Sara G. Murray
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Sara G. Murray has a total of 39 co-authors. Here are the top co-authors by number of shared publications:
Gabriela Schmajuk
Department of Medicine, University of California, San Francisco, California, USA
4 shared publicationsJinoos Yazdany
Department of Medicine, University of California, San Francisco, California, USA
4 shared publicationsLaura Trupin
Department of Medicine, University of California San Francisco, San Francisco, California, United States of America
3 shared publicationsEdward H. Yelin
University of California San Francisco and Phillip R. Lee Institute for Health Policy Studies San Francisco California
2 shared publicationsPatricia P. Katz
University of California San Francisco and Phillip R. Lee Institute for Health Policy Studies San Francisco California
2 shared publicationsAlvin Rajkomar
Division of Hospital Medicine, Department of Medicine, University of California, San Francisco
1 shared publication
Latest Publications
- Open Access
Patient Safety Risks from AI Scribes: Signals from End-User Feedback
AI scribes are transforming clinical documentation at scale. However, their real-world performance remains understudied, especially regarding their impacts on patient safety. To this end, we initiate a mixed-methods study of patient safety issues raised in feedback submitted by AI scribe users (healthcare providers) in a large U.S. hospital system. Both quantitative and qualitative analysis suggest that AI scribes may induce various patient safety risks due to errors in transcription, most significantly regarding medication and treatment; however, further study is needed to contextualize the absolute degree of risk.
Open AccessPatient Safety Risks from AI Scribes: Signals from End-User Feedback
AI scribes are transforming clinical documentation at scale. However, their real-world performance remains understudied, especially regarding their impacts on patient safety. To this end, we initiate a mixed-methods study of patient safety issues raised in feedback submitted by AI scribe users (healthcare providers) in a large U.S. hospital system. Both quantitative and qualitative analysis suggest that AI scribes may induce various patient safety risks due to errors in transcription, most significantly regarding medication and treatment; however, further study is needed to contextualize the absolute degree of risk.
- Open Access
- 11 January 2024
- Published by Elsevier in Academic Pathology
- Vol. 11 (1) , 100101
Establishing priorities for implementation of large language models in pathology and laboratory medicine
- 11 January 2024
- Published by Elsevier in Academic Pathology
- Vol. 11 (1) , 100101
Open AccessEstablishing priorities for implementation of large language models in pathology and laboratory medicine
- 16 April 2021
- Published by Wiley in Journal of Hospital Medicine
- Vol. 16 (7) , 404-408
Automating Measurement of Trainee Work Hours
BACKGROUND: Medical training programs across the country are bound to a set of work hour regulations, generally monitored via self-report. OBJECTIVE: We developed a computational method to automate measurement of intern and resident work hours, which we validated against self-report. DESIGN, SETTING, AND PARTICIPANTS: We included all electronic health record (EHR) access log data between July 1, 2018, and June 30, 2019, for trainees enrolled in the internal medicine training program. We inferred the duration of continuous in-hospital work hours by linking EHR sessions that occurred within 5 hours as “on-campus” work and further accounted for “out-of-hospital” work which might be taking place at home. MAIN OUTCOMES AND MEASURES: We compared daily work hours estimated through the computational method with self-report and calculated the mean absolute error between the two groups. We used the computational method to estimate average weekly work hours across the rotation and the percentage of rotations where average work hours exceed the 80-hour workweek. RESULTS: The mean absolute error between self-reported and EHR-derived daily work hours for first- (PGY-1), second- (PGY-2), and third- (PGY-3) year trainees were 1.27, 1.51, and 1.51 hours, respectively. Using this computational method, we estimated average (SD) weekly work hours of 57.0 (21.7), 69.9 (12.2), and 64.1 (16.3) for PGY-1, PGY-2, and PGY-3 residents. CONCLUSION: EHR log data can be used to accurately approximate self-report of work hours, accounting for both in-hospital and out-of-hospital work. Automation will reduce trainees’ clerical work, improve consistency and comparability of data, and provide more complete and timely data that training programs need.
- 16 April 2021
- Published by Wiley in Journal of Hospital Medicine
- Vol. 16 (7) , 404-408
Automating Measurement of Trainee Work Hours
BACKGROUND: Medical training programs across the country are bound to a set of work hour regulations, generally monitored via self-report. OBJECTIVE: We developed a computational method to automate measurement of intern and resident work hours, which we validated against self-report. DESIGN, SETTING, AND PARTICIPANTS: We included all electronic health record (EHR) access log data between July 1, 2018, and June 30, 2019, for trainees enrolled in the internal medicine training program. We inferred the duration of continuous in-hospital work hours by linking EHR sessions that occurred within 5 hours as “on-campus” work and further accounted for “out-of-hospital” work which might be taking place at home. MAIN OUTCOMES AND MEASURES: We compared daily work hours estimated through the computational method with self-report and calculated the mean absolute error between the two groups. We used the computational method to estimate average weekly work hours across the rotation and the percentage of rotations where average work hours exceed the 80-hour workweek. RESULTS: The mean absolute error between self-reported and EHR-derived daily work hours for first- (PGY-1), second- (PGY-2), and third- (PGY-3) year trainees were 1.27, 1.51, and 1.51 hours, respectively. Using this computational method, we estimated average (SD) weekly work hours of 57.0 (21.7), 69.9 (12.2), and 64.1 (16.3) for PGY-1, PGY-2, and PGY-3 residents. CONCLUSION: EHR log data can be used to accurately approximate self-report of work hours, accounting for both in-hospital and out-of-hospital work. Automation will reduce trainees’ clerical work, improve consistency and comparability of data, and provide more complete and timely data that training programs need.
- Open Access
- 22 November 2018
- Published by Oxford University Press (OUP) in Journal of the American Medical Informatics Association
- Vol. 26 (1) , 61-65
Automated and flexible identification of complex disease: building a model for systemic lupus erythematosus using noisy labeling
Accurate and efficient identification of complex chronic conditions in the electronic health record (EHR) is an important but challenging task that has historically relied on tedious clinician review and oversimplification of the disease. Here we adapt methods that allow for automated “noisy labeling” of positive and negative controls to create a “silver standard” for machine learning to automate identification of systemic lupus erythematosus (SLE). Our final model, which includes both structured data as well as text processing of clinical notes, outperformed all existing algorithms for SLE (AUC 0.97). In addition, we demonstrate how the probabilistic outputs of this model can be adapted to various clinical needs, selecting high thresholds when specificity is the priority and lower thresholds when a more inclusive patient population is desired. Deploying a similar methodology to other complex diseases has the potential to dramatically simplify the landscape of population identification in the EHR. MeSH terms: Electronic Health Records, Machine Learning, Lupus Erythematosus, Phenotype, Algorithms
- 22 November 2018
- Published by Oxford University Press (OUP) in Journal of the American Medical Informatics Association
- Vol. 26 (1) , 61-65
Open AccessAutomated and flexible identification of complex disease: building a model for systemic lupus erythematosus using noisy labeling
Accurate and efficient identification of complex chronic conditions in the electronic health record (EHR) is an important but challenging task that has historically relied on tedious clinician review and oversimplification of the disease. Here we adapt methods that allow for automated “noisy labeling” of positive and negative controls to create a “silver standard” for machine learning to automate identification of systemic lupus erythematosus (SLE). Our final model, which includes both structured data as well as text processing of clinical notes, outperformed all existing algorithms for SLE (AUC 0.97). In addition, we demonstrate how the probabilistic outputs of this model can be adapted to various clinical needs, selecting high thresholds when specificity is the priority and lower thresholds when a more inclusive patient population is desired. Deploying a similar methodology to other complex diseases has the potential to dramatically simplify the landscape of population identification in the EHR. MeSH terms: Electronic Health Records, Machine Learning, Lupus Erythematosus, Phenotype, Algorithms
- Open Access
- 1 December 2017
- Published by American Medical Association (AMA) in JAMA Internal Medicine
- Vol. 177 (12) , 1863-1865
Using Spatial and Temporal Mapping to Identify Nosocomial Disease Transmission of Clostridium difficile
Hospital-acquired Clostridium difficile infection (CDI) is associated with significant morbidity and mortality.1 While risk factors like antibiotic exposure modulate susceptibility, infection control efforts aimed at reducing contact with infectious spores are critical to prevent nosocomial transmission.2-5 During hospitalization, patients visit many procedural and diagnostic common areas, presenting opportunities for contact with contaminated surfaces. However, these potential exposures are not typically captured in analyses evaluating disease transmission.6 Electronic health record (EHR) data allow us to track patients in time and space, but these data are not typically leveraged for infection control quality improvement efforts. We evaluated whether using a room within 24 hours of a patient with CDI was associated with increased risk of CDI in specific areas across our hospital.
- 1 December 2017
- Published by American Medical Association (AMA) in JAMA Internal Medicine
- Vol. 177 (12) , 1863-1865
Open AccessUsing Spatial and Temporal Mapping to Identify Nosocomial Disease Transmission of Clostridium difficile
Hospital-acquired Clostridium difficile infection (CDI) is associated with significant morbidity and mortality.1 While risk factors like antibiotic exposure modulate susceptibility, infection control efforts aimed at reducing contact with infectious spores are critical to prevent nosocomial transmission.2-5 During hospitalization, patients visit many procedural and diagnostic common areas, presenting opportunities for contact with contaminated surfaces. However, these potential exposures are not typically captured in analyses evaluating disease transmission.6 Electronic health record (EHR) data allow us to track patients in time and space, but these data are not typically leveraged for infection control quality improvement efforts. We evaluated whether using a room within 24 hours of a patient with CDI was associated with increased risk of CDI in specific areas across our hospital.
- Open Access
- 5 January 2016
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 11 (1) , e0144918
National Lupus Hospitalization Trends Reveal Rising Rates of Herpes Zoster and Declines in Pneumocystis Pneumonia
Infection is a leading cause of morbidity and mortality in systemic lupus erythematosus (SLE). Therapeutic practices have evolved over the past 15 years, but effects on infectious complications of SLE are unknown. We evaluated trends in hospitalizations for severe and opportunistic infections in a population-based SLE study. Data derive from the 2000 to 2011 United States National Inpatient Sample, including individuals who met a validated administrative definition of SLE. Primary outcomes were diagnoses of bacteremia, pneumonia, opportunistic fungal infection, herpes zoster, cytomegalovirus, or pneumocystis pneumonia (PCP). We used Poisson regression to determine whether infection rates were changing in SLE hospitalizations and used predictive marginals to generate annual adjusted rates of specific infections. We identified 361,337 SLE hospitalizations from 2000 to 2011 meeting study inclusion criteria. Compared to non-SLE hospitalizations, SLE patients were younger (51 vs. 62 years), predominantly female (89% vs. 54%), and more likely to be racial/ethnic minorities. SLE diagnosis was significantly associated with all measured severe and opportunistic infections. From 2000 to 2011, adjusted SLE hospitalization rates for herpes zoster increased more than non-SLE rates: 54 to 79 per 10,000 SLE hospitalizations compared with 24 to 29 per 10,000 non-SLE hospitalizations. Conversely, SLE hospitalizations for PCP disproportionately decreased: 5.1 to 2.5 per 10,000 SLE hospitalizations compared with 0.9 to 1.3 per 10,000 non-SLE hospitalizations. Among patients with SLE, herpes zoster hospitalizations are rising while PCP hospitalizations are declining. These trends likely reflect evolving SLE treatment strategies. Further research is needed to identify patients at greatest risk for infectious complications.
- 5 January 2016
- Published by Public Library of Science (PLoS) in PLOS ONE
- Vol. 11 (1) , e0144918
Open AccessNational Lupus Hospitalization Trends Reveal Rising Rates of Herpes Zoster and Declines in Pneumocystis Pneumonia
Infection is a leading cause of morbidity and mortality in systemic lupus erythematosus (SLE). Therapeutic practices have evolved over the past 15 years, but effects on infectious complications of SLE are unknown. We evaluated trends in hospitalizations for severe and opportunistic infections in a population-based SLE study. Data derive from the 2000 to 2011 United States National Inpatient Sample, including individuals who met a validated administrative definition of SLE. Primary outcomes were diagnoses of bacteremia, pneumonia, opportunistic fungal infection, herpes zoster, cytomegalovirus, or pneumocystis pneumonia (PCP). We used Poisson regression to determine whether infection rates were changing in SLE hospitalizations and used predictive marginals to generate annual adjusted rates of specific infections. We identified 361,337 SLE hospitalizations from 2000 to 2011 meeting study inclusion criteria. Compared to non-SLE hospitalizations, SLE patients were younger (51 vs. 62 years), predominantly female (89% vs. 54%), and more likely to be racial/ethnic minorities. SLE diagnosis was significantly associated with all measured severe and opportunistic infections. From 2000 to 2011, adjusted SLE hospitalization rates for herpes zoster increased more than non-SLE rates: 54 to 79 per 10,000 SLE hospitalizations compared with 24 to 29 per 10,000 non-SLE hospitalizations. Conversely, SLE hospitalizations for PCP disproportionately decreased: 5.1 to 2.5 per 10,000 SLE hospitalizations compared with 0.9 to 1.3 per 10,000 non-SLE hospitalizations. Among patients with SLE, herpes zoster hospitalizations are rising while PCP hospitalizations are declining. These trends likely reflect evolving SLE treatment strategies. Further research is needed to identify patients at greatest risk for infectious complications.
- 5 November 2015
- Published by Massachusetts Medical Society in New England Journal of Medicine
- Vol. 373 (19) , 1865-1870
A Breakthrough Diagnosis
A 19-year-old woman with multiple sclerosis presented with altered mental status, respiratory distress, and abdominal distention after a flight from Jordan to the United States.
- 5 November 2015
- Published by Massachusetts Medical Society in New England Journal of Medicine
- Vol. 373 (19) , 1865-1870
A Breakthrough Diagnosis
A 19-year-old woman with multiple sclerosis presented with altered mental status, respiratory distress, and abdominal distention after a flight from Jordan to the United States.
- Open Access
- 20 October 2014
- Published by Wiley in Arthritis Care & Research
- Vol. 67 (5) , 673-680
A Population‐Based Study of Infection‐Related Hospital Mortality in Patients With Dermatomyositis/Polymyositis
Objective: Dermatomyositis (DM) and polymyositis (PM) are debilitating inflammatory myopathies associated with significant mortality. We evaluated the relative contribution of infection to hospital mortality in a large population‐based study of individuals with PM/DM.Methods: Data derive from the 2007 to 2011 Healthcare Cost and Utilization Project National Inpatient Samples and include all hospital discharges that met a validated administrative definition of PM/DM. The primary outcome was hospital mortality. Variables for infections and comorbidities were generated from discharge diagnoses using validated administrative definitions. Logistic regression was used to investigate the relationship between infection and mortality in individuals with PM/DM, adjusting for sociodemographics, utilization variables, and comorbidities. Relative risks (RRs) were calculated to compare the overall prevalence of specific infections and associated mortality in PM/DM hospitalizations with those seen in the general hospitalized population.Results: A total of 15,407 hospitalizations with PM/DM met inclusion criteria for this study and inpatient mortality was 4.5% (700 deaths). In adjusted logistic regression analyses, infection (odds ratio [OR] 3.4, 95% confidence interval [95% CI] 2.9–4.0) was the strongest predictor of hospital mortality among individuals with PM/DM. Bacterial infection (OR 3.5, 95% CI 3.0–4.1), comprised primarily of pneumonia and bacteremia, and opportunistic fungal infections (OR 2.5, 95% CI 1.5–4.0) were independently associated with hospital mortality. The overall burden of infection in hospitalizations with PM/DM was significantly increased in comparison with the general hospitalized population (RR 1.5, 95% CI 1.4–1.6).Conclusion: Among hospitalized individuals with PM/DM, infection is the leading cause of mortality. Strategies to mitigate infection risk in both the clinic and hospital settings should be evaluated to improve disease outcomes.
- 20 October 2014
- Published by Wiley in Arthritis Care & Research
- Vol. 67 (5) , 673-680
Open AccessA Population‐Based Study of Infection‐Related Hospital Mortality in Patients With Dermatomyositis/Polymyositis
Objective: Dermatomyositis (DM) and polymyositis (PM) are debilitating inflammatory myopathies associated with significant mortality. We evaluated the relative contribution of infection to hospital mortality in a large population‐based study of individuals with PM/DM.Methods: Data derive from the 2007 to 2011 Healthcare Cost and Utilization Project National Inpatient Samples and include all hospital discharges that met a validated administrative definition of PM/DM. The primary outcome was hospital mortality. Variables for infections and comorbidities were generated from discharge diagnoses using validated administrative definitions. Logistic regression was used to investigate the relationship between infection and mortality in individuals with PM/DM, adjusting for sociodemographics, utilization variables, and comorbidities. Relative risks (RRs) were calculated to compare the overall prevalence of specific infections and associated mortality in PM/DM hospitalizations with those seen in the general hospitalized population.Results: A total of 15,407 hospitalizations with PM/DM met inclusion criteria for this study and inpatient mortality was 4.5% (700 deaths). In adjusted logistic regression analyses, infection (odds ratio [OR] 3.4, 95% confidence interval [95% CI] 2.9–4.0) was the strongest predictor of hospital mortality among individuals with PM/DM. Bacterial infection (OR 3.5, 95% CI 3.0–4.1), comprised primarily of pneumonia and bacteremia, and opportunistic fungal infections (OR 2.5, 95% CI 1.5–4.0) were independently associated with hospital mortality. The overall burden of infection in hospitalizations with PM/DM was significantly increased in comparison with the general hospitalized population (RR 1.5, 95% CI 1.4–1.6).Conclusion: Among hospitalized individuals with PM/DM, infection is the leading cause of mortality. Strategies to mitigate infection risk in both the clinic and hospital settings should be evaluated to improve disease outcomes.
- Open Access
- 1 May 2012
- Published by Wiley in Arthritis Care & Research
- Vol. 64 (9) , 1328-1333
Cardiovascular disease and cognitive dysfunction in systemic lupus erythematosus
Objective: Cognitive dysfunction and cardiovascular disease are common and debilitating manifestations of systemic lupus erythematosus (SLE). In this study, we evaluated the relationship between cardiovascular events, traditional cardiovascular risk factors, and SLE‐specific risk factors as predictors of cognitive dysfunction in a large cohort of participants with SLE.Methods: Subjects included 694 participants from the Lupus Outcomes Study (LOS), a longitudinal study of SLE outcomes based on an annual telephone survey querying demographic and clinical variables. The Hopkins Verbal Learning Test‐Revised and the Controlled Oral Word Association Test were administered to assess cognitive function. Multiple logistic regression was used to identify cardiovascular events (myocardial infarction, stroke), traditional cardiovascular risk factors (hypertension, hyperlipidemia, diabetes mellitus, obesity, smoking), and SLE‐specific risk factors (antiphospholipid antibodies [aPL], disease activity, disease duration) associated with cognitive impairment in year 7 of the LOS.Results: The prevalence of cognitive impairment as measured by verbal memory and verbal fluency metrics was 15%. In adjusted multiple logistic regression analyses, aPL (odds ratio [OR] 2.10, 95% confidence interval [95% CI] 1.3–3.41), hypertension (OR 2.06, 95% CI 1.19–3.56), and a history of stroke (OR 2.27, 95% CI 1.16–4.43) were significantly associated with cognitive dysfunction. In additional analyses evaluating the association between these predictors and severity of cognitive impairment, stroke was significantly more prevalent in participants with severe impairment when compared to those with mild or moderate impairment (P = 0.036).Conclusion: These results suggest that the presence of aPL, hypertension, and stroke are key variables associated with cognitive impairment, which may aid in identification of patients at greatest risk.
- 1 May 2012
- Published by Wiley in Arthritis Care & Research
- Vol. 64 (9) , 1328-1333
Open AccessCardiovascular disease and cognitive dysfunction in systemic lupus erythematosus
Objective: Cognitive dysfunction and cardiovascular disease are common and debilitating manifestations of systemic lupus erythematosus (SLE). In this study, we evaluated the relationship between cardiovascular events, traditional cardiovascular risk factors, and SLE‐specific risk factors as predictors of cognitive dysfunction in a large cohort of participants with SLE.Methods: Subjects included 694 participants from the Lupus Outcomes Study (LOS), a longitudinal study of SLE outcomes based on an annual telephone survey querying demographic and clinical variables. The Hopkins Verbal Learning Test‐Revised and the Controlled Oral Word Association Test were administered to assess cognitive function. Multiple logistic regression was used to identify cardiovascular events (myocardial infarction, stroke), traditional cardiovascular risk factors (hypertension, hyperlipidemia, diabetes mellitus, obesity, smoking), and SLE‐specific risk factors (antiphospholipid antibodies [aPL], disease activity, disease duration) associated with cognitive impairment in year 7 of the LOS.Results: The prevalence of cognitive impairment as measured by verbal memory and verbal fluency metrics was 15%. In adjusted multiple logistic regression analyses, aPL (odds ratio [OR] 2.10, 95% confidence interval [95% CI] 1.3–3.41), hypertension (OR 2.06, 95% CI 1.19–3.56), and a history of stroke (OR 2.27, 95% CI 1.16–4.43) were significantly associated with cognitive dysfunction. In additional analyses evaluating the association between these predictors and severity of cognitive impairment, stroke was significantly more prevalent in participants with severe impairment when compared to those with mild or moderate impairment (P = 0.036).Conclusion: These results suggest that the presence of aPL, hypertension, and stroke are key variables associated with cognitive impairment, which may aid in identification of patients at greatest risk.