• Open Access

The association of infant feeding with parent-reported infections and hospitalisations in the West Australian Aboriginal Child Health Survey

Authors

  • Wendy H. Oddy,

    1. Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia, and School of Public Health, Curtin University of Technology, Western Australia
    Search for more papers by this author
  • Cheryl Kickett-Tucker,

    1. Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia
    Search for more papers by this author
  • John De Maio,

    1. Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia
    Search for more papers by this author
  • David Lawrence,

    1. Centre for Developmental Health, Curtin University of Technology, Western Australia
    Search for more papers by this author
  • Adele Cox,

    1. Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia
    Search for more papers by this author
  • Sven R. Silburn,

    1. Centre for Developmental Health, Curtin University of Technology, Western Australia and Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia
    Search for more papers by this author
  • Fiona J. Stanley,

    1. Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia
    Search for more papers by this author
  • Stephen R. Zubrick

    1. Centre for Developmental Health, Curtin University of Technology, Western Australia and Telethon Institute for Child Health Research, Centre for Child Health Research, University of Western Australia
    Search for more papers by this author

Correspondence to:
Dr Wendy Oddy, Telethon Institute for Child Health Research, PO Box 855, West Perth, Western Australia 6872. Fax: (08) 9489 7700; e-mail: wendyo@ichr.uwa.edu.au

Abstract

Objective: To examine infant feeding associations with parent-reported infections and hospitalisations in Western Australian Aboriginal infants and children.

Method: Families in Western Australia with children under 18 years of Aboriginal or Torres Strait Islander descent were included. A stratified multi-stage sample using an area-based sampling frame was compiled. Survey weights produced unbiased estimates for the population of families with Aboriginal children. Data were collected on demographic variables, maternal and infant characteristics and parent-reported recurring chest, ear and gastrointestinal infections. The data were linked to the Hospital Morbidity System to identify hospitalisations for infections for the same children.

Results: Twenty-seven per cent of Aboriginal children were breastfed for less than three months. Parent-reported recurring chest, ear and gastrointestinal infections were reported in 47% of the 0–3 age group. Hospitalisations due to upper respiratory and gastrointestinal infections were most common in the older children, but wheezing lower respiratory infections were most common in younger children. Breastfeeding for less than three months and birth weight less than 2,500 g were risk factors for parent-reported chest infections and hospitalisations for upper and wheezing lower respiratory infections (p<0.05).

Conclusion: Rates of parent-reported chest infections and hospitalisations due to these infections continue to be high in Aboriginal infants and children. Because breastfeeding for less than three months and low birth weight are risk factors for these infections, interventions to reduce the prevalence of low birth weight and to increase breastfeeding rates should be primary health goals in Aboriginal communities for the benefits of Aboriginal infants and children.

Since white settlement, Australian Aboriginal people have experienced social change and disadvantage that has had major implications for their health.1–3 The areas of disadvantage include hospitalisation rates and prevalence of infectious diseases.4 Respiratory infections are a major health problem in Aboriginal communities, with high rates of lower respiratory infections in remote areas.5 Aboriginal children have significantly higher admission rates to hospital, stay longer and are more likely to die in hospital than non-Aboriginal children. Infections (mainly respiratory and gastrointestinal) were the most common reason for hospitalisation, with higher rates in Aboriginal (1,114 per 1,000 live births) than non-Aboriginal children under the age of two years (242 per 1,000) (p<0.001).6 Although mortality rates in Aboriginal infants have fallen in the past 30 years, respiratory diseases are one of the most common causes of morbidity and mortality in Aboriginal people.7–10 However, the aetiology of respiratory illness in Aboriginal people has not been widely explored.11,12

In the recent past, malnutrition was prevalent in Aboriginal infants and young children in association with gastroenteritis, respiratory tract infections, ear infections and other infections, and infant and young child mortality rates were much higher than for other Australians.13 Much of this was due to poor housing, inadequate hygiene, environmental contamination and weaning practices that were likely to be complicated by the synergistic interactions of infection and malnutrition.14,15 Further, there is consistent evidence of a risk of not breastfeeding against diarrhoeal disease in the first 4–6 months of life16 and acute otitis media in the first three months.17,18

The objective of our study was to examine early infant feeding effects on recurrent parent-reported infections and any hospitalisations for these infections in Western Australian Aboriginal infants and children living in remote and non-remote Aboriginal communities. To meet this objective, we analysed questionnaire data regarding breastfeeding duration in association with health outcomes measured in the West Australian Aboriginal Child Health Survey, a population study of Aboriginal infants and children.

Methods

Population and sample

Families eligible to participate in the West Australian Aboriginal Child Health Survey were families in the State of Western Australia with at least one child under the age of 18 years who was identified as being of Aboriginal or Torres Strait Islander descent. The design used was a stratified multi-stage sample using an area-based frame. Area-based sampling was used because there was no list of Aboriginal children from which a sample could be selected. The sampling frame was compiled from the 1996 Australian Census and included all Census Collection Districts (CDs) in Western Australia where there were at least two children of Aboriginal or Torres Strait Islander descent enumerated in the 1996 Census. A special feature of the survey process was the need to search each sampled Census District by going door to door to identify eligible families. This followed the National Aboriginal and Torres Strait Islander Survey (NATSIS) methodology and allowed a comprehensive identification of eligible families who lived in a wide range of circumstances, including those Aboriginal families living in areas with otherwise ‘low prevalence’ Aboriginal representation.8

Stratification, sampling and non-response

The frame was stratified into four regions that were combinations of Australian Bureau of Statistics (ABS) Statistical Divisions (SDs). Within strata, CDs were selected with unequal probability without replacement, where selection probabilities were based on a cost model. Within selected CDs, dwellings were selected using systematic sampling without replacement. All in-scope children within selected in-scope families were included in the sample. The survey interviewers knocked on 139,000 doors to find the survey sample families.

Between May 2000 and June 2002, a total of 2,386 families were selected to participate in the survey, of which 1,999 (84%) participated. There were 5,513 in-scope children in these families and survey data were gathered on 5,289 of them. The survey questionnaire, administered to mothers and carers, was the tool employed to collect data. The face-to-face data collection method was used to increase accuracy in reporting from the study population.

Estimation

Survey weights were used to produce unbiased estimates for the population of families with Aboriginal children living in Western Australia. Using the unweighted data we included 4,287 carers who were also birthmothers (4,287/5,289=81%). Discussion of the population throughout the text of the manuscript refers to the weighted estimates derived by weighting the survey sample to reflect the population of Aboriginal and Torres Strait Islander children and their carers in scope for the survey (24,000/29,800=80.5%). Standard errors for survey estimates of totals were produced using the Ultimate Cluster Variance estimation technique.19 Standard errors for estimates of proportions, percentages and ratios were calculated using a modified form of the Jack knife variance estimation technique.20

Interviews

The interviews took considerable time, and multiple visits were often necessary to ensure complete data and to minimise respondent fatigue. Interviews were budgeted for a three-hour time period per household in which no more than 90 minutes would be used in formal data collection. Questionnaire content varied depending on the age of the child. Questions were asked about early infant nutrition, family size, child and youth development; health and well-being; functional impairment and disability; and use and access to health, education and social services. These data were collected from interviews with the carers in the household who were the most knowledgeable about the survey children. In addition to the information collected on children, separate interviews were undertaken with up to two carers per child to gather information about the demographic and social characteristics of the household and family and to ask questions about the dwelling, neighbourhood and community.

Ethics approval

The entire survey was conducted under the approval of institutional ethics committees for mainstream and Aboriginal research. Consent was obtained from carers and young people to collect separate health and well-being information from young people aged 12–17 years. Consent was also obtained from carers to link their survey records and those of their children with birth and hospital admission records. All aspects of the survey are accountable to the Western Australian Aboriginal Child Health Survey Steering Committee, ensuring appropriate Aboriginal governance and ownership of the process and data. Comprehensive methodological details are available elsewhere.21,22

Outcomes and exposures

Only children aged 0–17 years whose birth mother reported data were included for analysis of breastfeeding data (80.4%) as breastfeeding data reported by carers who were not the birth mother of the child was not considered sufficiently reliable for analysis. All analyses adjusted for level of relative isolation.

Outcomes

There were two outcomes for the purpose of this analysis. The first outcome was parent-reported infections that included recurring chest infections, recurring gastrointestinal infections and recurring ear infections. For these questions the mother was told: “I am going to read you a list of health problems that some children have. Please tell me if (study child) has ever had any of them: recurring chest infection; recurring gastro infection; recurring ear infection.” Because maternal/carer report of these problems was not precise, we included hospitalisations for infection as a second outcome. Hospitalisations were grouped based on the ICD-10 classification system.23 Hospitalisations for respiratory illness and infection were combined as upper respiratory tract infection (upper respiratory tract infections, tonsillitis), wheezing lower respiratory tract illness (wheezing associated respiratory illness, bronchiolitis, bronchitis, bronchospasm or asthma) and non-wheezing lower respiratory tract infection or illness (chest infection, pneumonia, whooping cough, chronic cough, croup). Gastrointestinal infection and ear infection (otitis media, otitis media with effusion) were grouped separately.

Hospitalisation data

The WA Record Linkage System is unique in Australia and one of only a handful of similar data collections in the world. It links together birth and death registrations with administrative hospital data from several sources to give a comprehensive record of health services contacts for the population of Western Australia.24 As there are no unique identifying numbers, probabilistic record linkage has been used to link the files together. This operates on matching names, dates of birth, hospital names and addresses. The procedure allows for possible changes in the matching fields by calculating the probabilities of records being correct matches. Records that are potential links are clerically reviewed, and overall the error rate is less than 1%.25

Key components of the record linkage system for use with the WAACHS were the birth records and the Hospital Morbidity Data. The birth registrations were combined with data from forms filled in by midwives to produce the Maternal and Child Health Research Data Base (MCHRDB).24 For every midwife-attended birth in Western Australia wherever it occurs, the midwife submits a completed form indicating the characteristics of the infant, its condition at birth, details of the pregnancy and delivery together with some demographic details. Birth weight in grams is among the data gathered on this form. The Hospital Morbidity Data System (HMDS) records every admission to private and public hospitals in WA since 1980. The system records demographic information about the patients, diagnostic information (coded using the 9th and 10th revisions of the International Classification of Diseases)23 and information about service use including length of stay, any procedures performed in hospital, and transfers between hospitals. While almost all mothers/carers gave consent for record linkage to occur, it was not always possible to match records.

Approximately 5% of survey children were born outside of WA and thus could not be linked to their birth records. Overall, 4,637 of the 5,289 survey children were successfully linked to their birth records (87.7 %). While several key components of the WA Record Linkage System date back to 1980 or earlier, the oldest survey children were born in 1982. Where population comparisons have been derived from the full set of linked data for the total population, the period 1982–2001 has been used as the relevant reference period.

The hospital morbidity system will only identify those children who had an infection requiring admission, thus the validation of the data using the hospital morbidity system is more likely to reflect more severe infections that require hospitalisation rather than validating the presence of a carer-reported infection.

Exposures

The main exposure was the duration of any breastfeeding, although we also considered exclusive breastfeeding. The questions asked about breastfeeding were as follows: “Was your child breastfed?”, “How long was your child breastfed?”, “Is your child still being given only breast milk?” and “For how long was your child given only breast milk?” Data on infant feeding were collected on both the infant (ages 0–3 years) and child questionnaires (ages 4–17 years) (as binary and categorical variables: How long was the child breastfed? 0-<3 months, 3-<6 months, 6-<9 months, 9-<12 months, 12 months or more and still being breastfed?). Exclusive breastfeeding was only collected for children aged 0–3 years (26% of cohort; n=6,210/24,000) and was measured as duration in months (for how long was the child given only breast milk?). Due to the low sample size for this age group and loss of statistical power in analysis with the study outcomes we used only the breastfeeding questions that were asked of the entire cohort (How long was your child breastfed?).

Because the children's age range was 0 to 17 years, most of the breastfeeding data were retrospectively collected. The breastfeeding questions were dichotomised into ‘any breastfeeding for 3 months/any breastfeeding for 3 months or longer’. Those still breastfeeding were included in the breastfed for three or more months category in analysis.

In addition to infant feeding method, confounding measures that occurred around the time of breastfeeding were considered in multivariate analysis and included infant gender and birth weight, smoking during pregnancy, maternal education, age group of child, level of socio-economic disadvantage,26 and level of relative isolation. Asthma (measured as asthma ever; yes or no) and family size (as measured by number of children in the family) were further tested in the models.

Non-response characteristics

Analysis of non-response characteristics showed the achieved sample to be broadly representative of the population from which it was drawn. Comparisons with population benchmarks showed significant associations with region, age and socio-economic status. To adjust for differential non-response, post-stratification weighting was employed. However, because of the small size of the population and the number of factors involved, the weights were calculated using the generalised ranking procedure of Deville and Särndal.27 A full description of sampling methods, non-response characteristics and weighting is available elsewhere.21,22

Imputation

For most of the data items there was only a small amount of item level non-response. Random hot-deck imputation, a procedure that does not add extra information about the non-respondents but serves to fill out the dataset to make analysis and interpretation of the results more straightforward, was used for imputing non-response at the item level.28 Imputation classes were formed based on age, sex and relative isolation. Then within each imputation class, for each non-respondent a donor was chosen at random and the donor's response was used to impute the value for the non-respondent. The percentage of imputed records was less than 1% for each variable and based on this, imputation had no effect on the final analysis.

Statistical tests

Standard chi-square tests and chi-square tests for trend adjusted for the complex sample design were used to assess the difference between categorical breastfeeding exposures and the primary outcomes of the study. Analysis was conducted using logistic regression techniques for the outcome variables. Multilevel analysis, incorporating adjustments for the survey weights and survey design, was used to fit the logistic regression models29 and to adjust for survey design issues that include family clustering of outcomes within these models. Confounding was tested in the final multivariate analysis for those children with recurring infections and those without infections, those with hospitalisations for infections and those without hospitalisations, for gender, birth weight, level of relative isolation, age group, infant feeding method and socio-economic factors (exposure to tobacco smoke during pregnancy, maternal education level and level of socio-economic disadvantage). Asthma and family size were also tested in the models. Following adjustment, the factors remaining in the models included gender (male, female), age group, level of relative isolation (none, low, moderate, high, extreme), birth-weight (<2,500 g, 2,500+ g), infant feeding (<3 months, 3+ months) and level of socio-economic disadvantage (bottom 5%, >5% -10%, >10%-25%, >25%-50%, top 50%). Significance was set at α=0.05. The statistical package used for all analyses was SAS version.9 (SAS Institute Inc. SAS 9.1. Cary, NC: SAS Institute Inc., 2000–04).

Results

Of the children in the survey, 52% were male and 11% had a birth-weight of less than 2,500 g. In relation to level of relative isolation, 10% lived in areas of extreme isolation, 24% lived in areas of low isolation and 34% lived in areas of no isolation (see Table 1). In Table 1 data are given as to how many children were never breastfed (12%), breastfed for less than three months but more than none (14.7%), breastfed for 3-<6 months (11.6%), 6-<9 months (8%), 9-<12 months (11.5%), more than 12 months (34.5%) and those still breastfeeding (7.8%). Mothers of infants from birth to three years were asked how long they exclusively breastfed their infant, with 12.6% never breastfeeding, 44.3% giving other milk by three months of age and 72.4% giving other milk before six months of age.

Table 1.  Demographic, maternal and infant characteristics of the estimated population of Aboriginal children aged 0-17 in the West Australian Aboriginal Child Health Survey.
Primary exposures (cumulative per cent)% (n/N)a
  1. Notes:

  2. (a) Using the weighted data 24,000/29,800 carers were also birth mothers. The numbers are rounded as part of the rounding rules in the estimation procedure. The table title gives the weighted population as Aboriginal Children aged 0-17 for who primary carer is the birth mother (n=24,000). While this is true for the breastfeeding questions and smoking during pregnancy, the proportions for gender, birth weight, level of relative isolation, maternal education and socio-economic disadvantage are based on all children age 0-17 years.

  3. (b) The figure of 11.4% for birth weight <2,500 g has been derived by excluding the ‘not stated’ category (12.6%). The fgure of 46.5% for smoking during pregnancy has been derived by excluding the ‘not applicable’ category . The fgure of 24.7% for maternal education <10 years of schooling has been derived by excluding the ‘not stated’ category (2.7%).

Demographic variables 
Level of relative isolation 
 None34.1 (10,200/29,800)
 Low24.4 (7,270/29,800)
 Moderate21.4 (6,390/29,800)
 High10.6 (3,170/29,800)
 Extreme9.5 (2,830/29,800)
Socio-economic disadvantage 
 Bottom 5%26.3 (7,840/29,800)
 >5%-10%12.9 (3,860/29,800)
 >10%-25%25.2 (7,510/29,800)
 >25%-50%25.9 (7,720/29,800)
 Top 50%9.7 (2,890/29,800)
Maternal characteristics 
Smoking in pregnancy 
 Yes46.5 (11,100/23,940)
Maternal education 
 <10 years of schooling24.7 (7,370/28,990)
Those with family size 
 Four or more children46.2 (13,800/29,800)
Infant characteristics 
Gender 
 Male51.6 (15,400/29,800)
 Female48.4 (14,400/29,800)
Birth weightb 
 <2,500 g11.4 (2,970/25,940)
 2,500+ g88.6 (23,000/25,940)
Children who report asthma 
 Yes23.2 (6910/29,800)
Any breastfeedinga 
 Never breastfed12.0 (2,870/24,000)
 0-<3 months14.7 (3,520/24,000)
 3-<6 months11.6 (2,780/24,000)
 6-<9 months8.0 (1,910/24,000)
 9-<12 months11.5 (2,760/24,000)
 12 months+34.5 (8,250/24,000)
 Still breastfeeding7.8 (1,860/24,000)

We considered the prevalence of the primary infection outcomes (see Table 2). Table 2 relates to percentages surveyed at those ages who reported infections during their lifetime. Recurring chest, ear and gastrointestinal infections were common in the 0–3 age group but were also common at older ages. For ear infections, prevalence was high in the 0–11 age groups and gastrointestinal infections were also common in the younger age groups. Upper respiratory and gastrointestinal infections were the most common reason for hospitalisation due to infection in the older age groups, but wheezing lower respiratory infection was the most common reason for hospitalisation in the younger age groups. The older age groups had longer to have ever been hospitalised for any of these infections and for parent report the decline in percentages with age may be due to recall bias.

Table 2.  Prevalence of the parent-report and hospitalisation primary infection outcomes in the population estimates of Aboriginal children aged 0-17 years.a
 Prevalence %(n/N)  
Primary outcomes0-3 years n=6,9104-11 years n=13,80012-17 years n=9,100All children n=29,800
  1. Notes:

  2. (a) The table title gives the weighted and estimated population of all Aboriginal children aged 0-17 (n=29,800). The numbers are rounded as part of the rounding rules in the estimation procedure. The proportions for parent-reported infections and for hospitalisations are based on all children age 0-17 years, and the proportions are cumulative based on ever reported infection or hospitalisation.

Parent report    
 Recurring chest infections19.4 (1,340/6,910)11.5 (1,590/13,800)8.0 (730/9,100)12.3 (3,660/29,800)
 Recurring ear infections20.4 (1,410/6,910)19.9 (2,750/13,800)13.6 (1,240/9,100)18.1 (5,400/29,800)
 Recurring gastrointestinal infections7.2 (500/6,910)6.1 (850/13,800)3.5 (320/9,100)5.6 (1,670/29,800)
Hospitalisations    
 Upper respiratory12.9 (890/6,910)22.9 (3,160/13,800)27.3 (2,480/9,100)21.9 (6,530/29,800)
 Non-wheezing lower (includes pneumonia)9.2 (640/6910)15.9 (2,200/13,800)16.2 (1,480/9,100)14.5 (4,310/29,800)
 Wheezing lower13.5 (930/6,910)14.3 (1,970/13,800)15.1 (1,380/9,100)14.4 (4,280/29,800)
 Ear infections11.9 (830/6,910)21.5 (2,960/13,800)17.5 (1,600/9,100)18.1 (5,390/29,800)
 Gastrointestinal infections12.7 (880/6,910)29.5 (4,070/13,800)35.1 (3,200/9,100)27.3 (8,150/29,800)

In our models from multivariate logistic regression, we show that children breastfed for less than three months were at significant risk for recurring chest infections throughout childhood, in comparison to infants breastfed for three months or more (OR 1.49; 95% CI 1.14-1.89; p=0.002) (see Table 3). For recurring ear infections, being in areas of extreme isolation approached significance but no other exposure was associated with this outcome other than older age group. Living in areas of less social disadvantage was significantly protective against recurring chest infections (OR 0.64; 95% CI 0.42-0.98; p=0.040), hospitalisations due to non-wheezing respiratory (OR 0.59; 95% CI 0.35-1.00; p=0.052) and hospitalisations due to gastrointestinal infections (OR 0.59; 95% CI 0.38-0.94; p=0.025). Older age was a risk exposure for most reported recurring infections as was level of relative isolation and hospitalisation due to these infections.

Table 3.  Multivariate relationship between demographic factors, infant characteristics and parent-reported recurring chest, ear and gastrointestinal infections in children aged 0-17 years in the West Australian Aboriginal Child Health Survey.a
Primary exposureRecurring chest infectionsRecurring ear infectionsRecurring gastrointestinal infections
  OR (95% CI) p value 
  1. Notes:

  2. (a) Age is adjusted in the analysis

  3. (b) For each exposure, the stated odds ratio contrasts the odds of the recurring infection type in the column heading compared with one; 95% confdence limits are in brackets with the p value beside. Age group of child has been adjusted in the analysis.

Demographic variables   
Level of relative isolation   
 None1.001.001.00
 Low0.88 (0.66-1.17) 0.3701.15 (0.86-1.54) 0.3540.78 (0.48-1.28) 0.326
 Moderate0.85 (0.61-1.18) 0.3241.31 (0.92-1.88) 0.1361.28 (0.73-2.24) 0.381
 High0.73 (0.45-1.16) 0.1771.17 (0.78-1.75) 0.4580.99 (0.46-2.11) 0.980
 Extreme0.79 (0.53-1.20) 0.2701.50 (0.99-2.27) 0.0581.85 (1.07-3.22) 0.029
Socio-economic disadvantage   
 Bottom 5%1.001.001.00
 >5%-10%0.84 (0.56-1.27) 0.4110.92 (0.62-1.38) 0.6880.76 (0.41-1.39) 0.368
 >10%-25%0.77 (0.55-1.07) 0.1250.78 (0.56-1.09) 0.1440.62 (0.37-1.02) 0.061
 >25%-50%0.82 (0.58-1.14) 0.2370.88 (0.63-1.23) 0.4670.82 (0.53-1.27) 0.373
 Top 50%0.64 (0.42-0.98) 0.0400.77 (0.50-1.19) 0.2420.42 (0.17-1.07) 0.069
Infant characteristics   
Gender   
 Male1.001.001.00
 Female0.90 (0.75-1.07) 0.2430.86 (0.71-1.04) 0.1250.95 (0.69-1.31) 0.753
Birth weight   
 2,500 g+1.001.001.00
 <2,500 g1.32 (1.01-1.72) 0.0421.30 (0.95-1.77) 0.1011.54 (1.01-2.34) 0.044
Duration of any breastfeedingb   
 0-<3 months1.49 (1.14-1.89) 0.0021.12 (0.85-1.47) 0.4191.45 (0.97-2.17) 0.069
 3+ months1.001.001.00

Hospitalisations for upper respiratory infections were more prevalent in infants who were breastfed for less than three months (OR 1.27; 95% CI 1.02-1.56; p=0.032) (see Table 4). Other factors associated with upper respiratory infection were older age, living in low relative isolation and being <2,500 g at birth. The pattern was similar for hospitalisations due to wheezing lower respiratory infections with male gender, a shorter duration of breastfeeding, low to moderate isolation and lower birth weight being risk exposures. For hospitalisations due to non-wheezing lower respiratory infection, breastfeeding for less than three months approached statistical significance for risk (OR 1.27; 95% CI 0.93-1.72; p=0.130). Female gender, level of relative isolation and less socio-economic disadvantage were protective for non-wheezing lower respiratory infection but older age and lower birth weight were risk exposures with the same relationships true for hospitalisation due to gastrointestinal infection. Extreme isolation was a risk for hospitalisations due to pneumonia, ear and gastrointestinal infections. Female gender and birth weight more than 2,500 g were factors protective against hospitalisation due to all infections.

Table 4.  Multivariate relationship between demographic factors, infant characteristics and hospitalisation for respiratory infections, ear infections and gastrointestinal infections in children aged 0-17 years in the West Australian Aboriginal Child Health Survey.
Primary exposureHospitalisations for upper respiratory infectionsHospitalisations for wheezing lower respiratory infectionsHospitalisations for non-wheezing lower respiratory infectionsHospitalisations for ear infectionsHospitalisations for gastrointestinal infections
   OR (95% CI) p value  
  1. Notes:

  2. (a) Age is adjusted in the analysis.

  3. (b) For each exposure, the stated odds ratio contrasts the odds of the infection type in the column heading compared with one; 95% confdence limits are in brackets with the p value beside.

Demographic variables     
Level of relative isolation     
 None1.001.001.001.001.00
 Low1.54 (1.23-1.94) <0.0011.28 (0.97-1.68) 0.0791.62 (1.19-2.20) 0.0021.75 (1.33-2.31) <0.0011.48 (1.14-1.92) 0.003
 Moderate1.11 (0.81-1.54) 0.5111.69 (1.23-2.33) 0.0012.31 (1.55-3.44) <0.0011.69 (1.18-2.41) 0.0042.31 (1.62-3.28) <0.001
 High1.15 (0.68-1.92) 0.6031.36 (0.90-2.07) 0.1463.89 (2.49-6.07) <0.0012.65 (1.68-4.18) <0.0012.96 (2.09-4.21) <0.001
 Extreme1.10 (0.69-1.74) 0.6821.16 (0.71-1.91) 0.5613.42 (2.36-4.97) <0.0012.84 (1.97-4.08) <0.0013.59 (2.35-5.48) <0.001
Socio-economic disadvantage     
 Bottom 5%1.001.001.001.001.00
 >5%-10%0.79 (0.56-1.13) 0.2021.04 (0.69-1.55) 0.8570.78 (0.54-1.12) 0.1740.79 (0.55-1.13) 0.2020.74 (0.52-1.05) 0.096
 >10%- 25%0.88 (o.66-1.17) 0.3890.99 (0.71-1.38) 0.9580.62 (0.44-0.87) 0.0050.78 (0.54-1.12) 0.1720.66 (0.47-0.92) 0.016
>25%-50%1.00 (0.74-1.34) 0.9861.13 (0.83-1.53) 0.4320.65 (0.46-0.91) 0.0130.88 (0.65-1.21) 0.4420.70 (0.52-0.96) 0.028
 Top 50%0.97 (0.65-1.45) 0.8860.99 (0.66-1.49) 0.9490.59 (0.35-1.00) 0.0520.75 (0.48-1.17) 0.2060.59 (0.38-0.94) 0.025
Infant characteristics     
Gender     
 Male1.001.001.001.001.00
 Female0.81 (0.69-0.95) 0.0110.70 (0.57-0.87) <.0010.78 (0.63-0.96) 0.0200.76 (0.64-0.91) 0.0030.84 (0.72-0.98) 0.028
Birth weight     
 2,500 g+1.001.001.001.001.00
 < 2,500 g1.65 (1.24-2.19) <0.0012.40 (1.79-3.21) <0.0011.93 (1.44-2.59) <0.0011.61 (1.23-2.10) <0.0012.17 (1.64-2.87) <0.001
Duration of any breastfeedingb     
 0-<3 months1.27 (1.02-1.56) 0.0321.47 (1.12-1.92) 0.0051.27 (0.93-1.72) 0.1301.25 (0.97-1.61) 0.0780.99 (0.78-1.28) 0.963
 3+ months1.001.001.001.001.00

Compared with areas of no isolation, extreme relative isolation was a risk factor for recurring gastrointestinal infections (OR 1.85; p=0.029). Breastfeeding for less than three months approached significance as a risk against recurring gastrointestinal infection (OR 1.45; 95% CI 0.97-2.17; p=0.069) as was low birth weight (<2,500 g) for both chest (OR 1.32; 95% CI 1.01-1.72; p=0.042) and gastrointestinal infections (OR 1.54; 95% CI 1.01-2.34; p=0.044). Living in areas of less social disadvantage was significantly protective against recurring chest infections (OR 0.64; 95% CI 0.42- 0.98; p=0.040). Older age was a risk exposure for most reported recurring infections.

We tested for current asthma, family size and smoking during pregnancy in each of the models but the addition of these variables did not change our substantive conclusion about breastfeeding. Asthma was significant in the recurring chest infection model with an odds ratio >5, and breastfeeding for less than three months continued to be a significant risk (OR 1.37; 95% CI 1.05-1.79; p=0.019). Birth weight was no longer significant when asthma was included in the model. Asthma was also significant in the ear infections model, and although the effect was not as strong as for chest infections, the odds ratio was 1.35 (95% CI 1.03-1.76; p=0.029). Family size and smoking in pregnancy did not reach significance in these models. We tested asthma, family size and smoking in the gastrointestinal infections model, but none of these factors reached significance. Finally, we tested for interactions between low birth weight and breastfeeding but none were observed.

We re-ran the same multivariable models, but limited the sample to 0–3 year-olds, the rationale being that carers of children aged between 0–3 years would be more likely to remember how long they breastfed their children than carers of older children. Results of the further analysis on the 0–3 year-old subsample showed that less breastfeeding was significantly associated with hospitalisation for wheezing lower respiratory infections (OR 1.54; 95% CI 1.00-2.38; p=0.052); and hospitalisation for non-wheezing lower respiratory infections (OR 1.72; 95% CI 1.06-2.78; p=0.028).

We tested the effect of exclusive breastfeeding for three months or less (39.8%) compared with exclusive breastfeeding for greater than three months (40.7%) and those still breastfeeding (9.3%) on our selected outcomes (10.1% of the population were not the primary birth mother and therefore were not included in this analysis). Exclusive breastfeeding for longer than three months was not associated with increased protection for any of the reported recurring infections or hospitalisations. However, ‘still breastfeeding’ at the time of the survey showed significant protection against recurring ear infection (OR 0.29; 95% CI 0.09-0.91; p=0.035), gastrointestinal infection (OR 0.12; 95% CI 0.02-0.55; p=0.007) and hospitalisations for upper respiratory (OR 0.31; 95% CI 0.11-0.91; p=0.034), wheezing lower (OR 0.31; 95% CI 0.12-0.80; p=0.016), non-wheezing lower (OR 0.05; 95% CI 0.01-0.36; p=0.003) and ear (OR 0.08; 95% CI 0.02-0.38; p=0.002) infections. A longer duration of exclusive breastfeeding was not protective against any of the outcomes tested. When we adjusted for age of the child in the first three years of life the results for ‘still breastfeeding’ were weakened. Each additional year of age of the child significantly increased the risk for all of the infection and hospitalisation outcomes tested in this study.

Age-adjusted models demonstrated that age was a significant predictor of infections, therefore any association found in a model that does not include age is thus potentially confounded. Age was accounted for in the models that considered any breastfeeding, by including age group. Because children under one year of age make up a significant percentage of children in the aged 0–3 model and have a high probability of still being breastfed at the time of the survey, it is important to account for this. The important point to note about the exclusive breastfeeding models for children aged 0–3 years is that they have been fit to a substantially smaller dataset, which has reduced the statistical power of these models evident in the wider confidence intervals. It is thus much less likely that small effects would be detected even if they did exist.

Discussion

The objective of this study was to determine whether Aboriginal infants and children who were breastfed for longer than three months would have fewer recurring infections and hospitalisations due to these infections throughout childhood. We have shown that breastfeeding for less than three months was a risk factor against parent-reported recurring chest infections and hospitalisations for these infections, and that being born with low birth weight was a risk factor for infections throughout childhood and adolescence.

Traditionally, Aboriginal mothers breastfeed their infants frequently and exclusively for six months or more.30 In remote areas, Aboriginal mothers often breastfeed for two years or longer. However, most Aboriginal populations now live in cities and urban areas and have introduced Westernised practices of formula feeding. In Aboriginal groups in Perth, Western Australia, in which breastfeeding initiation and duration was measured, it was shown that breastfeeding rates were similar to that of other women in Perth.31 While fewer Aboriginal women who smoked were still breastfeeding at 24 weeks postpartum compared with non-smokers (58% vs. 64%), this difference was not significant.32 In a Melbourne study of Aboriginal infants, only 50% of babies were being breastfed at age three months.33

This earlier cessation of breastfeeding may have long-term health implications.34,35 In central Australia, Aboriginal children have very high incidence rates of invasive pneumococcal disease with infection rates 60–80 times higher than Anglo-European communities.5 Aboriginal children also have high rates of acute lower respiratory tract infections,12 and in north-eastern Australia rates for hospitalisation for pneumonia remain high.36 In 1983, a high prevalence of respiratory infections was recorded in a study of 1,287 Aboriginal children living in Western Australia.37 In this study, one in five children aged 0–4 years had a purulent nasal discharge and one in five children aged 5–9 years had a current loose cough. Our results for hospitalisations due to non-wheezing lower respiratory infections approached significance. However, because the effect of isolation and disadvantage were strong risk factors for these types of infections, any protection afforded by breastfeeding may have been overwhelmed by environmental factors in our study. We showed that increasing age was a significant risk for all the infection outcomes tested, lending support to the suggestion that the vicious cycle of endemic infections are perpetuated by high carriage rates of multiple species and multiple types of respiratory bacterial pathogens by high cross-infection rates and thus, by early age, of pathogen acquisition and prolonged carriage.18

In a meta-analytic review of the risk factors for otitis media, breastfeeding for at least three months was protective against the risk of acute otitis media (RR 0.87; 95% CI 0.79-0.95; p=0.003).17 However, we did not show a similar association in our study, and our finding is in agreement with the PROBIT randomised controlled study in Belarus that showed no significant reduction in upper respiratory tract infection (intervention group, 36.1%; control group, 36.2%; adjusted OR 0.87; 95% CI 0.58-1.30) or otitis media (intervention group, 6.2%; control group, 6.0%; adjusted OR 1.01; 95% CI 0.54-1.88).38

Although lower respiratory tract infections remain the leading cause of hospitalisations in Aboriginal infants and children, mortality rates are lower than those reported for developing countries.12 The severity of infections in Aboriginal children is decreasing, with fewer children admitted to hospital than previously, but there remains continuing concern about high rates of respiratory infections due to poor living conditions and inadequate vaccine delivery.39 In addition, the enormous burden of middle ear infections in Aboriginal children is well known, with the resulting hearing loss affecting school performance and social circumstances later in life.40 In our study, hearing loss was associated with poorer academic outcomes for Aboriginal children, but it was not the main predictor of poor school performance nor was there any observed association between ear infections and school performance.41

Human milk may confer several effects on the development of the respiratory tract and subsequent ability to fight infection and illness, providing protection against recurring chest infections and severe infections leading to hospitalisation. Specific nutritional, immuno-regulatory and bioactive components in milk may promote maturation of infant immune competence.42,43 Exclusive breastfeeding has been observed to slow the involution of the thymus gland during infancy,44 which is likely to have significant effects upon systemic T-cell function, given that this organ supplies the bulk of naïve T-cell precursors during early life. Furthermore, bioactive factors such as immunoglobulins and other proteins, glycoconjugates and oligosaccharides, lipids, anti-inflammatory components, cells, neutrophils, macrophages, nucleotides, enzymes, hormones and growth factors impart defence to the newborn infant.45 Since infants aspirate a small amount of milk during feeding, immunoglobulin A and other protective factors delivered to the respiratory tract might protect against these infection types46 and in another study47 we showed more Transforming Growth Factor-beta from milk, possibly through an effect on lung development, influenced wheeze at one year of age. Conversely, feeding of formula milk may be harmful (high bacterial load of formula milk, unhygienic preparation of formula milk, low rates of literacy among mothers using formula milk or poor quality water mixed with formula milk) as has been demonstrated in studies related to mother-to-child HIV transmission.48,49

Existing studies suggest that maternal recall is a valid and reliable estimate of breastfeeding initiation and duration, especially when the duration of breastfeeding is recalled after a short period (three years or less).50 To address the possibility of recall bias as a confounding element on the validity of the initiation and duration of breast feeding data in our study, we looked at the proportion of Aboriginal children ever breastfed, breastfed for six months and breastfed for 12 months or more by age at time of survey. We found that the proportion of ever breastfed children, children breastfed for six months or children breastfed for 12 months or longer by age at time of survey did not change over time, suggesting that these data are recalled with accuracy.

To build trust among the Aboriginal study respondents, we trained more than 130 screeners/interviewers, of whom 60% were Aboriginal. Training of the staff took seven days and after trial workloads and attrition, the field staff numbers dropped to 68, of whom 25% were Aboriginal. It is of note that having an Aboriginal interviewer was not always desirable due to the small Aboriginal population in Western Australia. This meant that many Aboriginal people were related and/or knew each other, which resulted in confidentiality issues not appropriate for the survey. So, while we did eventually employ 25% Aboriginal interviewers, there was a requirement that they had to be well qualified and deployed in areas where confidentiality issues were minimised.

Although this observational, retrospective analysis limits the conclusions that can be drawn, we have shown a significant association between a longer duration of breastfeeding and a reduced prevalence of recurring chest infections and hospitalisations due to these infections. Reporting of respiratory infections by the primary caregiver (in this case the parent) are subject to misclassification error due to lack of standardisation between culturally diverse communities and parental recall. We have minimised this bias by using aboriginal health workers to collect our data wherever possible. The comprehensive method in which Aboriginal infants and children were ascertained was a strength of our study. In addition, reliance on subjective questionnaire measurements can not always be reliable and by linking the study participant data to the Hospital Morbidity System we have been able to verify the association with breastfeeding and more severe infections. One of the limitations of our study was that we considered hospital admissions. From our clinical experience, there are a significant number of ear infections treated in the community with very few resulting in a hospital admission.

Breastfeeding for at least three months has been shown to reduce the incidence of chest infections in Aboriginal infants and children, but rates of infections and hospitalisations due to these infections continue to be high. Although additional studies are required to confirm these findings and to understand the mechanisms involved, public health interventions to promote breastfeeding for at least three months and up to two years may reduce the prevalence and subsequent morbidity of respiratory illness and infection in infancy. Interventions to increase breastfeeding and to reduce the prevalence of low birth weight should be primary health goals in Aboriginal communities for the benefits of Aboriginal infants and children.

Funding for the research

Dr W. H. Oddy is supported by a National Health and Medical Research Council Population Health Research Fellowship. The Western Australian Aboriginal Child Health Survey was funded by project and program grants from the Western Australian Health Promotion Foundation, LotteryWest, the Government of Western Australia, the Australian Government, and the Rio Tinto Aboriginal Foundation.

Acknowledgements

We thank the West Australian Aboriginal Child Health Survey Steering Committee for implementing this unique and comprehensive study of Aboriginal infants and children, and the study families who participated, without whom this research would not have been possible.

Ancillary