Survival Prediction Of Patients With Gastrointestinal Cancer In Home-based Hospice

Objective:This study aimed to investigate the survival time of patients with gastrointestinal cancers in home-based hospice,explore the influencing factors,and develop a survival prediction model for assessing their survival time,which could facilitate end-of-life planning and improve the quality of their final life span.Methods:We retrospectively studied the patients with gastrointestinal cancer from the Hospice Unit of Shantou University Medical College-affiliated First Hospital between 2008 and 2018.General baseline characteristics,disease-related characteristics,and related assessment scale scores were collected from the case records.The data were randomly split into a training set(75%)for developing a predictive model and a testing set(25%)for validation.We performed survival analysis and drew survival curve by Kaplan-Meier method.A stepwise regression method and LASSO regression method were used to select prognostic variables and establish Cox proportional hazard models respectively to analyze the influencing factors on survival time.The stepwise Cox regression model and LASSO Cox regression model were evaluated in terms of discrimination and calibration,and the better model was selected to predict the probability of survival at 30 and 60 days,and visualized in the form of nomogram.The above statistical analysis process was performed using R software.Results:A total of 1618 patients were included in this study,including 410 patients with liver cancer(25.3%),402 patients with esophageal cancer(24.8%),385 patients with colorectal cancer(23.8%),240 patients with gastric cancer(14.8%),98 patients with pancreatic cancer(6.1%),44 patients with biliary tract cancer(2.7%)and 39 patients with other digestive system cancers(2.4%).The median survival time for overall included patients was 35 days(IQR,17–66).Stratified random sampling was used to divide the data into two data groups: 1214patients(110 censored)as training dataset and 404 patients(33 censored)as testing dataset.The stepwise Cox regression model showed that 16 variables,including gender,age,awareness of the disease,tumor metastasis,history of radiotherapy,duration of pain before admission,effect of previous analgesic treatment,history of hypertension,history of diabetes,weight loss,nausea,abdominal distention,tachypnea,edema,KPS score and QOL score,were the influencing factors on survival time.The LASSO Cox regression model showed that 5 variables,including duration of pain before admission,abdominal distention,edema,KPS score and QOL score,were the main influencing factors on survival time.The stepwise Cox regression model showed an AUC of 0.725(95% CI: 0.675-0.775)at 30-day survival prediction and 0.740(95% CI:0.687-0.793)at 60-day,while the LASSO Cox regression model showed an AUC of 0.724(95%CI: 0.673-0.774)at 30-day prediction and 0.725(95% CI: 0.672-0.778)at 60-day,suggesting that both models have good discriminatory ability.However,the calibration evaluation showed that the LASSO Cox regression model was slightly better calibrated than the stepwise Cox regression model.Ultimately,the more parsimonious and efficient LASSO Cox regression model was selected as the survival prediction model for patients with gastrointestinal cancers in home hospice care and visualized in the form of nomogram.Conclusion:The survival time of patients with gastrointestinal cancers in home hospice care is generally short.Under the comprehensive consideration for results of stepwise regression and LASSO regression analyses,KPS score,abdominal distention,edema,duration of pain before admission and QOL score are the main predictors of survival.This is the first time to build a gastrointestinal cancer-specific predictive model,visualized in the form of nomogram,which may be a useful clinical tool providing more accurate survival assessment information.