This pre-registered, pragmatic, three-arm (1:1:1) patient-level randomized controlled trial with mixed-effects analysis at the encounter level tests two questions in real high-risk multidisciplinary clinical encounters at the Waymark clinically integrated network across three U.S. states (Ohio, Washington, Virginia): (1) does adding ANCHOR - a clinical AI structural verification layer - to a Gemini 3.1 Pro-assisted supervising-physician workflow reduce the rate of clinically meaningful safety failures, compared with the same Gemini 3.1 Pro-assisted workflow without ANCHOR? (2) does the Gemini 3.1 Pro-assisted workflow itself reduce the same safety endpoint compared with unassisted standard care in which the supervising physician writes their own SOAP assessment/plan from a blank template? ANCHOR is a single-call structural verification layer combining a Logical Neural Network (Riegel et al. 2020) certificate, six specialist agents, and concept-decomposed output with PMID citation provenance. ANCHOR is physician-facing only and is used by supervising physicians, not by the multidisciplinary clinical team they oversee. The trial randomizes 240 patients 1:1:1 across the Waymark clinically integrated network over a 12-week active-enrolment window (80 per arm). Eligible patients are adults (age 18+) identified as high-risk by combined claims-based and clinical criteria. Eligible encounters span three integrated Waymark service modalities: high-risk primary care, specialty care coordination, and real-time telemedicine urgent care. The primary endpoint is a per-encounter binary composite: any of (a) failure to mention a do-not-miss diagnosis, (b) under-triage, (c) contraindicated medication recommendation, (d) failure to recommend escalation when clinically warranted; adjudicated by a blinded panel of 3 board-certified physicians with majority-of-three scoring. The primary contrast is Arm 3 (LLM+ANCHOR) versus Arm 2 (LLM with safety prompt), isolating ANCHOR's marginal contribution over a deployment-equivalent LLM safety stack. The pre-specified secondary contrast is Arm 2 versus Arm 1. The trial is sized to the operational ceiling of the Waymark integrated-network workflow across the three states (240 enrollees over 12 weeks). At realistic effect sizes derived from the retrospective evaluation, the trial is underpowered for definitive efficacy declaration on either pairwise contrast and is reported as an initial deployment-feasibility validation cohort with effect estimates and 95 percent confidence intervals; full power calculations are pre-registered in the Statistical Analysis Plan. Single-blind outcome adjudication: 3 adjudicators score only the supervising physician's final clinical decision, so all three arms produce adjudication packets in identical format and arm allocation is structurally invisible. Statisticians remain blinded until database lock. A full waiver of informed consent is requested per 45 CFR 46.116(f)(3) with a companion HIPAA waiver of authorization under 45 CFR 164.512(i)(2)(ii). The study is registered on the Open Science Framework prior to first enrollment and reported under CONSORT-AI 2020.
DESIGN. Pragmatic, multicenter, encounter-level, three-arm parallel randomized controlled trial. 1:1:1 patient-level stratified permuted-block randomization, block size 6, stratified by site and acuity stratum. Once a patient is randomized at first eligible encounter, all subsequent encounters for that patient remain in the same arm. ARMS. * Arm 1 (Unassisted standard care, n=80): No LLM. No ANCHOR. The supervising physician opens a blank SOAP note template and writes their own assessment and plan from scratch. * Arm 2 (Gemini 3.1 Pro with safety prompt, n=80): Gemini 3.1 Pro generates the recommendation under a clinical-safety system prompt, content filters, and retrieval-augmented generation. The supervising physician reviews the LLM output directly. This stack is operationally equivalent to LLM-assisted clinical-decision-support deployments already in routine use at major U.S. health systems. * Arm 3 (Gemini 3.1 Pro + ANCHOR, n=80): Gemini 3.1 Pro generates the recommendation under the same safety prompt as Arm 2; ANCHOR augments that output with the Logical Neural Network certificate, specialist-agent verification, and concept-decomposed audit trail. OPERATIONAL SIZING. The Waymark integrated-network workflow across Ohio, Washington, and Virginia captures approximately 20 eligible high-risk multidisciplinary encounters per week. With 12 weeks of active enrolment, total enrolment is 240 encounters (80 per arm at 1:1:1). POWER CALCULATION. Anchored on the architectural-argument retrospective cohort. For each pairwise contrast at n=80 per arm, the plausible effect range is 5 to 11 percentage points absolute reduction (midpoint 8 percentage points). At the midpoint planning effect (Arm 2 event rate 25 percent versus Arm 3 17 percent), power is approximately 0.50 for the verification-layer-specific contrast (Arm 3 versus Arm 2) at alpha = 0.05. The trial functions as a calibration cohort with effect estimates and 95 percent confidence intervals. PRIMARY ENDPOINT. Per-encounter binary composite, adjudicated by 3 board-certified physicians blinded to allocation; final scoring by majority of three. Components: (a) Failure to mention a do-not-miss diagnosis; (b) Under-triage; (c) Recommendation of a medication contraindicated by documented allergies/conditions/comorbidities; (d) Failure to recommend escalation when clinically warranted. PRIMARY CONTRAST: Arm 3 versus Arm 2 (ANCHOR-specific). PRE-SPECIFIED SECONDARY CONTRAST: Arm 2 versus Arm 1 (AI-introduction). SECONDARY ENDPOINTS (hierarchical fixed sequence, Holm-Bonferroni step-down at family-wise alpha = 0.05): (1) Appropriate triage escalation rate; (2) Time-to-physician-decision; (3) Clinician acceptance rate of ANCHOR flags (Arm 3 only); (4) 30-day emergency-department visit rate (EXPLORATORY); (5) 30-day hospitalization rate (EXPLORATORY); (6) Social-determinants-of-health response composite. STATISTICAL ANALYSIS. Mixed-effects logistic regression with arm (3-level categorical) as primary fixed-effect predictor and random intercept for patient nested within supervising physician; fixed-effect adjustment for site and acuity stratum. Gatekeeping for the two pairwise contrasts. Pre-specified sensitivity analyses: cluster-robust standard errors at patient level; generalized estimating equations with patient cluster and exchangeable correlation; patient-level aggregated analysis; Bayesian logistic regression with informative prior on the ANCHOR-specific contrast from the retrospective evaluation; exact methods. Missing data via multiple imputation by chained equations (m=20). Intention-to-treat primary with treatment-policy estimand (per ICH E9(R1)); per-protocol sensitivity excluding API-failure encounters. No interim efficacy stop. Pre-specified safety-halt threshold check at midpoint (n=120 cumulative across arms) for greater than 5 percentage points absolute increase in over-escalation in Arm 3 relative to Arm 2. ANCHOR DIAGNOSTIC-TEST CHARACTERIZATION (independent of trial primary endpoint). ANCHOR's certificate (operational state: hazard flagged / no hazard flagged / out of scope) is characterized as a diagnostic test against adjudicated ground truth: sensitivity, specificity, positive predictive value, negative predictive value, area under the receiver-operating-characteristic curve, and calibration metrics (Expected Calibration Error, Brier score, reliability diagram). PRE-SPECIFIED ALGORITHMIC FAIRNESS AUDIT. Equalized odds gap, equal opportunity gap, predictive parity gap, and within-subgroup calibration (Brier-difference tolerance 0.05) across race/ethnicity, sex assigned at birth, age band, urban/rural per RUCA 2024, and primary language. Multiplicity correction: Benjamini-Yekutieli false discovery rate with q=0.05. Disparity decision rule (locked pre-unblinding): any pairwise equalized-odds gap exceeding 0.10 with the lower bootstrap 95 percent confidence interval bound greater than 0.05 triggers a Limitations section, mitigation proposal, and a hold on broader deployment until mitigation is validated. DATA CAPTURE. The 30-day endpoints use Waymark's integrated electronic-health-record + real-time admission/discharge/transfer feed + claims pipeline. ADT events are detected in near-real time across the full network of acute and ambulatory facilities; claims are reconciled at day 30 to confirm capture completeness. Pre-trial registration locked at the Open Science Framework prior to first enrolment; reported under CONSORT-AI 2020.
Study Type
INTERVENTIONAL
Allocation
RANDOMIZED
Purpose
HEALTH_SERVICES_RESEARCH
Masking
SINGLE
Enrollment
240
Gemini 3.1 Pro generates the care-management recommendation under a clinical-safety system prompt, content filters, and retrieval-augmented generation. Supervising physician reviews the LLM output directly without ANCHOR augmentation.
Same Gemini 3.1 Pro generation as Arm 2, with ANCHOR additionally applied: a single-call structural verification layer combining a Logical Neural Network safety certificate over a 3,206-rule clinical logic library, six concurrent specialist agents (drug interaction, lab interpretation, guideline compliance, citation verification, safety net, differential-diagnosis breadth), and a concept-decomposition module with PMID-traceable provenance. Decision support only; clinician retains all clinical decision authority.
Waymark
San Francisco, California, United States
RECRUITINGPer-encounter clinical safety failure (adjudicated binary composite)
Adjudicated binary composite of any of: (a) failure to mention a do-not-miss diagnosis appropriate for the presentation; (b) under-triage (routine/semi-urgent when emergent/urgent appropriate); (c) recommendation of a medication contraindicated by documented allergies/conditions/comorbidities; (d) failure to recommend escalation when clinically warranted. Adjudicated by a blinded panel of 3 board-certified physicians (Internal Medicine, Family Medicine, or Medicine-Pediatrics); final scoring by majority of three. Reported as proportion of encounters with composite safety failure.
Time frame: At the encounter (encounter-level outcome adjudicated within 4 weeks post-encounter)
Appropriate triage escalation rate
Proportion of encounters in which the assigned triage level was appropriate to the clinical presentation, adjudicated.
Time frame: At the encounter
Time-to-physician-decision
Continuous duration in seconds from encounter start to supervising-physician final decision; log-transformed for analysis. Excludes API latency.
Time frame: Within the encounter (real-time)
30-day emergency-department visit rate (exploratory)
Any emergency-department visit within 30 days post-encounter, captured through Waymark's integrated electronic-health-record + real-time admission/discharge/transfer feed + claims pipeline. EXPLORATORY: underpowered at n=240; reported as effect estimate with 95 percent confidence interval without significance claim.
Time frame: 30 days post-randomization
30-day hospitalization rate (exploratory)
Any hospitalization within 30 days post-encounter, captured through the integrated EHR + ADT + claims pipeline. EXPLORATORY: underpowered at n=240; reported as effect estimate with 95 percent confidence interval without significance claim.
Time frame: 30 days post-randomization
This platform is for informational purposes only and does not constitute medical advice. Always consult a qualified healthcare professional.