Our results are subject to several limitations. While the study designs are robust in the 4 national surveys, they are hampered by small proportion of participants reporting injection drug use. Because PWID are a small proportion of the general population, obtaining adequate numbers to produce stable estimates is difficult without very large sample sizes. This difficulty is exacerbated when estimates are stratified by sex, age, or race/ethnicity. In addition, the illicit nature of and stigma associated with injection drug use may have resulted in under-reporting of this behavior; however this bias should be mitigated in part by use of ACASI for most surveys included in our analysis. A second limitation is coverage bias. The surveys in the meta-analysis exclude individuals without stable housing. Given that a high proportion of PWID are unstably housed [27], [28], they are likely underrepresented in our analysis. This coverage bias would result in an under-estimate of the population proportion of PWID and an over-estimate of disease rates. A third limitation is the degree of heterogeneity among surveys. Although all surveys are population-based, the sampling methods,
