The Matthew effect: infant mortality in Canada and internationally.

OBJECTIVE To examine whether the magnitude of improvement in the health status of a population over time is dependent on the previous health status of that population. DESIGN AND SETTING A study of infant mortality rates in Canada's 12 provinces and territories between the periods 1961-1965 and 1991-1995, and of infant mortality rates in 133 countries between 1960 and 1995. MAIN OUTCOME MEASURES Spearman's rank correlations, relative risks, and risk differences to measure the relationship between infant mortality in the 1960s and changes in infant mortality between the 1960s and 1990s. RESULTS In Canada, regional rankings based on infant mortality rates in 1961-1965 were strongly correlated (inversely) with rankings based on the percent change in infant mortality between 1961-1965 and 1991-1995 (correlation coefficient = -.85). In contrast, internationally, rankings based on infant mortality rates in 133 countries in 1960 were positively correlated with percent change between 1960 and 1995 (correlation coefficient =.56). Regional differences in infant mortality rates, measured using relative risks, declined in Canada (highest relative risk: 4.2, compared with Ontario in the 1960s; highest relative risk: 2.2, compared with Ontario in the 1990s) but increased globally (highest relative risk: 5.0, compared with industrialized countries in 1960; highest relative risk: 15.1, compared with industrialized countries in 1995). CONCLUSIONS Canadian regions with higher infant mortality rates in 1961-1965 achieved larger improvements compared with regions with initially lower infant mortality rates. The pattern observed within Canada is unlike the pattern observed internationally.