Determining sample size and length of follow-up for smartphone-based digital phenotyping studies

Abstract

Objective Studies that use patient smartphones to collect ecological momentary assessment and sensor data, an approach frequently referred to as digital phenotyping, have increased in popularity in recent years. There is a lack of formal guidelines for the design of new digital phenotyping studies so that they are powered to detect both population-level longitudinal associations as well as individual-level change points in multivariate time series. In particular, determining the appropriate balance of sample size relative to the targeted duration of follow-up is a challenge. Materials and Methods We used data from 2 prior smartphone-based digital phenotyping studies to provide reasonable ranges of effect size and parameters. We considered likelihood ratio tests for generalized linear mixed models as well as for change point detection of individual-level multivariate time series. Results We propose a joint procedure for sequentially calculating first an appropriate length of follow-up and then a necessary minimum sample size required to provide adequate power. In addition, we developed an accompanying accessible sample size and power calculator. Discussion The 2-parameter problem of identifying both an appropriate sample size and duration of follow-up for a longitudinal study requires the simultaneous consideration of 2 analysis methods during study design. Conclusion The temporally dense longitudinal data collected by digital phenotyping studies may warrant a variety of applicable analysis choices. Our use of generalized linear mixed models as well as change point detection to guide sample size and study duration calculations provide a tool to effectively power new digital phenotyping studies.

Sociodemographic Characteristics of Missing Data in Digital Phenotyping

“…Some research questions may necessitate high-density, continuous GPS or accelerometer data, in which case it is often more statistically efficient to utilize a within-subject design with longer follow-up than a wide range of participants with limited follow-up. 37 Second, we found substantial individual-level variability in sensor non-collection. This finding suggests that the observed large differences in sensor non-collection are not due to systematic study-related issues but are rather due to high between-person variability.…”

“…Some research questions may necessitate high-density, continuous GPS or accelerometer data, in which case it is often more statistically efficient to utilize a within-subject design with longer follow-up than a wide range of participants with limited follow-up. 37…”

Digital tools for the assessment of pharmacological treatment for depressive disorder: State of the art

“…Stakeholders and Privacy impact statement (PIA). In addition to obvious stakeholders, such as the clinician and the patient, the expertise from the statistician performing the increasingly complex analyses ( Lydon-Staley et al, 2019 ;Washington et al, 2020 ), power calculations ( Barnett et al, 2020 ) has become indispensable too. The privacy or network information security experts should be invited as well in the development of new applications and study protocols.…”

Remote measurement technologies for depression in young people: A realist review with meaningful lived experience involvement and recommendations for future research and practice

“…It also highlighted gaps where very limited evidence was available, as well as several common issues that appeared across much of the literature reviewed including: convenience sampling and selection bias with those young people more interested in RMT more likely to take part; reliance on self-report in community samples with few reporting the number above / below clinical cut-offs; potentially inappropriate statistical analyses for the low sample sizes but large time series data sets 147 ; short study duration (7 days to 12 months at most) or snapshot monitoring periods (2 -3 weeks) in multiyear studies despite estimated relapse rates to be only 5% within the first 6 months, 12% by 12 months, 40% by 2 years, then 70% by 5 years [148][149][150] ; low number of predictor variables and little consideration of the many factors that could have confounding, mediating or moderating effects over the length of the monitoring period; potentially conflicting indicators of depression symptom severity due to the diversity of experience overlooked 76,77,122 ; heterogeneity, artificial inflation, or non-report of quantitative measures of feasibility making comparison across studies difficult, especially with regards to adherence 27,44 ; limited investigation of specificity, sensitivity, and impact on time to treat and depression outcomes; and overall little attempt to assess and explain variability in adherence, accuracy and predictive performance across individuals 74 . With the evidence base as it stands and important insights from those with lived experience, we make the following recommendations for the use of RMT for depression in young people:…”