Marcel Slockers

81 Changes in the causes of death of homeless people in Rotterdam tuberculosis), ischaemic heart disease, substance misuse, and external causes including unintentional injuries, suicides, homicides and accidental poisoning (from medication and illicit substances). Our study confirmed highest excess mortality from infectious diseases, psychiatric disorders and unnatural death, with high excess mortality for accidental poisoning (including drug overdoses) and intentional injury (suicide and homicide). The proportion of unnatural deaths, i.e. 26%, in our homeless cohort almost equals the results of a large Danish study that found 28% unnatural deaths (as a proportion of almost 5000 fatal cases). 8 We obtained our results in a comprehensive sample of homeless men and women using a wide range of services provided by both local communities and the church. Another strength of our study is the availability of reliable and valid data on unnatural causes of death. In the Netherlands, all deaths due to accidents, suicide and violence are assessed and confirmed by forensic specialists of the municipality. Statistics Netherlands uses a standard procedure for this aim, which has remained unchanged between 2001 and 2010 and justifies the absence of registration bias in our comparison of two time periods. But our study has some limitations as well. Homeless people who did not use any local service in 2001 were not included, not all institutes could provide data for the year 2001, persons without a legal status could not be matched, and homeless persons dying in other countries after migration were missed. Moreover, we had a small study sample for an analysis of mortality by specific causes of death. For several causes of death we observed rather wide confidence intervals in the Standardized Mortality Ratios and Hazard Ratios, and due to low numbers we were not allowed to conduct separate analyses for men and women or in-depth analyses for small causes of death (e.g. homicide). It should also be considered that our study design, i.e. an uncontrolled before-after study, has shown ecological associations between social policy measures and mortality reductions, which may generate hypotheses but does not allow conclusions on causal relationships. Another limitation is that we compared the hazard of mortality in the period 2001–2005 with the hazard of mortality in the period 2006–2010 for a cohort of homeless persons that were identified as such in 2001, without knowing whether they were still homeless in 2006 and without knowing the level of enrolment in the social programmes that were provided. Because of these limitations it is not possible to attribute the difference in hazard rates between periods to the social policies. It should also be considered that we did tests for all presented causes of death categories and spurious results due to multiple testing can therefore not be excluded.