Dr Michel Odent brings up an important issue. He suggests that
ecological data should be used for testing a cause and effect relationship
by comparing suicide rates in different countries. Such studies on
aggregated data (for example whole populations in countries) have
important applications but also limitations.
The origin of the present study was a serendipitous finding twenty
years ago of a remarkable regularity in the suicide rates of five-year
birth cohorts in the United States during 1937-75. After an extensive
search considering some thirty variables, of which eleven were thoroughly
tested, I found a possible explanation. Path analysis showed for example
that the suicide rate 1971-75 for ages 15-24 in the 48 contiguous states
was significantly correlated only with three variables during relevant
periods for the studied birth cohort: parental alcoholism, broken homes
and birth injury. Variables such as income, poverty, unemployment,
perinatal 'asphyxia' and congenital malformations were not significant. Of
the three significant variables birth injury had the largest path
coefficient. This led to the hypothesis, which could not be rejected on
the basis of the data. Ecological data are useful for developing
hypotheses.
But an association observed between variables at an aggregate level
(states in this case) does not necessarily represent the association that
exists at an individual level. Not reckoning this difference is generally
referred to as the 'ecological fallacy' (1). Hence, suicide rates in whole
countries or states might be associated with certain obstetric practices
without having any relevance for the individual cases. It is necessary to
make studies on individual cases to reach any conclusion about
causalities.
Dr Odent rightly draws attention, however, to the interesting
circumstance that the suicide rate is exceptionally low in the Netherlands
and that the organisation of obstetrics and midwifery is unique. This
observation certainly warrants further studies, particularly at the
individual level.
Reference
1. Last JM. A dictionary of Epidemiology. Oxford University Press
1995
Rapid Response:
Applicability of ecological data
Dr Michel Odent brings up an important issue. He suggests that
ecological data should be used for testing a cause and effect relationship
by comparing suicide rates in different countries. Such studies on
aggregated data (for example whole populations in countries) have
important applications but also limitations.
The origin of the present study was a serendipitous finding twenty
years ago of a remarkable regularity in the suicide rates of five-year
birth cohorts in the United States during 1937-75. After an extensive
search considering some thirty variables, of which eleven were thoroughly
tested, I found a possible explanation. Path analysis showed for example
that the suicide rate 1971-75 for ages 15-24 in the 48 contiguous states
was significantly correlated only with three variables during relevant
periods for the studied birth cohort: parental alcoholism, broken homes
and birth injury. Variables such as income, poverty, unemployment,
perinatal 'asphyxia' and congenital malformations were not significant. Of
the three significant variables birth injury had the largest path
coefficient. This led to the hypothesis, which could not be rejected on
the basis of the data. Ecological data are useful for developing
hypotheses.
But an association observed between variables at an aggregate level
(states in this case) does not necessarily represent the association that
exists at an individual level. Not reckoning this difference is generally
referred to as the 'ecological fallacy' (1). Hence, suicide rates in whole
countries or states might be associated with certain obstetric practices
without having any relevance for the individual cases. It is necessary to
make studies on individual cases to reach any conclusion about
causalities.
Dr Odent rightly draws attention, however, to the interesting
circumstance that the suicide rate is exceptionally low in the Netherlands
and that the organisation of obstetrics and midwifery is unique. This
observation certainly warrants further studies, particularly at the
individual level.
Reference
1. Last JM. A dictionary of Epidemiology. Oxford University Press
1995
Competing interests: No competing interests