Study design and setting
This was a retrospective cross-sectional analysis study of medical records on RTI outpatients in four randomly selected health facilities of different service level in Mbarara Municipality, Southwestern Uganda. The Municipality has thirteen government health facilities at different health care levels [14]. According to the 2014 National Population and Housing Census, the municipality had a population of 195,318 people (Statistics [15]. The selected health facilities included (i) Mbarara Regional Referral Hospital (MRRH) (ii) Mbarara Municipal Health Centre IV, (iii) Kakoba Health Centre III, and (iv) Nyamityobora Health Centre II.
Study population
The study population involved all outpatients diagnosed with RTI at selected Health facilities in Mbarara Municipality from 1st April 2019 to 31st March 2020 (prior to the Corona virus disease-COVID-19 pandemic). We excluded medical records missing vital information such as age and gender.
Sample size
Based on the WHO recommendation on how to investigate drug use in Health facilities i.e., at least 600 encounters to be included in a cross-sectional survey of medication use, and reviewing of at least 100 encounters per Health facility to describe or compare drug use by individual facilities [16].
We used a sample size of 780 encounters of RTI because it allowed the lowest facility to have at least 100 encounters reviewed as recommended by WHO [17].
Facility sample size
The individual facility sample size was calculated to put into consideration the difference in patient load. The regional referral had a highest patient load as such a higher sample was taken compared to other health facilities.
$$= \frac{Average\; monthly\; for\; facility}{{Monthly\; total\; of\; all \;facilities}} \times study\; sample \;size$$
MRRH
= (392/1030) 780 = 297
Mbarara Municipal Health Centre IV
= (274/1030) 780 = 208
Kakoba Health Centre III
= (230/1030) 780 = 175
Nyamityobora Health Centre II
= (133/1030) 780 = 101
Sampling technique
Multi-stage sampling at the facility and the medical record was used. The facilities were selected by level of service they included Mbarara Regional Referral Hospital and Mbarara Municipal Health Centre IV are purposively selected being the only ones at that level.
A simple random sampling technique was used to select Nyamityobora Health Centre II (7 health Centre II available) and Kakoba Health Centre III (4 health Centre III available).
Specific medical records were selected using a systematic random sampling method using the medical record number as a reference for the sampling frame
Sampling interval
$$SAMPLE\; INTERVAL = \frac{Annual \;Total \;RTI\; patients \;in \;facility}{{weighted\; sample \;size\; of\; the\; facility}}$$
MRRH = 4104/297 = 13
Health Centre IV = 2894/208 =13
Kakoba Health Centre III = 2052/175= 12
Health Centre II =1596/101= 15
The first record was chosen at random from the first sample frame by computer generated number and then every nth medical record encounter from this initial number was included in the study.
Data collection
Data was collected by the research team from hard copy medical records using a pretested web based google data extraction form with multiple checks to reduce entry errors. The form captured information from the out patients register which included: age, gender, month of treatment for RTI, route of administration, name of antibiotic prescribed, diagnosis, comorbidities, attendance (new vs old), malaria laboratory test, other medications, level of health facility, use of a generic name and if drug prescribed is on the essential drug list.
The clinician defined diagnosis was reclassified into three categories basing on the anatomical location i.e., upper respiratory tract infection, lower respiratory infection, and unspecified respiratory tract infection. Pharmacological classification was used to categorize the antibacterial drug prescribed.
Ethical consideration
The study was conducted in accordance with the Declaration of Helsinki. The study received ethics approval from research ethics committee of Mbarara University of Science and Technology (Approval Number: MUSTREC 09/7). Permission to collect data was sought from the district health officer and the head of each facility. Patient name or other identifiers were not captured.
Data analysis
A Microsoft excel sheet of the entered data was cleaned and data was then imported into Stata package 16.0 for statistical analysis.
Descriptive statistics was used to summarize the characteristics of patients with respiratory tract infection and presented as a percentage in each category. Logistic regression was used in the analysis to determine the association between the primary outcome (antibiotic prescribing) and the explanatory factors (age, gender, comorbidities, the period of visit, the total number of drugs prescribed, Upper respiratory tract infection diagnosis, health facility level, and laboratory test). Results of the analysis were presented as Odds Ratio (OR) at a confidence interval of 95% with P-value ˂ 0.05 considered significant.