TB routine examinations in Yogyakarta lung clinics had high sensitivity and specificity. Only 4.1% of 339 consecutively enrolled study participants who were initially not diagnosed with TB later turned out to have TB. To our knowledge this study is the first report addressing sensitivity and specificity of TB diagnosis under routine conditions in lung clinics in Indonesia. Follow-up as a part of a composite reference standard to assess diagnostic test characteristics has been successfully used in many different settings [16, 17, 20, 21].
The diagnostic sensitivity and specificity in this study are higher compared to the study from Kwazulu-Natal, South Africa . All 12 patients in our study who were bacteriologically negative but clinically diagnosed with TB, had clinical improvement during the TB treatment. Boehme et al. found that only 67 out of 138 patients who were clinically diagnosed with TB despite negative bacteriology improved after TB treatment . Using culture as reference test, a study in smear negative TB patients in Pakistan showed that the clinical diagnosis was correct in 80% . In a retrospective study in Taiwan, 87% of smear negative patients who were clinically diagnosed with TB were confirmed to have TB based on their response to therapy . Some factors may explain this: Indonesia has higher TB prevalence which results in higher pretest probability; the lung clinics conducted multiple tests in the first examination, thus the physicians had more data to support correct diagnosis; and Indonesia has lower HIV prevalence , which contributes to the higher sensitivity of sputum microscopy and CXR.
In Yogyakarta, Indonesia, the lung clinics had the facilities needed to establish a TB diagnosis, meanwhile some of the primary health centers did not have laboratories facilities for sputum smear microscopy or CXR, thus they needed to send samples or patients for microscopy or CXR examinations to the lung clinics or higher-level hospitals. In addition, the lung clinics conducted both microscopy and CXR examinations in the first visit, while primary health centers followed the national diagnostic algorithm that recommends CXR examination only when sputum smears are negative. Therefore, the duration of diagnostic delay for TB diagnosis in primary health centers was significantly longer than in lung clinics , and the primary health centers may not have sufficient tools to establish the correct diagnosis. However, there have been no studies that investigated the sensitivity and specificity of TB diagnosis in the primary health centers in Indonesia, thus we could not compare the results with the lung clinics. Furthermore, the proportion of patients lost in primary health centers was higher than in lung clinics, due to this longer duration of the diagnostic process , although generally the distances between patients’ houses and primary health centers were closer than distances between their houses and lung clinics. Often patients needed to come more frequently to the primary health centers than to lung clinics to get diagnosed, which means more costs and time spent, thus causing patients’ reluctance to finish the diagnostic procedures . Indeed, there is a high gap (estimated around 47%) between notification and the estimation of incident TB cases due to undetected and underreported cases in Indonesia, including Yogyakarta . Therefore, if the TB diagnostic process in the lung clinics could be replicated in the primary health centers, it would improve the detection rate and reduce further TB transmission in the community.
Presence of all symptoms showed fairly good sensitivity and specificity as reported earlier from Tanzania , and it helped to establish an accurate diagnosis. Persistent cough of at least 2 weeks duration had the highest sensitivity, although other studies have reported low sensitivity and high specificity for persistent cough [17, 24]. Night sweats were associated with the highest specificity. Indeed, night sweats have generally been considered as a specific symptom for TB though the pathophysiology is poorly understood; [25, 26] night sweats are limited to the upper trunk in TB .
It is reasonable that a revision of diagnosis from non-TB into TB was associated with a presence of TB index cases in patient’s surroundings, as the presence of exposure to TB index cases is one of risk factors to contract TB [1, 9]. Culture examination is indeed considered the gold standard for the diagnosis of TB; 10 out of 14 patients whose diagnoses were revised had positive culture but negative smear. Nonetheless, the solid media used for culture in our study typically has a slow turn-around time, considerably longer than cultures in liquid media that have the additional advantage of improved sensitivity. Liquid culture may however have lower specificity due to higher contamination rates [28,29,30]. To increase sensitivity of LJ culture, 2–3 specimens per patient should preferably be examined to improve the diagnostic yield [31, 32]. In only 34 participants in our study, multiple specimens were submitted for culture. CXR has higher specificity and sensitivity compared to clinical symptoms, which corresponds with previous studies [24, 33]. CXR in TB may resemble other pulmonary conditions [4, 5], however, almost all non-TB patients whose diagnoses were revised into TB and who had CXR suggesting TB, developed TB after a short time. Our data illustrate that CXR is a helpful tool to screen for active TB. A previous study reported that ≥80% of patients with pulmonary TB have at least one among five different radiographic appearances .
According to the TB national guidelines in Indonesia, patients with negative smear should return or be followed-up to investigate their responses to the antibiotics given . However, in reality, patients often do not return to the health centers, and the patients who are lost are not tracked . Many negative-sputum-smeared patients in our study did not visit the lung clinics or other health centers after their initial diagnosis. In this situation, we note that follow-up helped to establish a correct TB diagnosis. A national study in 2017 revealed that around 20% of new cases were not diagnosed . Follow-up also provides an affordable way to evaluate the sensitivity and specificity of the diagnostic process in low resource settings.
This study is the first report on the proportion of false positive cultures in Indonesia. We suspect that the rate of false (positive and negative) culture results was 11.2%, while our estimate of false positive culture rate is 2%, which is comparable with earlier reports in high and middle/low-income countries [34,35,36]. To confirm suspected cross-contamination resulting in false-positive diagnoses, fingerprinting techniques help to identify similarity of strains from patients that have no epidemiological link. For resource-poor settings like Indonesia, these techniques are currently unaffordable but clinical laboratories can improve operational procedures by enhancing adherence to protocols , and participating in External Quality Assessment programs. Despite unavailability of fingerprinting analysis, we assessed the possibility of false-positive culture in our study through all available data. The lack of quality and quantity of samples might influence the quality of culture in these suspected false culture cases. Therefore, laboratory and other health care workers should instruct patients how to produce sufficient samples to increase the sample quality and detection rate . In a similar setting in Java, Indonesia, education provided for sputum collection for the patients was suboptimal .
To address the diagnostic delay, a rapid, reliable, and point-of-care diagnostic tool is needed. However, clinicians need to be conscious of the possibility of false-positive test results of any diagnostic tool, and clinical judgment remains important. Some tools are now under development, such as the loop-mediated isothermal amplification test for TB , and electronic nose . In the settings of lung clinics and primary health centers in Yogyakarta and elsewhere in Indonesia, Gene-Xpert TB/RIF is not readily available; it is available in some hospitals , and although the use of Gene-Xpert could increase the TB positivity rate , the relatively high cost of cartridges precludes extensive use if suspicion for drug resistance is low. False-positive test results for TB may also occur with PCR  or Gene-Xpert .
While waiting for an implementation of a rapid, reliable, and point-of-care TB diagnostic tool, optimizing the current routine TB diagnostic work is a reasonable option. The current national diagnostic algorithm that recommends CXR examination when sputum smear results are negative, was proven to delay the TB diagnosis, and one of the factors causing patient’s loss, although it was originally developed to improve the sensitivity and specificity of TB diagnosis . Our study showed that the diagnostic process in the lung clinics which employs clinical evaluation, sputum smear microscopy, and chest radiography all together had high sensitivity and specificity, and in the same time reduced the delay time of TB diagnosis, as indicated by a previous study . Therefore, an attempt should be done to replicate the diagnostic process in the lung clinics to other settings, such as primary health centers. If laboratory facilities could not be provided in the primary health centers, there should be a prompt referral system that enables patients to get microscopy and CXR examinations on the same day. This effort will reduce the rate of patient loss thus reducing the number of undetected cases and ongoing TB transmission.
There are some limitations in this study; in particular, it was performed in Yogyakarta province alone. The organization of lung clinics in Yogyakarta is however typical and representative for lung clinics in Indonesia and we therefore assume that the diagnostic sensitivity and specificity of TB routine examination is comparable to other lung clinics in Indonesia. Another limitation was that we used solid (LJ) culture medium, which has lower sensitivity compared to liquid (e.g. Mycobacteria Growth Indicator Tube) culture, and we used only one specimen for culture. We did not perform fingerprinting so that we could not further confirm the suspected false positive culture results, and we did not perform Drug Susceptibility Testing (DST) on every specimen; however, we did not detect study participants failing to respond to TB treatment and the MDR-TB prevalence in Indonesia is low , thus for that matter, the lack of DST did not confound our assessment.