A cross-sectional study was conducted on a sample of adults (≥ 18 years of age) with stage 5 CKD-ND The kidney care program (KCP) is composed of two senior CKD nurses and three nephrology nurse practitioners working collaboratively with nephrologists and allied health to provide a single point of contact for patients with stages 4–5 supporting decision-making through information about the different kidney treatment modalities. The kidney care program at the time of this study included 350 patients.

Patients were invited to take part in this study during a six-month period between July and December 2022. Adults with a diagnosis of stage 5 CKD-ND (defined as an estimated glomerular filtration rate ≤ 15 ml/min/1.73 m2) were invited via phone, email or face-to-face during clinic attendance. Patients were excluded if they were admitted as an inpatient in hospital, had severe cognitive impairment and/or inability to complete the surveys written in English. Study participants completed the 13-item Patient Activation Measure (PAM-13) and 8-item Morisky Medication Adherence Scale (MMAS-8) questionnaires. Sociodemographic and clinical data were extracted from participant’s electronic medical records and entered into Research Electronic Data Capture (REDCap) software version 1 21.09.2022 [19].

This project received ethical approval from Central Adelaide Health Local Network (CALHN) (reference number 16067), Northern Adelaide Health Local Network (NALHN) (SSA reference number 22-040) and University of South Australia (application ID: 205151) Human Research Ethics Committee and Research Governance. The study was registered with the Australian New Zealand Clinical Trials Registry (ANZCRT) #12622000451707. The PAM-13 instrument is protected by Insignia Health license #1654066536–1685602536. This study complies with the STROBE Checklist for observational cross-sectional studies [20] (Supplemental Table 1).

Patient activation was evaluated using the PAM-13 with a 5 point Likert scale (strongly disagree to strongly agree). A fifth, not applicable (NA) response, was also offered. The scores for this tool were then converted to an overall activation score between 0 and 100, with higher values reflecting higher activation. Levels are designated by cut-off points based on the 2013 PAM license scoring rules (Insignia®) [21] and used by Hussein et al. [18]. PAM-13 scores were categorised into four levels. Level 1 (0.0–47.0) reflects low activation and suggests that the person does not yet understand their role in healthcare; Level 2 (47.1–55.1) indicates the person does not yet have the knowledge and confidence to take action; Level 3 (55.2–72.4) indicates the person is beginning to engage in positive health behaviours; Level 4 (72.5–100) reflects high activation and suggests the person is proactive and engaged in recommended health behaviours [18, 21].

Sociodemographic characteristics operationalised as categorical variables: age (younger than 64 years vs older than 65 years [older adults], classification based on the Australian Bureau of Statistics [22], sex (female vs male), marital status (considered in terms of single [never married], partnered [married or de facto relationship] and separated [divorced or widowed]), current living situation (people living alone, living with family [partner, children or parents] or others [supported by carers or friends]), educational level (below high school [did not finish primary school, did not go to school, did not complete high school], completed high school and higher than high school [college, university degree or postgraduate degree]), and ethnicity (assessed using the Australian Bureau of Statistics (ABS) [23]. Socioeconomic status was calculated using the Socio-Economic Indexes for Areas (SEIFA) Australian Bureau of Statistics [24]. The SEIFA disadvantage score, a quintile based on residential postcode data from the Australian Bureau of Statistics census, classified the postcodes following the Index of Relative Social Disadvantage (IRSD). This IRSD index summarises the socio-economic conditions of people living in an area. Each postcode of the IRSD was categorised into five quantiles: the first quantile represents the most disadvantaged and the fifth quantile represents the least disadvantaged population.

Treatment adherence measures were the number of missed appointments at kidney care and nephrologist clinic in the previous 12 months, and Morisky Medication Adherence Scale (MMAS-8) [25, 26]. The MMAS-8 consists of seven items with the binary response and one item with the Likert scale response. Cumulative scores based on eight items were used to obtain a final adherence score ranging from 0 to 8. Adherence was defined accordingly as low (score 0–5), medium (score 6–7) and high (score 8). MMAS-8 has been validated in studies with good reliability and predictive value [25].

Healthcare utilisation was measured by the number of hospital emergency department visits over the past 12 months and the number of hospital admissions over the prior 12 months.

All patients with stage 5 CKD in the kidney care program were identified and listed in alphabetic order by patient’s last name. Individual patient eligibility was assessed, and those eligible were selected sequentially from the list, assuming simple random sampling has occurred. The overall proportion of each patient activation level in the target population, patients in the kidney care program, was estimated. Sub-sample estimation of each patient activation level, where a sub-sample was defined by the characteristics concerning a sociodemographic factor was also conducted. The strength of evidence for the association between each sociodemographic factor and patient activation level was tested in a modified chi-squared test for contingency table, which considered survey weights. Due to ethics restrictions, no information was collected for non-respondents, hence only crude non-response rate was considered in the post-estimation weight construction.

The relationship between patient activation, treatment adherence and healthcare utilisation variables were examined via regression method after adjusting for confounders. Confounders were informed by expert clinical content knowledge in CKD (LL, PB, SJ) and existing literature [27]. Age and education level were confounders for hospital emergency department visits, hospital admissions and missed renal appointments over the last 12 months. Age, education level, and participant’s home status (self/family/others) were confounders for medication adherence.

A model-based approach, as opposed to a design-based approach, was used in the analysis to allow the findings to be generalised to CKD populations in other settings. Negative binomial distribution was used for count outcome measures, hospital admissions, emergency visits, and missed renal appointments, to account for the overdispersion. Linear regression was used for outcome measure medication adherence score. In modelling each outcome measure, different functional forms of patient activation score were fitted, and the selection of the model was guided by the information criteria and the ease of interpretability of the functional form itself. Statistical significance was indicated by a P value of < 0.05.