Using Smartphones to Reduce Research Burden in a Neurodegenerative Population and Assessing Participant Adherence: A Randomized Clinical Trial and Two Observational Studies

Abstract

Background: Smartphone studies provide an opportunity to collect frequent data at a low burden on participants. Therefore, smartphones may enable data collection from people with progressive neurodegenerative diseases such as amyotrophic lateral sclerosis at high frequencies for a long duration. However, the progressive decline in patients' cognitive and functional abilities could also hamper the feasibility of collecting patient-reported outcomes, audio recordings, and location data in the long term.

Objective: The aim of this study is to investigate the completeness of survey data, audio recordings, and passively collected location data from 3 smartphone-based studies of people with amyotrophic lateral sclerosis.

Methods: We analyzed data completeness in three studies: 2 observational cohort studies (study 1: N=22; duration=12 weeks and study 2: N=49; duration=52 weeks) and 1 clinical trial (study 3: N=49; duration=20 weeks). In these studies, participants were asked to complete weekly surveys; weekly audio recordings; and in the background, the app collected sensor data, including location data. For each of the three studies and each of the three data streams, we estimated time-to-discontinuation using the Kaplan-Meier method. We identified predictors of app discontinuation using Cox proportional hazards regression analysis. We quantified data completeness for both early dropouts and participants who remained engaged for longer.

Results: Time-to-discontinuation was shortest in the year-long observational study and longest in the clinical trial. After 3 months in the study, most participants still completed surveys and audio recordings: 77% (17/22) in study 1, 59% (29/49) in study 2, and 96% (22/23) in study 3. After 3 months, passively collected location data were collected for 95% (21/22), 86% (42/49), and 100% (23/23) of the participants. The Cox regression did not provide evidence that demographic characteristics or disease severity at baseline were associated with attrition, although it was somewhat underpowered. The mean data completeness was the highest for passively collected location data. For most participants, data completeness declined over time; mean data completeness was typically lower in the month before participants dropped out. Moreover, data completeness was lower for people who dropped out in the first study month (very few data points) compared with participants who adhered long term (data completeness fluctuating around 75%).

Conclusions: These three studies successfully collected smartphone data longitudinally from a neurodegenerative population. Despite patients' progressive physical and cognitive decline, time-to-discontinuation was higher than in typical smartphone studies. Our study provides an important benchmark for participant engagement in a neurodegenerative population. To increase data completeness, collecting passive data (such as location data) and identifying participants who are likely to adhere during the initial phase of a study can be useful.

Trial registration: ClinicalTrials.gov NCT03168711; https://ichgcp.net/clinical-trials-registry/NCT03168711.

Keywords: attrition; digital phenotyping; mobile health; mobile phone; smartphones; trial.

Conflict of interest statement

Conflicts of Interest: JPO is a cofounder and board member of a recently established commercial entity that operates in digital phenotyping. SP reports research grants from Amylyx Therapeutics, Revalesio Corporation, UCB/Ra Pharma, Biohaven, Clene, Prilenia, Seelos, The ALS Association, the American Academy of Neurology, ALS Finding a Cure, the Salah Foundation, the Spastic Paraplegia Foundation, the Muscular Dystrophy Association and reports personal consulting fees for advisory panels from Orion, Medscape and Cytokinetics. JDB reports equity in REACTNeuro. He reports consulting fees from Clene Nanomedicine, Biogen, Janssen, Sawai Pharmaceuticals, MT Pharma of America, MT Pharma Holdings of America. He reports research support from Amylyx, Alexion, Biogen, MT Pharma of America, Anelixis Therapeutics, Brainstorm Cell Therapeutics, nQ Medical, RAPA Therapeutics, NINDS, Muscular Dystrophy Association, ALS One, ALS Association, and ALS Finding A Cure. TMM reports licensing agreements with C2N and Ionis Pharmaceuticals, has served on advisory boards for Biogen and UCB Pharma, and is a consultant for Cytokinetics and Disarm Therapeutics.

©Anna L Beukenhorst, Katherine M Burke, Zoe Scheier, Timothy M Miller, Sabrina Paganoni, Mackenzie Keegan, Ella Collins, Kathryn P Connaghan, Anna Tay, James Chan, James D Berry, Jukka-Pekka Onnela. Originally published in JMIR mHealth and uHealth (https://mhealth.jmir.org), 04.02.2022.

Figures

Figure 1

Kaplan–Meier plot estimates of time-to-discontinuation for 3 data types. Each color denotes a different data type: audio data in red, GPS data in blue, and survey data in yellow. Participants that were censored before the end of the study are denoted by + signs. Each panel shows time-to-discontinuation in a different study: study 1 (top, a 12-week pilot study), study 3 (middle, a 20-week clinical trial), and study 2 (bottom, a 1-year observational study).

Figure 2

Boxplot of participants’ data completeness (in %) excluding the period after discontinuation. Data completeness was defined as percentage of days with any GPS data and percentage of weeks with a completed survey or audio recording.

Figure 3

Bar graph of data completeness per month in study (excluding the period after discontinuation), stratified by time-to-discontinuation of the participant (gray bar indicates time-to-discontinuation). Number of participants for each panel from left to right are as follows: N=7, 4, and 18 for study 1; N=20, 6, 10, 8, and 33 for study 2; and N=5, 1, 2, and 22 for study 3. Data completeness was defined as percentage of days with any GPS data and percentage of weeks with a completed survey or audio recording.