The main finding of this study is that the health status of immigrants from Eastern Europe, expressed by their health satisfaction, deteriorated to a much larger degree than that of native Germans, in spite of substantial improvements in all indicators of SES that the immigrants experienced.
Hypotheses H2 and H3 were supported by the findings. Immigrants were socioeconomically disadvantaged relative to native Germans, although this gap became smaller over time. Nevertheless, the decline in health satisfaction among the immigrants was much steeper than among native Germans over the same time period. The regression model shows that after multivariate adjustment, this decline is primarily associated with being an immigrant, and not with socioeconomic status. This finding possibly reflects the unfavourable health conditions and high prevalence levels of risk factors the immigrants were exposed to in Eastern Europe. These might include a high prevalence of smoking, hypertension, deficiencies in antioxidants due to a low consumption of fruits and vegetables or frequent binge drinking, which all have been discussed as determinants of the mortality crisis in Eastern Europe. Some immigrants might have already had subclinical disease upon arrival in Germany which only after some time became apparent and caused their health satisfaction to deteriorate. Immigrants who arrived after 1990 had an even higher risk of deteriorating health satisfaction, in line with the increasingly deteriorating health conditions in their countries of origin after the breakdown of the Iron Curtain in 1989. Alternatively, the observed decline could be due to factors associated with immigrant status but not reflected by standard socioeconomic indicators. Examples are psychological stress, discrimination, or a lack of informal support networks.
In contrast, hypothesis H1, postulating an initially lower health satisfaction among immigrants, was not supported by the findings. At the beginning of the observation period, younger immigrants even had a higher health satisfaction than native Germans; older immigrants had values similar to those of Germans. This phenomenon is known as "healthy migrant effect". One explanation is a selection effect at the time of migration. Recent explanatory models postulate a second effect, the so-called "late-entry-bias". This implies that immigrants with a poor health status may remain excluded from studies that start enrolling participants only years after the time of immigration.
Selection effects may have played a role especially in young people who migrated for economical reasons. This may cause a selection towards people who are part of the working force, who is normally healthier than the general population. Among older people of ethnic German origin, however, even those with poorer health status may have decided to migrate. Common motives are an affinity to Germany, seen as the original home country, and negative feelings towards Eastern European countries, experienced as oppressors during the Communist regime. This could explain the finding that initially there was no better health satisfaction in the oldest age groups.
Late-entry-bias could have played a role, too: Most of the participants had immigrated some years before recruitment into the GSOEP. Finally, the perceived health satisfaction in young people may still be high in spite of a high prevalence of risk factors because for many chronic diseases there is a long time lag between exposure and onset[12, 13]. In all probability, the initially higher health satisfaction among younger immigrants is caused by an interplay of the factors discussed.
The findings are in line with results from a previous study, which also used the GSOEP and looked into morbidity measures of Germans and migrants from Southern Europe with longitudinal comparisons between 1984 and 1991. The study found for two out of three measures a lower morbidity among migrants in 1984, but the increase in morbidity was faster among migrants than Germans. Initial morbidity among migrants was higher in women than in men, but the magnitude of the increase in morbidity was higher in male migrants. This sex-specific pattern was similar to the one found in the present study. One possible explanation are sex-specific differences in the perception of health.
The results of the present study are also supported by findings from Israel, a country which has experienced an immigration of 775,000 Jews from the former Soviet Union since 1989. Two cross-sectional surveys performed in 1998 found lower self-reported health among immigrants compared to veteran Israelis[28, 29]. In both studies, lower self-reported health was associated with a more frequent reporting of chronic disease. Adjustment for socio-economic variables did not change these results. One of the studies showed that more recent immigrants had a significantly higher risk of reporting sub-optimal health and having at least one disease. Earlier immigrants, in contrast, did not show differences in the prevalence of disease compared to veterans. This might be explained by a strong selection which caused the migration of more healthy people in the beginning of this big migration wave to Israel.
This study has some limitations. Although the GSOEP is representative of the population in Germany [13–16], this may not have been the case for the study population. The GSOEP provides weighting factors which allow adjustment for disproportionate sampling when using the complete GSOEP dataset. However, due to the inclusion criteria of the study, weighting was not possible.
Another drawback is that persons living in institutional settings, such as housing facilities for immigrants, are underrepresented in the GSOEP. Immigrants of ethnic German origin stay in such facilities for an average period of 1.5 years after arrival in Germany. These facilities are regarded as "foci of social problems" , so an under-representation of their inhabitants in the GSOEP might lead to an overestimation of health and socioeconomic status of the immigrant population in our study.
Excluding the GSOEP participants who did not answer the question on health satisfaction both in 1995 and 2000, or who dropped out of the panel, may also have introduced a bias. Persons with poor health status are probably underrepresented since they have a higher probability to be lost to follow-up. Drop-out rates in the immigrant group are higher, especially in older age groups. The selection bias that might possibly result, however, would not change our conclusions since it would lead to an underestimation of the decline in health satisfaction in the immigrant group.
Finally, health satisfaction is a subjective indicator, so its appropriateness as a proxy for the actual health status needs to be discussed. A review of twenty-seven studies in different cultural settings has shown measures of self-rated health, based on questions with different wording and using different scales, to be valid predictors of mortality. These findings were confirmed in a German national health survey and for the GSOEP. Self-rated health measures have been shown to have a high reliability  and to provide valid results in different ethnic groups. Nevertheless, it cannot be completely excluded that answers to the question on health satisfaction are influenced by the differing cultural backgrounds of immigrants and native Germans. The adaptation to a new environment which immigrants have to undergo may influence health satisfaction even though it doesn't affect health in an objectively measurable way. For example, differences between the image the immigrants had of Germany before arriving and the reality they experience afterwards could cause a general disappointment which could be expressed through deteriorating health satisfaction. Even poor knowledge of the German language could bias the results in the immigrant group towards higher or lower values. Nevertheless, there are convincing arguments for using self-rated health as an outcome. It can provide a more holistic view of health which may not be reflected by "objective" health measures such as quantifiable medical diagnoses.
It is not entirely clear which magnitude a decline in health satisfaction on a scale from 0 to 10 has to have in order to reflect a relevant deterioration in perceived health. The cut-off point for the regression model was chosen using the highest quartile of all values of deterioration reported in the study population. It is also not clear if a decline of a given magnitude has the same meaning independently of the absolute values of health satisfaction. Finally, a period of five years between the two measurements might also be too short to draw conclusions from changes in reported health satisfaction. A Swedish study, however, demonstrated a significantly higher mortality in men who had reported a deterioration of two or more points on a 7-point scale of self-perceived health over a period of seven years. This suggests that the degree of deterioration measured in the present study and the time interval are sufficiently large to mirror relevant changes in self-perceived health.