Optimal Vasopressin Initiation in Septic Shock: The OVISS Reinforcement Learning Study.

Summary

Using reinforcement learning across multi-institutional EHRs, an initiation rule recommended earlier and more frequent vasopressin use at lower norepinephrine doses in septic shock. Adherence to the rule was associated with lower in-hospital mortality, findings that were consistent across external validation cohorts.

Key Findings

• Rule recommended vasopressin in 87% vs clinicians’ 31%, earlier after shock onset (median 4 vs 5 hours).
• Recommended initiation occurred at lower norepinephrine doses (median 0.20 vs 0.37 μg/kg/min).
• Concordance with the rule associated with lower hospital mortality (adjusted OR 0.81; 95% CI 0.73–0.91); consistent across external datasets.

Clinical Implications

Consider earlier vasopressin initiation when norepinephrine doses are modest, especially if decision support can operationalize the rule. Treat as hypothesis-generating pending prospective trials.

Limitations

• Observational nature with residual confounding; causal inference limited
• Generalizability may vary with differing sepsis care patterns and vasopressin availability

Future Directions

Prospective pragmatic trials to test early vasopressin initiation rules and integration of AI decision support into sepsis pathways.