Design
This was a cross-sectional, descriptive study to develop a structural model of self-management among hospitalized older adults in long-term care hospitals.
Sample
Participants were selected through convenience sampling based on certain inclusion and exclusion criteria, from two long-term care hospitals with 300 or more beds in Jeolla province, South Korea. The selected long-term care hospitals have been certified as first grade by the authorized medical institution certification agency. The study period was from February 17 to March 10, 2021. The inclusion criteria were (a) hospitalized older adults aged 65 years with no psychiatric history and the ability to communicate and respond to the questionnaire; (b) diagnosis of two or more chronic diseases for over a year [6]; and (c) hospitalization in the general ward for 6 months or more (as patients often experience maladjustment periods of 3-6 months after being admitted to these facilities) [25,26]. The exclusion criteria were (a) diagnosis with severe dementia, (b) hospitalization in dementia wards, and (c) diagnosis of only a single chronic disease.
In the structural model, the sample size was estimated via maximum likelihood, while the minimum adequate ratio between parameter estimation and the sample size was 10:1 [27]. Thus, considering the 25 unknown parameters in this study, data were collected from 300 participants to satisfy the minimum requirement of 250 participants while accounting for a potential 20% dropout rate. In the final analysis, responses from 287 participants were used, with 13 incomplete responses excluded.
Ethics approval and consent to participate
To protect the subjects, this study was conducted with the approval of the Institutional Review Board at Yonsei University (Project No. Y-2020-0221). After explaining the study purpose, patients who submitted a signed consent form were enrolled.
Data collection
Data collection was conducted by convenience sampling based on eligible recommendations from directors of each hospital. To ensure that participants clearly understood the questionnaire, the research assistants read each question aloud, and gave participants time to respond or ask questions. The research assistants were three trained nurses with a minimum of five years of clinical experience in nursing older adults at care hospitals. In a meeting a week before data collection, the nurses were given a 30-minute training session on the study’s purpose, participants, precautions, and ethical considerations, as well as about each data collection tool.
All tools in this study were used with the permission of the corresponding authors. Prior to the structural model analysis, the tools were validated through confirmatory factor analysis. They satisfied the ≥.50 factor loading, confirming their validity. Data for various variables were collected using the following tools:
General characteristics
General data were collected regarding age, sex, religion, marital status, educational background, length of stay, insurance type, and number and types of chronic diseases; while cognitive function was measured by the Korean version of the Mini-mental state exam, K-MMSE.
Self-management
For self-management, 12 questions from the Partnership in Health (PIH) scale [28] translated and modified in Korean [29] were used. The subcategories were coping, partnership, awareness and management of symptoms, and knowledge of disease and treatment. Each question was rated on a nine-point Likert scale, with a higher score indicating a higher level of self-management [28]. Cronbach’s α for the reliability of the PIH was .86 in the Korean study [29], and .98 in this study.
Type D personality
For the type D personality, 14 questions from the type D Personality Scale-14 [22] translated and modified into Korean [30] were used. The subcategories were negative affectivity (NA, seven questions) and social inhibition (SI, seven questions). Each question was scored on a five-point Likert scale, with a higher score indicating a higher level of personal disposition [22,30]. The original Cronbach’s α was .87; .88 for NA and .86 for SI [22]. In this study, Cronbach’s α was .93 for both NA and SI, and .96 for the total score of the type D personality.
Cognitive illness perception
For cognitive illness perception, the Brief Illness Perception Questionnaire (Brief-IPO) [31] was used in its Korean translation [32]. Among the nine questions constituting the Brief-IPQ, six items that measure cognitive illness perception were used in this study, which were: consequences, timeline, personal control, treatment control, identity, and coherence [31,32]. Each question was rated on a 10-point Likert scale, in which an inverse scale was used for the questions on the consequences, identity, and timeline, with higher scores indicating lower levels of perceived risk. The Cronbach’s α in this study was .95 for the total score of cognitive illness perception.
Depression
For depression, the Center for Epidemiologic Studies Depression - 10 (CES-D 10) scale was used. The CES-D 10 is comprised of 10 questions rated on a five-point Likert scale, with higher scores indicating higher levels of depression. The total score ranges between 0 and 30, as inverse scoring was used for questions 5 and 8. Scores ≥10 indicate clinical depression [33]. Cronbach’s α was .71 in the original study [34] and .92 in this study.
Approach coping
For approach coping, 12 questions from the brief-coping orientation to problems experienced (COPE) [35] in their Korean translation [36] were used. The subcategories were active coping, emotional support, use of information, positive reframing, planning, and acceptance. Questions are scored on a four-point Likert scale (0–3), with higher scores indicating higher levels of coping [35,36]. In the original study, Cronbach’s α was .68 for active coping, .71 for emotional support, .64 for use of information, .64 for positive reframing, .73 for planning, and .57 for acceptance [35]. In this study, Cronbach’s α was .95 for the total score of approach coping.
Data analysis
Data were analyzed using SPSS version 23.0, AMOS 23.0, and SEM. Baseline characteristics were measured using descriptive statistics, skewness, and kurtosis parameters. The fitness of the proposed model with data was measured using the chi-square/degree of freedom ratio (X2/DF), Turker-Lewis index (TLI), comparative fit index (CFI), and root mean squared error of approximation (RMSEA). The validity of the model was determined based on CMIN/DF ≤3 [37], CFI and TLI ≥0.90 [38], and RMSEA ≥0.06 and ≤0.08 [39]. The direct, indirect, and total effects of the model were verified using bootstrapping with a sample size of 2,000 and a confidence level of 95% for testing statistical significance. Convergent validity, which is used to verify high correlations among measured variables constituting the potential variables, was measured using construct reliability (CR) (≥0.7) and average variance extracted (AVE)(≥0.5) [27]. Discriminant validity was measured by comparing the AVE and ρ2 values of each construct to determine whether the former exceeded the latter [37].