Development and Assessment of Artificial Intelligence-Driven Models to Predict Hernia Recurrence, Surgical Complications, and 30-Day Readmission Following Abdominal Wall Reconstruction

Abbas M Hassan, MD; Malke Asaad, MD; Anaeze C Offodile II, MD, MPH, FACS; Charles E Butler, MD, FACSDepartment of Plastic and Reconstructive Surgery, The University of Texas MD Anderson

Cancer Center, Houston, TX.

AWR is common but not without complications



>400,000 operations

performed annually in the US alone.



Costs >$3 billion



Complication rates remain far from acceptable:

Machine learning may improve AWR outcomes



ML: branch of artificial intelligence



ML learns relationships and patterns between complex variables.



ML can identifies patients at high risk of developing complications

Hernia Recurrence

42%

Surgical complications

44%

30-day readmission

13%

Complications following AWR are devastating



*1% reduction in HR alone >$30 million would be saved annually

↓ Quality

of life

↑ Morbidity

↑ Morbidity

↑ Healthcare

Costs*

Model development

80%

N=580

20%

N=145

•Demographics

•Comorbidities

•Preoperative assessment

•Surgical characteristics

New

data

Processed data

Predictive model

Algorithm

Development

ML Predictive Models

Model performance

Figure 1. Receiver operating characteristic curves for prediction of (A) hernia recurrence (HR), (B) surgical site occurrence (SSO), and (C) 30-day readmission. Generalized linear model achieved the highest area under the curve (AUC) for predicting HR (0.71). Support vector machine model achieved the highest AUC for predicting SSOs (0.75). Multiple adaptive regression splines model achieved the highest AUC for predicting 30-day readmission (0.74).

A

B

C

Figure 2. Decision curve analysis. The decision curves show the clinical usefulness of the machine learning (ML) models in predicting (A) hernia recurrence (HR), (B) surgical site occurrence (SSO), and (C) 30-day readmission in the validation dataset. Regardless of the threshold used, the net benefit of the ML models was greater than that of the reference models.

A

B

C

Predictive factors for complications

Figure 3. Feature importance analysis showing the impact of various factors in predicting (A) surgical site occurrence, (B) hernia recurrence, and (C) 30-day readmission. Higher values on the x-axis indicate higher variable importance.

Figure 4. Accumulated local effect (ALE) plot of predictive/protective variables for surgical site occurrence (SSO). ALE plots visualizethe effect of each of the variables on SSO. Higher values on the y-axis indicate a higher risk of SSO.

Clinical impact

Individualized risk

assessment

Preoperative

optimization

Surgical

planning

Shared-decision

making

Conclusion

ML accurately predicts the occurrence of complications

ML identified non-linear risk predictors of complications

ML should be incorporated in perioperative AWR care

References

Collaborate with us