diff --git a/domains/health/multi-agent-clinical-ai-adoption-driven-by-efficiency-not-safety-creating-accidental-harm-reduction.md b/domains/health/multi-agent-clinical-ai-adoption-driven-by-efficiency-not-safety-creating-accidental-harm-reduction.md
new file mode 100644
index 00000000..6e508a8c
--- /dev/null
+++ b/domains/health/multi-agent-clinical-ai-adoption-driven-by-efficiency-not-safety-creating-accidental-harm-reduction.md
@@ -0,0 +1,17 @@
+---
+type: claim
+domain: health
+description: The commercial and research cases for multi-agent architecture are converging accidentally through different evidence pathways
+confidence: experimental
+source: Comparison of Mount Sinai npj Health Systems (March 2026) framing vs NOHARM arxiv 2512.01241 (January 2026) framing
+created: 2026-04-04
+title: "Multi-agent clinical AI is being adopted for efficiency reasons not safety reasons, creating a situation where NOHARM's 8% harm reduction may be implemented accidentally via cost-reduction adoption"
+agent: vida
+scope: functional
+sourcer: Comparative analysis
+related_claims: ["human-in-the-loop-clinical-ai-degrades-to-worse-than-AI-alone", "healthcare-AI-regulation-needs-blank-sheet-redesign"]
+---
+
+# Multi-agent clinical AI is being adopted for efficiency reasons not safety reasons, creating a situation where NOHARM's 8% harm reduction may be implemented accidentally via cost-reduction adoption
+
+The Mount Sinai paper frames multi-agent clinical AI as an EFFICIENCY AND SCALABILITY architecture (65x compute reduction), while NOHARM's January 2026 study showed the same architectural approach reduces clinical harm by 8% compared to solo models. The Mount Sinai paper does not cite NOHARM's harm reduction finding as a companion benefit, despite both papers recommending identical architectural solutions. This framing gap reveals how research evidence translates to market adoption: the commercial market is arriving at the right architecture for the wrong reason. The 65x cost reduction drives adoption faster than safety arguments would, but the 8% harm reduction documented by NOHARM comes along for free. This is paradoxically good for safety—if multi-agent is adopted for cost reasons, the safety benefits are implemented accidentally. The gap between research framing (multi-agent = safety) and commercial framing (multi-agent = efficiency) represents a new pattern in how clinical AI safety evidence fails to translate into market adoption arguments, even when the underlying architectural recommendation is identical.
diff --git a/domains/health/multi-agent-clinical-ai-reduces-computational-cost-65x-while-maintaining-performance-under-workload.md b/domains/health/multi-agent-clinical-ai-reduces-computational-cost-65x-while-maintaining-performance-under-workload.md
new file mode 100644
index 00000000..ca3a14d6
--- /dev/null
+++ b/domains/health/multi-agent-clinical-ai-reduces-computational-cost-65x-while-maintaining-performance-under-workload.md
@@ -0,0 +1,17 @@
+---
+type: claim
+domain: health
+description: Specialization among agents creates efficiency where each agent optimized for its task outperforms one generalist agent attempting all tasks
+confidence: proven
+source: Girish N. Nadkarni et al., npj Health Systems, March 2026
+created: 2026-04-04
+title: Multi-agent clinical AI architecture reduces computational demands 65x compared to single-agent while maintaining performance under heavy workload
+agent: vida
+scope: structural
+sourcer: Girish N. Nadkarni, Mount Sinai
+related_claims: ["human-in-the-loop-clinical-ai-degrades-to-worse-than-AI-alone"]
+---
+
+# Multi-agent clinical AI architecture reduces computational demands 65x compared to single-agent while maintaining performance under heavy workload
+
+Mount Sinai's peer-reviewed study distributed healthcare AI tasks (patient information retrieval, clinical data extraction, medication dose checking) among specialized agents versus a single all-purpose agent. The multi-agent architecture reduced computational demands by up to 65x while maintaining or improving diagnostic accuracy. Critically, multi-agent systems sustained quality as task volume increased, while single-agent performance degraded under heavy workload. The architectural principle mirrors clinical care team specialization: each agent optimized for its specific task performs better than one generalist attempting everything. This is the first peer-reviewed demonstration of multi-agent clinical AI entering healthcare deployment at scale. The efficiency gain is large enough to drive commercial adoption independent of safety considerations.
diff --git a/entities/health/hasso-plattner-institute-digital-health-mount-sinai.md b/entities/health/hasso-plattner-institute-digital-health-mount-sinai.md
new file mode 100644
index 00000000..e287b9fc
--- /dev/null
+++ b/entities/health/hasso-plattner-institute-digital-health-mount-sinai.md
@@ -0,0 +1,21 @@
+# Hasso Plattner Institute for Digital Health at Mount Sinai
+
+**Type:** Research program  
+**Parent:** Icahn School of Medicine at Mount Sinai  
+**Director:** Girish N. Nadkarni, MD, MPH  
+**Focus:** Clinical AI systems, digital health infrastructure, healthcare workflow optimization
+
+## Overview
+The Hasso Plattner Institute for Digital Health at Mount Sinai is a leading clinical AI research program producing peer-reviewed studies on multi-agent AI architectures, misinformation detection, and healthcare workflow automation. The institute has strong health system connections and influences CIO-level technology architecture decisions.
+
+## Timeline
+- **2026-02** — Klang et al. published Lancet Digital Health study on LLM misinformation detection
+- **2026-03-09** — Published first peer-reviewed demonstration of multi-agent clinical AI showing 65x computational efficiency gain (npj Health Systems)
+
+## Research Output
+- Multi-agent AI architecture for clinical workflows
+- AI misinformation detection in healthcare
+- Clinical data extraction and medication safety systems
+
+## Significance
+First research group to publish peer-reviewed evidence of multi-agent clinical AI entering healthcare deployment. Research likely to be cited in health system technology architecture decisions through 2026-2027.
\ No newline at end of file