Effect of a participatory intervention in women’s self-help groups for the prevention of chronic suppurative otitis media in their children in Jumla Nepal: a cluster-randomised trial

Chronic suppurative otitis media (CSOM) is a multifactorial disease of poverty. Globally, CSOM affects 65 to 330 million people, of whom at least 50% suffer clinically significant hearing loss [1]. Hearing loss can interfere with early childhood development and decrease educational and social opportunities compounding the existing disadvantage of marginalised children [2]. Therefore, effective strategies for preventing CSOM in low resource settings are urgently needed.

Otitis media is a spectrum of disease beginning with an acute respiratory infection leading to acute otitis media and otitis media with effusion, which can result in a chronic perforation of the tympanic membrane, chronic inflammation of the middle ear cavity and otorrhoea or discharge (CSOM) [1]. Appropriate treatment of acute otitis media with oral antibiotics and early treatment of CSOM with topical antibiotics and ear mopping are simple, inexpensive and effective [3]. The prevalence of CSOM in Nepal is 5 to 10% in the available studies which mostly rely on cross-sectional groups of school children in less remote settings or patients attending ear, nose and throat specialist clinics [4‐7]. Every study that has been conducted in Nepal has revealed a prevalence above the 4% level satisfying the WHO definition of a ‘massive public health problem’ requiring ‘urgent attention’ [1].

Nepal is in the lower third of countries for human development (HDI 0.574, rank 149 out of 189) [8]. However, poverty remains ‘highly asymmetric’ in Nepal with the western regions and the mountains having poorer outcomes on every measure [9, 10]. Jumla is one of the most disadvantaged districts of Nepal, with an HDI of 0.409, a rank of 68 out of 75 districts [11]. CSOM is strongly associated with poverty and its social determinants including: low parental education level, low parental income, malnutrition, overcrowding, lack of access to clean water and sanitation [12, 13].

Until now, there has been little research into prevention of CSOM in low resource settings, leading to calls to have it added to the other 17 neglected tropical diseases [14]. Like the neglected tropical diseases, CSOM disproportionately affects people living in poverty causing significant morbidity, could be amenable to public health intervention and is neglected by research. A range of community based interventions have been successful in improving maternal and child health outcomes in low resource settings [15]. To our knowledge, the effectiveness of a community based educational intervention to improve the ear health of children has not previously been tested in a controlled trial. We hypothesised that in the disadvantaged mountain district of Jumla a community based intervention would improve the knowledge, attitudes and practices of women regarding ear disease and reduce the prevalence of CSOM in their children.

We recruited 30 groups, which comprised 508 women and 937 of their children between Jun 1, 2014 and Jul 31, 2014. Figure 1 presents the cluster and individual participant flow. Follow-up assessment was performed on 449 (88.4%) of the women and 748 (79.8%) of their children.

Table 1 shows the baseline characteristics of the women and children. The mean age of the women was 34.3 (SD 11.3) years, they had 3.4 children (SD 1.6) and more than half (52.8%) of the women had received no education. Intervention and control groups were similar on all measures except for caste. More women in the intervention groups belonged to Dalit caste (115, 42.8%) compared to control group (58, 24.3%). Women in 22 of the 30 clusters belonged to a single type of caste, either Brahman/Chhetri or Dalit, while three groups had one or two other caste members and two other groups had four or five. The completely Brahman/Chhetri groups were equally distributed in the intervention and control groups (eight in the intervention and nine in the control) but the completely Dalit groups were not (six in the intervention and two in the control). Therefore, since randomisation was performed by group, there were more women of Dalit caste in the intervention group than in the control group. Table 2 shows the baseline characteristics of the clusters. The intervention and control clusters were similar on all measures except for caste. Intervention groups had a mean of 7.7 Dalit women per group and control groups had a mean of 4.8 women per group.

We analysed the primary outcome at both the cluster and individual level (see Table 3). The main analysis using GEE and the null model gave non-significant results (mean difference = 0.14, 95% CI − 0.10 to 0.38, P = 0.25), as did the model that included geographical stratification (mean difference = 0.15, 95% CI − 0.09 to 0.38, P = 0.21). VDC 1 consistently had KAP lower scores in this model (mean difference = − 0.78, 95% CI − 1.0 to − 0.55, P < 0.0001). The ICC was 0.14, indicating a large degree of clustering.

Table 4 shows the GEE analysis of the covariates of the primary outcome demonstrating no significant difference between the KAP score in the intervention or control groups (mean difference = 0.14, 95% CI − 0.10 to 0.38, P = 0.27). Women of Dalit caste, who lived in VDC 2, and those with a larger number of children and number in household were all associated with a higher KAP score. Measures of socioeconomic status (amount of land and number of large animals owned) were not associated with the outcome; nor were smoking inside or the level of education reached by the woman or her husband.

The secondary outcome was prevalence of CSOM in the children at 12-month follow-up. Overall, 53 (7.1%) out of 748 children were suffering CSOM at follow-up assessment, 29 (7.4%) in the intervention group and 24 (6.8%) in the control group. There were 37 children with unilateral and 16 children with bilateral CSOM, and the prevalence increased with age. Forty of the children had experienced discharge for more than 12 weeks and 29 had done so for one year or more, and there was no difference in the mean duration of discharge between the intervention and control groups. (mean difference 2.83 weeks, 95% CI 62.52 to 68.19, P = 0.931). Table 5 shows the analysis of the secondary outcome. The unadjusted GEE showed OR 1.10 (95% CI 0.62 to 1.92, P = 0.75). When adjusted for stratification the GEE model produces an OR of 1.12 (95% CI 0.64 to 1.96, P = 0.70). The ICC for the secondary binomial outcome was 0.06.

Next, we analysed the secondary outcome covariates using GEE (see Table 6). The null model includes the variables of geographical stratification VDC and distance to the health post, as well as group type. In the null model there was no difference between the intervention and control groups in the prevalence of CSOM at follow-up, OR 1.12, 95% CI 0.64 to 1.96, P = 0.76 and for the model overall χ² (1, n = 748) = 0.15, P = 0.70. BMI-for-age z-score delivered an OR of 0.52, 95% CI 0.34 to 0.79, P = 0.003, and was the largest predictor in this model. The number of large animals was statistically significant but the odds ratio was very close to 1, which is a very small predictor in this model. Other measures of socioeconomic status, caste, education, smoking or geography did not make a unique contribution to the model.

We then compared the overall baseline and follow-up results. Firstly, for the primary outcome, we compared the mean of individual women’s baseline KAP score with the mean of their follow-up KAP scores using the paired samples t-test. The overall follow-up mean (mean = 7.72, SE = 0.48) was significantly greater than the overall baseline mean (mean = 7.21, SE = 0.08) KAP score (mean difference = − 0.51, 95% CI − 0.71 to − 0.31, t (446) = − 5.07, P < 0.0001). Secondly, we compared the overall baseline and follow-up prevalence of CSOM in the children. There were 106 (11.3%, n = 937) cases of CSOM in the baseline examination and 53 (7.1%, n = 749) in the follow-up examination. Using McNemar’s test to compare two related categorical variables, there was a significant reduction in the overall prevalence of CSOM at follow-up (P < 0.0001).

To our knowledge, this is the first cluster randomised trial to assess a community based intervention to prevent CSOM in a low to middle income country. In Jumla Nepal, health promotion in existing women’s self-help groups did not increase the women’s KAP for their children’s ear health or reduce the prevalence of CSOM in the children. However, there was significant overall improvement in KAP score and reduction in the prevalence of CSOM that was equal in both the intervention and control groups at 12 month followup. Our trial was powered to detect small effects and the intervention was delivered as planned so it is likely that the lack of increase in KAP and reduction of CSOM in the intervention group was a true null effect. Therefore, either our hypothesis that our health promotion would improve KAP and reduce CSOM was flawed or there were other confounding conditions.

Despite the null result from the intervention, there was a significant overall small increase in the women’s KAP score and a large decrease in the children’s prevalence of CSOM, from 11.3 to 7.1%. Although we cannot ascribe causality to this result with certainty the control group did receive an informal intervention by participating in the trial itself. The global research emphasises the difficulty of reducing CSOM, so it is most unlikely that a relative risk reduction of 37% in 1 year is a natural improvement [21]. The control group met the team, answered the survey questions twice in a 12-month period and allowed us to examine their children. When we found a child with any kind of ear disease we spoke to their parents, explained the disease, discussed treatment in detail and either gave them ciprofloxacin eardrops or referred them to the health post for oral antibiotics. There is evidence that just being asked about your behaviour can change it, a phenomenon called ‘mere measurement’, which may have affected women in the control group and this attribution effect has been found in many studies. In addition, the effect of this ‘much better than usual care’ might have been so powerful that it obscured the potential effect of the formal intervention [22].

Our study had important strengths. Our study was set among remote village women and children with a high burden of disease who are under-served by research. Our intervention was embedded in the local community, low-cost and easily reproducible in many contexts. We had a high follow-up rate and consistency in the delivery of the intervention. There was a significant equal increase in KAP scores and reduction in the prevalence CSOM in the both study groups suggesting that this was a genuine effect. Our study also had limitations. Our participants were unable to be blinded as to group allocation and our presence in the community and interest in ears was widely known. Some clusters were very near and could have contaminated the outcomes. We offered treatment to any child who presented with CSOM at any time during the study and referred any with acute otitis media to health services which, although ethically correct, potentially contaminated our findings.

The global research on prevention of CSOM is scanty, despite its morbidity and occasional mortality among disadvantaged people. Like our study, several promising interventions have not been able to demonstrate their effectiveness. CSOM is a complex condition that reflects the interaction of marginalisation, poverty, malnutrition, quality of health services, access to education and the inequity of health research. One reported trial tested the ‘Breathing, blowing, coughing’ exercise to clear mucus at the beginning of the school day which continues to be used in schools in remote Australian Aboriginal communities. Teachers reported ‘less snot’ there was no reduction in CSOM [23]. Similarly, the introduction of community swimming pools was hypothesised to reduce CSOM by passive ear toilet, but studies have shown no effect on the prevalence of CSOM [24]. Zinc supplementation has been unsuccessful [25] and even the screening program for Aboriginal children in New South Wales, Australia, has not provided evidence of a reduction in the prevalence of CSOM [26]. The failure of these studies to reduce the prevalence of CSOM demonstrates the difficulty of research into and management of CSOM. Therefore, our overall relative reduction of 37% is both meaningful and unique.

There are two interventions for the prevention of CSOM which have been successful and both use medication, so are very different to our community based intervention. In Nepal, Vitamin A was given to pre-schoolers for the prevention of blindness, and a sample were followed into adulthood. Schmitz et al. (2012) [27] found that malnourished pre-schoolers with discharging ears who were given Vitamin A had a 42% reduction in hearing loss in adulthood. The mechanism of this effect is not understood and all children in Nepal receive Vitamin A. The second intervention that has been shown to possibly reduce CSOM is pneumococcal immunisation which continues to be evaluated, with current reductions in acute otitis media of 6 to 43% reported from developed nations [28]. In addition, successful clinical treatment programs such as the Earbus in Western Australia report significant reductions in CSOM but need community engagement, skilled staff and intensive follow-up which is difficult to achieve in low resource settings such as Jumla [29].

The baseline prevalence of CSOM in our study (11.3%) was higher than in other studies in children in Nepal [4‐7]. However, there are no other recent studies measuring the prevalence of CSOM in children in similarly remote and disadvantaged places in Nepal. Similar to the Nepal studies, the prevalence of CSOM children in Bangladesh and India is 3 to 6%, while Indigenous children in remote Australia and Greenland have the highest rates of CSOM, 8 to 17% [30]. Therefore, the baseline prevalence in our study population was very high on world standards, probably explained by the degree of poverty and remoteness of Jumla.

Future research might explore adapting our materials to assess maternal knowledge, attitudes and practices at earlier time points to assess whether knowledge degraded over time. Local primary health service providers could be included in the study and usage could also be assessed to triangulate practice data. In addition, since CSOM is a chronic variable condition, longer term followup and a focus on younger children would be useful.

There was a significant improvement in the women’s KAP for ear health and a significant reduction in CSOM in their children at 12 months, equal in both study groups. We were not able to reject the null hypothesis that the intervention based on WHO Primary Ear and Hearing Care Resource [18] would improve outcomes. Our ‘contamination’ of the control group in ethically examining and treating children with CSOM, talking to their parents, stocking the local health post, this ‘much better than usual care’ may have contributed to the overall reduction in prevalence of CSOM, obscuring the effect of the formal intervention [22]. More research is urgently needed in low-resource setting to prevent the life-changing hearing loss of this neglected disease of disadvantage.