Hospital Quality Improvement · AI-Assisted Abstraction

Structured measure scoring from hospital chart documentation.

ClaritasQI applies a large language model to uploaded chart PDFs, evaluating each applicable quality measure against defined CMS and TJC criteria. Output is a structured compliance scorecard — not a narrative summary — with a verdict, rationale, and supporting chart citation for each measure element.

Methodology

How the abstraction process works.

ClaritasQI follows a structured, measure-driven evaluation sequence. For each applicable quality measure, the model is given the measure definition, inclusion and exclusion criteria, and the relevant chart content. It returns a verdict — PASS, FAIL, PARTIAL, or N/A — with a one-sentence rationale and the specific chart text that supports its finding. The process mirrors the logic of manual abstraction but applies it consistently across every uploaded record.

01

Chart Ingestion

Hospital chart PDFs are uploaded through a secure, HIPAA-compliant interface. The system extracts and parses structured and unstructured text — physician orders, nursing notes, laboratory results, vital signs, and discharge documentation.

02

Measure Evaluation

The model applies each relevant measure definition — drawn from CMS Core Measure specifications and TJC ORYX standards — against the extracted chart content. Each measure element is evaluated independently against its defined threshold and inclusion criteria.

03

Structured Output

Results are returned as a scorecard: a verdict for each measure element, a brief rationale citing the relevant chart finding, and an overall bundle or module compliance rate. Output is formatted for QI committee review and peer-review program documentation.

Traditional vs. AI-Assisted Abstraction

Where the approach differs from conventional QI review.

Manual chart abstraction is the current standard for most hospital QI programs. The following describes how ClaritasQI differs in practice — not as a replacement for clinical judgment, but as a structural alternative to the abstraction step itself.

Traditional Manual Abstraction
A trained abstractor reviews the chart and applies each measure criterion based on clinical knowledge and documentation review. Typically 30–60 minutes per chart.
Measure application is subject to inter-rater variability — different abstractors may interpret the same documentation differently against a given criterion.
Volume constraints typically require sampling. Comprehensive review of all applicable cases is rarely feasible for most QI programs.
Findings are recorded in spreadsheets or EHR-based forms. Aggregation and trend analysis require additional manual steps.
Chart citations supporting each abstraction decision are not always systematically recorded alongside the verdict.
ClaritasQI AI-Assisted Abstraction
The model applies each measure criterion to extracted chart text. Processing time per chart is minutes rather than hours, allowing broader case coverage within the same QI program.
Measure definitions and thresholds are applied from the same specification text on every evaluation, reducing criterion-level variability across cases and abstractors.
All uploaded cases are evaluated. Batch mode supports comprehensive review of a case population rather than a sampled subset, suitable for aggregate QAPI reporting.
Scorecards are generated in structured format. Aggregate compliance rates and trend data are computed automatically and exportable as PDF or Word for committee submission.
Each verdict includes the specific chart text that supports or contradicts the measure criterion, making the abstraction decision transparent and reviewable by a clinician.
Measure Library

Supported quality modules.

Each module maps to a defined regulatory standard. Phase 1 modules are available at launch; Phase 2 and 3 modules are in development and will be added in subsequent releases.

Module Regulatory standard Key elements evaluated Phase
SEP-1 Sepsis BundleCMS Core Measure SEP-13-hr bundle, 6-hr bundle, lactate, blood cultures, antibiotics, fluids, vasopressors, repeat lactatePhase 1
ACS / STEMICMS OP-2/OP-3, ACC/AHAAspirin on arrival and discharge, statin, door-to-balloon time, reperfusion strategy documentationPhase 1
StrokeCMS STK-1 through STK-10tPA eligibility and administration, DVT prophylaxis, antithrombotic therapy, rehabilitation assessment, smoking cessation counselingPhase 1
VTE PreventionCMS VTE-1 through VTE-6Prophylaxis ordered and administered, ICU VTE prophylaxis compliance, discharge instructionsPhase 1
Pneumonia / CAPCMS PN-6, IDSA/ATS guidelinesBlood cultures before antibiotic administration, appropriate antibiotic selection, time-to-first-dosePhase 2
Heart FailureCMS HF-1 through HF-3LVEF assessment, discharge instructions completeness, ACEI/ARB at dischargePhase 2
Perinatal SafetyTJC PC measures, ACOGElective delivery before 39 weeks, cesarean rate documentation, hemorrhage protocol adherencePhase 2
Surgical Care (SCIP)CMS SCIP / TJCProphylactic antibiotic timing, selection, and discontinuation; VTE prophylaxis; normothermia; glucose controlPhase 2
Pressure InjuryNDNQI, AHRQ NQF #0201Hospital-acquired Stage 3/4 PI, prevention protocol adherence, repositioning documentationPhase 2
Medication ReconciliationTJC NPSG.03.06.01Reconciliation on admission, transfer, and discharge; high-alert medication documentationPhase 3
Fall PreventionTJC NPSG.09.02.01Risk assessment performed, intervention implemented, post-fall review documentationPhase 3
Restraint UseCMS CoP 42 CFR §482.13Order compliance, monitoring frequency, 1-hour face-to-face, least restrictive alternative documentationPhase 3
Output Framing

Documentation designed for QI program use.

Hospital QI departments operate under peer review protections — Texas Health & Safety Code §161 and equivalent statutes in other states. The framing of QI documentation matters for how those protections apply.

ClaritasQI output uses compliance-oriented language throughout: measures are met or not met, variances are noted as improvement opportunities, and findings are framed as QI committee action items. The platform does not use language associated with litigation support tools — no deviations, no critical gaps, no breach findings.

This distinction is intentional. The output language is written to support classification of ClaritasQI records as peer review program documentation, consistent with the framing hospital counsel and QI committees expect.

Situation Litigation-framed language ClaritasQI framing
Missing documentationCritical gap identifiedImprovement opportunity noted
Protocol differenceDeviation from standardProtocol variance documented
Criterion not metBreach of standardMeasure not met
Reviewer actionFlagged for litigation reviewQI committee action item
Output document typeExpert reportQI program scorecard
Regulatory referenceChapter 74 / malpracticeCMS / TJC / QAPI

Platform access

ClaritasQI is a credentialed, access-gated platform. Sign in with your issued credentials, or contact us to discuss access for your department or QI program.

For access inquiries: info@claritasquality.com