There are many uncommon conditions in the United States, which are largely unstudied outside of clinical settings. The prevalence rates of these conditions are too low to be represented in even the largest health surveys. Consequently, there are no reliable estimates of the prevalence, disease or treatment characteristics of these populations. The population prevalence for one such condition – primary immune deficiency disease – was established by a national RDD telephone survey of 10,000 households in 2005. This survey identified 23 individuals with legitimate diagnoses of primary immune deficiency diseases out of approximately 28,000 persons in these households. The household sample yielded a precise estimate of the prevalence of these diagnoses in the United States, but the patient sample was too small to reliably characterize the disease and treatment. The current study used a large national Census-balanced online panel as an alternative approach to obtaining a larger, community based sample of patients with a rare disease. A total of 850,000 unique panelists were sent generic invitations to participate in a survey. Ultimately, the online panel produced a total of 159 respondents (with 174 qualifying patients in the household or immediate family) who met the diagnostic criteria for inclusion in the study. This survey provided the first national, community based estimates of the treatment of this rare disease with a large enough sample for reliable estimates. The findings from the non-probability online survey were remarkably consistent with those from the small, but probability prevalence survey. Although no substitute for probability samples, this study suggests that online panels may play a useful role in the monitoring of treatment and health outcomes in rare diseases where adequate community samples cannot be obtained from population based sampling frames.