Objectives: This study aimed to leverage FLAN-T5-Large, BERT, RoBERTa, and Gemma-2-2B, with fine-tuning, to identify instances of social isolation and social support within unstructured clinical notes. Materials and Methods: Annotated clinical note spans containing social context cues were used to fine-tune each model. Performance was evaluated using Accuracy, Precision, Recall, and Macro-F1 score. A structured prompt was used to instruct the model to perform classification task and mitigate overgeneralization. Performance comparisons across the models assessed sensitivity, robustness, and false positive reduction. Results: FLAN-T5-Large achieved highest performance, with Macro-F1 of 0.92{+/-}0.04, demonstrating balanced results across classes: social isolation (F1 = 0.91{+/-}0.03), no social isolation (F1 = 0.94{+/-}0.05), and social support (F1 = 0.90{+/-}0.04). Gemma-2-2B produced comparable results, with Macro-F1 score of 0.89{+/-}0.10. BERT and RoBERTa achieved lower Macro-F1 scores of 0.77{+/-}0.17 and 0.80{+/-}0.21 respectively, with variability across categories. Discussion: A major contribution of this work is precise identification of multiple concepts related to social connectedness. By integrating annotated examples of both true and false positives, including negations and contextually ambiguous terms, the model better distinguished relevant social context cues from noise. Training on both social isolation and support provided a dual framework for comparative analyses and patient stratification. Conclusion: Transformer-based NLP models, particularly FLAN-T5-Large, demonstrated potential for identifying social isolation and social support in clinical text. These findings support the use of generative AI techniques to enhance detection of social isolation from EHRs, advancing context-aware healthcare analytics.
L. Chinthala, C. Lemon, A. Shaban-Nejad et al.· medRxiv· 0 citations
COVID-19 has placed a monumental burden on the health care system globally. Although no longer a public health emergency, there is still a pressing need for effective treatments that prevent adverse outcomes associated with this disease. Nirmatrelvir/ritonavir (NMV-R) is a promising and potentially effective antiviral, which until recently was under emergency use authorization. Our objective was to evaluate the real-world effectiveness of NMV-R in preventing severe illness, hospitalization, death and long-COVID in a large nationwide cohort of outpatients with COVID-19.
Population-based retrospective cohort study of patients with a SARS-CoV-2 positive test or diagnosis (index) date between December 2021 and February 2023 within the National COVID Cohort Collaborative (N3C), with at least one risk factor for severe COVID-19, no evidence of contraindicated medical conditions or medication use, and no hospital or emergency department visit or death within 24 hours of eligibility. We emulated a sequence of target trials beginning on each of the first five days of diagnosis with COVID-19. We identified 921,034 eligible person-trials (each representing a patient’s eligibility at a given diagnosis day across sequential emulated trials), of which 77,449 were initiators and 846,585 were non-initiators of NMV-R treatment. NMV-R Initiators were matched to non-initiators in each trial. The marginal hazard ratio between initiators and non-initiators was estimated for four acute outcomes: severe illness, hospitalization or death, hospitalization, and death; and the post-COVID condition or long COVID.
Of 921,034 eligible “person-trials”, 74,449 were initiators and 846,585 were non-initiators of NMV-R treatment. Pooled across trials, the hazard for severe illness (HR: 0.76, 95% CI: 0.71 to 0.81), hospitalization or death (HR: 0.50, 95% CI: 0.44 to 0.57), hospitalization (sdHR: 0.52, 95% CI: 0.46 to 0.60), death (HR: 0.33, 95% CI: 0.21 to 0.51), and long-COVID (sdHR: 0.85, 95% CI: 0.77 to 0.95) were significantly lower among NMV-R initiators compared to non-initiators. Results further indicated larger associations between NMV-R and reduced risk of both acute and post-acute outcomes with early versus delayed NMV-R treatment initiation, and among unvaccinated versus vaccinated patient subgroups.
NMV-R is overall effective at preventing the risk of severe acute outcomes including hospitalization and death, as well as long COVID. Results were robust across multiple sensitivity considerations.
Not applicable.
Steve R Makkar, Kristen Hansen, Arjun S. Yadaw et al.· BMC Infectious Diseases· 0 citations
Summary Background Vaccination is a vital tool in preventing acute COVID-19 and may confer additional protection against Long COVID, although it is unclear whether this protection wanes over time. Methods We assessed electronic health record (EHR) data from a national, retrospective cohort of patients, comparing the 12-month cumulative incidence of Long COVID (ICD-10 code U09.9) among (A) patients who were vaccinated versus unvaccinated (two or more versus zero doses) and (B) patients diagnosed with acute COVID-19 1–3 months, 3–5 months, or 5–7 months after vaccination. Findings In our binary cohort (n = 519,980), we found that patients who were vaccinated had a lower risk of Long COVID (adjusted risk ratio 0.84 (0.81, 0.88)) or mortality (adjusted risk ratio 0.83 (0.81, 0.86)) than patients who were unvaccinated. In our longitudinal cohort (n = 1,085,291), we did not find significant heterogeneity in Long COVID risk during the seven months following vaccination. Interpretation We found that COVID-19 vaccination was protective against Long COVID, and we did not observe a significant waning of this protection within seven months after vaccination. Funding This research was financially supported by the 10.13039/100000060National Institute of Allergy and Infectious Diseases (1K01AI182501 to Zachary Butzin-Dozier) and a Global Development grant (OPP1165144) from the 10.13039/100000865Bill & Melinda Gates Foundation to the 10.13039/100005595University of California, Berkeley, CA, USA. Individual authors were supported by the following funding sources: 10.13039/100000025NIMHR01131542 (PI Rena C. Patel), Jerrod Anzalone is supported by the 10.13039/100000057National Institute of General Medical Sciences, U54 GM115458, which funds the Great Plains IDeA-CTR Network. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.
Z. Butzin-Dozier, Yunwen Ji, Lin-Chiun Wang et al.· EBioMedicine· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
