Study design and aim
This study design was cross sectional, reviewing patient folders at one time point, and using cost-of-illness approach to estimate the value of resources expended on malaria treatments outside the STG, which could have been saved.
Sampling of facilities and folders
The country was stratified into three ecological zones to ensure a representation of the diversity of the issues across. These were: (1) Northern – Upper East, Upper West and Northern; (2) Central – Brong Ahafo and Ashanti; and (3) Southern – Volta, Eastern, Greater Accra, Central and Western. One region each was selected to represent one zone. Thus, a total of 3 regions were covered. A list of districts within each region was obtained from Centre for Health and Information Management/ Policy Planning, Monitoring, and Evaluation Division of Ghana Health Service (CHIMS/PPME – GHS). Two districts were randomly selected in each region (one rural, one urban). In each district, the different levels and ownership of facilities were considered, namely public, mission and private. A total of 27 facilities were selected for this study (Table 1).
Sample size for patient folders for all regions
This study was designed primarily to estimate the mean cost of inappropriate prescription in health facilities. Assuming that the cost follows the Gaussian distribution, the equation used in estimating the required sample size is given as follows:
$$n=\frac{{\sigma }^{2}{\left({Z}_{\frac{\alpha }{2}}\right)}^{2}(1+f)}{{e}^{2}}Design effect$$
Where\({Z}_{\frac{\alpha }{2}}=1.96\) is the standard normal deviation corresponding to 95% significance criterion,
\(f=10.0\%\) is the non-response rate,
\(e\) =0.05 is the margin of error and
\(\sigma =0.75\) is the standard deviation of cost from 2015 National Health Insurance Scheme tariff.
Since \(\sigma\) has not been estimated from previous studies in Ghana, we approximated it using
where r is the range of cost from public primary care hospitals (GHS7.20 - 10.20). Assuming a design effect of 1.709, our estimated sample size for the study was 1,625 patients. This sample size was then allocated to regions (first) and facilities based on population of patients proportional to size.
Once the total sample for each facility for the year was estimated, simple random sampling was used to allocate the sample across the months (since malaria OPD attendance was not uniform across months). This was done by first obtaining OPD attendance register for the year 2016. The sampling frame was OPD malaria attendance for the year 2016. Secondly, monthly breakdown of OPD malaria cases for the year 2016 was obtained from the register, and identification number for folders of malaria patients were recorded. Then, the sample for the month was calculated based on the formula:
Where,
A = malaria cases for month y in each facility
B = total malaria cases for the year 2016 in each facility
C = total sample size calculated for each facility
The calculation was done for all the months (n = 12) in the year 2016. Thirdly, after obtaining the sample size distribution by month for 2016, all the patients’ folder numbers for each month were listed and the sample was drawn randomly. The process was repeated for all 12 months of the year 2016.
Inclusion criteria: Patients’ folders containing diagnosis of malaria, and explicitly specified as such in the folder.
Exclusion criteria: Patient recorded to have malaria in OPD register, but malaria diagnosis not clearly written or missing in patient’s folder.
Data extraction and variables
A data extraction form was used to gather patient prescription information from their folders. Data from the folders were collected by trained research assistants guided by prescribers in each facility. Using the STG for uncomplicated malaria, prescriptions were assessed and classified into inappropriate and appropriate prescriptions. Variables recorded include data on age, sex, diagnosis and the antimalarial medicines prescribed to the patients.
Data analysis
The average number of medicines per encounter was estimated. This was calculated as: Average number of medicines prescribed per encounter (C) = Total number of medicines prescribed (B) / number of encounters surveyed (A). Further analysis of the data provided additional indices such as pattern of prescription by Standard Treatment Guideline [17] among prescribers, frequency of antimalarial prescriptions by level of facility and prescriptions by therapeutic groups such as antibiotic prescription, analgesic prescription pattern, antimalarial etc. Finally, two physicians, using the STG as the specified criteria, independently assessed all prescription patterns with malaria diagnosis to determine whether the guide was followed. Disagreements were resolved by discussion and, if necessary, a third independent person was involved.
Determination and definition of inappropriate malaria prescriptions
We determined appropriate prescriptions as those based on appropriate examinations to test for malaria, diagnosis and then prescription of ACTs per the recommendations in the STG. If prescription and treatment did not follow this procedure, then the prescription was defined as inappropriate.
Determination of proportion of inappropriate malaria prescriptions
Based on the records samples per facility type, the proportion of inappropriate prescriptions (i.e., not adhering to STG) was estimated from the total sample size and expressed as the percentage of prescriptions for all uncomplicated malaria cases. A national estimate was calculated based on the sample estimates, number of regions, number of districts and number and type of health facilities in the country.
Estimation of cost of inappropriate prescription (Institutional cost)
The institutional cost of inappropriate prescriptions for uncomplicated malaria comprised medication cost and personnel cost. The costs of medications prescribed outside the STG was determined using prices obtained from the Public Procurement Authority of Ghana (PPA) and the market. Total and average costs were calculated. In the case of personnel cost, the average prescribers and dispensers’ times were determined from interviews. These average times were then multiplied by the daily gross salaries of prescribers and dispensers to estimate the personnel costs by health facility type. The summation of the medication and personnel costs constituted the total institutional cost of non-adherence to prescriptions for uncomplicated malaria.
Results of all cost were reported in United States dollar using Bank of Ghana annual average interbank exchange rate, 2016 (US$1.00 equivalent to GHS4.3).
Ethical approval
As a programmatic review for the National Malaria Control Programme, permission for data collection was obtained from the Ghana Health Service Headquarters and Christian Health Association of Ghana, Regional and District Directors, and facility heads.