This study employed a quasi-experimental design with a mixed methods approach to evaluate the impact of the intervention. Figure 1 visually represents the study design. It illustrates the selection of CUs from the larger population and their assignment to intervention and comparison groups. This was a non randomized implementation evaluation where CUs were purposively selected in consultation with the Department of Health in Busia County to ensure representativeness and feasibility. The matching process is based on size, population characteristics, workforce availability, and health facility spread to minimize pre-existing differences that could confound the results. Table 1 outlines the digital communication channels and measurement tools used in the study. These included SMS, a hotline, USSD, and IVR platforms. The study's aim was to generate practical evidence on how well each channel performs in harder-to-reach, remote settings, and to identify which tools are most appropriate for different operational needs.

During project implementation, all participants provided informed consent, either verbally or in writing, depending on literacy levels. CHWs were trained on ethical practices, including data confidentiality and consent procedures. For minors aged 15–17 involved in the intervention (as eligibility was later expanded), assent was obtained alongside guardian consent in accordance with ethical guidelines. Additionally, study ethical approval was sought from the JKUAT ethical review committee.

The study was conducted in Teso North Sub County, Busia County, Kenya. Teso North was chosen due to the existing healthcare access challenges, especially concerning antenatal and postnatal care. The evaluation focused on 10 Community Health Units within Teso North. Six of these CHUs (3 peri-urban and 3 urban) were assigned to the intervention group, while 4 served as the comparison group.

As shown in Figure 1, ten community units were selected from a total of 26 in Teso North, Busia County, Kenya. Six of these CUs were assigned to the intervention group, while the remaining four served as the comparison group.

The study population consisted of households within the purposively selected Community Units (CUs) in Teso North, Busia County, Kenya, that had children under 6 weeks old and/or pregnant women of reproductive age (18–49 years). All eligible households within the chosen CUs were included in the study, forming the total sampling frame.

A cluster sampling approach was used, where the CUs represented the clusters, and all eligible households within those clusters were included. This targeted sampling ensured the evaluation focused on the population most relevant to postnatal care interventions. The study included 129 surveys completed by mothers (20 telehealth service non-users and 109 users of the intervention) and seven FGDs with mothers. Additional FGDs were conducted with CHWs and local government staff to gain broader perspectives on the intervention's impact.

The project aimed to design and implement an integrated virtual and physical 24-hour triage solution, allowing clients to self-screen and be directed towards appropriate care pathways. The goal was to enhance care-seeking behaviours, reduce delays in care, and improve efficiency in delivering care.

This intervention integrated a telemedicine solution into an existing community health platform, with the goals of:

Improving the uptake of essential healthcare services, specifically postnatal care.

The project followed a 24-month schedule organized into three phases: a six-month period of exploration and co-design, six months of prototyping and early testing, and a twelve-month pilot implementation and evaluation phase. The evaluation itself was carried out in three stages, as outlined in Table 2: Telehealth Intervention Timeline and Rollout Phases.

The data collection for this study employed a mixed-methods approach, gathering both quantitative and qualitative data across three distinct phases:

Both quantitative and qualitative data analysis techniques were applied. Quantitative data were analyzed using Stata® version 15 and R version version 4.5.1 (2025-06-13 ucrt), employing descriptive, exploratory, and inferential statistics. Impact on on-time PNC visits were quantified using the difference in difference, generalized Cohen's d effect size. To examine determinants of receiving on time postnatal care, the averaged logistic regression model using generalized estimating equations (GEE) with an AR(1) correlation structure was fitte. For qualitative data, the audio recordings were transcribed and imported into Dedoose for coding and further analysis. Analysis across all transcripts was conducted using a constant comparative method to identify emerging themes and their repetitions and variations.

The hybrid model and virtual platform supported 388 households, reaching 551 clients (388 mothers and 163 infants) through 39 Community Health Providers. Program data indicated that the majority of the registered mothers were young adults aged between 20 and 29 years at 59% (n = 228), with the majority enrolling during their second trimester at 27% (106 women). 113/129 (88%) of registered women surveyed were married, while 12% (16) self-reported as single mothers. Regarding education, more than half of the mothers had attained primary and secondary education at 62(48%) and 50(39%), respectively. By the end-line study in September 2024, 77% (99/129) had delivered. Among those surveyed, 88% (113/129) were married and 12% (16) were single mothers, with most having primary, 48% (62), or secondary, 39% (50) education. It was critical to note that almost half of these mothers had no source of income at 48% (n = 62), translating to 50% of the mothers reporting to have no independent income whatsoever (n = 65). By the end-line study in September 2024, a significant majority (77%, 99 women) had given birth, while 60 eligible women (either pregnant or within six weeks postpartum) remained enrolled at the time of reporting. These results highlight telehealth's potential to engage women earlier in their pregnancy, despite the project's primary focus being on postnatal care.

Almost half of the registered women and the same proportion of infants utilized the Telehealth services. High adoption of Telehealth services was noted in Moding and Komiriai CUs, which were the Phase one intervention sites. While the two phase two sites, almost at par with about 15% contribution each to the total adoption rates. High adoption was noted in Moding (Phase 1) and Komiriai (Phase 2) CUs, both accounting for 58% (111/189) of all adopters (Table 3).

According to Health X data 49% of women took up the service (engaged with the clinical hotline) at least once after registration. 85% (109/129) of women surveyed reported to be users of telehealth services (Table 4). Between June 2023 and August/September adoptions rates (proactive engagement by the register user at least once after registration was 49% (189/388 women registered). The definition of adoption (measured by clinical calls with HealthX) is as follows:

High adopters: Engagement 4 times or more after registration were classified as high users

CHWs observed that high adopters of the virtual health service were typically individuals with a high perceived need, such as first-time mothers, adolescents, or those with a history of obstetric complications. They also tended to have slightly higher literacy and education levels, owned phones, or had supportive partners willing to share their phones and access to the service. A key driver of adoption was the quality of virtual customer service. Many users reported receiving more personalized care, longer consultations, and access to broader medical expertise than they typically received from local health facilities or CHWs.

In contrast, non-adopters often cited limited time and low perceived value of the service, believing they would be referred to a facility for care regardless. CHWs reported that these mothers often had low literacy levels, poor initial customer service experiences (such as missed call-backs or long wait times), difficulty using the platform independently, and limited phone access often sharing devices with CHWs, partners, or in-laws, or using inactive SIM cards. CHWs also noted that higher-income or working-class mothers were more likely to have health insurance or prefer private health services.

Overall, while the majority of the virtual CHWs delivered relatively well in enrolment, a strong backbone of performance management by supervisors was required to maintain momentum. CHW marketing of the virtual service primarily focused on the hotline as an emergency or clinical counselling service. Majority, 62% (24/39 CHWs) trained as virtual ambassadors onboarded a satisfactory level (10–28) of clients since project launch.

Based on Health X consultation records, strong client preference for the toll-free clinical calls remains the predominant channel accessed and used by households, with very limited use of the other USSD or IVR enabled functions and features (hail your CHW, danger sign risk assessment, report a delivery). SMS reminders likely prompted positive behaviour change, with 64% (83/129) of women reporting the reminders improved their timely clinical ANC and PNC attendance. On exploring how the respondents used Telehealth. The majority of the respondents (n = 90, 83%) used the platform to speak to/consult a doctor when they or their baby were ill. This was followed by 68 respondents who reported registration of themselves or their baby on delivery, while very few (n = 6) used the platform to hail their CHW.

Figure 5 illustrates utilization trends and service demand. The figure demonstrates that consultations for maternal and infant danger signs were the dominant use case (90%), confirming the model's relevance for emergency triage and maternal decision support. Secondary uses such as registration, follow-up, and nutrition consultations illustrate the gradual expansion of user engagement from reactive to proactive health behaviors. This progression aligns with the project's outcome framework, which aimed to move clients along a continuum from basic awareness to self-directed service use.

Both in IVR and USSD, the risk assessment tool (users can hear a list of danger signs associated with new mothers or infants in the PNC period) and the “Hail your CHW” had very limited uptake. In part early insights suggest this may have been due to a) low perceived value (clients already have the CHW phone number) b) a lack of active marketing by CHWs of these features c) that calling the virtual Doctor directly via the hotline was a more reassuring and direct experience for the user and d) that the virtual triage (e.g., danger signs) would escalate to a virtual Dr consultation).

The average number of PNC touchpoints between CHW and client within 6 weeks of delivery was, on average, statistically (p-value: 0.003) slightly higher in the intervention site at approximately 5 visits (4.99) compared to approximately 4 visits (3.96) in the control community units. However, these averages mask a greater monthly fluctuation of visits in the intervention sites during pilot implementation. Using the 20% treamed means and Winsorized standard deviations, the generalized Cohen's d comparing the intervention and control was 3.67 (95% CI: 1.64;8.02), indicating a very statistically meaningful effect. This implies Even after accounting for non-normality and outliers, the Intervention produced a substantial difference in with six weeks compared to the Control.

Survey with mothers revealed that the experiment group, on average, has a significantly higher number of PNC visits (4.99) compared to the control group (3.96) but with more variability. Hence, while the intervention in the experimental group may lead to more visits on average, the results are more variable and less consistent. Across June 2023–September 2024, trends for the unique women/infants visited at least once by their CHW within 48 h were on average 85% compared to the control of 87%. Women/infants receiving all three PNC touchpoints’ on time within 6 weeks, was not positively impacted by the virtual platform when we observe the trends between intervention and control site (334/625 (53%) vs. 965/1589 (61%), however due to the difference in the total number of women being served, the intervention site experienced significantly fewer cases of women/infants lost to follow-up during that period (291 compared to 625).

Three variables in Table 6 emerged as statistically significant predictors of on time PNC: previous PNC status (LAG PNC), previous follow-up activity (LAG follow up counts), and maternal age.

First, prior on time PNC status was the strongest predictor of current PNC uptake. CHWs who had done timely PNC visit in the previous period had markedly higher odds of doing another timely PNC visit again (OR = 55.21; 95% CI: 7.00–435.59; p < 0.001). This substantial temporal persistence suggests that timely PNC may reflect underlying behaviors and interactions patterns that CHW continue over time. Once a CHW woman engages successfully with the woman during the postnatal care, this behavior appears highly likely to repeat.

Second, higher follow-up activity in the previous time period was associated with increased odds of on time PNC in the current period (OR = 9.71; 95% CI: 1.08–86.91; p = 0.042). Although the confidence interval is wide, indicating variability, the direction and significance suggest that consistent or intensified CHW engagement may help facilitate timely PNC in subsequent periods.

Third, maternal age was positively associated with on time PNC, with each additional year of age increasing the odds of timely PNC by approximately 16% (OR = 1.16; 95% CI: 1.04–1.29; p = 0.007). This indicates that younger mothers remain a high-risk group for missed or delayed PNC services.

According to FGDs and KIIs with CHWs and Supervisors, the introduction of the hybrid virtual model resulted in several system efficiencies and benefits:

Improved follow-up by making it easier for CHWs to monitor pregnant and post-partum mothers through notifications

The telehealth intervention exceeded its registration targets, enrolling 388 households and 551 clients, demonstrating strong community engagement and improved access to maternal and newborn health services. The significant proportion of women enrolling in their second trimester indicates that the platform effectively encouraged earlier engagement in antenatal care. Moreover, the observed increase in postnatal care (PNC) touchpoints reflects a positive shift toward consistent maternal and neonatal health utilization, aligning with evidence from a retrospective study that reported similar increases in PNC attendance and postpartum depression screening following telemedicine adoption (8).

Qualitative findings from the FGDs and KIIs highlighted significant socio-cultural barriers that influenced women's ability to fully participate in virtual consultations. Gender dynamics were particularly pronounced, with reports of spousal refusal, male control over household phones, and restrictions on women's independent decision-making, patterns consistent with broader evidence showing that women often have limited agency over mobile phone access and use in low-resource settings. Low general and digital literacy further constrained women's confidence and capacity to navigate teleconsultation platforms, echoing previous studies demonstrating how literacy gaps disproportionately affect women's adoption of digital health tools (9). Together, these findings underscore the importance of designing telemedicine interventions that explicitly address gender norms, household power dynamics, and literacy barriers to ensure equitable access and meaningful participation.

Comparable findings have been documented in other African contexts, further validating the role of telehealth in improving maternal and child health outcomes. In Uganda, the Safe Delivery mHealth project demonstrated significant increases in skilled birth attendance and timely referral completion through mobile-based decision-support tools for community health workers (10). Similarly, in Nigeria, the HelloMama and Smart Start mHealth initiatives improved antenatal visit compliance and immunization uptake by providing structured voice and SMS messaging tailored to maternal needs (11). The use of mobile health tools by community health workers in sub-Saharan Africa was linked to higher uptake of antenatal care, more facility-based deliveries, and improved postnatal care attendance. To maximize the impact of these programs, it is important to recognize and address both the social and technical challenges that affect mHealth adoption (12).

Evidence from other continents also supports these trends. In Bangladesh, the Aponjon mobile health service reported increased antenatal care attendance and early postpartum check-ups, largely driven by personalized mobile reminders and remote health counselling (13). In Peru, telemedicine-supported maternal care in rural Andes communities enhanced follow-up adherence and early complication detection, demonstrating the effectiveness of hybrid care models in geographically challenging environments (14). Models such as “Mobile for Mothers” and the ReMIND mHealth intervention in India demonstrate that digital solutions can expand access to antenatal, intrapartum, and postnatal care, especially for low-literacy users. These studies emphasize that digital innovations must be carefully conceptualized, co-designed with communities, and rigorously evaluated to ensure equity and reach among vulnerable populations (15).

Collectively, these studies complement the Kenyan telehealth model's findings by confirming that hybrid, community-linked digital interventions can bridge geographical and socioeconomic barriers, improve service continuity, and promote proactive health-seeking behaviors. The convergence of evidence across Africa, Asia, and Latin America underscores telemedicine's global potential as a transformative enabler of equitable maternal and newborn care.

The intervention achieved a high referral completion rate of 92%, with significantly more referrals completed in intervention sites compared to control sites (3,630 vs. 1,873). Although control sites exhibited a slightly higher percentage of completed referrals (97% vs. 92%), they also recorded a larger number of missed referrals (329 vs. 62), indicating that telehealth facilitated more efficient triage, case prioritization, and timely follow-up. This efficiency likely contributed to reduced facility congestion and improved continuity of care for maternal and neonatal clients.

These findings align with previous evidence showing that telehealth interventions can strengthen referral linkages and continuity of care by enabling faster communication and remote monitoring during prenatal and postnatal periods (16). Similar outcomes have been reported in Ethiopia, where an mHealth referral tracking system increased referral completion among community health workers, real-time data exchange and timely pregnancy registration contributed to increased utilization of maternal health services across all stages of care (17), and in Ghana, where mobile decision-support tools improved timely referrals for obstetric emergencies (18).

SMS reminders and digital alerts also played a vital role in promoting appointment adherence and proactive client engagement. Evidence from Tanzania demonstrated that automated SMS reminders increased antenatal visit attendance by 30% (19), consistent with findings by Knop et al. (20), where targeted SMS interventions improved timely child immunization rates. Collectively, these results underscore that mHealth-enabled communication channels such as SMS, IVR, and teleconsultation platforms can effectively bridge service delivery gaps, improve follow-up adherence, and enhance patient-centered care in resource-limited settings. This reflects how behavioral nudges and low-cost communication tools can reinforce patient accountability and care continuity within telehealth ecosystems.

The study highlights the effectiveness of a hybrid telehealth model that integrates virtual consultations with community-based support from CHWs, fostering trust, community ownership, and continuity of care. However, challenges in integrating telehealth into routine CHW workflows underscore the need for sustained digital capacity-building, human-centred design, and supportive supervision (11). This aligns with broader evidence from Africa indicating that telemedicine can improve maternal outcomes, while persistent barriers such as infrastructure limitations and digital literacy gaps continue to constrain scale-up (8).

Complementary evidence from maternity care during the COVID-19 pandemic demonstrates that blended telehealth approaches, combining virtual group-based education with individual antenatal assessments, can be successfully embedded within routine clinic workflows. Buultjens et al. show that a hybrid, interdisciplinary telehealth model supported maternal psychological wellbeing when delivered by a small but committed clinical team, emphasizing the importance of continuity, relational care, and structured group engagement rather than telehealth delivery alone (21). Published in Women and Birth, this work reinforces the view that telehealth functions most effectively as part of a broader digital care ecosystem, rather than as a stand-alone operational substitute for in-person services. Together, these findings extend earlier telemedicine literature by positioning hybrid telehealth models as catalysts for health system redesign, enabling innovation diffusion, workforce role evolution, and patient empowerment. Comparable experiences from Indonesia, South Africa, and India further demonstrate that telemedicine success depends not only on technical infrastructure, but also on leadership support, digital adoption capacity, and user engagement frameworks (6, 20). For instance, the Indonesian hospital digital adoption model highlights how innovation readiness and data interoperability determine service performance in digitally transforming hospitals (20). Similarly, the study on Exploring consumer sentiments in telemedicine identifies that trust, perceived utility, and human connection drive sustained engagement (6). These insights expand the interpretation of Kenya's findings beyond access metrics, positioning telehealth as part of a larger digital innovation continuum that links community health with institutional transformation. Parallel experiences across Africa reinforce this trajectory. A randomized trial in Kenya and Uganda found that CHW-facilitated telehealth for hypertension achieved 86% control rates at 48 weeks, outperforming clinic-based care (22). In Malawi, CHWs adopting mHealth tools reported improved confidence and efficiency, though persistent barriers included connectivity and training gaps (22).

Systematic reviews also affirm telemedicine's benefits for chronic disease management, showing significant improvements in HbA1c and systolic blood pressure outcomes (25). Such global comparisons position telehealth as not only a service delivery tool but as a foundation for patient-centered digital transformation. These findings align with work showing that telehealth enhances chronic-disease management particularly diabetes by improving health outcomes, patient satisfaction and accessibility (3). Notably, in the hybrid model studied, the high reliance on hotline consultations as a preferred communication channel underscores the feasibility of integrating telehealth into community-health-service delivery. This alignment suggests that pairing digital consultations with trusted CHWs and local support systems offers a promising pathway for scale. Nonetheless, successful deployment at scale will require addressing device access, connectivity, training, supervision, data systems and sustainable funding models.

Comparable telehealth experiences from other low- and middle-income countries further demonstrate that blended digital–community health models can be successfully scaled. Telemedicine has become a central component of contemporary healthcare systems (3), especially as countries sought alternative care pathways during the COVID-19 pandemic. In India, the expansion of these services has been slow, largely because healthcare consumers have not adopted them at the expected rate (7). In Indonesia, platforms such as Halodoc and Alodokter supported millions of remote consultations, helping ease pressure on already strained health facilities (24). Likewise, Bangladesh and India (7) rapidly broadened public access to telemedicine services during the pandemic, demonstrating how digital health tools can maintain service continuity when in-person care is disrupted. Indonesia's Digital Midwife and the Philippines’ National Telehealth Service Program (NTSP) demonstrate how digitally supported CHWs extend clinical reach to underserved regions (25, 26). These experiences mirror the Kenyan pilot's outcomes, illustrating a shared path toward digitally enabled, equity-focused primary healthcare. Based on the findings of the studies, the initiative effectively reduced access barriers by allowing remote consultations, minimizing travel costs and delays. However, sociocultural dynamics, gender norms, and connectivity inequalities limited enrolment among certain groups. The study also emphasizes that bridging these inequities requires inclusive design, digital literacy training, and integration with national infrastructure to ensure equitable service reach.

In this study, the telehealth initiative effectively tackled healthcare access challenges in remote and underserved communities. The ability to consult healthcare providers via mobile phones reduced the need for physical travel, addressing both cost and distance barriers. However, sociocultural factors such as spousal approval and limited program awareness impacted enrolment rates, particularly among certain demographics. The study also identified gaps in phone access and network connectivity, which need to be addressed to ensure equitable access to telehealth services. This study supports the growing evidence that adopting telehealth technologies can enhance the antenatal care experience for women and improve access, while potentially reducing healthcare costs without negatively affecting maternal or neonatal health outcomes. However, as highlighted in recent literature (e.g., Telehealth in antenatal care: recent insights and advances), further research is needed to assess the impact of telehealth on rare but critical outcomes such as maternal and neonatal mortality (27). Additionally, the study echoes concern from (28) Berkley et al., which highlight that many high-risk children go undetected while low-risk children often receive unnecessarily intensive care. This imbalance underscores the importance of integrating telehealth solutions to ensure timely identification and escalation of high-risk cases, ultimately optimizing healthcare resource allocation and improving survival outcomes.

This study aimed to produce evidence that can inform future telehealth policy decisions, particularly those related to the broader socioeconomic effects of virtual care. Although rapid technological progress has strengthened telehealth delivery and motivated the development of federal reimbursement proposals, legislative action has slowed due to ongoing concerns among policymakers. Because policy debates often move quickly, decision-makers require timely, synthesized reviews of the most reliable evidence to determine whether telehealth legislation should be advanced, amended, or withheld (29). This research responds to that need by supplying relevant insights to support informed deliberation.

Optimize a digitized closed loop referral system (CLR): A comprehensive, integrated digital platform that links CHWs, supervisors, telehealth clinicians, and health facilities would streamline referral initiation, triage, tracking, and completion. Such a system would reinforce shared decision-making, reduce uncertainty in referral pathways, and improve feedback mechanisms to both clients and providers. By connecting all stakeholders in real time, a digitized CLR would support seamless coordination, enhance continuity of care, and advance a more responsive, data-driven telehealth ecosystem. Families often emphasize that high-quality referral decisions rely on collaborative communication and clear, specific information such as a shared understanding of why the referral is needed and which provider or subspecialist the client should see. Challenges arise when frontline providers experience uncertainty about whether referral is warranted or lack clarity on the most appropriate specialist for the case, which can limit meaningful shared decision-making and hinder timely escalation of care (30). Within our telehealth pilot, the hybrid model helped address several of these barriers by enabling CHWs and clients to access real-time clinical support through the hotline, improving clinical confidence and the precision of referral decisions. However, to fully capitalize on these gains and ensure consistency across all service delivery points, it is essential to strengthen the system infrastructure for managing referrals.