Predictive Analytics to Help Flag Deterioration in COVID-19 Patients

Introduction

In the past few months, there has been a surge in patients being admitted to the Emergency Department (ED) with respiratory illnesses related to the coronavirus disease 2019 (COVID-19). Assessing the risk of deterioration of these patients to perform triage is vital for clinical decision-making and resource allocation. While ED triage is tough under normal circumstances, during a pandemic, stressed and strained hospital resources add to the challenge. This problem is made even more complex due to our incomplete understanding of COVID-19. Data-driven risk evaluation based on AI and predictive analysis could, therefore, play a crucial role in streamlining ED triage.

To help with this is problem, researchers from Michigan Medicine have created a predictive analytics model that can correctly identify patient deterioration for both general ward and COVID-19 patients.

More About the Tool

This algorithm is more accurate than the Epic Deterioration Index (EDI), which is an existing tool used for patient deterioration investigation. The team has trained and authenticated a model termed as PICTURE model on a segment of COVID-19 patients who have been hospitalized using EHR data from the years 2014 to 2018. The team then applied this model to two test sets, comprising non-COVID-19 patients from 2019 and COVID-19-positive patients in 2020.

The result is that the predictive analytics algorithm can analyze a wide range of data, from vital signs and lab results to demographic information. Using this data, PICTURE has been able to identify patients who face the highest risk of health deterioration, as well as provide clarity on the risk factors that influence the prediction. This model can thus help health care experts to respond faster.

Using this PICTURE model, data from the electronic health record can be integrated and transformed into useful predictions based on the patient’s risk of suffering an adverse outcome,” states Brandon Cummings, a data scientist at MCIRCC.

“This is particularly important in the case of COVID-19 patients, whose health can decline unexpectedly and quickly. By forecasting these events before they occur, PICTURE can provide clinicians the time to react and stabilize the patient before drastic measures are needed.”

The Michigan Medicine researchers are currently working on testing the PICTURE model on other health systems and create specialized versions of it for individuals suffering from other health problems too.

As the COVID-19 cases continue to increase and wane, researchers believe that the PICTURE tool could act as a valuable resource for ICU staff that make triage decisions.

Predictive Analytics Tool to Predict Upcoming COVID-19 Surges

The COVID-19 Index is a predictive analytics tool designed to enable communities and healthcare organizations to see where cases of the virus are expected to increase and enable them to concentrate their resources on areas most impacted by the virus.

The COVID-19 Index displays data in a variety of formats and users can also see the data through tweets, chart books, APIs, chart books, and other summary analyses. Each of these formats provide early warnings to prevent the public and direct more efficient deployment of resources.

The ability to anticipate these events is extremely valuable while considering potential future waves of COVID-19 infections; however, the real the continued use of PICTURE will be crucial in all hospitalized patients no matter what the situation, since the virus itself is so unpredictable.

Conclusion

Today, we have very few tools at our hand to be able to predict what may happen. Some patients may deteriorate really quickly, while some may take longer. To be able to understand and see what is happening is extremely important.