1
|
Garcia P, Ma SP, Shah S, Smith M, Jeong Y, Devon-Sand A, Tai-Seale M, Takazawa K, Clutter D, Vogt K, Lugtu C, Rojo M, Lin S, Shanafelt T, Pfeffer MA, Sharp C. Artificial Intelligence-Generated Draft Replies to Patient Inbox Messages. JAMA Netw Open 2024; 7:e243201. [PMID: 38506805 PMCID: PMC10955355 DOI: 10.1001/jamanetworkopen.2024.3201] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [MESH Headings] [Track Full Text] [Journal Information] [Submit a Manuscript] [Subscribe] [Scholar Register] [Indexed: 03/21/2024] Open
Abstract
Importance The emergence and promise of generative artificial intelligence (AI) represent a turning point for health care. Rigorous evaluation of generative AI deployment in clinical practice is needed to inform strategic decision-making. Objective To evaluate the implementation of a large language model used to draft responses to patient messages in the electronic inbox. Design, Setting, and Participants A 5-week, prospective, single-group quality improvement study was conducted from July 10 through August 13, 2023, at a single academic medical center (Stanford Health Care). All attending physicians, advanced practice practitioners, clinic nurses, and clinical pharmacists from the Divisions of Primary Care and Gastroenterology and Hepatology were enrolled in the pilot. Intervention Draft replies to patient portal messages generated by a Health Insurance Portability and Accountability Act-compliant electronic health record-integrated large language model. Main Outcomes and Measures The primary outcome was AI-generated draft reply utilization as a percentage of total patient message replies. Secondary outcomes included changes in time measures and clinician experience as assessed by survey. Results A total of 197 clinicians were enrolled in the pilot; 35 clinicians who were prepilot beta users, out of office, or not tied to a specific ambulatory clinic were excluded, leaving 162 clinicians included in the analysis. The survey analysis cohort consisted of 73 participants (45.1%) who completed both the presurvey and postsurvey. In gastroenterology and hepatology, there were 58 physicians and APPs and 10 nurses. In primary care, there were 83 physicians and APPs, 4 nurses, and 8 clinical pharmacists. The mean AI-generated draft response utilization rate across clinicians was 20%. There was no change in reply action time, write time, or read time between the prepilot and pilot periods. There were statistically significant reductions in the 4-item physician task load score derivative (mean [SD], 61.31 [17.23] presurvey vs 47.26 [17.11] postsurvey; paired difference, -13.87; 95% CI, -17.38 to -9.50; P < .001) and work exhaustion scores (mean [SD], 1.95 [0.79] presurvey vs 1.62 [0.68] postsurvey; paired difference, -0.33; 95% CI, -0.50 to -0.17; P < .001). Conclusions and Relevance In this quality improvement study of an early implementation of generative AI, there was notable adoption, usability, and improvement in assessments of burden and burnout. There was no improvement in time. Further code-to-bedside testing is needed to guide future development and organizational strategy.
Collapse
Affiliation(s)
- Patricia Garcia
- Department of Medicine, Stanford University School of Medicine, Stanford, California
| | - Stephen P Ma
- Department of Medicine, Stanford University School of Medicine, Stanford, California
| | - Shreya Shah
- Department of Medicine, Stanford University School of Medicine, Stanford, California
- Stanford Healthcare AI Applied Research Team, Division of Primary Care and Population Health, Stanford University School of Medicine, Stanford, California
| | - Margaret Smith
- Stanford Healthcare AI Applied Research Team, Division of Primary Care and Population Health, Stanford University School of Medicine, Stanford, California
| | - Yejin Jeong
- Stanford Healthcare AI Applied Research Team, Division of Primary Care and Population Health, Stanford University School of Medicine, Stanford, California
| | - Anna Devon-Sand
- Stanford Healthcare AI Applied Research Team, Division of Primary Care and Population Health, Stanford University School of Medicine, Stanford, California
| | - Ming Tai-Seale
- Department of Family Medicine, University of California San Diego School of Medicine, La Jolla
| | - Kevin Takazawa
- Technology and Digital Solutions, Stanford Medicine, Stanford, California
| | - Danyelle Clutter
- Technology and Digital Solutions, Stanford Medicine, Stanford, California
| | - Kyle Vogt
- Technology and Digital Solutions, Stanford Medicine, Stanford, California
| | - Carlene Lugtu
- Nursing Informatics & Innovation, Stanford Healthcare, Stanford, California
| | - Matthew Rojo
- Technology and Digital Solutions, Stanford Medicine, Stanford, California
| | - Steven Lin
- Department of Medicine, Stanford University School of Medicine, Stanford, California
- Stanford Healthcare AI Applied Research Team, Division of Primary Care and Population Health, Stanford University School of Medicine, Stanford, California
| | - Tait Shanafelt
- Department of Medicine, Stanford University School of Medicine, Stanford, California
- WellMD Center, Stanford University School of Medicine, Stanford, California
| | - Michael A Pfeffer
- Department of Medicine, Stanford University School of Medicine, Stanford, California
- Technology and Digital Solutions, Stanford Medicine, Stanford, California
| | - Christopher Sharp
- Department of Medicine, Stanford University School of Medicine, Stanford, California
| |
Collapse
|
2
|
Parent C, Martinez DA, Venkataramani M, Yang C, Page KR. Racial and Ethnic Disparities in Glycemic Control Among Patients With SARS-CoV-2 in the Baltimore-Washington, District of Columbia Region. AJPM FOCUS 2024; 3:100156. [PMID: 38149079 PMCID: PMC10749874 DOI: 10.1016/j.focus.2023.100156] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Key Words] [Grants] [Track Full Text] [Download PDF] [Subscribe] [Scholar Register] [Indexed: 12/28/2023]
Abstract
Introduction Diabetes is a leading risk factor for COVID-19, disproportionally impacting marginalized populations. We analyzed racial/ethnic differences in glycemic control among patients who tested positive for SARS-CoV-2 in the Baltimore-Washington, District of Columbia region. Methods Glycemic control measured by HbA1c was compared by race and ethnicity among patients with a positive SARS-CoV-2 test at the Johns Hopkins Health System between March 1, 2020, and March 31, 2022. Risk factors associated with poor glycemic control (HbA1c≥8) were identified using logistic regression. Results Black, Latino, and Asian patients had a higher rate of prediabetes (HbA1c=5.7%-6.49%) and diabetes (HbA1c≥6.5%) than non-Hispanic White patients. Among patients with diabetes, poor glycemic control (HbA1c≥8%) was significantly higher among young adults (aged ≤44 years), Latino patients (AOR=1.5; 95% CI=1.1, 1.9), Black patients (AOR=1.2; 95% CI=1.0, 1.5), uninsured patients (AOR=1.5; 95% CI=1.2, 1.9), and those with limited English proficiency (AOR=1.3; 95% CI=1.0, 1.6) or without a primary care physician (AOR=1.6; 95% CI=1.3, 2.1). Conclusions Disparities in glycemic control among patients who tested positive for SARS-CoV-2 were associated with underlying structural factors such as access to care, health insurance, and language proficiency. There is a need to implement accessible, culturally and language-appropriate preventive and primary care programs to engage socioeconomically disadvantaged populations in diabetic screening and care.
Collapse
Affiliation(s)
- Cassandra Parent
- Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland
| | - Diego A. Martinez
- School of Industrial Engineering, Pontificia Universidad Católica de Valparaíso, Valparaíso, Chile
- Department of Emergency Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland
| | - Maya Venkataramani
- Division of General Internal Medicine, Johns Hopkins School of Medicine, Baltimore, Maryland
| | - Cui Yang
- Department of Health Behavior, Society and Policy, Rutgers School of Public Health, Piscataway, New Jersey
| | - Kathleen R. Page
- Department of Medicine, Johns Hopkins University School of Medicine, Baltimore, Maryland
| |
Collapse
|
3
|
Reproducibility and implementation of a rapid, community-based COVID-19 "test and respond" model in low-income, majority-Latino communities in Northern California. PLoS One 2022; 17:e0276257. [PMID: 36301834 PMCID: PMC9612491 DOI: 10.1371/journal.pone.0276257] [Citation(s) in RCA: 0] [Impact Index Per Article: 0] [Reference Citation Analysis] [Abstract] [Track Full Text] [Download PDF] [Figures] [Journal Information] [Subscribe] [Scholar Register] [Received: 05/09/2022] [Accepted: 10/03/2022] [Indexed: 11/24/2022] Open
Abstract
OBJECTIVE To evaluate implementation of a community-engaged approach to scale up COVID-19 mass testing in low-income, majority-Latino communities. METHODS In January 2021, we formed a community-academic "Latino COVID-19 Collaborative" with residents, leaders, and community-based organizations (CBOs) from majority-Latinx, low-income communities in three California counties (Marin/Merced/San Francisco). The collaborative met monthly to discuss barriers/facilitators for COVID-19 testing, and plan mass testing events informed by San Francisco's Unidos en Salud "test and respond" model, offering community-based COVID-19 testing and post-test support in two US-census tracts: Canal (Marin) and Planada (Merced). We evaluated implementation using the RE-AIM framework. To further assess testing barriers, we surveyed a random sample of residents who did not attend the events. RESULTS Fifty-five residents and CBO staff participated in the Latino collaborative. Leading facilitators identified to increase testing were extended hours of community-based testing and financial support during isolation. In March-April 2021, 1,217 people attended mass-testing events over 13 days: COVID-19 positivity was 3% and 1% in Canal and Planada, respectively. The RE-AIM evaluation found: census tract testing coverage of 4.2% and 6.3%, respectively; 90% of event attendees were Latino, 89% had household income <$50,000/year, and 44% first-time testers (reach), effectiveness in diagnosing symptomatic cases early (median isolation time: 7 days) and asymptomatic COVID-19 (41% at diagnosis), high adoption by CBOs in both counties, implementation of rapid testing (median: 17.5 minutes) and disclosure, and post-event maintenance of community-based testing. Among 265 non-attendees surveyed, 114 (43%) reported they were aware of the event: reasons for non-attendance among the 114 were insufficient time (32%), inability to leave work (24%), and perceptions that testing was unnecessary post-vaccination (24%) or when asymptomatic (25%). CONCLUSION Community-engaged mass "test and respond" events offer a reproducible approach to rapidly increase COVID-19 testing access in low-income, Latinx communities.
Collapse
|