Chronic Diseases Self-Efficacy Profiles and Healthcare Utilisation Patterns Among Older Adults with Multimorbidity in Albania: A Person-Centred Latent Profile Analysis

Background/Objectives: Multimorbidity renders self-efficacy a continuing partnership between older people, families, and healthcare and social care services. We examined whether nine domain-specific self-efficacy indicators were best understood as one overall level of confidence or as recurring configurations in which confidence was distributed differently across everyday tasks. The primary aim was to identify and interpret these configurations in older adults attending primary care in Vlora Municipality, Albania. Secondary aims were to evaluate the quality and reproducibility of the solution and to describe health literacy and six-month healthcare utilisation, without treating either as a cause or as a profile-defining variable. Methods: This verified secondary cross-sectional analysis used data from 464 community-dwelling adults aged 65 years or older with at least two chronic conditions recruited through five primary healthcare centres. After two questionnaire values were checked against the source records, 459 participants with complete nine-domain self-efficacy data formed the primary sample. We compared 40 Gaussian finite-mixture candidates: one to ten profiles under four covariance structures. Model choice used a combination of Bayesian information criterion (BIC), integrated completed likelihood (ICL), Akaike information criterion (AIC), entropy, average posterior probability, profile size, convergence, covariance diagnostics, interpretability, and reproducibility across 50 random starts. There is no single closed-form power calculation for profile enumeration; therefore, precision was assessed empirically with missing-item, domain-omission, and bootstrap checks. Health literacy and healthcare utilisation were kept outside profile formation and were described with conditional adjusted count models. Results: The retained exploratory Profile-specific full-covariance model (VVV) two-profile solution had 249 and 210 participants (mclust BIC = −8125.080; ICL = −8181.987; AIC = 7675.013). Assignment quality was reasonably precise (entropy = 0.814; average posterior probabilities = 0.961 and 0.931), but the profile solution was not simply a low-to-high severity scale. One pattern was stronger for finding information, obtaining support, and communicating with physicians; the other was stronger for exercise, household and social activities, symptom management, and emotional management. In 100 bootstrap refits, the adjusted Rand index ranged from 0.212 to 0.865 (median 0.584), so stability was moderate. After covariate adjustment, Profile 2 had a modelled general-practitioner visit rate that was 19% lower than Profile 1 [incidence rate ratio (IRR) = 0.809; 95% confidence interval (CI) 0.687–0.953; p-value = 0.011]; this is an exploratory association conditional on the fitted model, not evidence of a causal effect. Conclusions: A profile can show how confidence is organised across daily self-efficacy tasks even when two groups have similar overall means. For health and social care, the findings support assessment that links information and communication with activity, symptom management, emotional wellbeing, family support, and community resources. The profiles are exploratory conversation prompts rather than diagnostic categories, and the cross-sectional data do not show that self-efficacy causes healthcare use.