Machine Learning in Medicine: A New Approach to Emergency Admissions

Background on Emergency Admissions

The rise in emergency hospital admissions has put immense pressure on healthcare systems that are often operating at capacity. Many visits to the emergency room are deemed unnecessary, prompting policymakers to seek strategies to alleviate this burden. The goal is to minimize non-essential admissions while ensuring that medical professionals can focus on patients requiring urgent care.

Identifying Risk Factors for Emergency Care

To improve the identification of patients needing emergency intervention, researchers are working on developing a comprehensive list of risk factors. This list aims to replace the traditional hands-on assessments conducted by doctors or nurses in determining the necessity of emergency care. Historically, statistical models have been employed, requiring data input to analyze risk factors. However, these models have shown limitations in accuracy due to the multitude of variables involved and their rigid rule structures.

The Role of Machine Learning in Healthcare

In response to the shortcomings of traditional statistical methods, researchers are exploring the potential of machine learning in medicine, a subset of artificial intelligence. Similar to statistical models, machine learning systems input data; however, they possess the ability to enhance their predictive capabilities as they process more information.

Research Findings from Oxford

Researchers from Oxford, UK, utilized a substantial dataset comprising 4.6 million patients to evaluate two machine learning models against an established statistical model. Their study aimed to determine which approach more accurately predicted hospital admissions. The findings, published in PLOS Medicine, demonstrated that machine learning techniques surpassed the traditional statistical model in predicting which patients would require hospitalization.

Implications for Healthcare Policy and Practice

The insights gained from this research hold significant promise for informing healthcare policies and practices, potentially lessening the strain on emergency rooms. The researchers aspire to enable physicians to effectively monitor patients’ risk scores, allowing timely interventions to prevent unnecessary emergency admissions.

Reference

Cribb, N. (VetMB DVSc Dip.ACVS). Rahimian, F., Salimi-Khorshidi, G., Payberah, A., et al. (2018). Predicting the risk of emergency admission with machine learning: Development and validation using linked electronic health records. PLOS Medicine, 15(11), e1002695. doi:10.1371/journal.pmed.1002695.