Table 2.

Assessment of the evidence base using Joanna Briggs Institute Prevalence Critical Appraisal Tool, systematic review of HIV prevalence among female sex workers in the United States

Item	Criteria	Yes	No	Unclear	Comments
1	Was the sample representative of the target  population?	0	0	14	While all participants were female sex workers in the United States,  the degree to which their demographic and HIV risk behaviors are  representative is unknown due to the lack of information  characterizing this hidden population
2	Were study participants recruited in an  appropriate way?	2	12	0	Since female sex workers are a hidden population, probability  samples are costly and logistically challenging. Respondent driven  sampling and venue-based sampling are widely used sampling  methods for this population. Two studies conducted probability  sampling. All other studies used convenience sampling
3	Was the sample size adequate?	3	11	0	No studies described the process to estimate sample size. We  estimated that studies needed a sample size of 384 to estimate a  prevalence of 10 % (95 % CI 7–13 %). Only three studies satisfied  this criterion
4	Were the study subjects and the setting  described in detail?	5	9	0	To satisfy this criterion, we required the study to report the following  information about the population: (1) demographics (race and age),  (2)selected HIV risk factors (injection drug use; condom use) and  (3)setting (place or venue were sex work occurred). Only five  studies reported all criteria. Only 10/14 reported injection drug use.  We intended to use these factors as potential covariates in meta-  regression; while the lack of reporting does not influence the overall  prevalence per se, it impacts our ability to understand risk factors  associated with HIV and whether prevalence varies based on  characteristics and to what degree
5	Was the data analysis conducted with  sufficient coverage of the identified  sample?	8	2	4	To satisfy this criterion, we required that more than 90 % of those  included in the sample completed an HIV test. Eight of the 14  studies reported testing more than 90 % of participants
6	Were objective, standard criteria used for  the measurement of the condition?	10	1	3	To satisfy this criterion, HIV-positivity had to be determined by an  antibody screening test followed by a confirmatory test. Ten studies  fulfilled this criteria, one conducted only one screening test and  three did not report the testing algorithm used
7	Was the condition measured reliably?	10	1	3	Ten studies used an adequate HIV testing strategy with an antibody-  based screening test followed by confirmation. Three did not  specify the tests used. One study conducted only a screening test  using oral fluid without a confirmatory test. The oral test currently  commercially available has a sensitivity of 91.7 %
8	Was there appropriate statistical analysis?	13	1	0	The primary objective of this review was to determine prevalence  and then use meta-regression and sub-group analyses to explore  differences among studies and generate adjusted estimates as  appropriate. For convenience samples we only required that studies  report the number of participants with a positive HIV test and the  total number of individuals in the sample. For probability samples  we required for the studies to conduct weighted analyses. Of the  two probability-sampling studies, only one conducted weighted  analysis
9	Are all important confounding  factors/subgroups/differences identified  and accounted for?	6	8	0	To meet this criterion, we required for studies to report HIV  prevalence at least by one important sub-group including: race,  setting were sex work occurred, injection drug use or number of  partners. Four studies reported HIV prevalence by injection drug  use, one by number of partners and one by race
10	Were subpopulations identified using  objective criteria?	12	2	0	Sub-group membership was identified based on self-reports that  suffer from social desirability bias and recall bias. However, in  practice there is no other way to collect such behavioral data than  by self-report