Over a million developers have joined DZone.
{{announcement.body}}
{{announcement.title}}

Using Big Data to Reduce Drug Overdoses

DZone's Guide to

Using Big Data to Reduce Drug Overdoses

This research highlights how a big data approach can help hospitals understand which patients might progress to chronic opioid therapy after discharge.

· Big Data Zone ·
Free Resource

Hortonworks Sandbox for HDP and HDF is your chance to get started on learning, developing, testing and trying out new features. Each download comes preconfigured with interactive tutorials, sample data and developments from the Apache community.

The use of big data to identify at-risk groups is something that is showing considerable growth, both as more data is made available and greater computational power is available to make sense of the data.

A team from the University of Colorado highlights how this approach can help hospitals understand which patients might progress to chronic opioid therapy after discharge.

The issue is serious, as over 63,000 people died in the United States from a drug overdose last year, with opioids involved in around 75% of those deaths. What's more, national data suggests that there are over two million people in the US with an opioid use disorder.

"Doctors and patients are increasingly aware of the dangers associated with overprescribing of opioids," the authors say. "We can assist physicians in making informed decisions about opioid prescribing by identifying patient characteristics which put them at risk progressing to chronic opioid use."

Risk of Progression

The researchers aimed to build a prediction model to accurately identify the hospitalized patients who were at the highest risk of progressing to chronic opioid use following their discharge from the hospital.

The model was built using data from the electronic medical records at Denver Health Medical Center. Patients were classified as being on chronic opioid therapy (COT) due either to receiving a supply of oral opioids for 90 days or more or filling ten or more opioid prescriptions over a one year period.

The data contained in the medical record allowed the team to identify a number of variables that were strongly linked to a progression to COT. For instance, it might reveal a history of substance abuse or the receipt of a benzodiazepine.

The model was able to accurately predict chronic opioid therapy in 79% of patients and indeed was also able to predict no COT correctly in 78% of patients. The team believes that their work is the first of its kind to be developed for COT risk, and improves upon software such as the Opioid Risk Tool (ORT), which they claim has not been validated in a hospital setting.

"This prediction model could be incorporated into the electronic health record and would activate when a physician orders opioid medication. It would inform the physician of their patient's risk for developing COT and may impact their prescribing practices," the authors say.

With the data required to function already available in the system, there are no extra requirements placed upon the physician. As such, the team believes it would be fairly inexpensive to implement and particularly helpful support in the busy life of the doctor. Before that can happen, however, the team need to test the system more rigorously in other health care systems to determine that it works in a range of patient populations.

"Our goal is to manage pain in hospitalized patients, but also to better utilize effective non-opioid medications for pain control," the researchers conclude. "Ultimately, we hope to reduce the morbidity and mortality associated with long-term opioid use."

Hortonworks Community Connection (HCC) is an online collaboration destination for developers, DevOps, customers and partners to get answers to questions, collaborate on technical articles and share code examples from GitHub.  Join the discussion.

Topics:
big data ,healthcare ,data analytics ,predictive analytics

Published at DZone with permission of

Opinions expressed by DZone contributors are their own.

{{ parent.title || parent.header.title}}

{{ parent.tldr }}

{{ parent.urlSource.name }}