Psychological autopsy studies in high-income countries (1) have consistently found that at least 90% of persons who die of suicide are suffering from a mental disorder at the time of death. In China, however, there are now two well-designed independent psychological autopsy studies that have found substantially lower rates of mental disorders in suicide decedents. A nationally representative psychological autopsy study of 895 adult suicide victims conducted about a decade ago (2) reported a prevalence of mental disorders of 63%. In this issue, in a study of 392 rural suicide victims 15—34 years of age, Zhang and colleagues (3) report that only 48% had a current mental illness. The prevalence in the subsample of the earlier study who were 18—34 years of age (N=114) was 45% (4), similar to the rate seen in this new study.

Are these low rates of mental illness in Chinese suicide victims reliable? Zhang et al. used a Chinese version of the Structured Clinical Interview for DSM-IV (SCID) with proxy informants to make psychiatric diagnoses. All psychological autopsy studies employ proxy informants to make diagnoses, so this should not in itself explain the lower rate of mental disorders in Chinese suicide decedents. But in China a low proportion of individuals with mental illnesses ever receive treatment (5), so unlike in Western studies, the proxy-based diagnoses cannot be augmented by referring to information in clinical records. Studies have found that the SCID can be used reliably in China (5), but clinicians need to be flexible in their use of the SCID probes to make the instrument sensitive to cultural variation in the manifestation and expression of symptoms. Thus, the main methodological concern with the Zhang et al. study is whether or not the use of nonpsychiatric interviewers—who may be less able to flexibly change probes when respondents do not understand the standard SCID probes—led to an underestimation of the rate of mental illness. Interrater reliability was assessed by having interviewers code three taped mock interviews. But three interviews are too few to assess diagnostic reliability, and this method assesses interviewers' ability to code a standard interview conducted by someone else—it does not assess their ability to independently conduct the interview. And the much lower rate of mental illnesses in the 416 comparison subjects (3.8%) relative to that reported in community members 18—34 years of age (12.5%) who participated in a large psychiatric epidemiologic study in China in which the SCID was administered by psychiatrists (5) suggests that the interviewers in the Zhang et al. study may have been less thorough in their investigation of comparison subjects or that the proxy informants of living comparison subjects may have been more reluctant to report psychological problems in their associates. Despite these caveats, the close similarity of the study's reported prevalence of mental disorders in young suicide decedents to that of the previous independent psychological autopsy study (48% and 45%, respectively) gives us some reassurance that the overall prevalence reported in the suicide decedents is reasonably close to the actual rate.

There are many potential factors that could explain the relatively low reported prevalence of mental disorders among suicide decedents in China compared to rates reported in high-income countries. Although the prevalence of substance use problems has recently increased, it still remains much lower in China than in high-income countries, particularly among women (5). Psychological autopsy studies in China have not considered the full range of personality disorders—primarily because of doubts about the validity of the DSM-IV personality disorders in China. A study of 505 suicides in China showed that impulsive personality traits are an important risk factor for suicide (6), but the impulsive individuals among these suicide decedents did not have the serious psychosocial dysfunction that is required to diagnose a personality disorder. Another potential factor is the dichotomous nature of psychiatric diagnoses; the risk of suicide in China is linearly related to the severity of depressive symptoms (7), so many individuals without a diagnosis are still at substantial risk of suicide as a result of subsyndromal symptoms. This may be a more important factor in China and other cultures where the manifestation of dysphoric affect may not neatly match diagnostic criteria that were established using Western samples. And, finally, the frequent use of pesticides as a method of self-harm—used in 58% of suicides and 28% of suicide attempts in China (8)—increases the overlap between nonfatal and fatal suicidal behavior and thus could increase the proportion of suicide decedents without mental illnesses who carried out impulsive acts with little intention to die (6).

Why is this important? China accounts for up to one-third of global suicides (1), so the lower prevalence of mental disorders in Chinese suicide decedents has important implications for both the theoretical modeling of suicidal behavior and the development of international suicide prevention strategies. If a current mental illness is not a prerequisite for suicide, models of suicide in which risk and protective factors exert their influence via their effect on the occurrence and severity of mental disorders will need to be reconsidered. And if a substantial minority of suicide decedents do not have a mental illness, preventive approaches focused solely on the recognition and management of mental illnesses will need to be supplemented with other strategies. Moreover, 85% of suicides worldwide occur in low- and middle-income countries (9), and ingestion of pesticides is the most common method of suicide (10), so the situation in China may not be as atypical as it appears from the Western perspective.

Building more mental health centers in rural China or training more village doctors to identify and treat depression may be useful for other reasons, but these strategies will not prevent impulsive pesticide ingestion in persons who are in the midst of an intense argument with their spouse. The study by Zhang et al. highlights the importance of low social support both as an independent risk factor for suicide in young rural residents and as a magnifier of the risk due to mental illness. This new finding suggests that developing family-based and community-based social and psychological support networks should be a central component of suicide prevention activities. Given the high proportion of suicides by pesticide ingestion and the frequent occurrence of acute interpersonal conflicts in suicide decedents, other important suicide prevention strategies that should be considered in this setting are means restriction and enhancement of conflict resolution skills in children and young adults.

The conventional wisdom that suicide is almost always the outcome of mental illness will not be altered by one or two studies from China. Psychological autopsy studies, dependent as they are on retrospective data from proxy informants, have several fundamental problems, so there will always be some methodological nuance that supporters of the conventional wisdom can cite to question the validity of such studies. What we can do is gradually improve the methodology of these studies and increase the number of such studies conducted in rural areas of low- and middle-income countries, where the majority of suicides worldwide occur. Zhang et al. recommend increasing the case-control ratio in such studies, thus allowing for a better assessment of the relative importance of factors that may be uncommon in comparison subjects. I would also recommend several other steps: cultural adaptation of the probes used to assess symptoms (while maintaining the DSM or ICD diagnostic criteria); rigorous assessment of interviewer reliability by having instrument-trained clinicians conduct independent interviews with the same respondents; inclusion of continuous measures of symptom severity and personality traits in addition to the dichotomous diagnostic categories; assessment of the magnitude of the psychological impact of negative life events rather than simply recoding the number of negative life events; identification and assessment of culturally relevant protective factors; recording information about decedents' access to means; parallel studies of attempted and completed suicide to assess the degree of overlap between the two types of self-harm; and inclusion of qualitative in-depth interviews with informants to help clarify the local meaning of the quantitative results obtained.