CommuniCare MDS Optimization Case Study

Overview

AI-powered MDS optimization at national scale

CommuniCare Health Services needed to optimize MDS assessments across their skilled nursing facilities. Manual reviews took 3–4 hours per resident, and uncaptured diagnoses were directly impacting PDPM reimbursement and quality scores.

Digital Scientists built an AI-powered coding engine that ingests data from six clinical sources to surface uncaptured ICD-10 codes with confidence scoring and clinical evidence linking—reducing review time from hours to roughly 60 seconds per patient.

The platform has expanded to include ARD optimization (expected ~$20M additional annual impact), event scanning across 200+ clinical triggers, multi-facility operator dashboards, PCC real-time integration, and regulatory compliance automation.

Client

CommuniCare Health Services

Industry

Post-Acute Care / Skilled Nursing

Services

AI/ML, Data Engineering, Product Design, UX Research

Platform

PointClickCare

Engagement

2023 – Present

Method

Discover Experiment Engineer

Capabilities

AI / ML Data UX

The Challenge

Millions in revenue leakage from incomplete MDS assessments.

CommuniCare’s MDS coordinators were drowning. Each assessment took 3–4 hours, and with hundreds of residents across multiple states, diagnoses were going uncaptured, directly impacting PDPM reimbursement.

The problem wasn’t competence—it was capacity. Coordinators didn’t have time to review every clinical note, lab result, and medication record for every resident. Diagnoses were going uncaptured. Revenue was walking out the door.

Before DS

X

3–4 hours per MDS assessment

Manual review of clinical documentation, medications, lab results

X

Uncaptured diagnoses at scale

Not enough time to catch everything buried in the chart

X

Revenue leakage at scale

Hundreds of residents, millions in missed reimbursement

X

Quality scores suffering

Incomplete documentation affecting Five-Star ratings

User Research

We mapped the entire MDS ecosystem before writing code.

Before building anything, we conducted deep persona research to understand the MDS nurse’s world—their workflows, tools, challenges, and the network of stakeholders they coordinate with daily.

The MDS Coordinator

Also known as: MDS Nurse, RAI Coordinator, MDS Assessment Nurse, Clinical Reimbursement Specialist, PDPM Coordinator

Distinguished by dedication to quality resident care, trained in MDS assessments, synthesizes large quantities of data from resident interviews, health records, and clinical documentation.

Jobs to Be Done

• Ensure accurate, comprehensive resident assessments reflecting clinical acuity
• Identify and respond to significant changes in resident status promptly
• Adhere to RAI guidelines, ARD timelines, and state MDS regulations

Tools & Devices

PointClickCare (PCC), printed health records, laptop w/dual monitors

Key Challenges Uncovered

Documentation & data accuracy

Inaccurate coding, data entry errors, and incomplete documentation due to inconsistent, unavailable data or overwhelming information volume. Reliance on printed records makes verification time-consuming and error-prone.

Selecting the correct ARD date

Requires extensive RAI manual knowledge, years of experience, and excellent judgment—balancing precise care planning with optimal financial reimbursement amidst potential audits.

Coordination across disciplines

MDS nurses coordinate with nursing staff, physicians, therapists, dieticians, social workers, and admissions nurses—communication gaps and scheduling conflicts lead to assessment discrepancies.

Stakeholders Mapped

8

Roles across the MDS ecosystem

Shadowing

40+

Hours with MDS coordinators

User Interviews

17+

Facility & workflow validation

Feedback Loops

Weekly

Hotjar + champion nurse sessions

The Solution

AI-powered MDS assistant built with—not for—MDS coordinators.

We didn’t build an algorithm and hand it over. We embedded with CommuniCare’s MDS team, shadowed their workflows, and co-designed a solution that actually fit how they work.

MDS Coordinator reviewing AI recommendations

AI Diagnosis Recommendations

Our AI reviews clinical documentation, medications, and lab results to surface diagnoses that should be captured on the MDS. Every recommendation linked to supporting documentation—audit-ready from day one.

Pre-Submission Validation

Before MDS submission, AI flags potential errors, uncaptured diagnoses, and inconsistencies. Catch problems before they become denials or audit findings.

Click to zoom

Quality Analytics

Real-time visibility into quality measure performance across facilities. Identify improvement opportunities, track progress, and prepare for surveys.

Facility View

Actionable insights at every level

Each facility gets a dedicated overview showing quality score trends, CMI averages, total residents by status, and actionable alerts sorted by urgency. Operators can drill from regional overview to individual resident in seconds.

Facility overview with quality score trends and actionable insights

Click to zoom

Resident Management

Every resident, every deadline

The resident list gives MDS nurses at-a-glance status for every resident—overdue, due within 14 days, in progress, or discharged. Quick filters surface the most urgent work, and one click opens the full resident summary.

Resident list with MDS deadline status and quick filters

Click to zoom

Event Scanning & ARD Optimization

Detecting opportunities before they expire.

The system continuously scans clinical data to detect events—falls, new diagnoses, weight changes, isolation, and 200+ other triggers—that create MDS optimization opportunities. Each event is tied to a look-back period, and the ARD optimizer quantifies the financial impact of choosing different Assessment Reference Dates.

MDS event detection showing New Diagnosis and Isolation events with evidence trails

Click to zoom

Event Detection with Evidence

Each detected event surfaces the source data—physician orders, progress notes, lab results—so nurses can verify the finding and decide on action. Events that affect PDPM or quality measures are prioritized by urgency and look-back window.

ARD Optimizer comparing assessment reference dates with CMI impact and Five-Star implications

Click to zoom

ARD Optimization

Nurses can evaluate multiple ARD dates side-by-side, seeing the impact on Nursing CMI, case mix group, and Five-Star quality measures. In complex cases, higher payment can compete with quality ratings—the system helps nurses make the best decision for both the resident and the facility.

Quality Measures

Five-Star performance at your fingertips

The platform tracks 10+ quality measures in real time, showing numerators, denominators, current points, and Five-Star ratings for both short-stay and long-stay residents. Facilities can see exactly where they stand and where opportunities exist.

Quality score trends over time show the impact of optimization efforts, with CMI averages tracked monthly. This data drives both clinical decisions and financial planning.

Facility quality score trends and CMI average over time

Click to zoom

Our Approach

How the complete value chain delivered $12M+ in results.

This wasn’t just an AI project. It was a complete transformation—from messy data to measurable outcomes. Here’s how each step contributed to success.

Step 1

Data Foundation

Integrated clinical data from PointClickCare across multiple facilities. Cleaned and structured medication records, lab results, and clinical notes into AI-ready format.

Step 2

Responsible AI Build

Built AI using peer-reviewed CMS scoring models. Every recommendation explainable and linked to source documentation. 90%+ diagnosis accuracy validated.

Step 3

EHR Integration

Seamless PointClickCare integration. AI recommendations surface within existing workflow—no separate login, no duplicate data entry. Real-time validation before CMS submission.

Step 4

Co-Design

40+ hours shadowing MDS coordinators before writing code. Persona research across 8 stakeholder roles. Weekly feedback sessions throughout development.

Step 5

Implementation

Phased rollout starting with pilot facilities. June 2025 Verify release. On-site go-live support. Rapid issue resolution. No “deploy and disappear.”

Step 6

Change Management

Training programs for MDS coordinators. Champion nurse identification. Hotjar surveys and user interviews. Adoption from day one—not months later.

Step 7

Regulatory Automation

AI agents monitor CMS regulatory sites for changes. Versioning framework ensures MDS logic always uses the latest published rates, guidelines, and RAI manual updates.

Step 8

Ongoing Partnership

Regular executive reviews. Continuous optimization based on CMS updates and facility feedback.

Financial Impact

ARD optimization: the next ~$20M opportunity

ARD date optimization analysis identified substantial additional reimbursement opportunities across facilities by surfacing cases where alternative Assessment Reference Dates would yield higher PDPM payments.

The analysis dashboard captures PDPM payment differences, nursing CMI comparisons, and quality measure impacts—giving operators clear visibility into the value of optimization.

MDS 6-Month ARD Optimization Analysis Dashboard showing $24M extrapolated total and payment differences

Click to zoom

The Results

Validated against actual performance data.

Results validated against actual PDPM reimbursement and quality performance data.

Financial Impact

PDPM Revenue Recovered $10M+

Annualized. Previously uncaptured diagnoses now surfaced through AI-powered recommendations.

Quality Incentives $2M

Annual quality incentive payments earned through improved quality measure performance.

ARD Optimization (Expected) ~$20M

Additional annual earnings from ARD date optimization across facilities.

Average CMI Improvement +$120/day

Per patient benefit from more accurate case mix index classification.

Operational Impact

Time Per Assessment 40-60% ↓

MDS coordinators spend less time hunting through charts, more time on clinical judgment.

Five-Star QM Score 60/100

Quality measure performance improved through better documentation and coding accuracy.

Diagnosis Accuracy 90%+

AI recommendations validated against clinical documentation with high accuracy.

Event Scanning 200+

Clinical triggers monitored continuously, with expanded data sources and benchmarked accuracy.

Why It Worked

40+ hours shadowing before writing code.

Most healthcare AI projects fail because they’re built in conference rooms, not clinical settings. We took a different approach.

We embedded with MDS coordinators

Before writing a single line of code, our team spent 40+ hours shadowing MDS coordinators. Watching their actual workflow. Understanding their pain points. Learning their language. We mapped 8 stakeholder roles across the MDS ecosystem.

We built for audit defense

Every AI recommendation is linked to supporting clinical documentation. When auditors ask “why did you capture this diagnosis?”—there’s an answer ready. Defensible from day one.

We drove adoption, not just deployment

Change management isn’t an afterthought. We trained coordinators, identified champion nurses, launched Hotjar surveys and user interviews, and addressed resistance head-on. The solution works because people actually use it.

Related Work