{"schema_version":1,"research_id":"eoa_inverse_innovation_exp04_external_evaluation_20260802","source_assessment_id":"invariant_mode_decomposition_design__psychology:PROPOSAL_FIRST:v0","cell_id":"invariant_mode_decomposition_design__psychology","search_queries":["site:pmc.ncbi.nlm.nih.gov person-specific vector autoregressive psychological symptoms experience sampling personalized feedback","site:pmc.ncbi.nlm.nih.gov micro-randomized trial just-in-time adaptive intervention mental health","site:nimh.nih.gov strategic plan digital mental health personalized interventions measurement 2025","ecological momentary assessment participant burden reactivity systematic review psychology PMC","person-specific VAR experience sampling psychological symptoms dynamic network intervention personalized feedback PMC","idiographic vector autoregression ecological momentary assessment mental health dynamic network model PMC","within-person dynamic factor model psychological symptoms eigenvalues stability intervention ecological momentary assessment","N-of-1 personalized dynamic network intervention psychological symptoms experience sampling trial","site:who.int mental health at work 12 billion working days lost depression anxiety 1 trillion","site:hhs.gov ohrp 45 CFR 46 informed consent IRB minimal risk research official","site:nimh.nih.gov strategic plan 2024 personalized mental health digital tools objective behavior intervention effects","site:bls.gov occupational employment wages clinical counseling psychologists 2025","psychological vector autoregressive model eigenvalues stability roots experience sampling","idiographic VAR psychological data stability eigenvalues stationary model","dynamic factor model psychological intervention eigenvalues transition matrix person-specific","\"Time-Varying Network Models for the Temporal Dynamics\" JMIR 2024","\"Ecological Momentary Assessments and Automated Time Series Analysis\" JMIR 2016","\"Momentary Factors and Study Characteristics Associated With Participant Burden\" JMIR 2024","\"Effectiveness of network analysis-driven personalized digital interventions\" 2026","\"Development and initial implementation of the Dynamic Assessment Treatment Algorithm\" publisher","\"DATA-IN\" dynamic assessment treatment algorithm individual networks psychotherapy","\"Network Control Theory in Personalized Intervention Selection\" PDF"],"sources":[{"source_id":"S1","title":"Mental health at work","publisher":"World Health Organization","url":"https://www.who.int/vietnam/news/fact-sheets/detail/mental-health-at-work","source_class":"OFFICIAL_GUIDANCE","publication_date":"2024-09-02","accessed_at":"2026-08-02","claims_supported":["Excessive workload, low job control, job insecurity, and other workplace conditions are mental-health risks.","An estimated 15% of working-age adults had a mental disorder in 2019.","Depression and anxiety are associated with an estimated 12 billion lost workdays and US$1 trillion in lost productivity annually.","WHO recommends individual stress-management interventions alongside organizational measures and meaningful worker involvement."]},{"source_id":"S2","title":"Goal 3: Strive for Prevention and Cures","publisher":"National Institute of Mental Health","url":"https://www.nimh.nih.gov/about/strategic-planning-reports/goals/goal-3-strive-for-prevention-and-cures","source_class":"OFFICIAL_GUIDANCE","publication_date":"2024","accessed_at":"2026-08-02","claims_supported":["NIMH identifies mechanism-informed intervention development, target engagement, quantitative behavioral measures, real-world participant-reported outcomes, and digital tools in naturalistic environments as research interests.","NIMH explicitly supports efficient designs for tailoring interventions to individual needs.","NIMH states that failure to engage a target or failure of target engagement to improve clinical outcomes should disqualify the proposed target."]},{"source_id":"S3","title":"Informed Consent FAQs","publisher":"U.S. Department of Health and Human Services, Office for Human Research Protections","url":"https://www.hhs.gov/ohrp/regulations-and-policy/guidance/faq/informed-consent/index.html","source_class":"GOVERNMENT_OR_REGULATOR","publication_date":"undated current guidance","accessed_at":"2026-08-02","claims_supported":["Covered human-subjects research generally requires legally effective informed consent unless an IRB-approved exception applies.","Consent must support voluntary choice and minimize coercion or undue influence.","IRBs must consider additional safeguards for participants vulnerable to coercion or impaired decision-making, and relevant professional expertise is required for review."]},{"source_id":"S4","title":"Momentary Factors and Study Characteristics Associated With Participant Burden and Protocol Adherence: Ecological Momentary Assessment","publisher":"JMIR Formative Research","url":"https://formative.jmir.org/2024/1/e49512/PDF","source_class":"PRIMARY_RESEARCH","publication_date":"2024-04-24","accessed_at":"2026-08-02","claims_supported":["In an eight-day EMA study of 150 caregivers, average reported burden was low, but depressed mood, stress, and nonadherence were associated with greater burden.","EMA feasibility depends on tailoring the design and language to the participant population.","Assessment burden and adherence are material safety and data-quality outcomes for the proposed workflow."]},{"source_id":"S5","title":"Time-Varying Network Models for the Temporal Dynamics of Depressive Symptomatology in Patients With Depressive Disorders","publisher":"JMIR Mental Health","url":"https://mental.jmir.org/2024/1/e50136/PDF","source_class":"PRIMARY_RESEARCH","publication_date":"2024-04-18","accessed_at":"2026-08-02","claims_supported":["Person-specific time-varying VAR models have already been applied to depression, rumination, sleep, and social-contact data.","The study found marked between-person and within-person differences and evidence of nonstationarity in every participant.","The analysis used substantially longer series—mean 274 days among 20 participants—than the candidate's original 28-day design.","The study assessed estimate stability, prediction error, and variance explained, making drift and held-out validity established concerns rather than new concepts.","Its feasibility study involved participant reimbursement and provisioned phones, supplying a partial empirical anchor for pilot-resource assumptions."]},{"source_id":"S6","title":"Development and initial implementation of the Dynamic Assessment Treatment Algorithm (DATA)","publisher":"PLOS ONE","url":"https://journals.plos.org/plosone/article?id=10.1371/journal.pone.0178806","source_class":"PRIMARY_RESEARCH","publication_date":"2017-06-27","accessed_at":"2026-08-02","claims_supported":["DATA already converts intensive within-person symptom measurements, factor models, and dynamic factor models into ranked modular psychotherapy recommendations.","The study collected smartphone surveys four times daily for at least 30 days and required approximately 100–120 usable observations per participant.","The workflow explicitly considered factor persistence, cross-lagged influence, symptom severity, and mappings from items to interventions.","The work excluded suicidality from automated modeling for safety and used IRB approval, consent, and licensed clinical input."]},{"source_id":"S7","title":"Ecological Momentary Assessments and Automated Time Series Analysis to Promote Tailored Health Care: A Proof-of-Principle Study","publisher":"JMIR Research Protocols","url":"https://www.researchprotocols.org/2015/3/e100/","source_class":"PRIMARY_RESEARCH","publication_date":"2015-08-07","accessed_at":"2026-08-02","claims_supported":["AutoVAR already automates individual-level VAR analysis of EMA data and produces output intended for nonexperts.","The proof of principle found automated output comparable to manual analyses in four examples, while explicitly calling for additional validation.","The paper identifies equidistant timing, sufficient repeated observations, standardized data collection, expert interpretation, and integration with care infrastructure as implementation requirements.","The paper states that improved clinical outcomes from tailored time-series advice remained an empirical question."]},{"source_id":"S8","title":"Effectiveness of network analysis–driven personalized digital interventions versus standard intervention for depression: a proof-of-concept pilot randomized controlled trial","publisher":"Molecular Psychiatry, Springer Nature","url":"https://www.nature.com/articles/s41380-026-03467-w","source_class":"PRIMARY_RESEARCH","publication_date":"2026-02-06","accessed_at":"2026-08-02","claims_supported":["A preregistered pilot randomized 62 participants to EMA-network-personalized versus standardized intervention sequencing.","Both groups improved, but the group-by-time effect was near zero and imprecise (beta -0.01, 95% CI -0.96 to 0.94; p=0.987).","Existing evidence therefore does not establish an outcome advantage for network-personalized intervention selection.","The study provides a direct comparator and demonstrates a feasible IRB-reviewed pathway for prospective testing."]}],"problem_evidence":{"support":"MODERATE","rationale":"Work-related mental-health burden clearly exists and matters, and intensive longitudinal research shows that psychological variables have heterogeneous, time-varying within-person relationships. However, no relied-upon source demonstrates the proposal's exact problem: a stable coupled eigenmode that remains modest in every item yet reliably gives earlier, decision-relevant warning than item thresholds or direct prediction. Nonstationarity in all participants of S5 is counterpressure against that premise.","source_ids":["S1","S4","S5","S6"]},"stakeholder_evidence":{"support":"MODERATE","rationale":"NIMH is an identifiable prospective funder expressing need for mechanism-informed targets, digital naturalistic measurement, target-engagement tests, and personalized interventions. WHO identifies employers, workers, governments, and health services as responsible stakeholders. This is credible institutional pull for the research class, but there is no named clinic, employer, IRB sponsor, or funder commitment to this particular modal design.","source_ids":["S1","S2"]},"prior_art":{"proximity":"SUBSTANTIAL_COLLISION","closest_analogues":[{"name":"Dynamic Assessment Treatment Algorithm (DATA)","similarity":"Uses intensive smartphone self-reports, within-person factor and dynamic models, persistence and cross-lagged effects, and an explicit item-to-intervention map to prioritize modular treatment.","remaining_difference":"It ranks factors and items rather than eigenvectors of an observed-state transition operator and does not use spectral-gap, mode-rotation, or modal reconstruction gates.","source_ids":["S6"]},{"name":"Time-varying idiographic symptom networks","similarity":"Fits person-specific time-varying VAR models to depression, rumination, sleep, and social variables and evaluates stability and prediction error.","remaining_difference":"It is exploratory and does not decompose transition matrices into intervention-target modes or test prompts.","source_ids":["S5"]},{"name":"AutoVAR for tailored health care","similarity":"Automates individual-level VAR estimation from EMA and aims to translate temporal relationships into tailored health advice.","remaining_difference":"It does not establish eigenmode-specific target engagement, spectral governance, or clinical outcome superiority.","source_ids":["S7"]},{"name":"Network-personalized digital intervention sequencing RCT","similarity":"Uses pretreatment EMA-derived person-specific networks to choose intervention sequencing and compares personalization with a standardized sequence.","remaining_difference":"It targets network-defined core symptoms rather than invariant modes; its pilot found no detectable between-arm advantage.","source_ids":["S8"]}],"distinctive_claim_remaining":"Conditional on a participant having a prespecified, well-conditioned, reproducible, spectrally separated transition mode with unstructured held-out residuals, the mode coordinate will add out-of-sample predictive information beyond item thresholds, a fixed composite, and a direct regularized predictor; in a separately authorized micro-randomized test, a clinician-approved action selected for modal alignment will change the next-window mode coordinate or task/distress outcome more than generic, item-selected, or no-prompt controls without greater burden. Either part is falsified by unstable modes, no incremental prediction, no differential target engagement, structured residuals, or excess burden.","confidence":"HIGH"},"implementation_evidence":{"support":"MODERATE","rationale":"EMA collection, person-specific VAR estimation, automated analysis, dynamic-factor-based treatment mapping, and prospective network-personalized trials have all been implemented. The proposed eigendecomposition itself is computationally routine. Feasibility is weakened by ordinal/noisy self-reports, irregular timing, missingness, nonstationarity, and the observation burden needed to estimate six-variable individual models. Prior implementations used about 100–120 observations over at least 30 days or much longer series; a transition eigenvalue above one in bounded psychological ratings may indicate local nonstationarity or misspecification rather than a durable dangerous mode. Action vectors inferred observationally are not causal controls.","source_ids":["S4","S5","S6","S7","S8"]},"scores":{"meaningful_impact":{"score":3,"rationale":"The underlying work-stress burden is large, but the marginal impact of modal monitoring over simpler self-management is unmeasured.","source_ids":["S1","S8"]},"stakeholder_pull":{"score":3,"rationale":"NIMH and WHO express clear need for personalized, mechanism-tested digital mental-health research, but no operational partner has requested this design.","source_ids":["S1","S2"]},"incremental_advantage":{"score":2,"rationale":"Spectral-gap and mode-drift gates could add disciplined model-validity checks, but closely related personalized dynamic approaches exist and the available pilot RCT found no detectable advantage over standardized sequencing.","source_ids":["S5","S6","S7","S8"]},"distinctiveness_plausibility":{"score":2,"rationale":"The eigenmode framing is narrower than factor, node-centrality, and edge-based approaches, but most of the data, model, intervention-mapping, and validation workflow substantially collides with prior art.","source_ids":["S5","S6","S7","S8"]},"technical_implementability":{"score":3,"rationale":"The software and statistical components are implementable, but reliable participant-level estimation and interpretation under nonstationarity are unresolved.","source_ids":["S5","S6","S7"]},"adoption_authority_feasibility":{"score":3,"rationale":"A conventional participant-consent, clinician-supervision, and IRB-review pathway exists. Feasibility remains conditional on recruiting a licensed clinical partner and keeping employment and diagnostic decisions out of scope.","source_ids":["S2","S3","S6","S8"]},"evidence_readiness":{"score":2,"rationale":"The adjacent evidence base is mature enough to specify a rigorous test, but the candidate has no direct evidence of stable modes, earlier detection, incremental prediction, action target engagement, or acceptable false-alert rates.","source_ids":["S5","S6","S8"]},"safety_net_benefit":{"score":3,"rationale":"Participant-controlled, low-risk coping and explicit suspension rules could benefit adults before distress escalates, but assessment can burden precisely those reporting more stress or depressed mood and the model cannot replace crisis assessment.","source_ids":["S3","S4","S6"]},"scalability":{"score":2,"rationale":"Smartphone collection and automated VAR tools can scale technically, but dense per-person sampling, missingness, clinician review, repeated drift checks, and frequent model retirement constrain operational scale.","source_ids":["S4","S5","S7"]}},"score_confidence":"MODERATE","costs":{"first_evidence":{"band_2026_usd":"50K_TO_250K","scope":"One partnered, IRB-approved, no-prompt observational study of up to 24 adults over 42 days, including protocol design, participant compensation, clinical safety coverage, EMA infrastructure, quantitative analysis, preregistration, and reporting.","confidence":"MODERATE","assumptions":["Four scheduled assessments per day target at least 100 usable training observations per participant before held-out evaluation.","Existing survey or EMA infrastructure is adapted rather than built from scratch.","One quantitative scientist, one coordinator, and fractional licensed-clinician and principal-investigator effort are available.","No diagnosis, treatment delivery, wearable procurement, or employment-system integration is included."],"source_ids":["S4","S5","S6"]},"initial_deployment_startup":{"band_2026_usd":"250K_TO_1M","scope":"Build and validate a secure research-grade prompting service after observational success, add audit logs, consent and withdrawal controls, clinician review, security assessment, monitoring dashboards, and micro-randomized-trial protocols.","confidence":"LOW","assumptions":["Research deployment only, not a regulated diagnostic product.","Existing validated coping content is licensed or supplied by the clinical partner.","Includes engineering, privacy/security, clinical, statistical, and project-management work for roughly 9–18 months."],"source_ids":["S2","S3","S7","S8"]},"operational_launch":{"band_2026_usd":"1M_TO_5M","scope":"A multisite, separately reviewed micro-randomized trial with adequate participants, comparator arms, clinical safety coverage, data monitoring, independent analysis, and follow-up.","confidence":"LOW","assumptions":["Approximately 150–300 participants across multiple sites.","Includes participant compensation, site contracts, data and safety monitoring, statistical programming, and trial operations.","No commercial marketing, medical-device clearance program, or employer deployment is included."],"source_ids":["S2","S3","S8"]},"annual_recurring":{"band_2026_usd":"250K_TO_1M","scope":"Ongoing research operation after launch: hosting, security, participant support, licensed-clinician oversight, incident review, mode revalidation, software maintenance, governance, and audit retention.","confidence":"LOW","assumptions":["Limited research or health-system deployment with hundreds to low thousands of active consenting users.","Clinical review is exception-based rather than continuous surveillance.","Material expansion, multilingual validation, and regulated-product obligations would move costs upward."],"source_ids":["S3","S4","S7"]}},"verified_pipeline_gates":{"externally_supported_problem":{"status":"YES","reason":"Work-related mental-health burden is externally documented, and primary research confirms heterogeneous, coupled, time-varying within-person symptom dynamics; the exact hidden-mode failure remains unproven.","source_ids":["S1","S5","S6"]},"externally_credible_adopter_or_authorizer":{"status":"YES","reason":"NIMH is an identifiable potential funder expressing relevant research needs, while an IRB and licensed clinical research partner form a credible authorization route. This is not evidence of a commitment from either.","source_ids":["S2","S3"]},"distinct_testable_incremental_claim":{"status":"YES","reason":"Incremental held-out prediction and later differential target engagement can be tested against item thresholds, a fixed composite, direct regularized prediction, standardized or generic action selection, and no prompt.","source_ids":["S2","S6","S8"]},"bounded_next_evidence_step":{"status":"YES","reason":"A capped, no-prompt observational study can lock sampling, stability, residual, prediction, burden, and stop criteria before collection and can terminate without deploying recommendations.","source_ids":["S3","S4","S5","S6"]},"no_unresolved_safety_or_authority_stop":{"status":"YES","reason":"For the observational step only, informed consent, IRB review, clinician-approved safety thresholds, exclusion of diagnosis/employment/crisis authority, and immediate suspension are credible controls. Live prompting still requires a new review.","source_ids":["S3","S4","S6","S8"]},"credible_cost_scope_and_range":{"status":"YES","reason":"The four bands are broad resource-equivalent planning ranges tied to explicit study and deployment scopes; analogous studies document intensive sampling, participant compensation, provisioned devices, clinical review, and substantial data requirements. Partner quotes and local overhead remain unknown.","source_ids":["S4","S5","S6","S8"]}},"next_evidence_step":"Secure a licensed clinical or occupational-mental-health research partner and run one preregistered, IRB-approved, no-prompt pilot with at most 24 consenting adults, 42 days, and four brief assessments per day. Use the first 28 days for estimation and the last 14 for untouched evaluation. Before enrollment, simulate the fixed six-variable cadence and lock minimum usable observations, missingness, conditioning, mode-separation uncertainty, split-half or bootstrap mode-direction similarity, reconstruction error, residual-structure tests, and burden limits. Compare (1) independent item thresholds, (2) a fixed composite, (3) a persistence/no-change model, and (4) a direct regularized next-outcome predictor. Falsify the candidate for this setting if fewer than 70% of participants yield an estimable model, the selected mode is not reproducible across resamples or rotates in the test period, residuals retain outcome-relevant structure, the modal model fails to exceed the best comparator by the locked prediction margin, or burden and distress exceed the approved limits. Do not issue model-selected prompts; observational success only authorizes design and fresh review of a micro-randomized target-engagement study.","blocking_evidence":["No direct evidence shows that a coupled mode gives earlier or more actionable warning than item thresholds or direct prediction in adults managing work stress.","No named clinic, occupational-health program, employer-independent service, IRB sponsor, or funder has committed to the study.","The sampling cadence required for a reliable six-variable participant-level transition operator has not been justified by simulation or empirical reliability analysis.","No validated measurement model establishes that the brief ordinal items are comparable across time and sufficiently sensitive at the proposed cadence.","Nonstationarity, irregular intervals, missingness, near-degenerate modes, and model conditioning may prevent stable eigenmode interpretation.","No causal evidence shows that grounding, a two-minute task step, or contacting support acts as the proposed input vector or controls the target mode.","False-alert rates, alert timing advantage, assessment reactivity, subgroup accessibility, and participant burden are unknown in the target population.","Privacy, retention, breach response, jurisdiction-specific research requirements, and any later product-regulatory classification remain unspecified.","Local staffing rates, technology costs, insurance, indirect costs, and partner quotations are unavailable."],"research_disposition":"PARTNERED_RESEARCH_PROGRAM","world_novelty_boundary":"This evaluation measured only proximity within a bounded web search. It did not measure world novelty, patentability, freedom to operate, market size, or realized impact. The search found substantial prior art in idiographic dynamic-factor treatment algorithms, automated EMA-VAR analysis, time-varying symptom networks, and prospective network-personalized intervention trials. A highly relevant 2026 article on network-control-theory simulations for personalized psychological intervention selection was also surfaced, but its publisher full text could not be independently opened and it was therefore not used as a relied-upon source; it should be examined in a dedicated differentiation review. No exhaustive patent, dissertation, code-repository, conference, non-English, or proprietary-product search was performed.","arm":"PROPOSAL_FIRST","candidate_version":0,"controller_recommendation":{"action":"STOP_EMPIRICAL_RESEARCH_NEEDED","repairable":true,"material_progress_observed":true,"progress_targets":["Obtain a named licensed clinical or occupational-mental-health partner, IRB sponsor, and participant-safety owner.","Complete simulation-based cadence and identifiability analysis for the six-variable within-person operator before recruitment.","Preregister the no-prompt observational protocol, comparator implementations, locked prediction margin, mode-stability criteria, residual tests, burden thresholds, and participant-level stop rules.","Demonstrate on untouched observations that a reproducible mode adds decision-relevant information beyond item thresholds, a fixed composite, persistence, and direct regularized prediction.","Report every excluded or non-estimable participant and quantify missingness, false-alert proxies, mode rotation, conditioning, subgroup adherence, and burden.","If and only if observational gates pass, obtain separate ethics and clinical approval for a sufficiently powered micro-randomized comparison of modal-aligned, item-selected, generic, and no-prompt conditions.","Acquire partner-specific staffing, technology, security, participant-compensation, insurance, and indirect-cost quotations before deployment planning.","Differentiate the remaining spectral claim against 2026 network-control-theory work and other control-oriented psychological time-series research."],"reason":"Web evidence can establish the importance of work-related distress, the existence of institutional interest, technical precedents, substantial prior-art collision, and a lawful observational pathway. It cannot establish that stable coupled modes exist in the target population, outperform simpler models, provide earlier warning, or can be causally controlled by the proposed actions. Those decisive questions require consented longitudinal field data and, later, live randomized target-engagement testing."}}