External validation of larynx cancer outcome prediction model in an Australian cohort and impact of data curation on model performance.

Congress:

2021 ASM

Poster Number:

R-0429

Type:

Scientific Exhibit

Keywords:

Artificial Intelligence, Head and neck, Oncology, Experimental, Radiation therapy / Oncology, Outcomes

Authors:

F. F. Aly, M. Field, P. Sundaresan, S. Vinod, L. Holloway; Sydney, NSW/AU

DOI:

10.26044/ranzcr2021/R-0429

DOI-Link:

https://dx.doi.org/10.26044/ranzcr2021/R-0429

Conclusion

External validation of a larynx outcome prediction model demonstrated similar model discrimination performance to the original model for OS but inferior performance for LR in our validation patient cohort.
The LR model (both cohorts) and OS model (uncurated cohort) demonstrated poor calibration in the validation patient cohort, with calibration curves demonstrating differences between predicted survival compared to actual survival outcomes.
The accuracy of the model performance on our validation patient cohort may be affected by several factors:

Small sample size of the validation patient cohort
Differences in tumour characteristics to the original model development cohort
A greater amount of missing Hb data, which required imputation for model validation.

Data curation resulted in similar model performance, suggesting outcome prediction modelling on routinely collected data may be adequate in the clinical setting.
External validation is important to determine whether a published model is accurate in a local population. Model performance may be improved through updating model parameters or recalibrating the model to better suit the local population.

Fig 1: Table 1. Patient characteristics

Fig 2: Table 2. Model discrimination on uncurated and curated dataset

Fig 3: Figure 1. Calibration curves for 2 year Overall survival (OS) in a) uncurated...

Fig 4: Figure 2. Kaplan Meier survival curves for a) Overall survival (OS) and b)...