In Reply: We thank Dr. Cournos and her colleagues for their letter. We share their belief that clinicians and policy makers should interpret these (or any) findings with due regard for data limitations. In our article, we frequently noted the limitations of our data, and we welcome the opportunity to elaborate.

Although we view the existing literature as too limited and variable to draw firm conclusions about differences in HIV testing rates for the groups studied, we agree that rates are likely to be higher among persons with a substance use disorder, and we specifically cautioned that “the increased likelihood of HIV testing found among people with mental illness who have substance abuse problems” could have influenced our findings. However, although differential testing may have had some influence on the significant elevations of new HIV diagnoses found in the substance use disorder groups, this does not convince us to dismiss questions raised by the finding of notably low rates of new HIV diagnoses among persons with severe mental illness but no substance use disorder, particularly because the findings resemble the pattern found for schizophrenia and HIV in a prior study using administrative data (1). Even if people with a substance use disorder diagnosis are more likely to get tested for HIV, this does not explain why patients who had only serious mental illness were 23% less likely to acquire an HIV diagnosis than those with no serious mental illness and no substance use disorder diagnosis.

More generally, we do not find it difficult to imagine a subgroup of people with serious mental illness but without a substance use disorder who, perhaps as a result of stigma and social isolation, have such limited social contact that HIV transmission opportunities are few. Yet despite this possibility, which would be consistent with study findings, we specifically noted that we also find it clinically plausible that some symptoms of serious mental illness might in themselves increase risk for those without a substance use disorder. (It is possible, for example, that claims data might fail to detect risk elevations from a risk-increasing symptom that somehow also simultaneously decreased the odds of testing.) As we noted in the article, this clinical plausibility means that definitive conclusions should not be drawn without further research, and we several times stressed that HIV-related prevention and services should be part of treatment for all patients with serious mental illness.

New and different data can always clarify or correct claims made for previous findings, and we agree that routine testing can shed new light going forward. Institutional challenges of implementation will likely mean that these data will also have limitations to be considered, and they cannot provide information on the period we studied. However, we hope that testing data will tell us more about risk and infection rates among groups of persons with serious mental illness whose substance use is too limited, intermittent, or covert for them to acquire a substance use disorder diagnosis. More important, routine testing can speed effective treatment to people with HIV.