Stuart R Lipsitz

Massachusetts General Hospital, Boston, Massachusetts, United States

Are you Stuart R Lipsitz?

Claim your profile

Publications (463)2319.85 Total impact

  • [Show abstract] [Hide abstract]
    ABSTRACT: Trauma team training provides instruction on crisis management through debriefing and discussion of teamwork and leadership skills during simulated trauma scenarios. The effects of team leader's nontechnical skills (NTSs) on technical performance have not been thoroughly studied. We hypothesized that team's and team leader's NTSs correlate with technical performance of clinical tasks. Retrospective cohort study. Brigham and Women's Hospital, STRATUS Center for Surgical Simulation PARTICIPANTS: A total of 20 teams composed of surgical residents, emergency medicine residents, emergency department nurses, and emergency services assistants underwent 2 separate, high-fidelity, simulated trauma scenarios. Each trauma scenario was recorded on video for analysis and divided into 4 consecutive sections. For each section, 2 raters used the Non-Technical Skills for Surgeons framework to assess NTSs of the team. To evaluate the entire team's NTS, 2 additional raters used the Modified Non-Technical Skills Scale for Trauma system. Clinical performance measures including adherence to guidelines and time to perform critical tasks were measured independently. NTSs performance by both teams and team leaders in all NTS categories decreased from the beginning to the end of the scenario (all p < 0.05). There was significant correlation between team's and team leader's cognitive skills and critical task performance, with correlation coefficients between 0.351 and 0.478 (p < 0.05). The NTS performance of the team leader highly correlated with that of the entire team, with correlation coefficients between 0.602 and 0.785 (p < 0.001). The NTSs of trauma teams and team leaders deteriorate as clinical scenarios progress, and the performance of team leaders and teams is highly correlated. Cognitive NTS scores correlate with critical task performance. Increased attention to NTSs during trauma team training may lead to sustained performance throughout trauma scenarios. Decision making and situation awareness skills are critical for both team leaders and teams and should be specifically addressed to improve performance. Copyright © 2015 Association of Program Directors in Surgery. Published by Elsevier Inc. All rights reserved.
    Journal of Surgical Education 03/2015; DOI:10.1016/j.jsurg.2015.01.020 · 1.39 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Little is known about the sustainability and long-term effect of surgical safety checklists when implemented in resource-limited settings. A previous study demonstrated the marked, short-term effect of a structured hospital-wide implementation of a surgical safety checklist in Moldova, a lower-middle-income country, as have studies in other low-resource settings. To assess the long-term reduction in perioperative harm following the introduction of a checklist-based surgical quality improvement program in a resource-limited setting and to understand the long-term effects of such programs. Twenty months after the initial implementation of a surgical safety checklist and the provision of pulse oximetry at a referral hospital in Moldova, a lower-middle-income, resource-limited country in Eastern Europe, we conducted a prospective study of perioperative care and outcomes of 637 consecutive patients undergoing noncardiac surgery (the long-term follow-up group), and we compared the findings with those from 2106 patients who underwent surgery shortly after implementation (the short-term follow-up group). Preintervention data were collected from March to July 2010. Data collection during the short-term follow-up period was performed from October 2010 to January 2011, beginning 1 month after the implementation of the launch period. Data collection during the long-term follow-up period took place from May 25 to July 6, 2012, beginning 20 months after the initial intervention. The primary end points of interest were surgical morbidity (ie, the complication rate), adherence to safety process measures, and frequency of hypoxemia. Between the short- and long-term follow-up groups, the complication rate decreased 30.7% (P = .03). Surgical site infections decreased 40.4% (P = .05). The mean (SD) rate of completion of the checklist items increased from 88% (14%) in the short-term follow-up group to 92% (11%) in the long-term follow-up group (P < .001). The rate of hypoxemic events continued to decrease (from 8.1 events per 100 hours of oximetry for the short-term follow-up group to 6.8 events per 100 hours of oximetry for the long-term follow-up group; P = .10). Sustained use of the checklist was observed with continued improvements in process measures and reductions in 30-day surgical complications almost 2 years after a structured implementation effort that demonstrated marked, short-term reductions in harm. The sustained effect occurred despite the absence of continued oversight by the research team, indicating the important role that local leadership and local champions play in the success of quality improvement initiatives, especially in resource-limited settings.
    JAMA SURGERY 03/2015; DOI:10.1001/jamasurg.2014.3848 · 4.30 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Failure-to-rescue (FTR or death after postoperative complication) is thought to explain surgical mortality excesses across hospitals, and FTR is an emerging performance measure and target for quality improvement. We compared the FTR population to preoperatively identifiable subpopulations for their potential to close the mortality gap between lowest- and highest-mortality hospitals. Patients undergoing small bowel resection, pancreatectomy, colorectal resection, open abdominal aortic aneurysm repair, lower extremity arterial bypass, and nephrectomy were identified in the 2007 to 2011 Nationwide Inpatient Sample. Lowest- and highest-mortality hospitals were defined using risk- and reliability-adjusted mortality quintiles. Five target subpopulations were established a priori: the FTR population, predicted high-mortality risk (predicted highest-risk quintile), emergency surgery, elderly (>75 years old), and diabetic patients. Across the lowest mortality quintile (n = 282 hospitals, 56,893 patients) and highest-mortality quintile (282 hospitals, 45,784 patients), respectively, the size of target subpopulations varied only for the FTR population (20.2% vs 22.4%, p = 0.002) but not for other subpopulations. Variation in mortality rates across lowest- and highest-mortality hospitals was greatest for the high-mortality risk (7.5% vs 20.2%, p < 0.0001) and FTR subpopulations (7.8% vs 18.9%, p < 0.0001). The FTR and high-risk populations had comparable sensitivity (81% and 75%) and positive predictive value (19% and 20%, respectively) for mortality. In Monte Carlo simulations, the mortality gap between the lowest- and highest-mortality hospitals was reduced by nearly 75% when targeting the FTR population or the high-risk population, 78% for the emergency surgery population, but less for elderly (51%) and diabetic (17%) populations. Preoperatively identifiable patients with high estimated mortality risk may be preferable to the FTR population as a target for surgical mortality reduction. Copyright © 2015 American College of Surgeons. Published by Elsevier Inc. All rights reserved.
    Journal of the American College of Surgeons 03/2015; DOI:10.1016/j.jamcollsurg.2015.02.036 · 4.45 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Unlike majority of current statistical models and methods focusing on mean response for highly skewed longitudinal data, we present a novel model for such data accommodating a partially linear median regression function, a skewed error distribution and within subject association structures. We provide theoretical justifications for our methods including asymptotic properties of the posterior and associated semiparametric Bayesian estimators. We also provide simulation studies to investigate the finite sample properties of our methods. Several advantages of our method compared with existing methods are demonstrated via analysis of a cardiotoxicity study of children of HIV-infected mothers. © The Author 2015. Published by Oxford University Press. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
    Biostatistics 03/2015; DOI:10.1093/biostatistics/kxv005 · 2.24 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The test of independence of row and column variables in a (J×K) contingency table is a widely used statistical test in many areas of application. For complex survey samples, use of the standard Pearson chi-squared test is inappropriate due to correlation among units within the same cluster. Rao and Scott (1981, Journal of the American Statistical Association 76, 221-230) proposed an approach in which the standard Pearson chi-squared statistic is multiplied by a design effect to adjust for the complex survey design. Unfortunately, this test fails to exist when one of the observed cell counts equals zero. Even with the large samples typical of many complex surveys, zero cell counts can occur for rare events, small domains, or contingency tables with a large number of cells. Here, we propose Wald and score test statistics for independence based on weighted least squares estimating equations. In contrast to the Rao-Scott test statistic, the proposed Wald and score test statistics always exist. In simulations, the score test is found to perform best with respect to type I error. The proposed method is motivated by, and applied to, post surgical complications data from the United States' Nationwide Inpatient Sample (NIS) complex survey of hospitals in 2008. © 2015, The International Biometric Society.
    Biometrics 03/2015; DOI:10.1111/biom.12297 · 1.52 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Bernoulli (or binomial) regression using a generalized linear model with a log link function, where the exponentiated regression parameters have interpretation as relative risks, is often more appropriate than logistic regression for prospective studies with common outcomes. In particular, many researchers regard relative risks to be more intuitively interpretable than odds ratios. However, for the log link, when the outcome is very prevalent, the likelihood may not have a unique maximum. To circumvent this problem, a 'COPY method' has been proposed, which is equivalent to creating for each subject an additional observation with the same covariates except the response variable has the outcome values interchanged (1's changed to 0's and 0's changed to 1's). The original response is given weight close to 1, while the new observation is given a positive weight close to 0; this approach always leads to convergence of the maximum likelihood algorithm, except for problems with convergence due to multicollinearity among covariates. Even though this method produces a unique maximum, when the outcome is very prevalent, and/or the sample size is relatively small, the COPY method can yield biased estimates. Here, we propose using the jackknife as a bias-reduction approach for the COPY method. The proposed method is motivated by a study of patients undergoing colorectal cancer surgery. Copyright © 2014 John Wiley & Sons, Ltd.
    Statistics in Medicine 02/2015; 34(3). DOI:10.1002/sim.6348 · 2.04 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: To identify the timing and relative frequency of common postoperative complications in a contemporary, diverse surgical population and develop a mnemonic for teaching and clinical decision support. We enrolled a cohort of general and vascular surgical patients undergoing elective, inpatient surgery in the American College of Surgeons National Surgical Quality Improvement Program database between 2005 and 2011. Index complications were noted by postoperative day (POD). Timing and incidence were compared within each day. Among 614,525 patients, 51,173 (9.88%) experienced the following index complications over 30 days: pneumonia (n = 5947), urinary tract infection (n = 9459), superficial surgical site infection (sSSI) (n = 20,460), deep/organ space surgical site infection (dSSI) infection (n = 11,847), venous thromboembolism (n = 4478), kidney injury (n = 2620), and myocardial infarction (n = 1813). Median time to complication differed significantly for index complications (p < 0.0001). On POD 0, the most common complication was myocardial infarction (incidence 4.26/10,000 patient days; 95% CI: 3.75-4.78). On POD 1 and 2, pneumonia was the most common complication, with peak incidence on POD 2 (20.36; 95% CI: 19.22-21.51). On POD 3, pneumonia (16.3; 95% CI: 15.27-17.33) and urinary tract infection (15.5; 95% CI: 14.49-16.51) were significantly more common than other complications. On POD 4, the most common complication was sSSI (16.24; 95% CI: 15.20-17.28). From POD 5 to POD 30, sSSI and dSSI were the 2 most common complications. Risk of venous thromboembolism declined only slightly through POD 30. We propose a mnemonic for postoperative complication timing and frequency, independent of fever, as follows: Waves (myocardial infarction), Wind (pneumonia), Water (urinary tract), Wound (sSSI and dSSI), and Walking (venous thromboembolism) in the order of likelihood. Copyright © 2014 Association of Program Directors in Surgery. Published by Elsevier Inc. All rights reserved.
    Journal of Surgical Education 12/2014; DOI:10.1016/j.jsurg.2014.11.004 · 1.39 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Blood pressure (BP) control rates are suboptimal. We evaluated the effectiveness of 2 behavioral interventions to improve BP control via a 3-arm, randomized controlled trial of 533 adults with repeated uncontrolled BP, despite antihypertensive drug treatment for ≥6 months. The interventions were a tailored stage-matched intervention (SMI) or a nontailored health education intervention (HEI) of 6 monthly calls targeting diet, exercise, and medication. Control was usual care (UC). There were no baseline group differences. Baseline BP control was 42.6%, 40.6%, and 44.6% in SMI, HEI, and UC (P=0.74), respectively; systolic BP (with SEs) was 136 (0.89), 137 (1.33), and 137 (0.96) mm Hg. Six-month control was 64.6% (SMI), 54.3% (HEI), and 45.8% (UC) (P values for pairwise comparisons versus UC, 0.001 [SMI] and 0.108 [HEI]). At 6 months, systolic BP (SE) was 131.2 (1.05), 131.8 (0.99), and 134.7 (1.02) for SMI, HEI, and UC, respectively (P values for pairwise comparisons versus UC, 0.009 for SMI and 0.047 for HEI). SMI led to lower systolic BP and better BP control than UC. SMI constitutes a new, potent approach to assist patients with uncontrolled hypertension to reach BP goals. © 2014 American Heart Association, Inc.
    Hypertension 11/2014; 65(2). DOI:10.1161/HYPERTENSIONAHA.114.03483 · 7.63 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Miscommunications are a leading cause of serious medical errors. Data from multicenter studies assessing programs designed to improve handoff of information about patient care are lacking.
    New England Journal of Medicine 11/2014; 371(19):1803-12. DOI:10.1056/NEJMsa1405556 · 54.42 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Our purpose was to provide a metric by which evaluation criteria are prioritized during resident selection. In this study, we assessed which residency applicant qualities are deemed important by members of the American Association of Plastic Surgeons (AAPS).
    Journal of Surgical Education 09/2014; DOI:10.1016/j.jsurg.2014.07.013 · 1.39 Impact Factor
  • Journal of the American College of Surgeons 09/2014; 219(3):S112-S113. DOI:10.1016/j.jamcollsurg.2014.07.266 · 4.45 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: The relationship between timing of postoperative complications on mortality is unknown. We investigated the time-variable mortality risks of common surgical complications.
    Journal of Surgical Research 08/2014; DOI:10.1016/j.jss.2014.08.025 · 2.12 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Audits provide a rational framework for quality improvement by systematically assessing clinical practices against accepted standards with the aim to develop recommendations and interventions that target modifiable deficiencies in care. Most childbirth-associated mortality audits in developing countries are focused on a single facility and, up to now, the avoidable factors in maternal and perinatal deaths cataloged in these reports have not been pooled and analyzed. We sought to identity the most frequent avoidable factors in childbirth-related deaths globally through a systematic review of all published mortality audits in low and lower-middle income countries.
    BMC Pregnancy and Childbirth 08/2014; 14(1):280. DOI:10.1186/1471-2393-14-280 · 2.15 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Introduction Invasive procedures are resource-intense and may be associated with substantial morbidity. These harms must be carefully balanced with the benefits gained in life expectancy and quality of life. Prior research has demonstrated an increasing aggressiveness of care in cancer patients at the end of life. To better characterize surgical care in this setting, we sought to examine trends in the use of invasive procedures in patients diagnosed with metastatic cancer on presentation. Materials & Methods Using SEER-Medicare data, we identified invasive procedure claims from 1994-2009 for patients diagnosed with incident stage IV breast, colorectal, lung, and prostate cancer patients in 1995-2006. We grouped procedures into surgically relevant categories, using an adaptation of the Clinical Classifications Software (CCS), and measured utilization and relative changes over time. Results Of stage IV patients diagnosed in 2002-2006, 96% underwent a procedure during the course of their cancer care including 63% after the diagnostic period, and 25% in the last month of life. Between 1996 and 2006, minimal change was observed in utilization during the diagnostic period (+1.5%). However, there were significant increases during continuing care (+20.7%) and the last month of life (+21.5%). Procedures consistent with primary tumor resection decreased, while those with probable palliative intent and those unrelated to cancer increased. Conclusions Nearly all patients who present with metastatic cancer undergo invasive procedures. While overall utilization is increasing, the specific procedure types indicate that it may be appropriate, enhancing the quality of life in this vulnerable population.
    Journal of Surgical Research 08/2014; 193(2). DOI:10.1016/j.jss.2014.08.021 · 2.12 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Purpose Although demographic, clinicopathologic, and socioeconomic differences may affect treatment and outcomes of prostate cancer, the effect of mental health disorders remains unclear. We assessed the effect of previously diagnosed depression on outcomes of men with newly diagnosed prostate cancer. Patients and Methods We performed a population-based observational cohort study using Surveillance, Epidemiology, and End Results-Medicare linked data of 41,275 men diagnosed with clinically localized prostate cancer from 2004 to 2007. We identified 1,894 men with a depressive disorder in the 2 years before the prostate cancer diagnosis and determined its effect on treatment and survival. Results Men with depressive disorder were older, white or Hispanic, unmarried, resided in nonmetropolitan areas and areas of lower median income, and had more comorbidities (P < .05 for all), but there was no variation in clinicopathologic characteristics. In adjusted analyses, men with depressive disorder were more likely to undergo expectant management for low-, intermediate-, and high-risk disease (P <= .05, respectively). Conversely, depressed men were less likely to undergo definitive therapy (surgery or radiation) across all risk strata (P < .01, respectively). Depressed men experienced worse overall mortality across risk strata (low: relative risk [RR], 1.86; 95% CI, 1.48 to 2.33; P < .001; intermediate: RR, 1.25; 95% CI, 1.06 to 1.49; P = .01; high: RR, 1.16; 95% CI, 1.03 to 1.32; P = .02). Conclusion Men with intermediate-or high-risk prostate cancer and a recent diagnosis of depression are less likely to undergo definitive treatment and experience worse overall survival. The effect of depression disorders on prostate cancer treatment and survivorship warrants further study, because both conditions are relatively common in men in the United States. (C) 2014 by American Society of Clinical Oncology
    Journal of Clinical Oncology 07/2014; 32(23). DOI:10.1200/JCO.2013.51.1048 · 17.88 Impact Factor
  • Garrett M Fitzmaurice, Stuart R Lipsitz
    Biostatistics 07/2014; DOI:10.1093/biostatistics/kxu033 · 2.24 Impact Factor
  • 06/2014; 149(8). DOI:10.1001/jamasurg.2014.782
  • [Show abstract] [Hide abstract]
    ABSTRACT: For either the equivalence trial or the non-inferiority trial with survivor outcomes from two treatment groups, the most popular testing procedure is the extension (e.g., Wellek, A log-rank test for equivalence of two survivor functions, Biometrics, 1993; 49: 877-881) of log-rank based test under proportional hazards model. We show that the actual type I error rate for the popular procedure of Wellek is higher than the intended nominal rate when survival responses from two treatment arms satisfy the proportional odds survival model. When the true model is proportional odds survival model, we show that the hypothesis of equivalence of two survival functions can be formulated as a statistical hypothesis involving only the survival odds ratio parameter. We further show that our new equivalence test, formulation, and related procedures are applicable even in the presence of additional covariates beyond treatment arms, and the associated equivalence test procedures have correct type I error rates under the proportional hazards model as well as the proportional odds survival model. These results show that use of our test will be a safer statistical practice for equivalence trials of survival responses than the commonly used log-rank based tests.
    Statistical Methods in Medical Research 06/2014; DOI:10.1177/0962280214539282 · 2.96 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Often, the reader of a published paper is interested in a comparison of parameters that has not been presented. It is not possible to make inferences beyond point estimation since the standard error for the contrast of the estimated parameters depends upon the (unreported) correlation. This study explores approaches to obtain valid confidence intervals when the correlation [Formula: see text] is unknown. We illustrate three proposed approaches using data from the National Health Interview Survey. The three approaches include the Bonferroni method and the standard confidence interval assuming [Formula: see text] (most conservative) or [Formula: see text] (when the correlation is known to be non-negative). The Bonferroni approach is found to be the most conservative. For the difference in two estimated parameter, the standard confidence interval assuming [Formula: see text] yields a 95% confidence interval that is approximately 12.5% narrower than the Bonferroni confidence interval; when the correlation is known to be positive, the standard 95% confidence interval assuming [Formula: see text] is approximately 38% narrower than the Bonferroni. In summary, this article demonstrates simple methods to determine confidence intervals for unreported comparisons. We suggest use of the standard confidence interval assuming [Formula: see text] if no information is available or [Formula: see text] if the correlation is known to be non-negative.
    PLoS ONE 05/2014; 9(5):e98498. DOI:10.1371/journal.pone.0098498 · 3.53 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Repeated hospitalizations are frequent toward the end of life, where each admission should be an opportunity to initiate advance-care planning to high-risk patients. To identify the risk factors for having a 30-day potentially avoidable readmission due to end-of-life care issues among all medical patients. Nested case-control study. All 10,275 consecutive discharges from any medical service of an academic tertiary medical center in Boston, Massachusetts between July 1, 2009 and June 30, 2010. A random sample of all the potentially avoidable 30-day readmissions was independently reviewed by 9 trained physicians to identify the ones due to end-of-life issues. Among 534, 30-day potentially avoidable readmission cases reviewed, 80 (15%) were due to an end-of-life care issue. In multivariable analysis, the following risk factors were significantly associated with a 30-day potentially avoidable readmission due to end-of-life care issues: number of admissions in the previous 12 months (odds ratio [OR]: 1.10 per admission, 95% confidence interval [CI]: 1.02-1.20), neoplasm (OR: 5.60, 95% CI: 2.85-10.98), opiate medications at discharge (OR: 2.29, 95% CI: 1.29-4.07), Elixhauser comorbidity index (OR: 1.16 per 5-point increase, 95% CI: 1.10-1.22). The discrimination of the model (C statistic) was 0.85. In a medical population, we identified 4 main risk factors that were significantly associated with 30-day potentially avoidable readmission due to end-of-life care issues, producing a model with very good to excellent discrimination. Patients with these risk factors might benefit from palliative care consultation prior to discharge in order to improve end-of-life care and possibly reduce unnecessary rehospitalizations. Journal of Hospital Medicine 2014;. © 2014 Society of Hospital Medicine.
    Journal of Hospital Medicine 05/2014; 9(5). DOI:10.1002/jhm.2173 · 2.08 Impact Factor

Publication Stats

13k Citations
2,319.85 Total Impact Points

Institutions

  • 2015
    • Massachusetts General Hospital
      • Department of Surgery
      Boston, Massachusetts, United States
  • 2005–2015
    • Brigham and Women's Hospital
      • • Department of Medicine
      • • Department of Surgery
      • • Center for Surgery and Public Health
      Boston, Massachusetts, United States
    • New York University
      New York City, New York, United States
    • University of South Florida St. Petersburg
      St. Petersburg, Florida, United States
  • 2014
    • The University of Chicago Medical Center
      • Section of Urology
      Chicago, Illinois, United States
  • 1992–2014
    • Harvard University
      • Department of Health Policy and Management
      Cambridge, Massachusetts, United States
  • 1990–2014
    • Harvard Medical School
      • Department of Surgery
      Boston, Massachusetts, United States
  • 2012
    • St. Joseph Hospital, Orange
      Orange, California, United States
  • 1992–2012
    • Dana-Farber Cancer Institute
      • • Lank Center for Genitourinary Oncology
      • • Department of Biostatistics and Computational Biology
      Boston, Massachusetts, United States
  • 2011
    • Cedars-Sinai Medical Center
      • Cedars Sinai Medical Center
      Los Ángeles, California, United States
    • Georgetown University
      • Department of Urology
      Washington, Washington, D.C., United States
  • 2010–2011
    • Beverly Hospital, Boston MA
      BVY, Massachusetts, United States
    • University of Miami Miller School of Medicine
      • Department of Pediatrics
      Miami, Florida, United States
  • 2003–2011
    • Emory University
      • Goizueta Business School
      Atlanta, GA, United States
  • 1999–2010
    • Medical University of South Carolina
      • • Department of Medicine
      • • Division of Biostatistics and Epidemiology
      Charleston, South Carolina, United States
  • 2009
    • CUNY Graduate Center
      New York, New York, United States
    • Boston Children's Hospital
      Boston, Massachusetts, United States
  • 2008–2009
    • Partners HealthCare
      Boston, Massachusetts, United States
  • 2007
    • Texas A&M University
      • Department of Statistics
      College Station, Texas, United States
  • 1999–2005
    • University of Rochester
      • Division of Pediatric Cardiology
      Rochester, New York, United States
  • 2002
    • University of North Carolina at Chapel Hill
      • Department of Biostatistics
      North Carolina, United States
  • 1990–1998
    • Massachusetts Department of Public Health
      Boston, Massachusetts, United States