Costs and cost-effectiveness of management of possible serious bacterial infections in young infants in outpatient settings when referral to a hospital was not possible: Results from randomized trials in Africa

Introduction Serious bacterial neonatal infections are a major cause of global neonatal mortality. While hospitalized treatment is recommended, families cannot access inpatient treatment in low resource settings. Two parallel randomized control trials were conducted at five sites in three countries (Democratic Republic of Congo, Kenya, and Nigeria) to compare the effectiveness of treatment with experimental regimens requiring fewer injections with a reference regimen A (injection gentamicin plus injection procaine penicillin both once daily for 7 days) on the outpatient basis provided to young infants (0–59 days) with signs of possible serious bacterial infection (PSBI) when the referral was not feasible. Costs were estimated to quantify the financial implications of scaleup, and cost-effectiveness of these regimens. Methods Direct economic costs (including personnel, drugs and consumable costs) were estimated for identification, prenatal and postnatal visits, assessment, classification, treatment and follow-up. Data on time spent by providers on each activity was collected from 83% of providers. Indirect marginal financial costs were estimated for non-consumables/capital, training, transport, communication, administration and supervision by considering only a share of the total research and health system costs considered important for the program. Total economic costs (direct plus indirect) per young infant treated were estimated based on 39% of young infants enrolled in the trial during 2012 and the number of days each treated during one year. The incremental cost-effectiveness ratio was calculated using treatment failure after one week as the outcome indicator. Experimental regimens were compared to the reference regimen and pairwise comparisons were also made. Results The average costs of treating a young infant with clinical severe infection (a sub-category of PSBI) in 2012 was lowest with regimen D (injection gentamicin once daily for 2 days plus oral amoxicillin twice daily for 7 days) at US$ 20.9 (95% CI US$ 16.4–25.3) or US$ 32.5 (2018 prices). While all experimental regimens B (injection gentamicin once daily plus oral amoxicillin twice daily, both for 7 days), regimen C (once daily of injection gentamicin injection plus injection procaine penicillin for 2 days, thereafter oral amoxicillin twice daily for 5 days) and regimen D were found to be more cost-effective as compared with the reference regimen A; pairwise comparison showed regimen D was more cost-effective than B or C. For fast breathing, the average cost of treatment with regimen E (oral amoxicillin twice daily for 7 days) at US$ 18.3 (95% CI US$ 13.4–23.3) or US$ 29.0 (2018 prices) was more cost-effective than regimen A. Indirect costs were 32% of the total treatment costs. Conclusion Scaling up of outpatient treatment for PSBI when the referral is not feasible with fewer injections and oral antibiotics is cost-effective for young infants and can lead to increased access to treatment resulting in potential reductions in neonatal mortality. Clinical trial registration The trial was registered with Australian New Zealand Clinical Trials Registry under ID ACTRN 12610000286044.

General guidance is provided below.
Consult the submission guidelines for detailed instructions. Make sure that all The study was approved by the institutional ethics committees of each participating institution and the WHO Ethics Review Committee (Protocol ID NCH09008). Written informed consent was obtained from caregivers for each activity.
information entered here is included in the Methods section of the manuscript.

Introduction
Of the 5.3 million estimated child deaths in 2018, 2.5 million occurred in the neonatal period [1], with infections, intrapartum complications and preterm birth accounting for most. In 2016, infections including pneumonia, sepsis and meningitis were responsible for nearly 550 000 (21%) neonatal deaths, nearly all of them occurring in developing countries [2]. The World Health Organization (WHO) recommends postnatal home visits to increase coverage of care and improve newborn survival through identifying sick young infants, promoting appropriate care-seeking and early management. The proportion of deliveries at health facilities is increasing, but in many settings mothers and newborns are discharged within a few hours of birth and have little contact with a health provider until the 6-week postpartum and immunization visit. Births at home pose an even greater challenge for providing care during the first critical hours and days [3][4].
Based on a conservative estimate of 355 500-605 750 annual cases and 177 500-302 870 annual deaths due to neonatal sepsis in sub-Saharan Africa , 5.29-8.73 million disability-adjusted life years (DALYs) are lost each year, leading to an annual economic burden ranging from US$ 10 billion to US$ 469 billion. [5] WHO recommends referral to hospital and injectable therapy for management of PSBI in neonates (0-28 days old) and young infants (0-59 days old) [6][7].
However, in resource-limited settings 60-80% of the families of young infants with signs of severe infection do not accept referral to hospital due to reasons such as distance to the health facility, cost of hospitalization and cultural constraints, resulting in newborn deaths [8][9][10][11][12][13][14][15]. Four randomized controlled trials were conducted in five countries in Africa and Asia to evaluate the safety and effectiveness of simplified antibiotic regimens that could be given on an outpatient basis for treating PSBI in young infants when referral is not feasible [16][17][18][19]. In the two Asian trials in Bangladesh and Pakistan two simpler alternative antibiotic regimens were compared against a reference regimen, whereas in the African multicentre trial in five sites in three countries (DRC, Kenya and Nigeria), two additional regimens were evaluated (Box 1). These trials showed that simplified antibiotic regimens requiring fewer injections were equivalent in treatment outcomes to the reference regimen for young infants with signs of CSI and fast breathing without signs of critical illness when referral was not feasible [13][14][15][16]. The implementation of these findings will increase access to treatment for young infants with PSBI [20]. Limited evidence is available for the costs of PSBI and pneumonia management from lowand middle-income countries (LMICs) [21,22].
We report costs and cost-effectiveness analysis for outpatient treatment of PSBI with simplified antibiotic regimens in three African countries when referral is not feasible.

Methods
AFRINEST was conducted in a total of five sites in DRC (North and South Ubangi), Kenya (Eldoret) and Nigeria (Ile Ife, Ibadan and Zaria) to treat PSBI in neonates and young infants on an outpatient basis when referral was not feasible [13][14]. Five simplified antibiotic regimens were used (see Box 1). Pregnancy and birth surveillance and home visits were conducted. These sites were mainly rural, with some semi-urban and peri-urban areas. The study was conducted from April 2011 to June 2013, with the costs estimated for 2012. Details of the trial methodology have been reported elsewhere [23][24].

Box 1. Description of antibiotic regimens
Reference treatment: 1. Treatment regimen A The reference group received a gentamicin injection once daily and a procaine penicillin injection once daily for 7 days (14 injections in total) (as used in the AFRINEST and Simplified Antibiotic Therapy Trial [SATT] studies) [13][14][15][16].

Experimental treatments (intervention):
CSI: a young infant from 0-59 days of age presenting with any of these signs: severe chest indrawing, body temperature ≥ 38.0 °C or < 35.5 °C, stopped feeding well, or movement only when stimulated.
4. Treatment regimen D: gentamicin injection once daily and oral amoxicillin twice daily for 2 days, thereafter oral amoxicillin twice daily for 5 days (2 injections in total) (as used only in the AFRINEST study) [13].
Fast breathing pneumonia: A young infant from 0-59 days of age presenting with respiratory rate of 60 breaths or more per minute.
5. Treatment regimen E: oral amoxicillin twice daily for 7 days (as used only in the AFRINEST study) [14].
The direct costs of drugs, supplies and medical staff time for different types of services and by different providers were calculated for each mother-child dyad served. Indirect and administrative costs, such as for supervision, training, quality assurance, monitoring and evaluation, non-consumables and operational expenses did not change for every additional child served [25][26]. Some of the costs were considered as research and not included for costing, such as: development of material; initial training of staff; additional equipment and infrastructure; staff hired to monitor the effectiveness of the antibiotic regimens; baseline surveys; workshops, communication equipment and vehicles purchased. The cost of existing capacity, except for health workers' time, were also not included, but activities needed on a recurrent basis, such as refresher training, administration, monitoring meetings, communications and travel for screening were included. There were also non-consumable marginal costs, such as those for weighing scales, thermometers and timers for screening and diagnosis.
The effectiveness indicator was calculated using the risk difference in treatment failure and was estimated as the percentage of newborns who did not fail treatment after one week of enrollment with each of the regimens. Treatment failure was defined as any of the following: death; clinical deterioration or admission to hospital at any time after enrollment; persistence of fast breathing on day 4 or recurrence after day 4 up to day 8; and development of a serious adverse event related to the study antibiotics, such as organ failure, anaphylactic reaction, severe diarrhoea, or severe rash [13][14].

Ethical approvals
The study was approved by the institutional ethics committees of each participating institution and the WHO Ethics Review Committee (Protocol ID NCH09008). Written informed consent was obtained from caregivers for each activity.

Interventions/major activities and sub-activities
The interventions for which costs were derived were only linked to health services delivery in community and outpatient settings. The interventions were classified into four main categories and described below.
1. Home-based care ( Outpatient services by a nurse for sick young infants, such as administering injectables or assessment of non-response to treatment, was provided at government clinics in the DRC and Kenya. In about 10-15% of cases, a nurse made a home visit and administered the indicated injections. CHWs supervised administration of the first dose of oral amoxicillin every day at the home of the infant, while the second dose was given by the parent. In Nigeria, CHEWs initially identified sick young infants in the community, and referred them to the nurse at the clinic for assessment, enrollment, randomization and provision of the first injectable dose of treatment. Thereafter, the CHEW administered the first dose of oral amoxicillin daily and provided injectable therapy at the home of the child. The second dose of oral amoxicillin was given by the parent.

Data
Data were collected from each site on the number of visits for each sub-activity to identify pregnant women and young infants covered and treated under each intervention; time spent by different providers for each activity and salary of the provider; and quantities and prices of consumables and non-consumables for implementing the activities.
The number of young infants covered was estimated by adding those that had 1, 2, 3, or up to 10 postnatal visits. Treated young infants were estimated by adding those who had 1, 2, 3 or up to seven days of treatment for each regimen. For example, under regimen B in Kenya, if five young infants were treated for two days, three young infants for three days, 12 young infants for four days, two young infants for five days and 219 young infants for seven days, then the total number of young infants treated would be 241 (5+3+12+2+219) and visits would be 1610 (2*5)+(3*3)+(4*12)+(5*2)+(7*219). Table 1 provides the number of visits and treated young infants for each activity, and covered young infants for each site. A time and motion study was carried out on average time taken by CHWs, CHEWs and nurses per visit on pre-prepared forms for each activity, as self-recorded and also by an interviewer (who followed the health provider) on a random day to validate the self-recorded data [27]. Time spent on travel, waiting and visits where no contact could be made with the caregiver was recorded separately and split across the sub-activities depending on the purpose of the travel. Personal time was not included. Data on duration (in minutes) were recorded and estimated per visit per woman or child for which the activity was undertaken ( Table 2). If during a visit of 20 minutes, four women were surveyed in a household, then five minutes were estimated per visit per woman. If more than one provider was involved in an activity, then the weighted average cost was calculated based on the percentage of total visits covered by a provider for that activity (shown in brackets in Table 2). For example, in Nigeria the first dose of oral amoxicillin is given by the nurse at the clinic and the subsequent six daily doses by a CHEW at home. For a seven-day treatment with two doses per day, 1/7 (14%) was provided by the nurse and 86% was by the CHEW. The second dose was given by the parent each day, and no human resource costs were attributed. Table 2 also shows where the activity took place -in the clinic or as outreach. While the activities in Table 2 correspond to the activities in Table 1, the treatment under different regimens (activities 4.2-4.6) requires a combination of sub-activities (4a-4c in Table 2). Total human resource costs for each activity and regimen were determined based on the salary of the providers (Table 2) and total duration spent for a visit and an episode (estimated based on times per day and days per child) of treatment.
Data on quantities of drugs and consumables were collected for each regimen. While the injections were per administration, oral amoxicillin was calculated per treatment course. As unused amounts after dilution were not used later, the quantities for treatment with oral amoxicillin remained the same with 14, 10 or 2 doses under different regimens. Based on the number of administrations per day and number of days of treatment, total amounts of drugs and consumables required for a full seven-day treatment under different regimens are shown in S1 Table. The price of drugs and consumables was also calculated per dose and per full (seven days) or partial treatment (assumed to be three days). Drugs were centrally provided through WHO, and international procurement prices were used when local prices were not available. In the DRC and Kenya local generic prices were used while international procurement prices were used for Nigeria. Local prices for all consumables were available. The prices of drugs and consumables used across different sites are given in S2  3. Infrastructure and equipment: Equipment for power, computing/office, transport and communication were considered as health system costs and not included.

4.
Operational costs: Only 10% of the operational costs of transport, communication (airtime, internet and mobile allowances), and meetings with programme managers, nurses and CHWs/CHEWS were included. At least one monthly meeting in each community for all health workers was assumed. The balance of 90% was incurred for research, such as set-up costs of purchasing vehicles and office equipment or maintaining health systems, including utilities (such as water and electricity bills).

5.
Training: Costs were estimated for one refresher training course annually for all nurses, CHWs/CHEWS and supervisors considered essential for running the programme effectively.
Training of trainers of nurses and CHWs/CHEWS, initial training of CHEWS and nurses, and other costs such as files, training material, and job aids were considered as research costs.

Cost estimation for regimens
The costs per child treated were determined for five regimens and rescue treatment by adding the per child cost of human resources, drugs and consumables.

*Human resources across all sites includes the cost of daily assessment of young infants treated.
The costs of seven days of treatment per young infant treated showed lowest human resource costs in North and South Ubangi, DRC, primarily due to lower salaries and also because all injectables were provided in the clinics so that the travel and waiting time for nurses were reduced. In Kenya, the regimen that involved injectables administered by a nurse was more expensive due to higher salaries. In Nigeria, CHEWs replaced many functions performed by nurses, such as administering injectables, so the cost of treatment was lower due to lower salaries of CHEWs. However, this lower cost was offset as all treatment except the diagnosis and the first dose in Nigeria was an outreach activity by CHEWs, requiring more time than treatment at a clinic. The drug costs were similar across sites as in most cases international procurement prices were used. However, where generic medicines were used in the DRC and Kenya costs were lower. Kenya had higher costs for consumables (needles and distilled water vials [DWVs]) as compared to all other sites.
The direct costs of treatment for young infants treated for less than seven days were added to those treated for a full seven days to obtain the average cost per child treated for different regimens. The total costs of home-based care (identification, antenatal care, postnatal visits), links to facility, daily assessment and management were calculated based on human resource costs per visit and total number of visits. Only the marginal costs for non-consumables, operations, training and personnel for supervision and administration necessary for running the programme were used. Table 4 shows per child treated and per child covered costs for different treatment regimens including administration. Pre-OP treatment costs include home-based care (surveillance, antenatal and postnatal care visits) and links to the facility. 2 OP direct treatment costs include human resources for daily assessment of newborns and administration of drugs, medicines and consumables. 3 Indirect and administrative costs include personnel for supervision, management and operation,; training, nonconsumables and operations such as transport, communications and meetings. 4 The total costs may be slightly different from the sums due to rounding errors.  but is also less disruptive for families and carries less risk of hospital-acquired infections [22]. The reasons for refusal to accept referral to a hospital include lack of permission from concerned family members, lack of child care, religious and cultural beliefs, distance, cost of travel and treatment, concerns around quality of care and attitudes of health workers [13][14][15].

Interpretation of results
WHO guideline recommends using regimen B as option 1 and regimen D as option 2 for treatment of clinical severe infection in young infants 0-59 days old when referral is not feasible [20]. Using the risk difference in treatment failures reported in AFRINEST [16][17], and cost per treated child estimated above, the incremental cost-effectiveness ratio for clinical severe infection at average and low rates of treatment failures, regimen B comes out as most cost-effective, followed by regimen D. Compared to reference regimen A, while all three regimens are cost-effective at low and average risk difference, the results at high risk difference show lower cost and lower effect, and therefore need to be evaluated. For fast breathing, treatment with regimen E is more costeffective than reference regimen A with ICER=-1.1 for average risk difference in treatment failure. Manandhar and colleagues argued that an intervention that costs less than US$ 127 is costeffective [29]. Black and colleagues suggested that "interventions costing less than per capita gross national income per DALY averted can be termed "very cost-effective," and those costing less than three times per capita gross national income can be termed "cost-effective" [30]. The treatment for PSBI costing less than US$ 35 with either regimen used in our study is cost-effective at all sites by these criteria.
Our results show that administrative costs are an important component of managing PSBI on outpatient basis when referral is not feasible. Marginal administrative costs for effective implementation of the programme include management, supervision, meetings between health providers and supervisors, at least one refresher training course annually for all staff delivering services, basic equipment such as weighing scales, thermometers and timers for every nurse and CHW/CHEW, communications and travel. These vary from almost one third to one half of the total treatment costs at different sites, the highest being in the DRC.
Another important component of managing PSBI in outpatient settings is identification of sick young infants, antenatal and postnatal visits including education of the mother. These pre-enrollment and pre-treatment activities cost more in Nigeria as these are undertaken by CHEWs who are more qualified and paid more than the CHWs in other settings. Daily assessments of treated young infants were considered important for effective outcomes. The cost per child covered and treated also varies depending on the number of visits and per visit costs for the intervention. Larger numbers of visits to households (per child treated) had to be made for finding those with danger signs, especially in Ibadan. In the DRC, even though the salary for the providers was low, the costs increased due to a longer time taken by the health workers to reach the population in communities due to difficult terrain.

Strengths and weaknesses
Our study's strengths were that a randomized controlled trial was used to implement the study, along with standardized training of staff and data collection. The costs captured in research settings are normally higher than in routine work, but have been appropriately allocated with robust assumptions for the government programme. Some items, such as human resources, programmatic and administrative costs are more difficult to estimate in a government setting, where the providers are engaged in more than one activity. Most studies do not estimate and attribute programmatic and administrative costs for effective implementation of the programme. Our study not only estimates these, but also considers pre-treatment costs of actively screening children with danger signs. One potential weakness could be that the study uses the costs data for 2012. In order to make comparisons with recent studies or to use the cost data for advocacy, inflation factor must be applied.

Implications and conclusions
Outpatient management is cost-effective using a combination of injectable plus oral antibiotics for PSBI when referral is not feasible, and oral antibiotic therapy for those with only fast breathing. While the treatment cost for regimen D is slightly lower, regimen B with gentamicin injections once daily and oral amoxicillin twice daily for seven days is more cost-effective for