Learning Design, Learner Engagement and Clinical Skills Acquisition of Students at Kenya Medical Training College Campuses in Rift Valley, Kenya

Introduction

Clinical skills acquisition (CSA) is a fundamental component of healthcare education and is essential for ensuring patient safety and quality care delivery. Clinical skills remain indispensable to accurate diagnosis and patient care, however, their decline in modern training necessitates renewed emphasis on bedside teaching and robust assessment, with technology positioned as a supportive rather than substitutive tool to sustain high-quality care in the 21st century (Elder, 2018). In addition, inadequate clinical competence among nurses has been linked to knowledge gaps in medication management as well as limited proficiency in the use of technology (Schroers et al., 2021). Such deficiencies in training and experience among healthcare providers compromise the overall quality of care. Globally, particularly in low- and middle-income countries (LMICs), concerns persist regarding the preparedness of newly qualified health professionals. Evidence indicates that many graduates lack essential clinical competencies due to inadequate training and supervision (Faustinella & Jacobs, 2018). These competency gaps have significant implications for healthcare quality. Poor-quality care remains a major and often underrecognized challenge in LMICs worldwide, undermining health outcomes despite improvements in access (Kruk et al., 2018). A systematic review estimated poor-quality care to account for more than 8 million deaths annually in these settings (Kruk et al., 2018). Similarly, the National Academies of Sciences, Engineering, and Medicine [NASEM] (2018) reported that poor-quality care contributes to approximately 10–15% of total deaths in LMICs. Reinforcing this concern, evidence from multiple LMICs shows persistent deficiencies in the clinical competencies of newly qualified clinicians (Lewis et al., 2019). In sub-Saharan Africa, these gaps in the competencies of newly trained healthcare workers. remain pronounced. A training needs assessment conducted across four African countries identified substantial clinical skills gaps among mid-level health workers (Couper et al., 2018). Similarly, widespread deficiencies in clinical performance have been documented across seven African countries, including Kenya (Lewis et al., 2019). Pedagogical approaches contribute significantly to these outcomes, as teacher-centred models in medical education have been associated with reduced engagement and poorer clinical skill development compared to more interactive approaches (Grundgeiger et al., 2017). More interactive and learner-centred approaches, including gamified strategies, have been shown to improve motivation, engagement, and learning outcomes among nursing students (Andretta et al., 2025). In Kenya, inadequate clinical competence among healthcare workers continues to pose a significant challenge. Systemic constraints within the health sector have been associated with gaps in healthcare workforce capacity and suboptimal quality of care delivery (Kimathi, 2017). Empirical studies further demonstrate that suboptimal clinical performance among nursing students is associated with limitations in clinical skills training and instructional approaches (Kweya et al., 2020). Enhanced clinical teaching approaches improve clinical judgment among nursing students (Wachira et al., 2020). Blended learning (BL), which integrates face-to-face synchronous instruction with asynchronous e-learning, has emerged as a promising student-centred approach to address these challenges. BL environments promote flexibility, interaction, and learner-centred engagement (Attard & Holmes, 2020). The perceived usefulness and effectiveness of BL are strongly influenced by the quality of instructional design and alignment of learning activities (Kerzic et al., 2019). An effective blended learning approach empowers learners with greater autonomy and a more active role in their education, enhancing engagement and learning outcomes in large-group teaching settings (Herbert et al., 2017). BL has consistently demonstrated improved outcomes, including greater student satisfaction, enhanced clinical judgment, better Objective Structured Clinical Examinations (OSCEs) performance and superior transfer of clinical skills compared to fully online or traditional approaches. Empirical evidence demonstrates that BL improves clinical competency across affective, cognitive, and psychomotor domains (Enoch et al., 2022). In clinical training contexts, blended approaches enhance the acquisition of clinical skills by aligning theoretical instruction with practical application (He et al., 2024). Furthermore, BL has been associated with higher levels of student satisfaction and engagement compared to traditional teaching approaches (Leidl et al., 2020). Meta-analytic evidence also indicates that it is more effective than traditional methods in medical education (Vallée et al., 2020). The effectiveness of BL is partly explained by its ability to integrate multiple instructional modalities. Gong et al. (2021) showed that integrating online modules, virtual simulations and hands-on practice create synergistic learning experiences that promotes active learning and improves clinical performance. Similarly, integrating digital and practical learning components enhances knowledge transfer into clinical contexts and skill development (Sáiz-Manzanares et al., 2020). BL environments also support personalized and flexible learning pathways, enabling learners to progress at their own pace (Goradia, 2019). In addition, online components such as adaptive technologies and discussion forums facilitate reflective learning and enhance the overall learning experience (Shahkarami et al., 2025). Learner engagement (LE), a multidimensional construct reflecting students' complete investment in their learning, is a critical antecedent in CSA. LE conceptualized as students’ active involvement in learning processes through interaction, participation and meaningful use of instructional resources, is a key determinant of effective learning outcomes (Theelen & van Breukelen, 2022). Rameshkumar (2020) argued that LE shapes commitment, persistence, satisfaction and motivation which are key predictors of learning outcomes. Engaged learners are more likely to demonstrate self-regulated learning, active participation and deeper conceptual understanding (Regmi et al., 2024). In clinical education, increased engagement has been linked to improved preparedness for assessments such as OSCEs (Massey et al., 2017). Moreover, interactive and hybrid learning environments have been shown to enhance student participation and engagement (Battestilli et al., 2023). The use of online discussion platforms further supports engagement and deeper learning (Gasmi, 2022). Importantly, LE has been identified as a key mechanism through which BL influences learning outcomes (De Bruijn-Smolders & Prinsen, 2024). Consequently, LE may mediate the relationship between learning design (LD) in BL and CSA. Within Kenya, the Kenya Medical Training College (KMTC) plays a central role in training middle-level healthcare workers, producing the majority (over 85%) of the country’s healthcare workforce. The Rift Valley region hosts the highest concentration of KMTC campuses, making it a strategically important setting for examining educational practices and outcomes. Despite the widespread implementation of BL across KMTC campuses, there is limited empirical evidence evaluating its effectiveness in improving CSA within this context. Although BL and LE have been widely studied, no known study in Kenya has empirically examined LE as a mediator in the relationship between LD and CSA. This represents a critical gap in health professions education research in sub-Saharan Africa. Addressing this gap is essential for informing evidence-based instructional strategies aimed at improving the quality of healthcare training and, ultimately, patient care outcomes. The general objective of this study was to assess the relationship between LD and CSA and determine the mediating role of LE on the relationship. Specifically, the study focused on the following research objectives: (I) to assess the effect of LD on CSA of students at Kenya Medical Training College Campuses in Rift Valley, Kenya.; and (II) determine the mediating role of LE on the relationship between LD and CSA.

Methods

Study Design

Cross-sectional and retrospective cohort designs were used. In a cross-sectional study, exposures and outcomes are measured simultaneously (Setia, 2016), enabling the collection of large volumes of data within a relatively short period and facilitating the analysis of associations between variables (Wang & Cheng, 2020). The retrospective cohort design, in which the performance of second-year students (exposed to blended learning) was compared to their first-year performance (exposed to traditional methods) on the same Patient Assessment, allowed for objective comparison of outcomes across different instructional approaches, as retrospective cohorts enable evaluation of outcomes using pre-existing data (Kim, 2023; Hendrix, P., & Griessenauer, 2019). Data for the study were collected between January and May 2025.

Study Setting

The study was conducted in nine KMTC campuses located in the Rift Valley region of Kenya. The region was chosen because its highest number of campuses relative to other regions of the country, could provide a representative sample. KMTC campuses were used in the study because they train roughly 85% of all the country's middle level health workers, offer courses that require skill competencies, and have already implemented BL (KMTC, 2019). This enabled the analysis of the effects of BL on the acquisition of clinical skills among nursing and clinical medicine students. These courses were chosen because they train critical workers that are the first to interact with patients in hospitals.

Study Population

The target population was all the 13,174 students and 62 heads of departments (HODs) in 16 KMTC campuses in the Rift Valley region (KMTC Records, 2024). However, only nine of these campuses offered clinical medicine and nursing, which were the courses of interest (KMTC Records, 2024). The accessible population was, therefore, all the 5,519 students and 9 HODs in the nine KMTC campuses. Inclusion criteria: (1) second-year students; (2) nursing and clinical medicine students; (3) HODs from the two departments; (4) willingness to participate. Exclusion criteria: (1) non-nursing or clinical medicine students; (2) unwillingness to participate.

Sampling Procedure

The study used a mix of purposive and stratified random sampling to select participants. Purposive sampling was used to select the nine HODs, enabling the study to obtain requisite information from people in charge of implementing BL. As different campuses had varied number of nursing and clinical medicine students, they were selected using stratified random sampling to minimize selection bias (Makwana et al., 2023). First, the requisite number of students from each campus was determined, followed by the department type, weighted by each stratum’s target population to ensure proportionate representation. A sampling frame of the students was obtained prior to data collection, which was inflated by 20 more students in case the identified ones were absent. During sampling, students were randomly selected from the sampling frame, using Excel Random Number Generator function. Random sampling ensured that every student had an equal chance of being sampled (Creswell & Creswell, 2018).

Sample Size Determination

Sample size was calculated using Slovin’s formula (Ellen, 2020), used to determine an appropriate sample size when the population is known and finite n = N / (1 + N · e²) n = the sample size e = the desired level of precision/sampling error, which was +5 %. N = 5,519 Substituting: n=5,519/ (1+5,519 x 0.052) 5,519/14.7975=373 Thus, the study collected data from 373 students. For HODs, all 9 (census) were sampled. Each HOD was in charge of both nursing and clinical medicine.

Data Collection Tools

Data were collected using structured questionnaires, key informant interview (KII) schedules and observation checklist. Questionnaires and KIIs, used to collect data from students and HODs, respectively, were developed following literature review (He et al., 2024; Regmi et al., 2024; Gasmi, 2022). The questionnaire consisted of three sections: Sample characteristics; Learning design, comprising of: In-person learning (four items) and E-learning (five items); Learner engagement, consisting of three subcomponents, Level of student commitment (five items), Study habits (five items), and Ability to Adopt (six items); and Quality Clinical Skills, consisting of four domains, Decision Making Skills (five items), Knowledge Construction (five items), Knowledge Production (seven items), and Critical Thinking Skills (four items). The scale of items ranged from a minimum of unity (strongly disagree) and a maximum of 5 (strongly agree). Cronbach alpha values for In-person learning, E-learning, LE, and CS were 0.86, 0.96, 0.92, and 0.95, respectively, showing the reliability of the constructs. Principal components factor analysis using varimax rotation, extracted single components for each hypothesised construct, suggesting construct validity (Barnes & Forde, 2021). The observation checklist was used to record each student’s Objective Structured Clinical Examinations scores in Years One (OSCE1) and Two (OSCE2).

Data Collection Procedure

Data were collected by the lead author as part of a larger PhD study on the effects of BL on CSA. A week before actual data collection, the researcher visited the campuses to be sampled, with an introductory letter from the University and KMTC and research permits from NACOSTI and KEMU ISERC. The researcher explained to the principal and HOD of Nursing and Clinical Medicine the purpose of the study and sought permission to conduct it. The full list of students (sampling frame) was obtained and the specific students (including the inflated ones) to be sampled from each campus and department (strata) were determined using Excel’s RANDBETWEEN FUNCTION. The questionnaires in Google Form were distributed to each student via their official college email, to prevent multiple entries from a student (primary safeguard). Returned questionnaires were only accepted if they originated from a unique IP address (secondary safeguard). The questionnaires’ first page included a clear explanation of the study, participant’s rights, and researcher’s contact details and students were required to sign the consent before proceeding. Each student was given a week to fill and return the questionnaire or was replaced by the extra students. On specified days, determined during initial contact, the researcher conducted face-to-face interviews with HODs. This involved obtaining their consent, conducting the interview, and thanking them at the end. The researcher took notes during the interview and auto recorded the conversation for back up and clarification. The HODs also provided OSCE data.

Data Quality Control

A pilot study was done in three campuses from the neighbouring Central Region: Nyahururu, Nyandarua and Nyeri, by administering 30 questionnaires to second year nursing and clinical medicine students. Permission was obtained from the university’s ISERC, NACOSTI and KMTC research Committee. Items of constructs with Cronbach Alpha values <0.7 were revised while those above with Alpha values ≥ 0.7 were adjudged to be reliable (Creswell & Creswell, 2018). During the main study, incoming questionnaires were reviewed daily to ensure fidelity in filling. Questionnaires with missing values were discarded. Similar questionnaires arising from the same student email or IP address were deleted leaving only one for use in the study.

Data Management and Analysis

Qualitative data was analysed thematically. Quantitative data were coded by assigning numeric codes to answers, allowing statistical analyses. The data were transcribed into Microsoft Excel Worksheet-version 19 and exported into statistical programs. Statistical analyses were conducted using STATA-version 14 and SPSS (Statistical Package for Social Sciences)-version 26, with significant differences in the study recorded at p<0.05. LD was measured using in-person learning and e-learning. The strength of the relationships between LD, LE and CSA was assessed with Pearson’s Correlation Coefficient. Correlations coefficients of < 0.35, 0.36 – 0.67 and 0.68 - 1.0, were classified as weak, moderate, or strong, respectively (Schober et al., 2018). To assess the effect of LD on CSA (Objective 1), multiple linear regression (MLR), using ordinary least squares, was used to determine both the main and interaction effects, estimated by the following equation: Y= β0+ β1X1 + β2X2 + β3X1.X2 + ε Where, Y = CSA; β0 = constant; β1, β2, and β3 = coefficients to be estimated; X1 = In-person learning; X2 = e-learning; X1.X2 = interaction between in-person and e-learning; ε = error term. Interaction analysis was conducted according to guidelines outlined by Igartua and Hayes (2021). A repeated-measures ANOVA was conducted to assess the effects of LD in a BL environment on CSA by comparing OSCE1 and OSCE2. Following Igartua and Hayes (2021) guidelines, the mediating effect of LE was modelled as the indirect effect of LD on CSA (Objective 2) and conducted using the PROCESS macro (Model 5) and structural equation modeling (SEM). Confidence intervals (CI) for the mediating effect were constructed using a bias-corrected bootstrapping method.

Ethical Considerations

Permission for study was obtained from ISERC (Approval No: KeMU/ISERC/HSM/17/2024; 08/10/2024), NACOSTI (Permit No: NACOSTI/P/24/41390; 04/11/2024) and KMTC (Ref: KMTC/ADM/74/VOL.VI (537); 03/12/2024). Permission was then sought from all participants before data collection. The nature, objective, and methods of the study were explained completely to potential respondents first. Then, consent was sought from them. No material inducement was given to potential participants. Only when informed consent was obtained, did the study proceed to collect data from the respondents. The researcher ensured that the respondents had complete freedom to ask any question by way of clarification. The study observed strict privacy and confidentiality by ensuring that respondents did not write their identities on the questionnaire. The recorded audio from the interviews was only transcribed by the researcher.

Discussion

Findings showed significant, positive, and moderate or strong correlations between learning design (consisting of in-person learning and e-learning) and LE with CSA. Learning design significantly and positively predicted the acquisition of clinical skills. These results were reinforced by the ANOVA model, which showed that BL significantly improved student performance in OSCE test by 6.579 marks in second year. Interaction analysis revealed that the effect of in-person learning on CSA did not depend on the level of e-learning and vice versa. Hence, the total effect of LD on CSA could be estimated as a sum of in-person and e-learning effects. Learner engagement was found to partially mediate the relationship between LD and CSA, with a unit increase in LD indirectly influencing CSA. Learning design, encompassing both in-person learning and e-learning, significantly predicted clinical skills acquisition (CSA), suggesting that improvements in instructional design are associated with enhanced clinical competency development. These findings align with prior research by He et al. (2024), Sáiz-Manzanares et al. (2020), and Gong et al. (2021) and with qualitative results from this study. Several mechanisms may underlie this relationship. First, in-person and e-learning modalities may function complementarily, with each medium reinforcing the other. Sáiz-Manzanares et al. (2020) posited that blended learning (BL) provides a continuum of learning whereby online instruction establishes foundational knowledge prior to applied clinical practice. Second, well-designed blended instruction fosters collaboration, learner-centredness, active engagement, real-world application, and formative feedback (Leidl et al., 2020; Herbert et al., 2017), facilitating knowledge and competency development consistent with Constructivist theory, which holds that learners actively construct knowledge through experiential interaction with their environment rather than through passive knowledge reception. Third, e-learning may serve as a cognitive scaffold, while in-person sessions ensure psychomotor skill mastery, which is a proposition supported by the present finding that pairing e-learning with face-to-face instruction is associated with CSA. Interaction analysis revealed no moderating effect between in-person learning and e-learning on CSA; that is, each independently and significantly predicted CSA, indicating that the removal of either predictor would not attenuate the effect of the other. This was reasonable, since, before the emergence of digital technologies, training of health workers was entirely face-to-face, producing some good workers. On the other hand, it is possible to train health workers through online platforms (Shahkarami et al., 2025). Nevertheless, the present study found that e-learning augments in-person learning, such that CSA reflects the combined effect of both modalities — an effect greater than that of either component in isolation. This conclusion is corroborated by several studies: He et al. (2024) demonstrated that blended modules integrating online theoretical instruction with structured practical experiences improve learners' clinical performance outcomes; Sáiz-Manzanares et al. (2020) found that combining online modules, virtual simulations, and hands-on practical sessions yielded superior learning outcomes; and Gong et al. (2021) reported that a well-designed blended learning approach facilitates the transfer of theoretical knowledge to practical clinical skills. ANOVA results strongly suggested that blended learning is superior to traditional methods for clinical skills acquisition among KMTC students, indicating better skill retention and application after blended elements were introduced (He et al., 2024; Gong et al., 2021). Learner engagement (LE) partially mediated the relationship between LD and CSA. This could be one of the first empirical reports of the mediating role of LE on the relationship between LD and CSA in Kenya. This aligns with Battestilli et al. (2023), Gasmi (2022), and Vallée et al. (2020), who found that heightened LE through online modules, simulation exercises, and practical sessions enhances clinical skills acquisition and application. Engaged learners likely complete tasks, participate in discussions, and immerse themselves in practical sessions, achieving superior learning outcomes (Regmi et al., 2024; Vallée et al., 2020). This mediating role appears especially salient in BL environments, where greater student autonomy demands higher levels of self-directed engagement. Learners with low LE may inadequately interact with online modules, undermining the deliberate practice essential for procedural mastery and patient safety (De Bruijn-Smolders & Prinsen, 2024). Findings from this study suggests that curriculum developers at medical education institutions, such as KMTC, should implement a learning design containing well-aligned in-person and e-learning components. In-person activities should promote, experiential, practical competence, tactile precision, procedural accuracy, and clinical confidence while e-learning activities should aim to help understand complex procedures and provide flexible, self-paced learning. Methodological strengths of this study included an adequate sample size and high response rates of 85% and 89% among students and HODs, respectively. Potential limitations of self-reported questionnaire data were mitigated through triangulation with objective OSCE performance scores from Years 1 and 2, as well as KII. The complementary dual-design approach further enhanced rigor, with the cross-sectional and retrospective cohort designs potentially offsetting their respective weaknesses, for example, measurement variability inherent to retrospective designs (Makwana et al., 2023). The application of mixed-methods triangulation and bootstrapped mediation analysis strengthened both internal and external validity. Key limitations included the cross-sectional design's constraints on causal inference, partially addressed through the incorporation of pre- and post-OSCE scores. Self-reported data introduced the risk of response bias including social desirability, while student age, department and gender potentially confounded results from the retrospective study. Accordingly, these variables were controlled for in the ANOVA model. Spatial restriction to a single region and the focus on nursing and clinical medicine students may limit the generalizability of findings. Lastly, only eight of nine HODs participated, which may slightly affect qualitative completeness. Future studies could consider using other designs, such as experimental study, to test further the effect of LD on CSA. Multiple regions and countries in East Africa could be used to test further the relationships established in this study.