Abstract
Dear Editor
Thank you for asking me to comment on the letter by Dr Lovisi in response to my cross-national study examining the relationship between elderly suicide rates and the prevalence of internet users (Shah, 2010a). He is right in stating that not all the details of the studied countries were provided. The simple reason for this was that some of the variables that may influence the relationship between elderly suicide rates and the prevalence of internet users at a national level were ‘controlled’ for in the multiple regression analysis. These included: life expectancy, income inequality, gross national domestic product and the proportion of older people in the general population. Two important measures of socioeconomic status were used in the multiple regression analysis: income inequality and gross national domestic product. Thus, there was no need to use stratification by income suggested by Dr Lovisi. There are several problems with using such categorical stratification. First, the number of countries in each group would be greatly reduced resulting in a study with lower power and the consequent risk of type 1 and type 2 statistical errors. Second, income inequality and gross national domestic product are continuous variables, and thus by stratification not all the data set would be used in the analysis; the analysis would be fragmented. Third, stratifying socioeconomic status into different group requires agreed, valid and reliable definitions of such stratification groups. This can be problematic because there is no consensus on which measure of the two used is the better measure in this context and definitions of socioeconomic groups for stratification are fraught with difficulties. Many of characteristics of the studied countries are available in previous papers (Shah et al., 2007, 2008).
However, the findings were clear in that elderly suicide rates were positively associated with the prevalence of internet users independently of two different measures of socioeconomic status. These findings were similar to those observed for general population suicide rates also (Shah, 2010b). I am grateful that Dr Lovisi agrees with my view that this association does not establish a causal relationship. Dr Lovisi states:
The website, as the other suicide methods, will act as intervening variable, increasing the risk of suicide in vulnerable subjects. Therefore, on the causal pathway of suicide, the website will be an intermediate variable.
I could not agree more with this comment, provided we accept that there is an absence of epiphenomenon whereby a third variable independently influences both suicide rates and the prevalence of internet users in the same direction. For this reason, I suggested further research including: in-depth case studies of elderly suicide attempters addressing their use of the internet using qualitative methodology; examination of suicidal intent in elderly individuals who use ‘pro-suicide’ websites and chat rooms; and psychological post-mortem studies carefully examining the use of the internet by suicide victims – technological advances now allow investigation of the websites and chat rooms that may have been used by accessing the relevant computers. This will clearly help establish if the relationship is causal and whether or not the use of the internet is merely an event (and pehaps a mechanism) in the pathway to suicide just as purchasing paracetamol is.
I strongly disagree with Dr Lovisi’s statement:
In spite of being very important to implement legislation to regulate websites and chat rooms, for purposes of secondary prevention interventions for suicide, however, it is more relevant to know primary causes of suicide.
I believe it is equally important to understand the primary causes of suicide and the methods used for suicide. There clearly is good evidence that removal of methods of suicide can dramatically reduce suicide rates. They fell in England and Wales after detoxification of domestic gas.