A predictive method to determine incomplete electronic medical records
- Publication Type:
- Conference Proceeding
- SIGMIS-CPR 2018 - Proceedings of the 2018 ACM SIGMIS Conference on Computers and People Research, 2018, pp. 99 - 106
- Issue Date:
Copyright Clearance Process
- Recently Added
- In Progress
- Closed Access
This item is closed access and not available.
© 2018 Association for Computing Machinery. This paper is utilizing predictive models to determine missing electronic medical records (EMR) at general practice offices. Prior research has addressed the missing values problem in the EMRs used for secondary analysis. However, health care providers are overlooking the missing records problem that stores the patients’ medical visits information in EMRs. Our study provides a technique to predict the number of EMR entries for each practice based on their past data records. If the number of EMR entries is less than predicted, it warns the occurrence of missing records with the 95% confidence interval. The study uses seven years of EMRs from 14 general practice offices to train the predictive model. The model predicts EMR data entries and accordingly identified missing EMRs for the following year. We compared the actual visits illustrated by de-identified billing data to the predictive model. The study found auto-correlation method improves the performance of identifying missing records by detecting the period of prediction. In addition, artificial neural networks and support vector machines perform better than other predictive methods depending on whether the analysis aims at detecting missing EMRs or when identifying complete EMRs with no missing records. Results suggest that clinicians and medical professionals should be mindful of the potential missing records of EMRs prior any secondary analysis.
Please use this identifier to cite or link to this item: