Marcel Slockers

113 General discussion homeless people in the shelters of other welfare and religious organizations. We also used information from care workers who tracked people on the street. However, despite all our efforts, the study was unable to reach some groups with possibly even higher mortality risks than homeless people we monitored, such as undocumented immigrants and prostitutes with addiction problems. In addition, some of the homeless people from 2001 may have found housing - and as a result have lower mortality rates - during the period studied. Our data on the excess mortality may therefore still be an underestimate of the situation in Rotterdam during the period 2001-2010. We were able to link our homeless cohort on an individual level to the causes of death reg- istration of the Central Bureau of Statistics (CBS), in which there were no major changes in the period studied and in which unnatural causes are recorded based on findings of forensic physicians. Cause of death patterns among homeless people in Rotterdam and possible changes in these patterns are therefore well reflected by our study. However, we did have the limitation of a relatively small size of our homeless cohort, meaning we were not allowed to conduct further analyses around causes of death with small numbers of deaths. This prevented us, for example, from looking at the separate contributions of suicide and homicide to the reported decline in unnatural death following the introduction of the Action Plan for Social Support. The mortality, life expectancy, and cause of death patterns we found apply to documented homeless people in Rotterdam. Because of possible regional and local differences in homeless populations, the results cannot be simply generalized to all documented homeless people in the Netherlands. The different social setting in other countries, for example due to different policies toward alcohol and drugs and different health care systems, makes generalisability to foreign countries difficult. However, our results may well be reasonably comparable with mortality among homeless people in the other three major cities in the Netherlands. The Monitor Action Plan for Social Support 2012 report describes well how homelessness arises in the four different cities, Amsterdam, Utrecht, The Hague and Rotterdam. No major differences between the four cities were identified. For example, the percentage of evictions per 100,000 social housing units in the big cities varies between 0.35-0.56%. The shares of ex-prisoners who contact shelter services are also quite similar. The data used for the mortality study come from the period 2001-2010. These do not reflect the current situation because many social and policy developments have taken place since then, such as the financial crisis, the ending of the Action Plan for Social Support, the termination of health insurance for homeless people without an address, the uninsured schemes that followed, the sharp increase in the number of homeless people, and the corona crisis. As many of the developments after the period we studied likely had a negative