There is unequivocal evidence that universal newborn hearing screening (UNHS) leads to early detection of infants with sensorineural hearing loss well before three months of age,1 2 and thereby facilitates early intervention for improved developmental outcomes.3 4 Screening coverage and overall referral rate for diagnostic evaluation are two key measures of the effectiveness of UNHS programmes.5 6 For instance, the Joint Committee on Infant Hearing recommends that at least 95% of all eligible infants should be covered and overall referral rate should not exceed 4% within the first year of programme implementation.5 Where screening coverage in predominantly hospital-based UNHS programmes is generally satisfactory, referral rates at the early stages may be as high as 12–27%,7 8 resulting in high false-positives and the associated parental stress and high programme costs.9 However, referral rates have been known to decline over time as programme managers gain more experience in their respective settings.7 10 11 Referral rates can also be affected by the choice of screening technology and protocol.12 Other measures of programme effectiveness include but are not limited to the return rate for diagnosis, the age of confirmation of hearing loss and the age of intervention.5

Recognising the substantial contribution of developing countries to the global burden of hearing impairment, the World Health Assembly has recommended early detection and management of childhood hearing impairment among its member states.13 However, in most parts of the developing world, a significant proportion of babies are born outside hospitals thus making conventional hospital-based UNHS programmes of limited value for optimal coverage.14 Well-child clinics for routine childhood immunisation are reputed for attracting babies regardless of their place of birth for a diverse range of health interventions.15 The feasibility of community-based UNHS during routine immunisation clinics has been established but the associated multistage screening protocol, which requires mothers to make repeat visits in order to minimise referral rates, may lead to high drop-out rates.16 Moreover, current evidence on the performance of such UNHS programmes in developing countries is still limited in scope (small sample size) and duration (usually less than one year) largely due to resource constraints.17 18 This study therefore set out to evaluate the performance over a three-year period of a recently concluded pilot community-based UNHS programme and to determine factors associated with the risk of sensorineural hearing loss to guide community-oriented prevention initiatives.

METHODS

Setting

This study was conducted in an inner-city area of Lagos with a population of 243 777. The community is served by one general hospital, one children’s hospital, one specialist maternity hospital and seven primary healthcare centres, all of which are State-owned as well as several private hospitals and traditional maternity homes also known as herbal homes.

Study population

This study is an extension of an earlier pilot project on the feasibility of community-based infant hearing screening in Lagos, which was limited to 10 months.18 In the current study, all mothers attending four of the seven primary healthcare centres that administered routine Bacille Calmette-Guérin (BCG) immunisation from July 2005 to April 2008 were enlisted. Of all routine childhood vaccinations in the developing world, BCG immunisation clinics have the highest uptake with coverage above 90% in 101 countries and below 60% in only nine countries, thus attracting a high proportion of mothers who delivered outside hospitals.19 Uptake for BCG immunisation in this study location is about 75–98%, compared to the national average of 69%.20 The four immunisation clinics chosen accounted for over 75% of BCG vaccinations in this study location and vaccination was administered on separate weekdays, which allowed a single team to cover the four sites. Although BCG immunisation is given shortly after birth, it was not unusual for a few babies to present for vaccination after three months. However, the present screening programme was only offered to infants three months old or younger in order to facilitate comparison with other screening programmes. In addition, older babies were often easily restless and difficult to test in non-hospital based settings.17 Ethical approvals were obtained from the Lagos State Health Management Board, Nigeria and University College London, UK.

Screening protocol

The hearing screening protocol has been previously described.18 In summary, it consisted of a two-stage screening first with transient evoked otoacoustic emissions followed by automated auditory brainstem response for all those who failed the otoacoustic emissions test. The second-stage (automated auditory brainstem response) test was conducted at one of the four weekly clinics. Those who failed auditory brainstem response were referred to a diagnostic centre for further evaluation and appropriate intervention. Due to poor follow-up compliance and resource constraints for effective patient tracking, the present study has been limited to the outcomes of the two-stage screening tests. This two-stage screening protocol typically has a test sensitivity of 92%, specificity of 98% and positive likelihood ratio of 61 for screening newborns in hospital-based settings.21 Those who passed otoacoustic emissions or passed auditory brainstem response were classified as ‘pass’. Those who were referred with auditory brainstem response were classified as ‘fail’. Those who were referred with otoacoustic emissions and did not present for auditory brainstem response test were classified as ‘incomplete’. All services under this programme were offered at no charge to parents.

Main outcome variables

The outcome measures were screening coverage, first-stage referral rates and the second-stage screening results. Maternal and infant factors associated with the screening results were also evaluated.

Independent variables

Prior to screening and after obtaining mother’s written consent, the medical history and sociodemographic profile of each participant was documented with a structured questionnaire. Information of interest in this non-hospital setting was limited to those that could be reliably elicited or derived from parental reports such as maternal age, parity, antenatal clinic attendance for current delivery, place of delivery (hospital or non-hospital), type of birth attendants (skilled or unskilled) and social class. Hospital delivery was either government or private health facility, whereas non-hospital delivery comprised deliveries at family homes, herbal/traditional maternity homes, church-based facilities and, before arrival in hospitals, usually in cars or public transport. Unskilled birth attendants comprise traditional birth attendants, ‘untrained’ auxiliary nurses, neighbours and relations.22 Traditional birth attendants are community-based providers of care during pregnancy, childbirth and the postnatal period without any formal training.22 Social classes were derived from mother’s education and father’s occupation.23 Social class I was termed as ‘high’, II or III as ‘middle’ and IV or V as ‘low’. Infant characteristics included sex, gestational age and gestation type (singleton or multiple). Gestational age was estimated from the first day of the mother’s reported last normal menstrual period. All babies were weighed at the time of screening as the majority of babies were born outside hospitals without birthweight records. History of severe neonatal jaundice in the first week of life necessitating hospital admission for phototherapy and/or for exchange blood transfusion was also documented. Jaundice can be reliably identified by mothers in this population because of the familiar discolouration of the sclera and mucous membranes, as well as the traditional therapies commonly associated with this condition.24 25

Data management and statistical analysis

Data tracking and management software – HI*TRACK for Windows Version 3.5 Desktop (National Centre for Hearing Assessment and Management: NCHAM, Logan, UT, USA) – was used for monitoring the screening programme and tracking the mothers. Data was subsequently exported to SPSS (Windows version 16.0) for statistical analysis (SPSS Inc., Chicago, IL, USA). Distribution of the variables among the participants was examined by cross-tabulation with the outcome variables to provide a general overview. Where provided, mean values for continuous variables are stated with corresponding standard deviations (SD). Unconditional univariable multinomial logistic regression analysis was then performed for each independent variable against the dependent variable to examine the unadjusted association with the three screening outcomes using the Wald statistic. The strength of association was estimated by odds ratios (OR) and the corresponding 95% confidence intervals (CI). Factors with p<0.05 were entered into the multinomial multivariable logistic model to assess the effect of each variable independently on the dependant variable while controlling for the potential confounding effects of covariates. Collinearity and interaction effects were evaluated for the final logistic regression model. Given the wide variation in age at which infants were presented for BCG immunisation from earlier findings,18 and the reported correlation between otoacoustic emissions referral and age at screening,26 27 a survival analysis was conducted with a Kaplan-Meier proportional hazard model in which age at screening was used as the time scale, otoacoustic emissions referral was defined as the event and those who passed were treated as censored. This analysis is appropriate in situations where all individuals would not ultimately experience the event.28

RESULTS

A total of 7179 (96.7%) of the 7423 infants who attended the immunisation clinics were screened during the study period, but four infants abandoned at birth by their mothers and put under the care of social workers were excluded due to the lack of relevant data for this study. A summary of the characteristics of the study participants are presented in table 1.

Mean age of mothers was 27.87±5.13 years. Although the vast majority of mothers belonged to the middle or high social class, over half (51.7%) delivered outside hospital facilities. About 56.6% were attended by skilled health personnel suggesting that a few mothers who delivered outside hospital were still attended by skilled health personnel. Traditional maternity homes accounted for the largest proportion (40%) of all deliveries or 77.5% (or 2873) of all non-hospital deliveries. Very few cases of consanguinity (n = 9) and family history of deafness (n = 11) were reported and all the children with positive histories of consanguinity and family history of deafness, except one and two respectively, passed the screening tests. The vast majority (98.5%) of the infants were born at term (⩾37 weeks) and were predominantly presented for vaccination in their first month (mean age: 16.35±17.30 days) of life. A total of 474 (6.6%) of the infants had a history of neonatal jaundice out of which at least 26.2% required phototherapy and/or exchange blood transfusion.

A total of 749 (10.4%) infants were referred after the first-stage screening but only 355 (47.4%) returned for auditory brainstem response, out of which 128 (35.9%) failed (table 1). The overall failure rate after the second-stage screening was 1.8%. The survival analysis showed that cumulative proportion of otoacoustic emissions failures was significantly higher (p = 0.045) among infants born outside hospitals compared to those born in hospitals (figure 1). The excess risk became apparent in infants who were older than 2 weeks at the time of screening, while the use of auditory brainstem response for second-stage screening consistently improved the initial otoacoustic emissions referral rates throughout the duration of the programme (figure 2). Otoacoustic emissions referral dropped from its peak of 30.9% in the first month to an average of 15.6% within the first year of programme initiation.

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

Cumulative proportion of infants born outside hospitals who failed the first-stage screening compared with infants born in hospitals based on the age at the time of screening. Note: A higher proportion of infants born outside hospitals failed otoacoustic emissions screening as from the age of about 15 days.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Trend in monthly referral rates of first-stage and second-stage screening outcomes over programme duration. Note: The introduction of second-stage screening with auditory brainstem response resulted in lower monthly referral rates consistently throughout the duration of the programme.

In the univariable analyses, maternal age, social class, parity were the only maternal factors associated with the two-stage screening outcomes, while age at screening, gestational age and hyperbilirubinaemia requiring exchange blood transfusion were the only infant factors associated with screening outcomes. The adjusted odds ratios for these factors after the multivariable multinomial regression analysis are presented in table 2.

When compared to infants who passed the two-stage screening, infants who failed were significantly more likely to belong to mothers in the low social class (OR 7.85; 95% CI 1.88 to 32.81), older than a month at the time of screening (OR 5.07; 95% CI 3.49 to 7.35), and not likely to be singletons (OR 2.64; 95% CI 1.36 to 5.12), but were likely to have a history of neonatal jaundice requiring exchange blood transfusion (OR 7.70; 95% CI 3.82 to 15.53). A similar pattern was observed when infants who failed were compared with those who did not complete the screening programme. In contrast, infants who did not complete the screening programme compared with those who passed, were significantly less likely to be first child (OR 1.33; 95% CI 1.06 to1.67), were more likely to be older than one month (OR 2.44; 95% CI 1.92 to 3.10) and were also more likely to have a history of neonatal jaundice with exchange blood transfusion (OR 2.53; 95% CI 1.40 to 4.56). In fact, infants with a history of neonatal jaundice requiring exchange blood transfusion had more than a twofold risk of screen failure compared to children who either passed or dropped out of the programme. A similar pattern was evident among older (above one month of age) infants. There was no evidence of collinearity or significant interactions among the variables in the final model.

DISCUSSION

Consistent with preliminary findings in the first 10 months of this pilot project, the majority of births in this inner-city community in the heart of the most urbanised town in Nigeria occurred outside hospitals.18 This finding clearly suggests that national estimates of countries like Nigeria showing predominant non-facility based deliveries may not be restricted to urban–rural dichotomy and underscores the need for policy adaptation in maternal and child health care for the urban population well served by several public and private health facilities. It is also noteworthy that the use of non-facility-based services was irrespective of social class or attendance at antenatal clinics, which recorded excellent uptake.

The screening coverage of 97% was satisfactory by international standards and none of the eligible mothers declined participation throughout the study duration.5 Although it was not possible to determine the proportion of eligible infants from the total population that were missed at the four centres used for this study, it is not expected that this figure will be substantial given the high uptake of BCG immunisation in this community. This finding supports the view that community-based UNHS programme at BCG immunisation clinics provided an effective platform for reaching the vast majority of babies who would have otherwise been missed by conventional hospital-based UNHS programme. Community-based UNHS also provided a cost-effective method of screening babies born in several hospitals where individual UNHS programmes would have been contemplated. The average age at presentation at the clinics also suggests that early intervention within the first year of life was achievable even with a few months delay in diagnostic evaluation. Recent evidence from the UK has also demonstrated the effectiveness of community-based UNHS programme by health visitors for developed countries.29

Although the otoacoustic emissions referral rate declined steadily from an average of 15.6% in the first year to 6.3% in the third year, this level was still above the recommended 4% threshold. The inclusion of a second-stage screening with auditory brainstem response substantially reduced the annual refer rate to a satisfactory range of 3.4% in the first year to 0.53% in the third year. Auditory brainstem response is a more expensive screening technology than otoacoustic emissions and programme managers would need to weigh the trade-off in direct operational costs and the indirect costs of high referral rates along with the associated false-positives. The need for auditory brainstem response should be considered against the backdrop of the increased risk of otoacoustic emissions referrals in a population with predominant non-hospital deliveries. A fuller discussion on the cost implications of alternative screening strategies has been reported elsewhere.30

Given the reported efficiency of the two-stage screening protocol, the referral rate of 1.8% is likely to translate to a prevalence rate of hearing impairment far in excess of the two to four per 1000 reported in developed countries,31 and comparable to rates from other developing countries.16 32 The higher risk of failure among older infants may in fact be indicative of a higher incidence of the hearing impairment usually associated with acquired causes during the postnatal period, such as infections in early childhood, ototoxic medications and severe neonatal jaundice, which would have been missed by conventional pre-discharge hospital-based UNHS. It is therefore not altogether surprising that a country like Nigeria is not only in the top league of countries with the highest neonatal and child mortality rates globally,14 33 but also has the highest proportion of developmentally disadvantaged children worldwide.34

The increased risk of failure associated with severe jaundice is consistent with previous findings in this population and other developing countries.32 35 36 It was found that all the 10 infants who returned for auditory brainstem response out of the 23 with severe jaundice who were referred with otoacoustic emissions actually failed the test. This would suggest that a high proportion of the 13 infants with severe neonatal jaundice who dropped out of the programme were likely to fail if they were tested with auditory brainstem response. Jaundice is the most common condition requiring evaluation and hospital admission in the first week of life and constitutes a significant cause of neonatal mortality and morbidity. The prevention of severe jaundice would necessitate improved maternal education particularly during antenatal clinics on the avoidance of haemolytic agents, recognition and timely response to early danger signs as well as prompt treatment with phototherapy and exchange blood transfusion.

Socioeconomic deprivation has been associated with childhood hearing impairment even in developed countries, which may explain the increased risk of screen failure among infants whose mothers are in the low social class.37 38 This finding is also in agreement with the well-established association between socioeconomic disadvantage and developmental status in early childhood.39 However, the relatively wide confidence intervals for the odds ratios call for a cautious interpretation of the related risk level in this population. The increased risk associated with multiple gestation accords with several studies linking this factor with increased neurodevelopmental disabilities.40 Multiparous mothers are more likely to drop out of the programme if they have never had children with developmental problems and this group of mothers needs to be educated on the risk and burden of late detection of hearing impairment in an apparently well-child. Although most of these factors are less modifiable through primary prevention compared with severe jaundice, it is however, important to alert mothers on these risk factors and encourage them to have their babies tested promptly to facilitate early intervention.41 42

Some limitations of this population-based study are worth noting to guide those who may wish to implement a similar programme in a developing country. Although uptake of BCG immunisation is generally high even among Nigerian mothers living abroad,43 nevertheless, the use of this platform excluded mothers without surviving newborns and those who did not seek immunisation for their children (though only a small proportion of live births may be so affected) compared to a house-to-house survey. Diagnostic outcomes for the vast majority of the babies were not available due to the high drop-out rate, a factor largely attributed to resource constraints for effective community-oriented follow-up services. Although the two-stage screening protocol has high sensitivity and predictive values, without diagnostic evaluation it was not possible to establish the true prevalence and pattern of hearing impairment or evaluate the screening programme by conventional efficiency measures.44 In fact, considering that the incidence of false-positives is likely to be lower in older infants under a community-based programme compared with screening before discharge under a hospital-based programme, the efficiency sensitivity and predictive values are unlikely to be worse than currently reported in studies done in developed countries.21 Nonetheless, the factors that have been associated with screening failure in this study should simply be regarded as indicative of the ‘risk of sensorineural hearing loss’ until their relationship with the hearing status of the infants based on diagnostic test results have been established. Thus, the present experience underscores the need for a better tracking system and maternal education for infants who failed the screening tests to enhance the overall effectiveness of the programme. This may require a dedicated team outside the screening staff to ensure better results. Additionally, some infants with auditory neuropathy/dyssynchrony (a type of hearing impairment in which normal outer ear cell function of the cochlea co-exists with abnormal or dyssynchronous auditory brainstem response) often associated with severe jaundice may have been missed by exiting infants who passed otoacoustic emissions.5 For example, the Joint Committee of Infant Hearing recommends that all graduates of neonatal intensive care unit should be tested with auditory brainstem response regardless of their otoacoustic emissions status, which would have been applicable to infants with severe jaundice in this population.5 Thus, the failure rate for the present two-stage screening may have been understated by such cases of false-negatives.

Conclusions

In summary, it was found that the majority of babies in this inner-city community well served by several health facilities were born outside hospitals. Community-based UNHS showed consistent improvement in key outcome measures of performance by international standards and BCG immunisation clinics provided an effective platform for screening the substantial proportion of infants born outside hospitals who would have been missed by conventional hospital-based UNHS programmes. Low social class, age at screening, multiple gestation and severe jaundice were significantly correlated with the risk of hearing loss indexed by failure of two-stage screening with otoacoustic emissions and auditory brainstem response. Multiparous mothers as well as infants older than one month at screening and those with severe jaundice were mostly likely to drop out of the screening programme. Community-oriented interventions to address the burden of this condition in a low-income country through appropriate policy initiatives are needed.