Abstract:
Background: In Ethiopia, one-third of tuberculosis (TB) cases are missed each year, and delay
in the diagnosis of TB is hampering the whole cascade of care. Recent surveys have showed
very high prevalence of confirmed TB, even among those with a low duration of cough with
more than 50% having a bacteriologically confirmed pulmonary TB (PTB). Furthermore, there
is an increase in the incidence of smear-negative PTB patients who may potentially serve as a
source of infection.
Aim: This study was aimed to evaluate active and passive case finding interventions aiming at
halting the burden of TB in Ethiopia, by assessing the burden of TB and challenges related to
screening and diagnosis; prevalence and predictors; impact of early chest X-ray (CXR) on delay
among TB cases; and diagnostic performance of CXR among health care-seeking people
screened for cough of any duration.
Methods: A two-phase implementation study was conducted in four administrative regions in
Ethiopia. During Phase I, a facility-based cross-sectional study was conducted in seven
healthcare facilities selected from four administrative regions selected by stratified random
sampling procedures to assess TB burden and identify challenges related to TB screening and
diagnosis. In this Phase, data of 1,059,065 patients were included from the outpatient
department, HIV clinic, diabetic, and reproductive and child health units. Data were collected
from October to December 2018 using a retrospective review of three years' facility data (2015
to 2017) supplemented by a semi-structured interview with purposively selected health care
workers and heads of the health facilities.
During Phase II, a facility-based cross-sectional study was conducted among all consented
patients who attended healthcare at outpatient departments, reproductive and child health units
in the study facilities. The patients were screened for cough of any duration and grouped into
two groups: group I included patients with cough ≥ 2 weeks, and group II included patients with
cough of < 2 weeks. Both groups underwent CXR followed by sputum Xpert MTB/RIF assay
or smear microscopy tests depending on the type of test available in the facility where
participants recruited. CXR results were read and interpreted by radiologists. Pregnant women,
HIV patients on antiretroviral therapy, and patients with diabetes were exempted from CXR;
xvii
thus, evaluated for TB as group III. TB prevalence was calculated across the three groups. For
Phase I, descriptive analysis was done using Excel. For Phase II, analysis was performed using
Stata 14.0. The TB case predictors were analyzed using modified Poisson regression to compute
adjusted prevalence ratio (aPR) with a 95% confidence interval (CI). Delays were calculated
using median and inter-quartile range (IQR) and summarized into the following; patient delay
(first onset of cough to first facility visit, ≥ 15 days), diagnosis delay (first facility visit to date
of PTB diagnosis, > 7 days), and total delay (first onset of cough to date of PTB diagnosis, > 21
days). Kruskal-Wallis and Mann-Witney tests were used to compare the delays among
independent variables. Evaluation was done to check if early CXR in patients with cough
irrespective of duration may reduce TB diagnostic and treatment delays. The diagnostic
performance of CXR as a TB triage test before Xpert MTB/RIF assay among people with
respiratory complaints. Area under the curve (AUC), sensitivity, specificity, and predictive
values of CXR against Xpert MTB/RIF as the reference standard and further comparison with
clinical symptoms.
Results: The Phase I initial assessment data showed that a total of 1,059,065 patients visited the
healthcare facilities in three years. Of these, 978,480 (92.4%) were outpatients among which
20,284 (2%) were presumptive TB cases (with 14 days or more cough) and of which, 12.2%
(2483/20,284) had TB. About a quarter, 604 (24.3%), were smear-positive pulmonary TB
(PTB), 789 (31.8%) were smear-negative PTB, and 719 (29%) were extra-pulmonary TB. The
study showed that, TB screening was integrated into HIV clinic, outpatient department, diabetic
clinic but not with the reproductive and child health clinics. High patient load, weak TB
laboratory specimen referral system, and shortage of TB diagnostic tools including Xpert
MTB/RIF assay and chest X-ray, were the major challenges in the screening and diagnosis of
TB.
During Phase II, the study screened 195,713 people who sought healthcare for cough of any
duration, of which, 2647 (1.4%) reported cough symptom of any duration, of whom 1853 (70%)
underwent further diagnostic tests as they fulfilled the criteria for presumptive TB. Overall,
309/1853 (16.7%) were diagnosed with PTB. Of the 309 positive cases, 81.2% (251/309) were
in group I (cough ≥ 2 weeks), 44/309 (14.2%) in group II (cough of < 2 weeks), and 4.5%
(14/309) in group III (CXR exempted). The majority of predictors of PTB were age group of
xviii
25-34 (aPR = 2.0 [95% CI 1.3-2.8]), history of weight loss (aPR = 1.2 [95% CI 1.1-1.3]), and
TB suggestive CXRs (aPR = 41.1 [95% CI 23.2-72.8]).
Among 309 diagnosed PTB cases, the median (IQR) of patient delay, diagnosis delay, and the
total delay was 30 (16-44), 1 (0-3), and 31 (19-48) days, respectively. Patients’ delay contributed
a great role in the total delay, 201/209 (96.2%). Median diagnosis delay was higher among those
who visited health center, diagnosed at a facility that had no Xpert MTB/RIF assay, radiologist,
or CXR (p<0.05). Factors associated with patients’ delay were history of previous TB treatment
(aPR [adjusted Prevalence Ratio] = 0.79, 95% CI: 0.63-0.99) and history of weight loss (aPR =
1.12; 95% CI: 1.0-1.25). Early CXR screening for cough of < 2 weeks duration significantly
reduced the patients’ delay and thus the total delay.
Of the 2647 cases reported of having a cough, 757 had both CXR results and Xpert, hence
included in this analysis. Overall, 152/757 (20%) were Xpert confirmed PTB. The sensitivity
and specificity of CXR findings were 77.6% (95% CI: 70-84%) and 97.5% (95% CI: 96-99%),
respectively, when compared with the Xpert standard. The AUC of the CXR was 87.6% (95%
CI: 84-91%). Inclusion of night sweats as initial symptom screening before CXR improved the
sensitivity at 91.7% (95% CI 84-96, with the AUC of 93% (95% CI: 90-97%).
Conclusion: The burden of TB was high in the study setting, and frequent interruption of
laboratory reagents and supplies hampered TB screening and diagnostic services during passive
and active screening. A large number of PTB cases were diagnosed among people who sought
health care when screened for cough of any duration. CXR is an important tool for triaging and
screening for PTB before Xpert MTB/RIF assay among people with respiratory complaints in
Ethiopia. Early screening using CXR minimized delays in the diagnosis of PTB among people
with cough of any duration. Patients’ delay was prevalent and contributed significant role in the
delay of TB cases. Screening by cough of any duration and/or CXR among people seeking
healthcare along with ensuring the availability of Xpert MTB/RIF assay and skilled human
resources at primary healthcare facilities are important to reduce patient and diagnostic delays
of PTB in order to end TB in Ethiopia.