This study was prospectively registered as a parallel-group RCT in the Chinese Clinical Trial Registry (ChiCTR2200060627; registered on June 6, 2022). The initial design aimed to evaluate intervention efficacy, with 24-hour urinary sodium excretion as the primary outcome and secondary outcomes including sodium-related knowledge, blood pressure, and quality of life (QOL). A 3-month follow-up duration was selected based on previous digital health implementation studies in CKD and hypertension, which have demonstrated that 8 to 12 weeks is sufficient to detect changes in intermediate physiological and behavioral outcomes in real-world settings [7,17]. This time frame balances the need to observe meaningful behavior change while minimizing attrition in longitudinal implementation research.

During the 3-month interim review, only 24.2% (24/99) of participants completed valid 24-hour urine collections, resulting in 75.7% (75/99) missing data for the primary outcome, far below the prespecified acceptable threshold of 60%. Because increasing the sample size could not improve data completeness, the study design was formally revised from an RCT to a single-center, exploratory mixed methods implementation study [24-26].

This design adaptation was consistent with reporting guidelines for pragmatic implementation research. The original sample of 99 participants remained adequate for mixed methods inquiry, which typically requires at least 80 quantitative participants and 15 to 30 qualitative participants [27]. Accordingly, the analytical focus shifted from confirmatory efficacy testing to exploratory implementation outcomes. The core objective was revised to evaluate implementation processes, participant engagement experiences, and barriers to a nurse-led digital sodium restriction intervention in patients with CKD.

During the 3-month interim progress review conducted in November 2022, the completion rate for the prespecified primary outcome measurement was merely 24.2% (24/99), far below the expected level. Concurrently, notable participant attrition was observed in both groups. Given the poor feasibility of continuing the original primary outcome evaluation, our research team comprehensively assessed the actual research progress and field conditions. After obtaining formal approval from the ethics committee, we decided to abandon the planned primary outcome efficacy analysis, redefine this project as an exploratory mixed methods implementation study, and shift our core research focus to exploring intervention implementation processes and real-life patient experiences.

The research design and study protocol were approved by the Ethics Committee of the Second Affiliated Hospital of Army Medical University (approval number 2022-Research-040-01). The revised study design was subsequently reviewed and approved by the same ethics committee (approval number 2022-Research-515-01). All participants provided written informed consent prior to enrollment. All study procedures were conducted in accordance with the Declaration of Helsinki.

Eligible participants were recruited and screened by trained clinical nurses from inpatients admitted to the nephrology department between June 2022 and August 2022. Inclusion criteria were as follows: age 18 to 65 years; diagnosis of CKD stages 1‐5 [28], including those with hypertension; an estimated glomerular filtration rate stable over the past 3 months; 24-hour urinary sodium excretion >130 mmol/day at baseline; Barthel Index score ≥60; Mini-Mental State Examination (MMSE) score ≥24; proficiency in using smartphones and mobile applications; and willingness to participate. Exclusion criteria included the following: recent acute kidney injury; receipt of renal replacement therapy; major cardiovascular event within 6 months; uncontrolled hypertension (systolic blood pressure >160 mm Hg or diastolic blood pressure >100 mm Hg); prior participation in similar digital health programs; or any condition likely to interfere with study compliance.

Eligible participants were randomly assigned (1:1) to either the intervention or control group using computer-generated random sequences in SPSS (version 24.0; IBM Corp). Allocation was concealed via sequentially numbered, opaque, and sealed envelopes prepared by an independent statistician. Due to the nature of the intervention, participants and the intervention delivery nurse could not be blinded. However, to minimize ascertainment bias, outcome assessors, laboratory technicians, and data analysts remained blinded to group allocation throughout the study. The 46:53 group size discrepancy resulted from random fluctuation of the computer-generated sequence with no manual modification. Such a degree of imbalance falls within the expected range of chance variation in simple randomization and does not compromise the validity of random allocation.

The sample size was calculated using PASS (version 15.0; IBM Corp) software based on the anticipated between-group difference in 24-hour urinary sodium excretion at the 3-month follow-up. Data from a previous randomized trial suggested a mean excretion of approximately 173.5 mmol/day in control groups and 147.5 mmol/day in intervention groups, indicating a potential difference of 26 mmol/day [31]. A common SD of 34.0 mmol/day was assumed based on the same study. Using a 2-sample, 2-tailed t test with α=.05% and 90% power, 37 participants per group were required. Accounting for an estimated 20% dropout rate, the final target sample size was set at 47 per group (total n=94).

Quantitative data were collected at baseline (T0) and at the 3-month follow-up (T1). Outcomes of interest included 24-hour urinary sodium excretion [32] and blood pressure [33] as potential indicators of physiological response. Blood pressure was defined as the average of 3 consecutive office blood pressure measurements. Additional outcomes comprised sodium control knowledge, QOL, and laboratory parameters. Sodium control knowledge was assessed using a 10-item questionnaire developed by the research team. The questionnaire was based on CKD sodium restriction guidelines, refined through 3 rounds of expert review, with a content validity index of 0.92, and validated via pilot testing with 20 participants and a Cronbach α of 0.78, and the questionnaire is provided in Multimedia Appendix 1. QOL was measured using KDQOL-SF (Kidney Disease Quality of Life-Short Form) 1.3 [34], and laboratory parameters included serum creatinine, urea, uric acid, electrolytes, hemoglobin, and albumin. Adherence to 24-hour urine collection was anticipated to present challenges; therefore, this measure was analyzed exploratorily and interpreted cautiously, with due consideration of completion rates.

Qualitative data were collected through semistructured in-depth interviews with a purposive subsample of 23 participants. Participants who discontinued the intervention were not included due to practical constraints related to time and research resources. This subsample comprised only intervention completers; accordingly, all qualitative findings were limited to participants who completed the full intervention program.

All interviews were conducted in Mandarin Chinese by 2 nurse researchers with master’s degrees who had received systematic training in qualitative research (Checklist 2). These researchers were not involved in intervention delivery to minimize potential bias. Face-to-face interviews were conducted in the outpatient CKD management clinic, with each session lasting 20 to 40 minutes. All interviews were audio-recorded, transcribed verbatim, and anonymized. The transcripts were then translated into English using dual translation and back-translation to ensure linguistic and conceptual fidelity. Regular debriefing sessions were held throughout data collection to enhance reflexivity and analytic rigor. The interview guide was developed under the core framework of Orem’s Self-Care Deficit Nursing Theory, with the integration of key concepts from Empowerment Theory. The guide was designed to explore participants’ experiences across multiple dimensions, including task execution, emotional responses, barriers encountered, and evolving needs. Thematic analysis was used to analyze the interview transcripts. To ensure coding reliability, 2 researchers independently coded 20% of the data, yielding strong intercoder agreement with a Cohen κ value of 0.72. Discrepancies were resolved through consensus discussion. The finalized codebook, which included code definitions, thematic hierarchies, representative quotes, and decision rules, was then applied to the full dataset by the primary researcher. The codebook is provided in Multimedia Appendix 2. Sensitivity analyses were conducted to assess the potential influence of selection bias on qualitative interpretations. We hypothesized that participants who discontinued the intervention may hold distinct perspectives related to higher intervention burden, digital fatigue, and competing time demands. Such systematic differences may lead to overestimation of participant engagement and underestimation of real-world implementation barriers if noncompleters are excluded from analysis.

Of 303 screened patients, 99 were randomized: 46 to the enhanced digital intervention group and 53 to the minimal digital control group. Baseline characteristics were similar between groups (Table 1), confirming balanced randomization. At the 3-month follow-up, 52 (98.1%) in the control group and 45 (97.8%) in the intervention group completed blood pressure and blood component analysis, with the exception of 1 death in the intervention group and 1 withdrawal in the control group for personal reasons. However, only 13 (24.5%) in the control group and 11 (23.9%) in the intervention group completed the 24-hour urine collection (Figure 2), resulting in an overall completion rate of 24.2% (24/99) for this outcome measure.

Baseline demographic and clinical characteristics were balanced between groups (Table 1). The median age was 48 (IQR 37‐54) years in the control group and 44 (IQR 31-52) years in the intervention group (P=.21). There were no significant differences in sex distribution, CKD stage, comorbidities, or laboratory parameters, confirming successful randomization.

As shown in Table 2, the completion rate for 24-hour urinary sodium excretion was critically low (24/99, 24.2%), precluding reliable estimation of between-group differences or definitive conclusions about behavioral change. Among the small subset of completers (n=24), median sodium excretion appeared lower in the intervention group (97.3, IQR 73.7‐124.0) compared to controls (154.4 mmol/day, IQR 123.2‐190.9). However, inferential testing was not conducted due to insufficient power and selection bias concerns.

However, a significant improvement was observed in the secondary outcome of sodium control knowledge: intervention group participants scored 8 (IQR 7-8) versus 6 (IQR 6-8) in controls (z=2.84, P=.005). No significant differences were found in blood pressure, QOL (36-Item Short Form Health Survey [SF-36] and kidney disease targeted areas), or other physiological markers.

Integrated analyses were performed using a joint display matrix (Table 4) following GRAMMS guidelines for mixed methods reporting. Three patterns of integration were identified. First, convergent findings demonstrated significant improvements in sodium-related knowledge in the enhanced digital intervention group (z=2.84, P=.005). These quantitative improvements were supported by qualitative accounts of clear, structured learning and enhanced confidence in sodium management, providing consistent support for the improvement in knowledge gains observed in the quantitative analysis. Second, supplementary findings highlighted a critical measurement paradox: completion of 24-hour urinary sodium collection was only 24.2% (24/99). Qualitative data further identified substantial procedural burden and poor compatibility with daily routines as key barriers, indicating that the gold-standard assessment itself impeded adherence. Third, explanatory findings clarified the absence of significant between-group differences in blood pressure or QOL (P>.05). Identified barriers, including family resistance, digital fatigue, and limited long-term support, explained why improvements in knowledge did not translate into measurable changes in clinical or patient-reported outcomes. No divergent findings were observed. The integrated synthesis indicates that although the intervention improved sodium-related knowledge, translation into behavioral and clinical changes was limited by implementation barriers and measurement burden.

This study found that the nurse-led digital intervention improved participants’ sodium-related knowledge as a secondary exploratory outcome. However, the completion rate of the primary outcome (24-h urinary sodium collection) was only 24.2% (24/99), which precluded evaluation of the intervention’s true efficacy. These findings are consistent with Orem Self-Care Deficit Nursing Theory. Qualitative interviews identified a 3-phase patient engagement journey: initial enthusiasm, active participation, and subsequent disengagement driven by digital fatigue and family barriers. Overall, while digital interventions can enhance health knowledge, high-burden assessment tools and poor adaptation to daily life significantly hinder sustained behavioral changes.

The improvement in knowledge without obvious behavioral changes supports Orem theory, which holds that self-care agency requires adequate knowledge, skills, motivation, and environmental support [36-39]. This reflects a typical knowledge-action gap. Digital tools can efficiently disseminate standardized health information on a large scale, but knowledge alone cannot drive long-term behavioral changes [40-43]. Sodium restriction is affected by family dynamics, cooking habits, and taste preferences—factors far more complex than simple information delivery [44,45]. Therefore, future interventions should shift from one-way knowledge popularization to multidimensional empowerment, comprehensively addressing both structural and interpersonal barriers [46-48].

This study revealed a measurement paradox. Although 24-hour urine collection is the gold standard for sodium intake assessment [49], its complicated operation and heavy burden lead to low patient adherence and high measurement error [50,51]. This phenomenon aligns with Orem's view that self-care interventions must match patients’ actual capabilities, which limits its application in clinical practice and research. By contrast, the spot urine sodium-to-creatinine ratio is simple to operate and has been validated to predict disease progression in multiple studies [52,53]. AI-based real-time urine testing may balance accuracy and convenience; however, its reliability across different patient groups requires further verification before large-scale application [54-56]. We therefore suggest researchers prioritize outcome measures that offer both scientific rigor and real-world feasibility, rather than pursuing methodological purity at the expense of practicality, to improve the clinical translation of research findings.

Family resistance, lifestyle constraints, and digital fatigue are key barriers to long-term intervention effects [57-59]. Digital health success relies on continuous patient engagement rather than short-term adoption [60-62]. Based on Orem's self-care theory, we propose a stage-matched intervention framework that includes 3 core components: deficit identification, compensatory support, and self-care capacity improvement [63]. In the initial stage, medical staff identify patients’ knowledge gaps and practical obstacles to build trust. In the middle stage, personalized digital guidance is used to improve patients’ self-efficacy and bridge capability gaps. The later stage helps patients maintain healthy behaviors through community interaction and independent self-care management. This framework transforms static, one-size-fits-all intervention modes into adaptive, dynamic support that addresses the multidimensional self-care deficits in CKD sodium management. Future mixed methods studies should further explore the mechanisms of patient engagement to inform the development of truly patient-centered digital intervention solutions.

This study adopted a parallel mixed methods design. Combining quantitative data and qualitative narratives clarifies intervention effects and their underlying mechanisms—an integrated approach that remains underused in digital therapeutics research. Conducted in routine outpatient settings without additional resource input, the study demonstrates good ecological validity and practicality. The proposed patient empowerment pathway also provides a new perspective for exploring digital health engagement among patients with CKD.

Several limitations of this study should be acknowledged. First, the 24.2% (24/99) completion rate of 24-hour urine collection resulted in massive missing data, rendering ITT analysis and primary outcome testing invalid. Consequently, the true intervention effect could not be confirmed. Second, qualitative participants were limited to intervention completers, which may introduce selection bias, potentially overestimating engagement and underestimating implementation barriers, even after sensitivity analysis. Third, the single-center setting and small sample size limit the generalizability of findings across geographic, demographic, and statistical dimensions. Fourth, we did not analyze outcome differences across CKD stages 1 to 5, further limiting the applicability of the results.

This exploratory mixed methods study indicated that the nurse-led digital sodium restriction intervention was associated with modest gains in participants’ sodium-related knowledge as a secondary outcome. However, this study encountered prominent practical difficulties in implementing high-load outcome monitoring, with the completion rate of 24-hour urine collection reaching only 24.2% (24/99). The misalignment between idealized trial protocols and patients’ daily realities underscores the need for more feasible, patient-centered outcome measures and adaptive intervention designs. Rather than drawing conclusions about efficacy, this study provides critical insights into implementation barriers and informs the development of future person-centered digital tools. Digital health interventions may be improved via technical enhancements and better adaptation to patients’ daily lives. Future research should use implementation science and mixed methods to explore intervention mechanisms, thereby narrowing the gap between health knowledge and self-management in chronic disease management.