Sydney Virginia Lewis
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Sydney Virginia Lewis has a total of 34 co-authors. Here are the top co-authors by number of shared publications:
- Job Gideon Godino
Laura Rodriguez Research Institute, Family Health Centers of San Diego, San Diego, CA, United States
4 shared publications - Natalie Lauren Vawter
University of California San Diego, San Diego, CA, USA
3 shared publications - Adam Northrup
Laura Rodriguez Research Institute, Family Health Centers of San Diego, San Diego, CA, United States
2 shared publications - Earl Marc Felisme
Laura Rodriguez Research Institute, Family Health Centers of San Diego, San Diego, CA, United States
2 shared publications - Jacqueline Yareli Poblete
Laura Rodriguez Research Institute, Family Health Centers of San Diego, San Diego, CA, United States
2 shared publications - Jennifer Shea
Laura Rodriguez Research Institute, Family Health Centers of San Diego, San Diego, CA, United States
2 shared publications
Latest Publications
- Open Access
- 10 November 2026
- Published by JMIR Publications Inc. in Journal of Medical Internet Research
- Vol. 28, e81708
Labeling Clinical Notes With Large Language Models: Retrospective Benchmarking Study
Background: Clinical notes are rich sources of clinical information and are routinely used in clinical practice to infer past diagnoses and other comorbidities. Large language models (LLMs) may simplify the task by extracting this information from clinical notes. Objective: The study aims to evaluate the performance of state-of-the-art LLMs on the task of labeling English clinical notes. Methods: This institutional review board–approved retrospective study comprised 380 clinical notes obtained from 2 institutions in the United States. The notes were manually labeled for the presence or absence of 29 conditions (simplified Elixhauser Comorbidity Index) by 2 human readers—a radiology resident and a medical student under the supervision of a senior radiologist. Forty-six LLMs, including 4 closed-source commercial models from the GPT and Claude families, and 42 open-weights LLMs from the Llama (13 models) and the Qwen (29 models) families were used to label each note. The GPT-o3-mini-high reasoning model was used to parse the free-text model responses to binary yes/no answers. Using the manual labels as the reference standard, the recall, precision, and F 1 -scores for the models were evaluated. The effect of factors such as the length of the notes and the family, size, and release recency of the LLMs was analyzed using generalized linear models. Results: The top-performing LLMs included the GPT and Claude models, and the larger open-weights models included Llama 3 70B, Llama 3.1 70B, Llama 3.1 405B, Llama 3.3 70B, Qwen 1.5 72B, Qwen 1.5 110B, Qwen 2 72B, and Qwen 2 VL 72B, which achieved F 1 -scores above 70%. The most recent and higher-capacity releases of the LLMs have been improving mainly in their precision. The evaluation of 29 conditions showed a large variance in performance depending on the condition. Conclusions: In this evaluation with 46 LLMs, larger open-weights LLMs such as the Llama 3.3 70B performed on par with or better than the GPT and Claude models, achieving F 1 -scores above 70%. The performance of open-weights models improved significantly with model size and recency. These results suggest that locally run open-weights LLMs can be potentially used to label clinical notes.
- 10 November 2026
- Published by JMIR Publications Inc. in Journal of Medical Internet Research
- Vol. 28, e81708
Open AccessLabeling Clinical Notes With Large Language Models: Retrospective Benchmarking Study
Background: Clinical notes are rich sources of clinical information and are routinely used in clinical practice to infer past diagnoses and other comorbidities. Large language models (LLMs) may simplify the task by extracting this information from clinical notes. Objective: The study aims to evaluate the performance of state-of-the-art LLMs on the task of labeling English clinical notes. Methods: This institutional review board–approved retrospective study comprised 380 clinical notes obtained from 2 institutions in the United States. The notes were manually labeled for the presence or absence of 29 conditions (simplified Elixhauser Comorbidity Index) by 2 human readers—a radiology resident and a medical student under the supervision of a senior radiologist. Forty-six LLMs, including 4 closed-source commercial models from the GPT and Claude families, and 42 open-weights LLMs from the Llama (13 models) and the Qwen (29 models) families were used to label each note. The GPT-o3-mini-high reasoning model was used to parse the free-text model responses to binary yes/no answers. Using the manual labels as the reference standard, the recall, precision, and F 1 -scores for the models were evaluated. The effect of factors such as the length of the notes and the family, size, and release recency of the LLMs was analyzed using generalized linear models. Results: The top-performing LLMs included the GPT and Claude models, and the larger open-weights models included Llama 3 70B, Llama 3.1 70B, Llama 3.1 405B, Llama 3.3 70B, Qwen 1.5 72B, Qwen 1.5 110B, Qwen 2 72B, and Qwen 2 VL 72B, which achieved F 1 -scores above 70%. The most recent and higher-capacity releases of the LLMs have been improving mainly in their precision. The evaluation of 29 conditions showed a large variance in performance depending on the condition. Conclusions: In this evaluation with 46 LLMs, larger open-weights LLMs such as the Llama 3.3 70B performed on par with or better than the GPT and Claude models, achieving F 1 -scores above 70%. The performance of open-weights models improved significantly with model size and recency. These results suggest that locally run open-weights LLMs can be potentially used to label clinical notes.
- Open Access
- 22 May 2025
- Published by Oxford University Press (OUP) in Open Forum Infectious Diseases
- Vol. 12 (6) , ofaf302
Simplified and Integrated Hepatitis C Virus Testing and Treatment Algorithm for Unhoused People Who Inject Drugs
Background: Hepatitis C virus (HCV) infection disproportionately affects people who inject drugs (PWID), particularly people experiencing homelessness, who face significant barriers to testing and treatment. Simplified testing and treatment algorithms, including the use of mobile medical units (MMUs), may improve access to care for this population. This study evaluated the effectiveness of a simplified HCV testing and treatment algorithm with integrated care delivered via MMU to unhoused PWID in an urban United States setting. Methods: This pragmatic, randomized controlled trial compared a simplified MMU care model with usual care among 201 PWID who tested positive for HCV antibody. Patients were randomized into simplified MMU or usual care. The primary outcome was treatment initiation within 6 months. Secondary outcomes included treatment completion, sustained virologic response at 12 weeks (SVR12), and initiation of medication-assisted treatment for opioid use disorder. Results: Of 98 patients with confirmed HCV RNA, 33% in the MMU group initiated treatment compared to 24% in the usual care group (P = .4). Treatment completion and SVR12 rates were higher in the usual care group (16% vs 10% and 12% vs 4%, respectively), but differences were not statistically significant. MMUs facilitated faster treatment initiation but encountered challenges with patient retention and follow-up. Conclusions: While the MMU model demonstrated potential for improving rapid access to HCV treatment, overall treatment initiation, retention, and cure rates remained low. A critical need remains to develop approaches that more closely approximate a test-and-treat model to increase treatment initiation and completion and achieve elimination in unhoused PWID. Clinical Trials Registration: NCT04741750.
- 22 May 2025
- Published by Oxford University Press (OUP) in Open Forum Infectious Diseases
- Vol. 12 (6) , ofaf302
Open AccessSimplified and Integrated Hepatitis C Virus Testing and Treatment Algorithm for Unhoused People Who Inject Drugs
Background: Hepatitis C virus (HCV) infection disproportionately affects people who inject drugs (PWID), particularly people experiencing homelessness, who face significant barriers to testing and treatment. Simplified testing and treatment algorithms, including the use of mobile medical units (MMUs), may improve access to care for this population. This study evaluated the effectiveness of a simplified HCV testing and treatment algorithm with integrated care delivered via MMU to unhoused PWID in an urban United States setting. Methods: This pragmatic, randomized controlled trial compared a simplified MMU care model with usual care among 201 PWID who tested positive for HCV antibody. Patients were randomized into simplified MMU or usual care. The primary outcome was treatment initiation within 6 months. Secondary outcomes included treatment completion, sustained virologic response at 12 weeks (SVR12), and initiation of medication-assisted treatment for opioid use disorder. Results: Of 98 patients with confirmed HCV RNA, 33% in the MMU group initiated treatment compared to 24% in the usual care group (P = .4). Treatment completion and SVR12 rates were higher in the usual care group (16% vs 10% and 12% vs 4%, respectively), but differences were not statistically significant. MMUs facilitated faster treatment initiation but encountered challenges with patient retention and follow-up. Conclusions: While the MMU model demonstrated potential for improving rapid access to HCV treatment, overall treatment initiation, retention, and cure rates remained low. A critical need remains to develop approaches that more closely approximate a test-and-treat model to increase treatment initiation and completion and achieve elimination in unhoused PWID. Clinical Trials Registration: NCT04741750.
- Open Access
- 24 July 2023
- Published by JMIR Publications Inc. in JMIR Human Factors
- Vol. 10 (1) , e46313
Digitally Based Blood Pressure Self-Monitoring Program That Promotes Hypertension Self-Management and Health Education Among Patients With Low-Income: Usability Study
Background: According to evidence-based clinical guidelines, adults with hypertension are advised to self-monitor their blood pressure (BP) twice daily. Self-measured BP monitoring is a recommended strategy for improving hypertension management. Objective: We aimed to determine the feasibility and acceptability of a digitally based BP self-monitoring program that promotes hypertension self-management and health education among low-income patients. We hypothesized that the program would be highly feasible and acceptable and that at least 50% of the patients would use the monitor at the rate required for the reimbursement of the device’s cost (16 days of measurements in any 30-day period). Methods: Withings BPM Connect was deployed to patients at Family Health Centers of San Diego. Program elements included training, SMS text message reminders, and physician communication. Compliance, use, mean BP, and BP control status were calculated. A Kaplan-Meier time-to-event analysis was conducted to compare time to compliance between a strict definition (≥16 days in any rolling 30-day window) and a lenient definition (≥1 day per week for 4 consecutive weeks). A log-rank test was performed to determine whether the difference in time to compliance between the definitions was statistically significant. Mean systolic BP (SBP) and diastolic BP (DBP) before the intervention and after the intervention and mean change in SBP and DBP across patients were calculated. Paired sample t tests (2-tailed) were performed to assess the changes in SBP and DBP from before to after the intervention. Results: A total of 179 patients received the monitors. The mean changes in SBP and DBP from before to after the intervention were +2.62 (SE 1.26) mm Hg and +3.31 (SE 0.71) mm Hg, respectively. There was a statistically significant increase in both SBP and DBP after the intervention compared with before the intervention (P=.04 and P<.001). At the first and last measurements, 37.5% (63/168) and 48.8% (82/168) of the patients had controlled BP, respectively. During the observation period, 83.3% (140/168) of the patients had at least 1 controlled BP measurement. Use decreased over time, with 53.6% (90/168) of the patients using their monitor at week 2 and only 25% (42/168) at week 11. Although only 25.6% (43/168) achieved the strict definition of compliance, 42.3% (71/168) achieved the lenient definition of compliance. The median time to compliance was 130 days for the strict definition and 95 days for the lenient definition. The log-rank test showed a statistically significant difference in time to compliance between the compliance definitions (P<.001). Only 26.8% (45/168) complied with the measurement rate that would result in device cost reimbursement. Conclusions: Few patients used the monitors at a rate that would result in reimbursement, raising financial feasibility concerns. Plans for sustaining costs among low-income patients need to be further evaluated.
- 24 July 2023
- Published by JMIR Publications Inc. in JMIR Human Factors
- Vol. 10 (1) , e46313
Open AccessDigitally Based Blood Pressure Self-Monitoring Program That Promotes Hypertension Self-Management and Health Education Among Patients With Low-Income: Usability Study
Background: According to evidence-based clinical guidelines, adults with hypertension are advised to self-monitor their blood pressure (BP) twice daily. Self-measured BP monitoring is a recommended strategy for improving hypertension management. Objective: We aimed to determine the feasibility and acceptability of a digitally based BP self-monitoring program that promotes hypertension self-management and health education among low-income patients. We hypothesized that the program would be highly feasible and acceptable and that at least 50% of the patients would use the monitor at the rate required for the reimbursement of the device’s cost (16 days of measurements in any 30-day period). Methods: Withings BPM Connect was deployed to patients at Family Health Centers of San Diego. Program elements included training, SMS text message reminders, and physician communication. Compliance, use, mean BP, and BP control status were calculated. A Kaplan-Meier time-to-event analysis was conducted to compare time to compliance between a strict definition (≥16 days in any rolling 30-day window) and a lenient definition (≥1 day per week for 4 consecutive weeks). A log-rank test was performed to determine whether the difference in time to compliance between the definitions was statistically significant. Mean systolic BP (SBP) and diastolic BP (DBP) before the intervention and after the intervention and mean change in SBP and DBP across patients were calculated. Paired sample t tests (2-tailed) were performed to assess the changes in SBP and DBP from before to after the intervention. Results: A total of 179 patients received the monitors. The mean changes in SBP and DBP from before to after the intervention were +2.62 (SE 1.26) mm Hg and +3.31 (SE 0.71) mm Hg, respectively. There was a statistically significant increase in both SBP and DBP after the intervention compared with before the intervention (P=.04 and P<.001). At the first and last measurements, 37.5% (63/168) and 48.8% (82/168) of the patients had controlled BP, respectively. During the observation period, 83.3% (140/168) of the patients had at least 1 controlled BP measurement. Use decreased over time, with 53.6% (90/168) of the patients using their monitor at week 2 and only 25% (42/168) at week 11. Although only 25.6% (43/168) achieved the strict definition of compliance, 42.3% (71/168) achieved the lenient definition of compliance. The median time to compliance was 130 days for the strict definition and 95 days for the lenient definition. The log-rank test showed a statistically significant difference in time to compliance between the compliance definitions (P<.001). Only 26.8% (45/168) complied with the measurement rate that would result in device cost reimbursement. Conclusions: Few patients used the monitors at a rate that would result in reimbursement, raising financial feasibility concerns. Plans for sustaining costs among low-income patients need to be further evaluated.
- Open Access
- 9 March 2023
- Published by JMIR Publications Inc. in JMIR Preprints
Digitally Based Blood Pressure Self-Monitoring Program That Promotes Hypertension Self-Management and Health Education Among Patients With Low-Income: Usability Study (Preprint)
BACKGROUND According to evidence-based clinical guidelines, adults with hypertension are advised to self-monitor their blood pressure (BP) twice daily. Self-measured BP monitoring is a recommended strategy for improving hypertension management. OBJECTIVE We aimed to determine the feasibility and acceptability of a digitally based BP self-monitoring program that promotes hypertension self-management and health education among low-income patients. We hypothesized that the program would be highly feasible and acceptable and that at least 50% of the patients would use the monitor at the rate required for the reimbursement of the device’s cost (16 days of measurements in any 30-day period). METHODS Withings BPM Connect was deployed to patients at Family Health Centers of San Diego. Program elements included training, SMS text message reminders, and physician communication. Compliance, use, mean BP, and BP control status were calculated. A Kaplan-Meier time-to-event analysis was conducted to compare time to compliance between a strict definition (≥16 days in any rolling 30-day window) and a lenient definition (≥1 day per week for 4 consecutive weeks). A log-rank test was performed to determine whether the difference in time to compliance between the definitions was statistically significant. Mean systolic BP (SBP) and diastolic BP (DBP) before the intervention and after the intervention and mean change in SBP and DBP across patients were calculated. Paired sample t tests (2-tailed) were performed to assess the changes in SBP and DBP from before to after the intervention. RESULTS A total of 179 patients received the monitors. The mean changes in SBP and DBP from before to after the intervention were +2.62 (SE 1.26) mm Hg and +3.31 (SE 0.71) mm Hg, respectively. There was a statistically significant increase in both SBP and DBP after the intervention compared with before the intervention (P=.04 and P<.001). At the first and last measurements, 37.5% (63/168) and 48.8% (82/168) of the patients had controlled BP, respectively. During the observation period, 83.3% (140/168) of the patients had at least 1 controlled BP measurement. Use decreased over time, with 53.6% (90/168) of the patients using their monitor at week 2 and only 25% (42/168) at week 11. Although only 25.6% (43/168) achieved the strict definition of compliance, 42.3% (71/168) achieved the lenient definition of compliance. The median time to compliance was 130 days for the strict definition and 95 days for the lenient definition. The log-rank test showed a statistically significant difference in time to compliance between the compliance definitions (P<.001). Only 26.8% (45/168) complied with the measurement rate that would result in device cost reimbursement. CONCLUSIONS Few patients used the monitors at a rate that would result in reimbursement, raising financial feasibility concerns. Plans for sustaining costs among low-income patients need to be further evaluated.
- 9 March 2023
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessDigitally Based Blood Pressure Self-Monitoring Program That Promotes Hypertension Self-Management and Health Education Among Patients With Low-Income: Usability Study (Preprint)
BACKGROUND According to evidence-based clinical guidelines, adults with hypertension are advised to self-monitor their blood pressure (BP) twice daily. Self-measured BP monitoring is a recommended strategy for improving hypertension management. OBJECTIVE We aimed to determine the feasibility and acceptability of a digitally based BP self-monitoring program that promotes hypertension self-management and health education among low-income patients. We hypothesized that the program would be highly feasible and acceptable and that at least 50% of the patients would use the monitor at the rate required for the reimbursement of the device’s cost (16 days of measurements in any 30-day period). METHODS Withings BPM Connect was deployed to patients at Family Health Centers of San Diego. Program elements included training, SMS text message reminders, and physician communication. Compliance, use, mean BP, and BP control status were calculated. A Kaplan-Meier time-to-event analysis was conducted to compare time to compliance between a strict definition (≥16 days in any rolling 30-day window) and a lenient definition (≥1 day per week for 4 consecutive weeks). A log-rank test was performed to determine whether the difference in time to compliance between the definitions was statistically significant. Mean systolic BP (SBP) and diastolic BP (DBP) before the intervention and after the intervention and mean change in SBP and DBP across patients were calculated. Paired sample t tests (2-tailed) were performed to assess the changes in SBP and DBP from before to after the intervention. RESULTS A total of 179 patients received the monitors. The mean changes in SBP and DBP from before to after the intervention were +2.62 (SE 1.26) mm Hg and +3.31 (SE 0.71) mm Hg, respectively. There was a statistically significant increase in both SBP and DBP after the intervention compared with before the intervention (P=.04 and P<.001). At the first and last measurements, 37.5% (63/168) and 48.8% (82/168) of the patients had controlled BP, respectively. During the observation period, 83.3% (140/168) of the patients had at least 1 controlled BP measurement. Use decreased over time, with 53.6% (90/168) of the patients using their monitor at week 2 and only 25% (42/168) at week 11. Although only 25.6% (43/168) achieved the strict definition of compliance, 42.3% (71/168) achieved the lenient definition of compliance. The median time to compliance was 130 days for the strict definition and 95 days for the lenient definition. The log-rank test showed a statistically significant difference in time to compliance between the compliance definitions (P<.001). Only 26.8% (45/168) complied with the measurement rate that would result in device cost reimbursement. CONCLUSIONS Few patients used the monitors at a rate that would result in reimbursement, raising financial feasibility concerns. Plans for sustaining costs among low-income patients need to be further evaluated.
- 7 November 2022
- Published by Elsevier in Annals of Epidemiology
- Vol. 75, 77
A1C control during the pandemic among patients with type-2 diabetes
- 7 November 2022
- Published by Elsevier in Annals of Epidemiology
- Vol. 75, 77
A1C control during the pandemic among patients with type-2 diabetes