Community case management of lower chest indrawing pneumonia with oral amoxicillin in children in Kenya

To determine the accuracy and effectiveness of community health workers (CHWs) when compared to trained nurses for management of pneumonia in Kenyan children.


INTRODUCTION
Despite the availability of low-cost treatment, pneumonia continues to be a major killer of children under five years of age, accounting for nearly 15% of child deaths globally; a large number are in sub-Saharan Africa (1,2). One major contributor to this situation is the lack of access (including distance and cost) to prompt and effective treatment (3)(4)(5). The World Health Organization (WHO) and the United Nations Children's Fund (UNICEF) recommend that where access to health facilities is not feasible, community health workers (CHWs) treat pneumonia through integrated community case management (iCCM) (6)(7)(8). iCCM of pneumonia has been shown to be a feasible and effective strategy in Asia, where up to 35% of pneumonia-related mortality in children can be prevented (9,10).
World Health Organization recommends treatment of lower chest indrawing pneumonia with oral amoxicillin on an outpatient basis (11). Evidence for this recommendation was provided from hospital and outpatient settings (12)(13)(14). In Kenya, just half of caregivers seek help for their children with suspected pneumonia from an appropriate healthcare worker, and of these, less than 50% receive antibiotics for pneumonia (15).
In a Kenya Ministry of Health-and WHO-supported exercise in 2011, Kenyan experts (academics, clinicians and policy-makers) generated and systematically scored research questions for maternal, newborn and child health interventions to reduce child mortality in Kenya using Child Health and Nutrition Research Initiative methodology. During this exercise, iCCM of pneumonia was identified as a top priority for research in Kenya. iCCM of lower chest indrawing pneumonia using oral amoxicillin has been successfully demonstrated in Asia (16,17), but no such data exist for Africa (18). WHO and UNICEF currently do not recommend treatment of lower chest indrawing pneumonia with oral amoxicillin by CHWs as part of iCCM, as no evidence is available from Africa. To address this gap, we trained CHWs in rural western Kenya to manage lower chest indrawing pneumonia using oral amoxicillin as part of iCCM to increase access to pneumonia treatment. We conducted this implementation research in the existing county government health system.

Study design
This observational intervention study was conducted in Homabay County in western Kenya between January 2014 and March 2015. Our primary objective was to determine the accuracy and effectiveness of CHW classification, treatment and treatment outcomes of lower chest indrawing pneumonia when compared to trained nurses.

Study setting
Homabay is a rural county with a population of approximately one million and has eight administrative subcounties: Homabay, Rangwe, Ndhiwa, Mbita, Suba, Kabondo, Kasipul and Rachuonyo North. Children under five years account for 16% of the population. The county is characterised by high levels of poverty and the highest HIV prevalence in Kenya (19), as well as one of the highest under-5 mortality rates in the entire country (20). Despite an adequate number of health facilities in Homabay, almost half of the population finds distance and poor road access as serious barriers to health care. Those living on islands in Lake Victoria must travel very long distances over water and land to reach health facilities.

Study participants
Children aged 2-59 months with a cough and/or difficulty breathing during regular home visits by CHWs, or whose caregiver sought care from a CHW, were assessed, classified and treated according to the WHO algorithm for pneumonia for health facilities (Box 1) (21). The CHW recorded the findings on a WHO standardised sick child recording form.

Study procedures Training
The study procedures were implemented through alreadyexisting CHWs in 241 community health units linked to existing health facilities in all of Homabay County. The CHWs had been previously selected and trained over six weeks using the national community health strategy curriculum to provide a wide variety of preventive and promotive health services. Ministry of Health Community Health Extension Workers (CHEWS) supervised the CHWs. CHWs were trained in iCCM for six days to enable them to identify, classify and treat pneumonia, malaria and diarrhoea in children 2-59 months of age. The CHWs gained skills that allowed them to counsel caregivers, manage stocks of drugs and supplies, and accurately complete relevant registers and data collection forms. CHEWs were trained on how to use these forms in supervision. The national Kenyan iCCM training programme, adapted from WHO/UNICEF materials, was used. A team of CHW trainers of trainers experienced in iCCM and integrated management of childhood illness (IMCI) conducted the training. The methodologies used included lectures, presentations, discussions, case studies, clinical demonstrations, practice, role plays, video shows and facility-based clinical attachments. Pre-and posttraining assessments were conducted. In order to support, supervise and evaluate the CHWs during the implementation of the study, diploma holding nurses in the study area underwent training in the WHO adapted Kenya IMCI curriculum as well as the iCCM curriculum. The training was performed at the link health facilities and was facilitated by national, county and sub-county IMCI trainers.

Classification and management
Children were eligible for inclusion in the study if they were 2-59 months of age and presented with fast breathing and/or lower chest indrawing pneumonia. Children who had WHO-defined severe pneumonia (with any danger sign) were not enrolled. For a child classified with chest indrawing pneumonia, the CHW treated the child with oral amoxicillin (90 mg/kg per day in two divided doses) for five days. The CHW educated the caregiver on danger signs and advised the caregiver to take the child to the nearest health facility if any danger sign became apparent or if the child was unable to take the oral antibiotic. For children with severe pneumonia (with danger signs), the CHW referred the patient using a written referral slip to the nearest health facility for further management, and where possible accompanied the patient. If the child was able to take the oral antibiotic, a prereferral dose of oral amoxicillin was given. Children were followed up by the CHWs either in the patient's home or at the CHW's house on day 3 (after 48 hours) and day 6 as per national iCCM guidelines. At each visit, caregiver self-reported adherence to the oral antibiotic was assessed, as well as the child's clinical status and the presence of any danger sign and adverse events. Malnutrition was assessed using a colour-coded mid upper arm circumference (MUAC) strip. Normal nutrition status was coded as green, moderate malnutrition as yellow and severe malnutrition as red. A child was regarded as lost to follow-up if contact could not be made after day 3 of initiation of treatment.

Data collection
As this was the first time CHWs were involved in curative services, each case of pneumonia needed to be tracked and verified within 24 hours. We created a free 24/7 decision support hotline to give assistance to the CHWs treating pneumonia. The hotline was a specialised telephone service that provided an effective way to communicate with and link the pneumonia case to a trained nurse for physical case confirmation. The hotline utilised the existing widespread cellular phone coverage linked to a web-based platform, which had a preprogrammed diagnostic algorithm for pneumonia assessment, classification and treatment based on the WHO sick child recording form. Two attendants staffed the hotline. Whenever a CHW identified and treated a case of pneumonia, she/he sent a 'Please Call Me' message to a study hotline number. The attendant called the CHW to record what the CHW had diagnosed in an online study database. The online system then compared the CHW's classification against the system algorithm and regardless of the diagnosis, sent an automatic SMS with the child's key location details to the nearest nurse for confirmation. The nurse traced the child within 24 hours and independently assessed and classified the child. After determining eligibility, the nurse obtained written consent from the caregiver to conduct subsequent follow-ups and document treatment outcomes. The nurses recorded their pneumonia classification through a mobile phone application, and their data were synchronised into the central online database. The nurse classification was used as the gold standard to measure the accuracy of classification and appropriateness of treatment administered by the CHW. Children who were not eligible for enrolment were provided appropriate care, including facilitated referral to hospitals as needed.
The study nurse subsequently followed up enrolled children on days 4 and 14 to assess for treatment outcomes (Box 2), and used a standardised questionnaire to gather information on follow-up visits by CHWs and assess caregivers' perceptions on the quality of care provided. If the nurse suspected treatment failure or relapse during the follow-up visit, he/she referred the child to the nearest health facility. Adverse drug reactions were also monitored during these visits.

Sample size
Sample size was calculated to assess the primary outcome of treatment failure at day 4 of treatment. We assumed (i) no seasonality of pneumonia incidence; (ii) average duration of pneumonia symptoms of one week; (iii) estimated pneumonia incidence of 300 cases per 1000 children per year in low-and middle-income countries, which translated to a two-week period prevalence of approximately 2% (22,23); (iv) at least 20% of children with pneumonia have lower chest indrawing pneumonia; and (v) approximately 10% of children suffer treatment failure (24). A precision of 0•5% around the treatment failure estimate and 95% confidence interval (CI) was used. The level of implementation was at the sub-locations (clusters) in the community, so necessary adjustments in were made using an intracluster correlation of 0.16, resulting in a sample size of 1673. A factor of 15% loss to follow-up yielded a total sample size of 1924.

Exposure and outcomes
The primary outcomes were (i) concordance between CHWs and nurses in identification and classification of lower chest indrawing pneumonia; and (ii) treatment failure at day 4. The explanatory variables included child's age, sex, comorbidity, moderate or severe malnutrition and CHW characteristics.

Statistical analysis
Frequencies with percentages were used to summarise categorical variables. Means and medians reporting the respective standard deviations and interquartile ranges (IQR) were used to summarise continuous variables. The comparison of pre-and postintervention outcomes was carried out using cluster-adjusted t-tests for continuous and chi-square tests for categorical variables. Univariate, bivariate and multivariate logistic regression models were adjusted for clustering. Variables significant at p < 0.2 in the bivariate models were included in the multivariate (adjusted) model. Proportions, coefficients or odds ratios were reported as appropriate with respective 95% CIs. Sampling weights were incorporated in all the survey data analysis as appropriate to account for the survey sampling strategy. Cohen's kappa statistic (k) test was utilised to measure the level of agreement between the CHWs and nurses (25). Analysis was carried out using Stata version 13.1 College Station, TX: StataCorp LP.

Ethical clearance
The study protocol was reviewed and approved by the KEMRI National Ethical Review Committee, Kenya National Pharmacy and Poisons Board and WHO. All serious adverse events were reported to the KEMRI Ethical Review Committee and WHO. Community leaders and village headmen were approached through village health committees, briefed on the project, and their verbal consent and approval solicited. Written consent in the preferred local language (Dholuo or Swahili) was obtained from caregivers of children who were followed up after receiving treatment from the CHWs to allow collection of data to assess the effectiveness of the intervention.

Quality assurance
Community health workers underwent an accreditation and standardisation process 8-10 weeks after the first training. This process included evaluation of CHW performance and standardisation of assessment, classification and treatment. Monitoring of adherence to the protocols, including human subject's protection, was conducted by a designated KEMRI ethics officer not directly involved in the study as well as by an independent WHO monitor experienced in performing such work in low-resource settings.

Main outcomes
Concordance in identification and classification of lower chest indrawing between CHWs and nurses Comparison between CHWs and nurses for concordance in identification, classification and treatment of lower chest indrawing showed an agreement of 88.7% with a Kappa coefficient of 0.65 (95% CI: 0.57-0.62; exact p-value <0.001).

Factors associated with treatment failure at day 4 and day 14
The multivariate model did not identify any factors associated with treatment failure on day 4 (  (Table 3).
We documented five deaths in the course of the study. No serious adverse events related to amoxicillin were identified.
Factors associated with incorrect chest indrawing classification by CHW Children with moderate malnutrition (OR 1.68; 95% CI: 1.22-2.30), with comorbidities such as diarrhoea or malaria (OR 1.55; 95% CI: 1.32-1.81) or those with an additional day of delay in presenting to the CHW (OR 1.06, 95% CI: 1.02-1.10) were more likely to have an incorrect classification of lower chest indrawing by the CHW.

DISCUSSION
Our study demonstrates that CHWs were able to identify, classify and treat lower chest indrawing pneumonia in children aged 2-59 months with a five-day course of oral amoxicillin with low treatment failure rates in a real programme setting in Kenya. Our observed treatment failure rate was lower (2.1%) than the 9% reported from Haripur, Pakistan, and 8% from Matiari, Pakistan, the only two community trials that have reported iCCM for lower chest indrawing pneumonia (12,13,16,17). Our loss to follow-up was comparable to the 5.3% reported from Matiari and 2.4% from Haripur (12,13,16,17). Taking a very conservative view, even if we were to combine the 5.6% loss   Critics of iCCM for pneumonia have noted that CHWs may not be able to correctly identify chest indrawing (26). In our study, the concordance for classification of lower chest indrawing between CHWs and nurses showed 88.7% agreement (k = 0•595), which was comparable to the 93•7% in Haripur (16). We can attribute this success to the close supervision of the CHWs as well as to the periodic accreditation and standardisation of skills. CHWs who had more experience (seen more than five children) were better at identifying and classifying lower chest indrawing. As in other higher-level trained healthcare workers, clinical competence is often a function of selfconfidence and self-efficacy (27). It is likely that CHWs who had seen five children or fewer with pneumonia lacked experience, which might have led to missed or incorrect classification of lower chest indrawing. Trainers and supervisors need to identify strategies to help CHWs build self-confidence as well as motivation. Such strategies may include supervisory mentoring, standardisation exercises for clinical skills and accreditation, and sessions in which CHWs share experiences with each other.On exploratory analysis we found that CHWs were more likely to misclassify lower chest indrawing in children with comorbidities such as malaria and diarrhoea, those with moderate malnutrition or those whose caregivers delayed care seeking. This subset of children likely represents sicker children for whom extra care is needed in disease assessment and classification. This challenge in classification of sicker children has also been reported among higher-level trained healthcare workers (28,29). This finding has implications for programmatic implementation and scale-up of iCCM. CHWs should be able to identify and refer children with any danger sign and those with malnutrition. Training of CHW needs to emphasise prompt identification and referral to nurses and other trained health care workers. Additionally, caregivers need to be educated and motivated to seek appropriate care early.
Our study had several strengths. First was its research design, its innovative use of mobile phone technology and implementation in a rural high mortality context, with its potential to be scaled up and catalyse a policy change in Kenya. Second, the study was integrated into the existing community health delivery structures and used mobile phones already owned by CHWs to obtain real-time notification of all pneumonia cases in the county and to trigger case verification within 24 hours. Finally, Ministry of Health staff and administrators developed the implementation, monitoring and evaluation framework, trained CHWs and their supervisors and supervised implementation, thus increasing the generalisability and sustainability of our findings to other parts of the country.The inherent limitation of this study was that although the nurses were only provided with the name and location of the child, it is still possible that the nurse was able to deduce the CHW assessment and classification based on the caregiver report about treatment. To avoid this bias, one could have had CHWs send a 'Please call me' message for all children regardless of their classification. However, this would have increased the number of children requiring follow-up and the number of nurses to consult, which would be challenging in a setting of limited human resources. The second limitation was that no laboratory or radiological investigation was undertaken for the enrolled children. It is likely that this criteria selected children with respiratory signs of those children with probable pneumonia who would benefit from antibiotics as well as those without probable pneumonia who would not benefit from antibiotics. However, this iCCM approach to classification and management of pneumonia is similar to the current standard of care at most primary health care facilities where classification of pneumonia is algorithm based. In many of these primary health care facilities, both laboratory and radiological services are not available. Moreover, even when available, radiological confirmation of pneumonia is difficult and only 5-20% of cases have bacteraemia. In addition, viral pathogens, which preclude use of antibiotics, are probable causes of