# Cascading Clinical Data Search

*/Opportunities/Cascading_Clinical_Data_Search*

## Opportunity Overview

**Wedge**: The initial beachhead is retrospective chart review for oncology clinical trials at mid-sized research networks. Oncology trials contain complex inclusion criteria buried in unstructured pathology reports, creating acute pain and high failure rates for manual recruiters. Once the system accurately surfaces eligible oncology patients, the product expands horizontally into cardiology trials and eventually into automated disease registry reporting for the hospital itself.
**Timing**: Large language models with deep context windows and specialized medical embeddings now accurately map unstructured clinical narratives to standard ontologies without manual tagging. Additionally, recent mandates for standardized FHIR APIs provide a structured data baseline that an AI agent queries directly before falling back to unstructured text.
**Why This I C P**: Clinical trial recruiters at mid-sized Clinical Research Organizations face intense financial pressure to enroll patients quickly because delayed trials cost sponsors millions per day. They possess high patient data volume but lack the massive internal data-engineering teams of top-tier pharma, driving them to buy off-the-shelf workflow agents.
**Size Of Prize**: The addressable market comprises approximately 5,000 clinical trial sponsors and major research sites globally. Multiplying these 5,000 entities by an average annual spend of $100,000 on manual cohort identification and data abstraction labor yields a $500M annual prize.
**Gap Narrative**: Clinical researchers and trial coordinators manually cross-reference structured EHR data with unstructured physician notes to identify patient cohorts. Existing search tools require exact keyword matches or complex SQL queries, missing eligible patients whose conditions are described using non-standard medical terminology. A cascading search system interprets the clinical intent, queries structured databases first, and falls back to semantic search over unstructured notes to find exact clinical matches.
**Defensibility**: Defensibility stems from deep workflow lock-in and a compounding evaluation dataset of accepted versus rejected patient matches. As coordinators use the system, it builds a proprietary mapping of local clinical shorthand to standard trial criteria. Because the core retrieval technology is largely commoditized, long-term survival depends entirely on integrating directly into the trial management systems rather than relying on algorithmic superiority.
**Why This Thesis**: The Agent approach fits perfectly because cohort identification is fundamentally a multi-step reasoning task rather than a static database query. An autonomous agent recursively adjusts search parameters by querying structured databases, evaluating the patient yield, and semantically mining unstructured text to fill enrollment gaps.

## Opportunity Linked Thesis

**Thesis**: [Software](/Theses/Software)

## Opportunity Linked I C P

**Icp**: [Clinical Research Organization](/CompanyTypes/Clinical_Research_Organization)

## Opportunity Market Sizing

_Illustrative — target and order-of-magnitude estimate figures, not an achieved track record (this Thing is concept-stage)._

**S A M**: ~$500M-800M (Mid-to-large CROs managing multi-center Phase II-IV trials)
**S O M**: ~$15M-30M (Capturing ~60-120 mid-market North American CROs)
**T A M**: ~8,000 global clinical research organizations and sponsors × ~$250k/yr ≈ ~$2.0B
**Growth Rate**: ~12-18%/yr, driven by rising decentralized trial adoption and expanding multi-modal clinical data fragmentation
**Paid Comparable Spend**: ~$300k-500k/yr per CRO spent on manual data abstraction labor, fragmented CTMS queries, and custom reporting scripts

## Opportunity Incumbents

- [Epic Cosmos](/Products/Epic_Cosmos) — Tool
- [TriNetX Clinical Network](/Products/TriNetX_Clinical_Network) — Tool
- [Manual Chart Review](/Products/Manual_Chart_Review) — Service
- [Custom SQL Queries](/Products/Custom_SQL_Queries) — DIY
- [i2b2 Data Repository](/Products/i2b2_Data_Repository) — Open-Source
- [OHDSI Atlas Platform](/Products/OHDSI_Atlas_Platform) — Open-Source
- [Exported Excel Extracts](/Products/Exported_Excel_Extracts) — Spreadsheet

## Opportunity Win Conditions

**Kill Thresholds**:
- Initial data ingestion and indexing takes > 14 days per trial site
- Accuracy of search results against manual chart review < 95%
- Fewer than 5 active queries per week per onboarded data manager
- Pilot-to-paid conversion rate < 20% after 90 days
**Leading Metrics**:
- Time-to-first-query execution
- Weekly active queries per data manager
- Ratio of natural language queries to manual SQL fallbacks
- Number of disparate data sources indexed per trial
- False-positive rate on cohort inclusion criteria matches
**What Proves Right**: Clinical data managers log in daily to execute multi-modal queries across EDC and eSource platforms without writing SQL. Mid-market CROs replace at least half of their manual chart review hours within the first two trial phases. Customers sign six-figure annual contracts after a 30-day pilot demonstrates immediate labor savings.
**What Proves Wrong**: Data fragmentation proves too idiosyncratic per site, requiring bespoke engineering for every new trial connection. Search results return low-confidence matches for unstructured clinical notes, forcing users to manually verify every hit. Data managers abandon the platform and revert to exported Excel extracts because the tool fails to handle complex inclusion criteria.

## Opportunity Build Profile

**Hardest Part**: Consistently parsing and linking heavily abbreviated, negated, and unstructured physician notes to standardized medical ontologies without generating false positives from historical family mentions or negated symptoms.
**Min Viable Scope**: Focus exclusively on retrospective cohort identification for clinical trials within a single therapeutic area like oncology. Deliberately leave out point-of-care clinical decision support, automated billing workflows, and cross-institution federated search.
**Cold Start Problem**: Training robust clinical retrieval models requires real patient data, which is locked behind institutional firewalls and HIPAA restrictions. Break this by running initial model training on open datasets like MIMIC-IV, then securing a single design partner to deploy on-premise for a specific research cohort.
**Time To First Value**: 2–4 weeks to integrate and index historical EHR data before the first reliable cohort query executes
**Data Moat Available**: true
**Technical Difficulty**: High

## Neighborhood

### Incumbent in

- [Manual Chart Abstraction](/Products/Manual_Chart_Abstraction) — incumbent in · Products
- [Excel Data Export](/Products/Excel_Data_Export) — incumbent in · Products
- [Epic Cosmos](/Products/Epic_Cosmos) — incumbent in · Products
- [i2b2 Data Repository](/Products/i2b2_Data_Repository) — incumbent in · Products
- [Custom SQL Queries](/Products/Custom_SQL_Queries) — incumbent in · Products
- [OHDSI Atlas Platform](/Products/OHDSI_Atlas_Platform) — incumbent in · Products
- [TriNetX Clinical Network](/Products/TriNetX_Clinical_Network) — incumbent in · Products

### Applies thesis

- [Clinical Research Organization](/CompanyTypes/Clinical_Research_Organization) — applies thesis · CompanyTypes

### Embodies

- [Software](/Theses/Software) — embodies · Theses

### Similar Opportunities

- [Protocol Compliance Agent](/Opportunities/Protocol_Compliance_Agent) — similar · Opportunities
- [Discharge Compliance Agent](/Opportunities/Discharge_Compliance_Agent) — similar · Opportunities
- [Care Coordination Agent](/Opportunities/Care_Coordination_Agent) — similar · Opportunities
- [IRB Protocol Preparation](/Opportunities/IRB_Protocol_Preparation) — similar · Opportunities
- [Clinical Trial Artifact Parsing](/Opportunities/Clinical_Trial_Artifact_Parsing) — similar · Opportunities
- [Overlay Resolution Agent](/Opportunities/Overlay_Resolution_Agent) — similar · Opportunities
- [AI Chart Reconciliation for Hospitals](/Opportunities/AI_Chart_Reconciliation_for_Hospitals) — similar · Opportunities
- [AI Diagnostic Review](/Knowledge/Medicine_and_Dentistry/Opportunities/AI_Diagnostic_Review) — similar · Opportunities
- [Clinical Evidence Extraction for Community Hospitals](/Opportunities/Clinical_Evidence_Extraction_for_Community_Hospitals) — similar · Opportunities
- [Continuous HIPAA Remediation](/api/md.md.md/Opportunities/Continuous_HIPAA_Remediation) — similar · Opportunities
- [Hospital Clinical Record Parsing](/Opportunities/Hospital_Clinical_Record_Parsing) — similar · Opportunities
- [IRB Proxy](/Opportunities/IRB_Proxy) — similar · Opportunities
- [Automated Onboarding for Holistic Clinics](/Opportunities/Automated_Onboarding_for_Holistic_Clinics) — similar · Opportunities
- [Post-Discharge Triage](/Opportunities/Post-Discharge_Triage) — similar · Opportunities
- [EHR Normalization Engine.md](/api/md.md.md/Opportunities/EHR_Normalization_Engine.md) — similar · Opportunities
- [Traceability Matrix Generator](/Metrics/Requirements_Traceability_Index/Opportunities/Traceability_Matrix_Generator) — similar · Opportunities
- [Care Transition Navigator](/Opportunities/Care_Transition_Navigator) — similar · Opportunities
- [Clinical Criteria Mapper](/Opportunities/Clinical_Criteria_Mapper) — similar · Opportunities
- [Clinical Documentation Agent](/Opportunities/Clinical_Documentation_Agent) — similar · Opportunities