Validation of the Treatment Related Impact Measure for Diabetes Treatment and Device: TRIM-Diabetes and TRIM-Device
Authors Meryl Brod President, The Brod Group Mette Hammer Associate Director, Health Outcomes, Novo Nordisk Torsten Christensen Associate Director, Health Outcomes, Novo Nordisk Donald Bushnell Associate Director, Health Research Associates
Background Diabetes is one of the most debilitating common illnesses and often requires lifelong drug management Treatments can be varied in administration mode (oral, syringe, pen, pump) and anti-diabetic agent type (oral, GLP-1 or insulin) Impact of both treatment drug and treatment delivery system is multifaceted Most patient-reported outcome (PRO) measures currently available are specific to Type 1 or Type 2 diabetes treatment modality or delivery systems and are designed to be either a health-related quality of life (HRQoL) or a treatment satisfaction measure
Background To be truly comprehensive, defining these impacts should cross traditional domain boundaries of HRQoL, treatment satisfaction, and treatment behavior
Treatment Related Impact Measures The Treatment Related Impact Measure-Diabetes (TRIM-Diabetes) and Treatment Related Impact Measure-Diabetes Device (TRIM-Device) measures were developed to address this gap These measures are intended to capture the full spectrum of impacts of treatment and cannot be classified as strictly HRQoL, or as treatment behavior and satisfaction measures We suggest that these measures be referred to as measures of treatment impact
The TRIM-Diabetes/Device PRO Development Developed according to FDA draft guidance Designed to capture the full range of impacts of diabetes treatment on patients functioning and well-being across: Type 1 and Type 2 diabetes All currently available delivery systems and treatments (oral agents, GLP-1 pens, inhaled or pump delivered insulin and insulin delivered with syringe/pens) 105 persons with diabetes and 8 expert clinical interviews in US, Australia and UK were interviewed in focus groups and by telephone
The TRIM-Diabetes/Device PRO Development Preliminary version had 60-item self-report questionnaire assessing six hypothesized domains: Productivity (Daily Activities), Productivity (Work), Psychological, Device Satisfaction, Efficacy and Burden Five-point Likert response options Scored so that a higher score indicates a better health state
Web Based Validation Study Methods Subjects were over 18, currently on diabetes treatment and had either Type 1 or Type 2 diabetes diagnosed by a physician (self-report) A multi-sourced panel recruitment strategy was employed which used a wide range of permission email recruitment, affiliate networks and website advertising A stratified sample procedure was employed using invitation selection criteria to account for disproportional response rates between stratification categories
Statistical Methods Validation procedures were conducted according to an a priori developed statistical analysis plan (SAP): 1. Item level psychometric and conceptual criteria were used to refine and reduce the preliminary item pool and reduce redundancy between items 2. Factor analysis to identify structural domains was performed 3. Reliability and validity testing was then performed All reliability and validity tests were performed on both the total scores and for each domain
Statistical Methods Item Characteristics/Measurement Model (Scaling) Item reduction: both item psychometric properties and conceptual importance were taken into consideration in making retention/deletion decisions Factor structure: exploratory principle component factor analysis using Varimax orthogonal rotation with Kaiser normalization Internal consistency reliability: Cronbach s alpha Test-retest reliability: assessed at two weeks post initial completion of the battery
Statistical Methods Convergent validity and known-groups validity was evaluated by testing the a priori hypotheses using a twotailed Pearson s correlation coefficient with significance at the p<0.05 level Minimal Important Difference (MID) calculation (exploratory) used self-report items also included in the battery, one per domain of the TRIM-Diabetes/Device, as anchors: The MID considered changes in scores of TRIM-Diabetes domains between responses of Slightly and Somewhat
Results of Validation Study 507 subjects participated in web-based survey The age of the study sample ranged from 18 to 80 years Mean age of 51 years 53% Female 84% White and 6% African-American 74% Type 2 diabetes
Scaling Properties Item Characteristics and Measurement Model (Scaling) The response distributions showed no missing data (online data collection study not allowing for) Nine items showed a ceiling effect (higher than 50%) Several pairs of items were found to be correlated at above 0.70, indicating possible redundancy Several items were also revealed to be unclear in their fundamental concept, and thus the items did not fit into the conceptual framework Based on these indicators 24 items were dropped
Factor Structure Diabetes related Items in seven distinct domains (confirmed by scree plot): Treatment Burden Daily Life Diabetes Management Psychological Health Compliance Device satisfaction items in 2 domains: Device Function Device Bother Determined the 2 device domains formed their own independent measure of device satisfaction and could be considered a separate stand-alone measure of device impact (TRIM-Diabetes Device)
Reliability and Validity Results Internal consistency reliability coefficients (total score and all subscales) are all in the acceptable range from 0.80 and 0.94 Test-retest reliability was analyzed in a subset of 56 subjects are in the acceptable ranging from 0.71 to 0.89 (total score and all subscales) All convergent validity hypotheses were met for the All known-groups validity hypotheses were met
Minimally Important Difference All domains and total score of TRIM-Diabetes met ½ SD criteria with exception of Compliance which was a new domain; there was no overall item to examine the MID: Burden ( =10.6, ½ SD=9.5) Daily Life ( =16.0, ½ SD=9.2) Diabetes Management ( =12.0, ½ SD=8.2) Psychological ( =17.8, ½ SD=8.7) Total score ( =17.6, ½ SD=7.8) For the TRIM-Device Function and Bother domains, the differences did not meet the ½ SD thresholds
Conceptual Model Key Determinant Diabetes Medication Domains/Modules MEDICATION Diabetes Management Side effects Efficacy Daily Life Lifestyle flexibility Productivity Treatment Burden Compliance Psychological Health DEVICE Bother Function Consequences Short term Adherence Willingness to start medication HRQoL Life Satisfaction Morbidity Social stigma Self image Willingness to recommend Long-term Employment status/ history Life satisfaction Morbidity/ mortality Persistence Key Moderators Fear of injection/insulin Co-morbid conditions Treatment naïve or not Type of diabetes Duration of diabetes Prior/current treatment Age Employment status Health insurance status Diabetes education Social support Financial status Life style/ activity level
Final Measures TRIM-Diabetes 28 items 5 Domains and Total Score Treatment Burden Daily Life Diabetes Management Psychological Health Compliance Response burden 5 minutes TRIM-Device 8 items 2 domains Bother Function Response burden 1 minute
Sample TRIM Items Domain Treatment Burden Daily Life Diabetes Management Sample item How convenient or inconvenient is it for you to take your medication at the right time [Not at all Extremely] Because of your medication, how often do you have to limit your daily activities [Never Always] When I take diabetes medication I feel worried about my blood sugar control [Never Always]
Study Limitations Validation study was web-based; may have a respondent selection bias based on Internet access - don t believe to be significant: It has been shown that computer use for those over the age of 50 in U.S. is increasing and approximately 75% of people in this age group have computers at home We did not have longitudinal data to examine change found with the TRIM-Diabetes The use of the TRIM-Diabetes in non-english speaking countries or in sub-groups of patients know to have characteristics which may influence PROs should be examined in future studies
Conclusions The TRIM measures are valid and reliable (both total score and domains) The overall score as well as each domain has been validated so that the measures and domains can be used independently Assessing these broad spectrum impacts should: Provide valuable insight to health care professions: Facilitate physician-patient interactions Improve adherence and persistence of treatments Lead to management plans tailored to the individual Allow researchers to better assess treatment impact