Evidence base: published literature, cited and verified. Validation studies: designed, IRB-gated.
This research began before the first line of code. As a licensed clinician learning to build software, the author investigated medication-related deaths and the evidence base behind current care, and those findings shaped every design decision in the FSL system. The work addresses five problems in mental-wellness education: stigma, shame, access, privacy and autonomy. It asks whether a wallet-signed, participant-controlled system can lower barriers to help-seeking while keeping consent verifiable and revocable. Six research streams organize the evidence: stigma and access; participant autonomy and data sovereignty; privacy and consent; orthomolecular and neurobiological evidence; informatics and interoperability; and decentralized architecture. The planned study designs are a randomized comparison of entry flows, within-subject consent-comprehension testing, formal protocol verification, an IRB-gated observational cohort, FHIR conformance testing and operational measurement. The doctoral contribution specifies bilateral wallet-signed consent: the participant and the Sovereign Guide each sign, and session keys are designed to derive from both signatures, so that a session can only be established with both parties’ consent.
Doctoral thesis: “Bilateral Wallet-Signed Consent: A Custom Signature Protocol for Decentralized Sovereign Data Governance, with Behavioral Health as Proving Ground.” Central research question: Can wallet-signed bilateral consent serve as a defensible governance primitive for sovereign behavioral health data? Full scope: Doctoral Scope.
157,149,246
people live in designated Mental Health Professional Shortage Areas (as of June 30, 2026).
Source: HRSA, 2026, HPSA quarterly report, Table 1, p. 2
96.3%
of people with a 12-month DSM-IV disorder who perceived a need and did not seek treatment reported an attitudinal barrier (24 countries, n = 4,583).
Source: Andrade et al., 2014, Table 3
290,054,222
individuals affected, reported as a combined total by the 741 breach filings affecting 500 or more individuals submitted to HHS in 2024 (portal tally, Sept 30, 2026).
Source: HHS OCR breach portal
52.1%
of trial participants understood randomization, the least-understood consent component (103 studies).
Source: Tam et al., 2015
Data and code: rv-evidence.csv · build-rv-charts.py (every figure in the charts comes from this file; the script regenerates the charts deterministically) · ocr-breach-export-2026-09-30.csv (the HHS OCR breach portal export behind the breach figures, taken 2026-09-30; SHA-256 a05a0eb2dcff81c316fcce9ff0e5683e49ea806a0cb448999cd25edc09e92084).
| Stream | Research question | Hypothesis (Planned) | Doctoral lane |
|---|---|---|---|
| Stigma and Access | Can an education entry point that asks for no name or disclosure, only a wallet signature, reduce the disclosure and attitudinal barriers that deter help-seeking for mental wellness? | Participants who enter through the wallet-signature entry point report lower BACE treatment-stigma subscale scores than participants who enter through a name-and-email registration. | VALIDATE |
| Participant Autonomy and Data Sovereignty | Can participants hold cryptographic proof of each grant and revocation of access to their wellness records, so that consent is demonstrable and as easy to withdraw as to give? | Participants who grant and revoke access with wallet signatures identify who can read their records and how to withdraw access at QuIC objective-understanding scores non-inferior to those after a conventional consent form (margin: 5 points on the 0–100 QuIC scale). | VALIDATE |
| Privacy and Consent | Can wallet signatures bound to single-use nonces yield sign-in and consent records that cannot be replayed, and what does EIP-712 domain separation add beyond EIP-191? | The nonce-bound EIP-191 sign-in and consent protocols satisfy replay resistance and authentication properties against a network attacker, and an EIP-712 variant adds a domain binding that the EIP-191 model lacks. | PROVE |
| Orthomolecular and Neurobiological Evidence | Which participants respond to specific evidence-based nutrients and plant-based and gluten-free dietary change, and can participant-held longitudinal self-report make that individual response observable? | Participant-reported mood trajectories over 12 weeks differ between participants who adopt an evidence-based nutrient or a plant-based and gluten-free dietary change and those who do not, with individual response varying across participants. Primary contrast: plant-based dietary change, consistent with the meta-analytic evidence that dietary improvement reduces depressive symptoms (Firth et al., 2019b); gluten-free adherence is a pre-specified exploratory moderator, because the evidence for gluten-free diets and mood is limited to people with gluten-related disorders (Busby et al., 2018). | VALIDATE |
| Informatics and Interoperability | Can device and laboratory results reach a participant-controlled record as standard FHIR R4 resources, with access governed by the participant’s own signed consent? | Device and laboratory data mapped by the FSL pipelines pass FHIR R4 validation and preserve every source value, unit and code. | VALIDATE |
| Decentralized Architecture | Can a small clinical practice operate decentralized payment and record-integrity infrastructure in production, a self-hosted Bitcoin node with a watch-only wallet and Merkle roots anchored to a public chain, and verify it end to end? | The self-hosted payment and anchoring stack records every settled payment exactly once and verifies every anchored record against its on-chain root. | BUILD |
Feeds: VALIDATE
Evidence
People’s own negative attitudes toward help-seeking (OR 0.80) and stigmatising attitudes (OR 0.82) are associated with less help-seeking while perceived public stigma is not (OR 0.97, 95% CI 0.93–1.02; Schnyder et al., 2017); among people with a 12-month DSM-IV disorder who perceived a need and did not seek treatment, 96.3% reported an attitudinal barrier against 22.6% for any structural barrier, and 157,149,246 people live in designated mental health shortage areas.
Source: Schnyder et al., 2017, 27 studies, n = 31,677
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Own negative help-seeking attitudes | 0.80 (OR) | 0.73 to 0.88 | 27 studies; 31,677 participants (review total) | Schnyder et al., 2017 |
| Stigmatising attitudes (personal stigma) | 0.82 (OR) | 0.69 to 0.98 | 27 studies; 31,677 participants (review total) | Schnyder et al., 2017 |
| Self-stigma | 0.88 (OR) | 0.76 to 1.03 | 27 studies; 31,677 participants (review total) | Schnyder et al., 2017 |
| Perceived public stigma | 0.97 (OR) | 0.93 to 1.02 | 27 studies; 31,677 participants (review total) | Schnyder et al., 2017 |
| General stigma (unspecified) | 0.98 (OR) | 0.84 to 1.15 | 27 studies; 31,677 participants (review total) | Schnyder et al., 2017 |
Source: Andrade et al., 2014, WHO World Mental Health surveys, 24 countries, n = 4,583 (Table 3)
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Any attitudinal barrier | 96.3% | — | 4,583 | Andrade et al., 2014 |
| Wanted to handle the problem on their own | 63.8% | — | 4,583 | Andrade et al., 2014 |
| Problem was not severe | 24.4% | — | 4,583 | Andrade et al., 2014 |
| Any structural barrier | 22.6% | — | 4,583 | Andrade et al., 2014 |
| Thought it would get better | 16.0% | — | 4,583 | Andrade et al., 2014 |
| Financial | 15.9% | — | 4,583 | Andrade et al., 2014 |
| Perceived ineffectiveness | 15.7% | — | 4,583 | Andrade et al., 2014 |
| Availability | 12.4% | — | 4,583 | Andrade et al., 2014 |
| Stigma | 7.7% | — | 4,583 | Andrade et al., 2014 |
| Inconvenient | 6.4% | — | 4,583 | Andrade et al., 2014 |
| Transportation | 5.4% | — | 4,583 | Andrade et al., 2014 |
Source: HRSA, 2026, HPSA quarterly report as of June 30, 2026, Table 1
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Facility designations | 4,925 | — | — | HRSA, 2026 |
| Geographic-area designations | 1,216 | — | — | HRSA, 2026 |
| Population-group designations | 968 | — | — | HRSA, 2026 |
A systematic review of 144 studies with 90,189 participants found that stigma has a small-to-moderate negative effect on help-seeking (median d = −0.27) and ranks fourth among barriers, with disclosure concerns the most commonly reported stigma barrier (Clement et al., 2015). A meta-analysis of 27 studies found that people’s own negative attitudes toward help-seeking (OR 0.80) and stigmatising attitudes toward people with mental illness (OR 0.82) were associated with less active help-seeking, while perceived public stigma was not, with substantial between-study heterogeneity (I² 58% to 91%) (Schnyder et al., 2017). Across WHO World Mental Health surveys in 24 countries (n = 63,678), low perceived need was the most common reason for not initiating treatment, a desire to handle the problem on one’s own was the most common barrier among those who perceived a need (63.8%), and attitudinal barriers were much more important than structural ones overall, although attitudinal barriers dominated for mild-to-moderate cases and structural barriers for severe cases (Andrade et al., 2014).
These studies measure how strongly disclosure concerns, personal attitudes and a preference for self-reliance are associated with less help-seeking, and Clement et al. (2015) state that the findings can inform the design of interventions to increase help-seeking. They do not evaluate an entry point that requires no identity disclosure and supports self-directed learning, so whether such a design changes help-seeking remains open.
Method (Planned)
Two-arm randomized comparison among adults seeking self-serve mental-wellness education, recruited online: each participant is randomly assigned to the wallet-signature entry or a name-and-email registration and completes the measures at entry and at four weeks, with the sample size set by an a priori power analysis. Stigma measures are administered anonymously in both arms, so the arms are designed to differ only in the entry method. Baseline BACE and ATSPPH-SF scores, age, gender, prior help-seeking and prior wallet use are recorded before randomization, and balance between arms is checked on each (standardized mean difference below 0.1).
Data and measures (Planned)
Barriers to Access to Care Evaluation, including its treatment-stigma subscale (BACE; Clement et al., 2012); Attitudes Toward Seeking Professional Psychological Help, short form (ATSPPH-SF; Fischer & Farina, 1995); entry-flow completion logged as a system measure.
Analysis plan (Planned)
Arm-by-time effects on BACE and ATSPPH-SF scores are tested with linear mixed-effects models adjusted for the baseline covariates (baseline score, age, gender, prior help-seeking, prior wallet use), analyzed by intention to treat, and entry-flow completion is compared between arms with logistic regression. Entry-flow completion is analyzed separately from the self-report measures.
Built component: Wallet-signature sign-in with a single-use nonce (EIP-191), with no name or email in sign-in: evidence.futuresystemslab.io (Connect Wallet)
Feeds: VALIDATE
Evidence
Given control over provider access, 45 of 105 participants (43%) restricted at least one provider, while 60 (57%) imposed no restrictions. Thirty-four (32.3%) blocked all personal information from all listed individuals of at least one provider or staff type. The share of individuals nationally who were offered and accessed their online medical records rose from 25% in 2014 to 65% in 2024.
Source: Schwartz et al., 2015, n = 105
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Blocked all personal information from all doctors, nurses and/or staff | 32.3% | — | 105 | Schwartz et al., 2015 |
| Blocked access by all doctors and/or nurses | 24.8% | — | 105 | Schwartz et al., 2015 |
| Denied access to all doctors, nurses and staff | 4.8% | — | 105 | Schwartz et al., 2015 |
Source: Richwine, 2025 (ONC Data Brief No. 77), HINTS 2024, n = 7,278 (national survey; not peer-reviewed)
Survey-weighted national estimates; wave sizes: 2014 3,677 and 2020 3,865 (Westat, 2015, 2020); 2024 7,278.
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Offered and accessed online records, 2014 | 25% | — | — | ONC Data Brief No. 77 (2014) |
| Offered and accessed online records, 2024 | 65% | — | — | ONC Data Brief No. 77 (2024) |
| Accessed records by app (portal users), 2020 | 38% | — | — | ONC Data Brief No. 77 (2020) |
| Accessed records by app (portal users), 2024 | 57% | — | — | ONC Data Brief No. 77 (2024) |
The GDPR sets principles for processing personal data, including limiting data to what is necessary (Art. 5), requires the controller to be able to demonstrate consent (Art. 7(1)), and gives the data subject the right to withdraw consent at any time, as easily as it was given (Art. 7(3)). Under HIPAA, an individual has a right of access to inspect and obtain a copy of protected health information in a designated record set, with exceptions that include psychotherapy notes (45 CFR § 164.524(a)(1)). Dynamic consent places a personalised digital interface between researchers and participants, letting individuals tailor and manage their own consent preferences and carry those preferences with their data (Kaye et al., 2015).
These instruments define participant rights to access, demonstrable consent and withdrawal, while the record of consent sits with the controller or covered entity. Dynamic consent has mainly been developed in biobanking, so participant-held, independently verifiable records of each grant and revocation for wellness data remain open.
Method (Planned)
Within-subject comparison: adult participants complete a conventional consent form and a wallet-signed grant-and-revoke task in counterbalanced order and then complete the measures, while the system logs task success and time.
Data and measures (Planned)
Quality of Informed Consent (QuIC; Joffe et al., 2001); System Usability Scale (SUS; Brooke, 1996); mHealth App Usability Questionnaire (MAUQ; Zhou et al., 2019); logged grant and revocation success and time to revoke.
Analysis plan (Planned)
QuIC objective-understanding scores (0–100 scale) are tested for non-inferiority against a pre-specified margin of 5 points: the wallet-signature condition is non-inferior if the lower bound of the one-sided 97.5% confidence interval for the paired difference (wallet minus conventional) lies above −5 points, using a paired t-test, or a Wilcoxon signed-rank test with a Hodges–Lehmann interval where distributions require, with condition order included to check for carryover; SUS and MAUQ are reported descriptively, and revocation success as a proportion with a 95% confidence interval.
Built component: Participant-signed consent grants for each Sovereign Guide, with one-step revocation of every grant: evidence.futuresystemslab.io/security
Feeds: PROVE
Evidence
The 741 breach filings submitted in 2024 reported a combined 290,054,222 individuals affected. Across 103 studies, understanding of informed consent ranged from 75.8% for freedom to withdraw to 52.1% for randomization, with no increase over 30 years.
Source: HHS OCR breach portal, tally of portal CSV exports taken 2026-09-30 (figures change as OCR updates records)
Totals count breaches by submission date to the HHS OCR portal and are revised as reports update. HHS’s Annual Report to Congress reports 742 such reports received in calendar year 2024 and, by occurrence date, 663 breaches affecting 242,908,056 individuals (Office for Civil Rights, 2026).
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| 2021 | 61,206,238 | — | — | HHS OCR breach portal (2021) |
| 2022 | 64,570,326 | — | — | HHS OCR breach portal (2022) |
| 2023 | 183,256,225 | — | — | HHS OCR breach portal (2023) |
| 2024 | 290,054,222 | — | — | HHS OCR breach portal (2024) |
| 2025 | 140,575,464 | — | — | HHS OCR breach portal (2025) |
Totals count breaches by submission date to the HHS OCR portal and are revised as reports update. HHS’s Annual Report to Congress reports 742 such reports received in calendar year 2024 and, by occurrence date, 663 breaches affecting 242,908,056 individuals (Office for Civil Rights, 2026).
Source: HHS OCR breach portal, same tally
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| 2021 | 715 | — | — | HHS OCR breach portal (2021) |
| 2022 | 718 | — | — | HHS OCR breach portal (2022) |
| 2023 | 745 | — | — | HHS OCR breach portal (2023) |
| 2024 | 741 | — | — | HHS OCR breach portal (2024) |
| 2025 | 804 | — | — | HHS OCR breach portal (2025) |
Source: Tam et al., 2015, 103 studies, 135 cohorts
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Freedom to withdraw | 75.8% | 70.6 to 80.3 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Nature of the study | 74.7% | 68.8 to 79.8 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Voluntary participation | 74.7% | 67.9 to 80.5 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Potential benefits | 74.0% | 65.0 to 81.3 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Purpose of the study | 69.6% | 63.5 to 75.1 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Potential risks and side-effects | 67.0% | 57.4 to 75.4 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Confidentiality | 66.2% | 55.3 to 75.7 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Alternative treatment if withdrawn | 64.1% | 53.7 to 73.4 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Knowing treatments were compared | 62.9% | 45.5 to 77.5 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| No therapeutic misconception | 62.4% | 50.1 to 73.2 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Naming at least one risk | 54.9% | 43.3 to 65.0 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Placebo | 53.3% | 38.4 to 67.6 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
| Randomization | 52.1% | 41.3 to 62.7 | 103 studies; 135 cohorts (review total) | Tam et al., 2015 |
EIP-191 defines the signed-data format used by personal_sign: version byte 0x45 and the prefix “\x19Ethereum Signed Message:\n” followed by the message length. EIP-4361 (Sign-In with Ethereum) specifies how Ethereum accounts authenticate with off-chain services by signing a standard message that carries scope, session details and security mechanisms such as a nonce. EIP-712 defines hashing and signing of typed structured data under version byte 0x01, binding each signature to a 32-byte domain separator. The GDPR requires that consent be demonstrable by the controller and withdrawable at any time (Art. 7(1), 7(3)).
The standards specify message formats and signature schemes, while replay protection and the keeping of demonstrable consent records are left to the implementing service. EIP-191 personal_sign messages carry no domain separator of the kind EIP-712 defines, so binding a signature to one application rests on the message text itself.
Method (Planned)
Formal protocol verification: the sign-in and consent message flows are modeled as symbolic protocols with the attacker controlling the network, and the EIP-191 and EIP-712 variants are modeled separately and compared.
Data and measures (Planned)
Protocol models in the Tamarin prover (Meier et al., 2013) or ProVerif (Blanchet, 2001), with lemmas for replay, authentication and domain binding; Quality of Informed Consent (QuIC; Joffe et al., 2001) for participant understanding of the consent text.
Analysis plan (Planned)
Each property is stated as a lemma and proved or refuted by the prover; any counterexample trace is reported together with the fix and the re-verification result.
Built component: Nonce-bound signature verification for sign-in (EIP-191, and EIP-4361 on the FSL Console) and for every consent grant: evidence.futuresystemslab.io/security
Feeds: VALIDATE
Evidence
The orthomolecular clinical literature (Walsh, Mensah, Hoffer and the Pfeiffer tradition) generated the hypotheses; meta-analyses reported statistically significant effects for EPA-predominant omega-3, methylfolate at 15 mg, vitamin D, zinc and N-acetylcysteine for depressive symptoms, of which Firth et al. (2019a) identify EPA-predominant omega-3 as the intervention with “the strongest evidentiary support” (SMD 0.61, reduced to 0.42 and still significant after adjusting for publication bias) and qualify the others (methylfolate only in few small trials; vitamin D and N-acetylcysteine with low confidence; zinc trials at high risk of attrition bias; “a lack of compelling evidence” for other vitamins, zinc, magnesium and inositol across mental disorders, with emerging evidence for vitamin D in major depression), alongside null results for DHA-predominant omega-3, folic acid, magnesium and inositol, a zinc deficit of 1.85 µmol/L in depression (Swardfager et al., 2013), and a placebo-controlled null result for adjuvant niacin (Ramsay et al., 1970); FSL’s participant-held longitudinal self-report is designed to test individual response.
Sources: 21 verified records, linked in the data table (PubMed, PMC, publisher and institute pages)
| Source | Year | Evidence level | Peer-reviewed | Note |
|---|---|---|---|---|
| Firth 2019a | 2019 | meta-analysis | yes | Meta-review of 33 meta-analyses of placebo-controlled RCTs (10,951 individuals) |
| Firth 2019b | 2019 | meta-analysis | yes | Meta-analysis of 16 RCTs of dietary improvement (corrected by 2020 and 2021 errata) |
| Busby 2018 | 2018 | meta-analysis | yes | Systematic review with meta-analysis: gluten-free diet and depressive symptoms in gluten-related disorders (13 prospective studies, 1,139 participants) |
| Sarris 2016 | 2016 | meta-analysis | yes | Positive replicated results for SAMe, methylfolate, omega-3 and vitamin D; mixed for zinc, folic acid, vitamin C, tryptophan; nonsignificant for inositol |
| Swardfager 2013 | 2013 | meta-analysis | yes | Zinc 1.85 µmol/L lower in depression (95% CI 1.19-2.51) |
| Papakostas 2012 | 2012 | RCT | yes | Two double-blind trials: 7.5 then 15 mg null (n = 148); 15 mg NNT about 6 (n = 75) |
| Ramsay 1970 | 1970 | RCT | yes | Placebo-controlled comparative study (randomization not stated in the abstract): no significant difference for adjuvant nicotinic acid |
| Petrie 1981 | 1981 | RCT | yes | CMHA collaborative 48-week study (randomization not stated in the abstract): single vitamins significant, combination not |
| Walsh 1997 | 1997 | cohort/case-control | yes | Median Cu/Zn 1.40 in 135 assaultive males vs 1.02 in 18 controls |
| Yao 2006 | 2006 | cohort/case-control | yes | Higher urinary isoprostane F2α-VI (oxidative stress) and platelet- and endothelial-activation markers in children with autism vs controls |
| Crayton 2007 | 2007 | cohort/case-control | yes | Serum copper 131 vs 111 µg/dL with vs without post-partum depression history |
| Osmond 1962 | 1962 | case series/clinical database | yes | Review of a nine-year niacin study (level from title; outcomes not in the PubMed record) |
| Hoffer 1964 | 1964 | case series/clinical database | yes | Ten-year follow-up of niacin treatment (level from title; outcomes not in the PubMed record) |
| Walsh 2004 | 2004 | case series/clinical database | yes | 207 consecutive cases, uncontrolled; 92% of compliant assaultive cases reported fewer assaults |
| Faber 2009 | 2009 | case series/clinical database | yes | Review of prior records for 230 children with ASD; mean Zn/Cu 0.608 |
| Hambly 2017 | 2017 | case series/clinical database | yes | Open-label micronutrient trial, 31 completers, MOAS d = 1.26 |
| Pfeiffer 1982 | 1982 | book/expert opinion | yes | Review: zinc, the brain and behavior |
| McGinnis 2008a | 2008 | book/expert opinion | yes | Narrative review, Discerning the Mauve Factor Part 1 (Altern Ther Health Med) |
| McGinnis 2008b | 2008 | book/expert opinion | yes | Narrative review, Part 2 |
| Walsh 2012 | 2012 | book/expert opinion | no | Book: Nutrient Power (Skyhorse) |
| Mensah 2019 | 2019 | book/expert opinion | no | Townsend Letter article listed by Mensah Medical; no peer-reviewed Mensah paper found |
Source: Firth et al., 2019a (33 meta-analyses, n = 10,951); Firth et al., 2019b (corrected values; 16 RCTs, n = 45,826)
Limitations: all zinc RCTs were rated at high risk of attrition bias; N-acetylcysteine heterogeneity was I² = 92.64%; Firth et al. (2019a) characterize the vitamin D evidence as emerging.
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Omega-3, MDD (13 RCTs) | 0.398 (SMD) | 0.114 to 0.682 | 1,233 | Firth et al., 2019a |
| Omega-3 >50% EPA | 0.61 (SMD) | 0.38 to 0.85 | 969 | Firth et al., 2019a |
| Omega-3 ≥50% DHA | -0.028 (SMD) | -0.21 to 0.16 | 469 | Firth et al., 2019a |
| Folate-based adjunct | 0.37 (SMD) | 0.01 to 0.72 | 904 | Firth et al., 2019a |
| Folic acid | 0.4 (SMD) | -0.08 to 0.88 | 657 | Firth et al., 2019a |
| Methylfolate 15 mg | 0.73 (SMD) | 0.28 to 1.19 | 99 | Firth et al., 2019a |
| Methylfolate incl. 7.5 mg | 0.34 (SMD) | -0.4 to 1.08 | 249 | Firth et al., 2019a |
| Vitamin D | 0.58 (SMD) | 0.45 to 0.72 | 948 | Firth et al., 2019a |
| Zinc, adjunct in MDD | 0.66 (SMD) | 0.26 to 1.06 | 104 | Firth et al., 2019a |
| Magnesium | 0.22 (SMD) | -0.17 to 0.48 | 538 | Firth et al., 2019a |
| N-acetylcysteine | 0.37 (SMD) | 0.19 to 0.55 | 574 | Firth et al., 2019a |
| Inositol | 0.35 (SMD) | -0.2 to 0.89 | 188 | Firth et al., 2019a |
| Dietary improvement, depression | 0.162 (g) | 0.055 to 0.269 | 45,826 (16 RCTs) | Firth et al., 2019b |
| Dietary improvement, anxiety | 0.085 (g) | -0.031 to 0.202 | 2,270 (11 RCTs) | Firth et al., 2019b |
Source: Hambly et al., 2017, 31 completers; within-group pre-post change in aggression (MOAS), not comparable with the controlled effects on depression and anxiety above
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Micronutrients, aggression (MOAS; open-label) | 1.26 (d) | — | 31 completed | Hambly et al., 2017 |
Source: Swardfager et al., 2013, 17 studies, n = 1,643 depressed and 804 controls
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| All studies | -1.85 (WMD) | -2.51 to -1.19 | 1,643 depressed; 804 controls | Swardfager et al., 2013 |
| Inpatients | -2.543 (WMD) | -3.522 to -1.564 | 1,643 depressed; 804 controls (review total) | Swardfager et al., 2013 |
| Community samples | -0.943 (WMD) | -1.563 to -0.323 | 1,643 depressed; 804 controls (review total) | Swardfager et al., 2013 |
| Higher-quality studies | -2.354 (WMD) | -2.901 to -1.807 | 1,643 depressed; 804 controls (review total) | Swardfager et al., 2013 |
A meta-review of 33 meta-analyses (10,951 individuals) found the strongest evidence for polyunsaturated fatty acids, particularly eicosapentaenoic acid, as an adjunctive treatment for depression, positive effects of high-dose methylfolate in major depressive disorder, emergent evidence for N-acetylcysteine, and good safety profiles across supplements (Firth et al., 2019a). A systematic review found primarily positive results in replicated studies of adjunctive S-adenosylmethionine, methylfolate, omega-3 and vitamin D alongside antidepressants and mixed results for zinc, folic acid, vitamin C and tryptophan; of these, only omega-3 (a significant, moderate-to-strong effect) and folic acid (not significant) were meta-analyzed, with marked heterogeneity and funnel-plot asymmetry indicating potential study bias (Sarris et al., 2016). A meta-analysis of 16 randomized trials (45,826 participants; 15 of the 16 in samples with nonclinical depression) found that dietary interventions reduced depressive symptoms (g = 0.162, 95% CI 0.055 to 0.269, as corrected), with no effect on anxiety (Firth et al., 2019b; corrections: “The Effects of Dietary Improvement …: Erratum,” 2020; “…: Second Erratum,” 2021). In people with gluten-related disorders (mostly biopsy-proven celiac disease, plus celiac-antibody-positive adults and one cohort with diarrhea-predominant irritable bowel syndrome), depressive-symptom scores improved on a gluten-free diet (8 studies, 953 participants: one randomized trial and seven single-arm before-and-after studies; SMD −0.37, 95% CI −0.55 to −0.20), and in two crossover trials of people with self-reported non-celiac gluten sensitivity a blinded gluten challenge showed a nonsignificant trend toward worse depressive symptoms (p = 0.25); effects in people without a gluten-related disorder were not studied (Busby et al., 2018).
Meta-analyses reported statistically significant effects for some adjunctive nutrients and dietary improvement at population level, with marked heterogeneity between studies. Which individuals respond, which dietary components matter and how to deliver these interventions are open questions the authors themselves identify.
Method (Planned)
Prospective observational cohort of adult participants using self-serve education, gated on IRB approval: participants log mood and nutrition in the app and complete the measures at baseline, 6 and 12 weeks.
Data and measures (Planned)
PHQ-9 (Kroenke et al., 2001) and GAD-7 (Spitzer et al., 2006), administered only after IRB approval; participant mood and nutrition logs. Participant safety: any PHQ-9 item 9 response other than “Not at all” immediately displays the 988 Suicide & Crisis Lifeline (call or text 988 in the US) and emergency-services guidance, and alerts the study’s qualified clinician, who reviews it within one business day and offers the participant contact through the app under the IRB-approved risk-management procedure; each positive response and the action taken are logged and reported to the IRB.
Analysis plan (Planned)
Mixed-effects growth models of PHQ-9 and GAD-7 scores with a random intercept and slope per participant estimate individual response alongside the group effect, adjusted for baseline symptom severity, age, medication use and baseline diet; the primary contrast is plant-based dietary change, and gluten-free adherence from the nutrition logs enters as a pre-specified exploratory moderator (interaction with plant-based change), reported as hypothesis-generating; the analysis is pre-registered before data collection.
Built component: Wallet-keyed participant mood and nutrition logs, readable only by the participant or a Sovereign Guide with an active grant: evidence.futuresystemslab.io/advisors/ghasemzadeh
Feeds: VALIDATE
Evidence
Hospitals engaged in all four interoperability domains (send, receive, find and integrate) rose from 23% in 2014 to 76% in 2025.
Source: ONC Health IT Quick Stat #66, non-federal acute care hospitals (2020 survey reported as 2021; no 2024 value)
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| 2014 | 23% | — | — | ONC Quick Stat #66 (2014) |
| 2015 | 26% | — | — | ONC Quick Stat #66 (2015) |
| 2016 | 29% | — | — | ONC Quick Stat #66 (2016) |
| 2017 | 41% | — | — | ONC Quick Stat #66 (2017) |
| 2018 | 46% | — | — | ONC Quick Stat #66 (2018) |
| 2019 | 55% | — | — | ONC Quick Stat #66 (2019) |
| 2021 | 62% | — | — | ONC Quick Stat #66 (2021) |
| 2022 | 70% | — | — | ONC Quick Stat #66 (2022) |
| 2023 | 70% | — | — | ONC Quick Stat #66 (2023) |
| 2025 | 76% | — | — | ONC Quick Stat #66 (2025) |
HL7 FHIR Release 4 (v4.0.1) defines the Observation resource for measurements and simple assertions, and its Vital Signs profile codes heart rate as LOINC 8867-4. FHIR R4 also defines a Consent resource for recording a healthcare consumer’s choices. SMART on FHIR lets applications run unmodified across health record systems using contemporary web standards for transport, authorization and interface, and its authors report the need for pragmatic constraints on base FHIR (Mandel et al., 2016).
SMART on FHIR standardizes how applications reach data in electronic health record systems, and FHIR defines how consent is represented. How a participant’s own wallet-signed consent can govern FHIR data held outside those systems remains open.
Method (Planned)
Conformance study: a fixed corpus of synthetic HL7v2 ORU^R01 messages and wearable readings runs through the bridge and the wearable route, and each output resource is validated and compared field by field with its source.
Data and measures (Planned)
HL7 FHIR R4 validator results against the R4 Observation and Vital Signs profiles; data completeness rate (source fields present in the output); mapping fidelity rate (values, units and LOINC codes matching the source).
Analysis plan (Planned)
Validation errors and warnings are counted per resource type, completeness and fidelity are reported as proportions with 95% confidence intervals, and every failure is traced to its mapping rule.
Built component: HL7v2 → FHIR R4 interface-engine bridge: evidence.futuresystemslab.io/hl7-fhir; Google Health → FHIR R4 wearable route: demo.futuresystemslab.io
Feeds: BUILD
Evidence
Of 65 papers in a systematic review of blockchain in healthcare, 45 were technical and 14 of those presented a proof-of-concept implementation.
Source: Agbo et al., 2019, 65 included papers
| Item | Value | 95% CI | n | Source |
|---|---|---|---|---|
| Technical papers | 45 | — | 65 papers | Agbo et al., 2019 |
| Proof-of-concept implementations | 14 | — | 65 papers | Agbo et al., 2019 |
A review for biomedical informatics sets out blockchain features, their benefits for health care applications compared with traditional distributed databases, current applications, and the challenges of adoption with proposed solutions (Kuo et al., 2017). A PRISMA systematic review found many proposed healthcare use cases for blockchain but a lack of adequate prototype implementations and of studies characterizing their effectiveness (Agbo et al., 2019). BIP-32 hierarchical deterministic wallets let account public extended keys be shared so that all incoming and outgoing transactions can be seen without exposing any secret key, the basis of a watch-only wallet.
The systematic review finds few working prototypes and little evaluation of effectiveness for healthcare uses of blockchain. A 2025 systematic review reports that adoption remains limited to conceptual work and proofs of concept, often on simulated datasets rather than validated real-world data (Shaikh et al., 2025). Evidence from deployed, measured systems operated by small practices therefore remains limited.
Method (Planned)
Operational measurement over a fixed observation window on production infrastructure, using scripted test invoices settled in Bitcoin through the self-hosted BTCPay server and anchored record batches on the Ethereum Sepolia testnet.
Data and measures (Planned)
Settlement correctness (recorded against BTCPay-settled invoices); anchoring latency (record creation to confirmed root) and cost (gas per batch and per record); verification success rate (records whose proof recomputes to the anchored root).
Analysis plan (Planned)
Correctness and verification success are reported as proportions with 95% confidence intervals, and latency and cost as medians with interquartile ranges per batch size.
Built component: Self-hosted BTCPay settlement to a watch-only wallet, with record Merkle roots anchored to MerkleAnchor on the Ethereum Sepolia testnet: evidence.futuresystemslab.io/sovereign-clinical
| Stream | Design | Sample | Instruments | Analysis | Status |
|---|---|---|---|---|---|
| Stigma and Access | Two-arm randomized comparison of entry flows | Adults seeking self-serve mental-wellness education, recruited online; size set by an a priori power analysis | BACE (Clement et al., 2012); ATSPPH-SF (Fischer & Farina, 1995), administered anonymously in both arms; entry-flow completion | Linear mixed-effects models (arm × time), adjusted for baseline covariates, by intention to treat; logistic regression for entry-flow completion | Planned · IRB-gated |
| Participant Autonomy and Data Sovereignty | Within-subject comparison, counterbalanced | Adult participants | QuIC (Joffe et al., 2001); SUS (Brooke, 1996); MAUQ (Zhou et al., 2019); logged grant and revocation success and time | Non-inferiority on QuIC (5-point margin; one-sided 97.5% CI) by paired t-test or Wilcoxon signed-rank with Hodges–Lehmann interval, condition order checked for carryover; SUS and MAUQ descriptive; revocation success as a proportion with 95% CI | Planned · IRB-gated |
| Privacy and Consent | Formal protocol verification; consent-text comprehension | Symbolic models of the EIP-191 and EIP-712 variants; adult participants for QuIC | Tamarin (Meier et al., 2013) or ProVerif (Blanchet, 2001); QuIC (Joffe et al., 2001) | Lemmas proved or refuted; counterexample traces with fix and re-verification | Planned (verification) · IRB-gated (QuIC) |
| Orthomolecular and Neurobiological Evidence | Prospective observational cohort (baseline, 6 and 12 weeks) | Adult participants using self-serve education | PHQ-9 (Kroenke et al., 2001); GAD-7 (Spitzer et al., 2006); mood and nutrition logs, including gluten-free adherence | Mixed-effects growth models adjusted for baseline severity, age, medication use and baseline diet; primary contrast plant-based dietary change; gluten-free adherence as a pre-specified exploratory moderator; pre-registered | Planned · IRB-gated |
| Informatics and Interoperability | Conformance study | Fixed corpus of synthetic HL7v2 messages and wearable readings | HL7 FHIR R4 validator (HL7 International, n.d.-b); completeness and fidelity rates | Error and warning counts; proportions with 95% CI | Planned |
| Decentralized Architecture | Operational measurement over a fixed window | Scripted test invoices settled in Bitcoin through the self-hosted BTCPay server and anchored record batches on the Ethereum Sepolia testnet | Settlement correctness; anchoring latency and cost; verification success rate | Proportions with 95% CI; medians with IQR | Planned |
Every human-participant study on this page is Planned and begins only after IRB approval. Any PHQ-9 item 9 response other than “Not at all” triggers the participant safety procedure in the orthomolecular study design. Platform sign-in collects a wallet address and a single-use nonce only. In the stigma study, the comparison arm registers with name and email, and study measures are administered anonymously in both arms. Consent is managed by the participant’s wallet: each grant or revocation of a single Sovereign Guide’s access is a participant wallet signature verified with a single-use nonce; the participant can also revoke every grant at once from the signed-in session, without a separate signature.
Data collection from external participants begins after IRB approval; every study on this page is Planned. The work has a single evaluator to date. Several outcomes rely on self-report measures. The orthomolecular cohort is observational, which limits causal inference. A self-selected online sample limits generalizability. Formal verification covers the modeled protocol only. System limitations are documented under Known Limitations on the portfolio home page (https://evidence.futuresystemslab.io/#limitations), and security findings and controls on the security page. Participants in the wallet conditions will sign with standard (EOA) wallets, the signature type EncryptHealth verifies.