Introduction
Clinical decision-making (CDM) forms the cornerstone of professional nursing practice,1 shaping patient safety, the effectiveness of care, and the overall quality of healthcare delivered.2 Nurses’ capacity to make timely, appropriate, and evidence-based clinical decisions is becoming ever more essential in today’s complex, patient-centred healthcare environment.3 CDM is not a single act but a continuous process of gathering and appraising information to guide reasoned judgments.4 Deficiencies in CDM among nurses can compromise the quality of patient care decisions, the completeness of patient data collection, the capacity to incorporate new clinical information, and the accuracy of patient data processing.5
Two key emotional factors that shape the quality of CDM among nursing students are self-confidence and anxiety.6,7 Oanh et al8 define self-confidence as the belief that a person can complete a task or achieve a desired goal. Bandura’s9 concept of self-efficacy, the belief in one’s capability to organise and execute the actions required to achieve specific outcomes, is widely applied in nursing literature to the related construct of self-confidence. Self-efficacy can be considered a situationally specific self-confidence.10 Self-confidence also has emotional regulation and communication aspects, however, which Gebreegziabher et al7 use to distinguish it from self-efficacy.
These emotional factors, in turn, influence CDM, the process through which nurses analyse patient information and select safe, effective, evidence-based interventions.11 Competence in this process is assessed with its own scales; accordingly, the present study measured self-confidence and anxiety only using a self-reported instrument. In clinical practice, according to Bibi et al6 self-confidence empowers students to trust their clinical judgment and reasoning, and to apply theoretical concepts to real-life scenarios. A lack of confidence can lead to hesitation and compromise patient care.12 Therefore, developing self-confidence is a key goal for nursing students, especially when working with real patients during clinical practice.
Anxiety can be described as an emotional state marked by a sense of impending danger, arising from the interplay of several feelings in situations where a negative outcome is possible, and further influenced by an individual’s confidence in their own capabilities.7 Gebreegziabher et al7 further argued that anxiety is more common among nursing students than among other healthcare professionals. Confidence and anxiety are strongly and inversely related: students with lower self-confidence are significantly more likely to experience high anxiety.6,7 Nursing students nonetheless frequently grapple with elevated stress and anxiety, which adversely affect their learning process.13
Anxiety in clinical decision-making (ACDM) is a common emotional state experienced by nursing students during clinical practice and can significantly impair the delivery of relationship-centred care, a core element of nursing practice.14 The clinical setting is inherently challenging for students given their limited experience: unlike students, expert nurses can discern patterns and trends in clinical situations, providing a general overview of patient issues that facilitates decision-making.15 Understanding how educational interventions influence these psychological dimensions of CDM is therefore important for preparing competent graduates.
In Malawi, these challenges are often compounded by high patient loads, limited resources and diagnostic equipment, and minimal supervision.16 In one high-dependency unit, for example, nurse-to-patient ratios ranged from 1:10 to 1:13 by day and 1:15 to 1:20 at night,17 and only 2.7% of sampled facilities had all basic equipment.18 Large student cohorts, staff shortages, and limited clinical sites also limit the clinical support students receive.19 In this context, novice nurses develop their decision-making skills over time, drawing on reasoning, available resources, and feedback from supportive environments.20 Self-confidence and low anxiety are therefore important, yet most strategies for building confidence and reducing anxiety, such as simulation debriefing, have been used elsewhere and are not widely accessible in Malawi. A study by Wheeles21 found a statistically significant difference in self-confidence with CDM in face-to-face clinicals compared to virtual clinical simulation. Evidence on teaching strategies that effectively improve confidence and reduce anxiety during actual clinical placements remains limited, particularly in sub-Saharan Africa.
Case-Based Learning (CBL) with structured post-clinical debriefing and reflection has shown promise in simulation settings.22–24 CBL provides practice with authentic patient problems,23 while debriefing and reflection offer feedback and structured time to process the experience,25,26 which may help to build confidence and ease anxiety. Kolb26 described learning as the process by which knowledge is created through the transformation of experience; reflective observation is one stage of his learning cycle. The 3D model of debriefing begins with defusing, in which learners discuss the emotional impact of the experience and recap events, and then moves to discovering and deepening understanding through analysis and application to future practice.25 Guided reflection was structured with the Clinical Reasoning Competency Framework (CRCF),27 whose four domains run from awareness of clinical cues to evaluation and reflection.
In terms of self-efficacy, Bandura9,28 identified verbal persuasion and physiological states as sources of efficacy beliefs. Facilitator feedback and the regulation of emotional arousal may therefore supplement the mastery experience that CBL provides with real patients. However, this combination has received little attention during actual clinical placements in Malawi.
This gap matters because simulation, the usual setting for debriefing, is limited in Malawi by several challenges, such as the absence of simulation in regulatory body syllabi, lack of formal training, demand for knowledgeable clinical instructors, inadequate resources, and resistance to its adoption.19
This study examined the changes in self-confidence and anxiety in CDM among undergraduate nursing students during clinical placements at three referral hospitals in Malawi following CBL with post-clinical debriefing and guided reflection. The study also examined how self-confidence and anxiety changed over time across three measurement points, and whether patterns of change differed across clinical placement sites or participant characteristics. It was hypothesised that, compared with traditional clinical teaching, CBL with post-clinical debriefing and reflection would be associated with a greater improvement in self-confidence and a greater reduction in anxiety in CDM from baseline (T1) to the final follow-up (T3), tested as a group × time interaction for each outcome.
Methods
Study Design
This study forms part of a larger doctoral research project evaluating the effectiveness of CBL with post-clinical debriefing and guided reflection among undergraduate nursing students. The present manuscript reports the anxiety and self-confidence in clinical decision-making outcomes from that study. The clinical reasoning outcomes from this study are reported separately in a companion manuscript currently under review in the Nurse Education in Practice journal.
A quasi-experimental, non-equivalent-groups, repeated-measures design was used to examine changes in self-confidence and anxiety in CDM among undergraduate nursing students in Malawi associated with a combined intervention of CBL, post-clinical debriefing and reflection.
Random allocation was not feasible because students had already been assigned to clinical groups through the institution’s routine placement scheduling process, which was managed independently by the clinical coordinator. Quasi-experimental designs are widely accepted in naturalistic educational settings where randomisation is not possible.29 Both groups completed the same two placements, paediatric and psychiatric nursing, over the twelve weeks, in a different order: the comparison group completed paediatrics first and the intervention group completed psychiatry first, with the intervention delivered during the paediatric placement. The comparison group had therefore finished its paediatric placement, under traditional teaching, before the intervention began, which reduced the opportunity for contact with intervention content.
Because group and placement order are confounded by design, their effects cannot be separated in the analysis.
Student orientation to the study was held from 15 to 21 December 2025, ahead of the academic break. The twelve-week clinical placement ran from 5 January to 29 March 2026, with the final follow-up assessment (T3) completed on 3 April 2026, once all students had returned to the main campus from their clinical sites. This study is reported in accordance with the Transparent Reporting of Evaluations with Nonrandomized Designs (TREND) statement, and the completed checklist is provided as Supplementary Table S1.30
Study Setting and Sample
The study took place within the undergraduate nursing and midwifery programme at Mzuzu University, across three affiliated referral hospitals in Malawi (Mzuzu Central Hospital (MCH), Kamuzu Central Hospital (KCH) and Zomba Central Hospital (ZCH)). These sites were selected for their comparable clinical learning environments, paediatric case mix, and availability of key paediatric units to achieve learning outcomes. Third-year students were selected because they had already completed earlier clinical placements under traditional teaching, providing a comparable baseline, and because self-confidence and anxiety in clinical decision-making are understood to develop progressively through clinical training experience.7,31
Eligibility Criteria
Students were eligible if registered as third-year nursing students at the study institution, assigned to paediatric and psychiatric placements during the data collection period, and willing to provide written informed consent. Students were excluded if they had a history of repeated unexcused absence from clinical placement (three or more absence reports recorded in departmental clinical reports during any preceding placement), because this would compromise exposure to the intervention and the completeness of outcome data; a known or self-reported mental health condition that could be adversely affected by the reflective and debriefing components, identified through institutional screening at programme entry, faculty observation or self-disclosure; or an existing nursing qualification, since prior qualification would introduce confounding clinical decision-making experience. No student met any of the exclusion criteria.
Eligible students were informed about the study in class, including its purpose and what participation would involve, at least two weeks before data collection began, to make an informed choice. The clinical coordinator facilitated access to the cohort.
Sample Size and Group Allocation
Total population sampling was used, as all eligible third-year students agreed to participate, so no student was lost through non-participation. This does not remove the risk of selection bias, however. Because students were not randomly allocated to clinical groups, the groups may differ in unmeasured characteristics, and allocation bias and baseline differences between groups cannot be ruled out; similar baseline scores on the measured variables do not exclude this.
Sample size was calculated a priori using G*Power 3.1 (F-tests: ANOVA, repeated measures, between factors) for two groups and three time points, with a moderate effect size (f = 0.25, chosen conservatively for a first test of this intervention in a low-resource setting), alpha = 0.05, power = 0.80, a correlation among repeated measures of 0.5 (the default) and a nonsphericity correction of 1. The calculation, which tests the between-group effect, yielded a minimum of 86 participants (actual power = 0.80). After two withdrawals, 84 students contributed data at T3.
Under the same assumptions (f = 0.25, ρ = 0.5), power was approximately 0.79 for the between-group effect and exceeded 0.99 for the group × time interaction, the primary hypothesis. The study was not designed to detect differences between the three hospital sites (about 14 students per group at each site), so non-significant site-level findings are interpreted with caution.
Group allocation followed the institution’s existing clinical placement schedule and was not influenced by the research team. The comparison group (n = 42) completed paediatric nursing first, then psychiatric nursing, both under traditional teaching; the intervention group (n = 44) completed psychiatric nursing first under traditional teaching, then paediatric nursing, during which the intervention was delivered. The groups had no joint classroom sessions or lectures during the study period.
Two intervention group students withdrew for reasons unrelated to the study (academic progression requirements). The primary analysis included all 86 enrolled participants and used all available data (255 of 258 possible observations, 98.8%), rather than complete cases only. Handling of the missing data is described in Data Analysis.
Participant flow from eligibility assessment to final analysis is presented in Figure 1. Figure 1Participant flow from eligibility assessment to final analysis. N = 86 students were enrolled.
All available data (255 of 258 observations, 98.8%) were analysed with generalised estimating equations (GEE). Allocation followed the institution’s placement schedule. Both withdrawals, in the intervention group, were due to academic progression requirements unrelated to the study.Flowchart sequence: Assessed for eligibility (n equals 86), Total population sampling; all eligible students agreed to participate, Enrolled and allocated (n equals 86), Allocation determined by existing institutional clinical placement schedule (not by research team), Comparison group (n equals 42), Paediatric placement (Weeks 1 dash 6): traditional teaching, Psychiatric placement (Weeks 7 dash 12): traditional teaching, No withdrawals, Completed T2 (n equals 42), No withdrawals, Completed T3 (n equals 42), GEE analysis: n equals 42, All available data included, Intervention group (n equals 44), Psychiatric placement (Weeks 1 dash 6): traditional teaching, Paediatric placement (Weeks 7 dash 12): CBL with post-clinical debriefing and reflection, 1 withdrawal before T2, Completed T2 (n equals 43), 1 further withdrawal before T3, Completed T3 (n equals 42), GEE analysis: n equals 44, n equals 44 at T1, n equals 43 at T2, n equals 42 at T3; All available data included.Flowchart of participant allocation into comparison and intervention groups, withdrawals and GEE analysis counts.Abbreviations: CBL, case-based learning.
Interventions
The intervention was a multicomponent package of three parts, CBL on real patient cases, structured group debriefing and guided reflective journaling, delivered together with facilitator support. It ran across the six weeks of the intervention group’s paediatric placement (weeks 7 to 12 of the twelve-week rotation). The intervention group attended a one-day briefing on the intervention on the Saturday after the psychiatric placement ended and T2 data had been collected (Friday), so that students knew what they would be doing.
The first week of the paediatric placement was then used for orientation and practice, and the remaining five weeks comprised full delivery by trained clinical nurse facilitators.
This intervention was grounded in Kolb’s Experiential Learning Theory, which provided structure for a weekly cycle of clinical experience, reflection and debriefing: real cases supplied concrete experience; debriefing and journaling supported reflective observation; Seven Jump analysis and CRCF-structured reflection supported abstract conceptualisation; and applying learning to the next case provided active experimentation. Bandura’s9 theory of self-efficacy informed the expectation that structured feedback would strengthen confidence and reduce anxiety: CBL with real patients offers mastery experience, debriefing offers vicarious learning and verbal persuasion, and reflection may help students regulate their emotional state. Because the components were delivered together, this study cannot determine the contribution of individual components.
The intervention was delivered by six clinical nurses (two per site), all holding at least a bachelor’s degree with five to over twenty years of paediatric experience (two with a master’s in child health). They delivered the intervention following two days of structured training in CBL, the 3D debriefing model, guided reflection, and CRCF integration, supported by a facilitation guide, question bank, and session report template. All scored above 80% on a post-training assessment and were confirmed competent by the first implementation week.
The facilitators were clinical nurses with paediatric expertise, as the intervention was delivered only in the paediatric placement. The psychiatric placement was delivered as traditional teaching for both groups. All facilitators used the same facilitation guide, question bank and session report template at all three sites and completed a session report after each weekly session.
Fidelity to the intervention protocol was monitored using a structured checklist at every session. The monitoring team comprised the lead researcher and nurse educators who did not teach the students and who had experience in simulation debriefing and CBL. At least two monitors observed sessions weekly at each site across all five weeks, rating each session across seven domains (general implementation, CBL using the Seven Jump process, CRCF integration, reflective practice, 3D debriefing, facilitator performance and participant engagement; 2 = fully, 1 = partially, 0 = not implemented).
Monitors gave feedback to the facilitators after each session to support delivery. The checklist ratings were used for this feedback and were not analysed quantitatively. No major barriers to implementation were identified.
Each of the five delivery weeks comprised three parts. (1) CBL used the Maastricht Seven Jump process,32 in groups of three, working on real patient cases followed for a minimum of three days or for the duration of the patient’s stay in the outpatient department before discharge or admission, whichever applied. Each week this component included about one hour of guided, facilitator-led CBL at the bedside, using a structured question bank to support students in managing their assigned patient.
(2) A structured group debriefing session used the 3D Model of Defusing, Discovering, and Deepening.25 Each week, one small group presented its case to the intervention students at the site for live facilitator-led debriefing and structured group reflection, lasting about three hours. Groups not presenting that week submitted a written reflection on their own case. Presentation responsibility rotated across groups, so that over the five weeks every group received a full debrief (five sessions in total) rather than a brief, same-week review.
(3) Guided individual reflective journaling was organised around the four domains of the CRCF: awareness of clinical cues, confirmation of clinical problems, determining and carrying out actions, and evaluation and reflection.27 Students also received the traditional clinical teaching required by the school. A week-by-week schedule of the intervention is provided as Supplementary Table S2, and in-person attendance of the monitoring team at the weekly debriefing sessions is shown in Supplementary Table S3.
The comparison group received traditional clinical teaching during both the paediatric and psychiatric placements. This comprised patient assignments under clinical nurse supervision, faculty-led teaching sessions held every two weeks, case study exercises in which students submitted a written case study of a patient using the nursing process, and written reflections on critical incidents. They did not have post-clinical debriefing or structured group reflection.
The traditional paediatric teaching was delivered by clinical nurses in the same units, including some of the clinical nurse facilitators who later delivered the intervention. At that time, they had not been oriented to the intervention and taught as they usually did. The group received no additional structured activity to match the extra time the intervention group spent with the facilitators.
To ensure fairness, the comparison group received two debriefing sessions, led by the intervention team, one week after the placement period ended, ensuring no intervention content was accessed during active data collection.
The intervention was reviewed by five expert validators (four nurse/midwife educators, one paediatric specialist, all master’s qualified). It used the Acceptability of Intervention Measure, Intervention Appropriateness Measure, and Feasibility of Intervention Measure.33 All three-domain means exceeded the scale midpoint of 3.00: acceptability M = 4.70, appropriateness M = 4.30, and feasibility M = 4.20, giving an overall mean of 4.40 out of 5.00. The scale-level Content Validity Index was 0.933, above the recommended threshold of 0.90.34 A planned pilot study was not possible because no students were on clinical placement at the intended site when ethical approval was granted, a limitation partially offset by the orientation week, validated instruments, and expert review.
To reduce researcher bias, the lead researcher had no prior teaching or supervisory relationship with any participant and did not teach paediatric nursing. The cohort was recruited from the intake year coinciding with the researcher’s commencement of doctoral study.
Outcome Measure and Data Collection
Self-confidence and anxiety in CDM were measured using the Nursing Anxiety and Self-Confidence with Clinical Decision-Making scale (NASC-CDM©).35 The instrument comprises two parallel 27-item subscales (six-point Likert scale, 1 = not at all to 6 = totally; range 27–162, higher = greater confidence or anxiety). Prorated totals were computed for participants completing at least 80% of items.
The original instrument demonstrated high internal consistency, with Cronbach’s alpha of 0.98 for confidence and 0.97 for anxiety, and exploratory factor analysis supported three factors for each subscale.35 The scale has since been adapted and tested among nursing students in Spain,10 Türkiye,36 and Korea,37 with good reliability in all three. The Korean version required a revised factor structure.37 Permission to use the NASC-CDM© was obtained directly from Dr. Krista White.
Internal consistency in the present sample remained excellent across all three time points (confidence α = 0.98/0.96/0.97; anxiety α = 0.99/0.96/0.96 at T1/T2/T3). To our knowledge, the NASC-CDM has not been formally validated in Malawi or elsewhere in sub-Saharan Africa, although it has been administered in English to undergraduate nursing students in Ethiopia.7 The English version of the instrument was administered without translation.
Data were collected at three time points: at baseline, prior to any clinical placement (T1, January 2026); midway, following the first six-week rotation (T2); and at the end of the study, once all students had returned to campus (T3, 3 April 2026). Hard-copy questionnaires were administered by an independent, blind data collector based at a different training institution with no prior familiarity with participants. Blinding of students and facilitators was not achievable given the intervention’s nature.
Students were assigned anonymous coded identifiers throughout, with approximately 25 minutes allocated for completion of both instruments at each time point.
Data Analysis
Data were analysed using IBM SPSS Statistics 26.0. The statistical analyst remained blind to group allocation throughout analysis. Baseline differences between groups were assessed using Mann–Whitney U and chi-square tests.
Shapiro–Wilk tests guided the choice between parametric and non-parametric tests. The primary GEE analyses are robust to non-normal distributions,38 and accounted for the correlation among repeated measures within participants, with a normal distribution, identity link, exchangeable working correlation (participant as subject, time as within-subject variable) and robust standard errors. The main predictor was the group × time interaction, with age, gender, and clinical site (categorical, with MCH as the reference) as covariates.
Placement order could not be entered separately because it is perfectly collinear with groups by design. GEE modelled the full trajectory across the three time points; ANCOVA gave a direct estimate of the between-group difference at T3 adjusted for T1 baseline scores, and Mann–Whitney U-tests provided non-parametric confirmation. Wilcoxon signed-rank tests, Kruskal–Wallis H-tests, Spearman’s rho correlations and placement-matched comparisons were supportive analyses.
Two primary hypotheses were tested (the group × time interaction for each outcome); both p values were < 0.001 and would remain significant after Bonferroni correction (alpha = 0.025). Supportive analyses were not adjusted. Significance was set at 0.05.
Three of 258 observations (1.2%) were missing, all from the two students who withdrew from the intervention group. Missing data were not imputed; the GEE used all available observations, which assumes the data were missing completely at random. The T3 ANCOVA and Mann–Whitney analyses, which used the 84 students with T3 data, served as a complete-case check.
Ethical Considerations
This study was performed in accordance with the principles of the Declaration of Helsinki. Ethical approval was granted by the Mzuzu University Research Ethics Committee (MZUNIREC), reference MZUNIREC/DOR/25/95, approved 20 October 2025, with permission from all three participating hospitals. The study was registered prospectively with the Pan African Clinical Trial Registry (PACTR202512666190013, 5 December 2025).
Written informed consent was obtained from all participants during the study orientation session, held from 15 to 21 December 2025, prior to the commencement of clinical placement on 5 January 2026; participation was voluntary, and students could withdraw at any time without penalty. Data were collected and stored using anonymous coded identifiers throughout. Minimal psychological risk was anticipated given the intervention’s reflective nature and students’ engagement with real patients; counselling support was arranged in advance, though no adverse effects occurred and no referrals were required during the study period.
A small token was provided to participants to cover internet data costs incurred during the study.
Results
Participant Characteristics and Baseline Compatibility
Of the 86 enrolled students, 84 contributed data at T3 following two withdrawals from the intervention group; all available data were used. No significant differences were observed between groups on any demographic, academic, or outcome variable at baseline (all p > 0.05). Baseline confidence and anxiety scores were closely matched, and all participants were generic undergraduate nursing students with no prior nursing qualification.
Full baseline characteristics are presented in Table 1, and outcome scores at each time point, with normality results, in Table 2.Table 1Baseline Demographic, Academic, and Outcome Characteristics by GroupVariableComparison (n = 42)M (SD)/n (%)Intervention (n = 44)M (SD)/n (%)χ2 / U / tpDemographic variablesAge23.10 (2.91)23.34 (3.17)864.500.603Gender0.440.507 Female24 (57.1)22 (50.0) Male18 (42.9)22 (50.0)Marital status1.830.401 Single38 (90.5)36 (81.8) In a relationship4 (9.5)7 (15.9) Married0 (0.0)1 (2.3)Source of income0.180.915 Parents29 (69.0)32 (72.7) Parents and government loan6 (14.3)6 (13.6) Other7 (16.7)6 (13.6)Monthly income (MK)0.950.621 ≤ 50,00024 (57.1)24 (54.5) 51,000–150,00017 (40.5)17 (38.6) Above 150,0001 (2.4)3 (6.8)Hospital site allocation0.100.953 MCH (Northern Region)13 (31.0)15 (34.1) KCH (Central Region)14 (33.3)14 (31.8) ZCH (South-Eastern Region)15 (35.7)15 (34.1)Academic variablesPrevious clinical grade HIV nursing65.74 (5.33)65.75 (6.31)918.000.958 Medical-Surgical nursing70.29 (7.98)71.23 (7.29)862.000.592Baseline outcome scores NASC-CONF99.31 (17.85)96.89 (19.27)0.6030.548 NASC-ANX99.73 (27.44)99.03 (27.95)0.1160.908Notes: Chi-square tests were used for categorical variables, Mann–Whitney U for age and prior grades, and independent-samples t-tests for NASC scores. NASC-CDM; range 27–162; higher = greater confidence or anxiety.Abbreviations: M, mean; SD, standard deviation; MK, Malawian Kwacha; MCH, Mzuzu Central Hospital; KCH, Kamuzu Central Hospital; ZCH, Zomba Central Hospital; NASC-CONF and NASC-ANX, confidence and anxiety subscales of the Nursing Anxiety and Self-Confidence with Clinical Decision-Making scale.Table 2Descriptive Statistics for Confidence and Anxiety by Group and Time PointOutcome/GroupT1 M (SD)T1 RangeT2 M (SD)T2 RangeT3 M (SD)T3 RangeNormality ViolationsaNASC-CONF (range 27–162; higher = greater confidence)Comparison (n = 42)99.31 (17.85)49–137107.79 (26.16)51–153115.51 (17.42)66–153NoneIntervention (n = 44/43/42)96.89 (19.27)67–15099.55 (29.28)51–161138.86 (14.22)112–156T3: W = 0.89, p = 0.001NASC-ANX (range 27–162; higher = greater anxiety)Comparison (n = 42)99.73 (27.44)36–14974.31 (28.68)34–14676.93 (22.17)31–123T2: W = 0.90, p = 0.002Intervention (n = 44/43/42)99.03 (27.95)39–15486.29 (34.52)33–14344.79 (11.56)30–81T2: W = 0.91, p = 0.003; T3: W = 0.87, p < 0.001Notes: Intervention group n = 44, 43 and 42 at T1, T2 and T3. Ranges are rounded to the nearest integer. a Shapiro–Wilk p ≤ 0.05; Mann–Whitney U replaced the t-test at that time point.
NASC-CDM; range 27–162; higher = greater confidence or anxiety.Abbreviations: M, mean; SD, standard deviation; T1, T2 and T3, baseline, mid-point and endpoint; NASC-CONF, confidence subscale; NASC-ANX, anxiety subscale (NASC-CDM).
Changes in Self-Confidence and Anxiety in CDM Associated with CBL, Post-Clinical Debriefing, and Reflection
GEE Analysis: Group Trajectories Over Time
The group × time interaction was significant for both confidence and anxiety (Table 3), indicating divergent trajectories across the study period, with no significant interaction at T2. By T3, the intervention group showed significantly greater confidence gains and anxiety reductions than the comparison group. Figure 2 shows the estimated marginal means and 95% confidence intervals at each time point.
Hospital site was a significant covariate for both outcomes and gender for anxiety, and the group × time interaction remained significant after adjusting for them. Table 3Generalised Estimating Equations: Model Effects and Parameter EstimatesSource/ ParameterConfidence (NASC-CONF)Anxiety (NASC-ANX)Wald χ2/βdf/SE95% CIpWald χ2 / βdf/SE95% CIpSection A: Model EffectsGroup2.9010.0894.2310.040Time152.332< 0.001151.132< 0.001Group × Timea38.122< 0.00150.372< 0.001Gender3.1710.0758.7810.003Hospital site12.3220.00221.272< 0.001Age0.0210.8830.1210.727Section B: Parameter Estimates Intervention × T3 vs T125.654.94[15.96, 35.34]< 0.001−31.646.43[−44.24, −19.04]< 0.001 Intervention × T2 vs T1−5.807.43[−20.36, 8.76]0.43512.408.18[−3.64, 28.43]0.130 Intervention × T3 vs T2b31.458.92[13.96, 48.93]< 0.001−44.0410.40[−64.43, −23.65]< 0.001Covariates Gender (Female vs Male)−5.072.85[−10.65, 0.51]0.07511.463.87[3.88, 19.04]0.003 Hospital site (KCH vs MCH)10.322.96[4.52, 16.12]< 0.001−19.084.22[−27.36, −10.81]< 0.001 Hospital site (ZCH vs MCH)7.033.20[0.75, 13.30]0.028−13.434.20[−21.65, −5.20]0.001 Age (years)−0.060.41[−0.86, 0.74]0.8830.220.62[−1.00, 1.43]0.727Notes: N = 86; 255 of 258 observations. Normal distribution, identity link, exchangeable correlation.
Reference: comparison group, male, MCH, T1. Bold, p < 0.05. a Primary effect of interest. b Contrast of T3 vs T1 minus T2 vs T1; SE = √(SE12 + SE2).2Abbreviations: β, unstandardised coefficient; SE, standard error; CI, Wald confidence interval; MCH, Mzuzu Central Hospital; KCH, Kamuzu Central Hospital; ZCH, Zomba Central Hospital; T1, T2 and T3, baseline, mid-point and endpoint; NASC-CONF, confidence subscale; NASC-ANX, anxiety subscale (NASC-CDM).Figure 2Estimated marginal means of (A) self-confidence and (B) anxiety in clinical decision-making at T1 (baseline), T2 (mid-point) and T3 (endpoint), by group. Points are model-based estimated marginal means from the generalised estimating equations (GEE) models, and error bars show 95% confidence intervals.
Confidence means were adjusted for age and hospital site; anxiety means were adjusted for age, gender and hospital site. The intervention group had 44, 43 and 42 students at T1, T2 and T3. CDM = clinical decision-making; NASC-CONF = confidence subscale; NASC-ANX = anxiety subscale.A) Line graph with error bars.
Horizontal axis label, T1 baseline, T2 mid-point, T3 endpoint. Vertical axis label, Self-confidence in clinical decision-making, NASC-CONF score, with ticks from 80 to 150. Legend entries: Comparison group, n equals 42, shown as a dashed line with open circles; Intervention group, n equals 44, 43, 42, shown as a solid line with filled squares.
Plotted values: Comparison group is about 99 at T1, about 108 at T2, about 115 at T3. Intervention group is about 97 at T1, about 100 at T2, about 139 at T3. Error bars are shown at each point.
B) Line graph with error bars. Horizontal axis label, T1 baseline, T2 mid-point, T3 endpoint. Vertical axis label, Anxiety in clinical decision-making, NASC-ANX score, with ticks from 30 to 120.
Legend entries match A. Plotted values: Comparison group is about 99 at T1, about 74 at T2, about 76 at T3. Intervention group is about 99 at T1, about 86 at T2, about 44 at T3.
Error bars are shown at each point.Two line graphs showing self-confidence and anxiety in clinical decision-making over three time points.
ANCOVA and Mann–Whitney U: Endpoint Comparisons
After controlling for T1 baseline, statistically significant between-group differences were observed for both outcomes at T3 with large effect sizes for confidence and anxiety, both substantially exceeding the large-effect benchmark of 0.8039 (Table 4). Levene’s test indicated heterogeneity of variance for anxiety (F(1, 82) = 13.91, p < 0.001), which was addressed by the equal group sizes at T3.40 Adjusted differences favoured the intervention group for confidence (23.81, 95% CI 16.85 to 30.76) and anxiety (32.38, 95% CI 24.78 to 39.99), and Mann–Whitney U-tests were concordant with the ANCOVA results.Table 4Endpoint Comparisons: ANCOVA and Mann–Whitney U Results (T3)OutcomeANCOVAMann–Whitney UFdfpPartial η2Adj. mean diff. (95% CI)UzprCohen’s dNASC-CONF46.351, 81< 0.0010.36+23.81 [16.85, 30.76]266.00−5.51< 0.0010.701.47NASC-ANX71.781, 81< 0.0010.47+32.38a [24.78, 39.99]191.00−6.18< 0.0010.781.82Notes: Bold, p < 0.05.
Adjusted mean difference: intervention minus comparison for NASC-CONF; aComparison minus intervention for NASC-ANX; positive values favour the intervention. Large-effect benchmarks: partial η2 ≥ 0.14, r ≥ 0.50, d ≥ 0.80 (Cohen).39Abbreviations: ANCOVA, analysis of covariance, adjusted for T1; CI, confidence interval; r, rank-biserial correlation; d, Cohen’s d; T3, endpoint; NASC-CONF, confidence subscale; NASC-ANX, anxiety subscale (NASC-CDM).
How Did Self-Confidence and Anxiety Change Over Time, and Did Patterns Differ by Site or Participant Characteristics?
Within-Group Trajectories
To examine the temporal pattern of change more closely, Wilcoxon signed-rank tests were conducted separately for each group across three periods: T1 to T2 (first placement, traditional teaching for both groups), T2 to T3 (second placement, comparison: traditional; intervention: CBL), and T1 to T3 (overall change) (Table 5, Section A). Neither group showed significant change in confidence during T1 to T2, though anxiety reduced significantly in both groups. During T2 to T3, the intervention group improved significantly on both outcomes, while the comparison group showed no significant change on either.
By T3, all 42 intervention group students showed higher confidence than at T1, and 40 of 42 showed lower anxiety, indicating that improvement in the intervention group occurred mainly during the period in which the intervention was delivered rather than accumulating gradually.Table 5Within-Group Change and Placement-Matched ComparisonsSection A: Within-Group Change (Wilcoxon Signed-Rank Tests)Period/OutcomeZprn ImprovedaZprn ImprovedaComparison (n = 42)Intervention (n = 42/43)T1 to T2 (first placement; traditional teaching both groups) NASC-CONF (NS both groups)−1.680.0930.2627/42−0.710.4800.1123/43 NASC-ANX−4.10< 0.0010.6331/42−2.090.0370.3227/43T2 to T3 (second placement; comparison: traditional; intervention: CBL) NASC-CONF−1.210.2280.1924/42−5.13< 0.0010.7937/42 NASC-ANX−0.660.5090.1018/42b−5.28< 0.0010.8136/42T1 to T3 (overall change) NASC-CONF−3.84< 0.0010.5932/42−5.65< 0.0010.8742/42 NASC-ANX−3.83< 0.0010.5931/42−5.61< 0.0010.8740/42Section B: Placement-Matched Comparisons (Mann–Whitney U)Comparison/OutcomeComparison group M (SD)Intervention group M (SD)Post-paediatrics: comparison T2 (traditional) vs intervention T3 (CBL) NASC-CONF107.79 (26.16)138.86 (14.22)U = 243.00, p < 0.001, r = 0.62 NASC-ANX74.31 (28.68)44.79 (11.56)U = 244.50, p < 0.001, r = 0.62Post-psychiatry control: both groups traditional teaching NASC-CONF115.51 (17.42)99.55 (29.28)U = 612.00, p = 0.011, r = 0.28 NASC-ANX76.93 (22.17)86.29 (34.52)U = 803.50, p = 0.382, r = 0.10Notes: Bold, p < 0.05. a Participants with improved scores (higher NASC-CONF, lower NASC-ANX). b 18 of 42 had lower anxiety at T3 than T2; 23 worsened; 1 tied. r benchmarks: 0.10 small, 0.30 medium, 0.50 large (Cohen).39Abbreviations: M, mean; SD, standard deviation; NS, not significant; r, rank-biserial correlation; T1, T2 and T3, baseline, mid-point and endpoint; NASC-CONF, confidence subscale; NASC-ANX, anxiety subscale (NASC-CDM).
Were the Improvements Associated with an Intervention Rather Than Placement Context?
To explore whether the observed improvements were consistent with the intervention rather than the placement context, two placement-matched comparisons were conducted using Mann–Whitney U-tests (Table 5, Section B). In the post-paediatric comparison, the comparison group at T2 (traditional teaching) was compared with the intervention group at T3 (CBL). The intervention group significantly outperformed the comparison group on both outcomes with large effect sizes.
This comparison holds placement type constant while varying the teaching approach, although timing, prior clinical exposure and placement sequence still differ between groups. In the post-psychiatry control analysis, where both groups received traditional teaching, no significant difference was observed for anxiety, suggesting no differential effect of clinical context on anxiety. A modest but significant difference favouring the comparison group was observed for confidence.
Taken together these comparisons are consistent with an intervention-associated difference but provide supportive evidence only.
Did Patterns of Change Differ by Site or Participant Characteristics?
Hospital site was a significant covariate for both outcomes and gender for anxiety in the GEE models (Table 3), reflecting differences in overall score levels between sites, pooled across all time points and placements; the group × time interaction remained significant after adjustment. Within-group Kruskal–Wallis tests, using each group’s own post-paediatric-placement measurement (T3 for the intervention group and T2 for the comparison group, because the groups completed paediatrics at different points), revealed no significant site differences in paediatric placement outcomes within the intervention group (confidence: H(2) = 0.00, p = 1.000; anxiety: H(2) = 0.03, p = 0.986) or the comparison group (all p > 0.30). This suggests that the outcomes were similar across MCH, KCH, and ZCH, although with only about 14 students per site within each group, these analyses had limited power.
Gender was a significant covariate for anxiety, with female students reporting higher anxiety overall than male students, but not for confidence; age was not a significant predictor of either outcome. No formal subgroup interaction analyses were conducted beyond the GEE covariate framework.
Relationship Between Self-Confidence and Anxiety at Endpoint
Spearman’s rho correlations were computed between confidence and anxiety total scores at T3 separately for each group, given the non-normal distributions observed in the intervention group at that time point (Table 6). Confidence and anxiety were significantly and inversely correlated in both groups. The inverse correlation was numerically stronger in the intervention group; however, the difference between the two correlation coefficients was not formally tested.Table 6Spearman’s Rho Correlations Between Confidence and Anxiety at T3 by GroupVariable PairComparison Group (n = 42)Intervention Group (n = 42)rsprspNASC-CONF — NASC-ANX−0.480.001−0.75< 0.001Notes: A negative correlation means higher confidence goes with lower anxiety.
Bold, p < 0.05.Abbreviations: rs, Spearman’s rho; T3, endpoint; NASC-CONF, confidence subscale; NASC-ANX, anxiety subscale (NASC-CDM).
Discussion
This study examined changes in self-confidence and anxiety in CDM following CBL with post-clinical debriefing and reflection among undergraduate nursing students in Malawi. Confidence increased more, and anxiety decreased more in the intervention group than in the comparison group, with the largest changes occurring during the period in which the intervention was delivered.
The magnitude of the confidence gain is consistent with Bandura’s9 self-efficacy framework, in which mastery experience is the most powerful source of efficacy belief. CBL gave students repeated opportunities to reason through real clinical decisions and receive feedback, while weekly debriefing may have supplied the vicarious learning and verbal persuasion that reinforce mastery experience into durable confidence. This effect size appears larger than that typically reported for CBL delivered without structured debriefing41–43 and the pooled effect reported across 22 controlled confidence-building interventions (SMD = 0.51).44 This suggests that debriefing and guided reflection may add to what CBL alone achieves, although this study cannot test that directly.
This is consistent with evidence that debriefing, rather than simulation itself, is the active ingredient in confidence development45 and that real clinical experience produces confidence gains that virtual simulation does not fully replicate.21
The anxiety findings may point to a related but distinct pathway involving the debriefing and reflection components, although this study cannot separate them from the other elements of the package. Reflective practice has independently been shown to reduce clinical anxiety by giving students structured time to identify feelings and process uncertainty,46 and a CBL study in a paediatric course using the same instrument reported smaller anxiety reductions.43 The sustained decline in the intervention group’s anxiety across the full placement, contrasted with partial rise in the comparison group during their subsequent psychiatric placement, may indicate that any benefit for anxiety accumulates with repeated debriefing rather than a single early exposure, although this cannot be tested here. This is consistent with the distinct emotional demands psychiatric placements are known to place on undergraduate students.47,48
The placement-matched comparisons provide supportive evidence that the pattern was not solely due to the clinical placement itself. Holding placement type constant, the intervention group significantly outperformed the comparison group on both outcomes. In the control comparison, where both groups received traditional teaching, no significant difference was observed for anxiety, suggesting no differential effect of clinical context alone.
A modest but significant difference favouring the comparison group was observed for confidence in this control comparison; this most plausibly reflects the comparison group’s earlier completion of the paediatric placement by this stage, since the intervention group encountered psychiatry as its first placement without the benefit of prior clinical exposure, rather than a differential teaching effect.
The effects observed were large and should be interpreted cautiously. Similar effects have been reported in other recent quasi-experimental research from Malawi, where standardised patient simulation produced a large improvement in practical clinical skill compared with lecture-based teaching (partial η2 = 0.465).49 This suggests that structured, hands-on teaching may outperform passive instruction in this setting. The effect could be smaller if the intervention were compared with another active, structured approach, or tested among students with greater access to structured clinical teaching.
Some of the differences may also reflect factors the design could not control: expectancy effects, motivation, contamination between groups, facilitator effects at each site, and placement sequence. Maturation cannot be ruled out, because the intervention was delivered in the second rotation, after students had completed a first placement and when more time had passed since baseline. Residual confounding is inherent to a non-randomised design.
The direction of change was similar across the three hospital sites, even though hospital site was a significant covariate reflecting differences in overall score levels between sites. This suggests that the benefit was not confined to one site, although the study was not designed to test site differences and a formal site-by-group interaction would be needed to confirm this. Gender was also a significant covariate for anxiety, with female students reporting higher anxiety than male students across the study period, consistent with the broader nursing education literature documenting higher anxiety among female nursing students.7 Age was not a significant predictor of either outcome.
The group difference in anxiety persisted after adjustment for gender in the GEE model.
At endpoint, the inverse correlation was numerically stronger in the intervention group, although the difference was not formally tested. This is compatible with Bandura’s9 framework, in which strengthened self-efficacy is understood to reduce anticipatory anxiety, and is consistent with evidence from Ethiopian undergraduate nursing students, where low confidence was strongly associated with elevated anxiety.7 These are correlations between score levels at T3 and do not show whether changes in confidence were associated with changes in anxiety over time. Longitudinal mediation analysis would be needed to test whether one change drives the other.
Implications for Practice, Research and Policy
These findings carry relevance for nursing education in resource-limited settings such as Malawi, where large cohorts, limited clinical placement capacity, and constrained numbers of qualified clinical educators restrict the individualised clinical exposure available to each student. CBL with structured debriefing and reflection does not require additional placement capacity, can be delivered to groups, and was associated with similar improvements across three hospital sites. This suggests that it may be a feasible addition to undergraduate nursing curricula in similar settings, although replication is needed.50 If replicated, the findings may inform consideration of post-clinical debriefing as a component of clinical placement standards within nursing education regulatory frameworks, backed by facilitator training and fidelity guidelines, and implementable within existing clinical infrastructure at comparatively modest cost relative to simulation-based alternatives.
Future research should consider replication across multiple institutions and sub-Saharan African contexts to establish generalisability, alongside randomised designs where ethically and logistically feasible to strengthen causal evidence. Longitudinal follow-up would help establish whether these improvements in confidence and anxiety are sustained over time. Further research into the mechanisms underlying debriefing’s effect, including metacognitive awareness, psychological safety, and facilitator quality, would advance theoretical understanding of how structured reflection produces these outcomes.
Limitations
Several limitations should be considered. No formal pilot study was conducted prior to implementation. The non-randomised, quasi-experimental design cannot rule out allocation bias or unmeasured confounding despite similar baseline scores.
Group and placement order were inseparable (the intervention group completed psychiatry first), so order effects, differences in timing and prior clinical exposure, maturation, expectancy and Hawthorne effects, and contamination between groups cannot be excluded. Students could not be blinded, and follow-up ended immediately after placement, so durability was not tested. Confidence and anxiety were assessed using self-report instruments, carrying an inherent risk of response bias.
The NASC-CDM has not, to our knowledge, been validated in Malawi, and no objective measure of CDM performance was included. The small participant token, provided to cover internet data costs, while modest, cannot fully rule out a degree of participation or compliance bias. Site-level subgroup analyses of paediatric placement outcomes (Kruskal–Wallis, about 14 students per group at each site) found no significant differences, in contrast to the significant hospital site effects detected in the primary GEE models; this discrepancy likely reflects the reduced power of the smaller subgroup analysis rather than a true absence of site-level influence.
The intervention was delivered as an integrated package of CBL, post-clinical debriefing and guided reflection, and the study was designed to evaluate that package. Facilitator expertise was not analysed as a factor in the outcomes, and fidelity was monitored with structured checklists, but the ratings were not analysed. Finally, despite having three hospital sites, the study was conducted within a single university, which may limit generalisability to institutions with different curricular structures or resource levels.
Conclusion
In this quasi-experimental study, a multicomponent intervention comprising CBL, post-clinical debriefing, and guided reflection was associated with greater improvements in self-confidence and greater reductions in anxiety in CDM than traditional clinical teaching among undergraduate nursing students at one Malawian university. Placement-matched comparisons provided supportive evidence, although differences in placement sequence, timing, and prior clinical exposure could not be excluded as alternative explanations. At T3, higher confidence was associated with lower anxiety in both groups; these endpoint associations do not establish a relationship between changes in the two outcomes or a causal mechanism.
These findings support further evaluation of the intervention package during clinical placements through larger, multicentre, preferably randomised studies with longer follow-up.