Related Experiment Video
Updated: Aug 26, 2025
05:58
Evaluation of a Point-of-Care Testing Analyzer for Measuring Peripheral Blood Leukocytes
Published on: March 22, 2022
4.2K
Turnaround time prediction for clinical chemistry samples using machine learning.
Eline R Tsai1,2, Derya Demirtas1, Nick Hoogendijk1
1Center for Healthcare Operations Improvement and Research (CHOIR), University of Twente, Enschede, The Netherlands.
Clinical Chemistry and Laboratory Medicine
|October 11, 2022
Summary
Accurate turnaround time (TAT) prediction in medical labs is now possible using the Extra Trees Regressor model. This helps improve healthcare efficiency by identifying factors influencing TAT and enabling timely interventions for prolonged testing times.
Area of Science:
- Medical Diagnostics
- Laboratory Medicine
- Health Informatics
Background:
- Turnaround time (TAT) is a critical performance metric for medical diagnostic laboratories.
- Accurate TAT prediction is essential for proactive management of delays and efficient healthcare resource allocation.
- Focusing on the automated pre-analytical and analytical phases is key to improving TAT.
Purpose of the Study:
- To develop and evaluate a predictive model for laboratory turnaround time (TAT).
- To identify key features influencing TAT in the pre-analytical and analytical phases.
- To enhance laboratory efficiency and healthcare planning through accurate TAT prediction.
Main Methods:
- Utilized a dataset of 90,543 clinical chemistry samples.
- Analyzed 39 features, including priority level and stage-specific workloads.
- Employed PyCaret to compare regression models, including Extra Trees (ET) Regressor, Ridge Regression, and K Neighbors Regressor.
- Evaluated models using relative residual and SHAP (SHapley Additive exPlanations) values.
Main Results:
- The Extra Trees (ET) Regressor model demonstrated superior performance with an R² of 0.63, MAE of 2.42 min, and MAPE of 7.35%.
- The average TAT was 30.09 minutes, with 77% of samples having a relative residual error of ≤10%.
- SHAP analysis revealed that pre-analysis workload and the number of modules visited were the primary drivers of TAT.
Conclusions:
- The ET Regressor model provides accurate TAT predictions for medical laboratories.
- Identified key TAT influencers enable laboratories to take timely actions on prolonged testing.
- Improved TAT prediction enhances healthcare provider planning and overall healthcare efficiency.
