Measuring sexual minority status in population surveys may result in misclassification due to the low prevalence of categories needed to capture a range of identities. While some response options clearly map to sexual minority (e.g., "gay") or non-sexual minority status (e.g., "heterosexual"), other options may not for respondents unfamiliar with the category names or whose preferred response does not appear. We use ancillary sociodemographic information from 2013-2014 National Health Interview Surveys to predict the probability of self-identifying as a sexual minority among those selecting "something else [SE]" or "don't know the answer [DK]". The odds of identifying as sexual minority among SE/DK respondents are higher for SE and urban respondents but lower for older, Hispanic, and low-income/missing-income respondents. We develop respondent-level probabilities of sexual minority status to represent sexual minorities without excluding or misclassifying potentially ambiguous responses and compare the effects of three health disparity estimation strategies: 1. using probabilities; 2. hard-classifying SE as sexual minority/DK as non-sexual minority; 3. dropping DE/SK cases.