Impact of the Ebola virus disease outbreak (2014-2016) on Tuberculosis Surveillance Activities in Guinea's National Tuberculosis Program: A Time Series Analysis

Tuberculosis (TB) is a major cause of disease and death worldwide. According to World Health Organization (WHO) estimates, Guinea is a country with a high incidence of tuberculosis (TB) and TB / HIV co-infection. In March 2014, the WHO announced the Ebola virus disease (EVD) outbreak in Guinea that caused a health system disruption. Our study aimed to assess the impact of the Ebola virus disease outbreak on the TB surveillance system through the main indicators of TB-related morbidity and mortality. This is a retrospective cohort study comparing TB trends using TB surveillance data from periods before (2011-2013), during (2014-2016) and after (2017-2018) the Ebola virus disease outbreak. A time-series analysis was conducted to investigate the link between the decrease in TB incidence and the Ebola virus disease through cross-correlation. We evaluated the lag observed in the cross-correlation test using the ANCOVA type II delayed variable dependent model. The current status of the surveillance system was compared to that of 2015 during the EVD through the Standards and benchmarks for TB surveillance and vital registration systems established by WHO. The EVD outbreak had a severe impact on TB surveillance in Guinea, with a significant decline in TB reporting rates during the epidemic period. The introduction of an early warning system would preserve the TB surveillance system ; which could encourage the implementation of interventions to ensure access to diagnosis, and enhanced surveillance for the treatment of TB.


Introduction
Although it can be cured in the majority of cases, tuberculosis (TB) is the leading cause of death from infectious diseases worldwide (1). This disease is a priority of the Ministry of  (5). Guinea reported a total of 3,811 cases of Ebola and 2,543 deaths nationwide during the outbreak. In addition to the devastating health effects, the EVD outbreak has also had a significant socio-economic impact in Guinea, Liberia, and Sierra Leone (6). According to World Bank forecasts for 2014(7), it is estimated that $2.2 billion was lost in 2015 in the gross domestic product (GDP) of the three countries. The Guinean health system was severely shaken during the EVD outbreak, primarily due to the lack of infrastructure and skilled workers (8). With a population of more than 11 million, the country had only one doctor and one nurse per 10,000 population, 25 times less than in the United States (9). This low ratio of health care providers to the population deteriorated further due to the high rate of Ebola infections and deaths among health care workers (192 Ebola infections, including 86 deaths).
Based on these findings, we hypothesized that the EVD outbreak impacted the TB surveillance system in Guinea and had led to an increase in TB-related morbidity and mortality. Our study therefore aimed to assess the impact of the EVD on TB surveillance activities conducted by Guinea PNT by analyzing trends in selected indicators before, during and after the EVD outbreak.

Location and period of study
The study was carried out at the NTP of Guinea from February 2019 to June 2019. This program is under the responsibility of the National Directorate of Great Endemics, which belongs to the Ministry of Health. This direction is the main body responsible for epidemiological surveillance of TB, notification, and treatment of TB cases.

Study design and population
We conducted a retrospective cohort study analyzing temporal trends in the incidence of TB and treatment outcomes from periods before (2011-2013), during (2014-2016) and after (2017-2018) the EVD outbreak. We also assessed the current status of the surveillance system against the 2015 assessment during the EVD outbreak.

Data used
The TB surveillance data used in this study were extracted from the National Health Information System (DHIS2), which contains a module for the collection and analysis of TB data. TB cases are collected and reported from the diagnostic and treatment centers available in all health districts. TB surveillance reports were collected quarterly and captured in the DHIS2 set up in 2016. Historical aggregate surveillance data (2009)(2010)(2011)(2012)(2013)(2014)(2015) from the NTP have been imported into this system, and data entry continues at the national level on a quarterly basis.
We have included all epidemiological surveillance data recorded in the National Surveillance System (DHIS2) from 2011 to 2018 that were validated and published by the NTP.
Population estimates were obtained from the National Statistics Institute (NSI), which conducted a general population census in 2014 and projected the population for subsequent years to 2020. The EVD outbreak surveillance data were obtained from the National Health Security Agency (NHSA), which monitors epidemics in the country. All surveillance data for TB and EVD in Guinea are fully anonymized and available for free access on the WHO website.

Operational definition of variables
The indicators analyzed in this study are calculated from the data elements collected from the quarterly report forms filled in by the agents of the diagnostic and treatment centers.
These indicators are consistent with the WHO revised TB reporting framework published in 2013 (10). We targeted two indicators, the TB case reporting indicator and the treatment outcome indicator as described below: Tuberculosis notification rate: number of reported TB cases per 100,000 population. This is based on reported TB cases that are either bacteriologically confirmed or clinically diagnosed. Therapeutic success rate: Percentage of reported TB patients who have been successfully treated.
Ebola notification rate: number of reported Ebola cases per 10,000 population.

Statistical analysis plan
We conducted descriptive analysis over each of the three periods evaluating the TB notification rates per 100,000 inhabitants. We used the WHO Standards and Benchmarks for TB surveillance and vital registration systems (11) to assess the status of the current surveillance system compared to the 2015 assessment during the EVD outbreak.
A time-series analysis was conducted to assess the effect of the EVD outbreak on the notification and treatment of TB cases by quarter. To do this, we used the autocorrelation test to examine the significance of the shifts observed in each time series separately and the cross-correlation coefficient explore the relationship between the time series of Ebola and those of TB.
Stationarity is necessary for the research of the cross-correlation between two timeseries, it is defined by a constant average and equal variance at any time and can be obtained by diversion or differentiation. We used the Dickey-Fuller test to check the stationarity of our time series, then we transformed the seasonal series into stationary series by differentiation (13). Differentiation is the sequential subtraction of the xt value of xt +1 from a time series to get subsequent changes over time (14). This technique helps to remove spurious correlations based on time dependencies between adjacent values in the input time series and removes these influences from the output time series (15). To confirm and elucidate the correlations observed between times series in the crosscorrelation test, we performed an analysis of interrupted time series (ITS) using the type II Sum Squares ANCOVA lagged dependent variable model (16). We included a default boot template, which executes 1,000 replications of the primary model with randomly drawn samples to drive the 95% CI bootstrap. An adjusted F-value (10% suppression) is reported, and a p-value initiated is derived from it. This model has been adjusted while estimating the mean difference of dependent variables ( TB cases notified) between interrupted periods (EVD, 2014 to 2016) and uninterrupted periods (2011 to 2013 and 2017 to 2018), taking into account the lag of the dependent variable and any other covariate specified.
The significance was defined as a value of p less than 0.05. The DHIS2, Excel, and R 3.5.1 software was used for data analysis.

Results
On average, 3027 ± 71 cases of TB are reported each quarter in Guinea, with a success rate of 83% ± 0.7 cases, for all periods of our series. Reported TB cases varied considerably during the period of EVD for all forms of TB including those clinically diagnosed (441 ± 139 cases) and confirmed bacteriologically (1849 ± 209 cases) on average (table1). The NTP TB case notification rate decreased from 120 cases per 100,000  Table   3).
The incidence of EVD rapidly changed-increasing and then decreasing, with the most significant proportion occurring before 2014 (more than 500 cases). The incidence of TB This increase is quite remarkable considering the therapeutic success rate, which averaged 82% before the EVD outbreak and 89% the EVD outbreak (F-value = 21.9 95% CI [8.9-47.5]) and p-value <0.001).
Regarding the TB surveillance system, of the 13 standards and criteria developed by WHO, five were achieved by the NTP in 2019, compared to only three in 2015 (Table 5). This means that the surveillance system deserves targeted, long-term action to meet the challenge of screening and monitoring patients on treatment.

Discussion
The WHO estimates the number of cases each year, however the number of TB cases actually reported by the NTP remains low compared to these estimates. NTP notifications have declined considerably during the Ebola epidemic experienced by Guinea. Some TB treatment centers have been transformed into a health center for Ebola patients, which has resulted in a weakening of TB service provision in some places. A recent systematic review of the link between the Ebola epidemic in West Africa and the health systems in Guinea, Liberia, and Sierra Leone (17) revealed the poor performance of health facilities, in part because of the lack of staff in these health facilities during the epidemic, inadequate funding for health, lack of monitoring and communication. A study in Sierra Leone (18) also reported a break in the relationship between the health system and communities during the EVD outbreak, resulting in a significant reduction in the use of health facilities.
According to our study, the reporting rate of TB cases in the NTP decreased from 120 cases per 100,000 population in 2011 to 100 per 100,000 population in 2014, when the cases of Ebola were highest. Similar results were revealed by the study on the impact of Ebola on the results of TB screening and treatment in Liberia. This study showed that for all forms of TB stratified by category and by age group, more substantial decreases were observed in the last two quarters of 2014 (19). Despite the blow to TB reporting by the EVD outbreak, the therapeutic success rate has remained stable with little upward variation over 80%. This confirms that TB cases that had been diagnosed had been followed closely during the EVD outbreak in Guinea, consistent with several other studies (17,20,22), including one in Guinea (19), which had a higher success rate during the EVD outbreak.
Our data showed that reporting rates for new cases and relapses for all forms of TB began to rise immediately after the 2014 decline. This post-Ebola performance could be due to the positive post-epidemic effects, such as improved diagnostic capacity by the new GeneXpert devices converted for TB screening and the opening of new treatment sites and the staff training. Consistent with these improvements, the TB system score in Guinea improved from 2015 to 2019. The same return of services after the outbreak was also noted by a study on the public health impact of the 2014-2015 EVD outbreak in West Africa (8). It reports that despite its adverse effects on public health and beyond, the EVD outbreak has provided West African countries with many opportunities. These have enabled Guinea to increase health expenditure, recruit an additional 2,950 health workers and begin to prioritize community participation in addressing public health threats (23).
Decreases in reported TB cases may simply be due to randomness if statistical tests are not available. Cross-correlation tests between the Ebola virus disease outbreak and TB time series confirmed that the observed decline was statistically significant with offsets beyond the confidence intervals of the cross-correlation curve. The incidence of TB dropped approximately 1500 cases in 2015 before increasing the following year, and continuing until the end of the Ebola virus disease outbreak in 2016.

Conclusion
Our study shows a near-universal significant decline in TB notification rates for all forms between 2014 and 2016 during the EVD outbreak. This trend was similar regardless of TB diagnosis method or TB patient category. As evidenced by the results of cross-correlation analysis and ANCOVA model, this can be attributed to the VME epidemic that has disrupted the entire health system. The EVD outbreak did not, however, affect the outcome of treatment for patients followed during the same period.

Ethics approval and consent to participate
This study used aggregated surveillance data for tuberculosis and Ebola. The authorization of the Tuberculosis Control Program in Guinea was obtained for the analysis of the data.

Consent for publication
Not applicable

Availability of data and materials
The data is available upon authors request.  Cross-correlation test of TB and Ebola time series