《promisdataanalysisplan》由会员分享，可在线阅读，更多相关《promisdataanalysisplan（25页珍藏版）》请在金锄头文库上搜索。
1、PROMIS SCC Data Analysis Plan (August 17, 2005) Input into this document has been provided by Rita Bode, Jakob Bjorner, David Cella, Karon Cook, Paul Crane, Steve Durako, Richard Gershon, Karen Gold, Elizabeth Hahn, Ron Hambleton, Ron Hays, Jin-Shei Lai, Paul Pilkonis, Bryce Reeve, Steve Reise, Denn
2、is Revicki, Jeanne Teresi, David Thissen, Kevin Weinfurt, David WeissINTRODUCTION This document describes a proposed approach for analyzing patient responses to a pool of items hypothesized to measure health domains applicable to outcomes research, clinical trials, and/or clinical practice. Also des
3、cribed is the application of item response theory (IRT) to the calibration of item banks. These banks will be used for developing short form instruments and enabling CAT-based assessment. Based on decisions by the Steering Committee, the first five domains targeted by the PROMIS network are physical
4、 function, fatigue, pain, emotional distress, social/role participation. A sixth domain, general health, was recently added. Those who conduct analyses for PROMIS should be sure to read the domain hierarchy protocol document authored by Fries et al.The data analysis plan applies to two distinct sets
5、 of analyses. Initially secondary analyses of existing datasets will be performed. Later in the project, psychometric analyses of data collected under the PROMIS network will support the development and evaluation of item banks and instruments for measuring identified PROMIS domains.SECONDARY DATA A
6、NALYSES Analyses will be performed on multiple datasets including the Cardiovascular Health Study (CHS), Chronic Hepatitis C (CHC) Study, Digitalis Investigation Group (DIG) QOL Sub-study, Medical Outcomes Study (MOS), Q-Score, and the WHO-QOL datasets. Other datasets considered included: American A
7、cademy of Orthopedic Surgery (AAOS) dataset, Antiarrhythmics versus Implantable Defibrillators (AVID), Atherosclerosis Risk in Communities (ARIC) Study, BIOQOL, CHD-Offspring, Diabetes Control and Complications Trial, FAMS- Bayer ESIMS Trial, GERD Study, HIV Cost and Services Utilization Study (HCSU
8、S), Item Banking Social Bank, MEDTEP HIV Study, National Epidemiologic Survey on Alcohol and Related Conditions, National Spinal Cord Injury Database, OABq Study, PAGI-SYM, P-QOL, Pain Audit Collection System, RAND Health Insurance Experiment (HIE), Schizophrenia Study, UK-FIMFAM, Unified Medical Gr
9、oup Association, Veterans Health Outcomes Study, and Womens Health Questionnaire. The secondary data analyses are opportunities to explore a number of psychometric questions, the answers to which will inform the eventual development of the PROMIS measures. These questions including what is the: (a)
10、dimensional structure of the domains and subdomains identified by domain chairs?, (b) functional number of response categories persons discriminate in self-reporting various outcomes?, and (c) range of targeted outcomes effectively measured by items within secondary datasets. A fourth question is of
11、 particular importance: “Since different constructs have different levels of conceptual breadth, at what point along the construct hierarchy will it be appropriate to scale items using item response theory since IRT assumes 2unidimensionality (Reise, et al, 2000). For example, PROMIS domain map conc
12、eptualizes Physical Health as being comprised of Function, Symptoms, and Special Senses. In turn, Function has been defined as Mobility, Dexterity, Central, and Activities of Daily Living. Additional sub-categorizations may be made within each of these subdomains. At what level will the domains be u
13、nidimensional enough for IRT modeling? At the level of Physical Health? At the level of Function? At the level of Mobility? These are questions that must be addressed conceptually and empirically. It is anticipated that the analyses of the secondary data sets will inform the decisions.DEVELOPMENT OF
14、 PROMIS INSTRUMENTS Ultimately, this analysis plan will be implemented on the datasets collected within the PROMIS effort and used to calibrate the items for the PROMIS core and supplemental banks. This plan also describes data collection methods to allow co-calibration of items to a common metric.A
15、NALYTIC METHODS The scope of work of the PROMIS initiative is vast and methodologically challenging. Below we classify the analyses that will be undertaken and describe methodologies for accomplishing the analytic tasks. In many cases there are multiple alternatives for addressing the analytic quest
16、ions and no scientific consensus regarding the best approach or the criteria by which to evaluate the results. Below we describe important issues, review available analytic options and, when evidence supports it, suggest preferred methods and criteria.DESCRIBING THE DATA A range of descriptive statistics will be included to examine the performance of the individual items as well as how they function within the scale. Item statistics will include
