Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Effect of gestational age misclassification on the pattern of low birthweight in Aborigines

Effect of gestational age misclassification on the pattern of low birthweight in Aborigines Abstract: Are most births of Aboriginal babies with low birthweight preterm or full term? There is no consensus because of the difficulty in obtaining valid measurements of gestational age. In Queensland, between 1988 and 1992, there were 519 births of Aboriginal babies with low birthweight in excess of the number expected if Aborigines had the same risk of low birthweight as whites. Most of these were preterm (males 76 per cent, females 65 per cent). Sensitivity analyses were used to investigate whether this result was robust to gestational age misclassification. Implausibly large misclassification proportions were required to make preterm low birthweight an insignificant contributor to the low birthweight excess in Aborigines. Therefore, efforts to reduce the number of preterm births should be given high priority. Unfortunately, significant reductions in the number of preterm births will not be achieved by reducing the prevalence of traditional risk factors for full-term low birthweight (for example, maternal smoking, teenage pregnancy). More work is needed to identify potentially modifiable risk factors for preterm birth. (Aust NZ,jPublic Health 1997; 21: 84-8) LTHOUGH low birth weight (<2500 g) is a commonly used endpoint in perinatal epiemiology, the population of babies with low birthweight is pathophysiologically heterogeneous.' One simple categorisation is to classify babies with low birthweight according to whether they are preterm or full-term. This categorisation is useful because there are differences in risk factors for low birthweight depending on its type. For example, Kramer has argued that betweenpopulation differences in the proportion of fullterm low-birthweight babies can be explained largely by differences in the prevalence of traditional risk factors such as maternal smoking, poor maternal nutrition and teenage pregnancy.2 These factors are also related to preterm birth, but the strengths of the associations are much weaker.2Most preterm births remain unexplained, and several authorities have recommended that more research should be undertaken into the causes of preterm birth as distinct from full-term low birth~eight.',~ There have also been suggestions that groups of babies with different types of low birthweight (such as preterm versus full-term) have different outcomes. For example, the preterm group is at higher risk of developing conditions such as hyaline membrane disease and sepsis than the full-term low-birthweight group.2 Both types have a higher-thannormal risk of cerebral palsy; however, preterm birth appears to carry the greater risk.2 These differences in risk factors and outcomes for the two broad types of low birthweight show that it is important to have some understanding of the pattern of low birthweight among Aborigines. Categorisation of low-birthweight babies as either Correspondence to Michael Coory, Centre for Clinical Epidemiology and Biostatistics, David Maddison Clinical Sciences Building, Royal Newcastle Hospital, Newcastle, NSW 2300. Fax (049) 236 148. A d preterm or full-term is complicated by the fact that measurement of gestational age is prone to error. This is believed to be particularly the case for aborigine^.^ Previous studies have produced conflicting results. For example, a study using data from the Western Australian Midwives Collection found that 62 per cent of Aboriginal babies with low birthweight were preterm.j In contrast, a study of Aboriginal births at Darwin Hospital found that the proportion of all Aboriginal live births which were preterm was 7 per cent, but the proportion of all Aboriginal live births with low birthweight was 13 per cent.6That study did not look specifically at the pattern of low birthweight, but the relatively small proportion of preterm babies indicates that the excess risk of low birthweight for Aborigines is mainly because of a higher risk of full-term low birthweight. One reason for the conflicting results might be that different methods were used to measure gestational age. The Western Australian study used the best clinical estimate (based on the date of the last menstrual period, antenatal ultrasound examination or maturity scoring at birth).5 The Darwin Hospital study relied solely on maturity scoring at birth.6 It is possible that the Western Australian study erroneously categorised full-term low-birthweight babies as preterm. On the other hand, it is possible that the Darwin Hospital study erroneously categorised preterm babies into the full-term group. The best way of estimating the effect of misclassification is by means of a validation study. However, validation of gestational age measurements is problematical. Either ultrasound scanning or secure dates would be suitable as the gold standard in a validation study. Unfortunately, such an approach would mean that the group of babies whose gestational ages are subject to the most error would be excluded.' Therefore, the generalisability of such a study would be open to question. For example, Blair VOL. AUSTRALIAN AND NEW ZEALAND JOURNAL O PUBLIC HEALTH 1997 F NO. LOW BIRTHWEIGHT IN ABORIGINES et al. found that only 40 per cent of pregnant Aboriginal women had had an ultrasound dating scan before 24 weeks' gestation, despite a policy of universal scanning at first presentation.8 Maturity scoring could be carried out on a representative sample of babies but is too inaccurate and imprecise to be used as the gold ~ t a n d a r dGiven these diffi.~ culties, one way of obtaining some insight into the effect of gestational age misclassification is by means of sensitivity analysis. This study had two aims. First, the pattern of low birthweight for Aborigines in Queensland was examined. Such an analysis has not previously been undertaken on Queensland data and it was of interest to know whether the pattern was similar to that found in Western Australia or at the Darwin Hospital. Second, the sensitivity of the reported pattern of low birthweight to various levels of gestational age misclassification was investigated. The results showed that most of the excess of low birthweight in Aborigines was due to preterm birth. Moreover, implausibly large levels of gestational age misclassification were required to change this pattern significantly. f Pattern o low birthweight The form of the data is summarised in Table 1. The proportions in each category for white babies were applied to the total number of Aboriginal live births. This gave the expected number of live births in each category if Aborigines had the same risk of low birthweight as whites. The excess over the expected number of Aboriginal babies of each type was calculated as excessPdb = aptlb expectedptlh = excessftlh afdb expected,, These were expressed as percentages of the excess over the expected number of Aboriginal babies with low birthweight. This gave the relative contribution of each category to the higher overall proportion for Aborigines. = + excessfdh) x 100 1 cptlb (excesspdh/(excesspdh = + cftlh (excess,,/ (excesspdh excessptlh)x 100 Methods The study used five years of data (1988 to 1992) from the Queensland Perinatal Data Collection. The collection includes all live births (and stillbirths) registered in Queensland and is similar to perinatal databases in the other Australian states and territories. The gestational age recorded in the Queensland collection is the best clinical estimate, whether based on the date of the mother's last menstrual period, ultrasound in early pregnancy, or maturity scoring of the neonate at birth. The method (or methods) used is not reported. The birthweight recorded is the first birthweight measurement, and is usually taken within an hour of birth. A baby's ethnic origin is based on the self-ascribed ethnic origin of the mother (the father's ethnic origin is ignored). Only the babies of mothers who were Aboriginal or white were included in the analysis. The babies of other mothers (Torres Strait Islanders, Polynesians, Asians) were excluded. All Aboriginal and white singleton livebirths of gestational age under 42 weeks were included in the analysis. Eighteen (0.3 per cent) of the Aboriginal and 380 (0.2 per cent) of white babies were excluded because information on birthweight or gestational age was missing. Complete data were available for more than 7400 Aborigmal babies and 179 000 white babies. Low-birthweight babies were defined as those weighing less than 2500 g. Because female babies are lighter on average than male babies, it could be argued that different cut-off points should be used. However, the 2500 g cut-off point is an accepted standard, and it was decided to use this for both males and females3 Separate analyses were carried out for males and females, since a fixed cut-off point for low birthweight (that is, 2500 g) will assign more full-term females to the low birthweight group. So cpdbis the excess over the expected number of Aborigines of low birthweight who were preterm expressed as a percentage of the total excess over the expected number of Aborigines of low birthweight. For this study, this was the numerical measure of main interest. For brevity, cptlhwill be called the peterm excess. Applying the formulas given above to the results from Western A~stralia,~value of 63 per cent was a obtained for the preterm excess (for males and females combined). Unfortunately, the Darwin Hospital paper did not present enough information for a calculation to be made for its dam6 However, because the proportion of preterm births among Aborigmal babies in the Darwin Hospital study was similar to that for whites (7 per cent), the preterm excess must be close to zero. Effect o misclasszjication o gestational age f f Misclassification matrices were used to relate the true (but unknown) numbers of babies of low birthweight who were preterm or full-term to the reported number^.^ Table 1 : Format of the data used in analysis of low birthweight of Aboriginal babies Category Preterm (c37 weeks) low birthweight Full-term (37 to 41 weeks) low birthweight Not low birthweight (c42 weeks) Total live births (c42 weeks) Aborigine White Total apdb ahlb wpdb tplb Wttlb Wnlb Wh, tftlb tnlb 'nib a , th, AUSTRALIAN AND NEW ZEALAND JOURNAL O PUBLIC HEALTH 1997 vot. 21 NO. 1 F COORY * t = r The symbol tl represents the true (but unknown) number of babies of low birthweight who were preterm, and tz represents the true number who were full-term. The symbols r, and K~ represent the corresponding reported numbers. For the misclassification matrix ( M ) , P,, is the proportion of babies of low birthweight who were truly preterm and were correctly reported as preterm. The other diagonal element (Pz2)is the proportion of babies of low birthweight who were truly full-term and were correctly reported as full-term. The off-diagonal elements (P.,, and PI?) give the probability of misclassifylng a truly preterm and a truly full-term low-birthweight baby. True gestational age Preterm Full-term Reported gestational age The reported proportions for the vector r were obtained from the Queensland Perinatal Data Collection. Various misclassification scenarios (various values for M ) were specified and the true numbers of preterm and full-term babies for each scenario obtained from t=M-'. r where M-' denotes the inverse of the misclassification matrix. Because Aboriginal babies are thought to be more likely to have their gestational age misclassified than whites, different misclassification matrices were used for Aborigines and whites. This meant that each misclassification scenario required the specification of four proportions: separate preterm misclassification proportions for whites and Aborigines and corresponding separate full-term misclassification proportions. The separate values for t obtained for Aborigines and whites were substituted for apt,,,, , aft,,, and wptlb,wftlb.For each misclassification scenario, the preterm excess was then calculated. Results were incorrectly were correctly Full-term The columns of M must add to 1. Hence, if one element in each column is specified, the whole matrix is specified. This is a more convenient way of describing M than writing out the whole matrix. It was decided to refer to M in terms of the off-diagonal elements as these give the proportion misclassified. For brevity, Rz1(the proportion of truly preterm babies who were incorrectly reported as full-term) will be called the peterm misclassajication poportion. Similarly, PI* (the proportion of truly full-term babies who were incorrectly reported as preterm) will be called the full-term misclassijication proportion. Table Pattern o low birthweight f The overall relative risk of low birthweight in singleton male Aboriginal babies compared with singleton male white babies was 2.4 (95 per cent confidence interval (CI) 2.2 to 2.7). For females, the corresponding relative risk was 2.6 (CI 2.4 to 2.9). Table 2 shows the relative risks for each category o f low birthweight. For males, there was very little difference between the percentage of Aborigines of low birthweight who were reported as preterm (74.1 per cent) and the corresponding percentage for whites (71.9 per cent) (' = 0.83, 1 df, P = 0.36). A similar pattern was x = 0.86, 1 df, P = 0.35), detected for females although the percentages of babies of low birthweight who were preterm were smaller (Aborigines: 63.4 per cent, whites: 61.2 per cent). This was expected, because full-term female babies are on average 150 g lighter than full-term male babies. In Queensland between 1988 and 1992, there were 519 low-birthweight Aborigines in excess of the (x2 2: Relative risk for preterm and full-term low birthweight in Aboriginal a n d white babies Relative risk 95% confidence interval Aborigines Whites Point estimate Males Preterm low-birthweight Full-term low-birthweighl Not low-birthweight Total live births Females Preterm low-birthweight Full-term low-birthweight Not low-birthweight Total live births NO. 2.1 to 2.7 1.7 to 2.6 2.4 to 3.0 2. I to 3.0 AUSTRALIAN AND NEW ZEAIAND JOURNAL O PUBLIC HEALTH 1997 vot. 21 F LOW BIRTHWEIGHT IN ABORIGINES Table 3: Estimates of the number of Aboriginal babies of low birthweight in excess of the expected number Estimated excess Males Females n % n % Preterm low-birthweight Full-term low-birthweight Total numbers expected if Aborigines had the same risk of low birthweight as whites (males: 223, females: 296). For males, the preterm excess was 76 per cent, and for females the preterm excess was 65 per cent (Table 3 ) . The first rows for males and females in Table 4 give the values for the preterm excess if it is assumed that there was no misclassification of gestational age in the data. These are just the observed values from Table 3. The other rows in Table 4 give the values for the preterm excess corresponding to different fullterm misclassification proportions for Aborigines. If the results of the Darwin Hospital study reflect the true situation in Queensland, then the preterm excess should be close to zero. Hence, the full-term misclassification proportion for Aborigines must be large. For example, for males, it would be necessary to postulate that 63 per cent of truly full-term, lowbirthweight Aborigines were incorrectly reported as preterm. Misclassification proportions larger than 50 per cent imply that classification of truly full-term low-birthweight babies as either full-term or preterm is worse than random assignment. This is unlikely. Misclassification of truly preterm Aboriginal babies was also considered (that is, the preterm misclassification proportion was varied). Such misclassification means that (in reality) there must be more preterm Aboriginal births than estimated in the analyses summarised in Table 4. Hence, it would be necessary to postulate even larger full-term misclassification proportions for Aborigines to make preterm low birthweight an insignificant contributor to the low-birthweight excess in Aborigines. Various misclassification scenarios for white babies were also considered. When the misclassification rates for whites were specified to be similar to those for Aborigines, the results were similar to those given in Table 4. The results changed significantly only if the misclassification pattern for whites was different from that for Aborignes. Table 5 summarises these simulations. The preterm misclassification proportion for white babies was set at 0.15 and the full-term misclassification proportion was Table 5:Preterm excess for misclassificationscenarios: Aborigines: preterm misclassificationproportion = 0; Whites: preterm misclassification proportion = 0.1 5, fuherm = 0.05 Full-term misclassification proportion far Aborigines Preterm excess Effect of gestational age rnisclassafication The data from the Queensland Perinatal Data Collection suggested that preterm low birthweight was the major contributor to the excess of low birthweight in Aborigines. If (in reality) this was not so, then many truly full-term Aboriginal babies must have been incorrectly reported as preterm (that is, the full-term misclassification proportion must be large). The effect of varying this proportion is shown in Table 4. To simplify these sensitivity analyses it was assumed that the preterm misclassification proportion for Aborigines was zero. For whites, it was assumed that there was no misclassification (that is, the misclassification matrix for white babies was the two-by-two identity matrix; preterm misclassification proportion = 0, full-term misclassification proportion = 0). Table 4: Preterm excess for misclassificationscenarios: Aborigines: preterm misclassification proportion = 0; Whites: preterm and full-term misclassification proportions =o Fuherm misclassification DroDortion for Aboriaines Preterm excess Males 0.30 0.40 0.50 0.60 0.63 Females Males Females AUSTRALIAN AND NEW ZEALAND JOURNAL O PUBLIC HEALTH 1997 VOL. 21 NO. 1 F COORY set at 0.05. This was considered an extreme scenario, because it implied that more than 90 per cent of white babies with low birthweight were truly preterm. Even for these scenarios, large full-term misclassification proportions for Aborigines were required to achieve a pattern similar to that reported from Darwin Hospital. Discussion If the problem of misclassification of gestational age is ignored, preterm birth is an important contributor to the high prevalence of low birthweight among Aborigines in Queensland. Not surprisingly, this result is similar to that from Western Australia.' Both studies used data from a whole-population database, and the methods of measuring gestational age were similar. The new information that this study provides is that implausibly large misclassification proportions were required to make preterm low birthweight an insignificant contributor to the low birthweight excess among Aborigines in Queensland. However, because sensitivity analysis (not a validation study) was used to assess the effects of gestational age misclassification, the possibility that the low birthweight excess in Aborigines is mainly due to full-term low birthweight cannot be ruled out. Improvements in our understanding of the poor neonatal health of Aborigines depend (to a large extent) on improving the validity of measurements on key variables. Gestational age is one of these variables. Meanwhile, sensible conclusions based on the currently available data must be made. The results of this study suggest that, at least for Aborigines in Queensland, a significant proportion of the lowbirthweight babies in excess of the expected numbers are preterm. Therefore, efforts to reduce the number of preterm births should be given a high priority. Unfortunately, significant reductions in the number of preterm births will not be achieved by reducing the prevalence of traditional risk factors for full-term low birthweight.2 Similarly, programs to detect and treat preterm labour do not represent a complete solution, because only 25 per cent of preterm births are theoretically preventable in this way.'O Blair et al. found that some of the shortening of gestation in Aboriginal women was due to maternal morbidity (for example, diabetes, urinary tract infections); however, most was unexplained.8 That study illustrates the fact that little is known about the aetiology of preterm birth. Some authorities have argued that the proportion of preterm babies with low birthweight in a population might be more amenable to change than the proportion of fullterm low-birthweight babies." If this is true, aetiological studies of preterm birth might hold the key to reducing differences in neonatal health between Aborigines and whites. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Australian and New Zealand Journal of Public Health Wiley

Effect of gestational age misclassification on the pattern of low birthweight in Aborigines

Loading next page...
 
/lp/wiley/effect-of-gestational-age-misclassification-on-the-pattern-of-low-3a1qXt8BrU

References (12)

Publisher
Wiley
Copyright
Copyright © 1997 Wiley Subscription Services, Inc., A Wiley Company
ISSN
1326-0200
eISSN
1753-6405
DOI
10.1111/j.1467-842X.1997.tb01659.x
Publisher site
See Article on Publisher Site

Abstract

Abstract: Are most births of Aboriginal babies with low birthweight preterm or full term? There is no consensus because of the difficulty in obtaining valid measurements of gestational age. In Queensland, between 1988 and 1992, there were 519 births of Aboriginal babies with low birthweight in excess of the number expected if Aborigines had the same risk of low birthweight as whites. Most of these were preterm (males 76 per cent, females 65 per cent). Sensitivity analyses were used to investigate whether this result was robust to gestational age misclassification. Implausibly large misclassification proportions were required to make preterm low birthweight an insignificant contributor to the low birthweight excess in Aborigines. Therefore, efforts to reduce the number of preterm births should be given high priority. Unfortunately, significant reductions in the number of preterm births will not be achieved by reducing the prevalence of traditional risk factors for full-term low birthweight (for example, maternal smoking, teenage pregnancy). More work is needed to identify potentially modifiable risk factors for preterm birth. (Aust NZ,jPublic Health 1997; 21: 84-8) LTHOUGH low birth weight (<2500 g) is a commonly used endpoint in perinatal epiemiology, the population of babies with low birthweight is pathophysiologically heterogeneous.' One simple categorisation is to classify babies with low birthweight according to whether they are preterm or full-term. This categorisation is useful because there are differences in risk factors for low birthweight depending on its type. For example, Kramer has argued that betweenpopulation differences in the proportion of fullterm low-birthweight babies can be explained largely by differences in the prevalence of traditional risk factors such as maternal smoking, poor maternal nutrition and teenage pregnancy.2 These factors are also related to preterm birth, but the strengths of the associations are much weaker.2Most preterm births remain unexplained, and several authorities have recommended that more research should be undertaken into the causes of preterm birth as distinct from full-term low birth~eight.',~ There have also been suggestions that groups of babies with different types of low birthweight (such as preterm versus full-term) have different outcomes. For example, the preterm group is at higher risk of developing conditions such as hyaline membrane disease and sepsis than the full-term low-birthweight group.2 Both types have a higher-thannormal risk of cerebral palsy; however, preterm birth appears to carry the greater risk.2 These differences in risk factors and outcomes for the two broad types of low birthweight show that it is important to have some understanding of the pattern of low birthweight among Aborigines. Categorisation of low-birthweight babies as either Correspondence to Michael Coory, Centre for Clinical Epidemiology and Biostatistics, David Maddison Clinical Sciences Building, Royal Newcastle Hospital, Newcastle, NSW 2300. Fax (049) 236 148. A d preterm or full-term is complicated by the fact that measurement of gestational age is prone to error. This is believed to be particularly the case for aborigine^.^ Previous studies have produced conflicting results. For example, a study using data from the Western Australian Midwives Collection found that 62 per cent of Aboriginal babies with low birthweight were preterm.j In contrast, a study of Aboriginal births at Darwin Hospital found that the proportion of all Aboriginal live births which were preterm was 7 per cent, but the proportion of all Aboriginal live births with low birthweight was 13 per cent.6That study did not look specifically at the pattern of low birthweight, but the relatively small proportion of preterm babies indicates that the excess risk of low birthweight for Aborigines is mainly because of a higher risk of full-term low birthweight. One reason for the conflicting results might be that different methods were used to measure gestational age. The Western Australian study used the best clinical estimate (based on the date of the last menstrual period, antenatal ultrasound examination or maturity scoring at birth).5 The Darwin Hospital study relied solely on maturity scoring at birth.6 It is possible that the Western Australian study erroneously categorised full-term low-birthweight babies as preterm. On the other hand, it is possible that the Darwin Hospital study erroneously categorised preterm babies into the full-term group. The best way of estimating the effect of misclassification is by means of a validation study. However, validation of gestational age measurements is problematical. Either ultrasound scanning or secure dates would be suitable as the gold standard in a validation study. Unfortunately, such an approach would mean that the group of babies whose gestational ages are subject to the most error would be excluded.' Therefore, the generalisability of such a study would be open to question. For example, Blair VOL. AUSTRALIAN AND NEW ZEALAND JOURNAL O PUBLIC HEALTH 1997 F NO. LOW BIRTHWEIGHT IN ABORIGINES et al. found that only 40 per cent of pregnant Aboriginal women had had an ultrasound dating scan before 24 weeks' gestation, despite a policy of universal scanning at first presentation.8 Maturity scoring could be carried out on a representative sample of babies but is too inaccurate and imprecise to be used as the gold ~ t a n d a r dGiven these diffi.~ culties, one way of obtaining some insight into the effect of gestational age misclassification is by means of sensitivity analysis. This study had two aims. First, the pattern of low birthweight for Aborigines in Queensland was examined. Such an analysis has not previously been undertaken on Queensland data and it was of interest to know whether the pattern was similar to that found in Western Australia or at the Darwin Hospital. Second, the sensitivity of the reported pattern of low birthweight to various levels of gestational age misclassification was investigated. The results showed that most of the excess of low birthweight in Aborigines was due to preterm birth. Moreover, implausibly large levels of gestational age misclassification were required to change this pattern significantly. f Pattern o low birthweight The form of the data is summarised in Table 1. The proportions in each category for white babies were applied to the total number of Aboriginal live births. This gave the expected number of live births in each category if Aborigines had the same risk of low birthweight as whites. The excess over the expected number of Aboriginal babies of each type was calculated as excessPdb = aptlb expectedptlh = excessftlh afdb expected,, These were expressed as percentages of the excess over the expected number of Aboriginal babies with low birthweight. This gave the relative contribution of each category to the higher overall proportion for Aborigines. = + excessfdh) x 100 1 cptlb (excesspdh/(excesspdh = + cftlh (excess,,/ (excesspdh excessptlh)x 100 Methods The study used five years of data (1988 to 1992) from the Queensland Perinatal Data Collection. The collection includes all live births (and stillbirths) registered in Queensland and is similar to perinatal databases in the other Australian states and territories. The gestational age recorded in the Queensland collection is the best clinical estimate, whether based on the date of the mother's last menstrual period, ultrasound in early pregnancy, or maturity scoring of the neonate at birth. The method (or methods) used is not reported. The birthweight recorded is the first birthweight measurement, and is usually taken within an hour of birth. A baby's ethnic origin is based on the self-ascribed ethnic origin of the mother (the father's ethnic origin is ignored). Only the babies of mothers who were Aboriginal or white were included in the analysis. The babies of other mothers (Torres Strait Islanders, Polynesians, Asians) were excluded. All Aboriginal and white singleton livebirths of gestational age under 42 weeks were included in the analysis. Eighteen (0.3 per cent) of the Aboriginal and 380 (0.2 per cent) of white babies were excluded because information on birthweight or gestational age was missing. Complete data were available for more than 7400 Aborigmal babies and 179 000 white babies. Low-birthweight babies were defined as those weighing less than 2500 g. Because female babies are lighter on average than male babies, it could be argued that different cut-off points should be used. However, the 2500 g cut-off point is an accepted standard, and it was decided to use this for both males and females3 Separate analyses were carried out for males and females, since a fixed cut-off point for low birthweight (that is, 2500 g) will assign more full-term females to the low birthweight group. So cpdbis the excess over the expected number of Aborigines of low birthweight who were preterm expressed as a percentage of the total excess over the expected number of Aborigines of low birthweight. For this study, this was the numerical measure of main interest. For brevity, cptlhwill be called the peterm excess. Applying the formulas given above to the results from Western A~stralia,~value of 63 per cent was a obtained for the preterm excess (for males and females combined). Unfortunately, the Darwin Hospital paper did not present enough information for a calculation to be made for its dam6 However, because the proportion of preterm births among Aborigmal babies in the Darwin Hospital study was similar to that for whites (7 per cent), the preterm excess must be close to zero. Effect o misclasszjication o gestational age f f Misclassification matrices were used to relate the true (but unknown) numbers of babies of low birthweight who were preterm or full-term to the reported number^.^ Table 1 : Format of the data used in analysis of low birthweight of Aboriginal babies Category Preterm (c37 weeks) low birthweight Full-term (37 to 41 weeks) low birthweight Not low birthweight (c42 weeks) Total live births (c42 weeks) Aborigine White Total apdb ahlb wpdb tplb Wttlb Wnlb Wh, tftlb tnlb 'nib a , th, AUSTRALIAN AND NEW ZEALAND JOURNAL O PUBLIC HEALTH 1997 vot. 21 NO. 1 F COORY * t = r The symbol tl represents the true (but unknown) number of babies of low birthweight who were preterm, and tz represents the true number who were full-term. The symbols r, and K~ represent the corresponding reported numbers. For the misclassification matrix ( M ) , P,, is the proportion of babies of low birthweight who were truly preterm and were correctly reported as preterm. The other diagonal element (Pz2)is the proportion of babies of low birthweight who were truly full-term and were correctly reported as full-term. The off-diagonal elements (P.,, and PI?) give the probability of misclassifylng a truly preterm and a truly full-term low-birthweight baby. True gestational age Preterm Full-term Reported gestational age The reported proportions for the vector r were obtained from the Queensland Perinatal Data Collection. Various misclassification scenarios (various values for M ) were specified and the true numbers of preterm and full-term babies for each scenario obtained from t=M-'. r where M-' denotes the inverse of the misclassification matrix. Because Aboriginal babies are thought to be more likely to have their gestational age misclassified than whites, different misclassification matrices were used for Aborigines and whites. This meant that each misclassification scenario required the specification of four proportions: separate preterm misclassification proportions for whites and Aborigines and corresponding separate full-term misclassification proportions. The separate values for t obtained for Aborigines and whites were substituted for apt,,,, , aft,,, and wptlb,wftlb.For each misclassification scenario, the preterm excess was then calculated. Results were incorrectly were correctly Full-term The columns of M must add to 1. Hence, if one element in each column is specified, the whole matrix is specified. This is a more convenient way of describing M than writing out the whole matrix. It was decided to refer to M in terms of the off-diagonal elements as these give the proportion misclassified. For brevity, Rz1(the proportion of truly preterm babies who were incorrectly reported as full-term) will be called the peterm misclassajication poportion. Similarly, PI* (the proportion of truly full-term babies who were incorrectly reported as preterm) will be called the full-term misclassijication proportion. Table Pattern o low birthweight f The overall relative risk of low birthweight in singleton male Aboriginal babies compared with singleton male white babies was 2.4 (95 per cent confidence interval (CI) 2.2 to 2.7). For females, the corresponding relative risk was 2.6 (CI 2.4 to 2.9). Table 2 shows the relative risks for each category o f low birthweight. For males, there was very little difference between the percentage of Aborigines of low birthweight who were reported as preterm (74.1 per cent) and the corresponding percentage for whites (71.9 per cent) (' = 0.83, 1 df, P = 0.36). A similar pattern was x = 0.86, 1 df, P = 0.35), detected for females although the percentages of babies of low birthweight who were preterm were smaller (Aborigines: 63.4 per cent, whites: 61.2 per cent). This was expected, because full-term female babies are on average 150 g lighter than full-term male babies. In Queensland between 1988 and 1992, there were 519 low-birthweight Aborigines in excess of the (x2 2: Relative risk for preterm and full-term low birthweight in Aboriginal a n d white babies Relative risk 95% confidence interval Aborigines Whites Point estimate Males Preterm low-birthweight Full-term low-birthweighl Not low-birthweight Total live births Females Preterm low-birthweight Full-term low-birthweight Not low-birthweight Total live births NO. 2.1 to 2.7 1.7 to 2.6 2.4 to 3.0 2. I to 3.0 AUSTRALIAN AND NEW ZEAIAND JOURNAL O PUBLIC HEALTH 1997 vot. 21 F LOW BIRTHWEIGHT IN ABORIGINES Table 3: Estimates of the number of Aboriginal babies of low birthweight in excess of the expected number Estimated excess Males Females n % n % Preterm low-birthweight Full-term low-birthweight Total numbers expected if Aborigines had the same risk of low birthweight as whites (males: 223, females: 296). For males, the preterm excess was 76 per cent, and for females the preterm excess was 65 per cent (Table 3 ) . The first rows for males and females in Table 4 give the values for the preterm excess if it is assumed that there was no misclassification of gestational age in the data. These are just the observed values from Table 3. The other rows in Table 4 give the values for the preterm excess corresponding to different fullterm misclassification proportions for Aborigines. If the results of the Darwin Hospital study reflect the true situation in Queensland, then the preterm excess should be close to zero. Hence, the full-term misclassification proportion for Aborigines must be large. For example, for males, it would be necessary to postulate that 63 per cent of truly full-term, lowbirthweight Aborigines were incorrectly reported as preterm. Misclassification proportions larger than 50 per cent imply that classification of truly full-term low-birthweight babies as either full-term or preterm is worse than random assignment. This is unlikely. Misclassification of truly preterm Aboriginal babies was also considered (that is, the preterm misclassification proportion was varied). Such misclassification means that (in reality) there must be more preterm Aboriginal births than estimated in the analyses summarised in Table 4. Hence, it would be necessary to postulate even larger full-term misclassification proportions for Aborigines to make preterm low birthweight an insignificant contributor to the low-birthweight excess in Aborigines. Various misclassification scenarios for white babies were also considered. When the misclassification rates for whites were specified to be similar to those for Aborigines, the results were similar to those given in Table 4. The results changed significantly only if the misclassification pattern for whites was different from that for Aborignes. Table 5 summarises these simulations. The preterm misclassification proportion for white babies was set at 0.15 and the full-term misclassification proportion was Table 5:Preterm excess for misclassificationscenarios: Aborigines: preterm misclassificationproportion = 0; Whites: preterm misclassification proportion = 0.1 5, fuherm = 0.05 Full-term misclassification proportion far Aborigines Preterm excess Effect of gestational age rnisclassafication The data from the Queensland Perinatal Data Collection suggested that preterm low birthweight was the major contributor to the excess of low birthweight in Aborigines. If (in reality) this was not so, then many truly full-term Aboriginal babies must have been incorrectly reported as preterm (that is, the full-term misclassification proportion must be large). The effect of varying this proportion is shown in Table 4. To simplify these sensitivity analyses it was assumed that the preterm misclassification proportion for Aborigines was zero. For whites, it was assumed that there was no misclassification (that is, the misclassification matrix for white babies was the two-by-two identity matrix; preterm misclassification proportion = 0, full-term misclassification proportion = 0). Table 4: Preterm excess for misclassificationscenarios: Aborigines: preterm misclassification proportion = 0; Whites: preterm and full-term misclassification proportions =o Fuherm misclassification DroDortion for Aboriaines Preterm excess Males 0.30 0.40 0.50 0.60 0.63 Females Males Females AUSTRALIAN AND NEW ZEALAND JOURNAL O PUBLIC HEALTH 1997 VOL. 21 NO. 1 F COORY set at 0.05. This was considered an extreme scenario, because it implied that more than 90 per cent of white babies with low birthweight were truly preterm. Even for these scenarios, large full-term misclassification proportions for Aborigines were required to achieve a pattern similar to that reported from Darwin Hospital. Discussion If the problem of misclassification of gestational age is ignored, preterm birth is an important contributor to the high prevalence of low birthweight among Aborigines in Queensland. Not surprisingly, this result is similar to that from Western Australia.' Both studies used data from a whole-population database, and the methods of measuring gestational age were similar. The new information that this study provides is that implausibly large misclassification proportions were required to make preterm low birthweight an insignificant contributor to the low birthweight excess among Aborigines in Queensland. However, because sensitivity analysis (not a validation study) was used to assess the effects of gestational age misclassification, the possibility that the low birthweight excess in Aborigines is mainly due to full-term low birthweight cannot be ruled out. Improvements in our understanding of the poor neonatal health of Aborigines depend (to a large extent) on improving the validity of measurements on key variables. Gestational age is one of these variables. Meanwhile, sensible conclusions based on the currently available data must be made. The results of this study suggest that, at least for Aborigines in Queensland, a significant proportion of the lowbirthweight babies in excess of the expected numbers are preterm. Therefore, efforts to reduce the number of preterm births should be given a high priority. Unfortunately, significant reductions in the number of preterm births will not be achieved by reducing the prevalence of traditional risk factors for full-term low birthweight.2 Similarly, programs to detect and treat preterm labour do not represent a complete solution, because only 25 per cent of preterm births are theoretically preventable in this way.'O Blair et al. found that some of the shortening of gestation in Aboriginal women was due to maternal morbidity (for example, diabetes, urinary tract infections); however, most was unexplained.8 That study illustrates the fact that little is known about the aetiology of preterm birth. Some authorities have argued that the proportion of preterm babies with low birthweight in a population might be more amenable to change than the proportion of fullterm low-birthweight babies." If this is true, aetiological studies of preterm birth might hold the key to reducing differences in neonatal health between Aborigines and whites.

Journal

Australian and New Zealand Journal of Public HealthWiley

Published: Feb 1, 1997

There are no references for this article.