Authors: Neslihan Suzen, Evgeny M. Mirkes, Damian Roland, Jeremy Levesley, Alexander N. Gorban, Tim J. Coats

Published on: February 09, 2024

Impact Score: 8.35

Arxiv code: Arxiv:2402.06563

Summary

• What is new: The research breaks new ground by showing that missing data in EPRs are likely non-random and explaining the link to health care professionals’ practice patterns. It also introduces an effective imputation method using k-Nearest Neighbour for paediatric emergency and trauma case data.
• Why this is important: Missing information in Electronic Patient Records (EPRs) can introduce bias in clinical data analysis, affecting the validity of clinical decisions.
• What the research proposes: The study proposes using statistical approaches and machine learning techniques, specifically Singular Value Decomposition (SVD) and k-Nearest Neighbour (kNN) for imputing missing data in EPRs.
• Results: Among the tested methods, the 1NN imputer yielded the best results, suggesting that imputing attributes from the most similar patients is an effective strategy.

Technical Details

Technological frameworks used: nan

Models used: SVD, kNN

Data used: Paediatric emergency data, TARN database

Potential Impact

Healthcare providers and institutions, Electronic Health Record (EHR) system vendors, companies specializing in health data analytics.

Want to implement this idea in a business?

We have generated a startup concept here: HealFill.