Optimizing the Efficiency of Patient Data Capture Using Smartphone Technology: Evaluation of the Correlation Between Promis Instruments for PRO Data Capture

Huifeng Yun¹, Jennifer Beaumont², Shuo Yang³, James Willig⁴, W. Ben Nowell⁵, Seth D. Ginsberg⁶, Kelly V. Clayton⁷, Shantana Hazel^7,8, Carole Wiedmeyer⁵ and Jeffrey Curtis⁹, ¹Epidemiology, University of Alabama at Birmingham School of Public Health, Birmingham, AL, ²Medical Social Sciences, Northwestern University Feinberg School of Medicine, Chicago, IL, ³Division of Clinical Immunology & Rheumatology, University of Alabama at Birmingham, Birmingham, AL, ⁴Med - Infectious Diseases, University of Alabama at Birmingham, Birmingham, AL, ⁵CreakyJoints/Global Health Living Foundation, Upper Nyack, NY, ⁶Global Healthy Living Foundation, CreakyJoints, Upper Nyack, NY, ⁷Global Healthy Living Foundation, Upper Nyack, NY, ⁸Sister Girl Foundation, Inc., Upper Nyack, NY, ⁹Division Clinical Immunology & Rheumatology, University of Alabama at Birmingham, Birmingham, AL

Meeting: 2016 ACR/ARHP Annual Meeting

Date of first publication: September 28, 2016

Keywords: Disease Activity, PROMIS and patient outcomes

Session Information

Date: Sunday, November 13, 2016

Title: Health Services Research - Poster I

Session Type: ACR Poster Session A

Session Time: 9:00AM-11:00AM

Background/Purpose: Patient-reported outcomes (PROs) are key to enabling the comprehensive assessment of patient-centered benefits in comparative effectiveness research (CER). However, the relationships between different PROMIS instruments and condition-specific disease activity measures in diseases such as rheumatoid arthritis (RA) have not been well studied. The objectives of this analysis were to evaluate the longitudinal relationship between different PROMIS instruments and the RAPID3, a measure of self-reported patient disease activity.

Methods: Four PROMIS instruments: pain interference, physical function, sleep disturbance, and fatigue as well as the RAPID3 were administered to participants in the PCORI-funded ArthritisPower patient registry. After descriptive analytics, we estimated multiple correlations between PROMIS instruments and the RAPID3. For each PRO instrument and with each assessment used as the unit of measure, we used model of fit, also called R-squared (the proportion of variance of outcomes that can be explained by the predictors), calculated from mixed models to understand how longitudinal PROs were related to each other. Using pain as an example, we evaluated the R-squared for each model with additional PROs and demographic factors including enrollment age, sex, race, twitter account, region and visit times.

Results: A total of 1,546 unique participants who answered the survey one or more times was included in the analysis, with mean (SD) age of 49 (12) years. The mean score for pain interference was 63.7 (SD: 7.0), physical function 37.5 (7.1), sleep disturbance 58.4 (8.7), fatigue 63.8 (8.8), and RAPID3 15.5 (5.7). Most PROMIS instruments were low to moderately (around 0.2) correlated with each other and the RAPID3. Using pain interference as an example, R-squared measures revealed a high total variance explained (R²=49%) between pain interference and physical function (Table); those involving pain, physical function, fatigue, sleep disturbance and RAPID3 also revealed a higher variance contribution with these additional PROs (66%). Additional adjustment for demographic factors added little variance explanation (1.4%).

Conclusion: PROMIS pain interference, physical function, sleep disturbance, fatigue instruments and RAPID3 are low to moderately correlated to each other. Age, gender, race and other demographic factors play little role in explaining variance in PROs. These results suggest potential efficiencies in using some measures to predict or impute the values for other measures and to optimize the frequency of patient data collection using at-home technologies including Smartphone Apps.

Table: Total variance explained by predictors (R-Squared) using PROMIS pain interference as the dependent variable
Models with PROMIS pain interference as the dependent variable	Pain variance that was not able to explained by predictors	Total pain variance that could be explained by predictors (R-Squared)
Baseline model: no predictors	42.6	N/A
Model with predictors: PROMIS physical function	22.6	46.9%

Model with predictors: PROMIS physical function, sleep	20.7	51.6%

Model with predictors: PROMIS physical function, sleep, fatigue	18.7	56.1%

Model with predictors: PROMIS physical function, sleep, fatigue, RAPID3	15.3	64.1%

Model with predictors: PROMIS physical function, sleep, fatigue, RAPID3 and others (age, gender, race, region, etc.)	15.1	64.7%

Model with only demographic predictors: age, gender, race, region, etc.	42.3	0.7%

Disclosure: H. Yun, Amgen, 2; J. Beaumont, None; S. Yang, None; J. Willig, None; W. B. Nowell, None; S. D. Ginsberg, None; K. V. Clayton, None; S. Hazel, None; C. Wiedmeyer, None; J. Curtis, None.

To cite this abstract in AMA style:

Yun H, Beaumont J, Yang S, Willig J, Nowell WB, Ginsberg SD, Clayton KV, Hazel S, Wiedmeyer C, Curtis J. Optimizing the Efficiency of Patient Data Capture Using Smartphone Technology: Evaluation of the Correlation Between Promis Instruments for PRO Data Capture [abstract]. Arthritis Rheumatol. 2016; 68 (suppl 10). https://acrabstracts.org/abstract/optimizing-the-efficiency-of-patient-data-capture-using-smartphone-technology-evaluation-of-the-correlation-between-promis-instruments-for-pro-data-capture/. Accessed .

« Back to 2016 ACR/ARHP Annual Meeting

ACR Meeting Abstracts - https://acrabstracts.org/abstract/optimizing-the-efficiency-of-patient-data-capture-using-smartphone-technology-evaluation-of-the-correlation-between-promis-instruments-for-pro-data-capture/