THE ROLE OF MACHINE LEARNING IN PREDICTING OUTCOMES OF GASTROINTESTINAL CANCER

Abstract

Background

Machine learning (ML) has become one of the most prevalent tools in healthcare since it presents new opportunities for both predicting and ameliorating the patient’s condition. Here, this research aims to identify the efficacy of the use of ML regarding algorithm accuracy and, data aggregation and synthesis in determining outcomes of GI cancer, a prospective view from healthcare workers.

Objective

To understand the performance of different ML algorithms to predict the outcomes of patients with GI cancer, to identify the critical factors that affect its efficacy, and to analyze the moderating factors in terms of data integration and data analysis skills.

Methods

This study used a quantitative, descriptive-correlational design. The data were collected by administering self-developed, structured questionnaires to healthcare professionals (355), secondary datasets from hospitals, and existing literature. Descriptive analysis used self-developed questionnaires from doctors and patients and statistical methods include normality tests, Cronbach’s Alpha reliability tests, regression analysis, and mediation analysis. Again, use of graphics including bar charts and scatter plots were utilized in the course of data analysis for easy interpretation.

Results

The study showed that the ML algorithms predict GI cancer outcomes by 62 % (R2 = 0.62, p < 0.001) with data integration at the center of the outcomes. The internal consistency of the questionnaires was also quite high (Cronbach’s α= 0.87 (α= 0.87α= 0.87). Most participants had a positive attitude toward the impact of ML. Further, the Shapiro-Wilk test was used to test for normality of data and yielded a result W=0.95,p>0.05W=0.95, p>0.05. Descriptive analysis pointed to most respondents as being representatives of the working population, with 90% of the target population being between 25–44 years – the ML tool users.

Conclusion

This work shows that machine learning is beneficial in enhancing the forecasts of GI cancer and facilitating patients’ precipitating prescriptions. The study also strongly supports the notion of data integration and clinician buy-in for the successful use of the approaches detailed in the paper. Recommendations for future research include designing ways to mitigate data privacy concerns and elucidating how to develop algorithms that are easier to interpret as well as training physicians to work effectively with the technology to optimize the utility of ML in oncology.