Care Management

AI Chronic Care Management Agent

Worklists, device triage, outreach and billing evidence for CCM and RPM programs, with clinicians setting every rule.

We build AI chronic care management agents that prepare care coordinator worklists, triage RPM device readings against care team thresholds, draft outreach, propose care plan updates for approval and log the time and evidence CCM, RPM and APCM billing require.

Workflow scoping · Evaluation design · Human oversight · Integration planning

Updated September 2026

AI Chronic Care Management Agent product illustration

What an AI chronic care management agent does

An AI chronic care management agent handles the coordination work around CCM and RPM programs. It reads device data and EHR events, ranks who needs attention today, drafts outreach from approved content, proposes care plan updates for a clinician to approve, and records the minutes, consent and transmission days each CPT code requires. The care team sets the thresholds and makes every clinical decision.

What an AI chronic care management agent does for the care team

Care coordinators spend most of their month finding who needs attention, chasing data and logging time. The agent takes that work; the care team keeps every clinical decision.

Builds the daily worklist

Ranks enrolled patients by out-of-range readings, overdue outreach, missed appointments, recent discharges and care gaps, so coordinators start the day with the patients who need a call.

Triages device readings against care team thresholds

Blood pressure, glucose, weight and SpO2 readings arrive as FHIR Observations. Thresholds per patient come from the care plan; the agent never sets or changes them. Readings outside range go to a nurse queue with the trend attached.

RPM data pipeline to FHIR Observation

Drafts outreach and check-ins

Prepares check-in messages, call scripts and education content from approved material, sends routine check-ins on consented channels, and summarizes replies for the coordinator.

Keeps the care plan current

Proposes updates to goals, problems and interventions from new results and visit notes. A clinician approves each change before it is written to the CarePlan.

Tracks time and billing evidence

Logs coordinator and clinician time per patient per month, device transmission days and interactive communication, and flags where a patient is close to a billing threshold or missing documentation.

CCM, RPM and RTM code stacking

Closes care gaps

Identifies overdue screenings and quality measure gaps, such as blood pressure control and glycemic status for diabetes, and drafts the outreach or order request for review.

CMS care management programs and CPT codes the agent supports

Each program has its own documentation rules. The agent records the evidence each code requires so the monthly claim holds up to audit.

Chronic Care Management (CCM)

CPT 99490 and 99439 for clinical staff time, 99491 and 99437 for physician or QHP time, and complex CCM 99487 and 99489. Requires two or more chronic conditions expected to last at least 12 months, documented consent, a comprehensive care plan and 24/7 access.

Remote Patient Monitoring (RPM)

CPT 99453 for setup, 99454 for device supply with 16 or more days of data in 30 days, and 99457 and 99458 for treatment management time with interactive communication. The CY2026 Physician Fee Schedule added 99445 for 2 to 15 days of data and 99470 for the first 10 minutes of management.

RPM software servicesCMS CY2026 fee schedule summary

Principal Care Management and RTM

PCM (99424 to 99427) for a single high-risk condition, and Remote Therapeutic Monitoring (98975 to 98981) for therapy adherence and musculoskeletal or respiratory status.

Advanced Primary Care Management and BHI

APCM codes G0556 to G0558 for practices delivering advanced primary care, and Behavioral Health Integration (99484) for behavioral conditions managed in primary care.

RPM billing

Data and EHR integration for chronic care management

FHIR resources

Patient, Condition, CarePlan, Goal, CareTeam, Observation for vitals and device data, MedicationRequest for the active list, Task for coordinator work and Communication for outreach records.

Devices and aggregators

Cellular blood pressure cuffs, glucometers, scales and pulse oximeters through device vendor APIs or aggregators, normalized to Observations with LOINC codes such as 8480-6 and 8462-4 for blood pressure, 29463-7 for weight and 2339-0 for glucose.

RPM data into Epic and Oracle Health

EHRs

Epic, Oracle Health, athenahealth and eClinicalWorks through FHIR R4 and HL7 v2 ADT feeds, so discharges and ED visits trigger outreach within the window your program sets.

Epic integrationathenahealth integration

Clinical boundaries and alert design

The care team sets the rules

Thresholds, escalation paths and care plan goals come from the responsible clinicians. The agent applies them and proposes changes; it does not adjust medications or treatment.

Alerts people act on

Readings are grouped by patient and trend, repeated out-of-range values are combined into one task, and each alert carries the context a nurse needs, which keeps alert volume manageable.

RPM alert fatigue

Urgent readings escalate at once

Critical values such as a very high blood pressure with symptoms go straight to the on-call clinician with the patient instructed to seek care, following the escalation rules in the care plan.

PHI and audit

We sign BAAs, encrypt data in transit and at rest, and log every agent action and approval. Nirmitee is ISO 27001:2022 certified.

Delivery plan and timeline

  1. 1

    Discovery (2 to 3 weeks)

    Programs in scope, enrollment rules, thresholds, escalation owners, devices and EHR access.

  2. 2

    Build (5 to 7 weeks)

    Device ingestion, worklist, outreach, care plan proposals, time tracking and EHR integration in a sandbox.

  3. 3

    Shadow period (3 to 4 weeks)

    The agent builds worklists and drafts beside coordinators; its priorities are compared with theirs.

  4. 4

    Go-live by program

    Start with one program such as RPM for hypertension, then extend to CCM and other conditions.

What Drives the Cost of an AI Chronic Care Management Agent

Programs and conditions

Each program adds documentation rules; each condition adds thresholds, content and care plan templates.

Device sources

Direct device APIs, aggregators and patient-entered readings each need their own ingestion and validation.

EHR write-back

Writing care plans, time logs and outreach records back to the EHR costs more than read-only context.

Outreach channels

SMS, voice and app messaging add consent handling and language work.

Proof and further reading

Care and monitoring platforms

Our teams build remote monitoring, care management and patient-facing AI products, including an oncology patient copilot that answers only from the patient's record and cited guidelines. Client names are withheld.

Security and contracts

We sign BAAs, and Nirmitee is ISO 27001:2022 certified.

AI Chronic Care Management Agent FAQs

What is an AI chronic care management agent?

It is software that supports a care team running chronic care programs: it builds the daily worklist, triages device readings against thresholds the care team sets, drafts outreach, proposes care plan updates for clinician approval, and logs the time and evidence CCM and RPM billing require.

Does the agent make treatment decisions?

No. Thresholds, goals and escalation paths come from the responsible clinicians. The agent applies those rules, prepares the work and proposes changes, and a clinician approves anything that changes the care plan. It does not adjust medications or treatment.

Which CPT codes can the agent help document?

CCM (99490, 99439, 99491, 99437, 99487, 99489), PCM (99424 to 99427), RPM (99453, 99454, 99457, 99458, plus 99445 and 99470 from the CY2026 fee schedule), RTM (98975 to 98981), APCM (G0556 to G0558) and BHI (99484). It records time, consent, device days and interactive communication for each.

What are the CMS requirements for chronic care management?

The patient needs two or more chronic conditions expected to last at least 12 months that put them at significant risk, documented consent, a comprehensive electronic care plan shared with the patient, 24/7 access to the care team and continuity of care. CCM code 99490 requires 20 minutes of clinical staff time in the month.

Can AI agents be used for remote patient monitoring?

Yes. The agent ingests device readings as FHIR Observations, checks them against each patient's thresholds, groups repeated values into one task, escalates critical readings, and tracks the 16-day transmission rule for 99454 and the interactive communication required for 99457.

Which devices and EHRs does it integrate with?

Cellular blood pressure cuffs, glucometers, scales and pulse oximeters through vendor APIs or aggregators, and Epic, Oracle Health, athenahealth and eClinicalWorks through FHIR R4 and HL7 v2 ADT feeds. Discharges and ED visits trigger outreach automatically.

How do you prevent alert fatigue in RPM programs?

Readings are grouped by patient and trend, repeated out-of-range values become a single task, each alert carries the context a nurse needs, and thresholds are tuned per patient by the care team. We review alert volume and action rates weekly during the first months.

Is an AI care management agent HIPAA compliant?

It is when every component runs under a BAA, PHI is encrypted in transit and at rest, access is role-based, and every action and approval is logged. We sign BAAs, use BAA-covered model endpoints with no training on your data, and Nirmitee is ISO 27001:2022 certified.

How do you measure whether the agent helps the care team?

We track outreach completed on schedule, time from out-of-range reading to nurse contact, missed escalations, alert action rate, documentation completeness for billing, and coordinator time spent per patient, all against the baseline measured in discovery.

How long does it take to launch an AI chronic care management agent?

About 10 to 14 weeks to a first program: 2 to 3 weeks of discovery, 5 to 7 weeks of build, and 3 to 4 weeks of shadow operation beside coordinators. Adding programs or conditions afterwards is configuration and content rather than a new build.

Scope your AI chronic care management agent

Bring your programs, devices, EHR and escalation model. We will map the worklist rules, integrations, billing evidence and shadow plan for your first program.

Talk to an engineer