With opioid abuse a raging epidemic in the United States, hospitals are looking for effective ways to track and manage opioid treatment effectively. In an effort to move in this direction, a group of researchers has developed a model which predicts the likelihood of future chronic opioid use based on hospital EHR data.
The study, which appears in the Journal of General Internal Medicine, notes that while opioids are frequently prescribed in hospitals, there has been little research on predicting which patients will progress to chronic opioid therapy (COT) after they are discharged. (The researchers defined COT as when patients were given a 90-day supply of opioids with less than a 30-day gap in supply over a 180-day period or receipt of greater than 10 opioid prescriptions during the past year.)
To address this problem, researchers set out to create a statistical model which could predict which hospitalized patients would end up on COT who had not been on COT previously. Their approach involved doing a retrospective analysis of EHR data from 2008 to 2014 drawn from records of patients hospitalized in an urban safety-net hospital.
The researchers analyzed a wide array of variables in their analysis, including medical and mental health diagnoses, substance and tobacco use, chronic or acute pain, surgery during hospitalization, having received opioid or non-opioid analgesics or benzodiazepines during the past year, leaving the hospital with opioid prescriptions and milligrams of morphine equivalents prescribed during their hospital stay.
After conducting the analysis, researchers found that they could predict COT in 79% of patients, as well as predicting when patients weren’t on COT 78% of the time.
Being able to predict which patients will end up on COT after discharge could prove to be a very effective tool. As the authors note, using EHR data to create such a predictive model could offer many benefits, particularly the ability to identify patients at high risk for future chronic opioid use.
As the study notes, if clinicians have this information, they can offer early patient education on pain management strategies and where possible, wean them off of opioids before discharging them. They’ll also be more likely to consider incorporating alternative pain therapies into their discharge planning.
While this data is exciting and provides great opportunities, we need to be careful how we use this information. Done incorrectly it could cause the 21% who are misidentified as at risk for COT to end up needing COT. It’s always important to remember that identifying those at risk is only the first challenge. The second challenge is what do you do with that data to help those at risk while not damaging those who are misidentified as at risk.
One issue the study doesn’t address is whether data on social determinants of health could improve their predictions. Incorporating both SDOH and patient-generated data might lend further insight into their post-discharge living conditions and solidify discharge planning. However, it’s evident that this model offers a useful approach on its own.