This study used patient-level data from Iguhu hospital, a sub-county hospital in Kakamega county, in the malaria endemic-prone Western Kenya highlands. The hospital is funded by the Kaka mega County government Health serving a population of 17,860 people. It has 21 beds and 4 cots catering for children, women, and men. Critically ill patients are transferred to Kakamega County Teaching and Referral Hospital in Kakamega township 16 kilometres away. The majority of the population work in small-scale subsistence farming keeping livestock, growing corn and tea or work in the informal sector.  Water catchments are used for agricultural purposes. Comprising both endemic and highland-epidemic prone areas, the topography, climate and farming methods favour high rates of malaria transmission. A sample of children in Kakamega county found levels of parasitemia in Kakamega to be 33%, well above the national average of 8%.
The sample comprised all infants, children and adults admitted with a confirmed laboratory diagnosis of malaria to Iguhu hospital between June 2016 and May 2017.
Cost data collection
An ingredients-based approach was used to estimate the health system and household costs of diagnosing and managing malaria hospitalisations. Patient and household costs were classified as: 1) direct costs (or out-of-pocket costs) subdivided into medical and non-medical costs; and 2) indirect costs. Direct medical costs included payments for hospital fees, medications and non-medical costs such as transport, meals and the costs of funerals for malaria related deaths. Indirect costs - including lost productive time due to travelling to hospital, being ill or providing care to a sick child – were estimated for adults over the age of 18 years using the human capital approach, taking a subsistence wage for an agricultural worker  and multiplying it by the lost time.
To estimate resource use, detailed data were collected from hospital admission records which included: patient age, sex, village of residence, sublocation, pregnancy status, type of malaria diagnostic test conducted, type of malaria treatment, any other diagnostic tests and treatments, length of stay, discharge destination; and any fees paid by patients for their inpatient stay. Key personnel from Iguhu hospital advised on specific costs for patients, including hospital charges, transport and food, provision and stockouts of medications in the past 12 months, and the approximate costs of these medications at local village vendors.
Health system costs
Health system cost data were extracted from patient-level medical records and included costs associated with hospitalisation (bed cost), laboratory tests, medications and related overheads (S.Table 1). Health system costs were classified into four categories: bed day cost and a diagnostic test for malaria on admission, management of malaria, management of comorbidities such as diarrhoea and anaemia excluding antibiotics; and antibiotics for management of infection.
Inpatient cost per bed day was obtained from World Health Organization (WHO) “choosing interventions that are cost-effective (WHO-CHOICE)” framework.  These country-specific estimates are based on a primary level hospital with few specialties and between 30-200 beds and include all personnel, capital and accommodation costs excluding medications and tests. The cost per inpatient day was inflated to 2020 prices and averaged over 12 months. This bed cost was multiplied by length of stay which was extracted from inpatient records.
Laboratory test for malaria
All inpatients had confirmed malaria and the cost of microscopic diagnosis was derived from previously published estimates for Kenya.
Medications for management of malaria and additional treatment
As the hospital records only listed the names of medications, dosage calculations were based on recommended dosage and duration documented in the Médecins Sans Frontières (MSF) Essential Drugs, and MSF Clinical Guidelines based on diagnosis. When recommended dosages were based on a child’s weight, approximate weights for age were calculated using the WHO weight for age charts and costs were estimated based on manufacturer formulations and pack size. Dosages were independently calculated by CW and JA. When an antibiotic was recorded but an additional diagnosis was not documented (10% cases), the most frequent diagnosis for persons of a similar age in the dataset who were also prescribed that antibiotic was used to estimate dosage.
The costs of medications and equipment were estimated using the Kenya Medical Supplies Authority (KEMSA) price list. KEMSA is funded through the Ministry of Health and is responsible for the procurement and sales of essential medicines and medical supplies to government health facilities. For drugs not included on the KEMSA list, a mean cost was calculated using the Kenya Drug-Index. The costs of microscopy to detect parasites in blood samples used to confirm a diagnosis of malaria were based on previously published estimates.
All medications listed in the hospital records and administration costs (disposable needles and syringes, IV giving sets) were included. For an additional diagnosis of dehydration, the cost for the insertion of an intravenous line and 24 hours of IV fluids was included. For anaemia, where blood was required, it was assumed that one unit of packed cells was used and that any oral supplements prescribed were administered for the duration of inpatient stay.
Patient and household costs
Direct medical costs
Direct patient costs included: registration books, laboratory test for malaria on admission and any hospital fees documented in hospital records. Children under five years of age and pregnant women were exempted from paying a registration fee and fees for malaria laboratory tests. Hospital inpatient charges to patients were included as recorded in the patient record. Any inpatient charges were assumed to be paid at discharge and deducted from the total health system cost.
Costs for medication following discharge were based on prices in the Iguhu hospital pharmacy medications price manual. Hospitals set their own price for medications dispensed to discharged patients and outpatients. If the drug was not listed in the manual, the price in the KEMSA price list was doubled to reflect the price ratio in the pharmacy manual.
Direct non-medical costs
Travel costs of patients and any accompanying persons were estimated for motorbike taxi transport. If the patient was a child under 6 years of age it was assumed that a caregiver travelled to and from the hospital with them and a meal was purchased each day. If a patient had to be transferred to Kakamega, an additional cost was estimated for a return trip. In the case of death, funeral costs were also included.
Discounting was not necessary as all costs were estimated over one year. Any costs from the published literature were converted to local currency (Kenyan Shilling) and adjusted for inflation to 2020. Changes over time in the prices of non-medical goods and services were adjusted using the Kenya consumer price index (CPI) , and medical costs were adjusted using the Kenyan CPI for Health , and then converted to USD.
Patient-level data were transferred securely in a Microsoft Excel format and analysis was performed using SAS 9.4 (SAS Institute, Cary NC). Patient characteristics are described with chi-square tests used to calculate differences in proportions. Sensitivity analysis was conducted to examine the impact on household costs of an increase in hospital costs due to inflation (5%); and the impact of stockouts of antibiotics as this was reported to be a common issue by hospital personnel. If antibiotics were not available from the hospital, it was assumed they would be purchased from a local vendor. The percentage difference between the public sector (hospitals) and private sector (pharmacies) procurement prices for locally produced and imported medicines (expressed as median price ratios)  was used to estimate the cost increase in antibiotics if they were purchased from a local vendor. To reflect usual availability of medicines, the proportion of imported medicines (55%) and locally produced medicines (45%) were also taken into account.  Based on these assumptions, the price of antibiotics in the private sector was estimated to be 158% higher than in the public sector. It was assumed antibiotics were in stock 66% of the time based on the reported availability of general medicines in the public sector.