We excluded conference abstracts because the STROBE and CONSORT criteria were designed to evaluate published manuscripts check details and abstracts lack the necessary data required for evaluation.[9, 10] Lastly, if the pharmacist participated in the study intervention, but the research was not designed to examine the HIV pharmacist’s contribution, the study was excluded. In these types of studies, the pharmacist’s involvement was often limited to antiretroviral dispensing only or dispensing directly observed therapy only. The authors considered several tools

to assess thoroughness of reporting in manuscripts and opted to apply STROBE criteria to observational studies and CONSORT criteria to randomized studies.[9, 10] Criteria were rated as present (yes), absent (no) or not applicable to the study. Compound criteria, which included multiple assessments, were separated for consistency of evaluation. Authors held a preliminary discussion of each criterion in the STROBE and CONSORT checklists to guide JQ1 the initial interpretation. A small list of additional criteria deemed important by the authors and

relevant to studies of HIV pharmacists were assessed separately from STROBE or CONSORT criteria. These included concordance of the declared study design with Cochrane classifications, description of HIV pharmacist training or prior experience and evaluation

of key outcomes measures such as adherence, CD4+ cell count and HIV-1 viral load.[6] Each study was independently reviewed by two authors (JC/PS or JC/BD) for presence or absence of the required STROBE, CONSORT or additional criteria. Inter-reviewer agreement on the criteria evaluated for each study was assessed using an unweighted kappa statistic, calculated for each study using STATA version 12.0 (StataCorp, College over Station, TX, USA).[11] If the two primary reviewers had different ratings (present versus absent versus not applicable) on a particular criterion within a manuscript, the reviewers met to see if the disagreement could be resolved through discussion. In the seven instances that a disagreement could not be resolved through discussion, a third author was asked to review the study and provide final input. Descriptive statistics were used to calculate the frequency at which studies satisfied various criteria and the overall proportion of criteria that each study satisfied. The criterion inclusion rate was summed across all observational studies and divided by 19, or summed across all randomized studies and divided by three, to obtain an average inclusion rate. Of the initial 1545 citations, 1477 were discarded after abstract review because they were duplicate or irrelevant. The remaining 68 articles were reviewed by two authors (PS/JC).