Study design, setting and study population
This was a quantitative study in which patient data was collected retrospectively while prescribers were interviewed by a questionnaire in 11 public health facilities of purposively selected service levels. The facilities included Jinja Regional Referral hospital 01, Health centers IVs (n = 4), and Health centers IIIs (n = 06) in Jinja City, Uganda. Data were collected from June1, 2022 to May 31, 2023. he public health system in Jinja City consists of 26 public health facilities, including Jinja Regional Referral Hospital, 13 Health Center IIs, 8 Health Center IIIs, and 4 Health Center IVs. The Uganda’s health care system consists of a hierarchy health facility service levels starting with village health team (VHTs) at community village level as the lowest to the health center II at parish level, health center III at sub county level, health center IV at County level and general hospital at district level, Regional Referral Hospital (RRH) at regional level and National Referral Hospital (NRH) at national level as the highest. The volume, complexity of services and specialties increases across the hierarchy such that the lowest level refer patients to the next highest level. Uganda Clinical guidelines (UCG) and Integrated Management for Child hood Illnesses (IMCI) are the standard treatment guidelines used in Uganda.
Inclusion and exclusion criteria
We included outpatient records of individuals diagnosed with RTIs from public health facilities in Jinja City between June 1, 2022, and May 31, 2023. Exclusion criteria comprised records with missing entries for age, sex, and diagnosis. Additionally, we excluded records from Health Center IIs and those of RTI patients with other infectious comorbidities.
Sample size and sampling procedure
In order to determine the prevalence of antibiotic prescription, the appropriateness for prescription of antibiotic, and patient factors associated with antibiotic prescription, we targeted to retrospectively abstract 1790 patient records from the outpatient register. The reason was to have a sample size which was above 600 and that would allow having more than 100 records from each of the 11 public health facilities as recommended by WHO/INRDU for studies assessing the quality of prescribing, [20].
To cater for patient load at different levels of the facilities so as to avoid oversampling and under sampling at individual facilities, we calculated the sample size for each facility based on the average rate of out-patients with RTIs at the facility during the study period.
Facility sample size = \(\frac{total number of RTI out patients for the facility}{total number of RTI outpatients for all facilities in the study. }\times study sample size\)
The total number of patients diagnosed with RTIs was obtained from the Health unit out-patient monthly HMIS 105 report. This report summarizes the total number of cases for each disease condition diagnosed at the health facility within a month. In order to establish health system factors that were associated with antibiotic prescription we targeted to interview all prescribers available in the 11 public health facilities.
We purposively selected public health facilities: one being Jinja Regional Referral Hospital premised on having the highest population of prescribers. The same method was used to select four health center IVs being the only ones in the Jinja City. Six out of eight health center IIIs in Jinja City were selected by simple random sampling. The prescribers were purposively sampled, on the premise of being persons who had prescribed antibiotics for more than 30 days during the study period.
At each health facility, patient records were selected from outpatient registers using systematic random sampling. The sampling interval (K) was calculated using the formula K = N/n, where K is the sampling interval, N is the total number of patient records available at a given health facility, and n is the facility sample size.
Study variables
The dependent variable in this study was the rate of antibiotic prescription for respiratory tract infections (RTIs), Additionally, we assessed the antibiotics prescribed based on the AWaRe grouping, which categorizes antibiotics into Access, Watch, and Reserve groups.
The independent variables in this study included patient-related factors, such as age and sex, as well as RTI-specific factors, such as the type of RTI, including acute bronchitis, acute otitis media, and acute upper respiratory tract infections (URTIs. The study also considered healthcare facility-related factors, such as the service level of the health facility and the availability of clinical guidelines. Furthermore, prescribers' training on antibiotic use was examined as a potential factor influencing antibiotic prescription practices.
Data management and analysis
The patient records and interview responses were collected into predesigned templates in Kobo collect software, downloaded as excel sheets. The responses in the excel sheet from the interview of prescribers were applied to patient’s records excel sheet. This was done in such way that if more than 75% of the prescribers accepted that a certain health system factor applied to their facility, then it was applied on all its patients. For-example if 75% of the prescribers accepted that there was no access to UCG then every patient at that facility was assumed to be diagnosed when there was no access to UCG by the prescriber. The resultant excel sheet was exported into Stata software version 14.0 (StataCorp, College Station, Texas, USA) for analysis.
To assess the appropriateness of antibiotic prescription for RTIs, we compared the prevalence of antibiotic prescription for individual and generalized RTIs against optimal values recommended by the World Health Organization (WHO) and Europe. For example, the optimal values for antibiotic prescription include less than 20% for acute tonsillitis, < 20% for acute otitis media, less than 20% for acute upper respiratory tract infections (ARTIs), < 30% for acute bronchitis, and 90–100% for pneumonia [21]. Rates above the optimal values were to indicate inappropriate prescribing of antibiotics for RTIs. We calculated the index for percentage encounter with antibiotics by a method described by [22]. Index of antibiotic prescription = WHO/INRUD Optimal value of antibiotic prescription/Observed value.
The rate of appropriate prescribing = observed indices/optimal indices) X 100 [22] the optimal indices for all WHO/INRUD prescribing indicators is 1. The rate of inappropriate prescription = 100- rate of appropriate prescribing.
The patient and institutional factors that can be used to predict antibiotic prescription were determined by modified Poisson regression analysis. A bivariate analysis was performed and the results for association between each individual variable with antibiotic prescribing were summarized into frequencies and crude prevalence ratio at a confidence interval of 95% and P-value. A multivariate analysis was then performed by including all the factors with p-value < 0.05, however all the sub-variables under type RTI diagnosis were included irrespective of their P-value at bivariate analysis. Any variables with a p value that is less than 0.05 after adjustment was considered to be associated with antibiotic prescription among RTI out-patients in public health facilities in Jinja City.