Infant mortality trends in Réunion (see appendix) are part of the broader movement in general mortality. In 1950, life expectancy at birth did not exceed 50 years for men and 53 years for women, 15 years less than in metropolitan France. Although mortality declined rapidly in metropolitan France during the second half of the twentieth century, the decline was faster in Réunion and the difference between the two territories has gradually narrowed, standing at only 3 years in 1990. Changes in infant mortality have been even more spectacular. Whereas fifty years ago the infant mortality rate was over 165 per 1,000 live births, it was already only 50‰ in the mid-1970s and is currently below 10‰. Progress in improving the survival of children was three times faster in Réunion than in metropolitan France. Several sources of statistical information make it possible to follow these changes in detail. After briefly describing these sources our article analyses the levels, trends and components of infant mortality and discusses the determinants of such a rapid change in the light of the available data.

Since the law of 1970 [1], certain preventive medical examinations have been compulsory. One such is the so-called 8th day examination, which results in the issuing of a health certificate that the physician sends to the local Maternal and Infant Protection service (PMI – Protection maternelle et infantile) [2]. These provisions apply to metropolitan France as well as to the overseas départements. In Réunion, these certificates have been computerized since 1991. Since that date the rate of coverage has approached 95%, which is higher than in metropolitan France. As in the case of civil registration, the 8th day certificates concern all births in Réunion and are the object of continuous registration. The certificate includes information about the conditions of delivery as well as the health status of the children at birth and during the first week of life (Rochat and Brodel, 1999; PMI, 2000).

The national perinatal survey (enquête périnatale) covers all French départements, including Réunion where it was conducted for the first time in 1995 and repeated in 1998. This national survey is a joint undertaking by the French National Institute for Health and Medical Research (Institut national de la santé et de la recherche médicale – INSERM), the General Department of Health (Direction générale de la santé – DGS) and the Department of Research, Studies, Evaluation and Statistics (Direction de la recherche, des études, de l’évaluation et des statistiques – DREES) of the French Ministry of Employment and Solidarity, plus the local Maternal and Infant Protection (PMI) services. Its main objective is to collect information on the health status of mothers and children, on medical practices during pregnancy and delivery, and on perinatal risk factors. It was designed to complete the information obtained from the 8th day certificates, which, since they are produced continuously for all births, are necessarily succinct. The perinatal survey of 1995 was based on a representative sample of 1,131 children born in Réunion during the month of December. The survey of 1998 used a smaller sample of 480 children born over a two-week period. The samples of these perinatal surveys are representative of all the children born alive or stillborn in public or private maternity clinics, as well as those born outside these facilities but transferred there after birth (INSERM et al., 1995; DRASS, 1999).

For this study, therefore, we used two exhaustive sources of data — civil registration material and 8th day certificates — for a detailed analysis of the levels, trends and components of infant mortality in Réunion, plus a source of period discrete information based on representative samples of the entire population concerned, namely the national perinatal surveys of 1995 and 1998 [3], with which to identify factors likely to influence changes in infant mortality. For all of these sources we used the reports (both published and internal) communicated to us by the agencies involved, without conducting new analyses on the individual data, to which we did not have access. Lastly, we also incorporated the results of the few published studies on health and mortality in Réunion.

The deaths of stillborn children are excluded from these calculations. The definition of late foetal mortality is complex and underwent an important change in France in 1993 (Blondel, 2000). Before this date confusion occurred between stillbirths defined by the World Health Organisation (WHO) as children showing no signs of life at delivery, and “false” stillbirths, children who die immediately after birth but have breathed. Since March 1993, to harmonize with international statistical practice, the civil registration of live births and of stillbirths is done according to signs of life at birth. Only those children who die without showing any sign of life are designated as stillborn, whereas those who have breathed are now included in early neonatal mortality statistics. Premature babies (under 28 weeks of pregnancy) who have died are counted differently. Those whose gestation duration was under 22 weeks of amenorrhoea (or whose birth weight was under 500 g.) are included among miscarriages (foetal mortality). For the others registration depends on the signs of life manifested at birth: those who have not breathed are counted with the stillborn whereas the children considered viable at birth by the physician are included among live births and counted among the cases of early neonatal mortality.

There is a special civil registration form for stillbirths. This form makes it possible to calculate a late foetal mortality rate which relates the number of stillbirths during a year to the total number of births (live births and stillbirths) in the same year. Late foetal mortality is grouped together with early neonatal mortality to form perinatal mortality, the rate of which is also calculated per 1,000 total births (Figure 1).

Definitional changes are important since they make international comparisons difficult and induce artificial fluctuations in the evolution of infant mortality and its components. Thus the change in the definition of stillbirths adopted in 1993, and the modification of medical and reporting practices that accompanied it, could partly explain the variations in late foetal and early neonatal mortality during the 1990s (Blondel, 2000). At that time, however, the level reached was already very low and therefore these problems do not call into question the considerable and very real change in infant mortality in Réunion over the past fifty years.

Table 1
Changes in infant mortality and its components and number of births, Réunion, 1951-1999

The two components of neonatal mortality have moved in parallel, with the mortality of the first six days remaining consistently two or three times higher than late neonatal mortality. As for late foetal mortality, its decline has followed that of early neonatal mortality. Initially very slow (with a rate still between 40‰ and 45‰ at the end of the 1960s) it accelerated during the 1970s, causing late neonatal mortality to be reduced by half in ten years, declining to only 20‰ around 1980, then to 10‰ in 1990 and 7‰ in 1999. It should be noted that the substantial fluctuations observed in the rates during the 1990s may be explained by the large random variations due to the calculations being based on a very small number of deaths, hardly 90 infant deaths per year.

What is immediately apparent when examining these graphs is the rapidity of the mortality reduction in Réunion. At the beginning of the 1950s the proportion of children who died before their first birthday was approximately twice as high as in Martinique, Guadeloupe and Mauritius, but by the 1970s Réunion had caught up with the other overseas départements. Thereafter the change followed a course parallel to that in the French West Indies, while Guyana, where infant mortality declined much more slowly, was left far behind. The 1970s also saw the closing of the gap with Mauritius. Because of a stabilization and possibly even a rise in the infant mortality rate in Mauritius between 1965 and 1975 and despite a resumption of the decline after this period, the difference between the two Indian Ocean islands has become almost constantly to the advantage of Réunion since the mid-1970s, and currently the rate is three times higher in Mauritius (19‰ compared with 6‰ in Réunion in 1999).

The comparison with metropolitan France, which constitutes the benchmark for all social and health services in Réunion, is equally favourable. Whereas in 1951 infant mortality rates in metropolitan France and Réunion were in a ratio of 1 to 3 (50‰ and 165‰ respectively), the risk of dying before the age of one had become almost identical in the two territories at the start of the 1990s. Since then the rates observed in Réunion have tended to fluctuate around a slightly higher level than in metropolitan France, especially for neonatal mortality, but the current levels are so low (5‰ in Réunion and 3‰ in metropolitan France) that not too much should be read into this disparity. The only significant difference remaining between the département of Réunion and metropolitan France is that observed for late foetal mortality, with rates of 7.1‰ and 4.6‰ respectively in 1999.

We can observe this evolution thanks to the medical cause-of-death statistics prepared by INSERM since 1981 in Réunion using an identical procedure to that implemented in metropolitan France since 1968. The coding of causes of death follows the rules of the International Classification of Diseases established by the World Health Organization. As pointed out earlier, the number of infant deaths registered each year in Réunion is very small, fewer than 100 on average over the last ten years, and for this reason we decided to present here a classification using only seven categories. This classification makes clear the main changes while avoiding the problem of subsequent changes to the detailed classification of the causes of death [5]. The seven categories are as follows: 1) infectious and parasitic diseases, 2) diseases of the respiratory system, 3) congenital anomalies, 4) conditions originating in the perinatal period, 5) symptoms, signs and ill-defined conditions, 6) accidents (external causes of injury and poisoning), 7) other diseases [6]. Table 2 presents the changes in cause-specific infant mortality rates between 1981 and 1997 for these seven categories. Because of the small number of infant deaths per year and the large random variations that are the consequence we have regrouped the years into four periods, each of which concerns between 450 and 500 deaths. Table 3 compares the rates in Réunion during the most recent period (1993-1997) with those of metropolitan France and Mauritius.

The cause-specific rates show that the epidemiological transition was largely achieved before the 1980s. Infectious and parasitic diseases and diseases of the respiratory system, which were responsible for 75% of infant deaths in the mid-twentieth century, currently represent fewer than 5% of infant deaths in Réunion. In the first period (1981-1983) the rates were already very low, 8 and 4.4 per 10,000 respectively (Table 2). Nevertheless the fight against these diseases continued to be effective since at the end of the period (1993-1997) the rates had fallen to 2.7 and 2.1 per 10,000, 4 or 5 times lower than in Mauritius. Although they remain high compared with the rates in metropolitan France, which are at approximately 1.1 per 10,000 for both types of causes, they have dropped to such a low level in Réunion that any further progress that could be made against these diseases is unlikely to influence greatly the general level of infant mortality.

Table 2
Medical causes of infant mortality, Réunion, 1981-1997. Mean annual mortality rates per 10,000 live births

Table 3
Medical causes of infant mortality, 1993-1997. Indices of mean annual mortality rates in metropolitan France and Mauritius (base 100 = Réunion)

This is not the case, however, of congenital anomalies and perinatal disorders which represented approximately 70% of deaths between 1993 and 1997. Infant mortality rates for these two groups of diseases are 17.1 and 34.5 per 10,000 live births respectively. Perinatal disorders represent the second group of causes for which there is a large difference between Réunion and Mauritius, the corresponding rate being four times higher in Mauritius during the most recent period. These disorders include principally prematurity, respiratory distress syndrome and other respiratory conditions of the foetus and the newborn. Because these diseases occur mainly in utero and immediately after birth their preponderance is associated with an increase in the contribution of perinatal mortality to total mortality before one year. Contrary to the war against infectious diseases, which can be waged effectively with simple and inexpensive means (vaccinations and hygiene measures), combating genetic diseases and perinatal disorders requires complex and costly medical technologies which have only gradually been introduced into Réunion. This explains the substantial remaining difference for such causes of death between this département and metropolitan France where, over the same period, infant mortality rates for congenital anomalies and perinatal disorders were roughly 12 and 18.5 per 10,000 births respectively.

Some qualification of these results is necessary, however, to allow for a possible substitution effect between perinatal disorders and the category of symptoms and ill-defined conditions. This category of deaths with imprecise, unknown or unspecified causes is much smaller in Réunion than in metropolitan France. For these two territories infant mortality in this category was 6.0 and 12.5 per 10,000 births respectively in 1993-1997. A similar difference exists between Réunion and Mauritius, with Mauritius enjoying the better rate of only 3.2 per 10,000. These disparities in a proportion of one to two can only be explained by differences in how the causes of infant deaths are reported. Perinatal disorders constitute a category of choice in which to group deaths whose causes are difficult to establish and it is conceivable that the tendency to do this is greater in Réunion than in metropolitan France, for a reason that remains to be explained. Unfortunately this hypothesis cannot be verified by an analysis of causes of death during the perinatal period because we do not know the cause-specific structure of mortality for that period, the statistics being available only for the period of infancy as a whole without any distinction between its different components. Another possibility is that some of the deaths during the early neonatal period have been erroneously recorded among stillbirths for causes of death that are difficult to identify, as a recent book (Michel, Catteau and Hatton, 1995) has suggested. Such a phenomenon might partly explain the substantial difference in late foetal mortality observed between the two territories.

These points notwithstanding, the difference between the rates of infant mortality from perinatal disorders in Réunion and metropolitan France is certainly in part real, given the frequency of premature births and the proportion of children with a low birth weight on the island. The national perinatal survey of 1998 indicates that premature births (under 37 weeks of pregnancy) and children with low birth weight (under 2,500 g.) both accounted for 12.5% of births in 1999 (PMI, 2000), double their proportions in metropolitan France. Prematurity and low birth weight are risk factors for perinatal and early neonatal mortality (Bréart, 1996; Papoz, Schwager and Favier, 2001).

The infant mortality rate for deaths from accidents has varied between 3 and 6.5 per 10,000 since the beginning of the 1980s, approximately the same as in Mauritius and twice as much as in metropolitan France. The types of accidents responsible for the deaths reported in this category vary widely from year to year, making it hard to identify the factors that could explain the difference between the two territories.

Lastly, a declining trend in mortality from other diseases has been observed since the mid-1980s, with a rate very similar to that observed in metropolitan France at the end of the period (1993-1997). Diseases of the nervous system are largely predominant in this category since they represent nearly 80% of deaths, with a rate of 4.5 per 10,000 births in 1993-1997. Although this rate is approximately double that of metropolitan France it represents more or less the same proportion of infant deaths, approximately 5%, in the two territories.

Table 4
Infant mortality rate and its components by sex, and excess male mortality ratio, Réunion, 1993-1997

Excess male infant mortality concerns principally perinatal disorders, and notably intrauterine hypoxia and asphyxia at birth, whereas in metropolitan France it is respiratory diseases that have the most disproportionate effect on boys during the first year of life compared with girls. On the other hand, girls are more prone to accidents than boys, though the difference between the sexes is tending to narrow. The ratio between the mortality rates from accidents of boys and girls has gone from 0.5 in 1988-1992 to 0.8 in 1993-1997.

The medical follow-up of pregnant women has improved so much over the last thirty years that it now compares favourably with that in metropolitan France. Thus, according to the 8th day certificates, 87% of women having given birth in 1999 benefited from at least seven prenatal medical visits, 11% had between four and six medical visits, and fewer than 2% had three or fewer (PMI, 2000). The national perinatal survey of 1998 also revealed that the average number of prenatal visits was higher in Réunion than in metropolitan France (DRASS, 1999). Since the beginning of the 1990s, fewer than fifty women on average each year have given birth without having had a single prenatal visit, representing 0.3% of births in 1999. At the same time the proportion of women having three ultrasound scans during their pregnancy has increased. Between the two prenatal surveys of 1995 and 1998, this proportion increased by 8%, from 53% to 61% (DRASS, 1999).

Generalized access to health care extends to childbirth and the first days of life. Whereas in 1951 more than 75% of births took place at home, this proportion is currently down to 0.5%, which means that medical care is immediately available to almost all newborn babies (INSEE, 1957; INSERM et al., 1995). The 8th day certificates show, for example, that in 1999 nearly 95% of children born in Réunion had benefited from a medical check-up at the end of the first week of life and that in 99% of cases the check-up had been carried out by a paediatrician (PMI, 2000), a higher proportion than in metropolitan France (94% in 1994). Lastly, medical follow-up continues during infancy as indicated by a vaccination coverage that is equal to, if not better than, that in metropolitan France (Catteau, 2001).

Table 5
Distribution of live births by age of mother*. Réunion (1951 and 2000) and metropolitan France (2000)

Table 6
Distribution of live births by birth order*. Réunion (1951 and 2000) and metropolitan France (2000)

Analysis of the distribution of births by the age of the mothers shows, however, that taken overall the change in birth timing has not been especially favourable to the reduction in infant mortality. The proportion of births to women aged between 25 and 34, age groups associated with the highest survival chances for infants, increased only slightly, from 49% to 53%, between 1951 and 2000 (Table 5). On the other hand, changes in the order-specific distribution of live births have contributed considerably to improved infant survival. Table 6 presents the distribution of births by birth order in 1951 and in 2000 in Réunion and compares the latter with that observed at the same date in metropolitan France. It shows a reversal of the distribution, with a sharp fall in the relative proportion of fourth or higher order births to the benefit of first and second order births. The proportion of fourth or higher order births has fallen from 47% in 1951 to 10% today. The evolution is especially remarkable for fifth or higher order births, the proportion of which was divided by nine over this period (36% in 1951 as against 4% in 2000). An INSEE study shows that, compared with first and second order children, excess mortality reaches 15% for third order children, 40% for fourth order children and almost 55% for fifth or higher order children in metropolitan France (Dinh, 1998). Assuming that these ratios are applicable to Réunion, the changes in structure by birth order observed since the 1950s would be enough to account for over 20% of the reduction in infant mortality. If distribution by birth order in Réunion matched that of metropolitan France in 2000 it would cause a reduction of only 4% in the rate of infant mortality. The changes in reproductive behaviour that could still occur in Réunion will therefore probably not have much more effect on the level of infant mortality.

Diabetes and arterial hypertension are the other two serious conditions held responsible for the health problems experienced by women in Réunion during pregnancy. At the beginning of the 1990s, the mortality rate from diabetes among women in Réunion was still four times higher than in metropolitan France (Michel, Catteau and Hatton, 1995). Currently, 15% of pregnant women are still affected by diabetes or arterial hypertension, two risk factors for the unborn child (Papoz et al., 2001). These conditions are second to signs of premature birth as causes of hospital admissions during pregnancy (DRASS, 1999). They might also partly explain the more frequent hospitalization of pregnant women in Réunion compared with metropolitan France. The perinatal survey of 1998 showed that the proportion of women admitted to hospital during pregnancy was almost half as high again in Réunion as in metropolitan France (30% compared with 21%), and among these women the proportion who had to be admitted for more than 6 days was 20% in Réunion, almost double that (11%) in metropolitan France (DRASS, 1999).

In response to the concerns expressed by physicians, the Regional Agency for Health and Social Welfare (Direction régionale de l’action sanitaire et sociale – DRASS) implemented a prenatal plan in Réunion over a five-year period (1996-2001). This plan comprised a series of measures intended to improve medical supervision of pregnancy and the conditions of childbirth, in particular though the reorganization of the hospital system and the creation of a network of perinatal professionals. The first two perinatal surveys (1995 and 1998) were conducted in Réunion in view of an evaluation of this programme and a new survey is planned to be conducted shortly. The most recent civil registration data (1999) already indicate a resumption in the decline of perinatal and late neonatal mortality, though given the small number of events and large random fluctuations from year to year it will be several years before this positive result can be confirmed. It seems, however, that the precarious economic situation of part of the population of Réunion is slowing down progress in the fight against infant mortality.

Concerning pregnant women more specifically, the perinatal survey of 1998 revealed that only 60% lived in a household with at least some income from paid employment (compared with 90% in metropolitan France), while almost half received state benefits (unemployment benefits, social insertion income, or single parent allowance) compared with fewer than 20% in metropolitan France. Yet the same survey shows that women who are economically vulnerable benefit from fewer prenatal consultations than the others (DRASS, 1999) and we know from analysis of the 8th day certificates that the probability of having a baby of low birth weight, and therefore at a higher risk of dying, increases significantly when there are fewer than four prenatal visits (Rochat and Brodel, 1999). This result is confirmed by the perinatal survey which indicated that the frequency of prenatal hospitalizations was higher (33%) among economically vulnerable women and, after the birth, among their children (16%), than among the others (24% and 10% respectively) (DRASS, 1999).