Introduction

Mobile phones are simple but powerful devices that could enhance health services in Africa. Cell phone penetration in sub-Saharan Africa (SSA) has been steadily increasing [1]. At least 477 million people in SSA (45% of the population) subscribed to mobile phone services at the end of 2019 [2]. It is estimated that by the end of 2025, the number of smartphone connections in the region will almost double to reach 678 million as internet penetration and mobile phone ownership increase in the continent [3]. Public and private entities have used telephone hotlines to provide general health information and education and repurposed them as alert lines for outbreak response. For example, in Malawi, the Health Centre by Phone which was used to provide health information on nutrition and general health, was later repurposed for outbreak response during the COVID-19 pandemic [4,5]. There is no standard definition of a telephone hotline. However, in health care, a telephone hotline can be a toll-free call line or charged at a fixed cost to address the population’s health needs.

Despite the potential of telephone hotlines, evidence of their role and impact during outbreaks is lacking. Most reviews of m-health have in developing countries have focused on the use of mobile phones in chronic disease management and health behavior changes in non-outbreak settings [6,7]. One literature review assessed African telephone hotlines for risk communication and community engagement in 13 African countries during outbreaks [8]. However, this study was not a systematic review and did not assess study quality. There was also a narrow focus on communication and COVID-19 in a few countries.

Data is needed on the effectiveness of telephone hotline during outbreaks in Africa. This evidence is of particular importance to the Africa Centres for Disease Control (Africa CDC), which was established to safeguard the health of the continent by detecting and responding quickly and effectively to disease threats and outbreaks [9]. Before Africa CDC or any organization can support telephone hotlines across Africa, evidence of the usefulness of hotlines is needed. This systematic review and meta-analysis assessed the published evidence on the role and effectiveness of telephone hotlines for outbreaks in Africa.

Methods

Results

Search results

Fig 1 is a PRISMA flow diagram of our search. Our search yielded 1251 non-duplicate hits that were later screened using our inclusion and exclusion criteria. After screening, 41 full texts were retrieved. Finally, 21 studies were included in the narrative synthesis, while 12 were included in the meta-analysis.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. PRISMA flow diagram showing the selection of studies.

https://doi.org/10.1371/journal.pone.0292085.g001

General characteristics of included studies

Table 1 is a summary of the included studies. Of the included studies, two were from southern Africa [12,13], four were from central Africa [14–17], six were from east Africa [18–23] and nine were from west Africa [24–32]. We found no studies from northern Africa. In terms of coverage, six were local and six were national hotlines. The hotlines were set up for unusual respiratory events (one study), polio (one study), Ebola (10 studies), COVID-19 (two studies), malaria (one study), influenza-like illnesses (ILI) (one study) and rift valley fever in livestock (one study). The hotlines in Sierra Leone and South Sudan were set up for Ebola initially but extended for use in the COVID-19 response. The hotlines were mainly used as part of the event-based surveillance systems in the countries. In most studies (12 out of 19), data were collected and reported using paper-based systems. Seven studies used digital reporting. In terms of study design, eight studies were prospective cohorts, ten were cross-sectional, two were mixed method studies, and one was a before/after study. Two studies explored the use of telephone hotlines in rumour tracking, raising awareness, dispelling myths, and combatting misinformation during outbreaks [20,27].

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of included studies.

https://doi.org/10.1371/journal.pone.0292085.t001

Effectiveness of the hotlines

Table 2 shows the sensitivity of the hotlines in generating alerts. Fig 2 is a forest plot from the meta-analysis of the studies in Table 2. A total of 332,323 alerts were generated, and 67,658 met the case definition. These alerts corresponded to an overall pooled sensitivity of 38% (95%CI: 24–52%) I² = 99.99%. When stratified according to the coverage (local vs. national), the sensitivity was 41% (95% CI: 24–59%) I² = 99.94% for local hotlines and 26%(95%CI:5–47%) I² = 99.98% for national hotlines.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Forest plot showing pooled sensitivity in generating alerts by coverage of the hotline.

https://doi.org/10.1371/journal.pone.0292085.g002

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Effectiveness of hotlines: Proportion of alerts that met case definition.

https://doi.org/10.1371/journal.pone.0292085.t002

Table 3 shows the proportion of confirmed cases identified through follow-up with the hotline. Fig 3 is a forest plot from the meta-analysis of the studies in Table 3. Of the 2958 confirmed cases, 562 were identified through follow-up with the hotline. This corresponded to an overall pooled proportion of 47% (95%CI: 20–74%) I² = 99.53%. The meta-regression showed the heterogeneity was not explained by the sample size (S2 File).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Forest plot showing the pooled proportion of cases identified through follow up with the hotline.

https://doi.org/10.1371/journal.pone.0292085.g003

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Effectiveness of hotline: Proportion of confirmed cases generated using the hotline.

https://doi.org/10.1371/journal.pone.0292085.t003

Fig 4 is a funnel plot to assess for publication bias. From the figure, the points are more concentrated in larger effect sizes. The publication bias was high toward positive effects. The egger test also confirms the high publication bias (S3 File).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Funnel plot for publication bias.

https://doi.org/10.1371/journal.pone.0292085.g004

Other measures of effect reported were the rate of death reporting (from three studies) and vaccination confirmation (from one study). From the studies that used hotlines in reporting deaths and for mortality surveillance, the rate of reporting via the hotline ranged from 34% to 83%, depending on the stage of the outbreak. The only study that used a telephone hotline to follow up on vaccination coverage was conducted in Chad. This study showed that by using telephone calls to verify and confirm vaccinators’ visits, the vaccination coverage increased from 43% before intervention to 95% after the intervention. This corresponded to a significant attributable change of 0.52% (95% CI: 0.51–0.53) in the vaccination coverage for polio in children.

One study from Kenya used telephone hotlines in surveillance for rift valley fever (RVF) in livestock during El Nino rains and the threat of RVF. In this study, 69 herds met the case definition for RVF, and 24 were probable cases. There was no follow-up confirmation of cases.

In two studies, the effect reported was in terms of the number of rumors submitted. One of the studies, conducted in Ivory Coast, used the telephone hotline to submit 1,747 rumors coded, analyzed, and used to inform risk communication and community engagement activities.

Discussion

Telephone hotlines and other emerging digital technologies have been identified as an essential area of epidemic response [33]. Our study delved into the critical role played by telephone hotlines and the assessment of the effect. Our findings demonstrate that many African countries have used hotlines to enhance disease surveillance. This study extends the literature by examining hotlines in LMIC settings, using a systematic review, and meta-analysis of data from 12 studies.

Our study shows that telephone hotlines can play a crucial role in disease surveillance in the generation of alerts during outbreaks by providing a quick and easy means for individuals to report potential outbreaks and seek information and guidance. In the early stages of an outbreak when timely detection and response can limit the spread of disease, this ability of rapidly generating alerts is extremely useful. Other studies have been conducted outside Africa to examine the effectiveness of telephone hotlines in generating alerts and predicting outbreaks. In a large metropolitan area in Japan, Katayama et al showed that the hotline could be effective in predicting an outbreak of influenza [34]. They showed that there was a high level of correlation between the telephone triages and the number of influenza patients, making it possible to use the hotline for early alert generation and response. Other studies in the United Kingdom have shown that telephone hotlines as part an early warning system successfully detected influenza-like illness outbreaks and abnormal rises in vomiting and diarrhoea, prompting a timely response from the public health authorities [35]. However, the timeliness of the response depends on the availability of resources. In one study from a resource constrained country like Yemen, there was instead a delay of about 3 days in the response to cholera when the national telephone hotline system was used [36].

Even though the overall evidence suggest that telephone hotlines can be effective in generating alerts, it is worth noting that some studies have provided evidence to refute this finding. In the United Kingdom, one study found that the national health service hotline can overlook small, localized outbreaks and generate false positive alerts leading to the generation of unnecessary responses [37]. In addition, for diseases such as Cryptosporidiosis and Rotavirus, the hotline failed to detect outbreaks and resulted in bias reporting [37,38].

Overall, our study shows that while national hotlines can be effective at providing general information, local hotlines can be more effective at generating alerts. Taking into consideration specific disease outbreaks, resources available and the needs of the local community, local hotlines can be more effective at connecting individuals to local resources and facilitating communication between the public and local health departments. In Africa, at least 45 member states have national hotlines that receive calls and provide essential information to keep the public safe. In South Korea, the national hotline was used to provide reliable and accurate information about COVID-19 to the public [39]. These national hotlines easily become overwhelmed and lack adequate resources for follow up actions. The decentralization of national hotlines to local hotlines provides unique opportunities to get closer to the local communities, track cases, provide real-time alerts and easily adapted to implement effective disease control measures during the Ebola outbreaks Sierra Leone and South Sudan [18,28,31,32,39]. In China, during the COVID-19 pandemic, out of the 625 hotlines that were created across the country, 420 were rapidly adapted to meet the increasing the psychological demands [40].

Another important finding from our study is the relevance of telephone hotlines in reporting deaths. Telephone hotlines can be useful in mortality surveillance, particularly in low-resource settings where traditional surveillance systems may be inadequate. In many of these low-income countries, many deaths in remote areas can go unreported. The lack of complete and robust mortality data limits the ability of many countries in Africa to measure the impact of programmatic interventions and adequately plan to address their health needs [41]. While initiatives like the Rapid Mortality Mobile Phone Surveys(RaMMPS) consortium is developing innovative mobile phone approaches to generate national mortality estimates(including excess mortality) during outbreaks [42], we found that there is still a limited number of published literature on the role of telephone hotlines in mortality surveillance during outbreaks in Africa. All the studies included in our systematic review with data on deaths reported were from the Sierra Leone hotline. Out of Africa, researchers in Malaysia showed that it was feasible to use mobile phones to support mortality surveillance during the COVID-19 pandemic where face-to-face contacts were severely restricted [43]. Also, in non-outbreak settings, some studies have shown the usefulness of mobile telephones. One study in Papua New Guinea showed that health workers successfully used mobile phones for notification of deaths and verbal autopsy as part of their mortality surveillance programme adopted by the government [44]. In Malawi, another study found that more than half the deaths reported using mobile phones were not registered in the National Registration Bureau [45].

Despite the immense potential of mobile telephones in strengthening mortality surveillance in communities, their success will depend on a wide range of factors such as the availability of committed community health workers for routine reporting; the presence of established death management systems that involve contact with families; and the presence of reliable platforms that can support remote connectivity [46]. These factors also determine the quality and consistency of the data reported.

Regarding risk communication during pandemics, our study shows that by providing a direct line of communication between the public and trusted sources of information, telephone hotlines can be used to track rumours, dispel myths, and fight misinformation. Using telephones, individuals can report rumours that can be used to optimize effective public health messaging. Although countries like Comoros, Kenya, Madagascar, Malawi and Zambia used their telephone hotlines to track rumours during the COVID-19 pandemic [47], we found only one peer-reviewed study in Ivory Coast which systematically showed how rumours could be tracked using telephone hotlines [27]. Conventional ways of tracking rumours have been through social media applications like twitter, Facebook, and Instagram. The World Health Organization (WHO) has established a rumour tracking programme and a community engagement hub. These WHO hubs use the leading social media outlets in rumour tracking. However, not all mobile phone users in developing countries possess telephones that can accommodate social media apps, and many older phone users may not be actively engaged on social media [48]. Consequently, telephone hotlines provide a simple means to extend the reach for rumour tracking to involve wider demography of users and those who may not have smartphones.

In all the studies that were included in this systematic review, the objective and rigorous measurement of the effectiveness of the hotlines posed a significant challenge. The need for appropriate tools and the capacity of the hotline operators to conduct basic research that assesses the effect of hotlines were essential barriers. No tool was available for the objective assessment of the effect of the hotlines in rumour tracking, fighting misinformation, and promoting community engagement. For studies that measured the effect of the hotlines in terms of the alerts generated and alerts that were subsequently followed up and tested, significant methodological flaws were identified. The cross-sectional studies used secondary data to assess the effect and did not adjust for confounding factors such as the age distribution, location of the callers, differences in socioeconomic status and health systems. In addition, prospective cohort studies had no initial considerations to ensure that the design and data collection methods were robust, and the comparator groups were unclear in most studies. In these cohort studies, there was no consideration of the effect of low response rates which resulted from callers not calling back or the hotline operators not doing appropriate follow-up. Most studies needed to outline the selection of sites and sampling frame clearly. These limitations may arise because there is usually more emphasis on implementing the intervention than basic research to generate evidence that may optimize the intervention in the setting of outbreaks [49]. It is, therefore, necessary to develop simple research tools and frameworks that can be used by hotline operators who are not usually researchers, to help them design, collect, analyze, and use the data collected while the outbreak is ongoing. These data will help them understand how well they are performing and help funders adequately direct funding to appropriate components of the hotline.

Our study shows that many call centres still depend on paper-based tools to collect, analyze, and report data. However, some call centres have used electronic data capture and software to ensure they handle data more efficiently. Paper-based tools may lead to delays in collection, analysis, and interpretation. In one study from Ivory Coast, the authors demonstrated that software such as the District Health Information System 2 (DHIS2) could collect and analyze rumours. The DHIS2 is widely available in many health districts across Africa [50] and could be considered a potential open-access software that can be used in a network of African hotlines by the Africa CDC to standardize the collection, analysis, and visualization of data for a more coordinated outbreak response. Furthermore, when using telephone hotlines, a considerable amount of data can be generated that may require more innovative and more efficient ways of analysis and interpretation. Machine learning and artificial intelligence could increase efficiency by making sense of the data collected by the hotlines. None of the studies mentioned the use of artificial intelligence, but in countries such as Togo, machine learning and phone data have been used to improve the targeting of humanitarian aid [51].

In general, randomized controlled trials (RCTs) will not be a suitable study design to show the effectiveness of telephone hotlines because they are not practical for such population-wide interventions and not relevant for outbreaks since decisions need to be made quickly [52]. In peace times, though expensive, a pragmatic stepped-wedged RCT can be used with a new network expansion city by city. A before-and-after study design can be used to assess the effectiveness and impact of the hotlines. In Chad, Aluma et al. used a before-and-after study design to show that a polio hotline can increase polio vaccination uptake [15]. Even though there was no adjustment for confounders, the study shows that with training and attention to basic research, it is possible for research to generate quality evidence in call centres.

Conclusion

Our study suggests that telephone hotlines can be effective in outbreak disease surveillance in Africa. Further implementation research with robust quantitative and qualitative methods is needed to scale up telephone hotlines in rural areas.

Supporting information

Acknowledgments

The authors thank the London School of Hygiene and Tropical Medicine library for access the OVID search databases. The authors also thank the Dr Nafiisah Chotun and Dr Mohammed Abdullaziz of the division of disease control of Africa CDC, for their support and guidance during this project. Special thanks to Dr Hana Rohan for her guidance during this project.