Health system delay in the treatment of tuberculosis patients in Ethiopia: a systematic review and meta-analysis


 Background: Delay in diagnosis and initiation of effective treatment associated with increase in morbidity, mortality and on-going person-to-person transmission in the community at large. In Ethiopia, several studies have been conducted regarding health system delay among tuberculosis patients. However, studies assessing the health system delay in treatment of tuberculosis patients in Ethiopia had inconsistent and inconclusive findings. Therefore, this systematic review and meta-analysis aimed to determine the pooled median time of the health system delay in the treatment of tuberculosis and its determinants in Ethiopia. Methods: We systematically searched from different databases: Google Scholar, Science Direct, PubMed, Embase, Scopus and Springer link databases for studies published from June 6, 1997 up to December 20, 2020. The quality of the studies was assessed using the Newcastle-Ottawa scale adapted for observational studies. Heterogeneity was evaluated using I squared statistic. We conducted a meta-analysis for the pooled median time of health system delay and its determinants using random-effects model in R version 4.0.3 software(for median estimation) and Stata version 14 (for metan). The pooled estimates with 95% confidence intervals (CI) were presented using forest plots. Results: A total of 14 studies which comprising 6161 patients satisfying a priori set criteria were included. Our meta-analysis showed that, the estimated pooled median time of the health system delay was 15.29(95%CI: 9.94–20.64) days. In the subgroup analysis, studies conducted from 1997 to 2015 the pooled median health system delay was 21.63(95% CI: 14.38-28.88) days, whereas studies conducted after 2015 the pooled median time of 9.33(95% CI: 3.95-14.70) days. Living in rural area (pooled OR: 2.42, 95%CI: 1.16-5.02) was significantly associated with health system delay. Conclusions: In Ethiopia, patients are delayed more-than two weeks in the treatment of tuberculosis. Being from rural residence was more likely to lead prolonged health system delay. Implementing efforts by targeting rural residence may help to shorten the health system delay and important implications for the success of tuberculosis control.


Introduction
Despite remarkable progress in TB control that has been achieved over the past year, it remains a global public health challenge [1]. About one-fourth of the world's population were affected by Latent Tuberculosis (TB) and TB is the first infectious disease killer [2]. In 2019, an estimated 1.2 million people die due to tuberculosis, including 208 000 HIVpositive people [3]. It kills more than five thousand children, women, and men each day [1,4]. In addition, globally, 3.3% of new TB cases and 17.7% of previously treated cases were assumed to harbor drug resistant-TB [3]. According to the 2020 Global tuberculosis report, Ethiopia is among the 30 high TB, HIV, and MDR-TB burden countries, with an Saharan African countries and the Middle East Asia showed that travel time for the return visit and being female was associated with health system delay [37,38].
Massive efforts were implemented during the era of the stop TB strategy to substantially reduce the global burden of TB by 2015 through universal access to diagnosis and treatment regardless of socio-economic barriers. However, the health system's delay remains a major impediment to effective TB prevention and control [39]. Several studies have reported health system delays and its associated factors among TB patients in Ethiopia; however, the findings on health system delays have varied or been inconsistent and inconclusive. With these variations of reports, and as far as in our search, we could not find any studies conducted on systematic review and meta-analysis on the extent and associated factors of health system delay in Ethiopia. Therefore, this systematic review and meta-analysis aimed to provide a pooled national estimate of the median time of health system delay and to identify its associated factors in Ethiopia. This evidence will be helpful for healthcare professionals, health-policy makers, and program managers to apply efficient interventions, and to improve effective tuberculosis control programs.

Reporting
The Preferred Reporting Items for Systematic Reviews and Meta-analyses (PRISMA) guideline [40] was used to report the results of this systematic review and meta-analyses. This systematic review and meta-analysis was registered with PROSPERO Registration number CRD42020220820.
published and unpublished.
Publication date: The authors included articles published from June 6, 1997 until December 20 th , 2020.
Exclusion criteria: Despite the above-mentioned eligibility criteria, articles which we were unable to access the full text after two email contacts with the principal investigator of the particular study were excluded from the analysis.
Studies done before 1997 were excluded from the review since Ethiopia launched the Directly Observed Treatment, Short-Course (DOTs) strategy in 1997 as part of the National Tuberculosis and Leprosy Control Program (TLCP) to treat TB patients [1]. WHO developed DOTS strategies to address major constraints on the achievement of global TB control targets. One of the strategies was expanding access to diagnosis and treatment through community TB care and public-private mix approaches aimed at engaging all care providers in DOTS implementation [39]. This might have an impact on reducing the number of days for the initiation of tuberculosis treatment.

Databases and Search Strategy
The comprehensive search for potential studies were conducted by two of the authors (KT and FW). The studies were published in English from June 6, 1997 up to December 20, 2020. We exhaustively searched the following databases: Google Scholar, Science Direct, PubMed, Embase, Scopus, and Springer link databases for all available studies using the search strategy by combining the keywords, Medical Subject Heading (MeSH) and free text terms described in (Table 3). In addition, the manual search of reference list of included studies were also reviewed to retrieve further studies. The titles and abstracts of the identified reports were used to exclude studies that did not meet the inclusion criteria in the first step. In the next step, for studies potentially eligible for inclusion, both authors (KT and FW) selected full-text studies based on eligibility criteria independently. The full articles of selected studies were screened to confirm eligibility and the reviewers discussed whether the studies should be included until a consensus was reached. If not, the disagreement was resolved by the decision of the third coauthor (DE). To identify unpublished and ongoing studies, we contacted researchers and experts in the TB field. We also requested unpublished data from organizations such as the Federal Ministry of Health, Government of Ethiopia), and Non-governmental organizations (NGOs) working on tuberculosis, but no one responded.

Outcome Measures and Data Extraction
The outcome of interest in this study was health system delay in the treatment of tuberculosis, defined as the time between the first visit to a health care provider and the start of treatment [12,42]. Therefore, included studies define operationally health system delay from the first visit to the health care facility to the initiation of treatment and report the median health system delay with the inter quartile range (IQR). All observational studies of patients receiving treatment for TB that recorded at least the median health system delay were included. We converted all time measures to days for studies that did not report delay measures in days. Delay reported in weeks were transformed into days by a multiple of 7; months by a multiple of 30. We used the median and inter quartile range of studies as the primary measure for meta-analysis. Because, all the included studies reported a median number of days with an inter quartile range. Eligible studies were selected using the pre-specified inclusion/exclusion criteria. All relevant information from the included studies were extracted independently by two (KT and FW) authors after data extraction checklist development. Two authors (KT, FW) screened titles and abstracts obtained from the database search and decided on eligibility. Following identification of potential studies, KT and FW made the final selection through review of full articles including the study design, participants, outcome variable, and publication year. Co-authors (DE and YAB) closely supervised the selection process. Finally, potentially eligible full-text articles that fulfilled the inclusion criteria were included in the review. We extracted data from each included study into a Microsoft excel spreadsheet: first authors, publication year, the region of the study conducted, data collection period, study participants, sample size, and the median with IQR of health system delay. We also extracted the predictor variables from the included studies for health system delay. Further data were also extracted on the health system delay among exposure categories and the odds ratio for the meta-analysis.

Quality Assessment Tool
Two reviewers (KT and FW) independently assessed the quality of the articles before including in the analysis. The Newcastle-Ottawa Scale (NOS) adapted for cross-sectional studies was used to assess the quality of the included studies [43,44]. The quality of each study was assessed in three sections using the following criteria: participant selection, comparability of the groups, and description of outcome assessment with a maximum of ten scores. A star was assigned to each point of the scale to categorize the studies into good, fair, and poor quality based on the NOS criteria. The first section scored a maximum of five stars and focused on the representativeness of the sample. The second section concerned with how the confounding variables were controlled with a maximum of two stars. The third section focuses on the study's outcomes and statistical analysis, with the possibility of earning three stars (Table 4). Finally, the average score provided by two reviewers was taken. A third author (DE) was brought in to resolve their disagreement over the assessment result. Articles with a score of seven or higher were considered to be of high quality. This cut-off point was considered after referring to previous literature [45].

Data Processing and Analysis
Data were extracted from each of the original studies using Microsoft Excel and then exported to STATA version 14 and R software version 4.0.3 for analysis. The pooled estimates A Systematic Review and Meta-analysis with 95% confidence intervals (CI) were presented using forest plots. The pooled median health system delay (in days) was estimated using a random effect model. The heterogeneity of the included studies was evaluated using I squared statistic [46]. In this study, significant heterogeneity was observed among the included studies (I 2 = 99.88%, p < 0.0001). As a result, a random-effects meta-analysis model was used. The median time of health system delay from different studies was pooled in a meta-analysis using R software. We also conducted a meta-analysis on factors in health system delay in the treatment of TB using STATA version 14 statistical software. The overall effect of factors related to delays was estimated from studies by conducting a meta-analysis based on ORs and 95% CI. Sub-group analysis was performed to identify potential factors that could explain the inconsistencies between effect sizes across the primary studies based on different variables (i.e., year of study, region of study conducted, sample size, and types of TB).

Search Results
A total of 1031 articles were identified during our initial search from databases and all retrieved studies were exported to the Endnote version 9 (Clarivate, Thomson Reuters, George Mason University) reference manager and then 477 articles were excluded due to duplication. We excluded 692 articles after examining the titles and abstracts, primarily because they did not have a relationship with the outcome of interest. Full-text evaluations were performed on 48 studies; of which 34 studies were excluded after reading the full text with the reason (Table 5). Finally, a total of 14 studies that satisfied the eligibility criteria were included in this systematic review and meta-analysis (figure 1).

Characteristics of Original Studies
The included articles covered 9 regions with two city administrations; the majority of the studies were conducted in Amhara region [7,11,[47][48][49][50]. Five studies [35,[50][51][52][53] enrolled pulmonary tuberculosis patients, three studies [7,8,47] enrolled smear-positive PTB patients, and the remaining six studies [9-11, 48, 49, 54] enrolled both types of tuberculosis patients. All of the articles were cross-sectional studies, and the sample sizes of the individual studies included in our meta-analysis ranged from 129 [35] to 875 [51] with a total of 6161 study participants. The studies included in this review were published between 2003 and 2019 (Table 1).

Meta-analysis
The smallest median health system delay among studies included in the analysis was 4 days [11] and the longest was 34 days [35]. The overall estimated pooled median health system delay was 15.29 (95% CI: 9.94-20.64) days. We used a random-effect model because the overall results of the I squared statistic shown high heterogeneity among included studies (I 2 = 99.88%) for median health system delay estimation. Therefore, we conducted subgroup analysis through sample size, year of study, region of the study conducted, and type of tuberculosis in order to explore possible causes of heterogeneity.  Google scholar Patient* AND "diagnostic delay*" OR "treatment delay*" OR "patient delay*" OR "health system delay*" OR "health service delay*" OR "provider delay*" OR "total delay*" AND tuberculosis* OR "pulmonary tuberculosis*" OR AND "associated factor*" OR determinant* AND Ethiopia* 398 Science direct "diagnostic delay" OR "treatment delay" OR "health system delay" OR "total delay" AND tuberculosis OR "pulmonary tuberculosis" AND "associated factor" AND Ethiopia 156 A Systematic Review and Meta-analysis

Databases Search terms and strategy Number of studies
Scopus patient* AND "diagnostic delay*" OR "treatment delay*" OR "patient delay*" OR "health system delay*" OR "health service delay*" OR "provider delay*" OR "total delay*" AND tuberculosis* OR "pulmonary tuberculosis*" AND "associated factor*" OR determinant* AND Ethiopia* 78 Springer link Patient* AND "diagnostic delay*" OR "treatment delay*" OR "patient delay*" OR "health system delay*" OR "health service delay*" OR "provider delay*" OR "total delay*" AND tuberculosis* OR "pulmonary tuberculosis*" AND "associated factor*" OR determinant* AND Ethiopia* 96 Table 4. Quality assessment result of the studies included in the systematic review and meta-analysis (Newcastle-Ottawa quality assessment scale) for Cross-Sectional Studies. Author

Subgroup Analysis
In subgroup analyses, studies conducted with a sample size of less than 350 reports a pooled median health system delay of 16.34 (95% CI: 6. 33-26.34), Whereas, studies conducted with a sample size of more than 350 with a pooled median health system delay was 14.65 (95% CI: 9.03-20.28) days. Based on the type of TB, studies that enrolled PTB patients had the highest pooled median health system delay 16.63 (95% CI: 16.63 (7.46-25.81) days, followed by those studies that enrolled all forms of TB patients 16.01 (95% CI: 6.02-26.01) days. Similarly, studies conducted from 1997 to 2015 reported the highest pooled median health system delay of 21.63 (95% CI: 14.38-28.88) days as compared to studies conducted after 2015 with a pooled median time of 9.33 (95% CI: 3.95-14.70) days (Table 2).

Factors Associated with Health System Delay
The data for each of the five associated factors were exported to Stata SE version 14 after extracting data on excel spread sheet. Ten studies [7,9,10,35,[48][49][50][51][52][53] assessed associated factors including patients' socio-demographic characteristics, clinical characteristics, and diagnostic modalities. Six studies were analyzed the relationship between sex and health system delay; four studies assess distance to the health facility with health system delay. HIV serostatus was mentioned in three articles; from this one article mention that being HIV positive was decrease health system delay as compared to HIV negative patients [48]. Type of TB or form of TB mentioned in three articles and from these two articles showed that those who were smear-negative and EPTB patients had prolonged health system delay [10,50] (Table 6). Three studies reported the association between residence and health system delay; one article revealed that patients living in the rural area had prolonged health system delay as compared to patients living in urban residences [35].
Our meta-analysis also showed that patients from the rural areas were more likely to have a prolonged health system delay with pooled OR of the studies (OR: 2.42, 95%CI: 1.16-5.02). However, in our meta-analysis, there was no significant relationship between health system delay and distance from the health facility (OR: 1.36 (0.72-2.55)), form of TB (OR: 1.39 (0.99-1.95)), HIV serostatus (OR: 1.07 (0.82-1.39)) and sex (0.96 (0.80-1.15)) ( Figure 2). Other socio-demographic and clinical as well as diagnostic modalities factors were not evaluated due to the lack of data in the individual studies.

Discussion
Even though a systemic review and meta-analysis done on patient delay in the diagnosis of tuberculosis patients in Ethiopia [55]; health system delays in the treatment of tuberculosis patients were not previously conducted. Thus, this meta-analysis attempted to estimate the pooled median health system delay and to review contextual factors associated with health system delay. The findings could be helpful for TB prevention and control programs; to effectively reducing delays for seeking prompt diagnosis and treatment. Our finding showed that the pooled median health system delay in the treatment of tuberculosis patients was 15.29 (95% CI: 9.94-20.64) days. These results suggest that even when patients seek care in a timely manner, significant time can be lost after their first contact with the healthcare provider. Our pooled analysis also indicated that residence was significantly associated with health system delay. Patients from rural residences were more likely to delay in the treatment of tuberculosis compared to those patients from urban areas (pooled OR = 2.42; 95% CI = 1. 16-5.02). This might be due to the fact that, access to health care services was particularly low in rural areas of Ethiopia where the majority of the population lives. Together with uneven distribution of health care professionals, results in little availability and poor quality of health care services in rural areas [56].
Our finding showed that the pooled median health system delay was lower as compared to the previous systematic review and meta-analysis done in LMIC, estimated average health system delay was 28.4 days [24], India the pooled median health system delay was 31 days [22], and a systematic review done in 78 countries, the pooled health system delay was 39.3 days [57]. This finding also lower than primary studies done in Vietnam 42 days [26], Taiwan 29 days [58], China 26 days [59], and Turkey 64.1 days [60]. The possible reasons for such discrepancy might be related to differences in the accessibility of health care service or variation of the infrastructure from country to country and it could be also due to improvements in the diagnostic capacities like the introduction of rapid molecular diagnostic tests especially, Gene Xpert technology which effectively shortened the health system delays [34,61]. Our finding was in line with studies conducted in Hong Kong 20 days [62] and Nepal 18 days [33]. However, the pooled estimate median health system delay was higher than studies done in Vietnam 7 days [26], China 4 days [63], and Uzbekistan 7 days [64]. This might be due to the high costs variation A Systematic Review and Meta-analysis especially for the diagnosis of EPTB patients for pathological and x-ray diagnosis which were not free of a charge unlike the rest of TB services. Secondly, it might be due to the difference in our inclusion criterion regarding included studies dates of publication, study participants, and types of TB. Our systematic review and meta-analysis enrolled both PTB and EPTB but the above studies only include smearpositive PTB patients. Since smear-negative PTB and EPTB patients might be need a number of investigations that would be requested before confirming the diagnosis which might be leads to prolong the health system delay [32,65]. The other possible explanation could be due to having poor access and low health care service coverage, a study was done in Ethiopia showed that, the overall TB health care service coverage was 23% [66].
Our findings, in the subgroup analysis suggested that, studies conducted from 1997 to 2015 had the highest pooled median health system delay of 21.63 (95% CI: 14.38-28.88) days as compared to studies conducted after 2015 with a pooled median health system delay of 9.33 (95% CI: 3.95-14.70) days. This might be due to the establishment of the End TB strategy. Since it starts after 2015, and the main targets were 90% reduction in TB deaths and minimize TB incidence by 80% in 2030. To achieve this goal WHO sets as one of the main strategy was through early diagnosis and prompt treatment [5]. Therefore, to reduce the time delay for diagnosis and treatment of tuberculosis patients, WHO recommended rapid molecular diagnostic tests such as Xpert® MTB/RIF assay. Xpert provide results within 2 hours, this leads to a significant impact on reducing health system delay [5]. Furthermore, the national TB program of Ethiopia has also re-prioritized the key strategic interventions in the five-year national TB strategic plan that paves towards achieving to reach 90% of all people with tuberculosis diagnosed and treated [1]. The program is committed to improve access and equitable TB services to vulnerable and marginalized population groups where TB burden concentrates and most delays happen. In addition, the program also considered that the strategies could only be achieved if TB diagnosis, treatment, and prevention services are provided within the context of progressing towards universal health coverage [30]. This program-related consideration might be the contributing factor for the reduction of a pooled median health system delay for studies conducted after 2015 in Ethiopia.
In this systematic review and meta-analysis, we found that rural residence had a positive association with health system delay. Being from the rural residents were nearly two and half times more likely to delay in the treatment of tuberculosis compared to those patients from urban residents with a pooled OR of 2.42 (95% CI: 1.16-5.02). The finding of this review was consistent with a systematic review and meta-analysis done in 40 countries [23]. Similarly, this finding was supported by primary studies conducted in Tajikistan [67], India [68], Indonesia [69], and China [63]. The possible plausible reason for this situation could be patients living in rural areas had low access to health care facilities, as a result of the absence of better access to anti-tuberculosis drugs and diagnostic modalities [70]. For instance, most patients in rural areas of Ethiopia had primary access to the health post (first level of the health care system) where there were no TB diagnostic services [48], and patients might walk for some hours to access hospitals or health centers [71]. The other possible explanation might be, 85% of the rural population of Ethiopia not having access to health care, and lack of physicians available in the public sectors. As well as, appropriate and affordable TB services is still problematic in some rural areas of Ethiopia [72][73][74] that could lead to prolonged health system delay. In addition to this, initially those patients were come from rural residents repeatedly visited traditional healers before diagnosis and treatment [75]. Many patients from rural areas did not arrive at the health care facilities on time unless they were seriously sick [71]. Since, the presence of traditional healers, which are prominent in rural areas across developing countries had been suggested that seeing traditional healers influence tuberculosis treatment initiation [76]. A study conducted in a rural area of China, Tanzania, and South Africa showed that visiting traditional healers were significantly associated with health system delay [59,76,77]. Therefore, this may lead to longer health system delay.
In this systematic review and meta-analysis, distance from the health facility, HIV status, sex, and type of TB were not associated with the health system delay. Even though a number of included studies were reported that distance from the health facility more likely to lead prolonged health system delay, the pooled analysis indicated that distance from the health facility was not associated with health system delay. The possible explanation might be since health system delay starts from when the patients reach at the health care provider, factors which were related to the health system delay mainly related to health care system related factors or factors which drive within the health care system like a lack of availability of trained man-power and diagnostic modality services might be leads to prolonged health system delay [33].

Limitations
Our search strategy was rigorous and multiple sources were searched, and we obtain studies from all regions of Ethiopia, this allows the study to represent the burden of health system delay at the national level. Despite these, our systematic review and meta-analysis was not free from limitations that arose from either individual studies or the review process. The review was limited to only articles published in the English language, and all included studies were cross-sectional studies. Since cross-sectional studies had a number of inherent limitations that potentially bias the results, and lack of the ability to determine causality as do other observational studies. The other limitation, since WHO as well as Ethiopia at the national level had no cut-off point to say tuberculosis patients were delayed or not. Due to this, each included study had a different cut-off point and they were not directly suited for a meta-analysis. Rather than excluding studies that failed to meet strict criteria, we tried to extract all the studies to arrive at a more national understanding of factors contributing to the health system delay.

Conclusion
In Ethiopia, this review highlights that patients were delayed more than two weeks in the treatment of tuberculosis. Being a rural resident, was the contributing factor of health system delay. This finding sound that TB patients were waited too long time to initiate anti-TB treatment and this might increase the morbidity and mortality of tuberculosis patients. Furthermore, the extended delays in these patients reflecting that, since they were visiting the health care system; health care providers could pose a greater risk of disease transmission and the community at large. Therefore, this suggests that there need to be strengthen early diagnosis, and treatment strategies on tuberculosis mainly to the rural residents through availing the scaling up of rapid molecular diagnostic tests such as Gene expert, improving access to DOTS services, and enhancing the capacity of health care providers on tuberculosis in rural areas could be vital. These could be an effective strategy to overcome the barriers of health system delay and critical to interrupt transmission of this deadly disease in the country. For successful TB control, implementing efforts like providing regular health education to the community about TB emphasizes the rural community and enhancing the quality of care in TB treatment facilities in rural areas could have important implications to reduce health system delay. The information provided by this review could assist health policy-makers in devising suitable interventions in order to minimize health system delay and reduce transmission of infection in the community.