Session Information
Date: Wednesday, October 24, 2018
Title: 6W007 ACR Abstract: Patient Outcomes, Preferences, & Attitudes II: PROs (2910–2915)
Session Type: ACR Concurrent Abstract Session
Session Time: 9:00AM-10:30AM
Background/Purpose: Advances in reality mining combined with the pervasive use of smart phones have shown measurable changes in phone behavior due to changes in health. We sought to determine if passively collected digital measures could predict patient-reported outcome (PRO) measures in patients with rheumatic diseases.
Methods: Participants were a subset of adult smartphone users enrolled in Forward, The National Databank for Rheumatic Diseases during 2013-2015. They used a custom smartphone app that collected daily passive digital measures including GPS-tracked mobility (single path) and mobility radius (circle that included all daily mobility). Users with Android OS phones also provided number and duration of calls and texts. PROs included daily pain and global assessment and weekly HAQ-II and PAS-II. Disease flares were regularly self-reported. Statistics included linear mixed effect models (MRM) with random slope for days for PROs and logistic GEE (for flares). Best models were selected using stepwise. Moving averages (MA) of the prior week were applied to the log-transformed passive data. Possible confounders accounted for included patient demographics, clinic measures, and seasonal factors.
Results: Of the 446 participants, 292 (66%) had rheumatoid arthritis (RA). They were mostly female (91%), of middle age (54±12 yrs) with moderate disability (HAQ-II 0.8±0.6) and comorbid illness (RDCI 2.3±1.7); 44% were treated with biologics. The MA of text length was most strongly and inversely associated with PROs including weekly pain (β= -0.15, 95%CI -0.26, -0.05). MA of mobility measures were significantly associated with global, HAQ-II, and PAS-II, but not pain. Table 1 shows estimates in RA from the MRM model for passive phone data keeping other variables at mean values. Flare incidence was significantly associated with the MA of mobility radius (OR = 0.92, 95%CI 0.86, 0.98).
Conclusion: Digital measures collected via smartphone may be a less intrusive means to identify worsening in patients with RA and other rheumatic diseases. Additional studies should confirm and expand these findings.
|
Table 1 – Predictions of weekly PRO for ranges of passive data in RA patients |
||||
|
Passive variable |
Estimated*: Pain |
Global |
HAQ-II |
PAS-II |
|
SMS length |
|
|
|
|
|
20 characters |
3.46 (2.86 – 4.07) |
3.50 (2.98 – 4.02) |
0.98 (0.82 – 1.13) |
2.61 (2.18 – 3.04) |
|
50 |
3.32 (2.73 – 3.91) |
3.38 (2.88 – 3.88) |
0.96 (0.80 – 1.11) |
2.51 (2.09 – 2.93) |
|
100 |
3.21 (2.62 – 3.81) |
3.29 (2.78 – 3.79) |
0.94 (0.79 – 1.10) |
2.43 (2.01 – 2.85) |
|
150 |
3.14 (2.54 – 3.74) |
3.23 (2.71 – 3.74) |
0.93 (0.78 – 1.09) |
2.38 (1.95 – 2.81) |
|
500 |
2.96 (2.33 – 3.60) |
3.08 (2.52 – 3.63) |
0.91 (0.75 – 1.07) |
2.26 (1.81 – 2.70) |
|
Interaction diversity |
|
|
|
|
|
0 |
2.40 (0.92 – 3.88) |
0.69 (0.39 – 0.99) |
1.46 (0.51 – 2.41) |
|
|
1 person |
3.14 (2.54 – 3.75) |
3.25 (2.74 – 3.77) |
0.93 (0.77 – 1.08) |
2.38 (1.95 – 2.81) |
|
3 |
3.26 (2.67 – 3.85) |
3.33 (2.83 – 3.84) |
0.95 (0.79 – 1.10) |
2.46 (2.04 – 2.89) |
|
5 |
3.31 (2.72 – 3.91) |
3.37 (2.87 – 3.87) |
0.96 (0.80 – 1.11) |
2.50 (2.08 – 2.93) |
|
15 |
3.43 (2.82 – 4.04) |
3.45 (2.92 – 3.97) |
0.98 (0.82 – 1.14) |
2.59 (2.16 – 3.02) |
|
30 |
3.50 (2.87 – 4.13) |
3.49 (2.95 – 4.95) |
0.99 (0.83 – 1.15) |
2.64 (2.20 – 3.08) |
|
Reaction time |
|
|
|
|
|
0.5 hour |
3.27 (2.67 – 3.86) |
3.34 (2.84 – 3.84) |
|
2.47 (2.05 – 2.89) |
|
2.5 |
3.30 (2.71 – 3.89) |
3.36 (2.86 – 3.86) |
|
2.49 (2.07 – 2.92) |
|
7 |
3.32 (2.73 – 3.92) |
3.37 (2.87 – 3.88) |
|
2.51 (2.09 – 2.93) |
|
12 |
3.33 (2.74 – 3.93) |
3.38 (2.88 – 3.88) |
|
2.52 (2.10 – 2.94) |
|
Mobility |
|
|
|
|
|
0.5 mile |
|
3.31 (2.81 – 3.81) |
0.95 (0.79 – 1.10) |
2.47 (2.04 – 2.89) |
|
5 |
|
3.47 (2.96 – 3.97) |
0.97 (0.82 – 1.13) |
2.53 (2.11 – 2.95) |
|
10 |
|
3.51 (3.00 – 4.02) |
0.98 (0.83 – 1.14) |
2.55 (2.12 – 2.97) |
|
20 |
|
3.56 (3.04 – 4.08) |
0.99 (0.84 – 1.15) |
2.57 (2.14 – 3.00) |
|
Mobility radius |
|
|
|
|
|
1 mile |
|
3.31 (2.80 – 3.81) |
0.96 (0.81 – 1.12) |
|
|
5 |
|
3.34 (2.84 – 3.84) |
0.96 (0.80 – 1.11) |
|
|
10 |
|
3.36 (2.85 – 3.86) |
0.95 (0.80 – 1.11) |
|
|
50 |
|
3.39 (2.88 – 3.89) |
0.95 (0.79 – 1.10) |
|
|
250 |
|
3.42 (2.91 – 3.93) |
0.94 (0.78 – 1.09) |
|
|
*Adjusted for sex, ethnicity, total income, baseline pain, education, number of persons in household, rheumatic disease comorbidity index (RDCI), sleep scale, self-reported joint count, prior biologic and DMARD exposures, Medicare status, calendar year, season, region and season/region interaction.
|
||||
To cite this abstract in AMA style:
Michaud K, Pedro S, Schumacher R. Can Passively-Collected Phone Behavior Determine Rheumatic Disease Activity? [abstract]. Arthritis Rheumatol. 2018; 70 (suppl 9). https://acrabstracts.org/abstract/can-passively-collected-phone-behavior-determine-rheumatic-disease-activity/. Accessed .« Back to ACR/ARHP Annual Meeting 2018
ACR Meeting Abstracts - https://acrabstracts.org/abstract/can-passively-collected-phone-behavior-determine-rheumatic-disease-activity/