In the case of loneliness interventions

In the case of loneliness interventions, all of the reviews essentially confirmed the findings of previous reviews that social skills training and group-based interventions can succeed in reducing loneliness. Is this conclusion justified, or is this a case in which prior conclusions have been perpetuated in the manner Kuhn describes? To combat bias favoring results that confirm dominant theories, some scientists have argued that specific study criteria should be met to warrant an evaluation of the effectiveness of the intervention (18). These criteria include random assignment of study participants to receive the intervention, evidence that the intervention is more effective than no intervention, findings that are replicated by at least one independent research group, and results that are published in peer-reviewed journals. Previous reviewers of loneliness interventions have, in fact, placed a premium on randomized trials that contrast a group randomly selected to receive the intervention with a group randomly selected to receive no intervention. However, none has employed meta- analysis, a quantitative technique for calculating the average effect of diverse interventions designed to accomplish the same goal. Whereas qualitative reviews are subjective and vulnerable to confirmatory biases, quantitative reviews are objective and relatively impervious to bias as long as all relevant studies are included in the analysis. To minimize bias in our meta- analysis, we first combed the literature to identify all the intervention studies that specifically targeted loneliness. To further meet our criteria for inclusion in the meta-analysis, studies had to be published in a peer-reviewed journal or as a doctoral dissertation (to ensure the 133 Page scientific integrity of the findings), between 1970 and 2009 (to include and extend the time interval reviewed qualitatively in prior research), and had to measure loneliness quantitatively. Fifty-two intervention studies for loneliness met our inclusion criteria. These studies were divided into three categories based on the experimental design used to assess the effects of the intervention. Twelve studies used a single group pre-post design in which loneliness among participants was assessed at baseline and again after exposure to the intervention. The single pre-post design is weak in terms of measuring the effectiveness of an intervention, however, because individuals who have high scores on a loneliness measure on one occasion are likely to score less extremely on a second occasion even if no intervention had occurred. Said differently, people whose measurements suggest they are very lonely at one point in time, on average, appear to be less lonely when measured at a later point in time. Our meta-analysis of these studies indicated there was indeed a lowering of loneliness as measured before and after the interventions, but we cannot conclude from this evidence that the reductions in loneliness were due to the interventions. Eighteen studies utilized a non- randomized group comparison design in which some of the participants sought out the intervention (the experimental group) while others (the control group) did not. In this design, assignment of individuals to the experimental or control groups was based upon convenience, participant preference, or some other factor, which means the groups that did and did not receive the intervention may differ in ways that HOUSE_OVERSIGHT_021379