






Healthcare AI tools are typically built and technically validated, then deployed with no assessment of whether they were needed in the first place and no evaluation of what they changed.
We are developing a three-stage framework that treats AI as a sociotechnical intervention rather than a technical product.
Stage 1 defines the healthcare or public health problem, compares AI against feasible non-AI responses, and produces an auditable appropriateness decision before development begins.
Stage 2 develops and validates the tool within those boundaries and ends in a deployment-readiness decision.
Stage 3 evaluates effectiveness, implementation, and the distribution of benefits and burdens under real-world conditions, feeding findings back to Stage 1. Equity, trustworthiness, relational care, and proportionate AI use operate across all three stages.
Objective: Whether Boston Public Schools, the Office of Worker Empowerment, and the Office of Housing delivered an equitable share of benefits to Black male residents of Boston between July 1, 2021 and June 30, 2026.
Methods: Extraction of program rules, eligibility regulations, budget books against a barrier taxonomy separating policy, institutional, and operational constraints. A CPA-led review traces dollars allocated, dollars spent, and service population by race and gender for every program line item. Legal review tests the record against the strong-basis-in-evidence standard from City of Richmond v. J.A. Croson Co.
Analysis: Observed Black male utilization compared against expected utilization derived from American Community Survey 5-year estimates for the eligibility-matched population. Significance tested using z-tests for proportions, chi-square tests, and adjusted logistic regression with race-by-gender interaction terms. Cells under 30 are tested using Fisher's exact test and 95% confidence intervals.
Deliverables: A nine-section disparity study report delivered January 2027, comprising disparity findings for every program in the analytic universe, gap matrices by department, a resource-flow database disaggregated by race and gender, a legal memorandum on narrow tailoring and race-neutral alternatives, and recommendations classified by policy, institutional, and operational lever
Status: Ongoing.


Purpose: Develop a Massachusetts-ready system for obstetric transfer across Levels of Maternal Care, with clear guidance for hospitals, EMS, patients, and families.
Method: Rapid evidence review, Massachusetts system evaluation, pathway mapping, provider interviews, spatial analysis, and QI measure design.
Activities: landscape synthesis, stakeholder interviews, transfer pathway mapping, protocol development, EMS-hospital workflow review, drill planning, and public resource hub content development.
Outputs: statewide protocol suite, QI measure set, data dictionary, run-chart templates, drill and debrief playbook, resource hub content pack, draft report, and final report.
Capacity building: supports hospitals and EMS with practical tools for readiness, activation, stabilization, transport, reception, handoff, debrief, quality monitoring, and training.
Status: Ongoing.
Role: independent evaluator for the statewide MASStrong peer-support program.
Population: nursing-home aides, home-health workers, personal-care attendants, and other long-term-care staff.
Program Format: confidential 90-minute online groups plus facilitator training and supervision tracks.
Evaluation Framework: CDC 6-Step Program Evaluation Framework.
Scope: support groups, facilitator training, skill sessions, certification, continuing education, supervisor training.
Evaluation Plan: co-develop an evaluation plan with partners. Define questions, indicators, data sources, timelines, and inclusion strategies for non-English speakers.
Methods: mixed methods. Surveys, interviews, attendance and training records, and administrative data.
Outcomes tracked: reach, engagement, retention, facilitator readiness, participant wellbeing, and workforce measures.
Data management: deidentified datasets, codebooks, versioned protocols, and reproducible code.
Deliverables: evaluation plan, instruments, dashboards, interim briefs, and a final report.
Status: ongoing.

Objectives: To identify the suicide prevention programs operating in all 50 US states during 2026, to determine the population each program states it serves, and to establish whether that population corresponds to the subgroups carrying the highest suicide burden in the same state.
Method: A search of state agency web domains, conducted in July 2026. For each program, extract the implementing agency, program type, and stated target population, with duplicate extraction and a blinded reliability check. State suicide burden profiles from CDC WONDER, 2018 to 2024, and Department of Veterans Affairs state data sheets. Tailored Population Score to score each program against the three highest-burden population subgroups of its own state.
Intended outcomes: A per-state scorecard of program-to-burden correspondence, a program dataset covering all 50 states, state-level suicide burden and veteran burden datasets, and a scoping review manuscript in preparation.
Status: Ongoing.
Objective: map all providers serving Medford residents.
Catalog: list each organization, services offered, eligibility, and access details.
Methods: provider inventory, validation with agencies, and structured data collection.
Analysis: gap analysis by population, service type, language, hours, and location.
Deliverables: comprehensive directory, gap analysis, and a written report.
Outcome: delivered the directory and report to the City.
Status: completed.

Goal: inform the next procurement for gun violence prevention services.
Lens: racial equity across design, data, and recommendations.
Methods: quantitative and qualitative data collection and analysis.
Community engagement: plan and run sessions with youth, families, and providers.
Literature review: assess evidence on effective youth and young-adult supports.
Data synthesis: integrate engagement, reports, and prior program data.
Recommendations: identify needs, best practices, and funding priorities.
Deliverables: evidence brief, engagement summary, metrics for RFP, draft logic model.
Status: ongoing.


Purpose: identify priority needs for seniors, BIPOC residents, and unhoused residents in Amherst, MA.
Method: Deliberative Engagement sessions with quantitative and qualitative components.
Activities: instrument design, facilitated sessions, surveys, interviews, and rapid analysis.
Outputs: needs assessment report, evidence tables, validated instruments, training materials.
Capacity building: trained community responders to run Deliberative Engagement sessions.
Status: awarded and completed.
Goal: unify health and human services data across departments.
Platform: integrated data layer linking Smartsheet, Unite Us, Google Workspace, and legacy systems.
Monitoring and evaluation: automated dashboards for KPIs and DEI metrics. Versioned code and data dictionaries.
Governance and security: HIPAA alignment, role-based access, audit logs, encryption, backup and recovery.
Interoperability: standardized schemas, API connectors, and ETL pipelines.
Accessibility: simple user workflows, forms, and reports for staff and leadership.
Scalability: cloud-ready deployment and load-tested performance.
Status: ongoing. Current work includes system assessment, data-flow mapping, draft data model, and a migration plan for a single source of truth.

Purpose: plan and launch a first-of-its-kind research program on Black male equity in Boston.
Focus areas: economic opportunity, health, education.
Literature review: systematic search to map gaps and priorities.
Methods: deliberative engagement sessions, stakeholder interviews, resident interviews, surveys.
Participants: Boston stakeholders and residents, with focus on Black males.
Data management: deidentified datasets, codebooks, versioned protocols.
Outputs: needs assessment report, evidence map, sampling plan, instrument bank.
Status: ongoing.
Focus: outcomes and equity effects of alternative public safety models, including diversion.
Population: service recipients with attention to BIPOC groups and trauma from over policing.
Methods: site visits, surveys of service recipients and grantee staff, interviews with service recipients, quantitative and qualitative analysis.
Outputs: final report, cleaned datasets, codebooks, reproducible analysis files, executive brief.
Use: informed program decisions on equity and effectiveness.
Status: completed.


Scope: process, outcome, impact, and cost effectiveness
Context: program launched in 2020; no evaluations after 2021
Approach: concurrent process and outcomes evaluation
Methods: site visits; surveys of service recipients and grantee staff; interviews with service recipients
Status: ongoing
Focus: health equity analytics, training automation, and compliant authoring.
Products:
Real-world data analytics for public health and equity research.
AI chatbot to train sales representatives with scenario practice and feedback.
LLM platform for structured authoring of regulatory documents.
Responsibilities:
Product design, workflows, and data pipelines.
Model selection, prompt design, and evaluation harnesses.
Human-in-the-loop review, red-teaming, and test plans.
Validation datasets, QA scripts, and benchmark reports.
Privacy, security, and audit logging.
Methods and features:
Deidentification and PHI scrubbing.
Retrieval-augmented generation with source citation.
Guardrails, role-based access, and version control.
Structured outputs using schemas and templates.
Outcomes:
Faster evidence synthesis and transparent provenance.
Consistent training with measurable knowledge gains.
Reduced authoring cycle time and fewer review rounds.
Clear traceability for regulatory and internal audits.
Tury Research Institute completed a formative evaluation of the Massachusetts EAPS program to assess implementation fidelity and equity in participation and outcomes.
The Equitable Approaches to Public Safety (EAPS) program looks to identify and implement alternative approaches to traditional public safety models that are effective and equitable to the residents of the Commonwealth that best serve their community.
Tury Research Institute continues to provide biostatistical services for Samsung Software as a Medical Device (SaMD) and digital health.


Copyright © 2026 Tury Research Institute - All Rights Reserved.