Clinical documentation of patient-reported medical cannabis use in primary care: Toward scalable extraction using natural language processing methods.

Background: Most states have legalized medical cannabis, yet little is known about how medical cannabis use is documented in patients' electronic health records (EHRs). We used natural language processing (NLP) to calculate the prevalence of clinician-documented medical cannabis use among adults in an integrated health system in Washington State where medical and recreational use are legal. Methods: We analyzed EHRs of patients ≥18 years old screened for past-year cannabis use (November 1, 2017-October 31, 2018), to identify clinician-documented medical cannabis use. We defined medical use as any documentation of cannabis that was recommended by a clinician or described by the clinician or patient as intended to manage health conditions or symptoms. We developed and applied an NLP system that included NLP-assisted manual review to identify such documentation in encounter notes. Results: Medical cannabis use was documented for 16,684 (5.6%) of 299,597 outpatient encounters with routine screening for cannabis use among 203,489 patients seeing 1,274 clinicians. The validated NLP system identified 54% of documentation and NLP-assisted manual review the remainder. Language documenting reasons for cannabis use included 125 terms indicating medical use, 28 terms indicating non-medical use and 41 ambiguous terms. Implicit documentation of medical use (e.g., "edible THC nightly for lumbar pain") was more common than explicit (e.g., "continues medical cannabis use"). Conclusions: Clinicians use diverse and often ambiguous language to document patients' reasons for cannabis use. Automating extraction of documentation about patients' cannabis use could facilitate clinical decision support and epidemiological investigation but will require large amounts of gold standard training data.

Prevalence of Medical Cannabis Use and Associated Health Conditions Documented in Electronic Health Records Among Primary Care Patients in Washington State.

Importance: Many people use cannabis for medical reasons despite limited evidence of therapeutic benefit and potential risks. Little is known about medical practitioners' documentation of medical cannabis use or clinical characteristics of patients with documented medical cannabis use. Objectives: To estimate the prevalence of past-year medical cannabis use documented in electronic health records (EHRs) and to describe patients with EHR-documented medical cannabis use, EHR-documented cannabis use without evidence of medical use (other cannabis use), and no EHR-documented cannabis use. Design, Setting, and Participants: This cross-sectional study assessed adult primary care patients who completed a cannabis screen during a visit between November 1, 2017, and October 31, 2018, at a large health system that conducts routine cannabis screening in a US state with legal medical and recreational cannabis use. Exposures: Three mutually exclusive categories of EHR-documented cannabis use (medical, other, and no use) based on practitioner documentation of medical cannabis use in the EHR and patient report of past-year cannabis use at screening. Main Outcomes and Measures: Health conditions for which cannabis use has potential benefits or risks were defined based on National Academies of Sciences, Engineering, and Medicine's review. The adjusted prevalence of conditions diagnosed in the prior year were estimated across 3 categories of EHR-documented cannabis use with logistic regression. Results: A total of 185 565 patients (mean [SD] age, 52.0 [18.1] years; 59% female, 73% White, 94% non-Hispanic, and 61% commercially insured) were screened for cannabis use in a primary care visit during the study period. Among these patients, 3551 (2%) had EHR-documented medical cannabis use, 36 599 (20%) had EHR-documented other cannabis use, and 145 415 (78%) had no documented cannabis use. Patients with medical cannabis use had a higher prevalence of health conditions for which cannabis has potential benefits (49.8%; 95% CI, 48.3%-51.3%) compared with patients with other cannabis use (39.9%; 95% CI, 39.4%-40.3%) or no cannabis use (40.0%; 95% CI, 39.8%-40.2%). In addition, patients with medical cannabis use had a higher prevalence of health conditions for which cannabis has potential risks (60.7%; 95% CI, 59.0%-62.3%) compared with patients with other cannabis use (50.5%; 95% CI, 50.0%-51.0%) or no cannabis use (42.7%; 95% CI, 42.4%-42.9%). Conclusions and Relevance: In this cross-sectional study, primary care patients with documented medical cannabis use had a high prevalence of health conditions for which cannabis use has potential benefits, yet a higher prevalence of conditions with potential risks from cannabis use. These findings suggest that practitioners should be prepared to discuss potential risks and benefits of cannabis use with patients.

The Association between Opioid Discontinuation and Heroin Use: A Nested Case-Control Study.

BACKGROUND: Opioid prescribing guidelines recommend reducing or discontinuing opioids for chronic pain if harms of opioid treatment outweigh benefits. As opioid discontinuation becomes more prevalent, it is important to understand whether opioid discontinuation is associated with heroin use. In this study, we sought to assess the association between opioid discontinuation and heroin use documented in the medical record. METHODS: A matched nested case-control study was conducted in an integrated health plan and delivery system in Colorado. Patients receiving opioid therapy in the study period (January 2006-June 2018) were included. Opioid discontinuation was defined as ≥45 days with no opioids dispensed after initiating opioid therapy. The heroin use onset date represented the index date. Case patients were matched to up to 20 randomly selected patients without heroin use (control patients) by age, sex, calendar time, and time between initiating opioid therapy and the index date. Conditional logistic regression models estimated matched odds ratios (mOR) for the association between an opioid discontinuation prior to the index date and heroin use. RESULTS: Among 22,962 patients prescribed opioid therapy, 125 patients (0.54%) used heroin after initiating opioid therapy, of which 74 met criteria for inclusion in the analysis. The odds of opioid discontinuation were approximately two times higher in case patients (n = 74) than control patients (n = 1045; mOR = 2.19; 95% CI 1.27-3.78). CONCLUSIONS: Among patients prescribed chronic opioid therapy, the observed increased risk for heroin use associated with opioid discontinuation should be balanced with potential benefits.

Modifying and Evaluating the Opioid Overdose Knowledge Scale for Prescription Opioids: A Pilot Study of the Rx-OOKS.

OBJECTIVE: To develop a validated instrument that measures knowledge about prescription opioid overdose. METHODS: Within an integrated health care system, we adapted, piloted, and tested the reliability and predictive validity of a modified Opioid Overdose Knowledge Scale (OOKS) instrument specific to prescription opioids (Rx-OOKS) with a patient population prescribed long-term opioid therapy and potentially at risk of opioid overdose. We used an interdisciplinary team approach and patient interviews to adapt the instrument. We then piloted the survey on a patient sample and assessed it using Cronbach's alpha and logistic regression. RESULTS: Rx-OOKS (N = 56) resulted in a three-construct, 25-item instrument. Internal consistency was acceptable for the following constructs: "signs of an overdose" (10 items) at α = 0.851, "action to take with opioid overdose" (seven items) at α = 0.692, and "naloxone use knowledge" (eight items) at α = 0.729. One construct, "risks of an overdose" (three items), had an α of 0.365 and was subsequently eliminated from analysis due to poor performance. We conducted logistic regression to determine if any of the constructs was strongly associated with future naloxone receipt. Higher scores on "actions to take in an overdose" had nine times the odds of receiving naloxone (odds ratio [OR] = 9.00, 95% confidence interval [CI] = 1.42-57.12); higher "naloxone use knowledge" scores were 15.8 times more likely to receive naloxone than those with lower scores (OR = 15.83, 95% CI = 1.68-149.17). CONCLUSIONS: The Rx-OOKS survey instrument can reliably measure knowledge about prescription opioid overdose recognition and naloxone use. Further, knowledge about actions to take during an opioid overdose and naloxone use were associated with future receipt of naloxone.

Association Between Opioid Dose Variability and Opioid Overdose Among Adults Prescribed Long-term Opioid Therapy.

Importance: Attempts to discontinue opioid therapy to reduce the risk of overdose and adhere to prescribing guidelines may lead patients to be exposed to variability in opioid dosing. Such dose variability may increase the risk of opioid overdose even if therapy discontinuation is associated with a reduction in risk. Objective: To examine the association between opioid dose variability and opioid overdose. Design, Setting, and Participants: A nested case-control study was conducted in a large Colorado integrated health plan and delivery system from January 1, 2006, through June 30, 2018. Cohort members were individuals prescribed long-term opioid therapy. Exposures: Dose variability was defined as the SD of the milligrams of morphine equivalents across each patient's follow-up and categorized based on the quintile distribution of the SD in the cohort (0-5.3, 5.4-9.1, 9.2-14.6, 14.7-27.2, and >27.2 mg of morphine equivalents). Main Outcomes and Measures: Opioid overdose cases were identified using International Classification of Diseases, Ninth Revision and International Statistical Classification of Diseases and Related Health Problems, Tenth Revision codes. Each case patient with overdose was matched to up to 20 control patients using risk set sampling. Conditional logistic regression models were used to generate matched odds ratios and 95% CIs, adjusted for age, sex, race/ethnicity, drug or alcohol use disorder, tobacco use, benzodiazepine dispensings, medical comorbidities, mental health disorder, opioid dose, and opioid formulation. Results: In a cohort of 14 898 patients (mean [SD] age, 56.3 [16.0] years; 8988 [60.3%] female) prescribed long-term opioid therapy, 228 case patients with incident opioid overdose were matched to 3547 control patients. The mean (SD) duration of opioid therapy was 36.7 (33.7) months in case patients and 33.0 (30.9) months in control patients. High-dose variability (SD >27.2 mg of morphine equivalents) was associated with a significantly increased risk of overdose compared with low-dose variability (matched odds ratio, 3.32; 95% CI, 1.63-6.77) independent of opioid dose. Conclusions and Relevance: Variability in opioid dose may be a risk factor for opioid overdose, suggesting that practitioners should seek to minimize dose variability when managing long-term opioid therapy.

Patient characteristics associated with treatment initiation and engagement among individuals diagnosed with alcohol and other drug use disorders in emergency department and primary care settings.

Background: Treatment initiation and engagement rates for alcohol and other drug (AOD) use disorders differ depending on where the AOD use disorder was identified. Emergency department (ED) and primary care (PC) are 2 common settings where patients are identified; however, it is unknown whether characteristics of patients who initiate and engage in treatment differ between these settings. Methods: Patients identified with an AOD disorder in ED or PC settings were drawn from a larger study that examined Healthcare Effectiveness Data and Information Set (HEDIS) AOD treatment initiation and engagement measures across 7 health systems using electronic health record data (n = 54,321). Multivariable generalized linear models, with a logit link, clustered on health system, were used to model patient factors associated with initiation and engagement in treatment, between and within each setting. Results: Patients identified in the ED had higher odds of initiating treatment than those identified in PC (adjusted odds ratio [aOR] = 1.89, 95% confidence interval [CI] = 1.73-2.07), with no difference in engagement between the settings. Among those identified in the ED, compared with patients aged 18-29, older patients had higher odds of treatment initiation (age 30-49: aOR = 1.25, 95% CI = 1.12-1.40; age 50-64: aOR = 1.42, 95% CI = 1.26-1.60; age 65+: aOR = 1.27, 95% CI = 1.08-1.49). However, among those identified in PC, compared with patients aged 18-29, older patients were less likely to initiate (age 30-49: aOR = 0.81, 95% CI = 0.71-0.94; age 50-64: aOR = 0.68, 95% CI = 0.58-0.78; age 65+: aOR = 0.47, 95% CI = 0.40-0.56). Women identified in ED had lower odds of initiating treatment (aOR = 0.80, 95% CI = 0.72-0.88), whereas sex was not associated with treatment initiation in PC. In both settings, patients aged 65+ had lower odds of engaging compared with patients aged 18-29 (ED: aOR = 0.61, 95% CI = 0.38-0.98; PC: aOR = 0.42, 95% CI = 0.26-0.68). Conclusion: Initiation and engagement in treatment differed by sex and age depending on identification setting. This information could inform tailoring of future AOD interventions.

The prevalence of Healthcare Effectiveness Data and Information Set (HEDIS) initiation and engagement in treatment among patients with cannabis use disorders in 7 US health systems.

Background: Cannabis use disorders (CUDs) have increased with more individuals using cannabis, yet few receive treatment. Health systems have adopted the Healthcare Effectiveness Data and Information Set (HEDIS) quality measures of initiation and engagement in alcohol and other drug (AOD) dependence treatment, but little is known about the performance of these among patients with CUDs. Methods: This cohort study utilized electronic health records and claims data from 7 health care systems to identify patients with documentation of a new index CUD diagnosis (no AOD diagnosis ≤60 days prior) from International Classification of Diseases, Ninth revision, codes (October 1, 2014, to August 31, 2015). The adjusted prevalence of each outcome (initiation, engagement, and a composite of both) was estimated from generalized linear regression models, across index identification settings (inpatient, emergency department, primary care, addiction treatment, and mental health/psychiatry), AOD comorbidity (patients with CUD only and CUD plus other AOD diagnoses), and patient characteristics. Results: Among 15,202 patients with an index CUD diagnosis, 30.0% (95% confidence interval [CI]: 29.2-30.7%) initiated, 6.9% (95% CI: 6.2-7.7%) engaged among initiated, and 2.1% (95% CI: 1.9-2.3%) overall both initiated and engaged in treatment. The adjusted prevalence of outcomes varied across index identification settings and was highest among patients diagnosed in addiction treatment, with 25.0% (95% CI: 22.5-27.6%) initiated, 40.9% (95% CI: 34.8-47.0%) engaged, and 12.5% (95% CI: 10.0-15.1%) initiated and engaged. The adjusted prevalence of each outcome was generally highest among patients with CUD plus other AOD diagnosis at index diagnosis compared with those with CUD only, overall and across index identification settings, and was lowest among uninsured and older patients. Conclusion: Among patients with a new CUD diagnosis, the proportion meeting HEDIS criteria for initiation and/or engagement in AOD treatment was low and demonstrated variation across index diagnosis settings, AOD comorbidity, and patient characteristics, pointing to opportunities for improvement.

Prediction Model for Two-Year Risk of Opioid Overdose Among Patients Prescribed Chronic Opioid Therapy.

BACKGROUND: Naloxone is a life-saving opioid antagonist. Chronic pain guidelines recommend that physicians co-prescribe naloxone to patients at high risk for opioid overdose. However, clinical tools to efficiently identify patients who could benefit from naloxone are lacking. OBJECTIVE: To develop and validate an overdose predictive model which could be used in primary care settings to assess the need for naloxone. DESIGN: Retrospective cohort. SETTING: Derivation site was an integrated health system in Colorado; validation site was a safety-net health system in Colorado. PARTICIPANTS: We developed a predictive model in a cohort of 42,828 patients taking chronic opioid therapy and externally validated the model in 10,708 patients. MAIN MEASURES: Potential predictors and outcomes (nonfatal pharmaceutical and heroin overdoses) were extracted from electronic health records. Fatal overdose outcomes were identified from state vital records. To match the approximate shelf-life of naloxone, we used Cox proportional hazards regression to model the 2-year risk of overdose. Calibration and discrimination were assessed. KEY RESULTS: A five-variable predictive model showed good calibration and discrimination (bootstrap-corrected c-statistic = 0.73, 95% confidence interval [CI] 0.69-0.78) in the derivation site, with sensitivity of 66.1% and specificity of 66.6%. In the validation site, the model showed good discrimination (c-statistic = 0.75, 95% CI 0.70-0.80) and less than ideal calibration, with sensitivity and specificity of 82.2% and 49.5%, respectively. CONCLUSIONS: Among patients on chronic opioid therapy, the predictive model identified 66-82% of all subsequent opioid overdoses. This model is an efficient screening tool to identify patients who could benefit from naloxone to prevent overdose deaths. Population differences across the two sites limited calibration in the validation site.