Trial Designs

In this section, we describe different types of experimental designs that can be used by behavioral intervention scientists to optimize different types of Adaptive Interventions.

  • Sequential Multiple Assignment Randomized Trials (SMARTs) can be used to develop optimized Adaptive Interventions (AIs)
  • Micro-randomized Trials (MRTs) can be used to develop optimized Just-in-Time Adaptive Interventions (JITAIs)
  • Hybrid designs such as SMART-MRTs can be used to develop optimized Multimodal Adaptive Interventions (MADIs).
  • Multilevel Implementation SMARTs (MI-SMARTs) can be used to develop optimized Multilevel Adaptive Implementation Strategies (MAISYs)

Sequential, Multiple-Assignment, Randomized Trials

Scientists often have multiple questions, across multiple stages, about how best to construct an Adaptive Intervention (AI). Sequential, multiple-assignment, randomized trials (SMART) are a type of experimental design used for developing an optimized AI. In a SMART, the unit of intervention (e.g., the patient) is randomized at more than one stage. Each stage of randomization corresponds to scientific questions about the selection and adaptation of intervention options (e.g., treatments) at particular points in time.

Micro-Randomized Trials

Micro-randomized trials (MRTs) are a type of experimental design used to construct an optimized Just-in-Time Adaptive Intervention (JITAI). In MRTs, the unit of intervention (e.g., patient) may be randomized many times (e.g., hundreds or even thousands of times) over the course of a study at points in time when it is plausible that an intervention is needed. MRTs help behavioral intervention scientists answer scientific questions about the dynamic conditions (e.g., internal, environmental) under which it is best to deliver particular intervention options in a JITAI.

Hybrid Experimental Designs

Hybrid designs, such as hybrid SMART-MRTs, are a type of experimental study design that can be used to construct an optimized Multimodal Adaptive Intervention (MADI). A MADI is an adaptive intervention that integrates human delivered (e.g., coaching sessions) and digital components (e.g., a prompt from a mobile device) that can be delivered and adapted on multiple timescales—slow (e.g., every few weeks or months) and fast (e.g., every few days or hours). In Hybrid SMART-MRTs, the unit of intervention (e.g., individual), is randomly assigned to different intervention components at multiple timescales—slow and fast. These randomizations help us answer scientific questions about how best develop a MADI, that is how to best combine intervention components that can be delivered and adapted by different modalities at different timescales.

For example, consider optimizing a MADI that includes coaching sessions delivered by a clinician, coupled with a smartphone (digital) app that is designed to encourage daily engagement in healthy behaviors. A Hybrid SMART-MRT may randomize patients: (i) at program entry and 4 weeks later, to understand the effect of delivering (vs. not delivering) coaching initially and 4 weeks later, as well as (ii) every day to understand the effect of sending a digital prompt to engage in healthy behaviors. Besides the hybrid SMART-MRT, other kinds of hybrid experiments are available to study the effects of intervention components that can be delivered and adapted at different timescales. For example, a hybrid experiment combining a standard factorial and a SMART can be used to study the separate and combined effects of multiple first-stage components along with at least one component that can be delivered and adapted at a later time point. A hybrid combining a standard factorial and a MRT can be used to study the separate and combined effects of multiple components that can be delivered once (typically at the beginning of the program) and a just-in-time component that can be delivered rapidly (e.g., every day). In each case, the goal is to provide information for designing an effective and resource efficient MADI.

Multilevel Implementation SMARTs

Multilevel Implementation SMARTs (MI-SMART) are a type of experimental study design used to construct an optimized Multilevel Adaptive Implementation Strategy (MAISY). In a MI-SMART, there are multiple levels (e.g., clinic and providers within clinic) and multiple time points of randomization. These randomizations help us answer scientific questions about how best to sequence and adapt different strategies across multiple levels of implementation.

For example, consider optimizing a two-stage MAISY for implementing cognitive behavioral therapy (CBT) at high schools across Michigan, which often have multiple school psychologists (SPs) on staff. A MI-SMART may randomize: (i) schools to understand the effect of a school-wide CBT skills coaching strategy in the first stage, and (ii) SPs within schools to understand the effect of a self-efficacy promotion strategy in the second stage.