Marcel Slockers

77 Changes in the causes of death of homeless people in Rotterdam to non-public microdata from Statistics Netherlands. We were unable to link 4 deaths to the cause of death database of Statistics Netherland We used the International Classifica- tion of Diseases (ICD 10) to classify all fatalities in 8 main categories: unnatural death (V010-Y899), infectious diseases (A000-B999), cardiovascular diseases (I000-I999), cancer (C000-D489), psychiatric disorders (F000-F999), respiratory diseases (J000-J999), gastrointestinal diseases (K000-K939) and,- other diseases (all other ICD 10 codes). Unnatural deaths could be further subdivided into: intentional injury (suicide and homicide combined),- accidental poisoning, -other accidents (all accidents other than poisoning combined, e.g. traffic accidents, accidental falls, accidental drowning, burns). To prevent personal identifications within our small dataset (n = 261), further distinctions were not allowed. We could therefore not separately study homicide deaths, but in addition to the combined category ‘intentional injury’ separate results for suicide could be produced. Data on mortality by cause of death for the population in Rotterdam were obtained from Statistics Netherlands (http://statline.cbs.nl) . Statistical analysis For the total period 2001–2010 we calculated the proportion of death by cause of death. We also calculated Standardized Mortality Ratios (SMRs) for this period to estimate excess mortality by cause of death among homeless people compared to the general population of Rotterdam. In the Netherlands, social policy measures to improve the living conditions of homeless people in the four largest cities were introduced in 2006. We calculated Hazard Ratios adjusted for age and sex to compare mortality rates by cause of death among homeless people in the period after (2006–2010) vs. before (2001–2005) implementing these measures. We started with a model without period, including only age classes and gender as independent variables, and compared this with a model within addition period (dichotomous). A significant HR for period indicates that the mortality hazard differs sig- nificantly between the two periods. Results The presented cohort consisted of 1870 men and 260 women. The mean age was 40.3 years. The mean follow up was 8.4 years and a total of 17909 persons years were observed. In total 265 homeless persons died during follow up (232 men-and 33 women). For 261 deaths the cause of death information was available.