Summary
- The 2027 clock is data-driven. For start-of-care dates on or after January 1, 2027, CMS intends to count OASIS data from patients of every payer toward a home health agency’s annual payment update.
- Capacity is already tight. 3% of home health providers reported turning down referrals in a 2026 Homecare Homebase report, and referral conversion has fallen 13%.
- AI use runs ahead of scale. 79% of home health leaders say they use AI, while 77% name integration and compatibility as barriers to scaling technology (Stratix research, 2026).
- Five gates decide whether pilots survive: identity, intake data, assessment data, integration and governance.
- One workflow, 90 days, seven KPIs gives a controlled pilot a defensible baseline.
Introduction
Home health agencies are heading into 2027 with a bigger data job than they’ve ever had. For start-of-care dates on or after January 1, 2027, CMS intends to fold OASIS assessments from patients of every payer into the annual payment update calculation. Today that calculation draws on Medicare and Medicaid patients only.
Patients feel the change first. An assessment with missing items delays the care plan built from it, and a referral stuck in intake waits while someone chases the packet. Referring physicians and discharge teams feel it next. In a 2026 Homecare Homebase report covered by Home Health Care News, 63.3% of home health providers said they had turned down referrals, and referral conversion had dropped 13%.
AI promises relief, and adoption is already broad. Stratix research reported in April found that 79% of home health leaders use AI, yet 77% cited integration and compatibility problems as barriers to scaling technology. Anyone who has watched a pilot stall will recognize it. The model works. The data underneath it doesn’t line up. This playbook covers the five data gates that decide whether post-acute care AI readiness holds up, the 2027 dates that set the clock, and a 90-day plan to start.
What AI readiness means in post-acute care
AI readiness is the point at which an organization’s data, systems and governance can support one specific AI workflow safely and measurably. In post-acute care that means a verified patient record, structured referral and assessment data, systems that exchange information, and named owners for oversight. Teams assess readiness one workflow at a time.
Scale matters here. Deloitte’s February 2026 survey of 100 US health care technology executives, split between health systems and health plans, found that 61% were already building agentic AI or had secured budgets, and 85% planned to raise investment. The report also observed that larger organizations tend to have stronger data foundations, governance and change capacity. Smaller post-acute providers feel that gap sooner.
Hospice leaders will see the same pattern in our earlier look at why hospice AI projects fail without data readiness. The five gates below extend that thinking across home health, home care and hospice.
The 2027 data clock
Five dates and proposals shape the data workload. Check the status of each one before you commit a budget line.
| When | What changes | Data consequence | Source |
| April 1, 2026 | OASIS-E2 takes effect | Update assessment forms, mappings and validation rules | CMS OASIS-E2 manual |
| Start of care on or after January 1, 2027 | CMS intends to include all-payer OASIS data in the annual payment update calculation; the 90% compliance threshold applies | Completeness and timeliness tracked for every payer, including commercial and managed care | CMS QTSO Q&As, July 2026 |
| Proposed July 1, 2026 | CY 2027 home health rule: calendar-year OASIS and HHCAHPS reporting windows and revised submission deadlines | Dashboards and close processes built on a calendar-year cycle | CMS fact sheet |
| Pending in Congress | H.R. 8883 would raise the reduction for missed quality reporting from 2 to 15 percentage points; reported out of Ways and Means as amended, not law as of August 31 | Raises the price of any reporting gap | OASIS Answers explainer |
| January 1, 2027 | Payers covered by CMS-0057-F must have Prior Authorization APIs in place; decisions due in 72 hours (expedited) or 7 days (standard) | Authorization status can arrive electronically; intake workflows should be ready to receive it | CMS-0057-F fact sheet |
Money adds pressure. The proposed CY 2027 rule estimates a 2.4% aggregate increase in Medicare payments to home health agencies, about $420 million, while continuing a 3.0% temporary reduction that recoups past overpayments. Efficiency has to come from the operation itself.
The five readiness gates
Each gate has a pass test, a failure signal and a starter KPI. A workflow is ready for an AI pilot when the gates it touches are green.
| Gate | Pass test | Failure signal | Starter KPI |
| 1. Identity | Governed match rules and a review queue for uncertain matches | One patient under two charts | Duplicate patient rate |
| 2. Intake data | Every referral becomes a structured, time-stamped record | Packets chased by phone and fax | Share of referrals complete at first review; referral-to-acceptance time |
| 3. Assessment data | Item-level validation at entry, an exception queue and an audit trail | Late fixes right before submission | OASIS completed on time and complete, by payer |
| 4. Integration | A documented interface and a failure alert for each handoff | Staff retyping data between systems | Manual touches per referral |
| 5. Governance | Approved-tool register, BAAs on file, logged human review | AI tools adopted without IT review | Share of AI tools with a BAA and access review |
Gate 1: One patient, one record
Referrals, visits, scheduling and billing each carry their own patient identifier. When two records describe one person, an assessment lands on the wrong chart and any model reads a partial history. A passing agency runs a governed master patient index with documented match rules and a human review queue for the matches the system can’t settle.
Clinical teams see the payoff in a single patient timeline across the EMR, monitoring and risk tools, as we described in our home health data visibility piece.
Gate 2: Referral data in structured form
Referral packets still arrive as faxes, portal exports and emails. AI can extract fields from them, but extraction only helps when staff can see what is missing before the clock runs. The initial assessment visit must happen within 48 hours of referral, within 48 hours of the patient’s return home, or on the ordered start-of-care date under 42 CFR 484.55. Time stamps for received, accepted and assigned turn that rule into data you can alert on.
Hospice teams already use automated eligibility screening to check every referral against the same criteria within minutes. Home health intake can borrow the pattern.
Gate 3: Assessment data you can defend
All-payer reporting extends the same completeness and timeliness expectations to commercial and managed care patients. CMS measures the annual payment update against a 90% compliance threshold under its Quality Assessments Only methodology. A passing agency validates items as the clinician enters them, routes exceptions to a named owner, and can trace every submitted item back to the visit note.
That traceability is the same discipline behind an audit-ready data lineage. Clinicians keep responsibility for every assessment. Software handles the completeness checks.
Gate 4: Systems that exchange data without a person in the middle
Most agencies run separate platforms for referrals, scheduling, clinical documentation, billing and payroll, and staff carry data between them. A passing agency documents the interface for each handoff, whether API, HL7, FHIR or secure file exchange, and gets an alert when a feed fails. Our guide to EMR and referral data integration walks through the connections that matter most.
At scale this looks like the work behind a national home care provider’s data platform spanning more than 40 source systems. Within 10 weeks, every acquired entity was integrated and reporting-ready with a single authoritative source for clinical, workforce, financial, and operational data.
Gate 5: Governance and HIPAA controls for AI
HHS’s Office for Civil Rights explains in its business associate FAQ that vendors who create, receive, maintain or transmit protected health information on behalf of a covered entity generally qualify as business associates, and that a written agreement with required elements comes first. The FAQ lists a third-party chatbot on a provider’s patient portal among its examples. Confirm your reading with counsel.
Six questions keep the governance gate honest:
- Which AI tools are approved, and who approved them?
- What patient information can each tool read or write?
- Is a business associate agreement signed for each vendor that touches PHI?
- How are access, retention and audit logs managed?
- Who reviews generated documentation before it enters the record?
- How are errors and security incidents found, reported and fixed?
Build these answers into the workflow design on day one. Our overview of data governance and CMS compliance frameworks shows how named owners and tested rules make that routine.
Where AI pays off after the gates turn green
Referral intake and eligibility checks
AI-assisted extraction pulls patient, payer and clinical fields from referral packets, flags missing documents and routes each referral to the right queue. The intake coordinator and clinician keep the acceptance decision. Track referral-to-acceptance time and manual touches per referral against the baseline.
Documentation completeness checks
Models can spot inconsistent or empty OASIS items before clinician sign-off and organize supporting records for review. Generated content needs validation, and licensed clinicians own the assessment. Track the share of assessments submitted complete and on time, by payer.
Scheduling and capacity
Homecare Homebase reported 2026 turnover of 25.46% for home health RNs and 34% for aides, and tied it to documentation burden and scheduling friction. Connecting scheduling, clinical and payroll data lets teams see capacity before a referral lands. Our home health staffing analytics guide covers the planning side.
Authorization and revenue cycle
Many revenue problems start at intake: missing orders, incomplete authorizations, patient details that differ between systems. As payers roll out the Prior Authorization APIs required by CMS-0057-F, agencies that already track authorization status as data can act on decisions faster. Teams weighing the technical options can start with our comparison of FHIR prior authorization APIs.
Eight questions for any AI vendor before a pilot
| Question | Why it matters | Green flag |
| Will you sign a BAA, and which subcontractors touch PHI? | HIPAA obligations follow the data | A named list and a signed template |
| Which of our systems do you read from and write to, and through what interface? | Integration drives cost and failure risk | Documented APIs, HL7 or FHIR support |
| Where does a person review output, and is each review logged? | Clinical accountability stays with licensed staff | Review steps built into the workflow |
| How do you handle OASIS-E2 items and all-payer data? | Rules changed in 2026 and change again in 2027 | A dated update process |
| Can we export our data and audit logs if we leave? | Protects against lock-in | Export in open formats |
| What accuracy and exception rates do you show on documents like ours? | Faxed referrals vary widely | Numbers from comparable customers |
| How do you monitor behavior after go-live? | Models and source systems drift | Named owner and alert thresholds |
| What does the pilot cost to run and maintain beyond the license? | Total cost decides scale | A written estimate for years one and two |
A 90-day plan for one workflow
| Days | Work | Deliverable | Exit test |
| 1 to 30: Assess | Map one workflow end to end, count manual handoffs, baseline the KPIs, grade the five gates for that workflow | Workflow map and KPI baseline | Every touched gate has a grade |
| 31 to 60: Design | Choose the workflow, define interfaces and human review points, name an owner, draft BAA and access rules, set stop criteria | Pilot charter | Owner and stop criteria signed off |
| 61 to 90: Pilot | Run a controlled pilot on a defined referral volume, compare to baseline weekly, gather staff feedback | Pilot readout | Decision to scale, adjust or stop |
Seven baseline KPIs: referral-to-acceptance time, share of referrals complete at first review, manual touches per referral, OASIS completed on time and complete by payer, authorization turnaround, clean claim rate, and days in accounts receivable. Measure outcomes. A count of deployed tools tells you little.
Final thoughts
AI will reach post-acute care’s daily work whether agencies are ready or not. The ones that benefit will have complete records, systems that exchange data, and staff who know who reviews what. The 2027 dates make that preparation measurable, and the 90-day plan gives it a starting line for home health agencies and their partners.
Pick one workflow this month. Baseline it. Grade the five gates against it. What that first pilot teaches you will shape the rest of the roadmap, for home health, home care and hospice teams alike.
