Corresponding Author Kathryn Schnippel, Themba Lethu Wing, Helen Joseph Hospital, Perth Road, Westdene 2092 South Africa: Health Economics and Epidemiology Research Office, Department of Internal Medicine, School of Clinical Medicine, Faculty of Health Sciences, University of the Witwatersrand, Johannesburg, South Africa. Tel.: +27 11 276 8896; E-mail: email@example.com
In South Africa, patients with multi-drug-resistant tuberculosis (MDR-TB) are hospitalised from MDR-TB treatment initiation until culture conversion. Although MDR-TB accounts for <3% of incident TB in South Africa, 55% of the public sector TB budget is spent on MDR-TB. To inform new strategies for MDR-TB management, we estimated the per-patient cost (USD 2011) of inpatient MDR-TB treatment.
All resources used by patients admitted to the MDR-TB hospital with confirmed MDR-TB from March 2009 to February 2010 were abstracted from patient records for up to 12 months after initial admission or until the earliest of final discharge, abscondment or death. Costs of hospital stay/day were estimated from hospital expenditure records and costs for drugs, laboratory tests, radiography and surgery from public sector sources. 133 patients met study inclusion criteria of whom 121 had complete cost records.
By 12 months, 86% were discharged with culture conversion, 8% died in hospital, 2% were still admitted, and 3% had absconded. The mean hospital stay was 105 days. The mean total cost per patient was $17 164, of which 95% were hospitalisation costs (buildings, staff, etc.) and ≤ 2% each for MDR-TB drugs ($380); TB laboratory tests, including drug susceptibility testing ($236); and other costs.
The inpatient cost per patient treated for MDR-TB is more than 40 times the cost of treating drug-susceptible TB in South Africa. There is potential for substantial cost savings from improved management of drug-susceptible TB and shifting to a model of decentralised, outpatient MDR-treatment.
En Afrique du Sud, les patients atteints de tuberculose multirésistante (TB-MDR) sont hospitalisés depuis le début du traitement de la TB-MRD jusqu’à la conversion de la culture. Bien que la TB-MDR représente <3% de la TB incidente en Afrique du Sud, 55% du budget de la tuberculose du secteur public est dépensé pour la TB-MDR. Afin d'informer les nouvelles stratégies de prise en charge de la TB-MDR, nous avons estimé le coût (USD 2011) du traitement par patient TB-MDR hospitalisé.
Toutes les ressources utilisées par les patients admis à l'hôpital de TB-MDR avec une TB-MDR confirmée, à partir de mars 2009 à février 2010 ont été extraites des dossiers des patients pour jusqu’à 12 mois après l'admission initiale ou jusqu’à la première sortie, l'abandon, ou le décès. Les frais de séjour hospitalier par jour ont été estimés à partir des dossiers des dépenses hospitalières et les coûts des médicaments, examens de laboratoire, radiographie et la chirurgie, à partir des sources du secteur public. 133 patients répondaient aux critères d'inclusion de l’étude dont 121 avec un relevé complets des coûts.
A 12 mois, 86% sont sortis avec une conversion des cultures, 8% sont décédés à l'hôpital, 2% étaient encore admis et 3% avaient abandonné. La durée moyenne d'hospitalisation était de 105 jours. Le coût total moyen par patient était de 17164 USD, dont 95% étaient des frais d'hospitalisation (bâtiments, personnel,… etc) et ≤ 2%, soit pour les médicaments de la TB-MDR (380 USD), soit pour les tests de laboratoire de TB, y compris les tests de sensibilité aux médicaments (236 UDS), soit pour d'autres coûts.
Le coût d'hospitalisation par patient traité pour la TB-MDR est plus de 40 fois le coût du traitement de la TB sensible aux médicaments en Afrique du Sud. Il est possible de réaliser des économies substantielles à partir d'une meilleure prise en charge de la TB sensible aux médicaments et le passage à un modèle de traitement décentralisé, ambulatoire de la MDR.
En Sudáfrica, a los pacientes con tuberculosis multirresistente (MDR-TB) se les hospitaliza desde el comienzo del tratamiento de la MDR-TB y hasta la conversión del cultivo. Aunque la MDR-TB es responsable de < 3% de incidencia de la TB en Sudáfrica, un 55% del presupuesto del sector público se gasta en MDR-TB. Con el fin de facilitar información para la elaboración de nuevas estrategias para el manejo de MDR-TB, hemos calculado el coste del tratamiento para MDR-TB por paciente ingresado (USD 2011).
Se obtuvieron datos sobre todos los recursos utilizados por pacientes admitidos en el hospital con una MDR-TB confirmada entre Mar 2009–Feb 2010, a partir de las historias clínicas de los pacientes, desde el momento de la admisión hospitalaria y hasta 12 meses después del primer ingreso, o tras haber sido de alta, haber abandonado o haber muerto. Se calcularon los costes de la estancia en el hospital/día utilizando los registros de gastos del hospital, mientras que los costes de los medicamentos, las pruebas de laboratorio y de cirugía se calcularon utilizando fuentes del sector público. 133 pacientes cumplían con los criterios de inclusión del estudio, de los cuales se tenía la información de costes completa para 121.
A los 12 meses, un 86% había sido dados de alta con conversión del cultivo, un 8% había muerto en el hospital, un 2% continuaba hospitalizado, y un 3% había abandonado. La estancia media en el hospital era de 105 días. El coste medio total por paciente era de $17,164, del cual un 95% correspondía a costes de hospitalización (edificios, personal sanitario, etc.) y ≤ 2% a cada una de las siguientes partidas: medicamentos para la MDR-TB ($380); pruebas de laboratorio para TB incluyendo pruebas de susceptibilidad a los medicamentos ($236); y otros costes.
En Sudáfrica, el coste de hospitalización por paciente tratado para MDR-TB es más de 40 veces el coste de tratar una TB susceptible a medicamentos. Es posible conseguir un ahorro sustancial con un manejo mejorado de la TB susceptible a medicamentos y con el paso a un modelo de tratamiento descentralizado y extra-hospitalario para la MDR-TB.
In recent years, following on severe global epidemics of both HIV and drug-susceptible tuberculosis, multi-drug-resistant tuberculosis (MDR-TB) has emerged as a serious health challenge. In 2009, South Africa had 9070 confirmed cases of MDR-TB, second highest globally in total case load (WHO 2011a) and the highest global case load for extensively drug-resistant TB (XDR-TB) at 572 reported cases (Zignol et al. 2012). MDR-TB requires 18–24 months of treatment and is associated with very high early mortality (Gandhi et al. 2010). Second-line TB treatment is expensive and complicated to administer, with potential for severe adverse events and success rates just above 60% (Orenstein et al. 2009). An estimated 60% of patients with TB in South Africa are co-infected with HIV (WHO 2011b). Although there is limited and conflicting global evidence of an association between HIV and MDR-TB (Suchindran et al. 2009), observational studies from South Africa have reported worse outcomes for MDR-TB patients co-infected with HIV (Brust et al. 2011; Farley et al. 2011).
Until late 2011, all patients with MDR- and XDR-TB in South Africa were treated as inpatients at specialised, provincial-level MDR-TB treatment centres. Patients with MDR-TB were hospitalised for 6 months or until they achieved culture conversion, defined as two consecutive months with culture-negative sputa (Directorate Tuberculosis Control 2007). Discharge was delayed if the patient was in poor clinical condition or had a previous history of treatment interruption, complications or major adverse drug events.
While drug-resistant TB of all types comprises less than 5% of incident TB in South Africa, control of drug-resistant TB accounted for 55% of the National Tuberculosis Program (NTP) budget in 2010 (WHO 2011a). A large share of this cost is likely due to the long hospital stays of patients with MDR-TB, which also results in treatment delays due to a shortage of some 740 MDR-TB treatment beds (Directorate Drug- Resistant TB 2011a). There are no empirical estimates, however, of costs of inpatient care for MDR-TB in South Africa, and very limited cost data are available globally. A recent systematic review of the costs of MDR-TB treatment found only four published estimates that met inclusion criteria for the review, and none were from sub-Saharan Africa (Fitzpatrick & Floyd 2012).
Although new South African guidelines (Directorate Drug- Resistant TB 2011b) recommend that smear-negative MDR-TB patients be treated through an outpatient model of care, most patients with MDR-TB will likely continue to require substantial hospital stays. Without up-to-date information on the cost of treating MDR-TB, it will be difficult for policy makers and programme managers to estimate the cost-effectiveness of alternative treatment models, examine the impact of shorter treatment regimens or plan for implementation of new guidelines. To fill this knowledge gap, we used patient-level data from a provincial MDR-TB treatment centre to estimate the cost of providing inpatient MDR-TB treatment in South Africa.
The study was conducted at the Klerksdorp/Tshepong Hospital Complex in North West Province, South Africa, the provincial referral hospital for MDR- and XDR-TB. Initially, patients with MDR-TB and XDR-TB were hospitalised in a 32- and 6-bed ward, respectively, within Klerksdorp Hospital. During the study period, funding from The Global Fund to Fight AIDS, Tuberculosis and Malaria assisted the North West Department of Health to construct new, separate structures for both inpatient wards and an outpatient clinic. Patients are now housed in a 36-bed MDR-TB ward and 20-bed XDR-TB ward at Tshepong Hospital. Full medical records for admitted patients, including laboratory results and drug administration charts, are stored in a file room within the specialised MDR- and XDR-TB wards.
TB treatment guidelines in effect during the study recommended that phenotypic first-line drug susceptibility testing (DST) be performed for all TB suspects with a history of TB treatment and for contacts of patients with MDR-TB. DST was also performed if patients with TB remained sputum smear positive at the end of 2, 3 or 5 months of first-line TB treatment (TB Control Programme 2008). DST for resistance to second-line drugs was performed for all viable culture isolates that were found to be resistant to both isonaizid (INH) and rifampicin (RIF). Patients with INH and RIF resistance were registered as MDR-TB cases and admitted to the study site wards to initiate treatment.
Similar to drug-sensitive TB treatment, MDR- and XDR-TB treatment have an intensive phase and continuation phase. Culture conversion, defined as two consecutive negative culture sputa taken 30 days apart, was used as an indicator of treatment efficacy. This determined the duration of hospitalisation, with hospital discharge at culture conversion. The intensive phase of treatment coincided with hospitalisation, but also continued for 4 months after culture conversion. The standardised regimen for the intensive phase of MDR-TB treatment was daily kanamycin or amikacin injections and terizidone or ethambutol, pyrazinamide, ofloxacin and ethionamide (Directorate Tuberculosis Control 2007). Upon discharge from hospital, all patients received outpatient care including directly observed treatment from a local clinic, with monthly outpatient visits to the MDR-TB hospital. Injections are not part of the continuation phase; all other drugs in the standardised regimen are continued at the same dose. Capreomycin injection (during intensive phase), para-aminosalicylic acid and moxifloxacin were available if patients were resistant to a second-line drug.
Most laboratory tests ordered for patients with MDR-TB were performed on site, at the National Health Laboratory Services (NHLS) facility at Tshepong Hospital. This included fluorescent sputum smear microscopy, liquid media cultures and DST for first-line drugs (INH, RIF, ethambutol and streptomycin). DST for resistance to second-line drugs (ethionamide, kanamycin and ofloxacin) was performed at NHLS's central TB referral laboratory in Johannesburg, a 2-h drive from the study site. During the study period, the central referral laboratory began using a line probe assay (LPA) for the identification of mycobacterium tuberculosis (MTB) in culture isolates while also providing rapid genotypic testing for INH and RIF resistance. Xpert MTB/RIF was not available at the study site during the study period.
The study was a medical record review of resource utilisation and outcomes for all patients with MDR-TB admitted between 1 March 2009 and 28 February 2010 with confirmed MDR-TB and for whom the complete patient medical record was available. Patients were excluded if they were under age 18 at the time of admission, were diagnosed with drug allergies or with mono- or poly-resistant TB but not MDR-TB, had initiated MDR-TB treatment at a different site or transferred out of the study site to a different inpatient treatment facility within 12 months of admission to the study site. All patients were followed from their admission until the earliest of the date of final discharge, abscondment or death, up to 12 months after admission. Patients who progressed to a diagnosis of XDR-TB during the 12 months after initiating MDR-TB treatment were retained in the sample.
Cost data and analysis
For each study subject, we reviewed the medical record to collect data on resources utilised from the date of admission to 12 months after admission. Resources captured included inpatient days; TB and non-TB drugs and fluids, including antiretroviral (ART) medications for HIV; specialist consultations; surgical procedures; chest x-rays and other scans; and TB monitoring tests and other laboratory investigations. For patients who had multiple admissions and intermediate discharges during the 12-month follow-up period, the total number of days admitted over the period was used. These data were used to estimate the total number of units of each resource used by each study subject.
The average inpatient cost of MDR-TB treatment was estimated using costing methods recommended by the WHO for TB control (Floyd 2002). Prices of medications including MDR-TB drugs, fluids and laboratory tests and investigations were collected from public sector suppliers of these products and services. Public sector hospital charges were used for specialist consultations, surgical procedures and scans, with the exception of x-rays for which costs were estimated from hospital expenditure records. Personnel costs, medical supply costs and costs for the purchase, maintenance and operation of infrastructure and equipment were collected from hospital expenditure and financial reports.
To estimate the cost per patient in our sample, we calculated three types of costs. First, for all patient-specific resource usage (drugs, laboratory tests, fluids, specialist consultations, scans, x-rays and surgical procedures), costs were calculated by multiplying unit costs by quantity of resources used. Second, DR-TB ward-specific costs that could not be attributed to individual patients were estimated as the sum of ward costs for personnel, supplies, equipment and infrastructure, and the total DR-TB ward cost divided by the total number of DR-TB inpatient days to obtain a cost per patient day for DR-TB ward-specific costs. Third, for shared services that the general hospital provides (e.g. hospital administration, laundry and security), we estimated the proportion of all inpatient days and outpatient days for the whole hospital that were spent in the DR-TB ward. We then multiplied the hospital's total cost for shared services by this proportion. For both DR-TB ward-specific costs and shared service costs, we divided the relevant totals by the total number of TB inpatient days to obtain a cost per patient day. The sum of ward-specific and shared costs per inpatient day is referred to as the ‘hotel’ cost. The total cost per patient in our sample is the cost of patient-specific resource usage plus the daily hotel cost for the length of the admission.
Buildings and equipment were discounted at 3% per annum. Costs are reported in 2011 USD, with prior year costs inflated using the South African consumer price index (Statistics South Africa 2012). The average exchange rate for 2011, ZAR 7.23/USD, is used (Oanda 2012). Patient-level data were captured in CSPro v. 4.1 (US Census Bureau). SAS v. 9.3 was used to generate descriptive statistics, analysis of the differences of proportions (chi-squared test) and differences of means (t-test).
Ethics approval for this study was received from the Human Research Ethics Committee of the University of the Witwatersrand, the Institutional Review Board of Boston University Medical Center and the Hospital's Research Committee.
During the 12-month period of study enrolment, 277 patients registered at the study site; 133 of these met inclusion criteria and were enrolled in the study (Figure 1). The main reason for exclusion was absence of a confirmed MDR-TB diagnosis, which typically indicated mono- or poly-resistant TB but not MDR-TB. The mean age of the study sample was 39.9 years, and 45% were female (Table 1). Nearly all (n = 123, 92%) reported being resident in North West Province at admission. 64% were unemployed. A large majority (n = 111, 83%) had a history of previous TB treatment. The mean interval from collection of sputum for DST to laboratory report MDR-TB diagnosis was 84 days; and from sputum collection to hospital admission was 111 days. By definition, all patients were TB culture positive at diagnosis, with INH and RIF resistance confirmed by either DST or LPA or both. 56% of patients were smear positive at diagnosis.
Table 1. Study population characteristics, by smear status at admission
Of 133 enrolled study subjects, four did not have a smear microscopy taken on admission and one patient's sputum sample was rejected, leaving a total of 128 subjects with known smear status at baseline. The sample was evenly divided between patients who were smear negative and smear positive at hospital admission. Three-quarters of smear-negative patients and more than half of smear-positive patients were HIV infected, of whom 43% were on ART at admission. Low body mass index (BMI) was common, with two-thirds of smear-positive patients and almost one-third of smear-negative patients having BMI <18.5. Anaemia (40%) and diabetes mellitus (10%) were also frequently diagnosed. 69% of smear-negative patients were culture negative at admission, and 72% of smear-negative patients had resistance to only first-line anti-TB drugs. In contrast, 53% of smear-positive patients had resistance to one or more second-line drugs.
Of the 133 subjects enrolled in the study, 121 had complete information on outcomes and resource utilisation in their medical records. This smaller sample of 121 is used for the remainder of this analysis.
After 12 months, 98% of patients who were smear negative at admission and 80% of those who were smear positive had been discharged upon culture conversion (Table 2). Three patients, all of whom were all smear positive at admission and resistant to one or more second-line TB drugs, were still hospitalised at the end of 12 months. Ten patients (8%) died while admitted, all of whom were HIV infected. Four (3%) absconded.
Table 2. Patient outcomes, resource utilisation and costs, by smear status at admission
All subjects (N = 121)
Smear positive (n = 55)
Smear negative (n = 61)
Smear not done (n = 5)
Significant difference between smear-positive and smear-negative subjects at P-value <0.05.
Sample size too small too stratify by these outcomes.
Patients were hospitalised for an average of 105 days (Table 2). 115/121 patients had at least one chest x-ray, usually at admission; the mean number of chest x-rays was 1.2 per patient. Sputum smears and cultures were, on average, taken monthly. Blood tests, including CD4 monitoring and liver and kidney function screening, were performed for patients without recent test results at admission and only repeated if admission was longer than 6 months or if results were abnormal.
Of the 74 patients (61%) with at least one positive TB sputum culture after admission, 64 had anti-TB drug resistance testing performed. Resistance testing was ordered as sets – phenotypic DST for four first-line TB drugs and, if the culture isolate was resistant to at least INH and RIF, then DST for three second-line TB drugs was completed by the laboratory. On average, patients were tested once for a set of seven drugs.
For 90% of sampled patients, treatment followed the standardised MDR-TB regimen. Patients received an average of 52 injections while hospitalised (i.e. 3.5/week). Reasons for not receiving injectables included a contra-indication identified by the attending doctor, adverse reactions and reduced frequency and dosage of injectables for patients who were underweight. Nearly all patients received daily vitamins B6 and B-complex. Pain and cough medication and hypnotics were commonly dispensed. Surgical procedures, other scans or investigations and specialist consultations (e.g. audiologist or ophthalmologists) were infrequent in this patient cohort. Two-thirds (66%) of the HIV-infected patients were maintained or initiated on ART during their hospitalisation, most (92%) on the standard first-line ARV regimen.
The average hotel cost per inpatient day, including room, board, clinical interactions and basic supplies to maintain the hospital wards but excluding all procedures and drugs, was $155 (Table 3). 58% of these costs were specific to the MDR- and XDR-TB ward. Shared services, including hospital administration, laundry and security, accounted for 21% of general hospital expenditure. MDR-TB inpatient days comprised only 8% of the total patient days for the hospital, and therefore, just 1.7% of general hospital expenditure was allocated to the MDR-TB inpatient costs.
Table 3. Hotel costs of inpatient treatment for MDR-TB, per day
Ward staff, including medical officer and nursing staff
Ward buildings and grounds
Ward equipment and furnishings
Supplies, including laundry, office, cleaning and patient meals
General hospital expenses allocated to MDR-TB patient costs
Non-ward hospital staff, including administration, pharmacy, admissions and medical records
Non-ward-specific hospital buildings and grounds
Non-ward-specific equipment and furnishings
Total hotel costs per day
While the new hospital was being constructed, some admitted patients who had negative smears but who had not yet culture converted were moved to an unused ward at a nearby private mining hospital. A monthly flat fee was paid to the private hospital for the service, with the drugs, laboratories and physician management continuing to be provided by the MDR-TB hospital. The average cost per patient day for this service was $188, or 21% higher than the cost per day at the specialised MDR-TB hospital.
The average cost of inpatient treatment was $17 164 per patient in the first 12 months after initial admission (Figure 2, Table 2). Cost per day in the hospital – ‘hotel costs’ including salaries, buildings, equipment, consumables and supplies – accounted for 95% of total costs. Patients who were smear negative at admission cost less than patients who were smear positive, with the difference due largely to the shorter inpatient stays of smear-negative patients (average 95 vs. 125 days). Costs for smear-positive patients were more positively skewed and had a wider variance than those for smear-negative patients. The average cost per patient for MDR-TB laboratory monitoring, including DST, was $236, 1% of the total. Drugs for treating MDR-TB, at $380 per patient, accounted for only 2% of the total. Total patient costs did not differ significantly by HIV status (results not shown). ART costs were a small component of total costs, with a mean cost of $66 per patient on ART. CD4 counts and viral load tests did contribute to HIV-infected patients having higher non-MDR-TB-related costs, at $180 compared to $132; however, these costs account for approximately 1% of the total inpatient costs.
In this study of the costs of inpatient treatment for MDR-TB in South Africa, we found that the average cost of treating a patient who is ultimately discharged with culture conversion — that is, a ‘successful’ treatment outcome — exceeds $17 000. This is nearly 40 times the average cost of treating drug-susceptible TB (Sinanovic et al. 2003) and nearly 25 times the cost of a year of first-line ART (Long et al. 2010). Management of drug-resistant TB already consumes a majority of the country's total TB control resources. The roll-out of Xpert MTB/RIF is estimated to increase the number of MDR-TB cases diagnosed by as much as 70% (Meyer-Rath et al. 2012), making it even more urgent to both prevent MDR-TB and to reduce the cost of its treatment.
Success rates for treating drug-susceptible TB in South Africa are far below both international and local targets, at 60% for retreatment, 64% for smear-negative TB and 73% for smear-positive TB cases (WHO 2011b). In 2011, the South African National TB Program adopted new guidelines allowing MDR-TB patients meeting specific criteria — primarily smear-negative disease and good clinical condition — to be treated as outpatients. While this is likely to reduce costs for these patients dramatically, the National Department of Health estimates that only 30–40% of patients with MDR-TB will meet these criteria (Directorate Drug- Resistant TB 2011a). Hospital stays will be shortened for many others, as the new guidelines allow discharge upon smear conversion, rather than culture conversion. Once the new guidelines are fully in effect, substantial cost savings should result for many but not all patients. The findings we report lend urgency to the effort to implement the new guidelines, which will require large-scale training of staff, improved patient monitoring and records systems, patient and community education on infection control, infrastructure improvements, accelerated laboratory processing capacity and decentralisation of the MDR-TB drug supply.
We note three main limitations to our findings. First, MDR-TB treatment typically lasts 18–24 months, with most of this interval spent in outpatient care. Our cost estimates capture only the inpatient component of the treatment regimen. While culture conversion is considered a good interim indicator and inpatient costs are usually far higher than outpatient costs, further research is needed to evaluate the total cost per patient of MDR-TB treatment. Second, by enrolling the study cohort at admission to an MDR-TB hospital, an average of 111 days after they were tested for MDR-TB, there is likely to be a survivor bias. Culture and DST can only be performed at centralised laboratories in South Africa, and both take weeks to generate results. For patients in our sample, the interval between sputum collection and availability of test results averaged 84 days, accounting for most of the delay in admission. A study in KwaZulu Natal, South Africa, that followed patients during the time from sputum collection to hospital admission found that 40% of patients with MDR-TB and 51% of patients with XDR-TB died in the first 30 days following sputum collection (Gandhi et al. 2010). Accelerated diagnosis of MDR-TB made possible by the scale-up of Xpert MTB/RIF technology may change the profile of the inpatient population and thus of the costs incurred. Finally, the data we present are from a single site in South Africa. There is variation in estimated hospital per patient day equivalent expenditure across South Africa, with an average $213 per day and standard deviation of $45 (Health Systems Trust 2012), and it may be that there is also variation across the MDR-TB facilities.
Although these limitations should be kept in mind, the cost estimates presented here provide the best evidence yet of the high cost of treating MDR-TB using an inpatient model of care. These results can be used to estimate the cost-effectiveness of alternative models of care and budget for the large proportion of patients who will continue to require hospitalisation. Our results also have relevance to other high-burden MDR-TB countries. The WHO reports that 24 of 27 high-burden countries require hospitalisation during the intensive phase of treatment (WHO 2011a). In the review of MDR-TB treatment costs mentioned previously, the two countries included that had inpatient models of care, Estonia and Russia, had much higher costs than those that allowed outpatient treatment, Peru and the Philippines (Fitzpatrick & Floyd 2012). For South Africa and other countries, our results thus underscore the importance of implementing outpatient MDR-TB treatment as quickly as possible, so that the resources currently used for long inpatient stays can be re-allocated to improving first-line TB treatment and prevention.
We thank Dr M van Rensburg, JP Phepheng and the staff of Klerksdorp/ Tshephong Hospital for their assistance. Funding for this study was provided by the South Africa Mission of the US Agency for International Development. KSch and EV received research training funded from Fogarty International Center. The funders had no role in study design, analysis, decision to publish or preparation of the manuscript.