Group Health Research Institute (GHRI) hosts regular seminars where investigators from the Institute and our collaborating institutions present their research findings.
Upcoming scientific seminars and events
Thurs., April 17-Fri. April 18, 2014
2014 Latino Health Conference
Thurs., May 8, 2014
Breakfast and program 7:15-9 a.m., Sheraton Seattle Hotel
15th Annual Hilde and Bill Birnbaum Endowed Lecture and Panel Discussion
“Affordable health care for all: How will our nation and region deliver on the promise?”
Past seminars and events
Mon., Feb. 3, 2014
“Development and evaluation of prognostic models in chronic heart failure.”
Benjamin French, PhD, Assistant Professor, Biostatistics and Epidemiology, University of Pennsylvania
Abstract: Recent clinical research regarding chronic heart failure has focused on identifying prognostic models to predict future morbidity and mortality. Accurate models could be used to counsel patients more effectively and to guide personalized treatment strategies over time. Development of a prognostic model typically requires specification of an appropriate statistical model and is most frequently achieved using standard regression methods such as Cox regression. The prognostic model can be evaluated using time-dependent receiver operating characteristic methods, or risk reclassification methods adapted for censored survival outcomes. In this seminar, Dr. French will
- illustrate the application of these methods to derive a multi-biomarker risk score and evaluate its prediction accuracy for terminal events in chronic heart failure: death, cardiac transplantation, and placement of a ventricular assist device.
- compare the performance of alternative model accuracy methods using simulations, both to evaluate power and to quantify the potential loss in accuracy associated with use of a sub-optimal regression model to derive the multi-marker score.
- discuss ongoing research directions, including the development of new methods to evaluate prediction accuracy for recurrent events (e.g., hospitalization) and their linked outcomes (e.g., total cost and length of hospital stay).
Tues., Dec. 10, 2013
“Consistent estimation of covariate effects for some between-/within-cluster covariate decomposition methods when data are missing at random.”
John Neuhaus, PhD, Professor of Biostatistics, Division of Biostatistics, University of California, San Francisco
Abstract: Investigators often gather longitudinal data to assess changes in responses over time within subjects and to relate these changes to within-subject changes in predictors. Missing data are common in such studies and predictors can be correlated with subject-specific effects. Maximum likelihood methods for generalized linear mixed models provide consistent estimates when the data are “missing at random” (MAR) but can produce inconsistent estimates in settings where the random effects are correlated with one of the predictors. On the other hand, conditional maximum likelihood methods (and closely related maximum likelihood methods that partition covariates into between- and within-cluster components) provide consistent estimation when random effects are correlated with predictors but can produce inconsistent covariate effect estimates when data are MAR. Using theory, simulation studies, and fits to example data this talk shows that decomposition methods using complete covariate information produce consistent estimates. In some practical cases these methods, that ostensibly require complete covariate information, actually only involve the observed covariates. These results offer an easy-to-use approach to simultaneously protect against bias from either cluster-level confounding or MAR missingness in assessments of change.
Tues., Nov. 12, 2013
“Improving diagnosis and monitoring in primary care: new technologies, new methods.”
Matthew J. Thompson, MBChB, MPH, DPhil, MRCGP,
Helen D. Cohen Endowed Professorship in Family Medicine, Vice Chair for Research, Department of Family Medicine, University of Washington
Abstract: Diagnostic research has been an under researched area of primary care practice and is emerging from a focus on diagnostic accuracy to consider wider aspects and implications of testing. Meanwhile, many of the technological advances in diagnostics (speed, scope, size, availability, cost) have largely bypassed primary care in most countries. Dr. Thompson will be presenting results from a portfolio of research conducted in the UK at the Centre for Monitoring and Diagnosis in Oxford, which aims to improve the accuracy and efficiency of diagnosis and monitoring in primary care settings. Thompson plans to cover three main areas
- New diagnostic technologies in primary care—what to learn from prioritizing, assessing evidence, and conducting field studies across multiple clinical areas—where are the gaps in the ‘bench to bedside/clinic’ pathways, and why does the diagnostics industry need our help?
- Monitoring of chronic disease has emerged as a major component of workload in primary care, yet evidence on frequency and actions based on monitoring have not been evidence based. How can applying statistical processes to monitoring improve efficiency of care?
- An update on some new methodological issues in diagnostic research—including diagnostic systematic reviews, handling intermediate test results, and communicating diagnostic accuracy.
Tues., October 22, 2013
“Will the true BP (blood pressure) please stand up?”
Beverly B. Green, MD, MPH, Associate Investigator, GHRI and Family Physician, Group Health
Andrea Cook, PhD, Associate Investigator, GHRI
Melissa Anderson, MS, Biostatistician, GHRI
Abstract: The Electronic Communications and Home Blood Pressure Monitoring Trial (e-BP), four-year results, and differences in BP outcomes based on electronic health record data and research measurements.
The e-BP Comparative Effectiveness Trial compared home BP monitoring, this plus pharmacist team care delivered via secure e-mail to usual care and results were published in JAMA. The study was chosen by the British Medical Journal as a finalist for 1 of the 10 most important studies of 2008. We will discuss long-term results of the trial, methodological questions, and result implications.
Tues., October 8, 2013
“Impact of mandatory behavioral screening for children insured by Medicaid in Massachusetts.”
Robert Penfold, PhD, Assistant Investigator, GHRI
Abstract: In Massachusetts, as part of a settlement of a class action suit, Rosie D vs. Patrick, the state mandated behavioral health (BH) screening in primary care for all children up to 21 years of age who were covered by MassHealth (Medicaid). Clinicians were required to conduct behavioral health screening with validated tools at well-child visits. Use of a screen was to be reported using Current Procedural Terminology (CPT) code 96110 in billing claims for the visit, and providers were to receive payment for the procedure. While the recommendation was to screen at well-child visits, consistent with EPSDT regulations, screens were also reimbursed if used at non-well child visits.
We investigated the impact of the mandate on rates of newly identified behavioral health problems and rates of subsequent behavioral health utilization using logistic regression and interrupted time series analyses. Among continuously enrolled children with evidence of screening, 43 percent with positive screens had no BH history. This ‘newly identified’ group was more likely to be female, younger, minority, and from rural residences Factors predicting positive modifiers included, gender (male), age (older), foster care, BH history, and Hispanic ethnicity.
The mandate was also associated with dramatic increases in the rate of non-physician behavioral health care utilization. However, we observed no measurable impact on the rate of specialty mental health care use or use of psychotropic medications. There was also some indication that Emergency Department visits with a coded mental health diagnosis increased in the post-mandate period.
Tues., September 24, 2013
“Marginal structural modeling to estimate the effect of long-term physical activity on cardiovascular disease and mortality.”
Susan Shortreed, PhD, Assistant Investigator, Biostatistics Unit, GHRI
Abstract: The majority of studies estimating the effect of physical activity on cardiovascular disease (CVD) and mortality assess physical activity at a single time point. The impact of long-term physical activity is likely to differ. In this talk, Dr. Shortreed will
- review marginal structural models, which are used to estimate cumulative exposure effects from longitudinal data. This review will include a discussion of propensity scores.
- overview the differences between marginal structural models and traditional adjusted time-varying outcome models.
- describe and report the results of an analysis using data collected from the Framingham Heart Study. This analysis considered physical activity measured at three time points and used marginal structural modeling to estimate the effect of long-term physical activity on incident CVD, all-cause mortality and CVD-attributable mortality.
Weds., August 14, 2013
“The development, implementation, and evaluation of the Keele STarT BACK Tool for stratifying back pain patients in terms of clinical complexity.”
Nadine Foster, DPhil, Jonathan Hill, PhD, and Gail Sowden, MSc, Keele University, England
Abstract: Come hear about the creation, implementation, and evaluation of a paradigm-shifting approach to primary care for back pain developed by researchers at Keele University in England. Rather than focusing primarily on the degree of acuity/chronicity of back problems, the Keele model gives greater emphasis to consideration of the psycho-social factors that have consistently been found more predictive of patient outcomes than have clinical measures. In essence, this tool uses patient responses to nine simple questions to determine the level of each patient’s complexity and then uses this information to select a treatment option that is most appropriate for that complexity level. This approach has been found to improve patient outcomes and decrease costs of care in National Health Service general practices. Group Health is preparing to conduct the first evaluation of this risk stratification approach in a U.S. health care system.
Tues., July 23, 2013
“Individualizing prevention in older adults.”
Sei Lee, MD, MAS, Assistant Professor, Geriatrics, Department of Medicine, University of California, San Francisco and Physician, Division of Geriatrics, San Francisco VA Medical Center
Abstract: Preventive interventions, by definition, seek to prevent adverse events in the future. However, preventive interventions often impose risks on patients immediately. The time between the risks of preventive interventions (usually immediate) and the benefits is the “lagtime to benefit” and can be many years. For patients with limited life expectancy, preventive interventions may expose them to the risks, with little chance they would benefit. We will examine how to determine mortality risk in multimorbid older adults and quantify the lagtime to benefit for common preventive interventions to identify which patients are most appropriate for specific preventive interventions.
Tues., July 9, 2013
“Primary care transformation: Lessons from the Safety Net Medical Home Initiative.”
Katie Coleman, MSPH, Research Associate, MacColl Center for Healthcare Innovation, GHRI
Abstract: Interested in reducing health disparities? Improving patient health outcomes? Better coordinating care? So are we! Join us as we share lessons learned from our five-year effort to develop patient-centered medical homes in 65 safety-net practices across the country.
Tues., June 25, 2013
“Methodological challenges for sequential medical product safety surveillance using observational health care data.”
Jennifer Nelson, PhD, GHRI Associate Investigator
Abstract: Post-licensure drug and vaccine safety monitoring activities can range from uncovering new and unexpected adverse events (signal identification) to providing more conclusive evidence about a specific suspected adverse event (signal confirmation). Traditionally, safety signal identification has been accomplished by screening a large number of events (e.g., 100s or 1,000s) for potential association with many drug or vaccine exposures without analytic tailoring across different exposure-event pairs. In contrast, signal confirmation generally involves a detailed protocol-based epidemiological study that is specifically customized for a single or just a few exposure-event pairs. More recently, an intermediate type of evaluation has emerged that targets several (e.g., 5-10) pre-specified exposure-event hypotheses and conducts routine sequential monitoring over time, often during the initial uptake period of a new drug or vaccine. In this talk, Dr. Nelson will describe these newer surveillance systems, which include the Vaccine Safety Datalink and Mini-Sentinel pilot, and highlight the methodological challenges they face when prospectively monitoring drug and vaccine safety using longitudinal health care database information.
Thurs., May 16, 2013
“Fatal breast cancer risk in relation to use of unopposed estrogen and combined hormone therapy.”
Gaia Pocobelli, PhD, NRSA Institutional Research Training (T-32) Fellow, Group Health Research Institute
Abstract: Findings from meta-analyses of epidemiologic studies and a Women’s Health Initiative (WHI) randomized trial leave little doubt that use of combined hormone therapy (CHT) is associated with an increased risk of developing breast cancer. Less clear is to what degree, if at all, the increase in incidence translates into an increase in breast cancer mortality. Although the prevalence of use of hormone therapy has declined dramatically since the early 2000s, millions of women in the U.S. alone currently take unopposed estrogen hormone therapy (EHT) and CHT, which are effective at managing adverse symptoms of menopause. We conducted a large population-based nested case-control study in Saskatchewan, Canada, where a population-based prescription drug database has existed since 1975. Individual women’s records were linked across various population-based databases in addition to the prescription drug database, including the population registry, hospital and physician services databases, cancer registry, and the Screening Program for Breast Cancer database. In this seminar I will report our findings on fatal breast cancer risk in relation to recency and duration of use of CHT and EHT.
The “Topics in Women's Health Seminar and Discussion Series” is sponsored by Group Health Research Institute NRSA Institutional Research Training Grant (T32): “Health Care Improvement for Aging Women.”
Weds., May 15, 2013
2013 Cha Chi Ming Lecture
University of Washington School of Medicine “Acupuncture for low back pain: What’s the evidence?”
Karen Sherman, PhD, MPH, Senior Investigator, GHRI
Tues., May 14, 2013
“How should we study induction of labor? Methodological issues and associations with maternal and infant outcomes in California.”
Blair G. Darney, PhD, MPH
Post-Doctoral Fellow, Dept. of Medical Informatics and Clinical Epidemiology and Dept. of Obstetrics & Gynecology, Oregon Health & Science University
Abstract: Induction of labor is common but its impact on maternal and infant health outcomes is not clear. It remains difficult to synthesize the existing literature about the impacts of elective induction of labor on maternal and infant health outcomes. Studies use different approaches to defining indication without medical indication (elective induction), comparison groups, data sources, and focus on different gestational ages or outcomes. The lack of transparent, reproducible methods to classify inductions as medically indicated or not and to define appropriate comparison groups is a key contributor to the evidence gap about the health impacts of elective induction. This presentation will provide clinical and policy background on induction of labor, describe the key methods issues that complicate studying induction of labor, and present a case study analysis. Our analysis focuses on induction of labor without medical indication and expectant management at each term gestational week (37-40) in California. We use a transparent method to classify inductions as non-medically indicated and use the clinically relevant comparison group, expectant management. We stratify by gestational age and parity and test the association of induction without a medical indication and cesarean delivery (CD), operative vaginal delivery, 3rd or 4th degree perineal lacerations, macrosomia, and perinatal death.
Thurs., April 25, 2013
14th Annual Hilde and Bill Birnbaum Endowed Lecture and Panel Discussion
Tues., April 23, 2013
“The placebo effect in randomized controlled trials (RCT): What is it? How can RCT designs address it?”
Katherine M. Newton, PhD, Associate Director for Research and External Affairs and Senior Investigator, GHRI
Abstract: Where did the notion of placebo controlled trials arise? What is the placebo effect? What does a placebo control for? What is the value of placebo controls in randomized trials? What is the origin of the attention control group? What factors should be considered in choosing a control group? This seminar will address these and other issues related to the placebo effect in randomized clinical trials.
Tues., April 9, 2013
“Assessing emerging depression treatments: Describing the cat before it’s out of the bag.” Gregory Simon, MD, MPH, Senior Investigator, GHRI and Group Health Psychiatrist
Robert Penfold, PhD, Assistant Investigator, GHRI
Abstract: Effectiveness or pragmatic trials are often touted as the solution to the health care cost crisis: a rigorous method to evaluate whether newer (and more expensive) treatments actually deliver greater value. Experience to date, however, suggests that such trials cannot provide useful answers in time to guide practice or policy. Alternative methods are needed for rapidly assessing the effectiveness of emerging treatments or practices. This presentation will describe the promise and challenges of a surveillance system focused on early assessment of treatment effectiveness.
Tues., March 26, 2013
“Systems approaches to leverage colorectal cancer screening adherence and follow-up.”
Beverly Green, MD, MPH, Affiliate Investigator, GHRI and Group Health Family Physician
Tues., March 12, 2013
“Heterogeneity in action: The role of passive personalization in comparative effectiveness research.”
Anirban Basu, PhD, Associate Professor, Departments of Health Services and Economics, University of Washington and Director, Program in Health Economics and Outcomes Methodology
Abstract: Despite the goal of comparative effectiveness research (CER) to inform patient-centered care, most studies fail to account for the patient-centeredness of care that already exist in practice, which we denote as passive personalization (PP). Since CER studies describe the average effectiveness of treatments rather than heterogeneity in how individual patients respond to therapies, clinical or coverage policies that respond to CER results may undermine PP in clinical practice and generate worse outcomes. We study this phenomenon empirically in the context of use of antipsychotic drugs in Medicaid patients with schizophrenia using novel instrumental variable methods. We find strong support for PP in clinical practice and demonstrate that the average effects from a CER study cannot be replicated in practice due to the presence of PP. In contrast, providing physicians with evidence to further personalize treatment can produce significant benefits.
Tues., Feb. 26, 2013
“Interpretable patient-level predictive models.”
Tyler McCormick, PhD, Assistant Professor, Department of Statistics, University of Washington
Abstract: This talk presents statistical methods which generate patent-level predictions that are both accurate and highly interpretable to health care providers and patients. In this context, an interpretable model should be able to pinpoint exactly why a particular prediction was made, and provide the reason in a clear and natural way. The talk begins by introducing the Hierarchal Association Rule Model (HARM) which sequentially predicts a patient's possible future medical conditions given the patient's current and past history of reported conditions. The core of our technique is a Bayesian hierarchical model for selecting predictive association rules (such as “dyspepsia and epigastric pain imply heartburn”) from a large set of candidate rules. We next present a model for traditional classification problems based on decision lists, which consist of a series of if...then... statements (for example, if high blood pressure, then stroke). Decision lists discretize the high-dimensional, multivariate feature space into a series of simple, readily interpretable decision statements. Our Bayesian framework, known as the Bayesian List Machine (BLM), introduces a formal relationship between sparsity and interpretability through a prior structure over lists. We compare our model with the CHADS2 score, actively used in clinical practice for estimating the risk of stroke in patients that have atrial fibrillation. Our model is as interpretable as CHADS2, but more accurate. This is collaborative work with Cynthia Rudin, PhD; Ben Letham; and David Madigan, PhD.
Scientific seminars are held at the Metropolitan Park East Building (MPE), 1730 Minor Avenue, in Seattle. For directions to GHRI and contact information see our Contact Us page.