Developing and validating a natural language processing algorithm to extract preoperative cannabis use status documentation from unstructured narrative clinical notes

OBJECTIVE

This study aimed to develop a natural language processing algorithm (NLP) using machine learning (ML) techniques to identify and classify documentation of preoperative cannabis use status.

MATERIALS AND METHODS

We developed and applied a keyword search strategy to identify documentation of preoperative cannabis use status in clinical documentation within 60 days of surgery. We manually reviewed matching notes to classify each documentation into 8 different categories based on context, time, and certainty of cannabis use documentation. We applied 2 conventional ML and 3 deep learning models against manual annotation. We externally validated our model using the MIMIC-III dataset.

RESULTS

The tested classifiers achieved classification results close to human performance with up to 93% and 94% precision and 95% recall of preoperative cannabis use status documentation. External validation showed consistent results with up to 94% precision and recall.

DISCUSSION

Our NLP model successfully replicated human annotation of preoperative cannabis use documentation, providing a baseline framework for identifying and classifying documentation of cannabis use. We add to NLP methods applied in healthcare for clinical concept extraction and classification, mainly concerning social determinants of health and substance use. Our systematically developed lexicon provides a comprehensive knowledge-based resource covering a wide range of cannabis-related concepts for future NLP applications.

CONCLUSION

We demonstrated that documentation of preoperative cannabis use status could be accurately identified using an NLP algorithm. This approach can be employed to identify comparison groups based on cannabis exposure for growing research efforts aiming to guide cannabis-related clinical practices and policies.

Protocol of a Combined Cohort and Cross-Sectional Study of Persons Receiving Medical Cannabis in Florida, USA: The Medical Marijuana and Me (M3) Study

Significant knowledge gaps regarding the effectiveness and safety of medical cannabis (MC) create clinical challenges for MC physicians, making treatment recommendations and patients choosing treatment among the growing number of options offered in dispensaries. Additionally, data describing the characteristics of people who use MC and the products and doses they receive are lacking. The Medical Marijuana and Me (M³) Study was designed to collect patient-centered data from MC users. We aim to describe preferred MC use patterns that patients report as "most effective" for specific health conditions and symptoms, identify user characteristics associated with such use patterns, characterize adverse effects, including cannabis use disorder, identify products and patient characteristics associated with adverse effects, describe concurrent prescription medication use, and identify concomitant medication use with potential drug-MC interaction risk. Among MC initiators, we also aim to quantify MC use persistence and identify reasons for discontinuation, assess MC utilization pattern trajectories over time, describe outcome trajectories of primary reasons for MC use and determine factors associated with different trajectories, track changes in concomitant substance and medication use after MC initiation, and identify factors associated with such changes. M³ is a combined study comprised of: (1) a prospective cohort of MC initiators completing surveys at enrollment, 3 months, and 9 months after MC initiation and (2) a cross-sectional study of current MC users. A multidisciplinary committee including researchers, physicians, pharmacists, patients, and dispensary personnel designed and planned study protocols, established study measures, and created survey questionnaires. M³ will recruit 1,000-1,200 participants aged ≥18 years, with ∼50% new and ∼50% current MC patients from MC clinics across Florida, USA. Study enrollment started in May 2022 and will continue until the target number of patients is achieved. Survey domains include sociodemographic characteristics, physical and mental health, cannabis use history, reasons for MC use and discontinuation, MC products and use patterns, concurrent use of prescription medications and other substances, and side effects. Data collected in the M³ Study will be available for interested researchers affiliated with the Consortium for Medical Marijuana Clinical Outcomes Research. The M³ Study and Databank will be the largest cohort of current and new MC users in Florida, USA, which will provide data to support MC-related health research necessary to inform policy and clinical practice and ultimately improve patient outcomes.

Priorities for Medical Marijuana Research from the Perspective of Physicians, Dispensary Owners/Staff, and Patients: A Survey Study

<b><i>Objective:</i></b> More patients are turning to medical marijuana as an alternative treatment, yet there are apparent knowledge gaps on the risk benefit of medical marijuana for a variety of indications. This study aimed to determine the priorities for medical marijuana research from the perspective of multiple stakeholders including patients, clinicians, and industry representatives. <b><i>Methods:</i></b> An anonymous survey was administered to attendees of the 2019 American Medical Marijuana Physicians Association annual meeting in Orlando, Florida. Respondents completed the survey on paper or smartphone via Qualtrics. The survey included questions on demographics and medical marijuana research priorities under the following broad categories: clinical conditions, safety issues, marijuana types, populations, and others. <b><i>Results:</i></b> Forty-six participants (56.5% female, mean age = 51.6 ± 14.1) responded to the survey. A majority were medical marijuana qualified physicians in Florida (56.5%), 30.5% other physicians or clinicians, and 21.7% medical marijuana patients (multiple choices allowed). The top conditions prioritized for research by this group were chronic pain, cancer, and anxiety, and the top priority safety issues were dosing/product choice, complications from smoking/vaping, and drug interactions. Regarding marijuana types, the group prioritized research on THC/CBD ratios, different modes of consumption, and terpenes. <b><i>Conclusions:</i></b> Findings from this survey indicate that medical marijuana stakeholders perceived a broad range of research topics as priorities. More research is needed to advance the evidence in these areas and provide guidance to patients, physicians, and the medical marijuana industry.

A Mapping Literature Review of Medical Cannabis Clinical Outcomes and Quality of Evidence in Approved Conditions in the USA from 2016 to 2019

In 2017, a National Academies of Sciences, Engineering, and Medicine (NASEM) report comprehensively evaluated the body of evidence regarding cannabis health effects through the year 2016. The objectives of this study are to identify and map the most recently (2016-2019) published literature across approved conditions for medical cannabis and to evaluate the quality of identified recent systematic reviews, published following the NASEM report. Following the literature search from 5 databases and consultation with experts, 11 conditions were identified for evidence compilation and evaluation: amyotrophic lateral sclerosis, autism, cancer, chronic noncancer pain, Crohn's disease, epilepsy, glaucoma, human immunodeficiency virus/AIDS, multiple sclerosis (MS), Parkinson's disease, and posttraumatic stress disorder. A total of 198 studies were included after screening for condition-specific relevance and after imposing the following exclusion criteria: preclinical focus, non-English language, abstracts only, editorials/commentary, case studies/series, and non-U.S. study setting. Data extracted from studies included: study design type, outcome definition, intervention definition, sample size, study setting, and reported effect size. Few completed randomized controlled trials (RCTs) were identified. Studies classified as systematic reviews were graded using the Assessing the Methodological Quality of Systematic Reviews-2 tool to evaluate the quality of evidence. Few high-quality systematic reviews were available for most conditions, with the exceptions of MS (9 of 9 graded moderate/high quality; evidence for 2/9 indicating cannabis improved outcomes; evidence for 7/9 indicating cannabis inconclusive), epilepsy (3 of 4 graded moderate/high quality; 3 indicating cannabis improved outcomes; 1 indicating cannabis inconclusive), and chronic noncancer pain (12 of 13 graded moderate/high quality; evidence for 7/13 indicating cannabis improved outcomes; evidence from 6/7 indicating cannabis inconclusive). Among RCTs, we identified few studies of substantial rigor and quality to contribute to the evidence base. However, there are some conditions for which significant evidence suggests that select dosage forms and routes of administration likely have favorable risk-benefit ratios (i.e., epilepsy and chronic noncancer pain). The body of evidence for medical cannabis requires more rigorous evaluation before consideration as a treatment option for many conditions, and evidence necessary to inform policy and treatment guidelines is currently insufficient for many conditions.

Practice Patterns and Training Needs Among Physicians Certifying Patients for Medical Marijuana in Florida

BACKGROUND

Little is known about the clinical training or practice experiences among physicians who certify patients for medical marijuana. The objective of this study was to determine information sources, factors influencing recommendations, clinical practices in patient assessment, communications, and recommendations, and priority areas for additional training among physicians who certify patients for medical marijuana.

METHODS

A cross-sectional state-wide anonymous survey of registered medical marijuana physicians in Florida between June and October 2020 was administered. Numerical responses were quantified using counts and percentages. The frequencies for "often" and "always" responses were aggregated when appropriate.

RESULTS

Among 116 respondents, the mean (standard deviation) age was 57 (12) years old, and 70% were male. The most frequently used information sources were research articles (n = 102, 95%), followed by online sources (n = 99, 93%), and discussions with other providers and dispensary staff (n = 84, 90%). Safety concerns were most influential in patient recommendations (n = 39, 39%), followed by specific conditions (n = 30, 30%) and patient preferences (n = 26, 30%). Ninety-three physicians (92%) reported they "often" or "always" perform a patient physical exam. Eighty-four (77%) physicians provided specific administration route recommendations. Half (n = 56) "often" or "always" provided specific recommendations for Δ-9-tetrahydrocannabinol: cannabidiol ratios, while 69 (62%) "often" or "always" provided specific dose recommendations. Online learning/training modules were the most preferred future training mode, with 88 (84%) physicians "likely" or "very likely" to participate. The top 3 desired topics for future training were marijuana-drug interactions (n = 84, 72%), management of specific medical conditions or symptoms (n = 83, 72%), and strategies to reduce opioids or other drugs use (n = 78, 67%).

CONCLUSIONS

This survey of over 100 medical marijuana physicians indicates that their clinical practices rely on a blend of research and anecdotal information sources. While physicians report clinical factors as influential during patient recommendation, patient assessment practices and treatment regimen recommendations vary substantially and rely on experimental approaches. More research is needed to inform evidence-based practice and training, especially considering details on drug interactions, risk-benefit of treatment for specific clinical conditions, and strategies to reduce opioid use.

A systematic review of evidence for cannabis and cannabinoids as adjuvant therapy in palliative and supportive oncology care

12091

Background: Medical cannabis use is increasing significantly in the United States as states reduce restrictions. However, ambiguity concerning the evidence for medical cannabis efficacy and safety, especially in the field of oncology, is persistent. Clinicians therefore face challenges in examining benefits and risks of medical cannabis as adjuvant treatment for cancer patients. This study identifies and evaluates the most recent available evidence for the efficacy of cannabis and cannabinoids as adjuvant in supportive and/or palliative use in patients with cancer. Methods: Electronic databases searched included PubMed, Embase, Web of Science, and Cochrane Library to identify studies published following the latest available systematic review, between July 2016 through October 2019. Studies conducted outside the United States, studies not evaluating cannabis or cannabinoids in Oncology care, and preclinical studies were excluded. Findings were organized in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) framework. Lastly, qualitative synthesis was used to generate summary statements about the role of cannabis and cannabinoids as adjuvant in supportive and/or palliative cancer care. Results: We screened 2,267 articles and included 96 studies in our qualitative synthesis. Among those were 2 RCT’s (1 completed), 6 Systematic reviews with Meta-analysis, 4 Systematic reviews without Meta-analysis, 71 other types of reviews and 13 observational studies. The most frequently reported outcomes assessed were efficacy of cannabis and cannabinoids for: pain (40 of 96; 17 indicating improvement), nausea and vomiting (26 of 96; 20 indicating improvement), cachexia (22 of 96; 2 indicating improvement), and utilization patterns of cannabis and/or cannabinoids among cancer patients (8 of 96). Conclusions: Latest available prevalence estimates indicate that a significant proportion of patients in the United States with cancer use cannabis and/or cannabinoids (18.3-40.0%). There is substantial evidence for the effectiveness of cannabis and cannabinoids in treating cancer-related pain; specifically, oromucosal THC/CBD spray. There is conclusive evidence for the effectiveness of cannabis and cannabinoids in relieving chemotherapy-induced nausea and vomiting; specifically, oral THC. There is inconclusive evidence regarding the effectiveness of cannabis and cannabinoids in treating cancer-related cachexia.

Abstract WP245: Demographics, Risk Factors and Outcomes of Stroke in Young Adults Aged 18-45 Years in Comparison With Those Older Than 45 Years

Introduction: The prevalence of stroke in young adults have increased with the prevalence of cerebrovascular risk factors in this population. The objective of this study was to compare stroke prevalence in age groups 18-45 years versus more than 45 years by demographics, risk factors, and outcome.

Methods: We analyzed the National Inpatient Sample Database from the Healthcare Cost and Utilization Project for the years 2008-14. Patients with stroke were identified based on discharge DRG codes 61-66 for acute ischemic stroke with use of thrombolytic therapy, intracranial hemorrhage or cerebral infarction. Demographics, risk factors from comorbid condition codes, and outcomes were analyzed.

Results: Out of a total of 3,624,243 adult patients with stroke hospitalized during the years 2008-14, 189103 (5.21%) hospitalizations were of patients aged 18-45 years. The ratio of stroke hospitalizations to overall hospitalizations increased from 154 to 185 per 10,000 hospitalizations from the year 2008 to 2014 with increase in stroke hospitalizations in both age groups. Stroke in young adults aged 18-45 years was more common in men, Blacks and Hispanics compared to those older than 45 years. Cerebrovascular risk factors were common in young adults with stroke and the prevalence of obesity, arterial malformation, tobacco use or alcohol abuse was higher in stroke patients aged 18-45 years compared to those older than 45 years (Table 1). There was up to 6% increase in the prevalence of cerebrovascular risk factors from 2008 to 2014. Young adults with stroke aged 18-45 years were more likely to be discharged home while older adults were more likely to be discharged to skilled nursing facility.

Conclusions: There are gender and racial differences in stroke among adults aged 18-45 years compared to those older than 45 years age. Modifiable cerebrovascular risk factors are common in young adults with stroke. Most notably, both obesity and smoking were particularly higher in younger patients.