By Dr. Elizabeth Marshall, MBA, Director of Clinical Analytics, Linguamatics
Twitter: @Linguamatics
Observing the Dance
“You can’t save everyone” I painfully remind myself after hearing about the fentanyl overdose of the 29-year-old man that we so desperately tried to help over the years. A man that floated in and out of our lives – when he wanted to work (as a day-laborer for our construction projects,) stay sober and turn his life around. After four years of this sad dance, we had suspected he might be dancing his final dance after we discovered he was out of rehabilitation and nowhere to be found.
Statistically speaking we were unfortunately bound to know someone that abused opioids. You will notice I didn’t quote a statistic, and there is a good reason. The only opioid statistics I believe are close to accurate are the numbers on opioid overdose deaths. It’s not that I don’t think society makes a valiant effort in obtaining the statistics. The truth is opioid addicts are hard to find in the early stages, and they rarely self-report. Once the addiction is established they are somewhat easier to identify – but by that time they are barely aware of their surroundings and hardly capable of survey participation.
Early identification is key – more so than we ever realized. As a physician, I was taught that opioids are safe for short-term use. But the meaning of the term ‘short-term’ is shrinking drastically – this recent Newsweek article reports that according to the CDC, dependency starts within just a few days. As a former research scientist, I have reviewed thousands of patient charts – and the majority of opioids I have seen prescribed are for a minimum of 5-7 days.
How do we stop this musical roundabout?
Identify patterns
Yes, I know, easier said than done. However, with vigilance and informatics solutions – for example, analytical reporting techniques supported with robust data extracted by using Natural Language Processing (NLP) over a multitude of data sources – we can identify patterns that reveal possible abuse. This requires examination of content from both structured and unstructured free text sources.
Some identification efforts that are enhanced with NLP-supported analytical reporting include:
- Close monitoring of the clinician notes of chronic pain patients to identify symptoms that may indicate addiction or withdrawal from opioids.
- Creating pain registries within hospital systems that identify timelines for prescribed pain medications. Registries should include metrics such as dosage, reason prescribed, and factors that put patients at higher risk for abuse, such as feelings of hopelessness.
- Identifying people that seem to be ‘accident’ prone with histories of injuries requiring pain medications. Then consider the reason for accidents. For example: Does the patient have medical issues that are causing so many accidents? Should the patient have a referral to an otolaryngologist for balance? Or, are there signs of addiction within the clinician’s notes that suggest possible self-harm in order to receive medications?
- Identifying individuals that often visit urgent care settings and ask for pain relief medications for various reasons. By reviewing clinician observations from within chart notes, drug-seeking behaviors may become apparent.
To help with clinical decision-making, reports should include known opioid abuse risk factors, such as those included in the Opioid Risk Tool (ORT). A patient may be at higher risk for abuse if there is a personal or family history of alcohol abuse, illicit drug use, or sexual abuse, or if the patient has one or more of these common psychiatric disorders: Bipolar, mood, anxiety, personality or stress disorders. Most importantly, once institutions set up these reports, they need to utilize them regularly. It’s critical that organizations have a workflow in place for proper follow-up. I realize this seems like a logical statement- and it is, but communication errors are common and sometimes dire.
Addicts are creative, so we must be too. We can’t just rely on Prescription Drug Monitoring Programs (PDMPs), which are helpful but also have limitations. The key is to have collaborative efforts that go beyond single institutions, by building registries that combine prescribing information from different data pools (pharmacies, social media, health records.) Unfortunately, this is an even less realistic option, given the challenging legalities required. Regardless, efforts under the ‘roof’ of single entities such as hospital systems, national pharmacies, etc. are critical. Community efforts, such as this exemplar initiative in North Carolina where police help instead of prosecute addicts, are also vital for combating the epidemic.
The bottom line: We must all work together to do what we can and should do. While we cannot save everyone, we can put our best foot forward to lead this dance towards a more positive conclusion.