Why was a revised report issued?

In the report, "HIV-Associated Behaviors Among Injecting-Drug Users—23 Cities, United States, May 2005–February 2006" three errors occurred: 1) the variables "Shared injection equipment" and "Tested for HIV infection" were calculated incorrectly; 2) inaccurate data were used for the size of the injecting drug user (IDU) population in each metropolitan statistical area, which in turn was used to obtain the overall estimated prevalence of each behavior, and 3) the standard errors were incorrectly reported. In addition, during re-analysis, more rigorous standards for the respondent-driven sampling estimation were applied.

Where can I find the revised report?

The revised report is located here.

How were the variables “Shared injection equipment” and “Tested for HIV infection” calculated incorrectly?

The variable for shared injection equipment did not correctly classify all those who shared equipment. In re-examining the calculation, it was determined that some people who had shared injection equipment were classified as not having shared. Thus, the revised estimate of the prevalence of sharing is higher than was originally reported.

The variable for “tested for HIV infection” incorrectly excluded some persons from analysis who should have been included, but who had not been tested for HIV in the past 12 months (i.e., comprised part of the denominator). Thus, the revised estimate of the prevalence of testing is lower than was originally reported.

How were the data for estimating the size of the injecting drug user (IDU) population inaccurate?

There were two problems with the data we used for the size of the IDU populations. First, we used the population rate of IDUs in each of the metropolitan statistical areas (MSA) reported in Brady et al (2008) to multiply by the MSA population size to get an estimate number of IDUs in each of the MSA. During this process, the population size that was used for San Juan, PR, was only for the municipio (city) of San Juan, and not the entire MSA. Secondly, we originally used data from Brady et al. for the year 1999 (before MSA definitions changed); during re-analysis, we determined that the 2002 estimates, which were closer to the time of the NHBS data collection, would be a more accurate estimate of population size for our purposes. The data used in the revised report for the injecting drug user population size in each MSA come from the online appendix available with electronic versions of the Brady et al. paper.

Reference: Brady JE, Friedman SR, Cooper HL, Flom PL, Tempalski B, Gostnell K. Estimating the prevalence of injection drug users in the U.S. and in large U.S. metropolitan areas from 1992 to 2002. J Urban Health 2008;85:323–51.

Why were the standard errors reported in the April 2009 MMWR incorrect?

Because the calculation of some variables and the data used for the size of the IDU population in each MSA changed, the revised value of the standard errors differs from the April 2009 report. Also, in the original table the standard errors were incorrectly calculated due to a missing step in the aggregation process.

How were the analysis standards used in the re-analysis more rigorous than the original analysis?

In the re-analysis, we did not include certain problematic values when aggregating the estimates across the 23 cities. The aggregation excluded prevalence estimates for groups with small cell sizes (<10) because these data artificially reduced the standard errors, suggesting more precision in the data than was warranted. We also did not include in the aggregation certain cases where the Respondent-Driven Sampling Analysis Tool software could not produce accurate prevalence estimates or confidence intervals due to the nature of the data (high homophily). Overall, less than 3.5% of the data was not included in the aggregation for these reasons.

Why do so many of the prevalence estimates in the updated table differ from the original?

By changing the source of the size of the IDU population for each MSA and using the accurate (i.e., higher) IDU population size for the San Juan MSA, the overall prevalence estimates for every variable changed at least a little bit. The variables “Shared injection equipment” and “Tested for HIV infection” changed considerably due to the way they were re-calculated (see above) and because of the changes to the size of the IDU populations.

How did you verify the new analyses?

Two people independently analyzed the data. They compared their results and reconciled any differences.

How do the changes in the results affect the conclusions and recommendations?

The conclusions and the recommendations from the original MMWR report are still applicable. The revised data underscore the importance of addressing drug-related risk behaviors such as sharing injection equipment and the continued need to promote HIV testing among IDU.