Health Policy Statistics Section
Session Slot: 8:30-10:20 Monday
Estimated Audience Size: 75-125
AudioVisual Request: None
Session Title: Multiple Measurements/Multiple Levels in Health Care Research
Theme Session: Yes
Applied Session: Yes
Session Organizer: Christiansen, Cindy L. Harvard Medical School and Harvard Pilgrim Health Care
Address: Harvard Medical School and Harvard Pilgrim Health Care, 126 Brookline Ave., Suite 200, Boston, MA 02215
Phone: (617) 421-2430
Fax: (617) 859-8112
Email: cindy@hphc.org
Session Timing: 110 minutes total (Sorry about format):
3 speakers and 1 discussant Opening Remarks by Chair - 3 minutes First Speaker - 30 minutes Second Speaker - 30 minutes Third Speaker - 30 minutes Discussant - 10 minutes Floor Discusion - 7 minutes
Session Chair: McCaffrey, Dan RAND
Address: RAND, 1700 Main St., PO Box 2138, Santa Monica, CA 90407-2138
Phone: (310) 393-0411 x6735
Fax: (310) 451-7004
Email: daniel_mccaffrey@rand.org
1. Methods for the Analysis of Signal Data in Medicine
Land, Stephanie, Carnegie Mellon University
Address: Department of Statistics Carnegie Mellon University Pittsburgh, PA 15213
Phone: (412) 268-1880
Fax: (412) 268-7828
Email: steph@stat.cmu.edu
Abstract: Medical researchers often need to relate signal data (e.g. analog time series or frequently measured time series) to covariates such as the health or treatment status of patients. This talk will focus on regression estimation and inference for such problems.
2. Analysis of hospital quality monitors using hierarchical time series models
Aguilar, Omar, ISDS Duke University
Address: Duke Univeristy Box 90251, Durham NC, 27708
Phone: 919-684-8840
Fax: 919-684-8594
Email: omar@stat.duke.edu
West, Mike, ISDS Duke University
Abstract: The VA management services department invests considerably in the collection and assessment of data to inform on hospital and care-area specific levels of quality of care. Resulting time series of ``quality monitors'' provide information relevant to evaluating patterns of variability in hospital-specific quality of care over time and across care areas, and to compare and assess differences across hospitals. In collaboration with the VA management services group we have developed various models for evaluating such patterns of dependencies and combining data across the VA hospital system. This paper provides a brief overview of resulting models, some summary examples on three monitor time series, and discussion of data, modelling and inference issues. This work introduces new models for multivariate non-Gaussian time series. The framework combines cross-sectional, hierarchical models of the population of hospitals with time series structure to allow and measure time-variations in the associated hierarchical model parameters. In the VA study, the within-year components of the model describe patterns of heterogeneity across the population of hospitals and relationships among several such monitors, while the time series components describe patterns of variability through time in hospital-specific effects and their relationships across quality monitors. Additional model components isolate unpredictable aspects of variability in quality monitor outcomes, by hospital and care areas. We discuss model assessment, residual analysis and MCMC algorithms developed to fit these models, which will be of interest in related applications in other socio-economic areas.
3. Analyzing patients' satisfaction with care using multi-level random-effects models
Hui, Siu L., Indiana University School of Medicine
Address: Division of Biostatistics Indiana University School of Medicine Riley Research Wing, 135 699 West Dr. Indianapolis, IN 46202-5119
Phone: 317-274-2661
Fax: 317-274-2686
Email: siuhui@mako.biostat.iupui.edu
Zhou, Xiao-Hua (Andrew), Indiana University School of Medicine
Perkins, Tony, Regenstrief Institute for Health Care
Tierney, Williams M., Regenstrief Institute for Health Care
Abstract: Since patient satisfaction has become more and more an important indicator of health care quality, in this report, we studied patients' satisfaction among elderly sick patients in an academic outpatient clinic. As part of a study of advance directives in an academic outpatient clinic, we interviewed a cohort of more than a thousand elderly patients, many with significant chronic illness, after scheduled appointments with their primary care physicians. Each patient was followed for one year, and after each scheduled visit, the patient was interviewed by a research assistant who administered two patient satisfaction questionnaires: (1) the ten-item instrument created by the American Board of Internal Medicine (ABIM), which evaluates satisfaction with doctor-patient communication and (2) the seven-item questionnaire from the Medical Outcomes Study (MOS), which assesses satisfaction with the visits just completed. We were interested in identifying clinical factors that might affect patients' satisfaction with care. Due to the hierarchical nature of the data and multidimensional nature of patients' satisfaction with care, we considered covariates at three levels, visit, patient, and doctor. We proposed the use of multi-level random-effects regression models to study effects of three-level covariates on patients' satisfaction scores.
Discussant: Katz, Barry Indiana University School of Medicine
Address: Division of Biostatistics Indiana University School of Medicine Riley Research WING, 135 699 West Dr. Indianapolis, IN 46202-5119
Phone: 317-274-2674
Fax: 317-274-2678
Email: katz@mako.biostat.iupui.edu
List of speakers who are nonmembers: None