What question did this study set out to answer?

Develop and implement a machine learning algorithm using electronic health record data to predict severe sepsis and septic shock, evaluating its impact on clinical workflows and patient outcomes.

August 6, 2019Critical Care MedicineOpen Access

A Machine Learning Algorithm to Predict Severe Sepsis and Septic Shock: Development, Implementation, and Impact on Clinical Practice*

View Full Paper

Ask AI

Bookmark

Authors

HGH.M. GianniniLung Institute JGJennifer C. GinestraUniversity of Colorado Anschutz Medical Campus CCCorey ChiversUniversity of Pennsylvania

Discussion

Member takes

Implication

Pre-post implementation study reveals modest clinical practice changes from a predictive sepsis alert in non-ICU patients, suggesting high specificity but limited overall outcome benefits.

Key Points

Develop and implement a machine learning algorithm using electronic health record data to predict severe sepsis and septic shock, evaluating its impact on clinical workflows and patient outcomes.
Derived and validated a random-forest classifier in non-ICU encounters across derivation (n = 162,212) and validation (n = 10,448) cohorts in a tertiary teaching hospital system.
Conducted a prospective pre-post clinical impact evaluation comparing a silent monitoring phase (n = 22,280) against an active clinical alert phase (n = 32,184).
The predictive algorithm exhibited 26% sensitivity, 98% specificity, a 29% positive predictive value, and a positive likelihood ratio of 13.
The clinical alert led to small, statistically significant increases in lactate testing and intravenous fluid administration, along with a reduction in time-to-ICU transfer.
No statistically significant improvements were observed in hospital mortality, discharge disposition, or overall rates of intensive care unit transfer.

Cite This Study

Giannini et al. (2019) studied this question.

synapsesocial.com/papers/6a0ee6cb950456576347dd1e https://doi.org/10.1097/ccm.0000000000003891

View Full Paper

Ask AI

Bookmark

Implication

Pre-post implementation study reveals modest clinical practice changes from a predictive sepsis alert in non-ICU patients, suggesting high specificity but limited overall outcome benefits.

A Machine Learning Algorithm to Predict Severe Sepsis and Septic Shock: Development, Implementation, and Impact on Clinical Practice*

Discussion

Member takes

Implication

Key Points

Cite This Study

A Machine Learning Algorithm to Predict Severe Sepsis and Septic Shock: Development, Implementation, and Impact on Clinical Practice*

Discussion

Member takes

Implication

Key Points

Cite This Study

Also Consider

Also Consider