This study is significant for its potential to reduce neonatal mortality in Kenya, where neonatal deaths account for nearly half of all cases of child mortality under the age of 5 years, highlighting the urgent need for targeted interventions. Despite global and national efforts, progress toward SDG target 3.2, which aims to end preventable newborn deaths and deaths under the age of 5 years by 2030, remains slow, particularly in resource-limited settings. The study evaluates the feasibility, applicability, and usability of neonatal risk predictor variables incorporated into the ML model proposed by Shukla et al [13], which was initially validated in India. By testing these predictors in Kenyan MOH facilities, the study assesses their adaptability to local clinical and infrastructural contexts. Through a user-centered approach, it also explores health care professionals’ experiences, perceptions, and interactions with the tool. If applicable, the model could enable early identification of high-risk neonates, support timely interventions, inform evidence-based policies, promote digital health adoption, and strengthen neonatal care systems, accelerating progress toward SDG 3.2.

A mixed methods approach will be used, combining qualitative and quantitative methods to enhance generalizability and statistical rigor and strengthen the credibility of the findings [22]. The study will be conducted in 3 sequential phases. First, key informant interviews (KIIs) will be conducted to capture contextual insights. The second phase is an intervention period, during which a quantitative neonatal risk survey tool is integrated into clinical workflows to evaluate the contextual feasibility, applicability, and usability of the ML risk predictor variables proposed by Shukla et al [13]. Third, a postimplementation questionnaire will be administered to assess whether the survey tool’s intended outcomes were achieved. Phase 2 (intervention period) and phase 3 (postintervention survey) will follow a longitudinal design, observing the same research participants at 2 time points over a 4-month period to examine changes in outcomes associated with the intervention prior to implementation [23,24]. Therefore, the study will fully leverage the strengths of both qualitative and quantitative methods. These include triangulation, methodological flexibility, in-depth insights, and a comprehensive understanding of the research phenomenon, all of which enhance the validity of the study’s findings compared with a single-method approach [25-27].

This will be the preintervention phase, which will involve conducting KIIs with key informants (KIs), including facility heads and neonatal department heads at selected MOH facilities providing neonatal care. The KIIs will capture contextual insights into participants’ lived experiences, current workflows, perceptions of neonatal risk interventions, potential benefits, anticipated implementation barriers, and preferred mode of deployment (paper based or electronic). These insights will enable the researchers to better understand the study context and stakeholder expectations, thereby informing the study’s planning and execution.

This will be the intervention phase, which will involve providing frontline neonatal care workers at selected MOH facilities with a paper-based neonatal risk assessment tool integrated into routine clinical workflows. Nominal-scale surveys will be used to operationalize and assess the applicability of the risk predictor variables incorporated into the ML model proposed by Shukla et al [13]. The tool will be deployed during routine neonatal care activities over a 4|-month period to enable real-world evaluation of its impact [28]. Longitudinal outcomes will be assessed through observations made during both the intervention and postintervention surveys, focusing on the tool’s feasibility, applicability, usability, and utility. Findings from this phase will inform the potential feasibility of implementing the intervention across Kenyan MOH facilities.

This will be the postintervention phase, during which quantitative surveys will be conducted with neonatal health workers. The primary objective is to assess whether the intervention’s intended outcomes were achieved, gain a deeper understanding of implementation barriers, highlight areas for improvement, and inform future deployment strategies [29]. Both Likert and semantic differential scales will be used to evaluate the predictors’ usability (ease of use and efficiency), UX (satisfaction and engagement), utility (perceived applicability for identifying high-risk neonates), and feasibility (practicality of implementation) within Kenyan MOH facilities.

Both standardized and custom validation tools will be used to comprehensively assess the usability, UX, interaction satisfaction, and utility of the intervention [30-33]. The instruments will include the Usability Scale Questionnaire, Questionnaire for User Interaction Satisfaction, and a Post-Study Neonatal Risk Predictor Utility Questionnaire. The Usability Scale Questionnaire evaluates UX with the neonatal risk predictor variables using a 5-point Likert scale (1=“strongly disagree,” 2=“disagree,” 3=“neutral,” 4=“agree,” and 5=“strongly agree”). The Questionnaire for User Interaction Satisfaction measures UX and interaction satisfaction using a semantic differential scale with bipolar adjectives such as “terrible”-“wonderful” and “difficult”-“easy.” The Post-Study Neonatal Risk Predictor Utility Questionnaire, a custom questionnaire developed specifically for this study, assesses how well the tool meets user needs and its overall utility. It includes items on ease of use, clarity of instructions, and relevance of content and perceived effectiveness rated on a 7-point Likert scale (1=“strongly agree,” 2=“agree,” 3=“slightly agree,” 4=“neutral,” 5=“slightly disagree,” 6=“disagree,” and 7=“strongly disagree”).

The findings from these tools will provide a comprehensive understanding of the intervention’s applicability and performance and support evidence-based decisions regarding the adoption and implementation of the proposed solution.

The target population will consist of health care workers providing neonatal care services in Kenyan MOH facilities. Specifically, 3 health facilities will be selected from 3 different counties: Kibera Community Health Centre – Amref (Nairobi County), Nazareth Mission Hospital (Kiambu County), and Matuu Level 4 Hospital (Machakos County). To enhance the national representativeness of the study findings, these facilities are located in diverse contexts across Kenya, an LMIC setting, reflecting the unique environmental and socioeconomic conditions in which they operate.

Kibera Community Health Centre is a public, community-based facility located within Kibera, one of Africa’s largest informal settlements characterized by widespread poverty and limited access to basic services, including health care. Nazareth Mission Hospital is a nonprofit, level 5 tertiary referral hospital in Kiambu County. It serves patients from across Kenya and neighboring regions, offering a wide range of specialized services. Matuu Level 4 Hospital, located in Machakos County, is a public level 4 health facility operating in a semiarid region characterized by low-income households, extreme temperatures, water scarcity, and food insecurity.

Data will be collected using both qualitative and quantitative methods, including KIIs, neonatal risk assessment tool and questionnaire, and postintervention questionnaires.

Qualitative data will be analyzed using thematic analysis to identify patterns and themes emerging from the KIIs. The analysis will follow an inductive coding approach involving systematic identification and classification of concepts, categories, and subcategories to generate overarching themes grounded in the data [5,43,44]. Data analysis will be conducted using the NVivo qualitative data analysis software (Lumivero). Thematic analysis is particularly valued for its flexibility and accessibility, making it a versatile qualitative research method. It enables in-depth exploration of complex datasets and supports the generation of rich, nuanced insights while remaining relatively straightforward to learn and implement [45].

Quantitative data obtained from the surveys will be analyzed using descriptive statistical techniques, including calculating frequencies and percentages and presenting results in tables and charts. Descriptive statistics are particularly useful for summarizing complex numerical data into manageable and interpretable formats, thereby facilitating the identification of key patterns and trends [46]. Data analysis will be conducted using Stata (StataCorp). This approach enhances clarity, improves communication of findings, and supports timely, evidence-based decision-making.

Ethics approval (SU-ISERC2949/25) was received from the Strathmore University Institutional Scientific and Ethical Review Committee in Nairobi, Kenya, on July 28, 2025, prior to this study. Various ethical considerations related to key research stakeholders will be carefully addressed, including concerns involving the researcher, research participants, and sponsoring organization [47]. The study will strictly adhere to these ethical guidelines to ensure compliance and integrity throughout the research process. Informed consent was obtained before engaging participants, after explaining the aim, benefits, and risks associated with the study and the mechanisms put in place to address them. In addition, privacy and confidentiality were ensured by omitting identifying details.

The study was funded in March 2025. Six participants (frontline neonatal health workers) had been recruited by the time of initial submission of the manuscript (August 2025). Data analysis is ongoing and is expected to be concluded in March 2026. Publication of study findings is expected by June 2026.

The research will provide a clear understanding of whether and how the neonatal risk predictors developed by Shukla et al [13] will be practically integrated into existing triage systems within Kenyan health care facilities. This includes identifying both enablers of and barriers to implementation in LMIC settings, assessing health system readiness, the availability of resources such as skilled personnel and infrastructure, and the level of institutional support for adoption.

The study will explore the applicability and clinical impact of the tool in accurately identifying high-risk neonates. It will investigate whether early risk identification using the tool improves neonatal outcomes or enables more timely interventions. Ultimately, the research will assess the tool’s potential to reduce neonatal mortality, aligning with SDG target 3.2, which aims to end preventable deaths of newborns.

The study will assess the usability and UX of the ML model’s neonatal risk predictors by examining how health care professionals interact with and perceive the usefulness of these variables in real-world clinical settings. This assessment will evaluate the predictors’ learnability and efficiency and overall user satisfaction, as well as the tool’s seamless integration into routine clinical workflows. Additionally, user feedback will provide valuable insights into the relevance, clarity, and acceptability of the predictors, helping inform potential improvements and ensure that they meet the practical needs of their intended users.

Evaluating the utility of the ML tool’s predictors is crucial for understanding their applicability in identifying neonatal risks. This evaluation will inform strategies for potential implementation and support the tool’s sustained use, maximizing its impact and application in practice. The tool is expected to have a positive impact on neonatal care processes across health facilities in Kenya. Anticipated benefits include improved triage decisions, enhanced applicability in neonatal care delivery, increased operational efficiency, and greater productivity among health care providers.

Dissemination of the study findings will occur at 2 levels. First, preliminary results will be presented to the stakeholders at a local conference. Second, the results will be shared through 2 peer-reviewed scientific journal publications. Throughout the dissemination process, the project will adhere to key principles aligned with responsible publication practices. These include maintaining ethical research conduct throughout the study, respecting intellectual property rights by properly acknowledging original authors and their work, and upholding academic integrity. The latter involves conducting the publication process with honesty, fairness, transparency, respect, and accountability, ensuring that the research makes a credible contribution to the broader scientific and health care communities.

Neonatal health outcomes remain a significant public health concern in LMICs, where approximately 98% of global neonatal deaths occur due to preventable causes such as preterm birth, intrapartum complications, and infections [48]. SSA bears a disproportionate burden, with NMRs often exceeding 25 per 1000 live births [49]. Socioeconomic disparities are key contributors as infants born to mothers with low literacy and lower income levels are more vulnerable to poor health outcomes. Access to prenatal care and essential neonatal services is often hindered by financial constraints, transport challenges, and low health literacy [50].

Moreover, fragile health care systems in LMICs lack adequate infrastructure, including electricity, medical equipment, and essential medical commodities and supplies, which compromises the quality of neonatal care. Inadequate infection control further exacerbates neonatal risk, with sepsis remaining a leading cause of mortality. Despite these challenges, promising interventions have emerged, such as community health worker–led programs and decentralized neonatal care units, which have shown substantial reductions in mortality. Achieving SDG 3.2 will require integrated efforts focused on health system strengthening; equitable access to care; and targeted community-based strategies, including digital health interventions such as ML tools.

The impact of contextually informed usability studies in LMICs extends beyond tool optimization to strengthening health system resilience. Effective usability testing has enabled the scalable deployment of artificial intelligence–driven diagnostic aids in Uganda, where simplified interfaces have reduced clinician cognitive load and improved diagnostic accuracy by 25% [51,52]. Moreover, integrating usability feedback into national digital health policies, as piloted in Rwanda’s health information exchange, fosters sustainable innovation [53]. Future efforts must prioritize capacity building for local usability researchers and leverage mixed methods data by combining quantitative metrics (such as task success rates) with qualitative insights to advocate for resource allocation. Ultimately, embedding usability science in LMIC health initiatives bridges this significant gap in technology adoption, advancing equity in global health delivery.

Several well-established frameworks guide the measurement of the feasibility of health care interventions such as digital health tools. Notably, the reach, effectiveness, adoption, implementation, and maintenance framework comprehensively addresses feasibility by evaluating the real-world effectiveness of interventions [54]. Additionally, the Medical Research Council framework for complex interventions provides guidance on the iterative assessment of interventions’ practicality, effectiveness, and sustainability in diverse contexts [55]. Other frameworks include implementation mapping, which systematically connects implementation determinants with specific strategies, providing a structured approach to assessing feasibility and developing tailored solutions [56,57]. The Consolidated Framework for Implementation Research also provides a detailed framework for analyzing feasibility, considering elements such as intervention characteristics, inner and outer settings, and stakeholder engagement, often in conjunction with pragmatic tools such as the Pragmatic Context Assessment Tool and the Clinical Sustainability Assessment Tool [58].

Successful adoption of ML tools within health care systems is influenced by stakeholder engagement and organizational readiness. The literature highlights the critical role of clinician and organizational perceptions in determining the feasibility, acceptability, and appropriateness of new technologies [59-61]. Complex systems thinking frameworks emphasize that feasibility assessments must account for dynamic interactions among technology, workflows, and human factors within health care environments [47,62,63]. Research indicates that early and sustained stakeholder engagement across clinical, administrative, and IT teams, as well as end user populations such as patients and caregivers, significantly enhances the feasibility of adopting ML tools. Additionally, assessing organizational readiness by evaluating knowledge, attitudes, and contextual factors (such as existing infrastructure and resource constraints) ensures that ML tools are tailored to local needs, thereby enhancing their practical integration and sustained use within health care systems [64-66].

The proposed evaluation of the ML model by Shukla et al [13] for neonatal risk prediction in Kenya is vital in informing its adoption and alignment with SDG target 3.2. It must address local challenges such as delayed triage, limited resources, inadequate staff training, and weak digital infrastructure. Mobile health platforms, already used in Kenyan CHW programs, present a scalable solution [67,68]. Integrating the tool with Kenya’s electronic medical record systems could enable real-time alerts for high-risk neonates. A community validation in Kakamega County found that CHWs achieved 85% accuracy with digital triage tools after brief training [69]. Evaluating usability in Kenya may improve triage, referrals, and resource use, thus reducing neonatal mortality. This initiative can guide scale-up, infrastructure investment, and policy alignment, supporting digital health transformation in LMICs.

The project will adhere to established quality standards throughout its life cycle, encompassing planning, data collection and analysis, and reporting of findings. Quality assurance in research refers to the techniques and strategies used to ensure proper care and control during the research process [70]. To uphold quality assurance in research, the project will implement measures to enhance research credibility, accuracy, and transparency [71,72]. First, staff competence will be ensured by providing relevant training to equip personnel with the necessary skills to perform their roles effectively. Project planning and documentation will be transparent, with all essential information made accessible and understandable to stakeholders, promoting open communication. Standard operating procedures will be established to guide project execution, including clear objectives, prioritized activities, appropriate methodologies, clear communication, and the protection of participant and data confidentiality.

However, this study has several limitations inherent to its exploratory preimplementation design. Neonatal health care staff will collect data by integrating a paper-based checklist into routine workflows, which may introduce observer and Hawthorne effects. The absence of a control group and blinding due to limited staffing and purposive recruitment of active neonatal care providers restricts causal inference and may introduce bias, although participants’ prior experience allows for meaningful before-and-after comparisons. To mitigate these limitations, methodological triangulation using both qualitative and quantitative approaches [73], including KIIs, surveys, and expert reviews, will be applied, alongside defined data quality standards. Data will be securely managed through restricted access, routine verification, encryption, and locked storage in compliance with Kenyan data protection laws. Ethical oversight, participant validation, expert review, and peer-reviewed publication will further strengthen the study’s credibility and trustworthiness.