Study design and setting
This was a cross-sectional study conducted between April and June 2018 among adults who presented with symptoms suggestive of TB predefined according to the world health organisation (WHO) criteria exiting 10 high volume public health facilities in Wakiso district in Uganda. It had a population estimated to be 2.5 million in 201815. About 60% of this population live in the urban areas14. Since 2010, interventions have been implemented to improve the case detection rate in the district. The interventions included: the DETECT child TB project; and roll out of national TB/HIV guidelines16. However, TB case detection rate is still at 57%16.
Wakiso district has seven health subdistricts namely Entebbe, Busiro South, Busiro North, Busiro East, Kyadondo North, Kyadondo South, and Kyadondo East. It hosts 67 public health facilities that offer free comprehensive primary health care services including screening and testing for Tuberculosis (TB). TB screening and testing services are expected to be offered at all care entry points especially outpatient, HIV/ART clinic, and Maternal Child Health departments.
The study population included adults exiting the public health facilities at two care entry points; HIV/ART clinic and outpatient department. We included in all adults aged 18 years and above with at least one symptom suggestive of TB predefined according to WHO criteria (i.e. cough for more than 2 weeks, night sweats, weight loss, and fever). In addition, for people living with HIV, we included patients presenting with cough of any duration. Patients who had sputum sent for TB investigation prior to current visit and TB patients who were already on treatment were excluded from the study.
We calculated a sample size of 255 clients using Kish-Leslie (1965) formula for cross sectional studies17. According to a study to evaluate TB diagnostic practices at five primary care health facilities in Uganda for one year, proportion of patients with symptoms suggestive of TB offered sputum examination was 21%3. Hence p =0.21, q=0.79, d (acceptable degree of error) =0.05, z (standard normal value corresponding to 95% confidence interval) =1.96.
Four health sub-districts were randomly selected from seven health sub districts. We then purposively selected 10 high volume health centres from the four health sub districts. They included Entebbe Hospital, Kasangati Health Centre IV, Wakiso Health Centre IV, Kajjansi Health centre IV, Buwambo Health Centre IV, Bweyogerere Health Centre III, Kiira Health Centre III, Nabweru Health centre III, Nsangi Health centre III, and Nakawuka Health Centre III.Each high-volume facility received an average of 98 (range 61-160) clients per day. The number of patients to be interviewed at each facility was determined by proportionate to size sampling. This depended on the average number of daily outpatient attendance over the last three months.
Patients were screened consecutively for interviews as they exited the different clinicians’ rooms at OPD and HIV clinics. An interviewer administered structured questionnaire was used to collect data on demographics, TB symptoms, and other clinical data relevant to TB. Participants were also asked if they had sputum and/ or a chest x-ray (CXR) requested by a healthcare worker at that visit. If sputum and/ or CXR had not been requested, they were referred back to the clinic staff for appropriate investigations.
The questionnaire used was pretested in two public health facilities and these were not part of the study sample. Research assistants were trained and supervised during data collection. Filled questionnaires were reviewed daily to check for completeness and consistency
Data were entered in Epidata version 3.1 database (EpiData database, Odense, Denmark). Data were cleaned and exported to Stata v14 (StataCorp LP, College Station, TX, USA) for analysis.
Continuous variables were described using means, or medians and the corresponding standard deviations or the interquartile ranges respectively while categorical variables were described using frequencies and percentages. The proportion of missed opportunities in TB investigation was calculated by dividing the number of patients with symptoms suggestive of TB who did not have sputum examination and/or CXR requested to rule out TB by the total number of patients with symptoms suggestive of TB.
At bivariate analysis, modified Poisson regression was used to identify factors significantly associated with the outcome. A p<0.05 was used as level of significance at the 95% confidence interval (95% CI) to test this association. Prevalence ratio was used as the measure of association. At Multivariate analysis, factors associated with the primary outcome at bivariate analysis were included in a multivariable model and adjusted prevalence ratios and 95% CI were estimated. A p<0.2 was used as a cut off to determine which variables to carry for multiple modified Poisson regression model to build the final model. Forward regression technique was used to build the multiple modified Poisson regression model while assessing the model variables for significance at p<0.05 and 95% CI. An adjusted R2 was generated for the final model to determine to what extent the factors were associated with the outcome of interest.
Ethical approval was obtained from Makerere University School of Public Health Research and Ethics Committee (FWA00011353). Wakiso District Health Office provided approval and permission to perform the study in the public health facilities. Informed consent was obtained from all participants before divulging any information and collection of data. All methods were carried out in accordance with relevant guidelines and regulations.