The five-city, large-scale epidemiological study conducted by the U.S. government's National Institute of Mental Health (the NIMH-ECA study, 1982-1984) produced 1-year prevalence estimates of 0.8 - 2.2% across the five cities, with a mean of 1.6 ± 0.2% (S.E.). Lifetime prevalence estimates (excluding individuals with co-morbid major depression, schizophrenia or organic brain syndrome) were 1.7 ± 0.1% (S.E.). These estimates may be high, since both studies utilized trained lay interviewers to administer a structured diagnostic interview. Studies involving psychiatrists have reported lower prevalence rates.

The two largest epidemiolgical studies by Karno et al. in 1988 and Weissman et al. in 1994 utilized trained lay interviewers to administer a structured diagnostic interview. They found that the one year OCD prevalence rates ranged from 0.8% to 2.3% (mean = 1.6%). The reliability and validity of these rates have been called into question by several researchers. A reanalysis of Karno et al.'s data found that less than 20% of OCD cases met diagnostic criteria when reinterviewed by lay interviewers one year later. Since Weissman et al. utilized the same methods, those results are presumably similarly affected.

Stein et al. in 1997 designed a community prevalence study to overcome the problems inherent in using lay interviewers. Individuals identified by structured lay interviews as probably cases of OCD or of subclinical OCD were reinterviewed with structured instruments by a highly experienced research nurse. The nurse reviewed her findings with the principal investigator, who assigned all diagnoses, and sought additional information when so instructed. Only 24% of individuals identified as probable OCD cases were assigned a research diagnosis of OCD. The resulting weighted one month prevalence rate for DSM-IV OCD for the entire sample was 0.6% (95% confidence interval = 0.3% - 0.8%). Since subjects who reported no obsessions or compulsions to the lay interviewers were not reinterviewed, some OCD cases may have been missed, resulting in an underestimate of the true prevalence.

In 1998, Koran, Leventhal, Fireman and Jocobson (unpublished data) studied the prevalence rates for clinically recognized OCD in a large prepaid health plan, the Kaiser Northern California Health Plan, which has more than 1.8 million members. Chart reviews on all cases with an OCD diagnosis in the Plan's computerized data base produced a one year treated prevalence rate of 0.095% in adults aged 18 years or older. This is less than the 10% of Karno et al. and Weissman et al.'s reported rates, and only 15% of the more conservative rate reported by Stein et al..

The OCD prevalence rates reported in community and primary care studies far exceed the clinically recognized prevalence rate in the Kaiser data base. Despite possible reasons for under-estimation in this data base, and for overestimation of clinically significant OCD in earlier studies, the difference suggests that many Kaiser members with clinically significant OCD go untreated. The proportion of untreated individuals among those with other forms of health insurance is unknown, but, given the delay in seeking treatment noted earlier, it is reasonable to suspect that it is equally large.