San Keller, PhD, Principal Investigator, PROMIS Network American Institutes for Research

Transcription

1 San Keller, PhD, Principal Investigator, PROMIS Network American Institutes for Research sankeller@air.org June 24, 2014 Webinar for the Genetic Alliance Dynamic Tools to Measure Health Outcomes from the Patient Perspective

2 This Webinar Will Explain what patient-reported outcome measures are and why you would want to use them Describe the goal of PROMIS Describe Click to how edit PROMIS Master title measures style are made and tested Show you how to access & use PROMIS measures

3 Examples of Patient Outcome Measures 6-minute walk test Observer Rating - Functional Impact Measure Patient Report (e.g. 0-to10 pain scale) 3

4 Why PRO s have become important health data Survival not a sufficient measure of health Sky-rocketing health care costs require focus on outcomes Some health outcome data can only come from patient reports Some health outcome data not practical to collect by other means Some health outcome data less burdensome to collect through PRO s

5 Emotions Social Activity Activities of Daily Living PROMIS Asks Patients to Report on Their Physical Activities Cognitive Functioning

6 Source: PROMIS Network For example, PROMIS Pain Behavior item

7 Why are PROs Necessary? Reported Change in Overall Patient Quality of Life 100 % (n=75) Better QOL Worse QOL 0 Dr Patient Spouse (Jachuck et al., 1982) 7

8 Do PROs Provide Useful Information? % who lost job Percent of patients who lost job within 1 year at each level of self-reported health < > 55 8

9 Do PROs Provide Useful Information? Percent of patients who died within 5 years at each level of self-reported health % Dead within 5 Yrs < >54 Scores on Self- Reported Physical Health - From Poor to Good Health 9

10 Patient Reported Outcomes: Applications Translational Research T1 -- Clinical Trials T2 -- Comparative effectiveness T3 Quality improvement T4 Personalized care Population Sciences Registries National surveys Health and social policy evaluations

11 Understanding the Implications of the Data Requires Standardizing the Measure Same Number, Different Measure, Different Meaning 32 32

12 Dramatic Example of the Problem Frequency (%) 20 Same Respondents, But Different Questions = Different Meanings Vitality Frequency (%) 70 Energy Source: Brazier et al., Zero 12

13 Major Barrier to PRO Measurement Lack of standardization in PRO measurement CONTENT PDQ-39 PDQL PIMS PLQ Walking Self-Care 5 1 Speech 1 1 Cognition Tremor 1 1 Pain 2 1 Sleep 1 1 Depression Click to edit Master 1 title 1 style 1 2 Anxiety 1 1 Close Relationships 3 2 Work 1 Leisure Activities TIME FRAME 1 Month 3 months Unspec. 1 Week RATING Freq Freq Impact Varies

14 How did this Happen? Different levels of health require different ques=ons Different types of health require different ques=ons Different health condi=ons require different ques=ons Cannot prac=cally ask all of these ques=ons of everyone No centralized, coordinated, adequately funded/ staffed ini=a=ve to create standard measures

15 Patient Reported Outcomes Measurement Information System NCCA NCI NINDS NHLBI NIMH PROMIS NIA NIDDK NICHD NIAMS NIAID

16

17 PROMIS Covers Many Aspects of Health

18 Item Bank Adaptive Assessments Source: PROMIS Network Short, Fixed-Format Assessments

19 PROMIS: How Developed Full Bank Testing (N=7,005) Block Testing (N=14,128) Site (n=329) Demo. short form Global health Forms A-G Legacy Polimetrix (n=6,676) Demo. short form Global health Forms A-G Legacy Co-morbidity General population (n=329) Click to edit Master General population title style Site (n=1,203) Demo. extended form Global health Forms H-W Co-morbidity Polimetrix (n=12,925) Demo. extended form Global health Forms H-W Co-morbidity Samples General population (n=6,676) (n=400) Clinical sample (n=803) General population (n=5,845) Clinical sample (n=7,080) UNC=236 Stanford=93 UNC=304 Stanford=96 Pittsburgh=252 Stanford=150 Duke=401

20 Probability of specified response p=0.50 Easy item less Amount of Trait more Probability of specified response Hard item p=0.50

21 0 Items Positioned Along a Continuum Based on Analysis of Patient Responses Item 1 Item 2 Item 3 Item 4 Item 5 Item 6 Item 7 1. Are you able to eat? 2. Are you able to get in and out of bed? 3. Are you able to stand without losing your balance for 1 minute? 4. Are you able to walk from one room to another? 5. Are you able to walk a block on flat ground? 6. Are you able to run or jog for two miles? 7. Are you able to run five miles? Source: PROMIS Network 50 Item n 100

22 Logic of CAT 1. Begin with initial score estimate 2. Select & present optimal scale item 3. Score response No 5. Is stopping 4. Re-estimate health score and confidence interval rule satisfied Yes 6. End scale assessment 7. End of battery? No 8. Administer next scale Yes Source: Adapted from Wainer et al. (1990) 9. Stop 22

23 Estimate of Person s Place on the health concept continuum gets more precise with every question Q1: Standard Error of Measurement decreases with each question Q2: Q3: Q4: Q5: Source: PROMIS Network

24 CAT versus Short Forms Short Forms 4, 6, or 8 items targeted at clinical range CAT Standard error cutoff =.3 Minimum Test Length = 4 Maximum Test Length = 12 Items calibrated using IRT

25 CAT versus Short Forms P r e c i s i o n Standard Error item SF36/Vitality 4-item CAT 13-item FACIT-Fatigue 13-item CAT 98-item Bank No Fatigue US General Population mean Severe Fatigue

26 Score Interpretation Low High Example of low fatigue This patient s fatigue score is 40, significantly better than average (50). People who score 40 on fatigue tend to answer questions as follows: I have been too tired to climb one flight of stairs: SOMEWHAT I have had enough energy to go out with my family: VERY MUCH

27

28 Information about What s Included

29 Peer-Reviewed Research on Fatigue Measures

30 Information about What s Included

31