The record review and patient interviews confirmed that CXR is the preferred first investigation for TB in the three hospitals. This does not necessarily reflect a systematic approach to TB screening and diagnosis, but probably mirrors the fact that CXR is a common general screening and diagnostic tool for respiratory illnesses, which is widely available and used in general hospitals. About two thirds of all patients with abnormal CXR were further investigated for TB with SSM.
Among persons who had a CXR examination, the proportion with any abnormality was 65%. The proportion was highest among those over the age of 65
years and among men, consistent with findings of the first national TB prevalence survey in Vietnam [3
Using any CXR abnormality as screening criterium for further TB diagnosis in these hospitals implies TB testing of the majority of those who have a diagnostic CXR on any indication, and thus TB testing of a large proportion of all patients in respiratory and other out-patient departments (OPDs). On the one hand, this can be expensive and overburden laboratories. However, the TB yield among these patients is high, even among people who do not have typical TB symptoms, as demonstrated in the national TB prevalence survey in Vietnam and elsewhere [3
In this study, the proportion of TB cases among people with abnormal CXR was 20% and the sputum-smear positivity ratio was 13.5%. This is double the culture positivity ratio among people with abnormal CXR in the TB prevalence survey in Vietnam (6.6%) [3
]. There were large differences in the positivity of abnormal CXR who had sputum examination in the three hospitals, 27.3% in Hue, 4.5% in Bachmai hospital and 2.8 in Choray hospital. The reasons for higher proportions of smear positive among abnormal CXR who had sputum examination in Hue than in Bachmai hospitals and Choray hospital may be explained by the fact that Hanoi and HCMC have TB hospitals (National Lung hospital and Hanoi TB and Lung disease hospital in Hanoi, and Pham Ngoc Thach hospital in HCMC), while Hue has no provincial TB hospital, only a provincial social disease and prevention center, so many TB suspects go to general Hue hospital for examination. Another reason for the difference in positivity may be that the study sample came from the general OPD in Choray while the study sample came from the chest OPD in Hue and Bachmai.
The finding of more than 30% of people with abnormal CXR not being investigated with SSM, in combination with a high proportion of TB among those investigated suggests that there are still missed opportunities to improve early TB detection in the hospital sector in Vietnam. Aiming for SSM in close to 100% of those with abnormal CXR findings would likely improve early TB detection. Using a more sensitive test than SSM, such as GeneXpert MTB/Rif would improve case detection further, while also reducing the number of smear-negative patients going through a complicated diagnostic work-up, which has low diagnostic precision, especially if the final diagnosis is based on CXR finding and clinical picture only [7
Is it rational to do CXR before a bacteriological TB test in hospitals? There are a couple of arguments in favor: First, CXR is a good tool for ruling out pulmonary TB. The high sensitivity of CXR to detect culture positive TB has been demonstrated also in Vietnam [3
]. Second, TB symptoms are not specific [9
]. Therefore many possible differential diagnosis need to be considered, and CXR is a screening and diagnostic tool for several respiratory conditions. Third: final CXR results can normally be made available more quickly than results of 2–3 SSM examinations, minimizing the risk of delay and defaulting during the diagnostic process. Fourth, CXR screening can minimize the burden on laboratories. A large number of people come to hospital OPDs, and a high proportion of them have cough. If these hospitals were to strictly follow the diagnostic algorithm recommended by the Vietnam NTP, a large number of SSM would be required, and then CXR examination for those with negative SSM, which would be the vast majority. This would lead to a large burden for the laboratory of these hospitals and not much less burden for the radiology department. However, a similar argument could be made about the burden on the radiology department when CXR is the primary test. Furthermore, if a very high proportion of patients have an abnormal CXR, most patients would anyway need to have both CXR and SSM.
Arguments against using CXR as a screening tool include higher cost than SSM (at least for conventional X-ray), and lack of standards and quality control of x-ray machines, films and reading [11
]. In addition, in the absence of a highly sensitive and specific confirmatory TB test, it can generate false positive smear-negative TB cases, even if backed up by good diagnostic practices [12
]. However, the CXR pros-and-cons equation changes with the potential for introducing GeneXpert MTB/Rif, a more sensitive test than SSM, more specific than CXR, but much more expensive than both [7
]. Pre-screening with CXR to minimize use of Xpert MTB/Rif would save more costs than pre-screening before SSM. Also, the risk of generating many false positive smear-negative TB cases through initial CXR screening would diminish with the use of XpertMTB/Rif due to the capacity of the test to both rule in and rule out TB. There are already diagnostic committees in place in the three hospitals, which have probably contributed to the low proportion of smear-negative TB cases found in the present study. However, diagnosing smear-negative TB based on CXR and clinical picture is always imprecise. Our study has some limitations. Firstly, the study was designed to describe current practices, and not to assess accuracy, sensitivity and specificity of the different diagnostic pathways. The study did not include a validation of diagnosis. Therefore, the precision of the procedures and the proportion of false negative and false positive cases are unknown. This study did not assess treatment uptake and outcomes among TB patients diagnosed in hospitals. This will be part of the subsequent project evaluation.