Fine-Tuning Transformers for Identifying Self-Reporting Potential Cases and Symptoms of COVID-19 in Tweets

We describe our straight-forward approach for Tasks 5 and 6 of 2021 Social Media Mining for Health Applications (SMM4H) shared tasks. Our system is based on fine-tuning DistillBERT on each task, as well as first fine-tuning the model on the other task. We explore how much fine-tuning is necessary for accurately classifying tweets as containing self-reported COVID-19 symptoms (Task 5) or whether a tweet related to COVID-19 is self-reporting, non-personal reporting, or a literature/news mention of the virus (Task 6). © 2021 Association for Computational Linguistics.

The impact of diversity, equity, and inclusion training in an independent community oncology practice

Background: Historically, Diversity, Equity, and Inclusion (DEI) efforts have largely been left out of private practice medical facilities - mainly present in large academic hospital centers. Virginia Oncology Associates (VOA), a member of The US Oncology Network, is an independent community oncology and hematology practice specializing in the diagnosis and treatment of cancer and blood disorders. VOA has approximately 600 total employees consisting of physicians and clinical and non-clinical staff. In August of 2020, VOA launched an initiative to foster a culture of inclusion with the creation of its Inclusion Council (IC). Sixteen employees', both clinical (3 physicians) and non-clinical, were chosen to participate. The council reports to VOA's joint policy board. Methods: The council partnered with an outside organization, Virginia's Center for Inclusive Communities, to launch a practice-wide DEI training program focusing on unconscious bias and microaggression. The training was initially planned to be in person, but due to the constraints of the COVID pandemic, the training was performed using a virtual platform. The members of IC attended three two-hour sessions while other staff members and physicians were mandated to participate in at least one training session. After completion of the sessions, a survey was sent to all employees and physicians to measure the impact of DEI training. All employees were also given the opportunity to provide additional, anonymous, written feedback. Results: Table. A total of 169 employees responded. 72% of respondents agreed or strongly agreed that the program increased awareness of unconscious bias and microaggression, 67% felt that the program helps foster a culture of inclusion in the workplace, and 66% of respondents felt that the program met expectations. Conclusions: DEI efforts are vital in all aspects of health care delivery and oncology settings. DEI training met staff expectations and positively fostered a culture of inclusion by bringing attention to unconscious bias and microaggression in a community oncology practice.

PCR109 Improving ACEI/ARB Adherence Among Nonadherent Patients Using Motivational Interviewing Intervention

Objectives: Angiotensin-converting enzyme inhibitors (ACEIs)/angiotensin II receptor blockers (ARBs) are recommended for patients with comorbid diabetes mellitus and hypertension due to their favorable impact on microvascular and macrovascular events, cardiovascular outcomes, and mortality. Poor adherence to ACEI/ARB is a major public health concern. Motivational interviewing (MI) intervention is an effective patient-centered approach to improve adherence. The objective of this study was to evaluate the effectiveness of a telephonic MI intervention conducted by pharmacy students, tailored by the past ACEI/ARB adherence trajectories among nonadherent patients with comorbid diabetes mellitus and hypertension. Methods: Patients continuously enrolled in a Medicare Advantage Plan from July 2017-July 2021 with an ACEI/ARB prescription between July 2017-December 2017 were identified. Group-based trajectory modeling (GBTM) was used to identify 4 distinct patterns of ACEI/ARB adherence: adherent, gaps in adherence, gradual decline, and rapid decline in adherence. Patients from the 3 nonadherent trajectories were randomized either into MI intervention or control group. The intervention group received an initial call followed by 5 follow-up calls. The primary outcome was adherence measured as proportion of days covered (PDC) ≥ 0.80 for 1-year post-MI implementation. Multivariable logistic regression model evaluated the effect of the intervention on ACEI/ARB adherence. Results: This study included 240 patients who received intervention and 480 randomly selected controls. Patients who had received ≥ 4 calls were more likely to be adherent than those who received ≤ 3 follow-up calls (OR=2.01;P=0.01) despite the gaps in follow-up calls due to the COVID-19 outbreak. Other significant predictors of adherence were baseline adherence trajectories, number of other medications on the index date, prevalent users, regimen complexity, and CMS risk score. Conclusions: At 1 year, patients who received ≥ 4 calls had significantly better adherence than controls. MI intervention by pharmacy students has been demonstrated to improve adherence on a long-term basis.

Real-World Dispensing of Buprenorphine and Methadone from MAR 2019 to MAR 2021 in California

Objectives: The US is amid a national opioid crisis before and during the COVID-19 pandemic. The Food and Drug Administration has approved methadone, buprenorphine, and naltrexone as medications for opioid use disorder (MOUD). This study examined the real-world dispensing of MOUD. Methods: All dispensing pharmacies, clinics, or other dispensers of Schedule II-V controlled substances in California report to the Controlled Substance Utilization Review and Evaluation System (CURES) on the day of prescriptions refills. Leveraging the data of buprenorphine (schedule III) and methadone (Schedule II) prescriptions from Mar 2019-Mar 2021 employing California’s deidentified CURES database, this study examined real-world dispensing of methadone and buprenorphine before (03/19/2019-03/18/2020) and during the pandemic (03/19/2020-03/18/2021). We did not review naltrexone dispensing, which is not a controlled substance. Results: In Mar 2019-Mar 2021, 182,367 patients≥18 in California obtained 875,051 buprenorphine and methadone prescriptions: Before the pandemic, there were 482,965 MOUD prescriptions dispensed to 116,644 patients;since the pandemic, 97,887 patients received 392,086 prescriptions, of which 32,164 patients(as “non-naïve” patients) started their MOUD before Mar 2020. On average, patients refilled their prescriptions 4.1 times/year before the pandemic and 4.0 times/year since the pandemic. The MOUD non-naïve patients (n=32,164) received 8.1 prescriptions/year before Mar 2020 and 7.4 refills/year afterward. The MOUD medications most widely prescribed in Mar 2019-Mar 2021 were buprenorphine (473,206 (98.0%) and 383,297 (97.8%), respectively, before and after the pandemic), which included 802,936 counts of buprenorphine alone and 53,567 combination medications of buprenorphine and naloxone. The number of methadone prescriptions declined from 9,759 before Mar 2020 to 8,789 during the pandemic. Conclusions: Buprenorphine is the leading MOUD prescribed for patients in California. The decline in MOUD dispensing for non-naïve patients may indicate restricted access to medication-assisted treatment under the pandemic. Policymakers should maintain or modify the policy strategies to help support medication access.