[The healthy migrant effect: role of selection and late entry bias]

First-generation immigrants frequently have a lower mortality than the host population, in spite of a low socio-economic status. This is usually explained by (self-) selection into migration. If this were the case, the immigrants' mortality risk would increase with time under observation. A persistently low mortality could be due to a late entry bias: if migrants are enrolled in a study years after immigration, sick or socio-economically unsuccessful individuals may already have returned to their countries of origin. Mortality risk would then be inversely associated with length of stay in the host country before enrollment.

We assessed the mortality risk of immigrants from Mediterranean countries to Germany in the German Socio-economic Panel, in relation to time under observation (1-15 years) and length of stay in Germany before enrollment (0-34 years), using the Cox regression.

In 1984-98, 2624 immigrants aged 16-83 years accrued 21,858 person years; 59 died. The hazard ratio, adjusted for age, sex and marital status, for each additional year under observation was 0.93 (95 % CI: 0.87-0.99); and for each additional 10 years in Germany before enrollment 0.49 (95 % CI: 0.27-0.89) in the age group >/= 50 years.

We found no evidence for a mortality increase with time under observation, suggesting that the healthy migrant effect is not primarily due to (self-)selection. The initial mortality advantage could be due to international differences in mortality patterns. A late entry bias does contribute to the persisting mortality advantage of older immigrants.