Main findings
Overall, the pooled mean total delay in the diagnosis and treatment of pulmonary tuberculosis was 88 days. The most important and largest contributor to total delay in the diagnosis and treatment of pulmonary tuberculosis was patient delay with a pooled mean delay of 81 days followed by doctor’s delay and treatment delay with pooled mean delays of 30 and 8 days respectively. The other hybrid delays including diagnostic and health system delays reported pooled mean delays of 70 and 42 days respectively. All pooled mean delays were considerably heterogeneous.
Some study level characteristics may account for some of the extreme heterogeneity found among studies measuring delays. Excluding ePTB patients increased mean doctor’s delay by 45 days on average, non-use of chest x-ray and conducting studies in HIC decreased PMTxD by 20 and 22 days on average, respectively. Other potential factors that may increase delays were: non-use of sputum microscopy in studies assessing patient delay; increased proportion of males enrolled and using survey method of data collection in the assessment of diagnostic delay; and increased mean age of patient enrolled in the assessment of total delay.
Strength and weaknesses
We searched only PubMed implying a potential for missing studies that had not been indexed in PubMed. We however, believe that the number of studies included in this systematic review more than triple number of studies included in any other systematic review of delays in diagnosis and treatment of PTB and missing studies are likely to have minimal impact on findings from our study.
Our study tested covariates with high variability across studies, making interpretation of the meta-regression easier. However, there is still a potential for ecological bias. For characteristics with both study-level and within-study individual patient-level values, the results from the meta-regression may not translate to patient-level. For example, for every 1 year increase in mean age across studies, we found that there was a corresponding increase of about 8 days in mean total delay. This may not imply that older people were likely to exhibit a greater total delay unless this has been demonstrated within studies. Thus, some caution is warranted in extrapolating our findings in the subgroup and meta-regression analyses to individual patient level.
Furthermore, the lack of statistical significance may not imply a lack of relationship between covariate and means of delays. This may be attributable to study precision even though we have included a fairly large number of studies. Also, the number of covariates that we have included was relatively modest and missing covariate values were infrequent. Covariates were also pre-specified in the protocol thereby, avoiding data dredging.
Our study did not take into consideration co-morbidities for example, HIV infection. Co-morbidities may likely be relevant predictor of delays in the diagnosis and treatment of pulmonary tuberculosis.
Mechanisms and results in context of other studies
Tuberculosis is a chronic disease that requires a high index of suspicion by both the patient and the healthcare providers to enhance prompt diagnosis. It is conventionally recommended that cough lasting two or 3 weeks should arouse the suspicion of affected person to seek help, [
15] and for the healthcare provider to screen by at least conducting a sputum microscopy test. Our study reveals there exist a considerable average delay before patients seek help from the healthcare providers. This may not be unconnected with the insidious-onset nature of the disease. More than 10 weeks on average is wasted before patients consult the appropriate health facilities. In some settings, it could be as a result of self-medication or seeking help from traditional healers or other providers outside the formal health system. Sreeramareddy et al. reported a lower average of 4 weeks for patient delay, [
10] which could be accounted for by the differences in method of analysis. The authors presented a descriptive summary which did not weight the studies whereas our meta-analysis provided an average weighted by the inverse variance method. Secondly studies assessing delays that included ePTB was excluded from their primary analysis and thus may partly account for higher PMPD in our study although the influence of ePTB on PMDD remains largely unclear. We had postulated that inclusion of ePTB would likely delay diagnosis but our findings from the meta-regression analysis shows the reverse and supports lower doctors’ delay when other factors were kept constant. However, studies enrolling ePTB were disproportionately (80%) from LMIC which was associated with higher PMDD. Only our study was strict with the definition of delays compared to other systematic review [
6‐
10].
Only for PMToD did mean age of study participants significantly increased the delay. An increase in study participants mean age of 1 year increased the mean total delay by about 8 days. Consistent with our finding were reports of within-study relationship between age and delay measures generally, including total delay. This is possibly a reflection of poorer socio-cultural and economic supports for the older persons in the settings where the studies were conducted. One within-study evidence demonstrated longer health system and total delay for patients older than 60 years [
16]; also longer diagnostic and treatment delays for patients older than 65 years was reported in another study [
4].
There are however, inconsistencies regarding the influence of gender on delays from patient-level evidence demonstrated in previous studies. Even though an increase of 1 % in the proportion of males was associated with an increase of about 6 days in only mean diagnostic delay, several patient-level studies have demonstrated an association between male sex and longer patient delays, [
5,
17,
18] between female sex and longer patient, [
19‐
21] diagnostic, [
21,
22] health system, [
16,
23] total [
20,
21,
24] and treatment delays [
21]. Other studies reported no association between sex and patient, [
25‐
29] doctor’s, [
21,
26] health system [
21,
30] and total delays [
26,
31]. These discrepancies may reflect differences in health seeking behaviours and gender roles in different settings.
Inclusion of smear negative PTB patients did not independently account for variations in delay measures. We had postulated that smear negative PTB would likely increase the number of investigations that would be requested before confirming the diagnosis which would increase at least the diagnostic and/or health system delay. This was however, not strongly supported by our findings contrary to findings demonstrated in some patient-level studies. Although, smear negative PTB was associated with an increased pooled mean diagnostic delay of 62 days on average, this was not statistically significant in the model. Patient-level studies demonstrated that smear negative result was associated with diagnostic, [
4] and health system delays [
30]. However, use of chest x-ray may not have uniform influence on diagnostic delay. It could aid diagnosis and reduce delays in centres with relevant expertise. Only the mean treatment delay was significantly reduced by non-use of chest x-ray in the meta-regression. Some health facilities may not have Chest x-ray. Thus in such centres, diagnosis may be based solely on clinical acumen and sputum microscopy. In contrast to our finding, absence of initial chest x-ray examination was associated with longer health system and total delays in a patient-level study [
16].
Depending on the countries, studies conducted in the tertiary facilities may favour longer or shorter delays. For example, being diagnosed in a tertiary health facility may mean that patients had visited lower levels of healthcare delivery previously, implying longer delay. However, it could also mean that the tertiary facilities are more equipped to arrive at the diagnosis and institute treatment within a shorter duration, especially in settings where patients have direct access to the tertiary centres. The level of healthcare delivery did not significantly explain the variations in any of the delay measures but in the subgroup analysis, studies carried out in tertiary centres reported twice the PMDD compared to studies conducted in lower levels of healthcare delivery.
Studies conducted in the HICs were significantly associated with a decreased mean treatment delay. This may be a result of better access to anti-tuberculosis drugs in HICs than in the LMICs. However, Sreeramareddy et al. reported similar patient and health system delays for both high income and low income countries [
10]. The WHO regions where studies were conducted did not independently explain variations in the mean delays however there was a tendency for the EURO region to have a higher pooled mean diagnostic, treatment and health system delays than the high burden SEARO and AFRO regions in the subgroup analysis. Clinician’s familiarity with disease presentation and a higher index of suspicion may play an important role in early disease recognition in endemic settings compared to settings where disease prevalence is low.
Implications
Findings from this study implicates patient health seeking behaviour as the weakest link in attaining early diagnosis of pulmonary tuberculosis. Guidelines recommend that patients with a cough productive of sputum and lasting for two or 3 weeks should arouse a suspicion of pulmonary tuberculosis. The evidence demonstrates a 7 week gap between a target of 3 weeks and the average time it takes patients to turn up at the health facilities for proper assessment. Furthermore, the evidence also implicates the health systems which are expected to arrive at a diagnosis and commence treatment within a week.
The passive case detection strategy may not be optimal for early diagnosis of pulmonary tuberculosis. Considering the mode of disease spread, additional efforts such as high risk screening among high-risk exposed household relatives of index patients may improve early detection rate. Sputum sample assessment of contacts has been shown to detect PTB at an early stage in about 7 % of over a thousand contacts screened in a study [
32]. Moreover, the period of infectivity may be much higher than the patient delay because TB prevalence surveys revealed a higher burden of culture-positive or X-ray detected PTB in individuals without symptoms [
33]. There is need for further research on cost effective methods to enhance prompt reporting of chronic cough to accessible health facilities. Further operational research for example, on the use of simple and cost-effective algorithms for any cough presenting at health facilities may be warranted.