Approaches that my team developed in the early 2000's - and still refine to this day - have been found to be the among the most effective for the affected arm after stroke.
The authors reviewed findings of 176 trials of interventions targeting the stroke-affected arm. Altogether, the studies enrolled over 6700 participants.
They found that modified constraint-induced therapy - an outpatient, reimbursable, intervention that we pioneered in the early 2000's - has the "most likelihood of being effective." This was followed closely by mental practice, which we also introduced into the stroke rehabilitation literature.
Despite the growing number of stroke survivors exhibiting residual arm impairments, it had long been known that conventional therapies targeting the affected arm are not effective. And, newer therapies require extensive training, time, and resources that are not available at most clinics. For example, the authors reported that, "...robotic interventions were relatively less effective in comparison to other options (mCIMT, mental imagery...interestingly, they represent the largest group of trials in the UE stroke literature, with 181 RCTs having investigated some form of robotic therapy."
As the availability of robotic and other "technological advancements" quickened, I recall some therapists being concerned that they would be "replaced by a robot." The findings of this, and other supportive studies, underscore the continued importance (if not superiority) of skilled, learning based, therapies that can be easily administered in a variety of clinical contexts.