Arthur P Wheeler

Vanderbilt University, Нашвилл, Michigan, United States

Are you Arthur P Wheeler?

Claim your profile

Publications (97)1155.57 Total impact

  • Critical care medicine 11/2015; 43(12 Suppl 1):158. DOI:10.1097/ · 6.31 Impact Factor
  • Trevor Taylor · Matthew Semler · Arthur Wheeler · Todd Rice ·

    Critical care medicine 11/2015; 43(12 Suppl 1):66. DOI:10.1097/01.ccm.0000474087.31883.86 · 6.31 Impact Factor
  • Arthur P Wheeler ·

    Chest 07/2015; 148(1):3-4. DOI:10.1378/chest.15-0395 · 7.48 Impact Factor
  • Michael J Noto · Arthur P Wheeler ·

    Intensive Care Medicine 06/2015; 41(7). DOI:10.1007/s00134-015-3846-6 · 7.21 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Recent emphasis has been placed on methods to predict fluid-responsiveness (FR), but the utility of using fluid boluses to increase cardiac index (CI) in critically-ill patients with ineffective circulation or oliguria remains unclear. Retrospective analysis investigating hemodynamic responses of critically ill patients in the ARDS Network Fluid and Catheter Treatment Trial (FACTT) who were given protocol-based fluid boluses. FR was defined as ≥15% increase in CI after a 15ml/kg fluid bolus. A convenience sample of 127 critically-ill patients enrolled in FACTT was analyzed for physiologic responses to 569 protocolized crystalloid or albumin boluses given for shock, low urine output (UOP), or low pulmonary artery occlusion pressure (PAOP). There were significant increases in mean central venous pressure (9.9±4.5 to 11.1±4.8 mmHg, p<0.0001) and mean PAOP (11.6±3.6 to 13.3±4.3 mmHg, p<0.0001) following fluid boluses. However, there were no significant change in UOP, and clinically small changes in heart rate (HR), mean arterial pressure (MAP), and CI. Only 23% of fluid boluses led to a ≥15% change in CI. There was no significant difference in the frequency of fluid responsiveness between boluses given for shock or oliguria versus boluses given only for low PAOP (24.0% vs. 21.8%, p=0.59). There were no significant differences in 90-day survival, need for hemodialysis, or return to unassisted breathing between patients defined as fluid responders and fluid non-responders. In this cohort of critically-ill and previously resuscitated ARDS patients, the rate of fluid responsiveness was low and fluid boluses only led to small hemodynamic changes.
    Chest 05/2015; 148(4). DOI:10.1378/chest.15-0445 · 7.48 Impact Factor
  • Michael J Noto · Todd W Rice · Arthur P Wheeler ·

    JAMA The Journal of the American Medical Association 05/2015; 313(18):1863-1864. DOI:10.1001/jama.2015.3539 · 35.29 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Coordinating efforts across disciplines in the intensive care unit is a key component of quality improvement (QI) efforts. Spontaneous awakening trials (SATs) and spontaneous breathing trials (SBTs) are considered key components of guidelines, yet unfortunately are often not done or coordinated properly. To determine if a pharmacist-driven awakening and breathing coordination (ABC) QI program would improve compliance (ie, process measures) as compared with the previous protocol, which did not involve pharmacists. The QI program included pharmacist-led education, daily discussion on rounds, and weekly performance reports to staff. Using a pre-QI versus during-QI versus post-QI intervention design, we compared data from 500 control ventilator-days (pre-QI period) versus 580 prospective ventilator-days (during-QI period). We then evaluated the sustainability of the QI program in 216 ventilator-days in the post-QI period. SAT safety screens were performed on only 20% pre-QI patient-days versus 97% of during-QI patient-days (P < 0.001) and 100% of post-QI patient-days (P = 0.25). The rates of passing the SAT safety screen in pre-QI and during-QI periods were 63% versus 78% (P = 0.03) and 81% in the post-QI period (P = 0.86). The rates of SATs among eligible patients on continuous infusions were only 53% in the pre-QI versus 85% in the during-QI (P = 0.0001) and 87% in the post-QI (P = 1) periods. In this QI initiative, a pharmacist-driven, interdisciplinary ABC protocol significantly improved process measures compliance, comparing the pre-QI versus during-QI rates of screening, performing, and coordinating SAT and SBTs, and these results were sustained in the 8-month follow-up period post-QI program. © The Author(s) 2015.
    Annals of Pharmacotherapy 04/2015; 49(8). DOI:10.1177/1060028015582050 · 2.06 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: To determine whether addition of an electronic sepsis evaluation and management tool to electronic sepsis alerting improves compliance with treatment guidelines and clinical outcomes in septic ICU patients. A pragmatic randomized trial. Medical and surgical ICUs of an academic, tertiary care medical center. Four hundred and seven patients admitted during a 4-month period to the medical or surgical ICU with a diagnosis of sepsis established at the time of admission or in response to an electronic sepsis alert. Patients were randomized to usual care or the availability of an electronic tool capable of importing, synthesizing, and displaying sepsis-related data from the medical record, using logic rules to offer individualized evaluations of sepsis severity and response to therapy, informing users about evidence-based guidelines, and facilitating rapid order entry. There was no difference between the electronic tool (218 patients) and usual care (189 patients) with regard to the primary outcome of time to completion of all indicated Surviving Sepsis Campaign 6-hour Sepsis Resuscitation Bundle elements (hazard ratio, 1.98; 95% CI, 0.75-5.20; p = 0.159) or time to completion of each element individually. ICU mortality, ICU-free days, and ventilator-free days did not differ between intervention and control. Providers used the tool to enter orders in only 28% of available cases. A comprehensive electronic sepsis evaluation and management tool is feasible and safe but did not influence guideline compliance or clinical outcomes, perhaps due to low utilization.
    Critical care medicine 04/2015; Publish Ahead of Print(8). DOI:10.1097/CCM.0000000000001020 · 6.31 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Rationale Effective teamwork is fundamental to the management of medical emergencies and yet the best method to teach teamwork skills to trainees remains unknown. Objectives In a cohort of incoming internal medicine interns, we tested the hypothesis that expert demonstration of teamwork principles and participation in high-fidelity simulation would each result in objectively-assessed teamwork behavior superior to traditional didactics. Methods This was a randomized, controlled, parallel-group trial comparing three teamwork teaching modalities for incoming internal medicine interns. Participants in a single-day orientation at the Vanderbilt University Center for Experiential Learning and Assessment were randomized 1:1:1 to didactic, demonstration-based, or simulation-based instruction and then evaluated in their management of a simulated crisis by five independent, blinded observers using the Team Behavior Rater score. Clinical performance was assessed using the American Heart Association Advanced Cardiac Life Support algorithm and a novel "Recognize, Respond, Reassess" (3R) score. Measurements and Main Results Participants randomized to didactics (n=18), demonstration (n=17), and simulation (n=17) were similar at baseline. The primary outcome of average overall Team Behavior Rater score for those who received demonstration-based training was similar to simulation participation (4.40 ± 1.15 vs. 4.10 ± 0.95, p=0.917) and significantly higher than didactic instruction (4.40 ± 1.15 versus 3.10 ± 0.51, p=0.045). Clinical performance scores were similar between the three groups and correlated only weakly with teamwork behavior (Rs2=0.267, p < 0.001). Conclusions Among incoming internal medicine interns, teamwork training by expert demonstration resulted in similar teamwork behavior to participation in high-fidelity simulation and was more effective than traditional didactics. Clinical performance was largely independent of teamwork behavior and did not differ between training modalities.
    03/2015; 12(4). DOI:10.1513/AnnalsATS.201501-030OC

  • Annals of Pharmacotherapy 03/2015; 49(3):366-7. DOI:10.1177/1060028014564395 · 2.06 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Daily bathing of critically ill patients with the broad-spectrum, topical antimicrobial agent chlorhexidine is widely performed and may reduce health care-associated infections. To determine if daily bathing of critically ill patients with chlorhexidine decreases the incidence of health care-associated infections. A pragmatic cluster randomized, crossover study of 9340 patients admitted to 5 adult intensive care units of a tertiary medical center in Nashville, Tennessee, from July 2012 through July 2013. Units performed once-daily bathing of all patients with disposable cloths impregnated with 2% chlorhexidine or nonantimicrobial cloths as a control. Bathing treatments were performed for a 10-week period followed by a 2-week washout period during which patients were bathed with nonantimicrobial disposable cloths, before crossover to the alternate bathing treatment for 10 weeks. Each unit crossed over between bathing assignments 3 times during the study. The primary prespecified outcome was a composite of central line-associated bloodstream infections (CLABSIs), catheter-associated urinary tract infections (CAUTIs), ventilator-associated pneumonia (VAP), and Clostridium difficile infections. Secondary outcomes included rates of clinical cultures that tested positive for multidrug-resistant organisms, blood culture contamination, health care-associated bloodstream infections, and rates of the primary outcome by ICU. During the chlorhexidine bathing period, 55 infections occurred: 4 CLABSI, 21 CAUTI, 17 VAP, and 13 C difficile. During the control bathing period, 60 infections occurred: 4 CLABSI, 32 CAUTI, 8 VAP, and 16 C difficile. The primary outcome rate was 2.86 per 1000 patient-days during the chlorhexidine and 2.90 per 1000 patient-days during the control bathing periods (rate difference, -0.04; 95% CI, -1.10 to 1.01; P = .95). After adjusting for baseline variables, no difference between groups in the rate of the primary outcome was detected. Chlorhexidine bathing did not change rates of infection-related secondary outcomes including hospital-acquired bloodstream infections, blood culture contamination, or clinical cultures yielding multidrug-resistant organisms. In a prespecified subgroup analysis, no difference in the primary outcome was detected in any individual intensive care unit. In this pragmatic trial, daily bathing with chlorhexidine did not reduce the incidence of health care-associated infections including CLABSIs, CAUTIs, VAP, or C difficile. These findings do not support daily bathing of critically ill patients with chlorhexidine. Identifier: NCT02033187.
    JAMA The Journal of the American Medical Association 01/2015; 313(4). DOI:10.1001/jama.2014.18400 · 35.29 Impact Factor
  • Matthew W Semler · Arthur P Wheeler ·

    Chest 06/2014; 145(6):1181-2. DOI:10.1378/chest.14-0438 · 7.48 Impact Factor
  • Janna S Landsperger · Arthur P Wheeler ·

    Critical care medicine 03/2014; 42(3):731-2. DOI:10.1097/01.ccm.0000435683.06554.b0 · 6.31 Impact Factor
  • April N Kapu · Arthur P Wheeler · Byron Lee ·
    [Show abstract] [Hide abstract]
    ABSTRACT: Background Vanderbilt University Hospital's original rapid response team included a critical care charge nurse and a respiratory therapist. A frequently identified barrier to care was the time delay between arrival of the rapid response team and arrival of the primary health care team. Objective To assess the impact of adding an acute care nurse practitioner to the rapid response team. Methods Acute care nurse practitioners were added to surgical and medical rapid response teams in January 2011 to diagnose and order treatments on rapid response calls. Results In 2011, the new teams responded to 898 calls, averaging 31.8 minutes per call. The most frequent diagnoses were respiratory distress (18%), postoperative pain (13%), hypotension (12%), and tachyarrhythmia (10%). The teams facilitated 360 transfers to intensive care and provided 3056 diagnostic and therapeutic interventions. Communication with the primary team was documented on 97% of the calls. Opportunities for process improvement were identified on 18% of the calls. After implementation, charge nurses were surveyed, with 96% expressing high satisfaction associated with enhanced service and quality. Conclusions Teams led by nurse practitioners provide diagnostic expertise and treatment, facilitation of transfers, team communication, and education.
    Critical Care Nurse 02/2014; 34(1):51-9. DOI:10.4037/ccn2014847 · 1.56 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Barriers to executing large-scale randomized controlled trials include costs, complexity, and regulatory requirements. We hypothesized that source document verification (SDV) via remote electronic monitoring is feasible. Five hospitals from two NIH sponsored networks provided remote electronic access to study monitors. We evaluated pre-visit remote SDV compared to traditional on-site SDV using a randomized convenience sample of all study subjects due for a monitoring visit. The number of data values verified and the time to perform remote and on-site SDV was collected. Thirty-two study subjects were randomized to either remote SDV (N=16) or traditional on-site SDV (N=16). Technical capabilities, remote access policies and regulatory requirements varied widely across sites. In the adult network, only 14 of 2965 data values (0.47%) could not be located remotely. In the traditional on-site SDV arm, 3 of 2608 data values (0.12%) required coordinator help. In the pediatric network, all 198 data values in the remote SDV arm and all 183 data values in the on-site SDV arm were located. Although not statistically significant there was a consistent trend for more time consumed per data value (minutes +/- SD): Adult 0.50 +/- 0.17 min vs. 0.39 +/- 0.10 min (two-tailed t-test p=0.11); Pediatric 0.99 +/- 1.07 min vs. 0.56 +/- 0.61 min (p=0.37) and time per case report form: Adult: 4.60 +/- 1.42 min vs. 3.60 +/- 0.96 min (p=0.10); Pediatric: 11.64 +/- 7.54 min vs. 6.07 +/- 3.18 min (p=0.10) using remote SDV. Because each site had different policies, requirements, and technologies, a common approach to assimilating monitors into the access management system could not be implemented. Despite substantial technology differences, more than 99% of data values were successfully monitored remotely. This pilot study demonstrates the feasibility of remote monitoring and the need to develop consistent access policies for remote study monitoring.
    PLoS ONE 12/2013; 8(12):e81890. DOI:10.1371/journal.pone.0081890 · 3.23 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Introduction: Sepsis is common in both medical and surgical ICU patients and a better understanding of population-based course of sepsis coupled with the use of early protocolized care facilitated by an Integrated Electronic Management Tool, may lead to improved outcomes Methods: In this randomized trial, a sepsis tool, available via the EMR throughout the ICU stay in both medical and surgical ICUs, presented a divided display with one half graphing trend and goal for HR, MAP, CVP, Hgb, and lactate and the other half offering sepsis assessment and management tabs in a work-flow modeled on the Surviving Sepsis Campaign 2008 guidelines with one-click order entry for labs, cultures, imaging, antibiotics, and goal-directed resuscitation Results: Of 280 medical and 66 surgical patients with suspected sepsis, 164 (medical 133, surgical 31) were randomized to control vs 182 (medical 147, surgical 35) to the electronic sepsis management tool, The utilization of the management tool to enter orders based on best practice guidelines for the management of sepsis was 9.3% in the MICU vs. 27.3% in the SICU (p<0.001). 18% of the MICU patients came from an OSH and 71.2% of the SICU patients were post-operative with a higher acknowledgement of sepsis at admission. Source of sepsis was pulmonary in 29.6% of MICU patients vs. 11.5% of SICU patients, while an abdominal source of sepsis occurred in 54.1% of SICU patients and only 13% of MICU patients. ICU mortality in septic MICU patients was 16.5% vs. 4.5% in SICU Conclusions: This study highlights the population-based differences in medical and surgical sepsis with septic medical patients having significantly more pulmonary source of infection versus abdominal source for surgical patients. Surgical ICU patients had a higher acknowledgement of sepsis on admission to the ICU and increased utilization of the management tool to place outstanding orders, not already fulfilled in the Operative Room. The low overall utilization of the integrated electronic management tool in both units likely decreases its overall effectiveness.
    43rd Annual Critical Care Congress; 12/2013
  • Joanna Stollings · David Mitchell · Todd Rice · Arthur Wheeler ·
    [Show abstract] [Hide abstract]
    ABSTRACT: Introduction: The 2012 Surviving Sepsis Campaign guidelines recommend norepinephrine (NE) as the initial vasopressor followed by vasopressin (AVP) or epinephrine if needed to achieve a mean arterial pressure (MAP) >65 mm Hg. It is widely thought there is a "maximum" NE dose, a belief reinforced by pre-programmed "guardrail" NE infusion rates on some pumps. However, no convincing data indicate there should be a maximum NE dose or that multiple drugs at lower doses are better to higher dose NE. Methods: Our guidelines recommended: initial NE use without a maximum dose, followed by AVP if desired, and discontinuation of NE before stopping AVP. Monthly Pharmacist led education was provided to nurses and physician trainees. Using a pre- post-intervention design, we compared 50 prospective intervention patients (5/13-8/13) with retrospectively collected data from 50 controls (9/11 -3/13). Data are shown as mean (std) and proportions. Results: The number of non-AVP vasopressors was 1.2 (+0.48) in control and 1.2 (+0.45) in intervention patients respectively, (p=0.633). Goal MAP was obtained in 2.7h (+3.3) vs. 3.4h (+4.7) in control and intervention patients, respectively (p=0.39). The incidence of new arrhythmia after vasopressor initiation was 1 (2%) in controls and 4 (8%) in intervention patients (p=0.366). The maximum NE dose used in controls was 20 (+20) mcg/min vs. 40 (+63.1) mcg/min in intervention patients (p=0.035). However, the use of AVP increased non-significantly from 36% in controls to 48% following the intervention (p=0.311). Additionally, 8 patients in the control group received 3 and 1 patient received four different vasopressors while 6 patients in the intervention group received 3 vasopressors and 0 received 4 vasopressors. In the control group 28% of patients had NE weaned off before AVP compared to 29% in the intervention group (p=1.00). Conclusions: Despite increasing the maximum dose of NE used and increasing the use of AVP, our guidelines did not change overall number of vasoactive drugs used, time to reach goal MAP or the incidence of arrhythmias.
    43rd Annual Critical Care Congress; 12/2013
  • Supriya Srinivasa Gowda · Arthur Wheeler ·
    [Show abstract] [Hide abstract]
    ABSTRACT: Introduction: Over the years, Intensive Care Unit admissions have increased.The management of critically ill patients is multidisciplinary & requires utilization of lots of resources resulting in higher cost of care. Route of admissions to ICU seems to play a pivotal role in the resource utilization and outcome of patients admitted to ICU. Methods: Data extracted from electronic MICU admissions log, covering the period 2011 - 2012. Routes of admission were categorized as patients admitted Emergency Medicine (ER), In-Hospital Transfers (IHT) including patients from in-hospital ICUs, wards, operative room & patients transferred via Rapid Response Team( RRT) and Transfers from Outside (OSH) which included patients from other hospitals, nursing homes, home etc. Acute Physiologic Score (APS) was collected for each patient at the time of ICU admission. Results: Of 5612 admissions to Medical ICU, 2941(52%) were from ER with mean APS of 11.7, 1355 (24%) were from IHT with mean APs of 13.9, and the remaining 25% were from OSH with mean APS of 14.3. Average age at the time of MICU admission was 55 yrs and when compared against each group it was more or less similar. Females comprised of about 48% of overall admissions. Overall mortality was 11.9%, mortality rate among patients admitted from ER was 7.5 % as compared to 16.9% and 16.7% respectively among patients admitted from IHT & OSH. Conclusions: Even though, greater number of patients were admitted from ER to MICU, patients were relatively less sicker at the time of admission and mortality rate was also lower compared to patients admitted from other routes of admission.
    43rd Annual Critical Care Congress; 12/2013
  • [Show abstract] [Hide abstract]
    ABSTRACT: Introduction: Hyperglycemic emergencies (diabetic ketoacidosis [DKA] and hyperosmolar non-ketotic hyperglycemia [HNKH]) are commonly managed in ICUs. Increasingly non-physician providers, (acute care nurse practitioners [ACNPs]) are providing care independent of physician trainees. Methods: Retrospective chart review comparing characteristics, treatment and outcomes of patients with DKA and HNKH admitted to a tertiary university hospital MICU from 10/1/10 through 12/31/12. Data were collected using REDCAP and reported as mean +/- std or proportions. Results: 239 admissions for hyperglycemic emergencies occurred over 27 months. 56 of these admissions received partial treatment at an outside hospital prior to transfer to our institution and are excluded from this report. 152 patients, 49.7% females, averaging 37.7 +/- 16.4 years of age, were admitted 183 times (mean 0.2 admissions/day). In 8% of cases the admission was the initial presentation of diabetes. Noncompliance was the most common cause of admission (46%) followed by infection (22%). 82 of the 183 (45%) admissions were to the ACNP service. Resident service patients had a lower pH 7.13 +/- 0.15 vs. 7.18 +/- 0.15 (p=0.040) than ACNP service patients but similar APS scores 10.5 +/- 6.5 vs. 9.5 +/- 5.2 (p=0.265), anion gap 23.5 +/- 6.8 vs. 22.1 +/- 6.0 (p=0.136), creatinine 2.3 +/- 1.6 vs. 2.1 +/- 2.5 (p=0.474), and glucose 585 +/- 249 vs. 605 +/- 271 (p=0.611) on admission. Patients on both services received similar amounts of fluid within 6 hours (2.8 +/- 1.4 vs. 2.9 +/- 1.6 L, p=0.0821) and had similar durations of insulin infusion (9.8 +/- 7.6 vs. 10.2 +/- 6.9 h, p=0.710). ACNP service patients trended toward receiving less fluid in the first 24 hours (4.6 +/- 2.4 vs. 5.2 +/- 2.2 L, p=0.061), corrected their anion gap faster (6.6 +/- 5.7 vs. 9.6 +/- 7.6 h, p=0.004) and trended to a shorter ICU LOS (1.9 +/- 0.9 vs. 2.3 +/- 2.0 d, p=0.090). Time to anion gap closure for ACNP service patients remained significantly shorter in multivariate linear regression after correcting for age, APS, fluid administration, admission glucose and pH. There was no difference in the incidence of hypoglycemia. Conclusions: ACNPs provided quicker correction of anion gap and trended toward shorter ICU LOS. We conclude that overall, the ACNP and resident services provide comparable care to patients with diabetic emergencies.
    43rd Annual Critical Care Congress; 12/2013
  • [Show abstract] [Hide abstract]
    ABSTRACT: Introduction: Sepsis is common in ICU patients and despite a mortality benefit associated with early protocolized care, adherence to management guidelines remains inconsistent. Methods: In this randomized trial, patients admitted to the MICU and SICU with sepsis were randomized to usual care vs treatment with a novel electronic sepsis management tool. The sepsis tool, available via the EMR throughout the ICU admission, presented a divided display with one half graphing trend and goal for HR, MAP, CVP, Hgb, and lactate and the other half offering sepsis assessment and management tabs in a work-flow modeled on the Surviving Sepsis Campaign 2008 guidelines with one-click order entry for labs, cultures, imaging, antibiotics, and goal-directed resuscitation. Results: Of 346 patients, 164 were randomized to control and 182 to sepsis tool with no difference in age, gender, race, presence of sepsis on admission, or source of sepsis. Intention-to-treat analysis showed no difference in time to blood cultures ordered (2.54h vs. 3.82h, p=0.451), antibiotics administered (3.82h vs. 3.55h, p=0.585), CVP measured (6.00h vs. 5.61h, p=0.340), vasopressor ordered (4.84h vs. 5.25h, p=0.284), lactate measured (2.12h vs. 1.85h, p=0.472). There was no difference in percentage of patients who in the first 6h had blood cultures obtained, antibiotics administered, CVP measured, vasopressors ordered, or lactated measured. Per-protocol analysis of 44 patients on whom the tool was utilized to enter orders versus the remaining 302 patients showed no differences in age, gender, race, route of admission, sepsis at admission, or source of sepsis. There was no difference in the rate of or time to blood cultures, antibiotics, vasopressors, lactate measurement. Patients who had orders placed trended toward quicker CVP measurement (4.97h vs. 5.92h, p=0.123), higher rate of CVP in the first 6 hours (39.5% vs. 26.7%, p=0.081), with higher rate of CVP>=8 (34.1% vs. 19.9%, p=0.032). Conclusions: This study establishes the feasibility of instituting an integrated, electronic tool for sepsis management. Assessment of the effect on procedural and clinical outcomes is limited by low utilization of the tool.
    43rd Annual Critical Care Congress; 12/2013

Publication Stats

6k Citations
1,155.57 Total Impact Points


  • 1990-2015
    • Vanderbilt University
      • • Department of Medicine
      • • Division of Allergy, Pulmonary and Critical Care
      • • Center for Health Services Research
      • • Department of Biostatistics
      • • Department of Biomedical Engineering
      Нашвилл, Michigan, United States
  • 2007-2008
    • University of California, San Francisco
      • • Division of Nephrology
      • • Division of Hospital Medicine
      San Francisco, California, United States
  • 2005
    • Harvard University
      Cambridge, Massachusetts, United States
  • 2002-2005
    • Emory University
      • • Division of Pulmonary, Allergy and Critical Care Medicine
      • • School of Medicine
      Atlanta, GA, United States
  • 1995
    • University of Kentucky
      Lexington, Kentucky, United States