Suicide Research Based on Danish Registers

Abstract

In the last 5 years there has been a series of remarkable studies on suicide from Denmark based on their longitudinal case registers. These registers include the Danish Psychiatric Central Register, which since 1969 has recorded data on all psychiatric inpatients; the Integrated Data base for Labor Market Research, which contains sociodemographic data on all Danes, as well as market conditions; the Danish Civil Registration System, which contains a unique personal identifier which allows linkage to first degree relatives, the Cause of Death Register, and the Danish Twin Register (Qin et al., 2003; Tomassini et al., 2003). Although there had been descriptive data of over 50,000 suicides published in the mid-19th century in France by Lisle (1856), until the publication of this Danish work there had been nothing to match the sophisticated analyses on such large numbers of persons. For example, Qin et al. (2003) reported on 21,169 persons who died by suicide between 1981 and 1997 and 423,128 live comparison subjects. They used the population attributable risks (PAR) statistic, which allows for the delineation of the proportion of a condition that could be eliminated if certain risk factors could also be eliminated, thereby placing risk factors in perspective rather than simply acknowledging that any specific individual factor may be of statistical significance. By so doing they extended earlier work of Mortensen et al. (2000), and demonstrated that suicide risk was overwhelmingly associated with mental illness requiring hospitalization, which had a PAR of 40.3%. This was in contrast to the next significant factors being 25.8% for single marital status, 10.2% for being an aged pensioner, and 8.8% for having the lowest quartile income, factors which are hardly amenable to any intervention. These factors were in contrast to the PAR of 2.8% for unemployment, which, although still statistically significant and of importance in its own right, is probably of little practical relevance in terms of any broad community programs designed to prevent suicide. Mortensen et al. (2000) previously commented on the significance of such analyses, and they noted that “Reports of high relative risk and attributable risk associated with unemployment and other socioeconomic factors may be confounded and overestimated owing to the lack of adjustment for the association with mental disorders.” The PAR statistic allows for such adjustment, and it is only by having access to such large data bases and using newer statistical techniques that this perspective can be fully appreciated. In other studies examining those with mood disorders, Hoyer et al. (2004) reported that suicide was the cause of death for 6% of 53,466 patients who were admitted for an affective disorder between 1973 and 1993, and Kessing (2004) analyzed data more recent than the 10th version of the International Classification of Diseases (ICD) in Denmark in 1994 to demonstrate that there was an increasing gradation of suicide risk for those who had mild, moderate, and severe depression. This demonstrated the utility of the ICD diagnostic subcategories based on a dimensional view of depression, and confirmed clinical beliefs about the prognosis and suicide risk of those with increasing degrees of depression. A further demonstration of the powerful statistical capabilities of these registers is provided by a study examining the effect of children in modifying suicide risk in adults. It has generally been considered that the presence of children is a protective factor, and Qin and Mortensen (2003) have confirmed this in research on 18,611 suicides between 1981 and 1997. Indeed, the numbers allowed detailed analyses which demonstrated that parents with children who died or who had been hospitalized for psychiatric illness had a greater risk of suicide, and the risk was even higher if the death of the child was by suicide. Similar clinically useful findings confirming the value of family ties as a protection against suicide have emerged from the twin study of Tomassini et al. (2003), who followed 21,653 individual twins who were alive in 1943 for