Temporal trends and socioeconomic differences in acute respiratory infection hospitalisations in children: an intercountry comparison of birth cohort studies in Western Australia, England and Scotland

Objectives Acute respiratory infections (ARIs) are a global cause of childhood morbidity. We compared temporal trends and socioeconomic disparities for ARI hospitalisations in young children across Western Australia, England and Scotland. Design Retrospective population-based cohort studies using linked birth, death and hospitalisation data. Setting and participants Population birth cohorts spanning 2000–2012 (Western Australia and Scotland) and 2003–2012 (England). Outcome measures ARI hospitalisations in infants (<12 months) and children (1–4 years) were identified through International Classification of Diseases, 10th edition diagnosis codes. We calculated admission rates per 1000 child-years by diagnosis and jurisdiction-specific socioeconomic deprivation and used negative binomial regression to assess temporal trends. Results The overall infant ARI admission rate was 44.3/1000 child-years in Western Australia, 40.7/1000 in Scotland and 40.1/1000 in England. Equivalent rates in children aged 1–4 years were 9.0, 7.6 and 7.6. Bronchiolitis was the most common diagnosis. Compared with the least socioeconomically deprived, those most deprived had higher ARI hospitalisation risk (incidence rate ratio 3.9 (95% CI 3.5 to 4.2) for Western Australia; 1.9 (1.7 to 2.1) for England; 1.3 (1.1 to 1.4) for Scotland. ARI admissions in infants were stable in Western Australia but increased annually in England (5%) and Scotland (3%) after adjusting for non-ARI admissions, sex and deprivation. Conclusions Admissions for ARI were higher in Western Australia and displayed greater socioeconomic disparities than England and Scotland, where ARI rates are increasing. Prevention programmes focusing on disadvantaged populations in all three countries are likely to translate into real improvements in the burden of ARI in children.

we applied a hierarchical diagnosis algorithm [6] within the readmission set in order to code an overall principal diagnosis. This algorithm ranked diagnoses in order of disease severity: whooping cough, pneumonia, bronchiolitis, influenza, unspecified ALRI and bronchitis. Children with missing data on sex or deprivation were excluded from the analyses.

Exposure Measures
We assessed hospitalisations for ARI in infants aged less than 12 months and young children aged 1- people [24] In all jurisdictions socio-economic deprivation scores were based on mother's residential address at time of her child's delivery and were grouped into four levels based on a country level ranking with the lowest scores representing the most socio-economically deprived.

Statistical Analysis
Consistent methodology was applied to the assembled datasets in the three jurisdictions. We calculated hospitalisation rates per 1000 child-years at risk for each diagnostic grouping of ARI (as principal diagnosis). To assess the impact of including additional diagnosis codes, we compared hospitalisation rates derived using the principal diagnosis code only with rates derived from using the principal plus all additional diagnosis codes (any diagnosis). We used any diagnosis to assess ARI rates by socio-economic deprivation and year of admission. We present age-specific hospitalisation rates with 95% confidence intervals (CI) and where appropriate, rates were compared using incidence rate ratios (IRRs) with 95% CIs. To assess temporal trends, we plotted annual hospitalisation rates in the two age groups for each jurisdiction by admission year for all ARIs and bronchiolitis, pneumonia and unspecified ALRI's. We also used negative binomial regression models to assess linear temporal trends in infant hospitalisations from 2001-2012 (Western Australia and Scotland) and 2004.
Year of admission was included as a linear term in the models, and the natural logarithm of child-years at risk was included as an offset in the models. Trends over time in ARI admission rates were assumed to be statistically significant if the Wald test p-value for the coefficient for the linear year term was <0.05. Models were adjusted for sex and the 4-level socioeconomic indicator and we present IRR's with 95% CI's. In order to control for overall trends in hospitalisation we also adjusted the models for the number of all non-ARI emergency admissions. [25] All data analyses were conducted within each jurisdiction in Stata version 14.0. [26] Figure 1). The relative difference in ARI hospitalisation rates between the most and least deprived infants was 3.5 (95% CI: 3.2, 3.7) in Western Australia; 1.8 for England and 1.3 for Scotland with similar patterns in children aged 1-4 years ( Figure 1). In multivariable models, level of socio-economic deprivation was significantly associated with all ARI categories in all infants but most notably in Western Australia, and in particular, pneumonia (IRR 6.9, 95% CI: 5.6, 8.6) and unspecified ALRI (IRR 8.9, 95% CI: 6.7, 11.8; Table 2).
Overall, ARI hospitalisation rates have increased in England and Scotland, but declined (infants) or remained stable (children aged 1-4 years) in Western Australia ( Figure 2 Figure 2). Similar results were seen for bronchiolitis admissions in infants.
Diverging trends were seen with pneumonia and unspecified ALRI across the three jurisdictions with pneumonia hospitalisation rates in infants declining in Western Australia from 9.0/1000 in 2002 to 3.9/1000 in 2012 while rates remained steady around 3-4/1000 in England and 2-3/1000 in Scotland ( Figure 2). After adjusting for sex, socio-economic deprivation and non-ARI admissions, the annual decline in pneumonia hospitalisations was 6% in Western Australia (IRR 0.94, 95%CI: 0.93, 0.96), 2% in England and 3% in Scotland (Table 2). Unspecified ALRI declined in Western Australia annually by 5% but increased by 6% and 2% annually in England and Scotland (Table 2). The interpretation of hospitalisation trends across countries is complex. We have found higher rates of ARI admissions in Western Australia compared with England and Scotland which could mean a higher incidence in ARI, a higher risk of developing more severe symptoms, or differences in diagnostic coding or hospital admission thresholds. A recent study comparing admission rates between England and Ontario finding substantially higher rates in England was partly explained by differing admission thresholds from differential waiting practices and policies in emergency departments. [4] Comparisons of asthma admissions from national hospital data in Finland and Sweden noted diverging trends citing differences in national coding guidelines and subsequent altered admission thresholds. [27] In an attempt to control for changing admissions thresholds over time within each jurisdiction, we adjusted our multivariable models for the overall trend in non-ARI emergency hospital use. However we could not adjust for differing thresholds between countries.
Emergency hospitalisations are increasing at a faster rate in England compared to other parts of the United Kingdom [28] and our data here suggests that hospitalisations due to unspecified ALRI and bronchiolitis in England are contributing to that increase. It is also possible that diverging trends are a result of diagnostic shifts in that for the same clinical presentations, a diagnosis of unspecified ALRI is given in England while other non-specific codes (including codes we have not assessed) are given in Western Australia and Scotland. The use of additional diagnosis codes for ARI seemed more frequent in Western Australia compared with England and Scotland and should be taken into consideration for future comparative studies using ICD diagnosis codes.
Hospitalisation rates for ARI were significantly associated with level of socio-economic deprivation, consistent with an earlier analysis in England. [29] This association was strongest in Western Australia with IRRs for those in the most deprived level in the order of 3.9 for all ARIs, up to 8.9 for unspecified ALRI. There appeared a linear relationship with level of deprivation and rates of ARI in Western Australia while rates in all levels (bar the most deprived) not differing in England and Scotland. Western Australian data were inclusive of Aboriginal children, an Indigenous population with higher levels of socio-economic disadvantage[30] compared to their non-Aboriginal peers and a significantly higher burden of pneumonia worldwide, [6,31,32] despite reductions in the 2000's and further reductions seen in our results here, most likely due to the positive impact of pneumococcal vaccination. [6,33] This most likely explains the higher rates of pneumonia seen in Western Australia compared with England and Scotland. Aboriginal children also suffer a disproportionate burden of RSV,[34] the major cause of bronchiolitis which could explain the higher bronchiolitis rates in Western Australia than in England and Scotland. However level of socio-economic deprivation has been associated with hospitalisations for respiratory infections in both Aboriginal and non-Aboriginal children [9] so the contribution of Aboriginal children alone cannot explain the higher socioeconomic disparities seen here. Indeed, when Aboriginal children were removed from the analysis, the socio-economic disparities remained, although slightly lessened, and were still higher than England and Scotland (e.g. the IRR for most deprived children for all ARI reduced from 3.9 to 2.1 and for unspecified ALRI reduced from 8.9 to 2.9 (data not shown)). Respiratory infections continue to be a source of health inequalities among disadvantaged children worldwide. Geographical remoteness is more of an issue in Western Australia due to its sheer geographical size in comparison to England and Scotland. The lack of adequate primary care in rural and remote Australia [35] which is often coupled with lower socio-economic levels could be driving higher hospitalisation rates.
Nevertheless, these important findings highlight the need for targeted prevention programs such as smoking cessation, improved housing and timely vaccination for key respiratory pathogens for the most disadvantaged populations in all three jurisdictions.
Unlike the United Kingdom, Australia does not have a uniform policy for seasonal influenza vaccination. Relative to other ARI diagnoses, recorded influenza hospitalisation rates are low.
Assessing the impact of the universal childhood vaccination program for influenza in the United Kingdom introduced in 2013-14 is likely to be challenging without linking national-level birth cohorts to infection surveillance data. This has already been implemented in Scotland. [36] There is also renewed interest in preventing morbidity due to RSV with vaccination.
[37] Understanding the baseline hospitalisation rates for RSV-bronchiolitis and pneumonia prior to when vaccination is available is critical to aid in implementation and for its ongoing evaluation post implementation.
We conducted our analysis on near total population birth cohorts in each jurisdiction and thus our outcome measures have narrow confidence intervals and minimum selection bias. An additional strength of the population-based cohort design is standardisation of analysis protocols and the provision of large numbers allowing us to assess temporal trends and associations with less common

Consent for publication
Not applicable.

Conclusions:
Admissions for ARI were higher in Western Australia and displayed greater socioeconomic disparities than England and Scotland, where ARI rates are increasing. Prevention programs focusing on disadvantaged populations in all three countries are likely to translate into real improvements in the burden of ARI in children. ARI hospitalisations are more common among children from poorer socio-economic backgrounds. [8,9] In addition to access to inadequate health care, risk factors for developing severe symptoms of ARIs, including prematurity, low birth weight, congenital anomalies, exposure to environmental tobacco smoke, damp and mould, and household overcrowding are all more common among children growing up in more deprived families in both high and low income settings. [10,11] Understanding the impact of socio-economic disparities on ARI hospitalisations among children (both over time and between countries) can provide an estimate of the preventable proportion of ARI. Linkage of administrative health datasets provides a platform to investigate these trends in populations over many years. Additionally, the availability of comparable hospital admission datasets with similar coding systems using International Classification of Diseases, 10 th edition (ICD-10) diagnosis codes allows comparison of hospitalisation rates among children for ARI according to deprivation level. recently, seasonal influenza. Excluding influenza, vaccination coverage at age 12 months is >90% for all 3 jurisdictions. [12,13] Our aim was to compare population-based hospitalisation rates by ARI diagnosis, age and level of socio-economic deprivation, and assess how ARI hospitalisation rates have changed over time.

Data Sources and Study Populations
We conducted separate population-based birth cohort studies using administrative data from Western Australia, England and Scotland. Western Australia covers the western third of Australia, an area of 2.5 million square kilometres with a population of nearly 2.6 million, [14] 3.6% of whom identify as being Aboriginal and/or Torres Strait Islander (herein referred to as Aboriginal). [15] Births were identified from the Midwives' Notification System and Birth Register, deaths were identified from the Death Register and hospitalisations were recorded in the Hospital Morbidity Database Collection that provides full coverage of all hospital separations (hereafter referred to as hospitalisations). In the absence of a unique person identifier in Australia, extracted data were probabilistically linked by the Western Australian Data Linkage Branch using a series of demographic identifiers using an established best practice protocol. [16,17] Aboriginal status was derived using a England has a population of 53.9 million. [19] The birth cohort was established by linking hospital birth and delivery records from the Hospital Episode Statistics (HES) database. [20] Hospitalisations and deaths were identified via linkage to mortality registration data from the Office for National Statistics. [21] Data linkage in England was carried out by NHS Digital, using a deterministic algorithm

Outcome Measures
Our outcome measure was an ARI emergency hospitalisation for children in their first 5 years of life.
All inter-hospital transfers were collapsed into a single admission. We identified hospitalisations for ARI using a selection of ICD-10 diagnosis codes (ICD-10-AM for Western Australia). [25]  we applied a hierarchical diagnosis algorithm [6] within the readmission set in order to code an overall principal diagnosis. This algorithm ranked diagnoses in order of disease severity: whooping cough, pneumonia, bronchiolitis, influenza, unspecified ALRI and bronchitis. Children with missing data on sex or deprivation were excluded from the analyses.

Exposure Measures
We assessed hospitalisations for ARI in infants aged less than 12 months and young children aged 1-  class. The Carstairs Index is measured at postcode sector level, which contains an average of 5000 people [28] In all jurisdictions socio-economic deprivation scores were based on mother's residential address at time of her child's delivery and were grouped into four levels based on a country level ranking with the lowest scores representing the most socio-economically deprived.

Statistical Analysis
Consistent methodology was applied to the assembled datasets in the three jurisdictions. We calculated hospitalisation rates per 1000 child-years at risk for each diagnostic grouping of ARI (as principal diagnosis). To assess the impact of including additional diagnosis codes, we compared hospitalisation rates derived using the principal diagnosis code only with rates derived from using the principal plus all additional diagnosis codes (any diagnosis). We used any diagnosis to assess ARI rates by socio-economic deprivation and year of admission. We present age-specific hospitalisation rates with 95% confidence intervals (CI) and where appropriate, rates were compared using incidence rate ratios (IRRs) with 95% CIs. To assess temporal trends, we plotted annual hospitalisation rates in the two age groups for each jurisdiction by admission year for all ARIs and bronchiolitis, pneumonia and unspecified ALRI's. We also used negative binomial regression models  Table 2).
ARI hospitalisation rates were higher for children from the most socio-economically deprived areas.
The association with deprivation was greatest in Western Australia and more marked in infants compared to young children aged 1-4 years ( Figure 1). The relative difference in ARI hospitalisation rates between the most and least deprived infants was 3.5 (95% CI: 3.2, 3.7) in Western Australia; 1.8 for England and 1.3 for Scotland with similar patterns in children aged 1-4 years ( Figure 1). In multivariable models, level of socio-economic deprivation was significantly associated with all ARI categories in all infants but most notably in Western Australia, and in particular, pneumonia (IRR 6.9, 95% CI: 5.6, 8.6) and unspecified ALRI (IRR 8.9, 95% CI: 6.7, 11.8; Table 2).
Overall, ARI hospitalisation rates have increased in England and Scotland, but declined (infants) or remained stable (children aged 1-4 years) in Western Australia ( Figure 2 Figure 2). Similar results were seen for bronchiolitis admissions in infants.
Diverging trends were seen with pneumonia and unspecified ALRI across the three jurisdictions with pneumonia hospitalisation rates in infants declining in Western Australia from 9.0/1000 in 2002 to 3.9/1000 in 2012 while rates remained steady around 3-4/1000 in England and 2-3/1000 in Scotland in England and 3% in Scotland (Table 2). Unspecified ALRI declined in Western Australia annually by 5% but increased by 6% and 2% annually in England and Scotland (Table 2).

DISCUSSION
ARI, particularly bronchiolitis, continues to be an important cause of infant and childhood hospitalisation. The availability of linked administrative data in three economically similar jurisdictions with publicly funded healthcare systems afforded us the opportunity to compare ARI hospitalisation rates in children. Overall, admission rates were highest in Western Australia and decreasing or remaining stable but increasing in England and Scotland. The relative differences in ARI admission rates between children from the most socioeconomically deprived areas to the least deprived areas were largest in Western Australia.
The interpretation of hospitalisation trends across countries is complex. We have found higher rates of ARI admissions in Western Australia compared with England and Scotland which could mean a higher incidence in ARI, a higher risk of developing more severe symptoms, or differences in diagnostic coding or hospital admission thresholds. A recent study comparing admission rates between England and Ontario finding substantially higher rates in England was partly explained by differing admission thresholds from differential waiting practices and policies in emergency departments. [4] Comparisons of asthma admissions from national hospital data in Finland and Sweden noted diverging trends citing differences in national coding guidelines and subsequent altered admission thresholds.
[31] In an attempt to control for changing admissions thresholds over time within each jurisdiction, we adjusted our multivariable models for the overall trend in non-ARI  However, our study does have some limitations. The socio-economic deprivation scores used were jurisdiction specific and included different items to represent disadvantage. In addition, area-level socio-economic deprivation was only measured at birth. Therefore, the observed association between area-level socio-economic deprivation and the rate of ARI admissions may be subject to increasing measurement error as the child's age increases. How socio-economic deprivation is associated with morbidity due to ARI at the primary care level is unknown but perhaps likely to aid in explaining disparities in socio-economic deprivation that we have seen here. While primary care data is more readily available in England and Scotland, limited data with adequate diagnostic information is available for population-based studies in Western Australia. As previously alluded to, there also may be differences in admission thresholds across the three jurisdictions that may explain some higher admission rates across countries. A comparison of emergency department presentations in conjunction with hospitalisations for ARI could be useful here, although diagnostic information from emergency department data is limited [43] and no individual level data on emergency department visits exist in Scotland. Additionally, through our experience of linking routine laboratory data to hospital data in Western Australia, we are aware of unspecific ICD codes             ARI hospitalisations are more common among children from poorer socio-economic backgrounds. [8,9] In addition to access to inadequate health care, risk factors for developing severe symptoms of ARIs, including prematurity, low birth weight, congenital anomalies, exposure to environmental tobacco smoke, damp and mould, and household overcrowding are all more common among children growing up in more deprived families in both high and low income settings. [10,11] Understanding the impact of socio-economic disparities on ARI hospitalisations among children (both over time and between countries) can provide an estimate of the preventable proportion of ARI. Linkage of administrative health datasets provides a platform to investigate these trends in populations over many years. Additionally, the availability of comparable hospital admission datasets with similar coding systems using International Classification of Diseases, 10 th edition (ICD-10) diagnosis codes allows comparison of hospitalisation rates among children for ARI according to deprivation level.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  recently, seasonal influenza. Excluding influenza, vaccination coverage at age 12 months is >90% for all 3 jurisdictions. [12,13] Our aim was to compare population-based hospitalisation rates by ARI diagnosis, age and level of socio-economic deprivation, and assess how ARI hospitalisation rates have changed over time.

Data Sources and Study Populations
We conducted separate population-based birth cohort studies using administrative data from Western Australia, England and Scotland. Western Australia covers the western third of Australia, an area of 2.5 million square kilometres with a population of nearly 2.6 million, [14] 3.6% of whom identify as being Aboriginal and/or Torres Strait Islander (herein referred to as Aboriginal). [15] Births were identified from the Midwives' Notification System and Birth Register, deaths were identified from the Death Register and hospitalisations were recorded in the Hospital Morbidity Database Collection that provides full coverage of all hospital separations (hereafter referred to as hospitalisations). In the absence of a unique person identifier in Australia, extracted data were probabilistically linked by the Western Australian Data Linkage Branch using a series of demographic identifiers using an established best practice protocol. [16,17] Aboriginal status was derived using a  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   6 validated algorithm using Aboriginal identification information across all available records. [18] England has a population of 53.9 million. [19] The birth cohort was established by linking hospital birth and delivery records from the Hospital Episode Statistics (HES) database. [20] Hospitalisations and deaths were identified via linkage to mortality registration data from the Office for National Statistics.

Outcome Measures
Our outcome measure was an ARI emergency hospitalisation for children in their first 5 years of life.
All inter-hospital transfers were collapsed into a single admission. We identified hospitalisations for ARI using a selection of ICD-10 diagnosis codes (ICD-10-AM for Western Australia). [25] 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  we applied a hierarchical diagnosis algorithm [6] within the readmission set in order to code an overall principal diagnosis. This algorithm ranked diagnoses in order of disease severity: whooping cough, pneumonia, bronchiolitis, influenza, unspecified ALRI and bronchitis. Children with missing data on sex or deprivation were excluded from the analyses. Deaths due to ARI in these populations are rare and our data would be not sufficiently powered to assess mortality rates in this cohort, especially for Western Australia and Scotland. As such we do not report ARI-related mortality rates here and focus our outcome measure on ARI-related hospitalisations.

Statistical Analysis
Consistent methodology was applied to the assembled datasets in the three jurisdictions. We calculated hospitalisation rates per 1000 child-years at risk for each diagnostic grouping of ARI (as principal diagnosis). To assess the impact of including additional diagnosis codes, we compared hospitalisation rates derived using the principal diagnosis code only with rates derived from using the principal plus all additional diagnosis codes (any diagnosis). We used any diagnosis to assess ARI rates by socio-economic deprivation and year of admission. We present age-specific hospitalisation rates with 95% confidence intervals (CI) and where appropriate, rates were compared using incidence rate ratios (IRRs) with 95% CIs. To assess temporal trends, we plotted annual hospitalisation rates in the two age groups for each jurisdiction by admission year for all ARIs and bronchiolitis, pneumonia and unspecified ALRI's. We also used negative binomial regression models  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   F  o  r  p  e  e  r  r  e  v  i  e  w  o  n  l  y   9 time in ARI admission rates were assumed to be statistically significant if the Wald test p-value for the coefficient for the linear year term was <0.05. Models were adjusted for sex and the 4-level socioeconomic indicator and we present IRR's with 95% CI's. In order to control for overall trends in hospitalisation we also adjusted the models for the number of all non-ARI emergency admissions. [29] All data analyses were conducted within each jurisdiction in Stata version 14.0. [30]

Public and Patient Involvement
A community reference group located in Western Australia was consulted during the conduct of this study. No individual patients were involved.

RESULTS
A total of 337,909 (Western Australia), 5,939,009 (England) and 699,590 (Scotland) births were included in the study (Supplementary Table 1 Table 2).
ARI hospitalisation rates were higher for children from the most socio-economically deprived areas.
The association with deprivation was greatest in Western Australia and more marked in infants compared to young children aged 1-4 years ( Figure 1). The relative difference in ARI hospitalisation rates between the most and least deprived infants was 3.5 (95% CI: 3.2, 3.7) in Western Australia; 1.8 for England and 1.3 for Scotland with similar patterns in children aged 1-4 years ( Figure 1). In multivariable models, level of socio-economic deprivation was significantly associated with all ARI categories in all infants but most notably in Western Australia, and in particular, pneumonia (IRR 6.9, 95% CI: 5.6, 8.6) and unspecified ALRI (IRR 8.9, 95% CI: 6.7, 11.8; Table 2).
Overall, ARI hospitalisation rates have increased in England and Scotland, but declined (infants) or remained stable (children aged 1-4 years) in Western Australia (Figure 2 Figure 2). Similar results were seen for bronchiolitis admissions in infants.  in England and 3% in Scotland (Table 2). Unspecified ALRI declined in Western Australia annually by 5% but increased by 6% and 2% annually in England and Scotland (Table 2).

DISCUSSION
ARI, particularly bronchiolitis, continues to be an important cause of infant and childhood hospitalisation. The availability of linked administrative data in three economically similar jurisdictions with publicly funded healthcare systems afforded us the opportunity to compare ARI hospitalisation rates in children. Overall, admission rates were highest in Western Australia and decreasing or remaining stable but increasing in England and Scotland. The relative differences in ARI admission rates between children from the most socioeconomically deprived areas to the least deprived areas were largest in Western Australia.
The interpretation of hospitalisation trends across countries is complex. We have found higher rates of ARI admissions in Western Australia compared with England and Scotland which could mean a higher incidence in ARI, a higher risk of developing more severe symptoms, or differences in diagnostic coding or hospital admission thresholds. A recent study comparing admission rates between England and Ontario finding substantially higher rates in England was partly explained by differing admission thresholds from differential waiting practices and policies in emergency  However, our study does have some limitations. The socio-economic deprivation scores used were jurisdiction specific and included different items to represent disadvantage. In addition, area-level socio-economic deprivation was only measured at birth. Therefore, the observed association between area-level socio-economic deprivation and the rate of ARI admissions may be subject to increasing measurement error as the child's age increases. How socio-economic deprivation is associated with morbidity due to ARI at the primary care level is unknown but perhaps likely to aid in explaining disparities in socio-economic deprivation that we have seen here. While primary care data is more readily available in England and Scotland, limited data with adequate diagnostic information is available for population-based studies in Western Australia. As previously alluded to, there also may be differences in admission thresholds across the three jurisdictions that may explain