Addressing the Opioid Epidemic with an NLP-based Clinical Opioid Data Repository (nCODR)

Strategies to address opioid abuse and overdoses have had limited effectiveness in mitigating the high rate of emergency room visits and mortality due to this growing opioid epidemic. A multi-level approach to the opioid epidemic appears necessary to include access, education, prevention, early identification, and mitigation of this scourge, but more importantly we need an evidence-based approach. Prescription monitoring programs may identify patients using opioids and physicians prescribing them, but will not address sources of drugs through other channels.

Studies have shown a near-linear relationship and steady increase from overdose deaths representing a small but quickly growing number of patients with addiction subject to increasing opioid use with tolerance eventually reaching fatal levels. Emergency room visits for opioid overdoses occur at over 30 times the rate of fatal overdoses, suggesting the tip of this iceberg. The straight-line increases leading to overdoses make any proportion of this number grow over time until we can make an impact in the number of people exposed to opioids or mitigate the percent and rate at which they progress to abuse resulting in deaths.

The heroin addiction problem of the 1960’s and 70’s was only partially addressed by the war on drugs and the methadone program. It was only curtailed due to hepatitis and AIDS that quickly spread through the IV drug use community. Addressing the opioid problem requires much greater understanding of factors increasing the risk that any given patient may become addicted to opioids and exploiting any opportunity to intervene over what may be years before overdose occurs.

To address the opioid addiction problem effectively we need more data that is not traditionally captured or available for analysis. The data we need centers around patients, providers, medical conditions and changes occurring over time. Examples of questions we need answered include reasons for starting on these highly addictive drugs, such as legitimate medical pain management and specifics about medical conditions. Data needed includes identification of which patients are at high risk for addiction; when and for whom alternative non-addictive step approaches can work; and which treatment is best utilized effectively. In addition to widely implemented drug screening programs and profiling patients who exhibit behaviors and risk factors for abuse, healthcare providers can be educated and given software tools that help identify patients at risk for opioid abuse and afford opportunity for intervention.

Medicaid patients are known to be at high risk. Are these patients Medicaid eligible because of work issues due to opioid addiction? How many of them have medical conditions that required opioid pain management and are there certain medical conditions where this is more common? Do we have veterans with injuries and what is the best way to treat and monitor them to avoid addiction?

To address this massive problem, it will be most helpful to develop a data-rich clinical opioid data repository. With existing technology we can build this quickly and inexpensively. The Framingham Study was a general medical database that has yielded enormous amounts of information representing a snapshot of general data collected. A focused opioid clinical data repository will allow clinicians to manage individual patients within HIPAA security and allow deidentification of patients for population-based studies by researchers, epidemiologists, public health professionals, politicians and others to answer opioid-specific questions, evaluate cohorts, and scientifically validate outcomes with best treatments.

An opioid data repository will benefit from a ‘Big Data’ approach. Natural language processing (NLP) will allow standardization of the unstructured information into structured information that can easily be queried with simple tools allowing anyone with appropriate access to do their own research project. Activities such as exercise, work, sleep, accidents, alcohol use, relationships can all be classified with standardized SNOMED terminologies. Some relevant information such as prescribed medication history and medical diagnoses are available from most EHRs. However, data such as alcohol and non-prescription drug use, complaints common to the opioid population of constipation, irritability, depression, withdrawal symptoms and other related information is data is currently trapped in EHRs; it is unstructured and practically unavailable. This unstructured data can be made accessible using NLP. Dashboards can be used for common reports and custom queries can be made allowing use of the real-world data for multiple purposes including virtual trials.

Focusing our efforts on building a natural language processing Clinical Opioid Data Repository (nCODR) can help us scientifically measure the opioid addiction problem and then look at what is working and not working in addressing this epidemic. Natural language processing will accommodate both structured and unstructured data from numerous sources to quickly obtain sufficient information for individual and population health management with security and role based privileges. We have a clear mission and focus to build a scalable cloud-based platform suitable for everyone’s needs with low expense, low risk and short timetable.

Contact us to learn more or schedule a meeting.

James M. Maisel, M.D.
Chairman, ZyDoc/Medisapien
"Medical Knowledge Management Solutions"
Stakeholders can reach me via email.

Share this:


March 7, 2018 In " Blog " , " EHR Usability " , " EHRs "
March 7, 2018 In " Blog " , " EHR Usability " , " EHRs "
February 20, 2018 In " Blog " , " Transcription Services " , " speech recognition "

Posted in Analytics, Blog, MediSapien, NLP, Big Data