MI-SMART: Methods for Optimizing Multilevel Adaptive Implementation Strategies to Promote the Adoption of Effective SUD and HIV Services

PROJECT SUMMARY: PROJECT 1
Evidence-based prevention, treatment, and recovery services for substance use disorders (SUDs) and HIV
often fail to be implemented or sustained due to barriers at multiple levels, such as lack of time and skills at the
practitioner level; lack of support or established processes at the clinic level; or lack of effective technological
tools, financing, and policies at the system level. The long-term goal of this project is to develop a new
generation of effective, replicable multi-level strategies for implementing evidence-based SUD/HIV services in
real-world settings, thus improving the outcomes of individuals with or at risk for SUD or HIV. To achieve this
long-term goal, we will: (Aim 1) Develop principles for the design of Multi-level Adaptive Implementation
Strategies (MAISYs) to guide how best to adapt (e.g., augment, intensify) implementation strategies based on
the context and needs at multiple levels (e.g., clinic, practitioner); and develop a new experimental design—the
Multi-level Implementation Sequential, Multiple Assignment, Randomized Trial (MI-SMART), in which
sequences of randomizations occur at multiple levels—to inform the construction of optimized MAISYs; (Aim 2)
Develop new data analysis methods for use with longitudinal data in an MI-SMART to address a suite of
questions about the construction of a MAISY; (Aim 3) Develop sample size calculators for MI-SMARTs to
enable SUD/HIV implementation scientists to plan and calculate statistical power for different types of MI-
SMART designs; and (Aim 4) Work with the Dissemination and Training Core to disseminate the developed
methods so that they can be readily applied by SUD/HIV scientists. This includes the development of webinars,
workshops, tutorials, and free, user-friendly software for SUD/HIV scientists.

Grant Number: 5P50DA054039-05
NIH Institute/Center: NIH
Principal Investigator: Daniel Almirall