conducted with about 80 percent of those interviewed at the first wave. The DIS portion of the interview consisted of specified questions directly pertinent to diagnostic criteria from the third edition of the Diagnostic and Statistical Manual (DSM-III) of the American Psychiatric Association. The revised DSM-III-R is the best classification system available at the current time (DSM-IV is in press) and is the basis for the definitions used throughout the rest of this report. However, operational definitions of specific mental disorders as given in DSM-III, when major diagnostic criteria changes were made, were used in the ECA study and are used for determining onset in the analyses below. Onset is defined here as the first diagnosis of the disorder. Diagnoses were made from the DIS symptom data by means of computer algorithms that simulated the application of DSM-III criteria (Boyd, Burke, Gruenberg, Holzer, Rae, George et al., 1984). Because the ECA results depended on both the diagnostic criteria chosen and the method of ascertainment, the mental disorders as classified therein are referred to here as “DIS/DSM-III disorders.”
The absence of firm data on the validity of the DSM-III system for classifying mental disorders enjoins us to be careful about conceptualizing the process of disease onset. It is particularly difficult to establish the validity of a threshold for the presence versus the absence of disorder, because signs and symptoms of mental disorders are widespread in the population and do not always reflect the presence of a mental disorder. From the clinical standpoint, subtle differences in how behaviors are categorized may suggest quite varied thresholds for making diagnoses; from the epidemiological standpoint, subtle differences in threshold may produce widely varying prevalences.
A disorder that has a complex causal chain, where no particular cause is regarded as sufficient, may be preventable up to the point of onset, when the individual meets full criteria for diagnosis. The concept of attributable risk (see Chapter 4) allows quantitative comparison of risk factors in the population, which helps in selecting risk factors for intervention programs. Risk factors having high attributable risk for a specific disorder, or especially risk factors having high attributable risk for multiple disorders, would be prime targets for preventive interventions. But each risk factor may operate differently, and may be differentially malleable, that is, modifiable, at different times in life. Also, each risk factor may have a sensitive period in which an important contribution to the disorder may occur only at a particular phase of