Micro-randomized Trials for Adaptive Digital Interventions

This direction is merged into another direction: micro-randomized trial. Its last version stays readable and citable here.

Version 1 of 115 papers17 conceptsfrozen 2026-10-04

Traditional clinical trials are designed to answer a single, binary question: Does this treatment work better than a placebo? In the world of mobile health (mHealth), this "black box" approach is often insufficient. A smartphone app isn't a pill; it is a collection of hundreds of potential "nudges"—push notifications, reminders, or gamified challenges—delivered across weeks or months.

To optimize these digital tools, researchers have developed the Micro-randomized Trial (MRT). Instead of randomizing a participant once at the start of a study, an MRT randomizes them hundreds or even thousands of times at discrete "decision points." This allows researchers to move beyond asking if an app works to asking when, where, and for whom specific components are effective.

The Problem of Static Engagement

Digital Mental Health Interventions (DMHIs) face a significant "engagement problem." As Jessica M. Lipschitz and colleagues note in The Engagement Problem: a Review of Engagement with Digital Mental Health Interventions and Recommendations for a Path Forward, real-world completion rates for these apps often fall below 20%.

The issue is often timing. A notification sent when a user is busy, stressed, or already succeeding can be perceived as "digital nagware," leading to habituation or even abandonment. To solve this, Omer T. Inan and Pamela Tenaerts argue for Digitizing clinical trials, moving away from episodic clinic visits toward continuous, real-world data collection that can power "Just-In-Time Adaptive Interventions" (JITAIs).

How Micro-randomization Works

The foundational statistical framework for MRTs, established by Peng Liao, Susan Murphy, and others in Sample size calculations for micro‐randomized trials in mHealth, treats a single participant's journey as a series of many small experiments.

  1. Decision Points: These are times when an intervention could be delivered (e.g., every morning at 8:00 AM, or whenever a wearable detects a spike in heart rate).
  2. Randomization: At each point, the system randomly decides whether to deliver a nudge or which version of a nudge to send.
  3. Proximal Effects: Researchers measure the immediate, near-term response (e.g., did the user log their mood in the next 30 minutes? Did they take 500 steps in the next hour?).

By repeating this process thousands of times across a cohort, researchers can estimate the marginal effect of an intervention component while accounting for the user's current context.

Capturing Context and Receptivity

The power of the MRT lies in its ability to identify "moderators"—factors that change how well a nudge works. In Assessing Real-Time Moderation for Developing Adaptive Mobile Health Interventions for Medical Interns, Timothy NeCamp, Srijan Sen, and Ambuj Tewari found that push notifications for mood and sleep were highly beneficial when interns were struggling, but could actually have a negative impact when they were already performing well.

Similarly, Jitao Wang used a cluster MRT to study gamified team competition among interns. The trial revealed that while competition boosted activity, its effectiveness was a "race against time," waning as the interns' burnout increased over the course of their residency.

From Data to Algorithmic Decision Rules

The ultimate goal of an MRT is not just to analyze behavior, but to build the logic for future AI-driven health assistants. As Ashley E. Walton and Predrag V. Klasnja explain in The Micro-Randomized Trial for Developing Digital Interventions, MRTs provide the "experimental evidence base" needed to construct decision rules.

We are already seeing this transition to automated personalization. Christian Rauschenberg and colleagues recently tested AI4U training, a system that uses Recurrent Neural Networks (RNNs) to assign interventions. By using an MRT design to compare ML-driven assignments against random ones, they proved that algorithmic personalization significantly boosts "momentary resilience" in youth.

By treating digital health as a dynamic, sequential process rather than a static product, Micro-randomized Trials allow researchers like Amy L. Cochran and Alexandra M. Psihogios to bridge the gap between clinical psychology and machine learning, creating interventions that are truly responsive to the "just-in-time" needs of the user.

Papers in this version

History & citationVersion 1 · 2026-10-04

Merged into another direction · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_micro_randomized_trials_for_adaptive_dig_85182a1c,
  title = {Micro-randomized Trials for Adaptive Digital Interventions},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/micro-randomized-trials-for-adaptive-digital-interventions/rdr_85182a1c58df46d7a3a4bb614ccf2edb},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_micro_randomized_trials_for_adaptive_dig_85182a1c_v1,
  title = {Micro-randomized Trials for Adaptive Digital Interventions},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/micro-randomized-trials-for-adaptive-digital-interventions/rdr_85182a1c58df46d7a3a4bb614ccf2edb/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}