Machine Learning–based Random Forest For Predicting Decreased Quality Of Life In Thyroid Cancer Patients After Thyroidectomy

Objective: Thyroidectomy is the main treatment method for thyroid cancer.The physical function,emotional function,cognitive function,social function and overall health status of patients with thyroid cancer decline after thyroidectomy affect the quality of life(Qo L)in thyroid cancer patients.Decreased Qo L in thyroid cancer patients after thyroidectomy is a common,but there is a lack of predictive methods for decreased Qo L.This study aimed to construct a machine learning-based random forest for predicting decreased Qo L in thyroid cancer patients three months after thyroidectomy.Methods: Two hundred and eighty-six thyroid cancer patients after thyroidectomy were enrolled in this prospective cross-sectional study from November 2018 to June 2019.The information such as age,gender,education level,marital status,area of residence,economic income,smoking,drinking history,diabetes,hypertension,pathological type,clinical stage,distant metastasis,surgical type,radiotherapy,nerve injury,symptom,and thyroid function were collected.The European Organization for Research and Treatment of Cancer quality of life questionnaire version 3(EORTC QLQ-C30)questionnaire was used to assess the Qo L three months after thyroidectomy,and decreased Qo L was defined as EORTC QLQ-C30 < 60 points.The patients were divided into decreased Qo L group and non-decreased group according to EORTC QLQ-C30,and the baseline characteristics were compared between two groups.The patients were randomly assigned to training and validation cohorts at a ratio of 7:3.The random forest model was constructed for predicting decreased Qo L in thyroid cancer patients after thyroidectomy based on training cohort,the predictive value of random forest model was acquired by receiver operating characteristic curve using training cohort,and the predictive value was verified using validation cohort.Results: The mean Qo L three months after thyroidectomy was65.93±9.00,with 21.33%(61/286)patients had decreased Qo L.A statistic difference between two groups on clinical stage,histological type,surgery type,nerve injury symptom,marital status,thyroid stimulating hormone level,and smoking was observed.Symptomatic deficit was mainly fatigue(54.5points),while functional deficits was mainly cognitive(58.98 points).The out-of-bag estimate of error rate was lowest when seven variables were used to insert the random forest model.The out-of-bag estimate of error rate was16.9%.The relative importance of features which can be computed by the Gini index as a metric,thereby presenting the top seven most important variables affecting Qo L,namely,clinical stage,marital status,histological type,age,nerve injury symptom,economic income,and surgery type.Two hundred eighty-six thyroid cancer patients were divided into a training cohort(n=201)and a test cohort(n=85).The area under the curve was 0.834(95% CI:0.764–0.934)in the training cohort,with sensitivity of 91.8%,specificity of51.2%,and accuracy of 83.1%,respectively.The area under the curve in test cohort was 0.897(95% CI: 0.831–0.963),which is consisted with that of the training cohort.Conclusions:1.Thyroid cancer patients had decreased Qo L three months after thyroidectomy.2.Symptomatic deficit was mainly fatigue,while functional deficits was mainly cognitive.3.The present study demonstrated that random forest model for predicting decreased Qo L in thyroid cancer patients three months after thyroidectomy displayed relatively high accuracy.4.Given the rapidly increasing number of thyroid cancer survivors and the need for assessment tool development,integration,and testing,this predictive model discussed timely improvements in understanding the Qo L of thyroid cancer survivors.