Hannah J. Wong
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Hannah J. Wong has a total of 83 co-authors. Here are the top co-authors by number of shared publications:
- Howard Abrams
University Health Network
6 shared publications - Michael Caesar
Senior director of Information Management at the University Health Network. He is a leader in large scale, complex quality improvement initiatives at UHN and across the Province, and has been the co-chair of UHN's IT Downtime Committee since its inception
6 shared publications Dante Morra
5 shared publicationsRobert C. Wu
Division of General Internal Medicine, Toronto General Hospital University Health Network Toronto Ontario Canada
4 shared publicationsAnastasia Gayowsky
ICES Toronto Ontario Canada
2 shared publicationsChengsong Liu
Guelph-Waterloo Centre for Graduate Work in Chemistry and Biochemistry, Department of Chemistry, University of Waterloo, Waterloo, Ontario, N2L 3G1, Canada
2 shared publications
Latest Publications
- Open Access
- 23 October 2025
- Published by Wiley in Journal of the American Geriatrics Society
- Vol. 74 (1) , 107-118
Predictors of Advance Directive Changes in Ontario Nursing Home Residents: A Case–Control Study
Background: Goals of care (GOC) discussions between the clinical team and nursing home (NH) residents provide the basis for decision‐making on advance directives (AD) that include do‐not‐resuscitate (DNR) and do‐not‐hospitalize (DNH). Optimal timing and prompts for initiating GOC discussions are unclear. This study investigates recent emergency department (ED) use and clinical and demographic factors associated with subsequent AD changes. Methods: Nested case–control study within a population‐based retrospective cohort using linked administrative health care data of individuals admitted to NHs in Ontario, Canada between 2013 and 2017 and then followed up until 2019. Eligible cases and controls were residents with and without an AD change between 2013 and 2019, respectively. Cases and controls were matched 1:1 by sex, composite AD at NH admission, NH admission date (±90 days), and birthdate (±365 days). The primary outcome was incident AD change, and exposures included recent ED use (either ED visits discharged back to NH or ED visits that resulted in hospitalization) and clinical and demographic variables measured at the time of documented AD change. Conditional logistic regression provided adjusted odds ratios for associations between exposures and incident AD change. Results: The cases and controls (27,942 residents) had a mean age of 84 years at NH admission and 67.1% were female. 48.3% had a baseline AD of “DNR Only” while the remaining were evenly divided between “Full Code” and “DNR+DNH.” The estimated adjusted odds ratio of AD change was 2.01 (95% CI, 1.83–2.21) in residents with recent hospitalization, 1.89 (95% CI, 1.67–2.13) in those having end‐stage disease, and 1.82 (95% CI, 1.56–2.12) in residents who were mostly bedfast. Conclusions: A recent hospitalization, end‐stage disease, or being bedfast are significant predictors of AD change. These important predictors exhibited by NH residents present opportunities to reassess GOC.
- 23 October 2025
- Published by Wiley in Journal of the American Geriatrics Society
- Vol. 74 (1) , 107-118
Open AccessPredictors of Advance Directive Changes in Ontario Nursing Home Residents: A Case–Control Study
Background: Goals of care (GOC) discussions between the clinical team and nursing home (NH) residents provide the basis for decision‐making on advance directives (AD) that include do‐not‐resuscitate (DNR) and do‐not‐hospitalize (DNH). Optimal timing and prompts for initiating GOC discussions are unclear. This study investigates recent emergency department (ED) use and clinical and demographic factors associated with subsequent AD changes. Methods: Nested case–control study within a population‐based retrospective cohort using linked administrative health care data of individuals admitted to NHs in Ontario, Canada between 2013 and 2017 and then followed up until 2019. Eligible cases and controls were residents with and without an AD change between 2013 and 2019, respectively. Cases and controls were matched 1:1 by sex, composite AD at NH admission, NH admission date (±90 days), and birthdate (±365 days). The primary outcome was incident AD change, and exposures included recent ED use (either ED visits discharged back to NH or ED visits that resulted in hospitalization) and clinical and demographic variables measured at the time of documented AD change. Conditional logistic regression provided adjusted odds ratios for associations between exposures and incident AD change. Results: The cases and controls (27,942 residents) had a mean age of 84 years at NH admission and 67.1% were female. 48.3% had a baseline AD of “DNR Only” while the remaining were evenly divided between “Full Code” and “DNR+DNH.” The estimated adjusted odds ratio of AD change was 2.01 (95% CI, 1.83–2.21) in residents with recent hospitalization, 1.89 (95% CI, 1.67–2.13) in those having end‐stage disease, and 1.82 (95% CI, 1.56–2.12) in residents who were mostly bedfast. Conclusions: A recent hospitalization, end‐stage disease, or being bedfast are significant predictors of AD change. These important predictors exhibited by NH residents present opportunities to reassess GOC.
- Open Access
- 14 June 2024
- Published by Elsevier in Journal of the American Medical Directors Association
- Vol. 25 (8) , 105090
Advance Directives Change Frequently in Nursing Home Residents
- 14 June 2024
- Published by Elsevier in Journal of the American Medical Directors Association
- Vol. 25 (8) , 105090
Open AccessAdvance Directives Change Frequently in Nursing Home Residents
- Open Access
- 30 July 2021
- Published by Springer Nature in BMC Nursing
- Vol. 20 (1) , 1-14
Acute care nurses’ perceptions of leadership, teamwork, turnover intention and patient safety – a mixed methods study
Background: This study contributes to a small but growing body of literature on how context influences perceptions of patient safety in healthcare settings. We examine the impact of senior leadership support for safety, supervisory leadership support for safety, teamwork, and turnover intention on overall patient safety grade. Interaction effects of predictors on perceptions of patient safety are also examined. Methods: In this mixed methods study, cross-sectional survey data (N = 185) were collected from nurses and non-physician healthcare professionals. Semi-structured interview data (N = 15) were collected from nurses. The study participants worked in intensive care, general medicine, mental health, or the emergency department of a large community hospital in Southern Ontario. Results: Hierarchical regression analyses showed that staff perceptions of senior leadership (p < 0.001), teamwork (p < 0.01), and turnover intention (p < 0.01) were significantly associated with overall patient safety grade. The interactive effect of teamwork and turnover intention on overall patient safety grade was also found to be significant (p < 0.05). The qualitative findings corroborated the survey results but also helped expand the characteristics of the study’s key concepts (e.g., teamwork within and across professional boundaries) and why certain statistical relationships were found to be non-significant (e.g., nurse interviewees perceived the safety specific responsibilities of frontline supervisors much more broadly compared to the narrower conceptualization of the construct in the survey). Conclusions: The results of the current study suggest that senior leadership, teamwork, and turnover intention significantly impact nursing staff perceptions of patient safety. Leadership is a modifiable contextual factor and resources should be dedicated to strengthen relational competencies of healthcare leaders. Healthcare organizations must also proactively foster inter and intra-professional collaboration by providing teamwork educational workshops or other on-site learning opportunities (e.g., simulation training). Healthcare organizations would benefit by considering the interactive effect of contextual factors as another lever for patient safety improvement, e.g., lowering staff turnover intentions would maximize the positive impact of teamwork improvement initiatives on patient safety.
- 30 July 2021
- Published by Springer Nature in BMC Nursing
- Vol. 20 (1) , 1-14
Open AccessAcute care nurses’ perceptions of leadership, teamwork, turnover intention and patient safety – a mixed methods study
Background: This study contributes to a small but growing body of literature on how context influences perceptions of patient safety in healthcare settings. We examine the impact of senior leadership support for safety, supervisory leadership support for safety, teamwork, and turnover intention on overall patient safety grade. Interaction effects of predictors on perceptions of patient safety are also examined. Methods: In this mixed methods study, cross-sectional survey data (N = 185) were collected from nurses and non-physician healthcare professionals. Semi-structured interview data (N = 15) were collected from nurses. The study participants worked in intensive care, general medicine, mental health, or the emergency department of a large community hospital in Southern Ontario. Results: Hierarchical regression analyses showed that staff perceptions of senior leadership (p < 0.001), teamwork (p < 0.01), and turnover intention (p < 0.01) were significantly associated with overall patient safety grade. The interactive effect of teamwork and turnover intention on overall patient safety grade was also found to be significant (p < 0.05). The qualitative findings corroborated the survey results but also helped expand the characteristics of the study’s key concepts (e.g., teamwork within and across professional boundaries) and why certain statistical relationships were found to be non-significant (e.g., nurse interviewees perceived the safety specific responsibilities of frontline supervisors much more broadly compared to the narrower conceptualization of the construct in the survey). Conclusions: The results of the current study suggest that senior leadership, teamwork, and turnover intention significantly impact nursing staff perceptions of patient safety. Leadership is a modifiable contextual factor and resources should be dedicated to strengthen relational competencies of healthcare leaders. Healthcare organizations must also proactively foster inter and intra-professional collaboration by providing teamwork educational workshops or other on-site learning opportunities (e.g., simulation training). Healthcare organizations would benefit by considering the interactive effect of contextual factors as another lever for patient safety improvement, e.g., lowering staff turnover intentions would maximize the positive impact of teamwork improvement initiatives on patient safety.
- Open Access
- 14 August 2019
- Published by Springer Nature in Human Resources for Health
- Vol. 17 (1) , 1-9
Turnover intention of hospital staff in Ontario, Canada: exploring the role of frontline supervisors, teamwork, and mindful organizing
This study contributes to a small but growing body of literature on how context influences employee turnover intention. We examine the impact of staff perceptions of supervisory leadership support for safety, teamwork, and mindful organizing on turnover intention. Interaction effects of safety-specific constructs on turnover intention are also examined. Cross-sectional survey data were collected from nurses, allied health professionals, and unit clerks working in intensive care, general medicine, mental health, or the emergency department of a large community hospital in Southern Ontario. Hierarchical regression analyses showed that staff perceptions of teamwork were significantly associated with turnover intention (p < 0.001). Direct associations of supervisory leadership support for safety and mindful organizing with turnover intention were non-significant; however, when staff perceived lower levels of mindful organizing at the frontlines, the positive effect of supervisory leadership on turnover intention was significant (p < 0.01). Our results suggest that, in addition to teamwork perceptions positively affecting turnover intentions, safety-conscious supportive supervisors can help alleviate the negative impact of poor mindful organizing on frontline staff turnover intention. Healthcare organizations should recruit and retain individuals in supervisory roles who prioritize safety and possess adequate relational competencies. They should further dedicate resources to build and strengthen the relational capacities of their supervisory leadership. Moreover, it is important to provide on-site workshops on topics (e.g., conflict management) that can improve the quality of teamwork and consequently reduce employees’ intention to leave their unit/organization.
- 14 August 2019
- Published by Springer Nature in Human Resources for Health
- Vol. 17 (1) , 1-9
Open AccessTurnover intention of hospital staff in Ontario, Canada: exploring the role of frontline supervisors, teamwork, and mindful organizing
This study contributes to a small but growing body of literature on how context influences employee turnover intention. We examine the impact of staff perceptions of supervisory leadership support for safety, teamwork, and mindful organizing on turnover intention. Interaction effects of safety-specific constructs on turnover intention are also examined. Cross-sectional survey data were collected from nurses, allied health professionals, and unit clerks working in intensive care, general medicine, mental health, or the emergency department of a large community hospital in Southern Ontario. Hierarchical regression analyses showed that staff perceptions of teamwork were significantly associated with turnover intention (p < 0.001). Direct associations of supervisory leadership support for safety and mindful organizing with turnover intention were non-significant; however, when staff perceived lower levels of mindful organizing at the frontlines, the positive effect of supervisory leadership on turnover intention was significant (p < 0.01). Our results suggest that, in addition to teamwork perceptions positively affecting turnover intentions, safety-conscious supportive supervisors can help alleviate the negative impact of poor mindful organizing on frontline staff turnover intention. Healthcare organizations should recruit and retain individuals in supervisory roles who prioritize safety and possess adequate relational competencies. They should further dedicate resources to build and strengthen the relational capacities of their supervisory leadership. Moreover, it is important to provide on-site workshops on topics (e.g., conflict management) that can improve the quality of teamwork and consequently reduce employees’ intention to leave their unit/organization.
- Open Access
- 1 December 2018
- Published by American Psychiatric Association Publishing in Psychiatric Services
- Vol. 69 (12) , 1230-1237
Behavioral Health Factors as Predictors of Emergency Department Use in the High-Risk, High-Cost Medicare Population
This study measured the presence, extent, and type of behavioral health factors in a high-cost Medicare population and their association with the probability and intensity of emergency department (ED) use. Retrospective claims analysis and a comprehensive electronic medical record-based review were conducted for patients enrolled in a 65-month prospective care management program at an academic tertiary medical center (N=3,620). A two-part model used multivariable logistic regression to evaluate the effect of behavioral health factors on the probability of ED use, complemented by a Poisson model to measure the number of ED visits. Control variables included demographic characteristics, poststudy survival, and hierarchical condition category risk score. After analyses controlled for comorbidities and other relevant variables, patients with two or more behavioral health diagnosis categories or two or more behavioral health medications were about twice as likely as those without such categories or medications to use the ED. Patients with a diagnosis category of psychosis, neuropsychiatric disorders, sleep disorders, or adjustment disorders were significantly more likely than those without these disorders to use the ED. Most primary ED diagnoses were not of behavioral health conditions. Behavioral health factors had a substantial and significant effect on the likelihood and number of ED visits in a population of high-cost Medicare patients. Attention to behavioral health factors as independent predictors of ED use may be useful in influencing ED use in high-cost populations.
- 1 December 2018
- Published by American Psychiatric Association Publishing in Psychiatric Services
- Vol. 69 (12) , 1230-1237
Open AccessBehavioral Health Factors as Predictors of Emergency Department Use in the High-Risk, High-Cost Medicare Population
This study measured the presence, extent, and type of behavioral health factors in a high-cost Medicare population and their association with the probability and intensity of emergency department (ED) use. Retrospective claims analysis and a comprehensive electronic medical record-based review were conducted for patients enrolled in a 65-month prospective care management program at an academic tertiary medical center (N=3,620). A two-part model used multivariable logistic regression to evaluate the effect of behavioral health factors on the probability of ED use, complemented by a Poisson model to measure the number of ED visits. Control variables included demographic characteristics, poststudy survival, and hierarchical condition category risk score. After analyses controlled for comorbidities and other relevant variables, patients with two or more behavioral health diagnosis categories or two or more behavioral health medications were about twice as likely as those without such categories or medications to use the ED. Patients with a diagnosis category of psychosis, neuropsychiatric disorders, sleep disorders, or adjustment disorders were significantly more likely than those without these disorders to use the ED. Most primary ED diagnoses were not of behavioral health conditions. Behavioral health factors had a substantial and significant effect on the likelihood and number of ED visits in a population of high-cost Medicare patients. Attention to behavioral health factors as independent predictors of ED use may be useful in influencing ED use in high-cost populations.
- 21 July 2016
- Published by Wiley in Journal of Hospital Medicine
- Vol. 11 (12) , 824-831
Goals of care discussions among hospitalized long‐term care residents: Predictors and associated outcomes of care
INTRODUCTION: There are limited data on the occurrence, predictors, and impact of goals of care (GOC) discussions during hospitalization for seriously ill elderly patients, particularly for long‐term care (LTC) residents.METHODS: The study was a retrospective chart review of 200 randomly sampled LTC residents hospitalized via the emergency department and admitted to the general internal medicine service of 2 Canadian academic hospitals, from January 2012 through December 2012. We applied logistic regression models to identify factors associated with, and outcomes of, these discussions.RESULTS: Overall, 9.4% (665 of 7084) of hospitalizations were patients from LTC. In the sample of 200 patients, 37.5% had a documented discussion. No baseline patient characteristic was associated with GOC discussions. Low Glasgow Coma Scale, high respiratory rate, and low oxygen saturation were associated with discussions. Patients with discussions had higher rates of orders for no resuscitation (80% vs 55%) and orders for comfort measures only (7% vs 0%). In adjusted analyses, patients with discussions had higher odds of in‐hospital death (52.0, 95% confidence interval [CI]: 6.2‐440.4) and 1‐year mortality (4.1, 95% CI: 1.7‐9.6). Nearly 75% of patients with a change in their GOC did not have this documented in the discharge summary.CONCLUSION: In hospitalized LTC patients, GOC discussions occurred infrequently and appeared to be triggered by illness severity. Orders for advance directives, in‐hospital death, and 1‐year mortality were associated with discussions. Rates of GOC documentation in the discharge summary were poor. This study provides direction for developing education and practice standards to improve GOC discussion rates and their communication back to LTC. Journal of Hospital Medicine 2015;11:824–831. © 2015 Society of Hospital Medicine
- 21 July 2016
- Published by Wiley in Journal of Hospital Medicine
- Vol. 11 (12) , 824-831
Goals of care discussions among hospitalized long‐term care residents: Predictors and associated outcomes of care
INTRODUCTION: There are limited data on the occurrence, predictors, and impact of goals of care (GOC) discussions during hospitalization for seriously ill elderly patients, particularly for long‐term care (LTC) residents.METHODS: The study was a retrospective chart review of 200 randomly sampled LTC residents hospitalized via the emergency department and admitted to the general internal medicine service of 2 Canadian academic hospitals, from January 2012 through December 2012. We applied logistic regression models to identify factors associated with, and outcomes of, these discussions.RESULTS: Overall, 9.4% (665 of 7084) of hospitalizations were patients from LTC. In the sample of 200 patients, 37.5% had a documented discussion. No baseline patient characteristic was associated with GOC discussions. Low Glasgow Coma Scale, high respiratory rate, and low oxygen saturation were associated with discussions. Patients with discussions had higher rates of orders for no resuscitation (80% vs 55%) and orders for comfort measures only (7% vs 0%). In adjusted analyses, patients with discussions had higher odds of in‐hospital death (52.0, 95% confidence interval [CI]: 6.2‐440.4) and 1‐year mortality (4.1, 95% CI: 1.7‐9.6). Nearly 75% of patients with a change in their GOC did not have this documented in the discharge summary.CONCLUSION: In hospitalized LTC patients, GOC discussions occurred infrequently and appeared to be triggered by illness severity. Orders for advance directives, in‐hospital death, and 1‐year mortality were associated with discussions. Rates of GOC documentation in the discharge summary were poor. This study provides direction for developing education and practice standards to improve GOC discussion rates and their communication back to LTC. Journal of Hospital Medicine 2015;11:824–831. © 2015 Society of Hospital Medicine
- Open Access
- 1 March 2016
- Published by Taylor & Francis in Journal of Multidisciplinary Healthcare
- Vol. ume 9, 133-136
Conceptualizing type 2 diabetes and its management
Type 2 diabetes is growing worldwide due to population growth, increased rates of obesity, unhealthy diet, and physical inactivity. Risk assessment methods can effectively evaluate the risk of diabetes, and a healthy lifestyle can significantly reduce risk or prevent complications of type 2 diabetes. However, risk assessment alone has not significantly improved poor adherence to recommended medical interventions and lifestyle changes. This paper focuses on the challenge of nonadherence and posits that improving adherence requires tailoring interventions that explicitly consider the social determinants of health.
- 1 March 2016
- Published by Taylor & Francis in Journal of Multidisciplinary Healthcare
- Vol. ume 9, 133-136
Open AccessConceptualizing type 2 diabetes and its management
Type 2 diabetes is growing worldwide due to population growth, increased rates of obesity, unhealthy diet, and physical inactivity. Risk assessment methods can effectively evaluate the risk of diabetes, and a healthy lifestyle can significantly reduce risk or prevent complications of type 2 diabetes. However, risk assessment alone has not significantly improved poor adherence to recommended medical interventions and lifestyle changes. This paper focuses on the challenge of nonadherence and posits that improving adherence requires tailoring interventions that explicitly consider the social determinants of health.
- 8 April 2014
- Published by S. Karger AG in Nephron Clinical Practice
- Vol. 126 (3) , 124-127
Inpatient and Emergent Resource Use of Patients on Dialysis at an Academic Medical Center
Background/Aim: End-stage renal disease patients require resources for emergent and inpatient care in addition to ambulatory dialysis. There are two dialysis modalities and settings which patients switch between. Our aim was to characterize the patterns and reasons for switching, as well as the emergent and inpatient utilization of these patients at the University Health Network. Methods: Patients who received chronic dialysis between March 1, 2006, and April 30, 2011, were identified. Utilization was measured by emergency department (ED) visits, inpatient hospitalizations, and bed-days occupied per year. Results: Out of 576 patients identified, 18.6% switched modality and/or setting. The majority of switches occurred during the first year of dialysis. Patients who switched had increased utilization compared to those on a continuous modality/setting. Overall, patients had a median rate of 0.91 ED visits per patient-year, compared to 1.56 for patients who switched modality and setting. Median inpatient bed resource requirement was 4.46 bed-days/patient-year overall, compared to 8.91 for patients who switched modality and setting. Conclusions: Emergent and inpatient utilization is related to the setting and modality of dialysis, although differences are partly explained by comorbidities. Patients who switch modalities use more resources and may be a prime population for interventions.
- 8 April 2014
- Published by S. Karger AG in Nephron Clinical Practice
- Vol. 126 (3) , 124-127
Inpatient and Emergent Resource Use of Patients on Dialysis at an Academic Medical Center
Background/Aim: End-stage renal disease patients require resources for emergent and inpatient care in addition to ambulatory dialysis. There are two dialysis modalities and settings which patients switch between. Our aim was to characterize the patterns and reasons for switching, as well as the emergent and inpatient utilization of these patients at the University Health Network. Methods: Patients who received chronic dialysis between March 1, 2006, and April 30, 2011, were identified. Utilization was measured by emergency department (ED) visits, inpatient hospitalizations, and bed-days occupied per year. Results: Out of 576 patients identified, 18.6% switched modality and/or setting. The majority of switches occurred during the first year of dialysis. Patients who switched had increased utilization compared to those on a continuous modality/setting. Overall, patients had a median rate of 0.91 ED visits per patient-year, compared to 1.56 for patients who switched modality and setting. Median inpatient bed resource requirement was 4.46 bed-days/patient-year overall, compared to 8.91 for patients who switched modality and setting. Conclusions: Emergent and inpatient utilization is related to the setting and modality of dialysis, although differences are partly explained by comorbidities. Patients who switch modalities use more resources and may be a prime population for interventions.
- 1 September 2013
- Published by Radiological Society of North America (RSNA) in Radiology
- Vol. 268 (3) , 779-789
Use of Imaging in the Emergency Department: Physicians Have Limited Effect on Variation
To quantify interphysician variation in imaging use during emergency department (ED) visits and examine the contribution of factors to this variation at the patient, visit, and physician level. This study was HIPAA compliant and approved by the institutional review board of Partners Healthcare System (Boston, Mass), with waiver of informed consent. In this retrospective study of 88 851 consecutive ED visits during 2011 at a large urban teaching hospital, a hierarchical logistic regression model was used to identify multiple predictors for the probability that low- or high-cost imaging would be ordered during a given visit. Physician-specific random effects were estimated to articulate (by odds ratio) and quantify (by intraclass correlation coefficient [ICC]) interphysician variation. Patient- and visit-level factors found to be statistically significant predictors of imaging use included measures of ED busyness, prior ED visit, referral source to the ED, and ED arrival mode. Physician-level factors (eg, sex, years since graduation, annual workload, and residency training) did not correlate with imaging use. The remaining amount of interphysician variation was very low (ICC, 0.97% for low-cost imaging; ICC, 1.07% for high-cost imaging). These physician-specific odds ratios of imaging estimates were moderately reliable at 0.78 (95% confidence interval [CI]: 0.77, 0.79) for low-cost imaging and 0.76 (95% CI: 0.74, 0.78) for high-cost imaging. After careful and comprehensive case-mix adjustment by using hierarchical logistic regression, only about 1% of the variability in ED imaging utilization was attributable to physicians.
- 1 September 2013
- Published by Radiological Society of North America (RSNA) in Radiology
- Vol. 268 (3) , 779-789
Use of Imaging in the Emergency Department: Physicians Have Limited Effect on Variation
To quantify interphysician variation in imaging use during emergency department (ED) visits and examine the contribution of factors to this variation at the patient, visit, and physician level. This study was HIPAA compliant and approved by the institutional review board of Partners Healthcare System (Boston, Mass), with waiver of informed consent. In this retrospective study of 88 851 consecutive ED visits during 2011 at a large urban teaching hospital, a hierarchical logistic regression model was used to identify multiple predictors for the probability that low- or high-cost imaging would be ordered during a given visit. Physician-specific random effects were estimated to articulate (by odds ratio) and quantify (by intraclass correlation coefficient [ICC]) interphysician variation. Patient- and visit-level factors found to be statistically significant predictors of imaging use included measures of ED busyness, prior ED visit, referral source to the ED, and ED arrival mode. Physician-level factors (eg, sex, years since graduation, annual workload, and residency training) did not correlate with imaging use. The remaining amount of interphysician variation was very low (ICC, 0.97% for low-cost imaging; ICC, 1.07% for high-cost imaging). These physician-specific odds ratios of imaging estimates were moderately reliable at 0.78 (95% confidence interval [CI]: 0.77, 0.79) for low-cost imaging and 0.76 (95% CI: 0.74, 0.78) for high-cost imaging. After careful and comprehensive case-mix adjustment by using hierarchical logistic regression, only about 1% of the variability in ED imaging utilization was attributable to physicians.