How machine learning facilitates decision making in emergency departments: Modelling diagnostic test orders

OBJECTIVES

Since emergency departments (EDs) are responsible for providing initial care for patients who may need urgent medical care, they are highly sensitive to increased patient delays. A key factor that increases patient delays is ordering diagnostic tests. Therefore, understanding the factors increasing diagnostic test orders and proposing efficient models may facilitate decision making in EDs.

METHODS

Month and week of the year, day of the week, and daily numbers of patients encoded based on 21 different ICD-10 codes were used as input variables. Daily test frequencies of patients requiring tests from laboratory and imaging services were modelled separately by linear regression models. Although significance of the input variables was identified based on these models, obtained forecasts and residuals were further processed by machine learning techniques to obtain hybrid models.

RESULTS

Day of the week, and number of patients with ICD-10 codes of 'A00-B99', 'I00-I99', 'J00-J99', 'M00-M99' and 'R00-R99' were significant in both test types. In addition to these, although daily patient frequencies with 'H60-H95', 'N00-N99' and 'O00-O9A' were significant for laboratory services, 'L00-L99', 'S00-T88' and 'Z00-Z99' were significant for imaging services. Although prediction accuracies of regression models were, respectively, as 93.658% and 95.028% for laboratory and imaging services modelling, they increased to 99.997% and 99.995% with the machine learning-integrated hybrid model.

CONCLUSION

The significant factors identified here can predict increases in use of laboratory and imaging services. This could enable these services to be prepared in advance to reduce ED patient delays, thereby reducing ED overcrowding. The proposed model may also be efficiently used for decision making.

Machine learning in patient flow: a review

This work is a review of the ways in which machine learning has been used in order to plan, improve or aid the problem of moving patients through healthcare services. We decompose the patient flow problem into four subcategories: prediction of demand on a healthcare institution, prediction of the demand and resource required to transfer patients from the emergency department to the hospital, prediction of potential resource required for the treatment and movement of inpatients and prediction of length-of-stay and discharge timing. We argue that there are benefits to both approaches of considering the healthcare institution as a whole as well as the patient by patient case and that ideally a combination of these would be best for improving patient flow through hospitals. We also argue that it is essential for there to be a shared dataset that will allow researchers to benchmark their algorithms on and thereby allow future researchers to build on that which has already been done. We conclude that machine learning for the improvement of patient flow is still a young field with very few papers tailor-making machine learning methods for the problem being considered. Future works should consider the need to transfer algorithms trained on a dataset to multiple hospitals and allowing for dynamic algorithms which will allow real-time decision-making to help clinical staff on the shop floor.

Predicting waiting and treatment times in emergency departments using ordinal logistic regression models

BACKGROUND

Since providing timely care is the primary concern of emergency departments (EDs), long waiting times increase patient dissatisfaction and adverse outcomes. Especially in overcrowded ED environments, emergency care quality can be significantly improved by developing predictive models of patients' waiting and treatment times to use in ED operations planning.

METHODS

Retrospective data on 37,711 patients arriving at the ED of a large urban hospital were examined. Ordinal logistic regression models were proposed to identify factors causing increased waiting and treatment times and classify patients with longer waiting and treatment times.

RESULTS

According to the proposed ordinal logistic regression model for waiting time prediction, age, arrival mode, and ICD-10 encoded diagnoses are all significant predictors. The model had 52.247% accuracy. The model for treatment time showed that in addition to age, arrival mode, and diagnosis, triage level was also a significant predictor. The model had 66.365% accuracy. The model coefficients had negative signs in the corresponding models, indicating that waiting times are negatively related to treatment times.

CONCLUSION

By predicting patients' waiting and treatment times, ED workloads can be assessed instantly. This enables ED personnel to be scheduled to better manage demand supply deficiencies, increase patient satisfaction by informing patients and relatives about expected waiting times, and evaluate performances to improve ED operations and emergency care quality.