Scale Development
Stage 1: Constructs of the Scale
For the literature review, a total of 9,969 studies were retrieved from the search engines. Of the studies, 4,730 duplicate papers were excluded, resulting in a remaining 5,239 studies. After excluding 4,476 papers that were not related to this study, 763 papers were selected. Then, 731 articles were excluded based on the selection criteria. Of the remaining 32 papers, 3 were systematic review studies, which already had quality evaluation, and 29 papers were evaluated using the critical by Joanna Briggs Institute [25]. As a result, 13 studies with methodological flaws were excluded, resulting in the final 19 studies for review.
A review of the 19 studies showed that essential attributes of the domain of universal self-care were maintaining proper nutritional status, balance between rest and activity, assistance-independence balance, participation and support in social activities, improvement of the residential environment, and overcoming fear of falling. Important attributes of the domain of developmental self-care were self-efficacy, improvement of resilience, and stress management. In the domain of health-deviated self-care, the essential attributes derived from the literature review were health status recognition, problem-solving skills, knowledge acquisition, rehabilitation implementation, daily life change adaptation, and pain/discomfort management. A total of 14 factors and 52 priori items were developed from the literature review.
For patient interviews, 11 older adults of varying age, gender, education level, family type, religion, and time after operation were recruited. Through individual interviews, the domain of spiritual support was derived as an attribute of self-care after hip fracture surgery and was thus added into the item pool. As a result, a total of 15 factors and 45 codes were derived and validated from the individual interviews.
Stage 2: Item Composition
Similarities and differences among the 15 factors and 45 priori items were identified through a conceptual reasoning process, which resulted in 7 components with 73 preliminary items: 4 items for maintaining a normal nutritional status, 17 items for effort in rehabilitation, 13 items for symptom recognition and management, 11 items for a safe environment, 11 items for participation in and support for social activities, 13 items for positive mental health, and 4 items for spiritual support.
Stage 3: Determination of the Scale
A 5-point Likert scale was used to measure the levels of SC in each item, ranging from 1 (not at all) to 5 (strongly agree).
Stage 4: Test of Content Validity
For the primary content validity, only items with an I-CVI of 0.78 or higher were selected. If the meanings of the items were duplicated based on the expert opinions, they were revised by deletion, correction, and integration. Out of 73 preliminary questions, 11 questions were revised and 42 questions were deleted, resulting in a total of 31 preliminary questions. In the process of examining S-CVI, 3 items were deleted and 7 items were revised, leading to 28 items. The secondary content analysis test showed that the I-CVI of all items was 0.78 or higher, ranging from 0.83 to 1. The S-CVI/AVE was 0.95, which met the criterion value of 0.90 or higher, thus securing the content validity of the scale.
Stage 5: Item review
After the linguistic adequacy of the items was reviewed by a scholar with major in Korean literacy, 3 items were revised. The preliminary survey containing 28 questions was conducted on 25 older adults who underwent HF surgery. The average survey time was 19.6 min. Based on a Likert 5-point scale, the overall clarity score of the questions was 3.85±.92, and the overall understandability of the questions was 3.43±.78.
Scale Verification
Stage 6: Scale application
Table 1 shows the general characteristics of the subjects. The average age of the participants was 83.25 years old, and 58.33% of the individuals were female. Regarding the educational level, 44.67% of the participants were uneducated. The majority of subjects lived with family. Approximately half of the participants were not employed, and 65.33% were religious. The functional state before hip fracture was normal in 216 patients. Regarding the type of surgery, 49.33% had total hip arthroplasty. The overall characteristics were homogeneous, as there were no significant differences in general characteristics between EFA and CFA groups (Table 1).
Stage 7: Evaluation of the scale
Construct validity
When the value of the Kolmogorov–Smirnova test or the Shapiro–Wilk test is p>.05, it can be considered to satisfy normality. The items of this scale did not show normality with p < .001, so the normality was re-assessed by skewness and kurtosis values. The corrected item-total correlation coefficients ranged from .128 to .615 and the total Cronbach's alpha reliability for the 28 items was 0.82. A total of 24 items were used for EFA, as 4 items with a correlation coefficient less than .3 were deleted from the pool.
As the result of primary EFA on 24 items, five factors were extracted. After deleting two items showing the communality of less than .5, secondary factor analysis was conducted on the 22 remaining items. There was no problem in communality, but one item was not loaded into any factor and thus was deleted. Tertiary factor analysis was performed on the 21 items. The results showed that the Kaiser-Mayer-Olkin (KMO) value and Bartlett's test of sphericity were statistically significant, confirming that the data was suitable for factor analysis. The communality ranged from .571 to .812, and the factor loadings ranged from .514 to .870. The cumulative explanatory power of the factors was 66.32%.
In EFA, a total of five factors were extracted from the data. The first factor included 5 items and was named as ‘functional independence.’ This factor was composed of behaviors to enhance proper nutrition and rehabilitation and focused on functional aspects in the recovery of hip fracture surgery. Factor 2 included five items and was named ‘symptom recognition and managements.’ Factor 3 was composed of five items and was named ‘positive mental health.’ Factor 3 included constructs of positive mental health and spiritual support and focused on the older adults’ psychological capacity to adjust themselves to changes in daily life and overcome barriers to rehabilitation. Factor 4 included three items and was named ‘participation in and support for social activity.’ Factor 5 was composed of three items and was named ‘safe environment. ’ (Table 2).
CFA was conducted on 157 study participants to confirm the model fit of the five factors derived from the EFA and to verify the convergent and discriminant validity of the constructs. The initial model fit indices tested with 21 items did not meet the standard values. After removing three items with an Standardized regression weights (SRW) less than .5 and an Squared multiple correlation (SMC) value of less than .5, the secondary CFA was performed. In the analysis of the model of the secondary CFA, the value of χ2 did not fit the standard value. However, since χ2 value tends to be easily rejected due to its overt sensitivity to sample size [24], the overall model fit should be evaluated with other fitness indices along with χ2 value. Other goodness-of-fit indices showed that the model was adequate and well explained by the data. Although the GFI value was slightly below the standard of the fit index, the overall model fit was good in consideration of other fit indices such as TLI and CFI.
In terms of convergent validity, all ranges satisfied the recommended criteria of factor loadings, C.R, AVE, and CR, supporting the convergent validity of the scale. Discriminant validity was secured, as all of the AVE values were larger than .727 squared, .529, which is the largest correlation coefficient between the primary discriminant validity outcome factors. The secondary discriminant validity was also secured, as the value obtained by multiplying the correlation coefficient ±2 between the two factors by the standard error did not include 1 (Table 3, Fig. 2).
Concurrent validity was examined with the Pearson correlation coefficients between the HFS-SC scale and the EQ-5D-5L. There were significant positive correlations among the subdomains of the HFS-SC and the EQ-5D-5L. There was also a high correlation between the HFS-SC and the EQ-5D-5L total scores, supporting the concurrent validity of the HFS-SC.
Reliability
To examine test-retest reliability, 52 subjects were evaluated using the HFS-SC scale at four weeks after the baseline measurement point. The ICC of the HFS-SC scores at the two time points was .82. In internal consistency reliability, Cronbach's alpha was .91 for the HFS-SC. By subdomain, Cronbach's alpha reliabilities were .88 for factor 1, .89 for factor 2, .86 for factor 3, .83 for factor 4, and .85 for factor 5 (Table 4).
Stage 8: Scale optimization
The psychometric properties of the HFS-SC were examined and confirmed. The HFS-SC consists of 18 items with inclusion of two inverse items. The scale is composed of 5 sub-domains: functional independence, symptom recognition and management, positive mental health, participation and support in social activities, and safe environment. The total score ranges from 18 to 90, with a higher score indicating a higher level of self-care for older adults who underwent HF surgery.