EHR Cancer Risk Prediction With Missing-Data Imputation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional computer-implemented tools for predicting esophageal adenocarcinoma (EAC) and gastric cardia adenocarcinoma (GCA) are ineffective when key predictors are missing from electronic health records (EHRs, often due to unreported or misdiagnosed gastroesophageal reflux disease (GERD), leading to inaccurate or no predictions and missed opportunities for early detection.
Innovation Solution
A computer-implemented method using machine learning that imputes missing values of key predictors in EHRs through simple random sampling imputation, trains a model with an extreme gradient boosting algorithm and multiple decision trees, and tunes parameters for optimal performance, enabling accurate prediction of EAC and GCA risk.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional prediction tools require explicit risk factor indications (such as GERD diagnosis) in EHRs, then prediction accuracy is improved when data is complete, but prediction capability is lost when key predictors are missing
Solution Approach 1:
The system performs preliminary imputation of missing risk factor values before prediction. By pre-filling missing GERD and other risk factor data using random sampling from the training dataset, the system ensures that prediction models receive complete input data even when EHRs have gaps, thus maintaining both accuracy and capability.
Solution Approach 2:
The system introduces an intermediary imputation step between data collection and prediction. This intermediary process uses random sampling to generate plausible values for missing risk factors, acting as a bridge that allows conventional prediction tools to function effectively even with incomplete EHR data.
2Ease of operation
If EHRs are used as the sole data source for prediction, then data collection is simplified, but prediction reliability deteriorates due to missing or unreported risk factors
Solution Approach 1:
The random sampling imputation process serves as an intermediary that enhances EHR data without requiring additional data collection efforts. It generates plausible risk factor values based on population statistics from the training set, maintaining operational simplicity while improving prediction reliability.
Solution Approach 2:
The system changes the state of missing parameters by imputing them with sampled values. This parameter transformation converts incomplete EHR records into complete prediction-ready datasets, preserving the simplicity of EHR-based collection while enhancing reliability through statistical supplementation.
3Adaptability or versatility
If machine learning models are trained with imputed missing values, then prediction capability is enhanced for patients with incomplete records, but model complexity increases
Solution Approach 1:
The system performs imputation as a preliminary data preprocessing step before model training. By completing missing values in advance using random sampling from the training dataset, the system maintains relatively simple models while enhancing their capability to handle real-world incomplete EHR data.
Data Source
AI summary
Systems and methods for predicting esophageal adenocarcinoma (EAC) and gastric cardia adenocarcinoma (GCA) using machine learning are provided. An example system may obtain an electronic health record (EHR) dataset, identify missing values in the EHR dataset, and generate imputed values for the missing values using simple random sampling imputation. The system may train a model using an extreme gradient boosting algorithm and a training dataset including the EHR dataset to generate a trained model including multiple decision trees. Training the model includes tuning the model to achieve a greatest value of an area under a receiver operating characteristic curve associated with the model. The system may obtain a patient EHR dataset, generate a prediction associated with a risk of EAC and/or GCA by applying the trained model to the patient EHR dataset, and provide the prediction to a computing device to determine a patient treatment protocol.