The application discloses a
pulmonary tuberculosis morbidity risk
prediction system and application, and is suitable for
pulmonary tuberculosis morbidity
risk assessment of the elderly.
Baseline data of the
elderly population meeting the standard are collected, pretreated through cleaning,
resampling, coding
standardization and the like, a plurality of
machine learning models are constructed and performance is compared, a
random forest is screened out as an optimal
algorithm, 11 core factors of gender, educational level,
residence,
body mass index BMI, age, occupation, smoking status, complication condition, drinking status, marital status and previous
tuberculosis history are selected to construct a model, internal and external
verification is carried out, and the model AUC reaches 0.872 and 0.812 respectively, and the calibration degree is good. A
prediction system including data collection, analysis and the like units is built based on the model, a
Web application tool of a Shiny framework is developed, interactive input of risk factors is realized, morbidity probability is automatically calculated, and three-level risk grading is realized. The application is simple in operation, accurate in prediction, can improve the active screening efficiency of
community elderly
pulmonary tuberculosis, and provides a scientific tool for accurate prevention and control.