An additional challenge in psychosocial intervention research relates to compliance with study measurements. A balance must be reached between patient burden and comprehensiveness. Every effort should be made to encourage patients to comply with study measurements. The problem of missing data points has been a troublesome issue in psychosocial research. This is particularly challenging when patients are seriously ill, or undergoing complex medical treatments, or when multiple outcome measures are being used.

Blinding should be used in outcome assessments to the extent possible, given the study design. When outcomes are self-reported, patients should be unaware of the specific study hypothesis (if possible), and research staff who score the questionnaires should be unaware of subjects’ randomization allocation. When outcomes are interview-based, interviewers should be blinded to randomization allocation.

Finally, cost and feasibility assessments to allow the possible balancing of benefits against costs of psychosocial interventions or programs should be included insofar as is possible and consistent with the abilities of the investigators, time and money resources available, and capacities of the settings. It is recognized that this is easier said than done, and that with the exception of an occasional study or mention in reports is rarely accomplished (Koocher and Pollin, 2001; Simpson et al., 2001).

Each randomized trial should have a clearly stated hypothesis and a clearly stated primary outcome; sample size calculations and statistical analyses should reflect this primary study question. The description of the primary outcome should include a description of the specific questionnaire and/or questionnaire item/scale as well as the specific time-point(s) in the study that will be used to define this outcome. If more than one primary study outcome (or time-point) is selected, allowances must be made for multiple testing in statistical analysis and sample size calculations. At times, the issue of multiple time-points can be overcome using statistical methods, such as repeated measures analysis of variance or slopes analysis, that incorporate measurements performed at multiple times into a single statistical analysis. Thus, although studies may use multiple instruments administered at multiple time-points with a plan for multiple approaches to analysis, the primary endpoint, time-point, and analytic approach must be stated a priori, and the sample size and significance cut-points must reflect these decisions. Other endpoints and/or analyses should then be viewed as secondary or hypothesis-generating. This will overcome the common criticism of psychosocial intervention studies that use of multiple outcomes, all of which are treated equally in the analysis, leads to multiple, and at times conflicting,