Machine learning−derived multivariable predictors of postcardiotomy cardiogenic shock in high-risk cardiac surgery patients
2024

Predicting Postcardiotomy Cardiogenic Shock with Machine Learning

Sample size: 11493 publication Evidence: high

Author Information

Author(s): Soltesz Edward G. MD, MPH, Parks Randi J. PhD, Jortberg Elise M. MS, Blackstone Eugene H. MD

Primary Institution: Cleveland Clinic

Hypothesis

Can a machine learning model accurately predict postcardiotomy cardiogenic shock in patients with poor left ventricular function undergoing cardiac surgery?

Conclusion

The machine learning model can reliably predict the risk of postcardiotomy cardiogenic shock based on preoperative variables.

Supporting Evidence

  • Operative mortality was significantly higher in patients with postcardiotomy cardiogenic shock compared to those with ideal recovery.
  • Patients with postcardiotomy cardiogenic shock had a higher incidence of prolonged ventilation and renal failure.
  • The model achieved an area under the curve of 0.74, indicating good predictive power.

Takeaway

Doctors can use a new computer model to figure out which heart surgery patients might have problems after their operation, helping them get the right care.

Methodology

The study analyzed data from 11,493 patients using machine learning algorithms to identify predictors of postcardiotomy cardiogenic shock.

Potential Biases

Potential bias due to reliance on available registry data and exclusion of certain patient demographics.

Limitations

The study relies on registry data, which may not capture all relevant clinical details, potentially underestimating the incidence of postcardiotomy cardiogenic shock.

Participant Demographics

Patients were adults with left ventricular ejection fraction ≤35% undergoing isolated on-pump surgery.

Statistical Information

P-Value

p<0.0001

Statistical Significance

p<0.05

Digital Object Identifier (DOI)

10.1016/j.xjon.2024.10.002

Want to read the original?

Access the complete publication on the publisher's website

View Original Publication