SSRN. SSRN 6812338.
Adinig, C. (2026). The Primary Chronic Trigger Framework: A Mathematical Blueprint for Detecting Ignition Events and Modeling Burden in Infection-Associated Chronic Conditions. SSRN. https://doi.org/10.2139/ssrn.6892100
Adinig, C. (2026). Stage Zero: Preclinical Immune and Autonomic Instability Preceding Infection-Associated Chronic Conditions. SSRN. SSRN 6903458.
Adinig, C. (2026). US-CCUC: Correcting Surveillance-Based Prevalence Undercounting Across Infection-Associated Chronic Conditions. SSRN. https://doi.org/10.2139/ssrn.6967838
Adinig, C. (2026). Long COVID Prevalence in U.S. Military and Veteran Populations: Corrected National Estimates Using US-CCUC Military. SSRN. https://doi.org/10.2139/ssrn.7004898
Adinig, C. (2026). Corrected Lupus Prevalence Modeling. SSRN. SSRN 6837279.
Barnett, K., Mercer, S. W., Norbury, M., Watt, G., Wyke, S., and Guthrie, B. (2012). Epidemiology of multimorbidity and implications for health care, research, and medical education: A cross-sectional study. The Lancet, 380(9836), 37-43. doi:10.1016/S0140-6736(12)60240-2.
Choutka, J., Jansari, V., Hornig, M., and Iwasaki, A. (2022). Unexplained post-acute infection syndromes. Nature Medicine, 28(5), 911-923. doi:10.1038/s41591-022-01810-6.
Davis, H. E., Assaf, G. S., McCorkell, L., et al. (2021). Characterizing Long COVID in an international cohort: Seven months of symptoms and their impact. EClinicalMedicine, 38, 101019. doi:10.1016/j.eclinm.2021.101019.
Davis, H. E., McCorkell, L., Vogel, J. M., and Topol, E. J. (2023). Long COVID: Major findings, mechanisms and recommendations. Nature Reviews Microbiology, 21(3), 133-146. doi:10.1038/s41579-022-00846-2.
Fortin, M., Stewart, M., Poitras, M. E., Almirall, J., and Maddocks, H. (2012). A systematic review of prevalence studies on multimorbidity: Toward a more uniform methodology. Annals of Family Medicine, 10(2), 142-151. doi:10.1370/afm.1337.
Ho, I. S. S., Azcoaga-Lorenzo, A., Akbari, A., et al. (2022). Variation in the estimated prevalence of multimorbidity: Systematic review and meta-analysis of 193 international studies. BMJ Open, 12(4), e057017. doi:10.1136/bmjopen-2021-057017.
Institute of Medicine. (2015). Beyond Myalgic Encephalomyelitis/Chronic Fatigue Syndrome: Redefining an Illness. National Academies Press. doi:10.17226/19012.
Komaroff, A. L., and Lipkin, W. I. (2023). Insights from myalgic encephalomyelitis/chronic fatigue syndrome may help unravel the pathogenesis of Long COVID. Trends in Molecular Medicine, 29(8), 689-692. doi:10.1016/j.molmed.2023.04.002.
Kuan, V., et al. (2024). Identifying multi-resolution clusters of diseases in ten million patients with multimorbidity in primary care in England. Communications Medicine, 4, 102.
National Academies of Sciences, Engineering, and Medicine. (2024). Long-Term Health Effects of COVID-19: Disability Assessment Challenges and Opportunities. National Academies Press. doi:10.17226/27756.
Okamoto, L. E., Raj, S. R., Peltier, A., Gamboa, A., Shibao, C., Diedrich, A., Black, B. K., Robertson, D., and Biaggioni, I. (2012). Neurohumoral and haemodynamic profile in postural tachycardia and chronic fatigue syndromes. Clinical Science, 122(4), 183-192. doi:10.1042/CS20110200.
Prados-Torres, A., Calderón-Larrañaga, A., Hancco-Saavedra, J., Poblador-Plou, B., and van den Akker, M. (2014). Multimorbidity patterns: A systematic review. Journal of Clinical Epidemiology, 67(3), 254-266. doi:10.1016/j.jclinepi.2013.09.021.
Schäfer, I., Kaduszkiewicz, H., Wagner, H. O., Schön, G., Scherer, M., and van den Bussche, H. (2014). Reducing complexity: A visualisation of multimorbidity by combining disease clusters and triads. BMC Public Health, 14, 1285. doi:10.1186/1471-2458-14-1285.
Vernon, S. D., Zheng, T., Do, H., et al. (2025). Incidence and prevalence of post-COVID-19 myalgic encephalomyelitis: A report from the observational RECOVER-Adult study. Journal of General Internal Medicine, 40, 1085-1094. doi:10.1007/s11606-024-09290-9.
Willadsen, T. G., Bebe, A., Køster-Rasmussen, R., et al. (2016). The role of diseases, risk factors and symptoms in the definition of multimorbidity: A systematic review. Scandinavian Journal of Primary Health Care, 34(2), 112-121. doi:10.3109/02813432.2016.1153242.
Comments