Brendan G Carr

University of Pennsylvania, Filadelfia, Pennsylvania, United States

Are you Brendan G Carr?

Claim your profile

Publications (116)429.27 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: We sought to improve public health surveillance by using a geographic analysis of emergency department (ED) visits to determine local chronic disease prevalence. Using an all-payer administrative database, we determined the proportion of unique ED patients with diabetes, hypertension, or asthma. We compared these rates to those determined by the New York City Community Health Survey. For diabetes prevalence, we also analyzed the fidelity of longitudinal estimates using logistic regression and determined disease burden within census tracts using geocoded addresses. We identified 4.4 million unique New York City adults visiting an ED between 2009 and 2012. When we compared our emergency sample to survey data, rates of neighborhood diabetes, hypertension, and asthma prevalence were similar (correlation coefficient = 0.86, 0.88, and 0.77, respectively). In addition, our method demonstrated less year-to-year scatter and identified significant variation of disease burden within neighborhoods among census tracts. Our method for determining chronic disease prevalence correlates with a validated health survey and may have higher reliability over time and greater granularity at a local level. Our findings can improve public health surveillance by identifying local variation of disease prevalence. (Am J Public Health. Published online ahead of print July 16, 2015: e1-e8. doi:10.2105/AJPH.2015.302679).
    American Journal of Public Health 07/2015; DOI:10.2105/AJPH.2015.302679 · 4.23 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: International Classification of Disease, Ninth Revision (ICD-9) diagnosis codes have not been validated for identifying cases of missed abortion where a pregnancy is no longer viable but the cervical os remains closed. Our goal was to assess whether ICD-9 code "632" for missed abortion has high sensitivity and positive predictive value (PPV) in identifying patients in the emergency department (ED) with cases of stable early pregnancy failure (EPF). We studied females ages 13-50 years presenting to the ED of an urban academic medical center. We approached our analysis from two perspectives, evaluating both the sensitivity and PPV of ICD-9 code "632" in identifying patients with stable EPF. All patients with chief complaints "pregnant and bleeding" or "pregnant and cramping" over a 12-month period were identified. We randomly reviewed two months of patient visits and calculated the sensitivity of ICD-9 code "632" for true cases of stable miscarriage. To establish the PPV of ICD-9 code "632" for capturing missed abortions, we identified patients whose visits from the same time period were assigned ICD-9 code "632," and identified those with actual cases of stable EPF. We reviewed 310 patient records (17.6% of 1,762 sampled). Thirteen of 31 patient records assigned ICD-9 code for missed abortion correctly identified cases of stable EPF (sensitivity=41.9%), and 140 of the 142 patients without EPF were not assigned the ICD-9 code "632"(specificity=98.6%). Of the 52 eligible patients identified by ICD-9 code "632," 39 cases met the criteria for stable EPF (PPV=75.0%). ICD-9 code "632" has low sensitivity for identifying stable EPF, but its high specificity and moderately high PPV are valuable for studying cases of stable EPF in epidemiologic studies using administrative data.
    The western journal of emergency medicine 07/2015; 16(4):551-6. DOI:10.5811/westjem.2015.4.24946
  • [Show abstract] [Hide abstract]
    ABSTRACT: Racial and ethnic disparities have been previously reported in acute stroke care. We sought to determine the effect of telemedicine (TM) on access to acute stroke care for racial and ethnic minorities in the state of Texas. Data were collected from the US Census Bureau, The Joint Commission and the American Hospital Association. Access for racial and ethnic minorities was determined by summing the population that could reach a primary stroke centre (PSC) or telemedicine spoke within specified time intervals using validated models. TM extended access to stroke expertise by 1.5 million residents. The odds of providing 60-minute access via TM were similar in Blacks and Whites (prevalence odds ratios (POR) 1.000, 95% CI 1.000-1.000), even after adjustment for urbanization (POR 1.000, 95% CI 1.000-1.001). The odds of providing access via TM were also similar for Hispanics and non-Hispanics (POR 1.000, 95% CI 1.000-1.000), even after adjustment for urbanization (POR 1.000, 95% CI 1.000-1.000). We found that telemedicine increased access to acute stroke care for 1.5 million Texans. While racial and ethnic disparities exist in other components of stroke care, we did not find evidence of disparities in access to the acute stroke expertise afforded by telemedicine. © The Author(s) 2015.
    Journal of Telemedicine and Telecare 06/2015; DOI:10.1177/1357633X15589534 · 1.74 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Geographic variation in healthcare has been traditionally studied in large areas such as hospital referral regions or service areas. These analyses are limited by variation that exists within local communities.
  • Journal of the American Academy of Dermatology 05/2015; 72(5). DOI:10.1016/j.jaad.2015.02.1093 · 5.00 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: We examine differences in inpatient mortality and hospitalization costs at trauma and nontrauma centers for injuries of minor and moderate severity. Inpatient data sets from the California Office of Statewide Health Planning and Development were analyzed for 2009 to 2011. The study population included patients younger than 85 years and admitted to general, acute care hospitals with a primary diagnosis of a minor or moderate injury. Minor injuries were defined as having a New Injury Severity Score less than 5 and moderate injuries as having a score of 5 to 15. Multivariate logistic regression and generalized linear model with log-link and γ distribution were used to estimate differences in adjusted inpatient mortality and costs. A total of 126,103 admissions with minor or moderate injury were included in the study population. The unadjusted mortality rate was 6.4 per 1,000 admissions (95% confidence interval [CI] 5.9 to 6.8). There was no significant difference found in mortality between trauma and nontrauma centers in unadjusted (odds ratio 1.2; 95% CI 0.97 to 1.48) or adjusted models (odds ratio 1.1; 95% CI 0.79 to 1.57). The average cost of a hospitalization was $13,465 (95% CI $12,733 to $14,198) and, after adjustment, was 33.1% higher at trauma centers compared with nontrauma centers (95% CI 16.9% to 51.6%). For patients admitted to hospitals for minor and moderate injuries, hospitalization costs in this study population were higher at trauma centers than nontrauma centers, after adjustments for patient clinical-, demographic-, and hospital-level characteristics. Mortality was a rare event in the study population and did not significantly differ between trauma and nontrauma centers. Copyright © 2015 American College of Emergency Physicians. Published by Elsevier Inc. All rights reserved.
    Annals of emergency medicine 05/2015; DOI:10.1016/j.annemergmed.2015.04.021 · 4.33 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Objectives This study explored what smartphone health applications (apps) are used by patients, how they learn about health apps, and how information about health apps is shared.Methods Patients seeking care in an academic ED were surveyed about the following regarding their health apps: use, knowledge, sharing, and desired app features. Demographics and health information were characterized by summary statistics.ResultsOf 300 participants, 212 (71%) owned smartphones, 201 (95%) had apps, and 94 (44%) had health apps. The most frequently downloaded health apps categories were exercise 46 (49%), brain teasers 30 (32%), and diet 23 (24%). The frequency of use of apps varied as six (6%) of health apps were downloaded but never used, 37 (39%) apps were used only a few times, and 40 (43%) health apps were used once per month. Only five apps (2%) were suggested to participants by health care providers, and many participants used health apps intermittently (55% of apps ≤ once a month). Participants indicated sharing information from 64 (59%) health apps, mostly within social networks (27 apps, 29%) and less often with health care providers (16 apps, 17%).Conclusions While mobile health has experienced tremendous growth over the past few years, use of health apps among our sample was low. The most commonly used apps were those that had broad functionality, while the most frequently used health apps encompassed the topics of exercise, diet, and brain teasers. While participants most often shared information about health apps within their social networks, information was less frequently shared with providers, and physician recommendation played a small role in influencing patient use of health apps.
    Academic Emergency Medicine 05/2015; 22(6). DOI:10.1111/acem.12675 · 2.20 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Effective measurement of health care quality, access, and cost for populations requires an accountable geographic unit. Although Hospital Service Areas (HSAs) and Hospital Referral Regions (HRRs) have been extensively used in health services research, it is unknown whether these units accurately describe patterns of hospital use for patients living within them. To evaluate the ability of HSAs, HRRs, and counties to define discrete health care populations. Cross-sectional geographic analysis of hospital admissions. All hospital admissions during the year 2011 in Washington, Arizona, and Florida. The main outcomes of interest were 3 metrics that describe patient movement across HSA, HRR, and county boundaries: localization index, market share index, and net patient flow. Regression models tested the association of these metrics with different HSA characteristics. For 45% of HSAs, fewer than half of the patients were admitted to hospitals located in their HSA of residence. For 16% of HSAs, more than half of the treated patients lived elsewhere. There was an equivalent degree of movement across county boundaries but less movement across HRR boundaries. Patients living in populous, urban HSAs with multiple, large, and teaching hospitals tended to remain for inpatient care. Patients admitted through the emergency department tended to receive care at local hospitals relative to other patients. HSAs and HRRs are geographic units commonly used in health services research yet vary in their ability to describe where patients receive hospital care. Geographic models may need to account for differences between emergent and nonemergent care.
    Medical care 04/2015; 53(6). DOI:10.1097/MLR.0000000000000356 · 2.94 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Sudden hospital closures displace patients from usual sources of care and force them to access facilities that lack their prior medical records. For patients with complex needs and for nearby hospitals already strained by high volume, disaster-related hospital closures induce a public health emergency. Our objective was to analyze responses of patients from public versus private emergency departments after closure of their usual hospital after Hurricane Sandy. Using a statewide database of emergency visits, we followed patients with an established pattern of accessing 1 of 2 hospitals that closed after Hurricane Sandy: Bellevue Hospital Center and NYU Langone Medical Center. We determined how these patients redistributed for emergency care after the storm. We found that proximity strongly predicted patient redistribution to nearby open hospitals. However, for patients from the closed public hospital, this redistribution was also influenced by hospital ownership, because patients redistributed to other public hospitals at rates higher than expected by proximity alone. This differential response to hospital closures demonstrates significant differences in how public and private patients respond to changes in health care access during disasters. Public health response must consider these differences to meet the needs of all patients affected by disasters and other public health emergencies. (Disaster Med Public Health Preparedness. 2015;00:1-9).
    Disaster Medicine and Public Health Preparedness 03/2015; 9:1-9. DOI:10.1017/dmp.2015.11 · 1.14 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The location of comprehensive stroke centers (CSCs) is critical to ensuring rapid access to acute stroke therapies; we conducted a population-level virtual trial simulating change in access to CSCs using optimization modeling to selectively convert primary stroke centers (PSCs) to CSCs. Up to 20 certified PSCs per state were selected for conversion to maximize the population with 60-minute CSC access by ground and air. Access was compared across states based on region and the presence of state-level emergency medical service policies preferentially routing patients to stroke centers. In 2010, there were 811 Joint Commission PSCs and 0 CSCs in the United States. Of the US population, 65.8% had 60-minute ground access to PSCs. After adding up to 20 optimally located CSCs per state, 63.1% of the US population had 60-minute ground access and 86.0% had 60-minute ground/air access to a CSC. Across states, median CSC access was 55.7% by ground (interquartile range 35.7%-71.5%) and 85.3% by ground/air (interquartile range 59.8%-92.1%). Ground access was lower in Stroke Belt states compared with non-Stroke Belt states (32.0% vs 58.6%, p = 0.02) and lower in states without emergency medical service routing policies (52.7% vs 68.3%, p = 0.04). Optimal system simulation can be used to develop efficient care systems that maximize accessibility. Under optimal conditions, a large proportion of the US population will be unable to access a CSC within 60 minutes. © 2015 American Academy of Neurology.
    Neurology 03/2015; 84(12). DOI:10.1212/WNL.0000000000001390 · 8.30 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Emergency visits are rising nationally, whereas the number of emergency departments is shrinking. However, volume has not increased uniformly at all emergency departments. It is unclear what factors account for this variability in emergency volume growth rates. The objective of this study was to test the association of hospital and population characteristics and the effect of hospital closures with increases in emergency department volume. The study team analyzed emergency department volume at New York State hospitals from 2004 to 2010 using data from cost reports and administrative databases. Multivariate regression was used to evaluate characteristics associated with emergency volume growth. Spatial analytics and distances between hospitals were used in calculating the predicted impact of hospital closures on emergency department use. Among the 192 New York hospitals open from 2004 to 2010, the mean annual increase in emergency department visits was 2.7%, but the range was wide (-5.5% to 11.3%). Emergency volume increased nearly twice as fast at tertiary referral centers (4.8%) and nonurban hospitals (3.7% versus urban at 2.1%) after adjusting for other characteristics. The effect of hospital closures also strongly predicted variation in growth. Emergency volume is increasing faster at specific hospitals: tertiary referral centers, nonurban hospitals, and those near hospital closures. This study provides an understanding of how emergency volume varies among hospitals and predicts the effect of hospital closures in a statewide region. Understanding the impact of these factors on emergency department use is essential to ensure that these populations have access to critical emergency services. (Population Health Management 2015;xx:xxx-xxx).
    Population Health Management 02/2015; DOI:10.1089/pop.2014.0123 · 1.35 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: To complete a 30-year interrupted time-series analysis of the impact of austerity-related and prosperity-related events on the occurrence of suicide across Greece. Greece from 1 January 1983 to 31 December 2012. A total of 11 505 suicides, 9079 by men and 2426 by women, occurring in Greece over the study period. National data from the Hellenic Statistical Authority assembled as 360 monthly counts of: all suicides, male suicides, female suicides and all suicides plus potentially misclassified suicides. In 30 years, the highest months of suicide in Greece occurred in 2012. The passage of new austerity measures in June 2011 marked the beginning of significant, abrupt and sustained increases in total suicides (+35.7%, p<0.001) and male suicides (+18.5%, p<0.01). Sensitivity analyses that figured in undercounting of suicides also found a significant, abrupt and sustained increase in June 2011 (+20.5%, p<0.001). Suicides by men in Greece also underwent a significant, abrupt and sustained increase in October 2008 when the Greek recession began (+13.1%, p<0.01), and an abrupt but temporary increase in April 2012 following a public suicide committed in response to austerity conditions (+29.7%, p<0.05). Suicides by women in Greece also underwent an abrupt and sustained increase in May 2011 following austerity-related events (+35.8%, p<0.05). One prosperity-related event, the January 2002 launch of the Euro in Greece, marked an abrupt but temporary decrease in male suicides (-27.1%, p<0.05). This is the first multidecade, national analysis of suicide in Greece using monthly data. Select austerity-related events in Greece corresponded to statistically significant increases for suicides overall, as well as for suicides among men and women. The consideration of future austerity measures should give greater weight to the unintended mental health consequences that may follow and the public messaging of these policies and related events. Published by the BMJ Publishing Group Limited. For permission to use (where not already granted under a licence) please go to http://group.bmj.com/group/rights-licensing/permissions.
    BMJ Open 01/2015; 5(1). DOI:10.1136/bmjopen-2014-005619 · 2.06 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: As emergency care becomes increasingly regionalized, systems planners must determine how system expansion impacts existing system assets. We hypothesized that accrediting new Level II and III trauma centers impacted the trauma patient census and severity at a nearby Level I trauma center and estimated the magnitude of the impact. We conducted an interrupted time series analysis using monthly patient counts during the past 10 years for five trauma centers located near one another in Pennsylvania. The Level I center (TC-A) operated for the entire period. A Level II center 39 miles away was accredited after 70 months (TC-B), one Level III center 46 miles away was accredited after 95 months but lost accreditation after 11 months (TC-C), and two other Level III centers 40 miles and 45 miles away were accredited after 107 months (TC-D and TC-E). Monthly patient volume at the Level I center, which increased gradually over the study and summed to 25,120 patients, decreased by 10.8% (p < 0.05) when TC-B was accredited and decreased by an additional 12.9% (p < 0.05) when TC-D and TC-E were accredited simultaneously. No change stemmed from temporarily accrediting TC-C. Ultimately, the Level I center treated 1,903 fewer patients than expected over 51 months, an 11.9% volume reduction, and patient severity remained consistent but mortality decreased. Accrediting Level II and Level III trauma centers reduced patient volume and reduced overall mortality at a nearby Level I center. Strategic planning of statewide trauma systems can help balance rapid access to care with maintenance of adequate annual patient volumes of critically injured patients. Epidemiologic study, Level IV.
    Journal of Trauma and Acute Care Surgery 11/2014; 77(5):764-768. DOI:10.1097/TA.0000000000000430 · 1.97 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Introduction: Time to achieve target temperature varies substantially for patients who undergo targeted temperature management (TTM) after cardiac arrest. The association between arrival at target temperature and neurologic outcome is poorly understood. We hypothesized that shorter time from initiation of cooling to target temperature ("induction") will be associated with worse neurologic outcome, reflecting more profound underlying brain injury and impaired thermoregulatory control. Methods: This was a multicenter retrospective study analyzing data from the Penn Alliance for Therapeutic Hypothermia (PATH) Registry. We examined the association between time from arrest to return of spontaneous circulation (ROSC) ("downtime"), ROSC to initiation of TTM ("pre-induction") and "induction" with cerebral performance category (CPC). Results: A total of 321 patients were analyzed, of whom 30.8% (99/321) had a good neurologic outcome. Downtime for survivors with good outcome was 11 (IQR 6-27) min vs. 21 (IQR 10-36) min (p=. 0.002) for those with poor outcome. Pre-induction did not vary between good and poor outcomes (98 (IQR 36-230) min vs. 114 (IQR 34-260) (p=. ns)). Induction time in the good outcome cohort was 237 (IQR 142-361) min compared to 180 (IQR 100-276) min (p=. 0.004). Patients were categorized by induction time (<120. min, 120-300. min, >300. min). Using multivariable logistic regression adjusted for age, initial rhythm, and downtime, induction time >300. min was associated with good neurologic outcome when compared to those with an induction time <120. min. Conclusion: In this multicenter cohort of post-arrest TTM patients, shorter induction time was associated with poor neurologic outcome.
    Resuscitation 10/2014; 88. DOI:10.1016/j.resuscitation.2014.10.018 · 3.96 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Background and Purpose We examine whether the proportion of the US population with 60 minute access to Primary Stroke Centers (PSCs) varies based on geographic and demographic factors. Methods Population level access to PSCs within 60 minutes was estimated using validated models of prehospital time accounting for critical prehospital time intervals and existing road networks. We examined the association between geographic factors, demographic factors, and access to care. Multivariable models quantified the association between demographics and PSC access for the entire United States and then stratified by urbanicity. Results Of the 309 million people in the United States, 65.8% had 60 minute PSC access by ground ambulance (87% major cities, 59% minor cities, 9% suburbs, and 1% rural). PSC access was lower in stroke belt states (44% versus 69%). Non-whites were more likely to have access than whites (77% versus 62%), and Hispanics were more likely to have access than non-Hispanics (78% versus 64%). Demographics were not meaningfully associated with access in major cities or suburbs. In smaller cities, there was less access in areas with lower income, less education, more uninsured, more Medicare and Medicaid eligibles, lower healthcare utilization, and healthcare resources. Conclusions There are significant geographic disparities in access to PSCs. Access is limited in nonurban areas. Despite the higher burden of cerebrovascular disease in stroke belt states, access to care is lower in these areas. Selecting demographic and healthcare factors is strongly associated with access to care in smaller cities, but not in other areas, including major cities.
    Stroke 10/2014; 45(11). DOI:10.1161/STROKEAHA.114.006021 · 6.02 Impact Factor
  • Chest 10/2014; 146(4_suppl):e75S. DOI:10.1378/chest.14-0737 · 7.13 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Study aims: To assess whether increased use of targeted temperature management (If M) within an integrated healthcare delivery system resulted in improved rates of good neurologic outcome at hospital discharge (Cerebral Performance Category score of 1 or 2). Methods: Retrospective cohort study of patients with OHCA admitted to 21 medical centers between January 2007 and December 2012. A standardized TTM protocol and educational program were introduced throughout the system in early 2009. Comatose patients eligible for treatment with TTM were included. Adjusted odds of good neurologic outcome at hospital discharge and survival to hospital discharge were assessed using multivariate logistic regression. Results: A total of 1119 patients were admitted post-OHCA with coma, 59.1% (661 of 1119) of which were eligible for If M. The percentage of patients treated with TTM markedly increased during the study period: 10.5% in the years preceding( 2007-2008) vs. 85.1% in the years following (2011-2012) implementation of the practice improvement initiative. However, unadjusted in-hospital survival (37.3% vs. 39.0%, p = 0.77) and good neurologic outcome at hospital discharge (26.3% vs. 26.6%, p = 1.0) did not change. The adjusted odds of survival to hospital discharge (AOR 1.0, 95% CI 0.85-1.17) or a good neurologic outcome (AOR 0.94,95% CI 0.79-1.11) were likewise non-significant. Interpretation: Despite a marked increase in TTM rates across hospitals in an integrated delivery system, there was no appreciable change in the crude or adjusted odds of in-hospital survival or good neurologic outcomes at hospital discharge among eligible post-arrest patients.
    Resuscitation 08/2014; 85(11). DOI:10.1016/j.resuscitation.2014.08.014 · 3.96 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Background and Purpose-Only 3% to 5% of patients with acute ischemic stroke receive intravenous recombinant tissue-type plasminogen activator (r-tPA) and <1% receive endovascular therapy. We describe access of the US population to all facilities that actually provide intravenous r-tPA or endovascular therapy for acute ischemic stroke. Methods-We used US demographic data and intravenous r-tPA and endovascular therapy rates in the 2011 US Medicare Provider and Analysis Review data set. International Classification of Diseases-Ninth Revision codes 433. xx, 434. xx and 436 identified acute ischemic stroke cases. International Classification of Diseases-Ninth Revision code 99.10 defined intravenous r-tPA treatment and International Classification of Diseases-Ninth Revision code 39.74 defined endovascular therapy. We estimated ambulance response times using arc-Geographic Information System's network analyst and helicopter transport times using validated models. Population access to care was determined by summing the population contained within travel sheds that could reach capable hospitals within 60 and 120 minutes. Results-Of 370 351 acute ischemic stroke primary diagnosis discharges, 14 926 (4%) received intravenous r-tPA and 1889 (0.5%) had endovascular therapy. By ground, 81% of the US population had access to intravenous-capable hospitals within 60 minutes and 56% had access to endovascular-capable hospitals. By air, 97% had access to intravenous-capable hospitals within 60 minutes and 85% had access to endovascular hospitals. Within 120 minutes, 99% of the population had access to both intravenous and endovascular hospitals. Conclusions-More than half of the US population has geographic access to hospitals that actually deliver acute stroke care but treatment rates remain low. These data provide a national perspective on acute stroke care and should inform the planning and optimization of stroke systems in the United States.
    Stroke 08/2014; 45(10). DOI:10.1161/STROKEAHA.114.006293 · 6.02 Impact Factor

Publication Stats

1k Citations
429.27 Total Impact Points

Institutions

  • 2010–2015
    • University of Pennsylvania
      • • Center for Clinical Epidemiology and Biostatistics
      • • Department of Emergency Medicine
      Filadelfia, Pennsylvania, United States
    • Emory University
      • Department of Emergency Medicine
      Atlanta, Georgia, United States
    • Mayo Clinic - Rochester
      • Department of Emergency Medicine
      Рочестер, Minnesota, United States
  • 2007–2015
    • Thomas Jefferson University
      Filadelfia, Pennsylvania, United States
  • 2006–2014
    • William Penn University
      Filadelfia, Pennsylvania, United States
    • Carolinas Medical Center University
      Charlotte, North Carolina, United States
  • 2013
    • Treatment Research Institute, Philadelphia PA
      Filadelfia, Pennsylvania, United States
  • 2009–2013
    • The Children's Hospital of Philadelphia
      Philadelphia, Pennsylvania, United States
  • 2007–2011
    • Hospital of the University of Pennsylvania
      • Department of Emergency Medicine
      Philadelphia, Pennsylvania, United States
  • 2008
    • Robert Wood Johnson Foundation
      Princeton, New Jersey, United States
  • 2007–2008
    • Robert Wood Johnson University Hospital
      Нью-Брансуик, New Jersey, United States