# Leveraging ML algorithms and data integration techniques to improve efficiency of causal moderation analyses of micro-randomized trial data

> **NIH NIH R01** · UNIVERSITY OF MICHIGAN AT ANN ARBOR · 2024 · $305,996

## Abstract

PROJECT SUMMARY
Smart devices (e.g., smartphones, smartwatches) and other wearables are increasingly used to deliver
personalized, digital interventions to patients in real time to assist with management of chronic diseases.
Digital interventions provide accessible, low-cost intervention delivery in unlimited amount and duration
whenever and wherever needed, even for reticent or difficult-to-reach populations. Mobile health
systems are excellent tools for the delivery of just-in-time adaptive interventions (JITAI), a general
intervention design that aims to provide the right type/amount of support, at the right time, by adapting
to an individual’s changing context. Many JITAIs have been developed with minimal use of empirical
evidence. Micro-randomized trials (MRTs) are an increasingly common experimental design that are the
main source for data-driven evidence of JITAI effectiveness. Causal moderation analysis is the cornerstone
data analytic method for building effective JITAIs, allowing investigators to answer scientific questions
about when and how to deliver digital interventions.
Current data analytic methods for moderation analysis of MRT data require investigators to pre-specify
features of the intensively collected longitudinal data (ILD) and high frequency sensor and mobile data
(SMD) that are used in a linear model for outcome prediction. Machine learning (ML) algorithms are
better suited to generate these outcome predictions but their use to estimate causal effects can lead to
bias under model misspecification and do not allow for proper uncertainty quantification of treatment
effect estimates. Moreover, current methods are designed to handle data coming from a single MRT;
however, it is increasingly common for researchers to employ designs in which only a subset of study
participants are included in the MRT or to have access to summary-statistics from external mHealth
studies. Combining information from participants in other study arms or external studies with information
from a smaller but more detailed MRT may significantly improve health scientists’ ability to answer
important scientific questions about how intervention effectiveness may change over time or may be
moderated by individual characteristics, time-varying context, or past responses. We propose to combine
ML algorithms and data integration techniques with the standard statistical tools used to assess causal
effect moderation to improve tailoring strategies for prevention and treatment of chronic diseases. At its
conclusion, the proposal will provide comprehensive statistical tools for efficient inference of time-varying
causal effects. These methods are essential for building effective intervention packages to improve health
outcomes in populations managing chronic diseases.

## Key facts

- **NIH application ID:** 10777343
- **Project number:** 1R01GM152549-01
- **Recipient organization:** UNIVERSITY OF MICHIGAN AT ANN ARBOR
- **Principal Investigator:** Walter Dempsey
- **Activity code:** R01 (R01, R21, SBIR, etc.)
- **Funding institute:** NIH
- **Fiscal year:** 2024
- **Award amount:** $305,996
- **Award type:** 1
- **Project period:** 2024-03-01 → 2027-11-30

## Primary source

NIH RePORTER: https://reporter.nih.gov/project-details/10777343

## Citation

> US National Institutes of Health, RePORTER application 10777343, Leveraging ML algorithms and data integration techniques to improve efficiency of causal moderation analyses of micro-randomized trial data (1R01GM152549-01). Retrieved via AI Analytics 2026-05-28 from https://api.ai-analytics.org/grant/nih/10777343. Licensed CC0.

---

*[NIH grants dataset](/datasets/nih-grants) · CC0 1.0*