Inpatient Deterioration & Risk Monitoring
AI that watches an admitted patient continuously and predicts an adverse event early enough to act on: sepsis and clinical decompensation, respiratory and haemodynamic instability, falls, and obstetric deterioration in labour. The input is a stream rather than a study, which separates this category from imaging triage, where a single scan is read once. The setting is acute inpatient, which separates it from remote patient monitoring in the home and from fall detection in residential senior living. Two failure modes define the category and neither appears elsewhere in this index. The first is alert fatigue: this is the only category where the failure is the output being ignored rather than being wrong, because a model can be accurate and still clinically useless if it fires often enough that staff stop responding to it. The second is lead time, which trades against precision. An alert thirty minutes before an arrest is close to worthless and one six hours earlier is actionable, so a sensitivity figure means very little without the median lead time and the alert burden per patient day beside it. Ask for all three together, and ask separately whether any external party has validated the model on their own data, because the most consequential negative result in clinical AI came from exactly that kind of independent revalidation of a widely deployed proprietary sepsis model.
| Vendor | Category | AI Centrality | Website |
|---|---|---|---|
|
E
Epic Sepsis Model
The Epic Sepsis Model is a proprietary sepsis prediction model built into the Epic electronic health record, scoring hospitalised patients continuously and firing an advisory to clinicians above a threshold Epic recommends setting at 6. It has been deployed at hundreds of United States hospitals, which makes it almost certainly the most widely used clinical prediction model in existence. It is also the most externally validated, and that history is the reason this record matters. Epic reported hospital level performance for the original model at an area under the curve of 0.76 to 0.83. In 2021 Wong and colleagues at Michigan Medicine published an external validation in JAMA Internal Medicine covering 38,455 hospitalisations, and found an area under the curve of 0.63, sensitivity of 33 percent, specificity of 83 percent and positive predictive value of 12 percent. An accompanying editorial was titled The Epic Sepsis Model Falls Short. A separate study of 145,885 encounters across two county emergency departments later reported sensitivity of 14.7 percent and, more strikingly, a median alert lead time of zero minutes, meaning half of all alerts arrived at or after the point clinicians had already recognised sepsis. Epic subsequently rebuilt the model. In February 2026 the same lead author published a multicentre prospective validation of Epic Sepsis Model version 2 in JAMA Network Open, covering 227,091 inpatient encounters across four major United States health systems, reporting an area under the curve between 0.82 and 0.92. Discrimination improved substantially. The paper also found high institutional variability, low positive predictive value and high alert burden, and recommended that institutions run local validation, integrate workflows to handle false positives, and adopt alert silencing strategies. This record is scoped to the sepsis model and is not an assessment of Epic Systems or its other AI features.
|
Inpatient Deterioration & Risk Monitoring | C | epic.com |
|
B
Bayesian Health
Clinical risk platform running real time machine learning models inside hospital EHRs to detect deteriorating patients, with early sepsis warning as the flagship use case and additional configured uses spanning clinical deterioration, pressure injuries, palliative care, and transitions of care. A Johns Hopkins spinout founded on roughly a decade of academic research, it is one of very few clinical AI vendors whose deployed system has been evaluated in large prospective multi site studies published in peer reviewed journals.
|
Inpatient Deterioration & Risk Monitoring | A | bayesianhealth.com |