nahass.ai — Est. 2026
Physician · Informaticist · Researcher

Data, insight, action — in the order that saves lives.

I'm a critical-care physician and VP of Health Informatics who builds the things I write about. This site documents work across Clinical, Quality, Predictive, and Operational intelligence — shipped at scale inside a twelve-hospital health system, published in peer-reviewed journals, and presented nationally.

View work → Connect →
Selected outcomes
18–26%
ICU mortality reduction, enterprise deterioration-index deployment across 12 hospitals.
NEJM AI · Pending
>3%
Absolute sepsis mortality reduction — bundle compliance, coding, and source control.
12-hospital initiative · 2024–25
$50M
Incremental revenue from AutoCDI. CMI +0.4; mortality index from 1.2 to 0.7.
Documentation strategy · 2025
44%
IV fluid reduction during a national shortage — shorter LOS, published nationally.
J. Healthcare Quality · 2025
01 — Clinical Intelligence

CDS built by a clinician who rounds.

Most clinical decision support fails because it was designed by people who have never been paged at 2am. I build from the Epic backend outward — Physician Builder expertise, OPA logic, and notification design — and study the impact of every single change before declaring it a success.

Methodology: hypothesize, build, measure, iterate. Every intervention earns its place with data. Most impactful project: AutoCDI — a multi-phase documentation strategy that moved CMI by 0.4 points and captured ~$50M in incremental revenue. Most precise: notification-driven extubation reminders for CABG patients, reducing unnecessary ventilator days through targeted, evidence-backed nudges.

Epic ClarityCaboodle Physician BuilderMSSQL TableauPythonR
1.1 — AutoCDI

Documentation Intelligence

Multi-phase CDI strategy integrating real-time physician prompts, discharge clarification notes, and outcome attribution. Built to move the mortality index and CMI — not just compliant documentation.

  • Physician CDI no-response reduction
  • Discharge clarification workflows
  • DNR documentation accuracy +10%
  • CMI & mortality index tracking
1.2 — OPA Design

Our Practice Advisories & Order Sets

Alert architecture that clinicians respect. Trigger logic, response options, and tiered urgency — designed and measured for acceptance rates and downstream behavior change to earn every alert's continued existence.

  • CABG extubation reminders
  • Alert fatigue analysis & reduction
  • Sepsis & RRT trigger design
  • Order set optimization
1.3 — Sepsis & Early Warning

Clinical Early Warning Programs

Bundle compliance, source control workflows, and rapid response trigger design — connected into the EHR so the right person is notified before the patient deteriorates further. Every threshold validated before going live.

  • Sepsis bundle compliance & coding
  • Rapid response escalation logic
  • ICU transfer decision support
  • HRRP readmission pathways
1.4 — Physician Builder

Physician Builder Programs

Led a 25+ Physician Builder and 45 SmartUser program at RWJBarnabas Health — from recruitment and governance design through build execution and outcome tracking.

  • Physician Builder recruitment & training
  • SmartUser program design
  • Build governance & prioritization
  • Epic build execution & testing
02 — Quality Intelligence

Every index, every ranking, every opportunity.

CMS claims, Epic Clarity, Caboodle, Vizient, and clinical data connected into a unified analytical layer — producing reports that translate into decisions. Not one source. The whole picture, interpreted by a clinician who knows what each number means at the bedside and in the boardroom.

Work spans Vizient Mortality, LOS, and Direct Cost Index improvement; competitor and opportunity analysis; USNWR ranking strategy; and CMS star rating programs. Every report physician-authored, peer-benchmarked, and built to drive action — not fill a slide deck.

MSSQLPythonR TableauSnowflake Google CloudAzureAWS
2.1 — Vizient

Vizient Mortality, LOS & Cost

Observed-to-expected deep dives by MS-DRG, service line, and attending — with specific, prioritized action plans identifying whether a mortality index problem is coding, clinical, or both.

  • O/E mortality disaggregation
  • Direct cost index analysis
  • LOS outlier identification
  • Peer group benchmarking
2.2 — CMS Stars

CMS Star Rating Improvement

Systematic analysis across all star measure groups — HCAHPS, outcomes, safety — with intervention prioritization based on measure weight, current gap, and realistic ceiling.

  • Measure-level gap analysis
  • HRRP & HAC reduction strategy
  • VBP program optimization
  • Star trajectory modeling
2.3 — USNWR

US News Rankings Strategy

Ranking methodology decoded — identifying where points are lost, where the fastest gains live, and what clinical and operational changes move the number, with data that survives CMO and board scrutiny.

  • Specialty ranking deep dives
  • Structural & process measure gaps
  • Competitor quality comparison
  • Multi-year trajectory planning
2.4 — Market Intelligence

Competitor & Opportunity Analysis

Using CMS, Vizient, and public claims data to map competitive quality position across service lines and geographies — identifying where market-share gaps and quality performance create high-value opportunities.

  • Service-line market share analysis
  • Quality-adjusted positioning
  • Growth opportunity ranking
  • Strategic board briefings
03 — Predictive Intelligence

ML that goes from model to clinical workflow.

Operationalizing clinical AI is not a data science problem — it is a clinical workflow problem wrapped in a governance problem. I work across all three: threshold optimization, silent-mode validation, ambient AI monitoring, and enterprise-grade generative AI governance.

I've deployed deterioration indexes, readmission risk scores, and ICU mortality models across twelve hospitals — and built a HIPAA-compliant, auditable LLM infrastructure, multimodal and guardrail-enabled, for enterprise health system use.

PythonR SnowflakeGoogle Cloud AzureAWSMSSQL
3.1 — ML Operationalization

Model Deployment & Threshold Optimization

Taking models out of notebooks and into workflow. Deterioration index, readmission risk, ICU mortality — each validated in silent mode, thresholds optimized per population, escalation rules designed to change behavior.

  • Silent-mode → production sequencing
  • Threshold & sensitivity optimization
  • Subgroup & fairness validation
  • Post-deployment drift monitoring
3.2 — Generative AI

GenAI Implementation & Governance

End-to-end generative AI programs — from analyst tooling to physician-facing applications — with governance frameworks that get approved by legal and compliance, not just the innovation committee.

  • Use-case scoping & prioritization
  • Enterprise governance policy design
  • Clinical workflow integration
  • ROI measurement frameworks
3.3 — Ambient AI

Ambient AI Operationalization

Ambient documentation tools are being adopted fast — with almost no clinical or ROI accountability. I build measurement frameworks tracking actual value: time saved, note quality, downstream alert performance, and attribution to outcomes.

  • Vendor-agnostic evaluation
  • Clinical quality monitoring
  • ROI attribution modeling
  • Workflow integration design
Proprietary Infrastructure

HIPAA-Compliant Enterprise LLM Services

Built — not procured. A fully HIPAA-compliant cloud LLM infrastructure supporting multimodal enterprise use across GCP, Azure, and AWS. Every prompt logged. Every output traceable. Every guardrail configurable. Designed for health systems that need auditability and observability before clinical deployment — from analysts writing SQL to physicians reviewing summaries.

HIPAA-compliant by design
Full audit logging
Observability dashboard
Configurable guardrails
Multimodal support
GCP · Azure · AWS ready
04 — Operational Intelligence

Supply chain, utilization, and the data between the systems.

Operational intelligence is what happens when clinical data meets operational data and a physician interprets it. Not throughput for its own sake — utilization, supply, and staffing patterns analyzed through the lens of patient outcomes.

During the 2024 national IV fluid shortage, I built a novel integration of barcode scanning data, nursing administration records, and clinical outcome data — producing a real-time utilization dashboard that achieved a 44% fluid reduction while simultaneously shortening length of stay. Published in the Journal for Healthcare Quality.

MSSQLTableauPython RClarityCaboodle Snowflake
4.1 — Supply Chain Intelligence

Supply & Utilization Analytics

Supply chain, barcode scanning, nursing administration, and clinical outcome data connected to produce utilization intelligence that clinical and operational teams can act on together — not separately and too late.

  • Real-time utilization dashboards
  • Shortage response & substitution logic
  • Outcome-linked consumption analysis
  • Nursing workflow integration
4.2 — Throughput & Capacity

Throughput & Capacity Intelligence

LOS outliers, boarding patterns, discharge timing, and bed utilization — analyzed with clinical context so that the throughput recommendations don't inadvertently trade off patient safety for operational efficiency.

  • LOS driver analysis by DRG
  • Discharge barrier identification
  • Boarding & boarding-origin patterns
  • Capacity planning with clinical context
4.3 — Surveillance Dashboards

Real-Time Operational Surveillance

Built a respiratory virus dashboard detecting infection upticks two weeks before CDC data surfaced — enabling timely masking mandates. Surveillance tooling designed to give operational leaders early signal rather than lagging confirmation.

  • Infection & respiratory surveillance
  • Early-signal vs. lagging-indicator design
  • Policy-trigger threshold setting
  • MSSQL · Tableau · real-time refresh
4.4 — Cross-System Integration

Data Integration & Reporting Services

CMS claims, Epic Clarity, Caboodle, supply chain, and nursing data connected into a single analytical layer — producing the recurring reports, dashboards, and executive summaries that drive actual decisions.

  • Cross-dataset normalization
  • Custom dashboard & report builds
  • Executive & board summaries
  • Recurring intelligence delivery
— Principal

Physician. Informaticist. Builder.

Thomas A. Nahass, MD, MS, FAMIA
— Thomas A. Nahass, MD, MS, FAMIA

Thomas A. Nahass

MD · MS · FAMIA

Tom is a board-certified critical-care physician, Clinical Assistant Professor of Medicine, and the Vice President of Health Informatics (Central Region) at RWJBarnabas Health — a twelve-hospital health system where he executes numerous projects across clinical decision support, AI deployment, documentation integrity, quality programs, and operational analytics.

His perspective is deliberately integrated. Actionable intelligence is not any single tool — it is what happens when clinical data, operational data, quality data, and predictive models are synthesized by someone who still rounds on patients and sits in the steering committee the next morning. He has shipped every service this practice offers.

He trained in internal medicine at Rutgers and critical care at Memorial Sloan Kettering, earned an MS in clinical informatics from OHSU, and is a Fellow of the American Medical Informatics Association. Becker's Rising Stars: Healthcare Leaders Under 40, 2025.

Board Certifications
Critical Care Medicine · Internal Medicine · Clinical Informatics
Academic Appointment
Clinical Assistant Professor, Rutgers RWJMS · Pulmonary & Critical Care
Current Research
Heart failure deep phenotyping · ICU digital twins · Agentic AI & RAG
Tech Stack
MSSQL · Python · R · Tableau · Snowflake · GCP · Azure · AWS
— Selected work & publications

A record of shipped work.

2025
Mortality impact of an AI-enabled deterioration index integrated with automated rapid response
Original Research
NEJM AI · Submitted
2025
AutoCDI multi-phase documentation strategy — CMI +0.4, mortality index 1.2→0.7, ~$50M captured
CDS Program
RWJBarnabas Health
2025
Navigating an IV fluid shortage: barcode integration + nursing data drives 44% reduction, shorter LOS
Operational Intelligence
J. Healthcare Quality
2025
Sepsis mortality initiative — >3% absolute reduction across 12 hospitals via bundle compliance & coding
Quality Program
RWJBarnabas Health
2025
AI and ICU Design: enhancing critical-care efficiency
Oral Presentation
CHEST Annual Meeting
2024
Enterprise deterioration-index rollout: 18–26% ICU mortality reduction, recognized at Epic UGM
Predictive Intelligence
Epic UGM Feature
2024
Respiratory virus surveillance dashboard detecting upticks 2 weeks before CDC data
Operational Intelligence
RWJBarnabas Health
2017
Asynchronous automated electronic laboratory result notifications: a systematic review
Systematic Review
JAMIA

Open to advisory conversations, collaboration, and speaking.

I'm available for academic collaboration, advisory roles, conference presentations, and peer consultation. Not currently seeking new clinical or operational engagements outside my primary role.

Email[email protected]
LinkedInthomas-nahass
Webnahass.ai
BasedNew Jersey · New York