The CONSORT checklist guided the reporting of this study. This study was registered in the ‘Overview of Medical Research in the Netherlands’ (nr: 23167) and approved by the Medical Ethical Committee of the University Medical Center Groningen (nr: 201900534).

Design and setting

We conducted a two-armed, non-blinded, quasi-experimental study within general practices and nephrology clinics in the Netherlands. Most Dutch residents are registered with a general practitioner, who usually treats patients with mild to moderate CKD. Patients with more severe CKD are referred to nephrology clinics.

Participants and sampling

Two practices and three clinics provided GoYK, whereas two practices and two clinics provided care-as-usual. We recruited patients with mild to severe CKD (CKD stages 2–4) and professionals from December 2020 until September 2021. Eligible patients were ≥18 years, had ≥3 months of CKD based on estimated glomerular filtration rate (eGFR), and had regular consultations with participating professionals. Patients with kidney failure and major cognitive or life-threatening conditions, and professionals with previous health literacy training were excluded.

Professionals were approached by email and selected eligible patients from their electronic patients records. Four hundred thirty patients received an information letter and consent form during consultations or at their home addresses. At variance with the registered trial, we included patients of various health literacy levels, but oversampled patients with limited health literacy. This enabled to analyze the intervention effectiveness among health literacy subgroups, and to guarantee sufficient participation, during COVID-19.

Sample size

To achieve 80% power, an a priori calculation for multilevel analysis set our needed sample at 91 patients nested under 38 professionals in each group. First, the number of professionals was based upon the outcome ‘use of health literacy communication strategies’ and assumed an effect size of 0.65. Second, we estimated the number of patients needed to detect a change of 10.0 in the intervention-control group difference for the patient activation measure [8] with a standard deviation of 15.8 [17], and a two-tailed alpha of 0.05. To allow for 5% attrition among professionals and 10% attrition among patients, we aimed to enroll 40 professionals and 102 patients per group.

Intervention

GoYK is a multi-component intervention, offered on top of care-as-usual, developed in co-creation with patients and professionals [15]. For patients, the intervention enhances CKD knowledge, communication and self-management through simple text and visuals. Professionals followed an e-learning course and attended a lesson session on health literacy, based upon the effective health literacy training of Kaper et al. [18]. Supplementary file 1 contains a detailed description.

Care-as-usual

In line with guidelines [19], patients had two to four annual consultations with a general practitioner, nephrologist or a specialized nurse practitioner. If needed, patients in the nephrology clinics received additional care from a nurse practitioner, dietitian or social worker. Patients in CKD-stage 4 received more intensive care, often having consultations every four to six weeks. Protocolled consultation topics in all settings were lab results, medication, lifestyle and, if applicable, co-morbid diseases.

Procedures

For patients and professionals, we collected data at baseline (T0), and after 4 (T1) and 9 months (T2). We used paper and online questionnaires, sent to the participants’ e-mail or home address. Data collection was tailored to limited health literacy patients, for example by allowing help with the questionnaires. Reminders were sent if questionnaires were not returned, after one and two weeks. Professionals helped to collect data on the patients’ clinical parameters from the electronic patient records.

Randomization and blinding

Within-organization randomization was infeasible. Blinding was not possible due to intervention visibility. Patient groups were concealed until study start to prevent bias.

Measures

Below are the primary and secondary outcomes. Supplementary file 2 provides an overview of the references upon which these outcomes are based.

Primary outcomes for patients

Self-management of health behaviors; Salt intake – Item adopted from Humalda et al. (2020). We asked how many days a week patients consumed salty foods within nine different food groups. Alcohol – We asked how many days per week patients drank alcohol and the average number of doses on those days. Physical activity – We asked the number of days a week patients did at least 30 min of physical activity, based upon O’Halloran et al. (2020). We dichotomized into adequate (5–7 days) and inadequate (0–4 days). Fluid intake – We asked about the number of millilitres of fluid intake per day. We dichotomized it into inadequate (< 1500 mL) and adequate (≥1500 mL). Medication adherence – The Medication Adherence Report Scale of Chan et al. (2020), 5 items, was used. A total score between 5 and 25 was calculated. Patient activation –Measured with the Patient Activation Measure (PAM-13). The PAM-13 gives a score between 0–100 [8].

Secondary outcomes for patients

Health literacy – 10 items of the All Aspects of Health Literacy Scale (AAHLS) by Chinn et al. (2013). The AAHLS provides a total score between 10 and 30. Patients had limited health literacy when they scored ≤ 25. Quality of the consultation – Based on the communication framework of Haes and Bensing (2009), we developed four statements with 5-point Likert scales asking if patients felt understood, could ask questions and share emotions, and if professionals listened well. For each statement, we calculated the percentage of patients agreeing as a measure for quality. A fifth item asked to select topics discussed during a consultation. We calculated the number of lifestyle topics. CKD clinical parameters –We obtained eGFR, blood pressures and BMI. Patients with a systolic blood pressure ≥ 140 or diastolic blood pressure ≥ 90 were classified as hypertensive.

Outcomes for professionals

The primary outcome for professionals was self-reported use of health literacy communication strategies. We asked 31 questions within domains: gathering information, providing information, shared decision-making and enabling self-management. Each item contained 7-point Likert scales (never-always). Sixteen items were from Kaper et al. (2018). We added fifteen items, in line with our intervention objectives. The secondary outcome was health literacy knowledge. We asked 6 questions with 7-point Likert scales from Kaper et al. on the professionals’ knowledge regarding health literacy (2018).

Background characteristics

For patients, we collected data on sex, age, education, marital status, years of CKD, smoking, and co-morbidities. For professionals, data included sex, age, profession, health setting, experience, expected contact with limited health literacy patients, and communication training.

Process evaluation

In the intervention group, we asked patients about the use (yes/no) and the usefulness of (useful, neutral, not useful), and the satisfaction with (yes/no, and grading 1–10) GoYK. Professionals answered questions on usefulness and satisfaction, and statements with 1–7 Likert scales, to check if they expected to use the learned strategies in the future.

Analyses

First, missing patient data were imputed (% missing = 0.6–10.6%) by performing 20 imputations using fully conditional specification, and predictive mean matching for continuous variables. The model included baseline variables associated with the outcomes to improve the prediction. For professional outcomes, we did not perform imputation because missing seemed not at random, as four of the seven non-returned questionnaires came from physicians. Second, we determined whether a two-level structure was present, with patients nested under professionals, by assessing the intra-class correlations at the professional level. These were close to zero, meaning that one- and two-level analyses yielded similar results. We reported the one-level analyses.

We calculated baseline descriptive statistics and evaluated differences between the intervention group and care-as-usual. Then, we assessed the intervention effect by performing linear or multinomial logistic regressions, with significance set at p < 0.05. We adjusted the patient analyses for baseline eGFR given a difference between the intervention group and care-as-usual. As a sensitivity analysis, we reran patient analyses with the non-imputed dataset and without correction for eGFR. We performed subgroup analyses for (1) patients at risk (i.e. with inadequate health behaviors, hypertension and perceiving the consultation quality as low), (2) patients with limited health literacy, (3) patients from general practices and nephrology clinics, and (4) patients who used ≥2 intervention components. In the process evaluation, we calculated descriptive statistics for all outcomes and analyzed differences between patients with limited and adequate health literacy, and from general practices and clinics.