Statistical Innovations in Health Services Research
Principal Investigator: Joseph C Gardiner
Abstract: DESCRIPTION (Provided by Applicant): Studies of healthcare utilization from national and state administrative databases have been stymied by a lack of powerful methodological approaches to circumvent deficiency of specificity often common in these databases. New statistical and econometric techniques are needed to uncover hidden structures in healthcare utilization data. This research plan addresses the development and application of innovative regression models for analyzing jointly healthcare utilization and outcomes from clinical and observational studies. All models incorporate observed heterogeneity due to patient and system characteristics. In addition, we explicitly acknowledge unobserved heterogeneity due to omitted variables, endogeneity of explanatory variables, and censoring in outcome measures. Specifically, we develop and test several statistical models to analyze measures of utilization (eg, length of stay, cost) and health outcomes (eg, survival, quality of life). The complexity of the models depends upon the coarseness of the available data (eg, longitudinal, cross-sectional, hierarchical) and richness in observed covariates. Markov models are used with longitudinal data to account for the dynamics of movement of patients between health states (eg, relapse, remission) with covariate effects incorporated in transition intensities through multiplicative intensity and proportional hazards models. Heterogeneity due to unobserved or omitted variables is accommodated through random effects, frailties, and latent class models. We use Coxian phase-type models to elicit hidden Markov structures in cross-sectional data on healthcare utilization. Hierarchical models are applied to accommodate complex sampling designs and clustering (eg, patients within hospitals, hospitals within geographic units). All models will be rigorously tested in simulation and cross-validation studies. Application of our methods will be demonstrated in three studies with healthcare utilization and outcomes. (1) Using the Nationwide Inpatient Sample we jointly estimate total hospital charge and length of stay associated with procedures for two broad disease categories, heart disease and cancer. (2) Using a linked data set of Michigan Medicare, Medicaid and Cancer &Death Certificate Registries, we estimate the cost of treatment and survival in patients with colon, breast, lung and prostate cancer, controlling for observed covariates and unobserved heterogeneity. (3) Using a proprietary longitudinal data set of patient functioning while undergoing cancer treatments, we estimate costs of care, survival and physical function jointly, and assess the impact of changes in physical function on cost and survival. By expanding the repertoire of analytic tools for health services researchers, this project will provide methods for extracting valuable information on healthcare utilization and outcomes from administrative databases that can be used to inform cost-effectiveness analyses and health policy. By expanding the repertoire of analytic tools for health services researchers, this project will provide methods for extracting valuable information on healthcare utilization and outcomes from administrative databases that can be used to inform cost-effectiveness analyses and health policy
Funding Period: 2004-05-01 - 2010-09-29
more information: NIH RePORT
- Statin therapy in the reduction of cardiovascular events in patients undergoing intermediate-risk noncardiac, nonvascular surgeryManjunath G Raju
Department of Cardiovascular Medicine, Cleveland Clinic Foundation, Cleveland, Ohio Section of Vascular Medicine, Cleveland Clinic Foundation, Cleveland, Ohio
Clin Cardiol 36:456-61. 2013..However, there is paucity of data on the role of statins in patients undergoing intermediate-risk noncardiac, nonvascular surgery (NCNVS)...
- Maternal perceptions of help from home visits by nurse-community health worker teamsLee Anne Roman
Department of Obstetrics, Gynecology, and Reproductive Biology, College of Human Medicine, Michigan State University, East Lansing, 48824, USA
Am J Public Health 102:643-5. 2012..For both groups, assistance with health education ranked highest among the types of assistance received. A higher percentage of women in the nurse-CHW group than the CC group reported that they received psychosocial help...
- Modeling hospital length of stay by Coxian phase-type regression with heterogeneityXiaoqin Tang
Center for Health Research, Geisinger Health System, Danville, PA, USA
Stat Med 31:1502-16. 2012..The methodology is illustrated with application to hospital admissions for acute myocardial infarction in the 2003 Nationwide Inpatient Sample from the Healthcare Utilization Project...
- A nonparametric test for equality of survival mediansMohammad H Rahbar
Division of Epidemiology, Human Genetics and Environmental Sciences, The University of Texas School of Public Health at Houston, Houston, TX, USA
Stat Med 31:844-54. 2012..In these situations, our test statistic provides an alternative to the Brookmeyer-Crowley test...
- Reduction in the intensity rate of appropriate shocks for ventricular arrhythmias with statin therapyAbhimanyu Beri
Division of Internal Medicine, Michigan State University, East Lansing, MI 48824, USA
J Cardiovasc Pharmacol 56:190-4. 2010..Larger randomized trials are needed to confirm this relationship...
- Applying propensity score methods in medical research: pitfalls and prospectsZhehui Luo
Department of Epidemiology, Michigan State University, East Lansing, MI, USA
Med Care Res Rev 67:528-54. 2010..Appropriate PS applications can create experimental conditions using observational data when randomized controlled trials are not feasible and, thus, lead researchers to an efficient estimator of the average treatment effect...
- Alleviating perinatal depressive symptoms and stress: a nurse-community health worker randomized trialLee Anne Roman
Department of Obstetrics, Gynecology and Reproductive Biology, College of Human Medicine, Michigan State University, 226 West Fee Hall, East Lansing, MI, 48824, USA
Arch Womens Ment Health 12:379-91. 2009..A Nurse-CHW team approach to EPS demonstrated advantage for alleviating depressive symptoms in Medicaid eligible women compared to CC, especially for women at higher risk...
- Fixed effects, random effects and GEE: what are the differences?Joseph C Gardiner
Division of Biostatistics, Department of Epidemiology, Michigan State University, East Lansing, MI 48824, USA
Stat Med 28:221-39. 2009..The choice of model in specific applications would depend on the relevant questions being addressed, which in turn informs the type of design and data collection that would be relevant...
- Diagnostic accuracy of predicting somatization from patients' ICD-9 diagnosesRobert C Smith
Department of Medicine and Psychiatry, Michigan State University, East Lansing, Michigan, USA
Psychosom Med 71:366-71. 2009..We labeled these codes as having "somatization potential." Our earlier study demonstrated that ICD-9 codes and other data from the ADB effectively identified somatization...
- PTSD and postpartum mental health in a sample of Caucasian, Asian, and Pacific Islander womenJane M Onoye
University of Hawai i at Manoa, Honolulu, HI, USA
Arch Womens Ment Health 12:393-400. 2009..Further research is also needed to elucidate the role of ethnic or cultural differences in trauma and PTSD and perinatal health...
- Evaluating the Michigan SENSOR Surveillance Program for work-related asthmaPhilip L Reed
Biomedical Research and Informatics Center, Michigan State University, East Lansing, Michigan 48824 1327, USA
Am J Ind Med 50:646-56. 2007..Limited use has been made of illness data to target inspections and evaluation of such targeting is even more limited...
- Costs of an intervention for primary care patients with medically unexplained symptoms: a randomized controlled trialZhehui Luo
Department of Epidemiology, Michigan State University, East Lansing, MI 48824, USA
Psychiatr Serv 58:1079-86. 2007....
- A dynamic model for estimating changes in health status and costsJoseph C Gardiner
Department of Epidemiology, Division of Biostatistics, Michigan State University, East Lansing, MI 48824, USA
Stat Med 25:3648-67. 2006..Our method, a joint regression model, provides a flexible approach to assessing the influence of patient characteristics on both cost and health outcomes while accommodating heteroscedasticity, skewness and censoring in the data...
- Longitudinal analysis of censored medical cost dataOnur Baser
Thomson Medstat, Ann Arbor, MI 48108, USA
Health Econ 15:513-25. 2006..The methods are applied to a sample of 201 Medicare beneficiaries diagnosed with lung cancer between 1994 and 1997...
- Cancer, Medicaid enrollment, and survival disparitiesCathy J Bradley
Department of Health Administration, Virginia Commonwealth University, Richmond, Virginia 23298, USA
Cancer 103:1712-8. 2005..The current article examined survival for adults < 65 years old diagnosed with breast, colorectal, or lung carcinoma who were either Medicaid insured at the time of diagnosis, Medicaid insured after diagnosis, or non-Medicaid insured...