In this paper, we review challenges encountered when utilizing clinical data collected from non-specialists about persons with intellectual disabilities and psychiatric disorders. We draw on our experience with a dataset of adult inpatients and outpatients drawn from the nine psychiatric hospitals in Ontario (see Lunsky et al., 2003). First we discuss the problem of identifying patients with intellectual disabilities based on standard information collected in hospitals. Many hospitals categorize their patients by primary or secondary diagnosis only, so that a co-existing intellectual disability may or may not be recorded. Then we discuss problems related to the accuracy of psychiatric diagnoses made by individuals with limited training with the population. We found major differences in diagnostic patterns between specialized dual diagnosis programs and more generic programs. We conclude the paper with the argument that it is important to study dual diagnosis not just within the population of persons with intellectual disabilities but also within the broader mental health system. Doing so, however, can lead to problems as such data has its limitations.