Intervention Designs

Our team develops guidelines, principles and methodologies for optimizing new interventions that can be used in practice to address the unique and changing needs of individuals over time.

This includes the development of methods for optimizing:

Adaptive Interventions

Adaptive interventions guide how best to sequence intervention options (e.g., intervention, treatment or engagement type, intensity, delivery modality) over time based on both static and time-varying information about the individual.

A prototypical adaptive intervention is one that guides decisions about how best to treat patients in a clinic setting. Here, the unit of intervention is the individual patient, the sequential interventions are offered/provided by a clinician, and the intervention decision points are at clinic visits. However, adaptive interventions can be developed for application across a number of diverse settings, including in education settings, prevention settings, and even emergency department settings.

Learn More

Just-in-Time Adaptive Interventions

Just-in-Time Adaptive Interventions (JITAI) are a form of an adaptive intervention intended to address the rapidly changing needs of individuals. Specifically, JITAIs use (potentially fast) changing information about the individual’s internal state (e.g., craving, stress) and context (e.g., physical location) to recommend whether and how to deliver interventions in real time, in the person’s natural environment.

JITAIs often are used in the context of mobile health or digital health interventions. Typically in these settings, the unit of intervention is the individual, intervention is offered/provided by (or in conjunction with) a digital device (e.g., smartphone, smartwatch), and the intervention decision points can be very frequent (e.g., every minute, or multiple times per day).

Learn More

Multi-modal Adaptive Interventions

Multi-modal Adaptive Interventions (MADIs) are a hybrid adaptive intervention design that allows for the provision of a combination of services that are sequenced and adapted by different modalities (e.g., human delivered, digital) and at different time scales. MADIs are used in settings where there is an explicit effort to combine aspects of adaptive interventions (AIs) with just-in-time adaptive interventions (JITAIs).

A typical MADI is one that guides decisions about how best to treat individuals (the unit of intervention) via: (1) a digital mode, with relatively more frequent intervention decision points (e.g., daily via a smartphone), and (2) a human delivered mode, with relatively less frequent intervention decision points (e.g., monthly, at clinic visits).

Multi-level Adaptive Implementation Strategies

Multi-level Adaptive Intervention Strategies (MAISYs) guide the sequencing and adaptation of implementation strategies across multiple levels of implementation (e.g., system level, clinic level, practitioner level). MAISYs guide implementers in selecting which implementation strategy to offer at different levels of implementation, under what conditions, and at what times.

In a MAISY, there are multiple units of intervention, at each of the levels (e.g., clinic, and practitioner within a clinic); intervention at the multiple levels is delivered by an implementer (or implementation team/organization); and, typically, there are relatively infrequent decision points (e.g., quarterly or yearly).

2nd Generation Just-in-Time Adaptive Interventions

2nd Generation Just-in-Time Adaptive Interventions (JITAIs) use reinforcement learning to continually update and sustain intervention effectiveness over time. 2nd Generation JITAIs respond to societal changes and evolving population-based needs, as well as to each individual’s evolving needs.

As with JITAIs, 2G-JITAIs often are used in the context of mobile health or digital health interventions. Typically, in these settings, the unit of intervention is the individual, intervention is offered/provided by (or in conjunction with) a digital device (e.g., smartphone, smartwatch), and the intervention decision points can be very frequent (e.g., every minute, or multiple times per day).

A key distinction between a JITAI and a 2G-JITAI is that: In a JITAI, the decision rules are pre-specified. Whereas in a 2G-JITAI, a pre-specified reinforcement learning algorithm is used to continuously update the decision rules.