Research Methods Core

点击翻译按钮获取中文摘要

The mission of the Research Methods Core (RMC) is to advance the quality and impact of research in "real world settings" by addressing methodological challenges posed by the clinical and service delivery complexities of depressed elders with limited access to good care. Concerns raised by our own communitybased studies, and the field more broadly, guided the development of RMC Initiatives as well as one of our Developmental Studies and three of our Pilot Studies: Initiative 1: Predictors and Moderators of Outcomes in Intervention Development: Development of novel interventions for depressed seniors who may not respond to existing treatment approaches is a complex process. We will rely on exploratory analyses of our rich RCT databases and use predictors of outcomes and moderators of treatment response to sharpen the selection of target populations, improve our interventions, and accelerate the process of moving research into practice. Pilot 3 focuses on the feasibility of web-based training as a way of enhancing the dissemination of community interventions. Pilot 5 studies the cost of care management in home healthcare as a way of evaluating its feasibility. Initiative 2. Reduction of Bias: Interventions focusing on provider behavior as well as patient outcomes often use cluster (setting-based) randomization designs resulting in non-equivalent groups. Building on our study of propensity modeling for patient-RCTs, we will investigate the suitability of propensity modeling in adjusting for group differences in cluster based designs (also addressed in Developmental Project 4). Another source of bias is loss of data through subject attrition, a problem inherent in intervention studies of chronic disorders like geriatric depression. To address this problem, we propose a sequential, multiple imputation procedure that builds on our work in which we evaluated bias reduction using a predictive model based on multiple imputation for the analysis of incomplete, missing not at random (MNAR) data (Pilot 4). Initiative 3. Generalizability of Community-Generated Interventions: To address sampling bias in community-based studies, we propose procedures related to both within-agency generalizability (e.g., reducing the source and impact of sample bias introduced by patient or client recruitment processes) and across-agency variation (e.g., biased introduced in the process of selecting community-based research partners or in implementing interventions).