Interesting title in NEJM this week:
Evidence-Based Work Design — Bridging the Divide, by by Burden and Dirbye, 392:1044.
###
Chat GPT provides a simple "infographic," then a short summary, then answers a question about what might actually be envisioned.
####
AI CORNER
####
 |
| AI Infographic |
Evidence-Based Work Design — Bridging the Divide (NEJM, March 2025)
The article by Burden and Dyrbye argues that corporatization and financially driven decisions in health care are fueling burnout, moral injury, and inefficiencies among health care workers (HCWs). Much of the harm comes from modifiable but poorly designed work structures, which the authors label as “administrative harm.”
They call for a shift to evidence-based work design, using data-driven and iterative methods — similar to clinical evidence-based practice — to align job demands and resources for better outcomes. This includes using EHR metadata, AI tools, and predictive analytics to monitor and improve workload and team structures.
They advocate for interdisciplinary collaboration and innovative research (e.g., simulations, rapid trials) to guide best practices. The ultimate goal is to support health systems in achieving the quadruple aim: better health outcomes, improved patient experience, lower costs, and enhanced HCW well-being.
###
Can you give some more specific examples?
###
Great question — you're right that the article lays out a compelling framework but remains high-level on specifics. Based on the text and its implications, here are some specific changes that the authors likely envision as part of implementing evidence-based work design:
🔧 1. Redesigning Staffing Models
- Shifting from RVU-based productivity metrics to evidence-based staffing thresholds (e.g., optimal patient load per hospitalist or per nurse).
- More thoughtful deployment of team members (e.g., when to use MDs vs. nurse practitioners vs. both) based on empirical outcomes rather than cost or habit.
- Flexible team structures based on case complexity, not flat headcount.
🕒 2. Monitoring and Adjusting Workload in Real Time
- Use of EHR metadata (e.g., message volume, after-hours charting, click fatigue) to trigger automated flags when workload is approaching burnout thresholds.
- Early-warning systems for unsafe staffing levels, akin to patient deterioration alerts.
📊 3. Embedding Simulation or AI-Based Planning Tools
- Simulation models to predict the downstream effects of staffing cuts (e.g., how reducing a unit’s nurse coverage increases falls or LOS).
- AI-powered planning tools to optimize shift design, break timing, or patient handoff processes.
💰 4. Reframing Financial Incentives
- Moving from short-term cost-cutting to total cost of care modeling (e.g., recognizing that fewer doctors per shift may lead to higher readmission rates or burnout-driven turnover costs).
- Payer or CMS pilot programs that reimburse based on staff well-being metrics or organizational sustainability.
📉 5. Incorporating Work Design into Safety & Quality Dashboards
- Adding workforce strain indicators to existing quality dashboards (e.g., ratio of in-basket messages per hour; % of after-hours charting).
- Tracking clinician attrition and burnout as key organizational metrics, not just HR stats.
🧪 6. Redefining What Gets Studied in Health Services Research
- RCTs or robust observational studies comparing, say, 10 vs. 15 patient loads per hospitalist in terms of safety, satisfaction, and discharge efficiency.
- Studies assessing which team configurations (e.g., solo NP vs. NP + MD) yield better chronic disease outcomes or fewer ED visits.
🏥 7. Co-Designing Workflows with Frontline Staff
- Including clinicians in administrative design decisions through structured “co-design” processes.
- For example, redesigning the inbox triage system with those who use it daily, not just IT or finance staff.
🚨 8. Guardrails on AI Implementation
- Requiring evidence of time savings and safety outcomes before rolling out AI tools that increase clinician responsibilities (like documentation).
- Avoiding the trap of offsetting AI efficiency gains by increasing visit volume quotas.
These would represent a paradigm shift: from reactive, finance-first staffing to a data-informed, adaptive, and humane approach to organizing clinical work. It’s an attempt to bring the same level of rigor to workflow design that we expect in prescribing a medication or evaluating a clinical intervention.
Want a summary of this list in a visual format (like a table or infographic)?
##
WSJ: Doctors are as upset as patients. Here.
Why AI-driven productivity is hard to measure in general. Here. Same topic but, why healthcare AI productivity is hard to measure. Here.