Outcomes of Digital Training for Community Health Workers in Low- and Middle-Income Countries: Scoping Review

Introduction

Low- and middle-income countries (LMICs) face a shortage of professional health care workers (HCWs) [1]. To overcome this gap, lay HCWs or community health workers (CHWs) have been recruited and trained to carry out an increasing array of tasks [2]. CHWs are laypeople working within their own community, performing functions related to health care delivery and health promotion, but have not received formal professional or paraprofessional certificates or degrees [3]. CHWs are often trained for specific tasks such as HIV testing, disease screening, or provision of immunization and have been recognized as critical role-players within the primary health care setting [4], where task shifting to CHWs has been shown to be a cost-effective method of service delivery scale-up [5,6]. When provided with the correct resources, training, and support, CHWs have improved accessibility to basic health services, resulting in better health outcomes [7,8].

The World Health Organization (WHO) recommends that CHWs receive regular training and supervision to fulfill their roles successfully [9]. To increase the competency of CHWs in health care provision, there is a need for more effective, higher-quality, and easily accessible health training [10]. The design of these training approaches would ideally also minimize the burden on an already strained health care system by limiting the time CHWs spend away from workstations while attending training required to improve service delivery.

The use of digital technology in CHW education may help overcome these challenges. In line with this, the increasing availability of technology could address the shortcomings of in-service training provision to CHWs. Digital training may provide more accessible, standardized, relevant, timely, and affordable solutions [11,12]. Moreover, digital training may provide flexibility that would allow participants to balance their tasks and learning endeavors effectively [13,14]. Given the widespread availability of technology devices, digital training by itself or in combination with face-to-face classroom training (ie, blended learning) has been widely used for medical education in a variety of settings [5,9,15-17]. These training approaches have facilitated improvements in knowledge, skills, overall competence, and clinical performance of health professionals across various health care settings, such as clinicians, nurses, and public health practitioners [18,19]. In addition, the COVID-19 pandemic necessitated and reinforced the practice of digital training for health workers, including lay cadres [14,20]. However, regardless of the benefits associated with digital training, the evidence surrounding these trainings for CHWs has not yet been synthesized.

Given that there are over 8 million CHWs in LMICs [21,22] globally, ensuring that they all receive optimal and appropriate training efficiently and cost-effectively is a challenge but also a critical priority. Digital solutions may present a viable option given that digital training can be effective in high-income countries (HICs) with professional cadres. However, evidence on the impact of digital training on capacity-building for CHWs on factors including knowledge, skills enhancement, and improved health care needs to be explored and understood. It is important to know whether these solutions can be applied to nonprofessional cadres and LMIC settings where there are limited digital skills and infrastructures [23,24], and whether they will be found to be acceptable, feasible, and effective. To develop and implement successful digital training programs for CHWs in resource-constrained settings, a greater understanding is needed of the training attributes and the extent to which these training modalities enhance CHW capacity. The Kirkpatrick evaluation framework articulates four levels of training outcomes: (1) participant reaction to the training program experience; (2) learning outcomes; (3) participant’s change in behavior; and (4) impact on the clinical setting [25]. We undertook this scoping review with this framework in mind. The overall purpose of this scoping review was to synthesize the evidence from the literature surrounding digital and blended learning among lay HCWs in LMICs to highlight areas for future research and implementation. Specifically, we sought to (1) characterize populations using these digital tools by geography and cadres, (2) describe the nature and attributes of blended and digital learning tools, (3) organize types of training outcomes examined by the levels defined in the Kirkpatrick framework, and (4) examine preliminary learning, implementation, and clinical outcomes.

Methods

Overview

We conducted a scoping review in accordance with the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines [26] on the provision of digital and blended learning training for CHWs in LMICs. Our scoping review followed systematic and transparent research steps guided by the framework described by Arksey and O’Malley [27] and updated by Levac et al [28], to characterize the attributes and nature of digital and blended learning training for CHWs in LMICs, as well as the outcomes considered. An internal protocol was developed for the review to define the inclusion and exclusion criteria and the review methods before data extraction.

Search Strategy

The Cochrane Library and PROSPERO (International Prospective Register of Systematic Reviews) were searched to identify available or ongoing scoping and systematic reviews pertaining to the provision of digital or blended training for CHWs in LMICs. No previous or ongoing relevant reviews were identified.

We then designed a comprehensive search strategy to identify all relevant studies according to the PRISMA-S (PRISMA-Search Extension) guideline [29]. We did not use published search filters. There were no search strategies that were adapted from other studies. A publication date limit was not set. The initial search was conducted on November 26, 2024. Twenty relevant search terms for “Community Health Workers” and 21 search terms for “digital learning” and “blended learning” were developed. These were combined with the World Bank Group 2022 list of LMICs using Boolean operators to develop a master search query. Where appropriate, each index-linked Medical Subject Headings term was expanded to contain all relevant subheadings. In addition, synonyms were searched for each key term, along with wildcards and truncation for free-text words. The following databases were searched to identify primary, peer-reviewed studies published up to and including January 12, 2026, including Cumulated Index in Nursing and Allied Health Literature (CINAHL), Cochrane, Embase, Education Resources Information Center (ERIC), Global Health, Google Scholar, PsycINFO, PubMed, Scopus, and Social Science Research Network (SSRN). The search terms were adjusted accordingly for each database. An updated search of studies, rerunning the searches, was conducted on January 12, 2026, by the research team. Cited references were examined for any relevant studies by browsing reference lists. Study registries were not searched. A full record of the conducted search for each database is provided in the online supplementary material (Table S1 in Multimedia Appendix 1).

Eligibility Criteria

Studies were included in the review if the population comprised primary participants who were CHWs [3], the concept involved CHWs being trained using digital or blended learning modalities, and the context included the following: (1) the training occurred in a country defined as an LMIC according to the World Bank Group 2022 classification of economies [30]; (2) the primary data were collected; (3) primary focus of the training addressed a communicable or noncommunicable disease domain; and (4) the article reported a learning, implementation, or clinical outcome. Digital learning was defined as a practice of learning using technologies to deliver educational content and training programs [31,32]. Blended learning was an education approach that combines face-to-face and digital learning approaches [33]. The training outcomes needed either to have been evaluated within the same group through pre-post study design or with a control group through randomized or nonrandomized controlled trials.

The scoping review did not restrict based on study design and included both qualitative and quantitative outcomes. Studies had to qualify as an original, full-text research study to be considered for inclusion in the review. Review articles, commentaries, letters, policy briefs, protocols, needs assessments, and conference abstracts were excluded.

Outcomes

We assessed learning, implementation, or clinical outcomes and noted the studies that reported each outcome. The learning outcomes included knowledge, attitudes, and behaviors. “Knowledge” referred to information acquired through the training. Attitudes focused on confidence to perform tasks. Behaviors focused on whether trainees used what they learned. The implementation outcomes included acceptability, appropriateness, feasibility, and fidelity. Acceptability was defined as the CHW perception that the training approach was agreeable or satisfactory. Appropriateness was the CHW’s perceived fit of the training. Feasibility was defined as the extent to which the training was successfully carried out in the setting. Fidelity was defined as the degree to which the training was delivered as intended. Clinical outcomes included how well the training improved patients’ uptake in clinical practice [34].

Selection of Sources of Evidence

All articles identified via database searching and other methods, including citation searching were exported into EndNote and imported into Covidence review software, and duplicate references were removed. The screening process was performed according to PRISMA-ScR. Titles and abstracts of all the articles identified in the search were screened by 2 authors (TAT and FA) to determine whether they would be considered relevant for a full-text review. For title and abstract screening, interreviewer reliability with Cohen κ was 0.84. Based on Landis and Koch’s [35] threshold values reference, a κ of ≥0.81 is considered “almost perfect agreement.” The full text of all the articles identified as potentially relevant was then retrieved and reviewed in full against the inclusion and exclusion criteria by both reviewers. All articles that did not meet the eligibility criteria were excluded from the review database, and reasons for exclusion were recorded. At all stages of the review, discrepancies between the authors were resolved via discussion. Where appropriate, the authors of individual papers were contacted for further information. References and other sources were reviewed to identify any other articles for full-text review. A critical appraisal of individual sources of evidence was not performed.

Data Charting, Extraction, and Synthesis

Data were independently extracted and tabulated by 2 authors (TAT and FA) using a data charting form in a Microsoft Excel spreadsheet that was approved by the research team. The data charting form was used to extract information regarding the population characteristics, training attributes, and identified reported outcomes. The use of a data charting form or table was recommended by Arksey and O’Malley [27] and Levac et al [28]. The data extraction form was piloted by 2 authors on 5 studies prior to use to ensure that all necessary data were captured appropriately. Where necessary, the corresponding authors for relevant studies were contacted via email to clarify aspects of their work prior to final extraction.

The bibliographic data extracted included the first author, title, year of publication, and country in which the study took place. CHW descriptions included CHW cadre name, number of CHWs trained, sex, age, education, disease domain, employment type (full-time or part-time), duty station (community or facility-based), remuneration (presence or absence), and employer. Training features included type of training, modality, synchronicity, venue, device type, availability online or offline, technology medium, duration, number of sessions, and pedagogical approaches. Outcomes included measurements, results, and limitations. The study design information was extracted and categorized into cross-sectional cohort, longitudinal cohort, quasi-experimental, and randomized controlled trials. The analytic approaches were classified into qualitative, quantitative, and mixed methods.

Once the data had been transferred into the spreadsheet, one author (TAT) reviewed the information and organized it into one of the following categories: (1) learning outcomes, (2) implementation outcomes, and (3) clinical outcomes. The descriptions of the included studies were analyzed and organized in tabular format, accompanied by a narrative summary.

Results

Search Results

The initial search of the 10 databases yielded 892 articles (see online Table S2 in Multimedia Appendix 1). After the exclusion of 173 duplicates, 719 papers were identified for initial screening. After title and abstract screening, 32 studies were identified for full-text review. An additional 13 articles were selected from reference lists and other sources, totaling 45 articles for full-text review. After the full text review, 18 studies [36-53] were identified for data extraction, and 27 were excluded because they did not meet inclusion criteria. Reasons for exclusion at full-text screening can be found in the PRISMA flowchart (Figure 1).

‎

Characteristics of Included Studies, CHWs, and Training

All 18 [36-53] studies in this review were published from 2009 to 2024. A majority of studies were conducted in Asia (n=13) [37-43,45,47,48,51-53], especially in India (n=8) [37,39,41-43,51-53], a few were conducted in Africa (n=5) [36,44,46,49,50], and one study was conducted in the Caribbean [37]. None were conducted in Latin America, Eastern or Southern Europe, or Oceania (Figure 2). The countries where the studies were undertaken are summarized in Table 1.

‎

Six different CHW terms were identified across the 18 studies, with variations being noted between studies in terms of CHW sex, type of employment contract, duty station, remuneration, and employer. The terms “community health worker” (n=9) [40,43-50] and “accredited social health activist” (n=5) [41,42,51-53] were commonly used. “Lay counselor,” “primary counselor,” “female community health volunteers,” and “acute care provider” were used in one study each. The majority of studies reported employing female CHWs only (n=11) [37-43,45,51-53] and the others employed males and females. The age of CHWs varied from 18 to 54 years. Most CHWs worked part-time (n=10) [38,39,41-44,46,50-53]; were community-based (n=14) [37-46,50-53]; were paid a monthly salary, incentive, or honorarium (n=14) [36,38,41-43,45-53]; and were employed by the government (n=15; Table 1) [36,38,40-44,46-48,50-53].

A total of 4 study designs were identified. Studies were cross-sectional (n=9) [36,37,39-41,43,47,48,50], cohort (n=1) [46], quasi-experimental (n=5) [38,49,51-53], and randomized controlled trials (RCTs, n=3) [42,44,45]. With regard to the analytic approaches applied, most studies were quantitative (n=10) [37-40,44-49] and mixed methods (n=6) [42,43,50-53]. Two used qualitative methods only [36,41].

The training modalities described in the studies were digital learning (n=14) [37-46,50-53] and blended learning (n=4) [36,47-49]. Some training approaches made use of tablets or smartphones (n=6) [36,37,42,44-46]. This was followed by computers (n=5) [40,43,47-49] and basic or feature phones (n=1) [50]. There were some studies that reported the use of multiple technologies (n=5) [39,41,51-53]. The training approaches were all implemented by researchers (n=18) and supported by government (n=2) [46,50] and nongovernmental organizations (n=1) [49]. The training formats were synchronous (n=8) [38,41-43,49,51-53], asynchronous (n=5) [37,40,44,46,50], and both (synchronous and asynchronous) (n=5) [36,39,45,47,48].

The training venue, duration, focus, and pedagogical approaches varied between studies. The majority of the training occurred in the community (n=14) [37-46,50-53] compared to a classroom (n=3) [47-49] or both (n=1) [36]. The trainings focused on the following disease areas: COVID-19 prevention, testing, and management (n=1) [38]; diabetes prevention (n=2) [47,48]; general conditions care or management (n=2) [37,43]; HIV testing services (n=2) [36,49]; maternal and child health (n=5) [40,46,51-53], mental health or depression screening and counseling (n=5) [39,41,42,45,50]; and pneumonia recognition, treatment, and prevention (n=1) [44]. Most of the training was accessible online (n=8) [36,38,39,41-43,46,53] rather than offline (n=6) [37,40,44,45,49,50] or both (n=4) [47,48,51,52]. The duration ranged from a minimum of one day to a maximum of 8 months. They lasted from one to 48 hours. The main pedagogical approaches employed were individual learning (n=5) [39-41,44,50] and group discussion (n=4) [47,48,51,52]. Other approaches included practicing scenarios, modeling, and receiving feedback to reinforce information [36-38,42,43,45,49] (Table 2). Full details of the training are summarized in Table 3

Training Outcomes

Fourteen studies reported learning outcomes (Table 4). The learning outcomes reported in the studies were knowledge, attitude, and behaviors. Assessment of knowledge was made mainly through pretraining and posttraining tests (n=13) [37-40,42,44-48,50-53]. All 13 studies reported improved knowledge outcomes after the training. Two studies reported improved attitude. One study from India that implemented both synchronous and asynchronous learning approaches reported enhanced attitude (confidence) in delivering maternal mental health services [39] and another from Rwanda demonstrated asynchronous learning improvements in providing support to patients with mental health conditions in the community [50]. One study from Bangladesh reported that asynchronous learning improved behavior in counseling couples on all available contraceptive options and child spacing [40].

The reported outcomes varied across the studies. Most studies presented one outcome from one category. Table 5 summarizes the available literature, notes gaps of digital training outcomes where little or no research has been conducted, and highlights areas for future research. Ten studies reported implementation outcomes [36,37,41-44,49-53]. Seven studies used synchronous learning [41-43,49,51-53], 3 asynchronous learning [37,44,50] and 2 studies implemented both approaches [36,39]. The implementation outcomes reported were acceptability, appropriateness, feasibility, and fidelity. Acceptability included perceived effectiveness, perception of remote training, usability, and satisfaction (n=10) [36,39,41-44,49-52]. Acceptability was assessed posttraining mainly through questionnaires, written qualitative feedback, or in-depth interviews. Three studies reported that participants were satisfied with the training [39,47,50] and the other 7 reported that the training was agreeable. Two studies reported on appropriateness and found the training suitable for capacity building for CHWs [41,50].

Less than half of the studies reported on the feasibility of participating in digital training, and all found the training was successfully carried out (feasible) [37,39,41,42,51-53]. Feasibility assessment mostly involved posttraining surveys or qualitative interviews or feedback or reviewing training logs. This included responses to questions about digital device familiarity [41] to the review of log-in attempts and the number of hours spent in the training system [51]. Some studies reported technical challenges that affected training feasibility. Three studies reported poor internet connectivity challenges, including lag or slow internet [38,41,42]. Four studies reported poor cellular network infrastructure [36,51-53]. One mentioned other technical glitches [39] and one discussed power cuts hindering charging tablets [36].

One study from Rwanda reported high fidelity to the asynchronous training schedule [50]. Another study from Malawi found synchronous learning improved fidelity in providing counseling services to patients as intended based on a CHW checklist [49]. This study also reported improved effectiveness on index case testing indicators, such as sexual contacts elicited [49]. This was the only clinical outcome. Full details of the outcomes for individual studies are summarized in Table 3.

Discussion

Principal Findings

This scoping review brings together studies and reports of the characteristics and outcomes of digital training for CHWs in LMICs. Despite a rigorous search across multiple databases and broad inclusion criteria, we identified only 18 studies, which highlights the dearth of robust evaluations of digital training in LMICs. Yet, the review indicated that despite technological and infrastructural challenges, training CHWs using digital technologies was acceptable and feasible and led to knowledge acquisition. This body of evidence was largely descriptive, with few studies using experimental or quasi-experimental designs. Most reported only on learning outcomes, with very few reporting on implementation fidelity or clinical outcomes as a result of CHW training. Additionally, the majority of studies in this review only measured Kirkpatrick [25] Levels 1 and 2. A few measured Level 3 behaviors, and only one measured Level 4 results. This highlights the need for additional research evaluating additional level 3 and 4 outcomes.

This is the first review of digital training among CHWs in LMIC. Other reviews have focused on digital training primarily for professional health workers, mostly from HICs [15,16]. These reviews have similarly found relatively few studies, mostly descriptive in nature, with only knowledge outcomes evaluated. A scoping review that focused on digital training for rural professional cadres [16] identified few nonexperimental studies (n=5) and focused mainly on knowledge acquisition. Similarly, a systematic review that evaluated digital learning for medical education in LMICs [54] found two-thirds of the studies identified were from upper-middle-income countries, and only 4 were RCTs. Additionally, a review of digital learning for professional HCWs [15] found a few studies (n=14) conducted in LMICs; only 7 were RCTs and mainly reported knowledge outcomes.

Most of the evidence gathered in our scoping review showed that the digital or blended learning training resulted in significantly improved knowledge of CHWs after the training, compared to before, and is consistent with other learning modalities [55,56]. Similarly, our results are consistent with other studies in medical education reporting on the effectiveness of digital or blended learning in enhancing knowledge acquisition among certified health professionals in both LMICs [57,58] and HICs [59-61]. This supports the finding that digital training accounts for the knowledge needed to perform tasks across disease entities and health cadres in both high- and lower-economic settings. Furthermore, a systematic review comparing the effectiveness of blended learning to traditional learning for professional or certified HCWs [17] concluded that blended learning demonstrated consistently positive effects on knowledge outcomes when compared with traditional learning in medical education. Our work extends these findings to CHWs’ knowledge to perform delegated assignments.

A few studies included in this scoping review addressed acceptability and appropriateness [41,50]. Feedback from CHWs suggested that they found digital or blended training acceptable for learning new information and appropriate as a mode of training. This has also been reported in other studies where digital training was acceptable among CHWs in diverse settings for data collection training [62]. Acceptability and appropriateness are important factors for training completion and skills application.

Overall, digital training was feasible, though approaches differed considerably. This is consistent with other studies with professional cadres [15,63]. For example, studies in Kenya [14] and Nigeria [57] found digital training for HCWs was feasible at scale. These studies highlight the potential feasibility of implementing digital training interventions in LMICs. Digital training programs in this review included the use of various technologies, including tablets, mobile phones, or feature phones. Some training approaches did not require internet or physical infrastructure, which may be lacking or unreliable in low-resource settings. For example, CHWs could access the training on a feature phone in the community without needing internet access or being in centralized locations [50]. The use of mobile devices made it possible for synchronous or asynchronous learning models to occur. Both models appear to be of benefit. Asynchronous learning offered CHWs the opportunity to participate in the training in a flexible manner at a time and place that suited them without missing any material due to other demanding personal or work-related responsibilities. Synchronous learning offered CHWs the opportunity to participate in group discussions, ask questions in real-time, and practice and receive feedback. Furthermore, synchronous sessions provided the trainees with access to technical support periodically. With the increased use of smartphones in LMICs, digital training could successfully be implemented in LMICs [64].

This review also underlined that using digital training programs in LMICs will require technological challenges to be overcome. Technical challenges, including slow or poor internet and cellular networks, were frequently cited as barriers to training completion. For example, Muke et al [42] reported that due to the challenge of poor internet connectivity, some participants were unable to access all content from the training modules because videos did not play. Preloading training material could eliminate reliance on connectivity and may ensure availability offline [65]. In addition, using feature phones demonstrated the possibility of mitigating network challenges [50]. The design of the digital training program is imperative. Simpler platforms may be easier to use and scale, and good accessibility may provide efficiencies that enable greater impact of the training.

Policy and Practice Recommendations

Given the global shortage of professional health workers, recruiting and training CHWs is one solution to overcome this gap in LMICs. The integration of digital training programs into national health strategies could play a pivotal role in achieving high-quality health care outcomes. LMICs should consider investing in digital infrastructure, including technology devices and literacy for lay cadres, to enhance accessibility to training and scalability of digital solutions. Tailoring and co-designing digital training programs with CHWs to ensure relevance and practicality will encourage engagement and sustainability of the digital tools. These recommendations support the WHO’s suggestion of using digital training to supplement health workers’ continuous capacity-building efforts [66].

Limitations

Our review had important limitations. Some studies did not provide enough details on training characteristics and outcomes. There was a potential to give greater emphasis to the studies that provided more detailed information when using narrative synthesis in data analysis. To overcome this, we contacted all authors for more information and were able to obtain some additional data that were not published in the articles. Additionally, we extracted data according to a predesigned structure and had 2 authors (TAT and FA) extract data independently.

Conclusion

CHWs play an essential role in providing health care and improving health outcomes in LMIC environments. Training must continue to be a core component of improving CHW skills. In LMIC environments that often have lower technological literacy and infrastructure, we sought to understand what role digital training could play. Digital trainings had a broad range of attributes. Few evaluations of digital training for CHWs in LMICs were identified in this review, in spite of a proliferation of such trainings. Furthermore, most evaluations lacked experimental designs. The existing evidence suggests that digital training can impact knowledge in lay health workers in LMICs; more rigorous studies with a broader range of outcomes are needed. Such additional work is needed to ensure lay health workers are well-capacitated to deliver critical health services.