In this cross-sectional study, we found that a median time from admission until diagnosis was 6 days, with 33.2% of patients being diagnosed before 6 days, and 45.6% in more than 6 days. The main factors associated with diagnostic delay were extra-pulmonary TB and negative sputum smear.
There is no agreed definition of what constitutes an "acceptable" delay. Ward and others  suggests that this interpretation should depend on the local health services and the local epidemiological situation, with a shorter delay to be expected when incidence is high. In a systematic review of literature , the reported ranges of average (median or mean) health system delay were 2-87 days for both low and high income countries. However, considering only in-hospital studies, the median interval from admission to initiation of TB treatment is in agreement with our results, varying from 4 to 12.5 days [14–18].
Despite a high incidence of TB, as suggested by Ward and others , 45.6% of patients in our study had diagnosis delayed by more than a week. In a population-based study in Canada, treatment was initiated after a week or more in 30% of all patients with pulmonary TB , which could be attributed to the relative rarity of active TB in this country, resulting in a lack of awareness of TB. However, in our study delayed diagnosis was related to extra-pulmonary TB and smear-negative pulmonary TB, which frequently carry diagnostic difficulties. In accordance with other studies [9, 16, 18–21], extra-pulmonary TB, including pleural TB, and smear-negative pulmonary TB cases were found to be associated with diagnostic delay. A cross-sectional survey of newly identified adult patients with TB demonstrated that patients with extra-pulmonary TB or smear-negative disease are significantly more likely to be hospitalized and are also more likely to have experienced treatment delay . In a referral hospital in Rwanda, smear-negative pulmonary TB and extra-pulmonary TB were risk factors for a longer health system delay .
Current smoking was associated with an earlier diagnosis in our study. According to previously described in literature [22–24], smoking TB cases were significantly more likely to present with a cough and difficult or labored breathing, and less likely to present with only extrathoracic TB than never-smoking TB cases. They were significantly more likely to have upper-zone involvement, cavitation, positive bacilloscopy and sputum culture, and pulmonary TB than never-smoking TB cases.
Our analysis revealed that the presence of cough was associated with a faster diagnosis. In a retrospective study conducted in an emergency department, the presence of cough was related to more rapid isolation and treatment . However, in other studies [9, 26, 27] cough, especially chronic, is observed as a cause of late presentation of patients to the health system. Nevertheless, once patients are in the health care system, this kind of complaint may give rise to earlier suspicion of TB, and consequently, earlier diagnosis and treatment.
We also found that the complaint of night sweats was more frequent in the group of patients diagnosed less than 6 days after admission. In a cross sectional hospital based survey in Tanzania , not knowing that night sweat is one of the TB symptoms was associated with patient delay in females. Regarding to our finding, it is possible that the presence of night sweats raise awareness of TB among health care workers.
Cavitary disease and milliary pattern on chest x-ray were associated with an earlier diagnosis in our investigation. A similar study has also shown that radiographic changes typical of TB, like apical infiltrates and cavitation, were previously associated with more rapid treatment .
Patients with HIV infection included in our study experienced some delay in diagnosis of TB, although this finding was not statistically significant. HIV is known to delay TB diagnosis due to non-specific results and atypical clinical presentation [28, 29]. However, our findings are in agreement to other studies [5, 30, 31] that found no association between HIV positivity and delayed diagnosis. In a previous investigation, HIV-positive TB patients suffered more serious symptoms when TB was diagnosed, which prompted them to visit the hospital and increased the suspicion of TB by the clinician .
Our data did not show an increased mortality in patients with retard in diagnosis. Controversy still exists on the role of delay in TB diagnosis and treatment in the mortality rate. Although some authors demonstrated an association between diagnosis delay and mortality [32–35], in other studies delayed diagnosis was not a risk factor for mortality [16, 33, 36].
One of the limitations of this study is that we evaluated only health system delay, and not patient delay. Furthermore, patients were recruited from a single hospital, and the results may thus not apply to other settings. Besides, retrospective design has inherent limitations. The information obtained retrospectively from chart review is probably not as complete and accurate as when data collection is done prospectively. In spite of these concerns, the study of factors associated with delayed diagnosis is important as it has an impact in transmission dynamics of TB. Therefore, to identify the sources of delays is a critical issue for an effective TB control program.
In conclusion, we demonstrated that the median delay in TB diagnosis in this setting is 6 days, and the factors associated with this delay in multivariate analysis were extra-pulmonary TB and negative sputum smear. Reducing these delays may require increase of diagnostic awareness in health care professionals, and a review of health service practices. Future studies should focus on attempt to explain the reasons of diagnostic retard in the patients with the characteristics related to delay in this study. In addition, studies on health care seeking may be warranted in this setting.