Infant Mortality by Gestational Age and Birth Weight in Canadian Provinces and Territories, 1990-1994 Births

Shi Wu Wen, Michael S Kramer, Shiliang Liu, Susie Dzakpasu and Reg Sauvé,
for the Fetal and Infant Health Study Group^a

Abstract

We compared gestational age-specific and birth weight-specific infant mortality in the Canadian provinces (excluding Ontario) and territories using the linked birth and death records for 1990-1994 births. Compared with Quebec, early neonatal mortality rates were higher in Saskatchewan, Alberta and Newfoundland among extremely small and preterm infants and among infants with no information on gestational age and birth weight on their records. Post-neonatal mortality rates were higher in Prince Edward Island, Manitoba, Saskatchewan, Alberta, British Columbia and the Northwest Territories among preterm (and low birth weight) and term (and normal birth weight) infants. We suggest that differences in registration practices probably explain the substantial interprovincial variations in early neonatal mortality rates among extremely small and preterm infants, whereas differences in demographic profile and the quality of obstetric, neonatal and infant care probably explain interprovincial variations in infant mortality rates among less extremely small and preterm infants.

Key words: birth weight; gestational age; infant mortality

Introduction

In almost all countries throughout the world, infant mortality has dropped markedly over the last century with improvements in sanitation, nutrition, infant feeding, and maternal and child health care,¹ although the decline has been slower in recent years.² Disparities in the risk of infant death remain, however, even in countries like Canada, where universal health care has been in place since 1970.³ For example, according to Statistics Canada figures, infant mortality for the combined years of 1991-1995 was 5.63 per 1,000 live births in Quebec but 13.22 per 1,000 live births in the Northwest Territories.^4,5

Why should such disparities remain in Canada despite the availability of universal health care for three decades? Analysis of infant mortality rates in Canadian provinces and territories by gestational age and birth weight groups may provide some hints. Improvements in, and access to, high-risk obstetric and neonatal care have led not only to markedly increased survival among extremely preterm newborns but to changes in registration practices of extremely small, immature newborns as well. Thus, births that were previously registered as spontaneous abortions (if they were registered at all) are now being registered as births, and births that were previously registered as stillbirths are now being registered as live births.^6-10

If there were substantial regional differences in registration practices, it should primarily affect the calculated infant mortality rates in groups with extremely low gestational age and birth weight, rather than other gestational age and birth weight groups. On the other hand, if substantial differences in infant mortality rates were observed in less extremely small or preterm infants, disparities in factors related to maternal health and maternal, neonatal and infant care should be considered. To examine such potential disparities, we analyzed gestational age-specific and birth weight-specific infant mortality rates among Canadian provinces and territories.

Methods

We used the linked birth and death records in Canada for births occurring from 1990 to 1994. Data processing and linkage procedures have been described elsewhere.¹¹ In brief, information on live births and deaths in Canada is collected by the registrars of the 10 provinces and 2 territories. Paper and electronic copies of the registration documents are submitted to Statistics Canada. Live birth data are stored in the Canadian Birth Data Base, and death data are stored in the Canadian Mortality Data Base.

To allow analysis of the relation between birth characteristics and death outcomes, Statistics Canada created a linked Canadian birth-death file for 1985-1994 birth records and 1985-1995 death records through the Canadian Perinatal Surveillance System. This linkage was based on a probabilistic procedure, using the Generalized Record Linkage System.^12,13 The Generalized Record Linkage System compares common fields in the two files, assigns weights to the resulting links and calculates a total weight. A weight of less than -90 would be automatically rejected and a weight of greater than +300 would be automatically accepted as a match. Manual resolution is carried out to confirm all linked records with a total weight ranging from -90 to +300 and all links to multiple births.

To reduce potential bias caused by secular trends in mortality and its determinants, we restricted our analysis to infants born between 1990 and 1994. Ontario data were excluded because of concerns about data quality.¹⁴ Newfoundland data were restricted to births between 1991 and 1994, because data from this province were not available to Statistics Canada before 1991. In the provinces and territories studied, nine live births occurring between 1990 and 1994 with a gestational age of less than 22 weeks and a birth weight of less than 500 g appeared to have survived infancy. Because the probability of survival at these birth weights and gestational ages is negligible, the survival designation in these cases could have been due to missing death certificates. The vital status of such births was reclassified as death on the first day of life; this reclassification is identified in all analyses.

We calculated province-specific rates for early neonatal (0-6 days after birth), late neonatal (7- 27 days after birth) and post-neonatal (28-365 days after birth) deaths, using all live births, survivors to the 7th day after birth and survivors to the 28th day after birth, respectively, as the denominators. Province-specific mortality rates were further calculated for the following gestational age and birth weight groups: <22 weeks, 22-23 weeks, 24-25 weeks, 26-27 weeks, 28-31 weeks, 32-33 weeks, 34-36 weeks, 37-41 weeks, 42 weeks and over, and not available (missing) for gestational age; and <500 g, 500-749 g, 750-999 g, 1,000-1,249 g, 1,250-1,499 g, 1,500-1,999 g, 2,000-2,499 g, 2,500-3,999 g, 4,000 g and over, and not available for birth weight. Relative risks and 95% confidence intervals were estimated for each province. Quebec was used as the reference for the estimates because Quebec had both the largest population and the lowest infant mortality rate¹⁴ among the provinces and territories studied.

To further explore the reasons for interprovincial variations in infant mortality, we also compared the distribution of important characteristics recorded on the birth certificate. We used mother's province of residence at delivery to calculate the province-specific rates and relative risks; nine births could not be assigned to a province or territory according to this variable and were excluded from these calculations.

Finally, to assess the potential impact of unlinked deaths and misclassification of birth weight and especially gestational age on the comparison of gestational age-specific and birth weight-specific mortality rates, we conducted supplementary analyses. In these, unlinked infant deaths were included in the calculation of overall early neonatal, late neonatal and post-neonatal mortality rates, and for those infants with a birth weight that deviated from the mean for a given gestational age by more than 5 standard deviations (based on the standard deviation for 1992-1994 live births), gestational age and birth weight were reclassified as "not available."

Results

A total of 1,230,938 births were registered by Statistics Canada for the nine provinces and two territories during the five-year study period. Among these, 1,223,895 were live births, 1,219,788 were live births surviving to the 7th day after birth and 1,218,928 were live births surviving to the 28th day after birth. Overall early neonatal, late neonatal and post-neonatal death rates were respectively 3.4, 0.7 and 2.3 per 1,000 live births. Of the 7,847 infant deaths, 199 could not be linked to birth records (i.e. the matching rate was 97.6%).

The majority of infants born before 22 weeks of gestation and a large proportion of those born before 24 weeks of gestation died soon after birth. Mortality rates were very high among infants born before 28 weeks and declined rapidly at higher gestational ages. Mortality rates reached their lowest level at term and increased slightly post-term. This "U"-shaped distribution in mortality rates was consistently observed in all provinces and territories studied, although some fluctuations did occur because of small samples in certain provinces and territories. Birth weight-specific mortality rates showed a similar pattern, with very high mortality in extremely low birth weight groups and a steep reduction at higher birth weights. However, unlike the situation for post-term infants, infants with large birth weight (>=4,000 g) were not at increased risk of infant mortality as compared with normal birth weight (2,500-3,999 g) infants (for detailed numbers and rates by province and territory see the perinatal health report from the Bureau of Reproductive and Child Health, Health Canada).

Compared with the province of Quebec, risks of early neonatal death, late neonatal death and post-neonatal death were substantially higher in most of the provinces and territories for most of the gestational age and birth weight groups (Tables 1-6), although we observed marked fluctuations due to small samples in several small provinces and territories.

Relatively large (relative risk >2.00) and statistically significant increases in risk were observed for early neonatal death at 26-27 weeks in the Northwest Territories; for late neonatal death at 24-25 weeks in Saskatchewan, and at 26-27 and 34-36 weeks in New Brunswick; and for post-neonatal death at 24-25 weeks in Saskatchewan, at 32-41 weeks in the Northwest Territories and at 37-41 weeks in Prince Edward Island and Saskatchewan (Tables 1-3).

Relatively large and statistically significant increases in risk were also observed for early neonatal death at 1,250-1,499 g in Saskatchewan, at 1,500-1,999 g in Prince Edward Island and at >=4,000 g in Newfoundland; for late neonatal death at <500 g in New Brunswick, Manitoba and Saskatchewan, at 500-749 g in Nova Scotia and Saskatchewan, at 750-999 g and 1,500-1,999 g in Newfoundland, and at 1,500-1,999 g in New Brunswick and Manitoba; and for post-neonatal death at <500 g in New Brunswick, at 750-999 g in Prince Edward Island, at all birth weights >=1,250 g in the Northwest Territories, at 1,500-1,999 g, 2,500-3,999 g and 4,000 g and over in Saskatchewan, and at 4,000 g and over in Newfoundland (Tables 4-6).

TABLE 1 Relative risk of early neonatal (0-6 days) death in Canadian provinces/territoriesa by gestational age group, 1990-1994 births
Gestational age (weeks)	Relative risk (and 95% confidence interval)
	NFLD	PEI	NS	NB	MAN
<22	NE (NE)	NE (NE)	NE (NE)	NE (NE)	NE (NE)
22-23	0.78 (0.59-1.03)	NE (NE)	0.97 (0.84-1.11)	0.84 (0.66-1.06)	1.11c (1.05-1.19)
24-25	0.93 (0.66-1.31)	0.46c (0.13-1.56)	1.03 (0.81-1.30)	0.89 (0.66-1.21)	0.89 (0.69-1.16)
26-27	1.38 (0.78-2.44)	NE (NE)	0.75 (0.45-1.27)	0.98 (0.56-1.73)	1.15 (0.80-1.67)
28-31	1.35 (0.81-2.26)	2.14 (0.99-4.63)	1.37 (0.93-2.03)	0.87 (0.50-1.51)	0.77 (0.50-1.18)
32-33	1.70 (0.92-3.14)	1.86c (0.47-7.29)	0.25c (0.08-0.75)	1.48 (0.87-2.53)	0.55 (0.29-1.05)
34-36	1.19 (0.63-2.25)	1.49c (0.55-3.99)	0.90 (0.57-1.44)	1.27 (0.80-2.02)	1.08 (0.92-1.53)
37-41	1.32 (0.89-1.97)	0.44c (0.14-1.36)	1.06 (0.77-1.45)	1.22 (0.88-1.69)	1.18 (0.92-1.53)
42-47	1.71 (0.39-7.44)	NE (NE)	1.34 (0.59-3.03)	0.97 (0.28-3.33)	0.78 (0.31-2.00)
Not available	43.24c (6.48-288.70)	NE (NE)	NE (NE)	NE (NE)	NE (NE)
TOTALd	1.42 (1.18-1.71)	0.83 (0.55-1.24)	1.10 (0.95-1.27)	1.08 (0.91-1.27)	1.15 (1.02-1.30)
TOTALe	1.38 (1.14-1.67)	0.79 (0.53-1.18)	1.06 (0.92-1.22)	1.03 (0.88-1.22)	1.13 (1.00-1.27)
Numbers in bold are statistically significant.; NE = Not estimable because of cell(s) with value of 0; a Excluding Ontario; b Quebec is the reference for the relative risks.; c Small sample with cell(s) <5; d Not including unlinked cases; e Including unlinked cases

TABLE 1 Relative risk of early neonatal (0-6 days) death in Canadian provinces/territoriesa by gestational age group, 1990-1994 births
Gestational age (weeks)	Relative risk (and 95% confidence interval)					Deaths per 1,000 live births
	SASK	ALTA	BC	YUK	NWT	QUEb
<22	0.91c (0.77-1.08)	1.01c (0.96-1.07)	1.02c (0.96-1.09)	NE (NE)	NE NE	944.8
22-23	0.91 (0.79-1.04)	0.98 (0.90-1.05)	1.05 (0.98-1.12)	NE (NE)	0.82c (0.51-1.31)	870.8
24-25	0.97 (0.77-1.23)	0.73 (0.61-0.88)	1.03 (0.88-1.21)	NE (NE)	0.34c (0.06-2.06)	486.1
26-27	1.48 (1.02-2.15)	0.91 (0.66-1.24)	0.99 (0.74-1.32)	0.55c (0.08-3.61)	2.28 (1.18-4.39)	164.5
28-31	1.53 (1.08-2.18)	1.12 (0.85-1.47)	1.09 (0.83-1.43)	NE (NE)	0.98c (0.37-2.59)	52.9
32-33	1.11 (0.65-1.87)	0.96 (0.67-1.39)	0.84 (0.58-1.23)	NE (NE)	0.99c (0.25-3.94)	25.6
34-36	1.31 (0.90-1.93)	1.00 (0.76-1.32)	0.88 (0.66-1.18)	0.98c (0.14-6.99)	0.88c (0.28-2.73)	6.9
37-41	1.02 (0.77-1.36)	0.97 (0.80-1.18)	0.95 (0.79-1.15)	1.12c (0.28-4.51)	1.59c (0.82-3.08)	0.8
42-47	1.23 (0.48-3.15)	1.39 (0.66-2.93)	0.70 (0.32-1.50)	3.47c (0.46-26.01)	NE (NE)	1.4
Not available	NE (NE)	NE (NE)	6.20 (2.39-16.04)	NE (NE)	NE (NE)	2.3
TOTALd	1.20 (1.06-1.36)	1.11 (1.02-1.22)	1.05 (0.96-1.15)	1.35 (0.75-2.44)	1.25 (0.88-1.78)	3.1
TOTALe	1.14 (1.01-1.30)	1.06 (0.97-1.16)	1.06 (0.98-1.16)	1.28 (0.71-2.32)	1.38 (0.99-1.92)	3.3
Numbers in bold are statistically significant.; NE = Not estimable because of cell(s) with value of 0; a Excluding Ontario; b Quebec is the reference for the relative risks.; c Small sample with cell(s) <5; d Not including unlinked cases; e Including unlinked cases

TABLE 2 Relative risk of late neonatal (7-27 days) death in Canadian provinces/territoriesa by gestational age group, 1990-1994 births
Gestational age (weeks)	Relative risk (and 95% confidence interval)
	NFLD	PEI	NS	NB	MAN
<22	NE (NE)	NE (NE)	NE (NE)	NE (NE)	NE (NE)
22-23	1.75c (0.41-7.45)	NE (NE)	2.00c (0.48-8.32)	3.00c (0.92-9.75)	3.50c (0.70-17.44)
24-25	1.61c (0.62-4.22)	NE (NE)	1.39 (0.61-3.16)	1.09c (0.41-2.93)	0.79c (0.29-2.16)
26-27	1.41c (0.35-5.71)	2.39c (0.36-15.99)	1.21 (0.47-3.09)	2.52 (1.07-5.90)	1.35 (0.60-3.06)
28-31	1.23c (0.38-3.98)	NE (NE)	1.34 (0.56-3.18)	1.48 (0.58-3.77)	0.73 (0.29-1.88)
32-33	1.75c (0.41-7.58)	NE (NE)	NE (NE)	0.56c (0.07-4.18)	0.61c (0.14-2.63)
34-36	1.78c (0.55-5.78)	3.71 (0.89-15.41)	0.90c (0.32-2.53)	3.80 (1.97-7.35)	0.17c (0.02-1.23)
37-41	0.65 (0.29-1.47)	1.86 (0.82-4.20)	0.52 (0.28-0.99)	1.08 (0.66-1.78)	1.37 (0.96-1.95)
42-47	NE (NE)	NE (NE)	0.68c (0.14-3.28)	0.74c (0.09-6.01)	NE (NE)
Not available	NE (NE)	NE (NE)	NE (NE)	NE (NE)	NE (NE)
TOTALd	1.26 (0.82-1.95)	1.50 (0.77-2.92)	0.99 (0.44-2.23)	1.62 (1.20-2.18)	1.16 (0.53-2.54)
TOTALe	1.24 (0.80-1.91)	1.47 (0.76-2.86)	0.89 (0.63-1.26)	1.58 (1.17-2.14)	1.06 (0.80-1.40)
Numbers in bold are statistically significant.; NE = Not estimable because of cell(s) with value of 0;a Excluding Ontario;b Quebec is the reference for the relative risks.;c Small sample with cell(s) <5;d Not including unlinked cases;e Including unlinked cases

Quebec had the lowest rate of teenage (<20 years) mothers, and the Northwest Territories had the highest rate (Table 7). Noticeable interprovincial variations were also observed in the proportions of mothers aged 35 and over and of primiparae. On the other hand, proportions of male sex and multi-fetal pregnancies were quite similar across the provinces and territories (Table 7).

Similar results were obtained after including unlinked infant deaths in the calculation of overall early neonatal, late neonatal and post-neonatal mortality (bottom row of Tables 1-6) and after reclassifying gestational age and birth weight as "not available" for those infants with a birth weight that deviated from the mean for a given gestational age (week) by more than 5 standard deviations (data not shown).

Discussion

An understanding of data quality and comparability is crucial for interpreting regional variations in infant mortality rates. Close scrutiny of Canada's vital statistics data has identified errors in birth and death registries in the province of Ontario.¹⁴ To ensure high data quality, we therefore excluded Ontario from our study.

The Generalized Record Linkage System used by Statistics Canada for the birth and death record linkage is a well-established system.^11-13 The matching rate was more than 97% for the provinces and territories studied, and the calculated overall mortality rates were similar when unlinked infant death cases were included. The quality of the linked birth and death records has been assessed by comparing the agreements for variables collected by two independent systems run in parallel in the provinces of Nova Scotia and Alberta.¹¹ This assessment demonstrated close agreement not only in the survival status of the linked records, but in gestational age and birth weight information as well.

Misclassification of birth weight and especially gestational age can occur in birth registry data,¹⁵ which may affect comparisons of gestational age-specific and birth weight-specific mortality rates. However, results obtained after reclassifying gestational age and birth weight as not available for those infants with a birth weight for gestational age that deviated from the mean by more than 5 standard deviations did not differ from the original results. Misclassification appears to have had no major impact on our overall results.

Because of the profound effect on infant mortality of low birth weight, in general, and of extremely preterm birth, in particular,^16-20 differences in registration of extremely small and immature infants (i.e. <500 g) as live births can severely compromise the comparison of infant mortality.^14,20 Regional variations may also exist in the registration of spontaneous or therapeutic abortions as births.^6-10 The substantial interprovincial variations in fetal and early neonatal mortality rates in infants of extremely preterm (<22 weeks of gestation) or low birth weight (<500 g) infants observed in our data (Tables 1-6) may have been caused, at least in part, by such differences in registration practices among Canadian provinces and territories. A similar inference can probably be made for births with no available gestational age or birth weight, since it is reasonable to assume that gestational age and birth weight were less likely to be recorded when extreme immaturity rendered viability unlikely.

Interprovincial variations in such factors as maternal health risks (e.g. genitourinary tract infection, cigarette smoking and substance abuse), quality of and access to both obstetric care (e.g. timely access to cesarean section) and neonatal intensive care, and quality of infant care (e.g. infant feeding, sleep positioning and injury prevention) may have played an important role in the observed interprovincial differences in mortality rates among less extremely small and preterm infants.

The provinces and territories with substantially higher rates in post-neonatal mortality, such as Manitoba, Saskatchewan, Alberta and the Northwest Territories, have proportionally larger native populations. It is widely recognized that post-neonatal mortality is much higher among native populations.^21,22 However, it seems difficult to attribute all of the increased post-neonatal mortality to this factor. We speculate that interprovincial differences in other aspects of infant health and health care may have also played a role. For example, the proportion of infants born to teenage mothers, a known risk factor for post-neonatal mortality,^23,24 is lowest in the province of Quebec (Table 7).