Jefferson Health's AI Plan: 10 Million Clinician Hours Back to Care
Dr. Baligh Yehia says Jefferson Health will use AI to streamline scheduling, fix revenue cycle pain points, close care gaps, and return 10 million patient hours to clinicians over the next three years. That's a clear, measurable promise. For operations leaders, it sets the bar: deploy AI where it directly moves access, throughput, and margin.
Why this matters for operations
Access is constrained, costs are up, and teams are stretched. AI gives you leverage across repetitive, rules-heavy workflows without adding headcount. The target isn't novelty-it's capacity, speed, and consistency.
Four focus areas to prioritize
- Scheduling and access: Predict no-shows, auto-fill cancellations, suggest optimal slots, and route to the right site of care. Expect fewer idle slots and faster time-to-appointment.
- Revenue cycle: Automate eligibility checks, prior auth prep, coding assists, and denial prediction with recommended fixes. Aim for higher clean-claim rates and fewer reworks.
- Care gaps: Surface gaps from registries and notes, draft patient outreach, and queue tasks for care teams. Close more gaps without adding manual chart review.
- Clinician time: Ambient notes, in-basket triage, and smart order suggestions. The goal is fewer clicks and shorter documentation time, not workflow disruption.
What it takes to implement
- Integration first: Connect to your EHR, scheduling, RCM, and CRM so AI can read and write where work happens.
- Human-in-the-loop: Keep people reviewing high-impact decisions (coding, denials, outreach language) while the system handles the bulk.
- Clear ownership: Pair ops owners with clinical champions and data teams for each use case. One lead, one KPI, one rollout plan.
- Incremental rollouts: Pilot one service line or site, measure, then scale.
Metrics that signal it's working
- Scheduling: No-show rate ↓ 15-30%; slot utilization ↑ 3-7%; time-to-appointment ↓ 20-40%.
- Revenue cycle: Clean-claim rate ↑ 2-5 points; initial denial rate ↓ 10-25%; days in A/R ↓ 2-5 days; DNFB ↓ 10-20%.
- Care gaps: Outreach completion ↑ 30-60%; closed gaps per FTE ↑ 25-50%.
- Clinician time: Documentation time per note ↓ 30-60%; in-basket load ↓ 20-40%; clinician satisfaction ↑.
Risk, governance, and trust
- Privacy & security: Keep PHI in compliant environments; restrict data movement; audit access.
- Bias & safety: Test models across demographics; set escalation routes for edge cases.
- Quality controls: Version models, track drift, and monitor error rates with weekly review.
- Change management: Train teams, collect feedback in the first 30 days, and iterate fast.
90-day starter plan
- Weeks 1-2: Pick two use cases (e.g., no-show prediction, denial prediction). Define KPIs and baselines.
- Weeks 3-6: Integrate data, set guardrails, and run shadow mode. Validate accuracy and workflow fit.
- Weeks 7-10: Launch with one clinic and one service line. Daily huddles for feedback.
- Weeks 11-13: Compare to baseline, publish results, and plan scale-up.
Tooling hints
- Scheduling: Predictive models for demand and no-shows; optimization to fill gaps automatically.
- RCM: Rules engines + machine learning for eligibility, coding assists, and denial risk scoring.
- Care gaps: NLP over notes to find missing screenings and vaccinations; automated outreach drafts with staff review.
- Clinician assistance: Ambient documentation, smart order sets, and message triage with clear acceptance flows.
Why this approach works
It targets the highest-friction workflows with measurable outcomes. It keeps people in control while letting software handle repetitive tasks. And it builds capacity without waiting on hiring cycles.
Further resources
The headline is big: 10 million patient hours back to care. The path is straightforward-start where waste is visible, measure relentlessly, and scale what works.
Your membership also unlocks:
