Product Summary

The nursing home MDS audit process relied heavily on manual review of patient medical records, making it difficult for nurses to consistently identify missed diagnoses, documentation gaps, and potential PDPM/NTA opportunities before final submission. Reviewing progress notes, ED records, discharge summaries, and other clinical documents also created a significant audit workload.

smartData is building a HIPAA-compliant, AI-powered clinical audit and validation platform that enables MDS nurses to upload and securely review patient records in one system. AI extracts diagnoses, maps them to ICD-10 codes, determines diagnosis status, cross-checks conditions against PDPM/NTA criteria, and analyzes medications for potential diagnosis disconnects. The platform also generates structured provider-query recommendations and 2–3-page audit reports, allowing nurses to review findings and take appropriate action. The solution includes secure document management, multi-tenant facility configuration, audit logging, role-based access controls, and HIPAA-oriented security measures.

Features

• AI-powered clinical document analysis for diagnosis extraction, ICD-10 mapping, and identification of active, historical, or suspected conditions.
• PDPM/NTA intelligence to identify qualifying comorbidities, calculate NTA points, and surface potential documentation opportunities.
• Medication-to-diagnosis cross-checking to identify clinical disconnects and potential missing or under-documented diagnoses.
• Secure multi-document patient record management with PDF/Word/scanned-PDF uploads, OCR, batch processing, and encrypted storage.
• Structured AI-generated provider queries and audit reports enabling MDS nurses to review findings and take appropriate documentation actions.

Technical Challenges

• Processing mixed clinical documents, including PDFs, Word files, and scanned PDFs, while preserving clinically relevant information for downstream AI analysis. : Built a reusable document ingestion and OCR pipeline with batch upload, patient/encounter grouping, file validation, and encrypted storage to prepare documents for structured AI extraction.
• Extracting diagnoses from unstructured clinical records and reliably identifying ICD-10 codes and diagnosis status such as active, historical, or suspected. : Implemented an AI interpretation workflow that extracts structured diagnoses, maps them to ICD-10, determines status, and stores AI outputs with versioning and timestamps for review.
• Connecting clinical findings to PDPM/NTA rules and identifying missing or under-documented diagnoses that may represent reimbursement opportunities. : Built clinical intelligence workflows to cross-reference diagnoses against PDPM/NTA criteria, calculate NTA points, and identify missing or under-documented conditions for MDS nurse review.
• Detecting clinical disconnects between medications and documented diagnoses and converting AI findings into actionable recommendations. : Implemented medication–diagnosis cross-checking and structured provider-query generation covering the observation, documentation gap, requested action, responsible specialty, and potential impact.