Prevalence rate of undiagnosed tuberculosis in the community in Ethiopia from 2001 to 2014: systematic review and meta-analysis

This review sets using published studies to estimate the prevalence of undiagnosed smear positive PTB in community in Ethiopia. Studies were searched using MEDLINE/PubMed, ‘Cochrane’ library and Google scholar databases as well as reviewing reference list of previous studies on Tuberculosis in Ethiopia. The search was done following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement guideline [13]. The electronic search was performed for the MEDLINE/PubMed database using the following Medical Science Heading (MeSH) terms: (“tuberculosis”[MeSH Terms] OR “tuberculosis”[All Fields]) OR TB[All Fields] OR MTB [All Fields] OR (“Mycobacterium tuberculosis”[MeSH Terms] OR (“mycobacterium”[All Fields] AND “tuberculosis”[All Fields]) OR “Mycobacterium tuberculosis”[All Fields]) AND (“community”[MeSH Terms] OR “community”[All Fields]) OR (“population”[MeSH Terms] OR “population”[All Fields]) OR (“survey”[MeSH Terms] OR “survey”[All Fields] OR (“screening”[MeSH Terms] OR screening [All Fields]) AND (“Ethiopia”[MeSH Terms] OR “Ethiopia”[All Fields]).The study limited only English language studies and human being category.

Studies were selected for this review if conducted in Ethiopia. The studies were assessed by two independent researchers against the following inclusion criteria:-1) Community based cross sectional study done in the community or population; 2) The age groups > 14 years; 3) reported type of laboratory tests performed; 4) Study using sputum sample; 5) Pulmonary tuberculosis;7) studies that reported quality assurance methods. Other studies reported the knowledge and practice of individuals towards TB disease and investigated patterns of drug resistance only were excluded. It was also not included reviews that reiterated findings from the already included studies not to repeat used the same data in a single review which is not meaningful.

The data extraction was done by two researchers (BA, GM) using a standardized and pretested format. Data were extracted using Microsoft Excel and includes: title, first author, publication year, year of survey, design of the study, regions of study (study site in the country),study base (population-based or Community based), sample size, number of suspected case, type of laboratory test (Culture or Acid Fast Bacilli (AFB) Microscopy), age group of study participants, number of newly diagnosed smear positive PTB, number of bacteriologically confirmed (old Vs. new) tuberculosis, actively Vs. passively diagnosed (on treatment). The point prevalence rate of undiagnosed smear positive and total bacteriological confirmed point prevalence of PTB was determined for each study. Disagreements on data extractions between the two investigators were solved by discussion and consensus.

The quality of the included studies was evaluated by two reviewers using Joanna Briggs Institute (JBI) appraisal tool for prevalence studies [14]. These assessment criteria include 11 criteria with three potential responses: yes, no and not mentioned. The criteria are: − 1) was the sample frame appropriate to address the target population? 2) Were study participants sampled in an appropriate way? 3) Was the sample size adequate? 4) Were the study subjects and the setting described in detail? 5) Was the data analysis conducted with sufficient coverage of the identified sample? 6) Were valid methods used for the identification of the condition? 7) Was the condition measured in a standard, reliable way for all participants? 8) Was there appropriate statistical analysis? 9) Was the response rate adequate, and if not, was the low response rate managed appropriately All assessments were entered into pre-performed and standardized data extraction forms. Studies were evaluated for quality by using these indicators; those with medium (fulfilling 50% of quality assessment criteria) and high quality were included for analysis. No study was excluded because of poor quality rather excluded due to one or more of the exclusion criteria described above.

The extracted data were entered and analyzed using Compressive Meta-analysis version 2.2 software. The original articles were described using forest plot and table. Since there was heterogeneity among studies, random effect model was used to compute the pooled prevalence rate of undiagnosed smear positive and total bacteriological confirmed PTB. Random effect model is more conservative than fixed effect model and takes into account any heterogeneity inherent in the meta-analysis. The estimated pooled prevalence rate with its 95% confidence interval (CI) was presented.

A potential source of heterogeneity across studies was evaluated by Cochrane Q test (presence of heterogeneity) and I²statistics (amount of heterogeneity). The existence of heterogeneity was verified using Cochrane Q test (P < 0.10 indicates statistically significant heterogeneity) and I² test that measures level of statistical heterogeneity between studies (values of 25, 50 and 75% are to mean low, medium and high heterogeneity respectively). The Egger weighted regression and Begg rank correlation test methods were used to statistically assess publication bias (P < 0.05 was consider as suggestive of statistically significant publication bias).

The search strategy retrieved 439 potential articles, of which 23 were screened as full text articles and nine studies comprising of 237,648 individuals were found eligible and included in the analysis (Fig. 1). All were community based cross-sectional studies and including PTB cases with a study population ranging from 12,149 individuals in Addis Ababa [15] to 47,478 in Amhara region [16] and conducted from 2001 to 2014.

The studies were conducted in different regions of the country; Southern Nations, Nationalities and Peoples (SNNP) [17], Addis Ababa [15], Amhara [16, 18], Oromia [19, 20], Afar [21] and Tigray region [22]. One study was conducted across regions [23]. All studies used questionnaires for screening of symptoms of PTB followed by diagnostic testing of presumptive PTB patients [15‐22]. One study reported chest X-ray as additional screening tool [23]. Among the articles included in the analysis, four studies [15‐18] were use Microscopy (ZN or Florescence) for bacteriological confirmation of PTB and the rest five studies combined Microscopy with Culture [19‐23]. Three studies provide the information on median duration of cough ranging 13 week [19] to 52 weeks [21]. Only two studies assessed the TB-HIV co-infection one with 1.9% [22] and another with 5.5% [19]. Study characteristics are summarized in Table 1.

The included articles exhibited high heterogeneity according to Cochrane Q test (Q test p = 0.001) and I² test (I² = 88.3%), which is indicative to using random effects model. The Egger weighted regression statistics (p = 0.08) and Begg rank correlation statistics (p = 0.2) indicated no evidence of publication bias (Fig. 2). As the numbers of the eligible studies included were limited, subgroup analysis to explore the source of heterogeneity is not visible in this systemic review.

The pooled prevalence rate of smear positive undiagnosed PTB (active smear positive PTB detection) from the random effects method of all studies included was 0.11% (95% CI, 0.06–013%, p < 0.001) (Fig. 3). The point prevalence estimate (per 100,000) of undiagnosed smear positive PTB in all the studies included was 79.7 (95% CI; 56.3–112.8).