Assessing risk of bias in prevalence studies: modification of an existing tool and evidence of interrater agreement