Longitudinal analysis of ear infection and hearing impairment: findings from 6-year prospective cohorts of Australian children

Background Middle ear infection is common in childhood. Despite its prevalence, there is little longitudinal evidence about the impact of ear infection, particularly its association to hearing loss. By using 6-year prospective data, we investigate the onset and impact over time of ear infection in Australian children. Methods We analyse 4 waves of the Longitudinal Study of Australian Children (LSAC) survey collected in 2004, 2006, 2008, and 2010. There are two age cohorts in this study (B cohort aged 0/1 to 6/7 years N=4242 and K cohort aged 4/5 to 10/11 years N=4169). Exposure was parent-reported ear infection and outcome was parent-reported hearing problems. We modelled ear infection onset and subsequent impact on hearing using multivariate logistic regressions, reporting Adjusted Odds Ratios (AOR) and Confidence Intervals (95% CI). Separate analyses were reported for indigenous and non-indigenous children. Results Associations of ear infections between waves were found to be very strong both among both indigenous and non-indigenous children in the two cohorts. Reported ear infections at earlier wave were also associated with hearing problems in subsequent wave. For example, reported ear infections at age 4/5 years among the K cohort were found to be predictors of hearing problems at age 8/9 years (AOR 4.0, 95% CI 2.2-7.3 among non-indigenous children and AOR 7.7 95% CI 1.0-59.4 among indigenous children). Number of repeated ear infections during the 6-year follow-up revealed strong dose–response relationships with subsequent hearing problems among non-indigenous children (AORs ranged from 4.4 to 31.7 in the B cohort and 4.4 to 51.0 in the K cohort) but not statistically significant among indigenous children partly due to small sample. Conclusions This study revealed the longitudinal impact of ear infections on hearing problems in both indigenous and non-indigenous children. These findings highlight the need for special attention and follow-up on children with repeated ear infections.


Background
Middle ear infection (otitis media) is a common childhood illness. Estimates indicate that by the age of 3 years at least half of children have experienced at least one episode of otitis [1][2][3]. Because it is a childhood illness, it requires close monitoring of signs and symptoms from parents and practitioners [4][5][6], and it is often co-morbid with other infections of the upper or lower respiratory tract [7,8].
One of the main consequences of otitis media is conductive hearing loss [9][10][11]. Hearing loss may result in speech and language disorders as well as a delay in academic development [12,13]. Subsequent behavioural problems from hearing impairment could also increase the risk of mental health disorders [14]. Otitis media affects children's quality of life, and depending on the severity of infection, may also result in a wider impact on caregivers and health services [15,16].
In a comprehensive report on otitis media in Australia reported by Access Economics, the number of cases in 2008 were estimated to be at least 650,000 [17]. Treatment costs of visits to general practitioners and medical prescriptions were estimated between A$100 to A$400 million [18]. The cost of otitis media is more than $5 billion in the United States [19]. Reports on otitis media in children have shown that indigenous children are much more likely to be diagnosed with severe ear infection and to suffer repeated or multiple episodes than nonindigenous children [20][21][22]. Limited access to medical care, lower socioeconomic status, and remote living conditions all contributed to early childhood hearing loss among indigenous children [17,23].
However there is little longitudinal evidence which identifies the timing of otitis media onset and sequelae related to multiple or repeated infections. This paper examines the epidemiology of ear infection and subsequent hearing problems using unique 6-year prospective longitudinal Australian data in two cohorts of children (0/1 to 6/7 years and 4/5 to 10/11 years) including analyses for both indigenous and non-indigenous children.

Growing up in Australia, the longitudinal study of Australian children (LSAC)
The LSAC aims to identify the social, familial and individual factors influencing children in their early years of development, using a longitudinal study of two child age cohorts [24][25][26]. The study collected data from mothers, fathers, teachers, carers, as well as direct observations of children.
The LSAC used a two-stage clustered sampling design, stratified by state and by metropolitan/urban status. Children were randomly selected using the Medicare database. Detailed information on development of questions, study design, and non-response are reported in the literature [27,28]. Technical and discussion papers related to the project are also available to the public at http://www.aifs.gov.au/growingup. The author has obtained a license to use LSAC data approved by Department of Families, Housing, Community Services, and Indigenous Affairs (FaHCSIA) of the Australian Government.
The   single-parent households, non-English speaking households, and families living in rental properties [28,29].

Study population and statistical analyses
This study uses 4 waves of LSACs data currently available (2004, 2006, 2008, and 2010). There were 5,107 children at the baseline in the Baby or infant cohort (B cohort) at aged 3-19 months and 4,983 children at the baseline in the Kindergarten cohort (K cohort) aged 4 years 3 months to 5 years 7 months. At Wave 4 in 2010, there were 4,242 in the B cohort and 4,169 children in K cohort-83.1% and 83.7% response rate, respectively. Sociodemographic characteristics of the sample available in this study included age and sex of parents and children in both cohorts, indigenous status of the study child, employment and school completion of parent, and remoteness. Socio-Economic Indexes for Areas (SEIFA), developed by the Australian Bureau of Statistics to rank area economic resources was included in the analyses [30].
Generally, the B and K cohort children were comparable on most socio-demographic characteristics, as shown in Table 1. Data on children were primarily collected via interview and questionnaire responses from the parent who knew the child best, mostly the child's biological mother (over 97% for both cohorts). The mean age of the primary caregiver or parent was 31.0 years (SD ±5.5) in the B cohort and 34.7 years (SD ± 5.6) in the K cohort. Slightly more than half of children in both cohorts were males and the majority of children sampled spoke English as the main language at home. There was less than 5% of children who reported indigenous status in each cohort (230 children in the B cohort and 187 children in the K cohort). Parents in the B cohort were slightly better off socio-economically-66.7% reported school completion of year 12 and higher compared to 58.1% among the K cohort. Population and longitudinal sample weights were used in the analyses to correct for minor nonrepresentativeness and attrition. The analyses were also adjusted for household clustering. Children's ear infection (exposure variable) was assessed by parent ("Does child have any of these ongoing problems. . . ear infection (yes, no)?"). It is likely that the reported ear infections include acute otitis media, chronic suppurative otitis media, otitis media with effusion, and uncommon ear diseases. However, duration and severity of ear infections were not reported.
The main outcome variable used in this study was hearing problems as reported by parent ("Does child have any of these ongoing problems. . . hearing problems (yes, no)?"). There was no information on cause, duration, and severity of hearing problems. Ear infections and hearing problems were asked with identical questions between waves. However, there were omitted questions on hearing problems in wave 2 of the B and K cohort and on ear infection in wave 3 of the K cohort.
In view of existing literature on the high prevalence on ear infections among indigenous children, separate analyses and results are reported for non-indigenous and indigenous children [31][32][33]. Confounder variables were sex, employment, and school completion of parents, remoteness, and SEIFA economic resource. Results present ear infections at baseline for all children with data. For subsequent analysis we exclude children with hearing problems at baseline in order to establish longitudinal impacts of ear infection on hearing impairment. Multivariate logistic regressions estimate the adjusted odds ratios and 95% confidence intervals of the associations between reported ear infections and reported hearing problems, all analyses used Stata version 12.

Results
Reported ear infection and hearing problems during the 6-year follow-up were summarised in Table 2 (taking into account that some questions in selected waves had been omitted). Among the B cohort, reported ear infections of indigenous children were more than double compared to non-indigenous children between 0/1 to 4/5 years. Hearing problems were also more commonly reported among indigenous children in both B and K cohorts. Table 3 presented adjusted odds ratios between reported ear infection in preceding wave and subsequent ear infections between wave adjusting for possible confounders in Table 1 and stratified by indigenous status. Among the K cohort, strong associations were found between reported ear infection in preceding and subsequent wave after adjusting for possible confounders: AORs ranged from 6.5 to 17.7 among non-indigenous children and 9.8 to 17.5 among indigenous children. A similar magnitude of association was found among the B cohort but some were not statistically significant.
Longitudinal analyses of repeated ear infection and hearing problems during the 6-year follow up (Table 5) were: among non-indigenous children, comparing to those never reported ear infection, the association between number of repeated ear infections and hearing problems at age 6/7 years increased substantially and statistically significantly in both B and K cohorts (AORs ranged from 4.5 to 31.7 and 4.6 to 51.0, respectively). However, these associations were not statistically significant among indigenous children in both cohorts.

Discussion
This paper examines the epidemiology of ear infection and hearing problems among two cohorts of Australian children. By using 6-year prospective cohort data, we established the longitudinal impacts of ear infection on hearing problems at various ages. We also found that repeated ear infection was substantially and significantly associated with hearing problems at later age.
Our findings support an earlier study on impact of middle ear disease associated with hearing loss [34]. As well, we confirm that the onset of ear infection is associated with an increased likelihood of repeated episodes and these children are at a much higher risk of long-term negative impact [35]. Our findings provide clear evidence that ear infection leads to hearing impairment and the consequences of ear infections may persist throughout early childhood, potentially compromising children's language acquisition, learning ability, and social interactions [36].
Findings on the sequelae of repeated ear infection and subsequent hearing problems were both striking and concerning. A multi-centre cross-sectional study among children in the US reported correlation between frequency of otitis media, worsening overall physical health, and impact on caregivers' time and emotional concerns [15]. Our data on repeated ear infection between waves have shown that follow-up is necessary for children with repeated ear infection episodes if the problems persist [37,38]. The strength of this study is the use of the Longitudinal Study of Australian Children, a nationally representative study that includes a wide array of baseline and repeated exposures and outcomes. A large sample size allows sufficient statistical power to examine multiple episodes of ear infections and longitudinal design allows assessment of impact over time [26]. We acknowledge the possible limitation of the questions and the parentreported nature of self-reported ear infections on behalf of children and could include various symptoms. We also note that the question related to ear infections only allows a 'yes-no' response, which could only capture a maximum of one episode a year. This may underreport the magnitude of ear infections and do not take into account repeated ear infection episodes within the same year. A recent population-based study in Scandinavia has shown self-reported otitis media to be relatively reliable and suggested that any inconsistency in reporting is likely to be associated with less severe episodes [39]. As well, we also acknowledge the possible limitation of parent-reported hearing impairment with severity not known. However, another cross-sectional study based on the Longitudinal Study of Children Survey of cohort aged 4 to 5 years has used this hearing variable and reported association between hearing impairment and language development, educational outcome, and metal health in children [14].

Conclusion
This paper makes a contribution to the limited longitudinal evidence to date on the impact of ear infection on subsequent hearing problems in both indigenous and non-indigenous children. Our findings highlight the need for follow-up services and strategies to minimise substantial, long-term impacts on hearing associated with ear infection in early childhood.