PBL4 - Add AI to Improve Administrative or Clinical Workflow

Case Background

X Rural Health Network (XRHN) serves a geographically dispersed population across rural areas of Missouri and Kansas, who has been facing increasing operational pressures:

  • Rising clinician burnout - Expanding documentation requirements
  • Growing administrative workload
  • Persistent communication gaps across care teams
  • Delayed follow-up on abnormal test results - Patient dissatisfaction with slow response times

Recognizing these challenges, XRHN leadership is exploring whether AI agents (autonomous or semi-autonomous digital assistants) could:

  • Reduce administrative burden
  • Improve patient triage
  • Strengthen remote monitoring workflows
  • Detect high-risk patients earlier
  • Enable proactive outreach
  • Support care teams with limited staff

Your team has been tasked with conducting the system’s first structured evaluation of where and how AI agents could improve rural healthcare workflows.

Problem Statement

Your team must select one of the following rural or remote-care workflows and evaluate where AI agents could augment, automate, or support tasks.

Candidate Workflow 1: Remote Patient Monitoring Esclation

  • Issue: RPM data (glucose, BP, weight, SpO₂) generates thousands of alerts per month that nurses cannot manually triage.
  • Potential AI roles: risk stratification agent, summarization agent, trend-detection agent

Candidate Workflow 2: Rural Telehealth Intake & Visit Preparation

  • Issue: Patients often join telehealth visits unprepared, with missing vitals or incomplete histories; clinicians spend significant time manually reviewing charts.
  • Potential AI roles: chart summarization, pre-visit questionnaires, technical check agents, patient-prep coaching.

Candidate Workflow 3: Behavioral Health Crisis Escalation in Remote Areas

  • Issue: Crisis support is limited; waits for tele-psychiatry can be long; patients may not understand when to seek help.
  • Potential AI roles: risk detection, message triage, safety-monitoring agent, supportive conversational AI agent.

Candidate Workflow 4: Care Coordination for Rural Chronic Disease Management

  • Issue: Communication across primary care, specialists, and pharmacies is fragmented; rural patients often lack frequent touchpoints.
  • Potential AI roles: care-plan summarizer, referral-prioritization agent, cross-team communication facilitator.

Candidate Workflow 5: Patient Education and Self-Management Support for Rural Communities

  • Issue: Many rural patients lack access to clear, understandable health education. Barriers include: Limited face-to-face time with clinicians, Low digital literacy, Chronic conditions (diabetes, COPD, heart failure) needing continuous self-management, Confusion around medication regimens, Difficulty navigating health information online
  • Potential AI roles:medication explanation agent, self-management coaching agent, appointment prep agent, condition progress check agent, care navigation agent

Learning Activities

  1. Identify workflow of interests. Select a workflow of interest described above and map its current-state workflow, and determine one specific function within the workflow (e.g., interpret, triage, document, escalate, reconcile, remind) where AI agents may provide meaningful benefit.

  2. Propose AI-agent solution. Design a conceptual AI agent that supports, augments, or automates the selected function. Your proposal must include:

    • AI agent purpose: What the agent does and why it improves the workflow.

    • Inputs and outputs: What data or signals the agent receives (notes, vitals, messages, RPM data, patient questions). What the agent produces (summary, triage score, draft documentation, patient-friendly education content, escalation recommendations).

    • Integration points: Where the AI agent can be integrated within the current enterprise information architecture (e.g., EHR, REDCap, messaging system, RPM dashboard, telehealth system).

    • Human–AI interaction rules: When the AI acts autonomously and when human review is required (e.g., flagged uncertainty).

    • Constraints and rural adaptation considerations : What are some other rural-specfic factors that will need to taken into consideration for effective implementation (e.g. , bandwidth, limited staff availability, digital literacy, offline mode).

  3. Future-state workflow diagram. Building on the current-state workflow digram, create a revised, AI-enabled workflow using a flowchart or swimlane format, to show where the AI agent enters the workflow, what tasks it performs, and how it interacts with humans and other systems.

Expected Deliverables

  • Short report (3-5 pages) summarizing proposed AI-agent solution

  • Architecture Diagram. There are many free or open-source tools for diagramming, such as Lucidchart (https://www.lucidchart.com/pages), Diagrams.net (previously draw.io, https://app.diagrams.net/), etc. You may use whichever tool you are most comfortable with or that best fits your needs.