Breastfeeding practice, breastfeeding policy and hospitalisations for infectious diseases in early and later childhood: a register-based study in Uppsala County, Sweden

Objective To examine the association between breastfeeding practice and hospitalisations for infectious diseases in early and later childhood, in particular, to compare exclusive breast feeding 4–5 months with exclusive breastfeeding 6 months or more. Thereby, provide evidence to inform breastfeeding policy. Design A register-based cohort study. Setting A cohort was created by combining the Swedish Medical Birth Register, the National Inpatient Register, the Cause of Death Register, the Total Population Register, the Longitudinal integration database for health insurance and labour market studies, with the Uppsala Preventive Child Health Care database. Patients 37 825 term and post-term singletons born to women who resided in Uppsala County (Sweden) between 1998 and 2010. Main outcome measures Number of hospitalisations for infectious diseases in early (<2 years) and later childhood (2–4 years). Results The risk of hospitalisations for infectious diseases decreased with duration of exclusive breastfeeding until 4 months of age. In early childhood, breast feeding was associated with a decreased risk of enteric and respiratory infections. In comparison with exclusive breast feeding 6 months or more, the strongest association was found between no breastfeeding and enteric infections (adjusted incidence rate ratios, aIRR 3.32 (95% CI 2.14 to 5.14)). In later childhood, breast feeding was associated with a lower risk of respiratory infections. In comparison with children exclusively breastfed 6 months or more, the highest risk was found in children who were not breastfed (aIRR 2.53 (95% CI 1.51 to 4.24)). The risk of hospitalisations for infectious diseases was comparable in children exclusively breastfed 4–5 months and children exclusively breastfed 6 months or more. Conclusions Our results support breastfeeding guidelines that recommend exclusive breastfeeding for at least 4 months.

Results The risk of hospitalisations for infectious diseases decreased with duration of exclusive breastfeeding until 4 months of age. In early childhood, breast feeding was associated with a decreased risk of enteric and respiratory infections. In comparison with exclusive breast feeding 6 months or more, the strongest association was found between no breastfeeding and enteric infections (adjusted incidence rate ratios, aIRR 3.32 (95% CI 2.14 to 5.14)). In later childhood, breast feeding was associated with a lower risk of respiratory infections. In comparison with children exclusively breastfed 6 months or more, the highest risk was found in children who were not breastfed (aIRR 2.53 (95% CI 1.51 to 4.24)). The risk of hospitalisations for infectious diseases was comparable in children exclusively breastfed 4-5 months and children exclusively breastfed 6 months or more. Conclusions Our results support breastfeeding guidelines that recommend exclusive breastfeeding for at least 4 months.

BACKGROUND
Infectious diseases cause large harm and suffering in childhood. Globally, infectious diseases are a leading cause of death in children under 5 years of age. 1 In high-income countries, infectious disease remains a major cause of hospitalisations in young children. 2 3 Breast milk contains IgA antibodies and oligosaccharides that prevent microbes to adhere to the mucosa, lactoferrin and lysozyme act directly on the microbes, and growth factors, nucleotides, cytokines stimulate maturation of the infant's immune system. 4 5 Numerous studies have shown that breastfeeding (BF) reduces the risk of infectious diseases in infancy. 4 6 7 A few studies have also reported that BF reduces the risk of infectious diseases after infancy. [8][9][10] Although it is well known that BF protects young children in high-income countries against infections, the optimal BF practice is Strengths and limitations of this study ► This register-based cohort study included almost all (<10% missing data) term and post-term singletons born in a geographical region over 13 years, which reduces the risk of selection bias. ► Children were followed from birth until 5 years of age or censoring, which enabled us to examine the effect of breast feeding in both early (<2 years) and later childhood (2-4 years). ► Information on pregnancy, birth and sociodemographic characteristics were obtained from highquality health and administrative registers, which allowed us to adjust analyses for several potential confounders including congenital malformation, large/small for gestational age, maternal age, maternal smoking, parity, maternal education level and maternal country of birth. ► Infections were identified using deidentified hospital discharge data, cases could not be confirmed by information obtained from medical records. ► Breastfeeding practice was not continuously collected, information on breastfeeding practice was only available at 1 week, 2, 4 and 6 months.

Open access
still under debate. WHO recommends exclusive BF (EBF) until 6 months of age followed by BF along with complementary feeding. 11 The American Academy of Pediatrics recommends BF for at least 12 months with EBF for about 6 months. 12 In contrast, the European Society for Paediatric Gastroenterology, Hepatology, and Nutrition recommend EBF for at least 4 months with EBF or predominant BF for approximately 6 months. 6 The aim of this study was to examine the association between BF practice and hospitalisations for infectious diseases in early (<2 years) and later childhood (2-4 years), in particular, to compare EBF 4-5 months with EBF 6 months or more.

Patient and public involvement
This study was conducted without patient involvement.

Setting
In Sweden, all children are offered free primary and hospital care, a free preventive child healthcare programme and a free general immunisation programme; working parents are entitled to a generous parental leave scheme. 13 During the study period, Swedish guidelines for BF duration were consistent with WHO recommendations. Until 2002, EBF was recommended until 4-6 months of age. Thereafter, EBF was recommended for at least 6 months. 14

Study population
This is a register-based cohort study of term and postterm (gestational age 37 weeks or more) singletons born to women who resided in Uppsala County between 1998 and 2010. A database was created by combining data from nationwide and local health and administrative registers. The Swedish Medical Birth Register includes information on prenatal, delivery and neonatal care, with coverage of 98%-99% of all births. 15 The National Patient Register contains information on all hospitalisations including International Classification of Disease, 10th Revision (ICD-10) codes for primary diagnosis. 16 The longitudinal integration database for health insurance and labour market studies, includes information on maternal education. 17 The Cause of Death Register provides information on mortality. The Total Population Register contains information about migration and maternal country of birth. The Child Healthcare Quality database in Uppsala contains data on BF, which was collected by child health nurses during free routine health checkups. 13 These registers were linked using the unique personal identification number assigned to every Swedish resident, by the Centre for Epidemiology at the Swedish National Board of Health and Welfare. All data were anonymised and deidentified prior to analysis. In 2004 the WHO definition of EBF was adopted in Sweden, and tastes were no longer accepted to fulfil this criterion. 18 19 Covariates Analyses were adjusted for several child and maternal characteristics. Information on small for gestational age (yes or no), large for gestational age (yes or no), congenital malformation (ICD-10 codes Q00-Q99) and sex (male or female), maternal age (≤19, 20-24, 25-29, 30-34 and ≥35), parity (1, 2, 3 and ≥4), maternal smoking at the first antenatal care visit (yes or no) and year of birth were retrieved from the Medical Birth Register. Maternal education level at year of birth (secondary school or less (≤9 years), upper secondary school (10-12 years), short postsecondary education (13-14 years) and long postsecondary education (≥15 years)) was obtained from the Longitudinal integration database for health insurance and labour market studies. Information on maternal country of birth (Sweden, Other Nordic, Other Europe and North America, Asia, Africa and Other) was retrieved from the Total Population Register.

Outcomes
The main outcome was overall number of hospitalisations with a principal diagnosis of infectious disease recorded in the National Patient Register. Secondary outcomes were number of hospitalisations for respiratory tract and enteric infections. Hospitalisations were recorded using ICD-10 codes. We used a previously developed coding scheme to identify infectious disease codes and group them into infectious disease categories (online supplemental appendix A). 3

Statistical analysis
Crude and adjusted associations between BF practice and number of hospitalisations were estimated using negative binomial regression models. Log follow-up time (days) was used as an offset. Results were presented as adjusted incidence rate ratios (aIRRs) with 95% CIs. Correlations between siblings were accounted for using generalised estimating equations with robust standard errors. 20 All Open access analyses were adjusted for time trends (year of birth). The adjusted analyses were also controlled for small for gestational age, large for gestational age, congenital malformation, sex, maternal age, parity, maternal smoking, maternal education level and maternal country of birth. Crude and adjusted models were fitted for each outcome (overall, respiratory and enteric infections) and follow-up period (early childhood (<2 years) and later childhood (2-4 years)). In the later follow-up period, an additional adjusted model controlled for previous admissions (a binary variable indicating hospitalisation in the early follow-up period) was fitted for each outcome. All models were restricted to observations with complete data on BF and covariates. In sensitivity analyses, the respiratory category was divided into upper and lower respiratory infections, analyses were adjusted for maternal body mass index (BMI) during early pregnancy and analyses were stratified by year of birth (1998-2003 and 2004-2010).
Maternal BMI was excluded from the original analyses due to a large proportion of missing data (12%). All statistical analyses were performed using Stata V.14 (Stata, 2015. Stata Statistical Software: Release 14).

RESULTS
The Swedish Medical Birth Register included 41 825 term and post-term singletons born to women who resided in Uppsala County between 1998 and 2010. We excluded children with missing data on BF (n=2127) or covariates (n=1873), leaving 37 825 children (90% of the original cohort). In the first follow-up period (early childhood), all children were followed from birth until 2 years of age or censoring due to death (n=10) or emigration (n=194).
In the later follow-up period (later childhood), the remaining 37 621 children were followed from 2 until 5 years of age or censoring due to death (n=14) or emigration (n=573 was reported by 25%. Women over 25 years of age and women with a postsecondary education, were more likely to report EBF for 4 months or more. In contrast, women who reported smoking during pregnancy were more likely to report no BF at 1 week. Figure 1 shows the association between BF categories and overall hospitalisations for infectious disease in early and later childhood. The risk of hospitalisations for infections decreased with duration of EBF until 4 months of age. In comparison with children exclusively breastfed 6 months or more, the highest risk of hospitalisations for infectious diseases in early childhood was found in children who were not breastfed (aIRR 1.89 (95% CI 1.45 to 2.47)) and in children exclusive breastfed <4 months with BF <6 months (aIRR 1.41 (95% CI 1.25 to 1.59)). Similarly, the risk of hospitalisations for infectious diseases in later childhood was highest in children who were not breastfed (aIRR 1.52 (95% CI 1.00 to 2.33), controlled for previous admissions) and in children breastfed <6 months and EBF <4 months (aIRR 1.43 (95% CI 1.16 to 1.75), controlled for previous admissions). The risk of hospitalisations for infectious diseases was comparable in children exclusively breastfed 4 to 5 months and children exclusively breastfed 6 months or more. Online supplemental appendix B includes full regression results including sensitivity analyses. Figure 2 shows associations between BF categories and hospitalisations for respiratory tract and enteric infections in early and later childhood. In early childhood, the risk of both respiratory and enteric infections decreased with duration of EBF until 4 months of age. In comparison with EBF 6 months or more, the strongest association was found between no BF and enteric infections (aIRR 3.32 (95% CI 2.14 to 5.14)). In later childhood, the risk of respiratory infections decreased with duration of EBF until 4 months of age. In comparison with children exclusively breastfed 6 months or more, the highest risk of respiratory infections was observed in children who were not breastfed (aIRR 2.53 (95% CI 1.51 to 4.24), controlled for previous admissions). Online supplemental appendix C includes full regression results including sensitivity analyses.
Sensitivity analyses showed no substantial changes in point estimates after adjusting for pregnancy BMI. Moreover, sensitivity analyses showed similar associations between BF categories and overall hospitalisations for infectious disease in the first (1998-2003) and second (2004-2010) birth cohort. In sex-stratified analyses, associations between BF categories and overall hospitalisations for infectious disease were similar in boys and girls (data not shown).

DISCUSSION
We found that the risk of overall hospitalisations for infectious diseases in early childhood (<2 years) and the risk of hospitalisations for respiratory infections in later childhood (2-4 years) decreased with duration of EBF until Open access There is strong evidence that BF protects infants from infectious diseases. 4 6 7 However, the optimal BF practice is under discussion. Our results are consistent with a Spanish study reporting that full BF until at least 4 months would prevent 56.4% (95% CI 30.9% to 69.4%) of non-perinatal infection in the first year of life. 21 In comparison with no BF, a Dutch study found that EBF until 4 months followed by partially BF decreased the risk of physician-confirmed upper respiratory (adjusted OR (aOR) 0.65 (95% CI 0.51 to 0.83)) lower respiratory (aOR 0.50 (95% CI: 0.32 to 0.79)) and gastrointestinal infections (aOR 0.41 (95% CI 0.26 to 0.64)) during the first 6 months. However, the authors concluded that EBF until 6 months may reduce the risk of infectious diseases even more. 22 In contrast to our findings, a US study reported that full BF for 6 months was, in comparison with full BF for 4 months, associated with a decreased risk of respiratory tract infection during the first 2 years of life. 23 Moreover, observational analysis of the 'Promotion of Breastfeeding Intervention Trial' found that EBF for 6 months was, compared with EBF for 3 months and partial BF thereafter, associated with a decreased risk of gastrointestinal episodes; no significant association was found for gastrointestinal hospitalisations. 7 The Dutch and US studies included less severe infections, whereas our study only included infections requiring hospital care.
There is increasing evidence that BF also reduces the risk of respiratory infections after infancy. The mechanism is still unknown, however, it has been suggested that breastmilk influence the development of the immune system. 8 24 Our findings are similar to those reported in a recent Japanese study. In comparison with no BF, EBF at 6-7 months of age was associated with a reduced risk of hospitalisations for respiratory tract infections (aOR 0.76 (95% CI: 0.58 to 0.99)) between 30 and 42 months of age, whereas no association was found between BF and hospitalisations for diarrhoea. 9 In a Dutch cohort study, BF for at least 6 months was in comparison with no BF associated with a decreased risk lower respiratory tract infections (aOR 0.71 (95% CI: 0.51 to 0.98)) between infancy and 4 years of age. 10 Moreover, a recent US study of children Children with missing data on BF (n=2127) or covariates (n=1873) were excluded leaving 37 825 children. *ICD-10 codes: Q00-Q99. †Maternal education level at year of birth, categorised as 9 years or less, 10-12 years, 13-14 years or 15 years or more. BF, breast feeding; EBF, exclusive BF; LGA, large for gestation age; MO, months; SGA, small for gestation age. Open access aged 6 years, found that any BF for 9 months or more was in comparison with any BF for >0 to <3 months associated with a decreased risk of ear (aOR 0.69 (95% CI 0.48 to 0.98)), throat (aOR 0.68 (95% CI 0.47 to 0.98)), and sinus (aOR 0.47 (95% CI 0.30 to 0.72)) infections, but not with cold/upper respiratory tract infections and pneumonia. 8 Overall, these findings are consistent with our results, that BF reduces the risk of respiratory infections after infancy.
Our study has several strengths. First, the use of highquality health and administrative registers allowed us to adjust analyses for potential confounders including congenital malformation, large/small for gestational age, maternal age, maternal smoking, maternal education level and maternal country of birth. Second, the long follow-up period (until 5 years of age or censoring) enabled us to examine the effect of BF in both early and later childhood. Additionally, this allowed analyses in later childhood to be adjusted for hospitalisations in early childhood. Finally, our study included almost all children (<10% missing data) in one county, thereby reducing the risk of selection bias. However, our study has several weaknesses. In early childhood, infections episodes may influence BF practice leading to reverse causation. In later childhood, confidence intervals were wide due to the small number of events. Therefore, the associations between BF and infections in later childhood need to be interpreted with caution. Information on current BF status was not continuously collected. Compared with studies with complete information on BF practice since birth, for example, collected through daily recordings, our study is likely to overestimate the proportion of exclusively breastfed children. 25 Moreover, Sweden has familyfriendly policies that enable BF including a generous Figure 2 Crude analyses (blue circles), adjusted analyses (red squares) and adjusted analyses controlled for previous admissions (green triangles) of the associations between BF categories, and hospitalisations for respiratory tract and enteric infections, in early childhood (<2 years) and later childhood (2-4 years). Incidence rate ratios are presented on a logarithmic scale. Vertical lines represent 95% CIs around the point estimates. BF, breast feeding; EBF, exclusive BF; MO, months.
Open access parental leave scheme. 13 Consequently, our findings may not be generalised to other populations with less familyfriendly policies and lower BF rates. Finally, due to the observational nature of this study, we cannot rule out the risk of unmeasured or residual confounding.

CONCLUSIONS
Our study found that the risk of overall hospitalisations for infectious diseases in early childhood and the risk of hospitalisations for respiratory infections in later childhood decreased with duration of EBF until 4 months of age. Additionally, the risk of hospitalisations for infectious diseases was comparable in children exclusively breastfed 4-5 months and children exclusively breastfed 6 months or more. Thereby, it supports the current European Society for Paediatric Gastroenterology, Hepatology and Nutrition guidelines, recommending EBF for at least 4 months with exclusive or predominant BF for approximately 6 months. Moreover, it adds to the growing body of evidence suggesting a protective effect of BF on respiratory infections after infancy.