Evaluation of contact tracing performance during an Ebola virus disease outbreak in a complex security environment: the case of North Kivu province, Democratic Republic of the Congo, 2018–2020

Abstract

Background

Contact tracing remains a pillar public health strategy for containing Ebola virus disease (EVD). During the 2018–2020 EVD outbreak in the Democratic Republic of the Congo (DRC), contact tracing was implemented on an unprecedented scale. Here, we assessed performance of contact tracing implemented in affected health zones, and identified risk factors associated with incomplete follow-up.

Methods

We performed a retrospective descriptive data analysis of 129,749 contacts in the contact line lists of North Kivu province between August 1, 2018 to June 25, 2020. Coverage, completeness, timeliness, and duration of contact tracing were determined to assess the performance of contact tracing implemented by field actors. Bivariate and multivariate logistic regression models were used to identify factors associated with incomplete contact tracing.

Results

Overall, more than 90% of all contacts initially identified and listed were monitored. However, 9.1% of contacts who had monitoring initiated had completed the 21 days follow-up. The median days between identification and the start of follow-up and duration contact follow-up were 3 (1–6) and 17 (12–19), respectively. The risk of incomplete follow-up was higher among contacts from urban and conflict-affected health zones.

Conclusion

Our findings indicate the necessity of prioritizing contact tracing in urban areas. This can be achieved by engaging locally trusted stakeholders to build community confidence. Furthermore, integrating digital contact tracing solutions may enhance the efficacy of traditional manual contact tracing.

Background

Ebola virus disease (EVD) is a serious zoonosis caused by viruses of the genus Ebolavirus, family Filoviridae. The Democratic Republic of the Congo (DRC) is the country most affected by EVD, with a total of 15 outbreaks [1]. During the 2018–2020 period, the DRC experienced its tenth EVD outbreak, considered the second largest worldwide, resulting in 3,470 cases and 2,287 deaths (case fatality rate [CFR]: 66%) [2]. Of the three northeastern DRC provinces affected, North Kivu accounted for 87% and 96%, respectively.

Control strategies and measures implemented during the 2018–2020 EVD outbreak included surveillance and reporting of cases, expanding of laboratory capacities, isolation and case management, infection prevention and control in treatment centers and the community, setting up of screening points of mobile populations at numerous border crossings, communication and social mobilization, vaccination, and contact tracing [3, 4]. The latter is recognized to play a crucial role in detecting cases to monitor and breaking the chain of transmission [5]. In the context of EVD, contact tracing involves tracking down anyone who had direct or indirect exposure to any confirmed, probable, or suspected EVD case, or bodily fluids of a case, within the past 21 days [6]. The process consists of three fundamental steps: (i) Contact identification: all potential contacts are identified through interview with the EVD case; (ii) Contact listing: identified contacts are listed and interviewed; (iii) Contact follow-up: listed contacts are monitored for presence of infection or onset of disease through daily visits over a 21-day period from the date of the most recent exposure [6].

While enormous contact tracing efforts were made to identify and list contacts during the 10th EVD outbreak in eastern DRC, with more than 250,000 contacts recorded [7], counter-performance was observed during the contact follow-up phase, particularly in Beni health zone [8]. It should be noted, however, that the evaluation of control efforts relating to contact tracing has not been carried out for all affected health zones. The lack of information on the scope and characteristics of contact tracing implemented in all the health zones affected by the 2018–2020 EVD epidemic justified the need to explore the performance of this decisive public health strategy.

Thus, we determined the level of performance of contact tracing implemented by field actors at global and health zone scale, and identified the risk factors associated with incomplete contact tracing.

Methods

Study design and setting

We performed a retrospective descriptive data analysis of electronic line lists from collection forms for contact tracing that was conducted for the EVD outbreak in the affected health zones of North Kivu province from August 1, 2018 to June 25, 2020. This outbreak emerged in a complex and violent socio-political and security environment [9]. For almost three decades, the North Kivu province had been facing recurrent violence linked to the activities of foreign and local armed groups [10]. Over 120 armed groups are active in the eastern provinces of the DRC [11], prompting the political authorities to declare a state of siege in 2021 to combat armed groups and protect the civilian population [12]. Furthermore, the North Kivu province covers an area of 59,483 km2, and its population is estimated at 6.6 million. More than half the population lives in urban areas, and over 68% of the population is under 25 [13] (See Fig. 1).

Fig. 1

Study site

Data collection and sources

For this study, we considered basic information on sociodemographic characteristics and information on the status and the follow-up of all confirmed, probable, and suspected EVD case-patient’s contacts, obtained from the electronic contact line lists of the Ministry of Health, compiled from paper collection forms used for contact tracing. This included the following: ID of contact, age, sex, health zone of residence, type of contact, relation with the index case, and daily follow-up. As recommended, contact tracers should visit contacts twice daily (morning and afternoon) for 21 days post-exposure to evaluate the health status of the contact through self-report and physical observation [6].

We also considered contextual data associated with violence and demographic characteristics. Information on active conflicts and attacks on health care was extracted from two georeferenced and disaggregated event databases, the Armed Conflict Location and Event Data (ACLED) project [14] and Insecurity Insight [15]. ACLED captures conflict armed, organized political violence and demonstrations events using cross-checking of multiple information sources: media reporting, reports from non-governmental or international organizations, selected social media accounts (Twitter and Telegram), and partnerships with local conflict observatories in hard-to-access cases [16]. Insecurity Insight uses the ”Taback-Coupland model of armed violence” to generate data on the impact of armed violence and insecurity on aid workers, aid agencies, health workers, educators, internally displaced persons and refugees [17]. After acquisition, all these contextual data were aggregated at the health zone level. In addition, population density data derived from the quotient between DRC population statistics by health zone and the surface area of health zones extracted from shapefiles, all obtained from public data available on the open-access data platform “The Humanitarian Data Exchange” [18, 19]. Health zones were then classified according to population density classes [20]: < 250 people per km2 for rural areas; 250–800 people per km2 for suburban areas; >800 people per km2 for urban areas.

Data analysis

Before proceeding with the analysis, contacts with duplicate information were excluded (N = 735). Descriptive statistics were then achieved for baseline characteristics of study participants, as well as the proportion of traced contacts who were potential EVD cases and new confirmed EVD cases. All variables are presented as frequencies and percentages. It should be noted that less than 10% of each main variable was represented by missing values. In accordance with the most commonly employed approach, namely “complete records analysis” [21], these were excluded during the analysis performance.

To assess the performance of the contact tracing implemented by field actors, we first excluded contacts who were confirmed EVD cases. We then determined the following indicators:

• Coverage of contact tracing: number of listed contacts that have been monitored at least once / total number of listed contacts overall;

• Completeness of contact tracing: number of contacts who completed the 21 days follow-up period following the last exposure / total number of listed contacts overall;

• Timeliness of contact tracing: median number of days (interquartile ranges [IQR]) corresponding to the difference between the last day of exposure and first day of contact follow-up;

• Duration of contact tracing: median number of days (IQR) of contact follow-up.

Accordingly, differences between health zones based on these indicators were tested by the non-parametric Kruskal-Wallis test or the X² test, as appropriate.

Finally, we performed a stepwise multivariable logistic regression model, including explanatory variables that had a significant relationship with the outcome variable in bivariate analysis, to explore risk factors associated with incomplete follow-up of traced contacts (i.e., contacts who have been identified, listed and whose follow-up has been delayed or initiated, and who have not been found at any time during the 21-day follow-up period). Crude and adjusted odds ratios (ORs) with their 95% confidence interval (95% CI) were determined and a statistically significant association was found based on P-value < 0.05. In addition, multi-collinearity assessed between the independent variables using the variance inflation factor (VIF) was less than 5. The goodness-of-fit of the final logistic regression model also evaluated using the Hosmer–Lemeshow statistic was greater than 0.05. Different analyses were performed using R version.4.2.0.

Results

Overall, 129,749 records for EVD case-patient’s contacts were analyzed in the present study. Of these, their median age was 23 years (13–36), and over a third were aged between 15 and 29 (35.6%), followed by those under 15 (28.0%). Female contacts were slightly more numerous than male contacts (51.2% vs. 48.8%). The risk of exposure, determined by the nature of the relationship with the index case and the type of contact, was represented in majority by community (49%) and touching or cleaning the linens, clothes, or dishes of the patient (37.9%), respectively (Table 1).

Table 1 Baseline characteristics of study participants in the health zones affected by EVD outbreak, 2018–2020, North Kivu, DRC (N = 129,749)

Of 129,749 total contacts, 2,900 (2.2%) were potential EVD cases. Also, 648 (0.5%) confirmed EVD cases were identified (Fig. 2).

Fig. 2

Proportion of potential EVD cases and new confirmed EVD cases, 2018–2020, North Kivu, DRC (N = 129,749). (A) Potential EVD cases; (B) Confirmed EVD cases

Overall, the monitoring was initiated for 124,602 (96%) contacts previously identified and listed. At health zone level, over 90% of all contacts in each health zone had monitoring initiated. However, 9.1% of contacts who had monitoring initiated had completed the 21 days follow-up. Beyond the significant differences observed (P-value < 0.001 by X² test), the Alimbongo, Kyondo, Manguredjipa, and Mabalako health zones were the only ones to record more than 25% of contacts completing follow-up (Fig. 3).

Fig. 3

Comparison of health zones affected by EVD outbreak, 2018–2020, North Kivu, DRC (N = 129,101). (A) Coverage of contact tracing; (B) Completeness of contact tracing

The median days between identification and the start of follow-up and duration contact follow-up were 3 (1–6) and 17 (12–19) respectively. Significant differences were observed between health zones in terms of the timeliness and duration of contact tracing (Fig. 4). See additional file 1 for detailed results of timeliness and duration of contact tracing in health zones affected by 2018–2020 the EVD outbreak in North Kivu.

Fig. 4

Comparison of health zones affected by EVD outbreak, 2018–2020, North Kivu, DRC (N = 129,101). (A) Timeliness of contact tracing; (B) Duration of contact tracing

Bivariate logistic regression analysis revealed that risk factor variables such as the family relation with the index case, age groups under 15 and 15–29 years, active conflict, attacks on health care, and living in urban health zones, showed a statistically significant association with incomplete follow-up among traced contacts, as evidenced in the results from the COR with 95% CI with P-values of less than 0.05 (COR = 1.11, 95% CI: 1.04–1.18; COR = 1.14, 95% CI: 1.02–1.26; COR = 1.27, 95% CI: 1.14–1.41; COR = 1.74, 95% CI: 1.64–1.84; COR = 1.24, 95% CI: 1.15–1.33; COR = 2.81, 95% CI: 2.68–2.94), respectively (Table 2).

Table 2 Bivariate logistic regression analysis of associated factors for incomplete follow-up among traced contacts in the health zones affected by EVD outbreak, 2018–2020, North Kivu, DRC (N = 129,101)

The final multivariable regression analysis showed that contacts from active conflict health zones had significantly greater odds of incomplete follow-up (AOR = 1.28; 95% CI: 1.19–1.38; P-value < 0.001) than those from non-conflict health zones. Also, the risk of incomplete follow-up was more than two times greater among contacts from urban health zones (AOR = 2.56; 95% CI: 2.43–2.69; P-value < 0.001) than those from suburban and rural health zones (Table 3).

Table 3 Results of final multivariable analysis of predictors of incomplete follow-up among traced contacts in the health zones affected by EVD outbreak, 2018–2020, North Kivu, DRC (N = 129,101)

Discussion

Our findings show that at least 9 out of 10 EVD case-patient’s contacts have not only been identified and listed, but also monitored. This indicates a high level of effort in the transition from the contact identification phase to the implementation of contact follow-up. However, in our context, only 9% of identified contacts had completed the 21 days follow-up, as recommended [6]. Although some studies have reported completion rates of the 21-day monitoring period around 90% [23], this highlights the poor performance of contact tracing, which can result from significant gaps between contact identification and completion of daily contact follow-up [8, 24].

The low level of daily contact tracing is also reflected in the differences observed for the median duration between identification and the start of follow-up (3 days [1–6]) and duration contact follow-up [17 days (12–19)]. The consequences are significant, as infected contacts contribute to chains of transmission, which can lead to large-scale expansion of EVD outbreaks [25].

The results on contact tracing underperformance observed in this study are consistent with the modelling framework demonstrating that incomplete follow-up was likely higher among contacts from health zones with a deleterious security context. The tenth EVD outbreak occurred in a region characterized by recurrent and violent socio-political crises that have lasted for more than two decades, perpetrated by more than 120 armed groups [9, 11]. Mistrust and lack of community involvement, caused by misinformation and misperceptions about the potential financial return of massive efforts to support EVD control activities, were at the root of targeted attacks on Ebola treatment and isolation centers, supply lines and field investigation teams [10, 26, 27]. The emergence of this outbreak has not only led to a significant increase in the number of conflicts reported in health zones affected by the disease [28], but also in some areas that were conflict-free prior to the epidemic [27].

Targeted attacks on health care would have significantly hampered the implementation of epidemic response activities, including contact tracing, which would also have indirectly influenced the spread of EVD. There is substantial evidence of a plausible causal link between the occurrence of these targeted attacks, the disruption of epidemic response activities and the subsequent increase in EVD incidence during this tenth outbreak. Mueller and Rebmann found a significant increase in the number of EVD cases in the two to four weeks following violent events targeting humanitarian and medical aid operations at provincial level [29]. Similarly, an average increase of 370 additional cases, corresponding to 250 expected deaths, was recorded at health zone level [30].

The present study shows that incomplete contact tracing for EVD was significantly observed in urban health zones. Limitations in contact tracing performance in urban areas were also demonstrated during the large EVD epidemic in West Africa [31]. This suggests that contact tracing is less effective in urban areas, for a number of reasons. Urban areas are generally characterized by a combination of high population density and complex social networks, making it more difficult to identify all contacts. There is also a combination of low cooperation from the community and a much heavier burden and sometimes limited resources for those involved in the response to the epidemic [31]. This emphasizes the necessity to fully engage local stakeholders [9, 26], such as community-trusted leaders and service providers, in conjunction with responders who are not from the communities affected by the EVD outbreak, in order to foster community confidence and engagement in the success of the contact tracing process.

This study has a number of limitations. The analysis of contact tracing performance for EVD did not consider the different phases of the epidemic, which made it impossible to identify different performance profiles. Furthermore, given the determination of field actors’ performance carried out in this study, the inclusion of information on contact tracers in the analysis of contact tracing completeness and timeliness would have enabled us to further explore associated risk factors. Finally, our study design suggested associations between predictors and incomplete contact tracing for EVD, but unfortunately did not reveal causal relationships.

Conclusion

Our findings indicate that despite the implementation of contact tracing on an unprecedented scale, significant discrepancies are observed between identification phase and completion of daily follow-up. There were limitations to performance of contact tracing in urban and conflict-affected health zones. The results of this study suggest the necessity of prioritizing contact tracing in urban settings, with a specific focus on the active involvement of local trusted leaders and service providers to guarantee genuine community engagement and achieve optimal contact tracing outcomes. In addition, they suggest the need to consider the large-scale deployment of digital contact tracing solutions to improve the effectiveness of traditional manual contact tracing [32], after careful risk assessment including monitoring of conflict dynamics as a part of the integrated disease surveillance and response system [33].