Three Ways Natural Language Processing (NLP) Can Help

Conquer the Opioid Crisis



Nearly thirty years into a national public health crisis, solutions continue to be explored using free text and structured data. 

In this post, Melax Technologies reviews the public health crisis of substance abuse. A variety of solutions are being used to develop an understanding of the magnitude of the problem and possible solutions. National Language Processing (NLP) brings the crisis and answers into focus. 

In 2018, 128 individuals in the US died each day from opioid overdose. The use and misuse of prescription painkillers, leading further to the misuse of street drugs, such as heroin and fentanyl, has put an enormous strain on the US economy, especially within public and social health administration. The CDC estimates more than $78 billion a year is attributed to substance misuse leading to lost productivity, and added healthcare, rehabilitation and criminal justice costs.


Evolution of a crisis

In the 1990s, the pharmaceutical development of opioids, coupled with reassurance that these medications would not be addictive, drove up prescription rates. By the time the healthcare profession realized the addictive properties of these medications, overdose was on the rise. So much so that by 2017 47,000 individuals died of overdose from prescription and street drug abuse. Another 1.7 million Americans abused these prescription medications and more than 600,000 abused heroin (in some cases, concurrently).


The numbers behind drug abuse

The opioid crisis can be explained as simply as miscommunication between the pharma industry and the healthcare system. However, the complexity of the issue becomes evident in the numbers behind drug abuse. As many as 29% of patients prescribed opioids misuse them, with about 4 to 6 percent going on to use heroin. When asked, 80% of heroin users report becoming addicted after misusing prescription painkillers.

While apparently suffering alone, drug users pass their health problems on to society at large, with neonatal abstinence syndrome and the spread of HIV and Hepatitis C among the complications faced by the population.


Government Intervention

Both the Department of Health and Human Services (HHS) and the National Institutes of Health (NIH) began to explore recovery systems, overdose-reversing drugs, public health education and improved pain management programs to combat abuse. In 2018, the NIH launched HEAL (Helping to End Addiction Long-Term). This initiative was intended to bring together multiple agencies to contain the public health crisis.


Healthcare Challenges

Nevertheless, as these initiatives developed, a clear picture of possible drug abusers and the extent of the problem remained hidden. For example, opioid overdose frequently remained underdiagnosed if not explicitly recorded in billing (ICD) codes. However, more prescription detail was available in the free text of clinical records. Additionally, drug names, and attributes such as durations and dosage vary among clinical records, which need to be normalized to clinical data standards.

NLP allows researchers and clinicians to collect extracted information effortlessly in order to develop protocols and treatment plans with clear “apples to apples” intelligence. Below are described three ways NLP is being used now to curb misuse


Three Ways NLP is Used to Recognize Opioid Abuse

  1. Pulling prescription detail from free text within the clinical record. This detail includes medication and its eight attributes (reason, ADE, frequency, strength, duration, route, form and dosage). Combined with medical information stored in unstructured tables, researchers are able to calculate the accurate value of morphine milligram equivalents (MME). MME is the CDC measure of the amount of milligrams of morphine an opioid dose is equal to when prescribed.
  2. Recognizing symptoms and mentions of overdose from free text in clinical records. This capability is used to develop phenotype algorithms to screen patients for opioid misuse.
  3. Mapping clinical data standards (for example, Fast Healthcare Interoperability Resources - FHIR) to allow the transfer of patient information, including substance misuse, across healthcare systems.


Conclusion

Opioid misuse will not be easily resolved. However, with natural language processing researchers, clinicians, and the public will be able to see not only the current scope of the problem, but ways to improve outcomes using all data, regardless of how deeply it is hidden within the medical record.

To learn how Melax Technologies’ clients use NLP for this and other healthcare programs, visit www.melaxtech.com.


Reference:

www.drugabuse.gov/drug-topics/opioid-overdose-crisis