Many chronic or cyclical conditions and their associated health behaviors are characterized by symptoms which are dynamic and personal in nature. Treatments may be effective for some patients, but ineffective for other patients. Additionally, some patients may initially respond well to a treatment, but over time their engagement or adherence with, or clinical response to the intervention, deteriorates. Such dynamic symptoms or outcomes are now often captured digitally. For such situations, the best way forward may be to aim to engage with and treat the right person with the right treatment at the right time for them.

The tailoring of regimes to be optimally effective treatment policies presents some unique methodological and statistical challenges. Through the advantages of fit-for-purpose randomized trials and estimating causal excursion effects, we can begin to build adaptive interventions.

Additionally, clinical trials are increasingly collecting outcomes from wearables such as accelerometers and other digital devices. High-frequency measurements from digital technologies pose new problems for defining and handling missing data.

In this webinar, we will discuss two examples of how adaptive health behavior interventions were developed from a sequential multiple assignment randomized trial and micro-randomized trial. We will also discuss missing data challenges in accelerometer outcomes in a physical activity trial.