Geriatric Nursing 36 (2015) 301e305
Contents lists available at ScienceDirect
Geriatric Nursing journal homepage: www.gnjournal.com
Feature Article
Psychometric evaluation of the Korean Version of the Self-Efficacy for Exercise Scale for older adults Mona Choi, PhD, RN a, Sangwoo Ahn, BSN, RN b, Dukyoo Jung, PhD, RN c, * a
Nursing Policy Research Institute, College of Nursing, Yonsei University, Seoul, Korea College of Nursing, Yonsei University, Seoul, Korea c Division of Nursing, College of Health Sciences, Ewha Womans University, Seoul, Korea b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 26 November 2014 Received in revised form 21 March 2015 Accepted 30 March 2015 Available online 29 April 2015
We evaluated the psychometric properties of the Korean version of the Self-Efficacy for Exercise Scale (SEE-K). The SEE-K consists of nine items and was translated into Korean using the forward-backward translation method. We administered it to 212 community-dwelling older adults along with measures of outcome expectation for exercise, quality of life, and physical activity. The validity was determined using confirmatory factor analysis and Rasch analysis with INFIT and OUTFIT statistics, which showed acceptable model fit. The concurrent validity was confirmed according to positive correlations between the SEE-K, outcome expectation for exercise, and quality of life. Furthermore, the high physical activity group had higher SEE-K scores. Finally, the reliability of the SEE-K was deemed acceptable based on Cronbach’s alpha, coefficients of determination, and person and item separation indices with reliability. Thus, the SEE-K appears to have satisfactory validity and reliability among older adults in South Korea. Ó 2015 Elsevier Inc. All rights reserved.
Keywords: Aged Exercise Psychometrics Questionnaires Self-efficacy Rasch analysis
Introduction Physical activity helps to promote and maintain older adults’ health. Regardless of sex, people with high levels of physical activity are likely to have a longer life expectancy than people who are sedentary.1 Furthermore, physical activity can help reduce depression, particularly in older adults.2 Despite these facts, older adults appear to participate less in physical activity compared with adults younger than 65 years old. For instance, in the United States, 11.9% of older adults aged 65 and over in 2012 met the recommended guidelines for physical activity, contrary to the 17.2% of adults aged 45e64.3 In the United Kingdom, the number of people who do not participate in at least moderateintensity activity increases proportionally with age, with an especially steep rate among people aged 65 or older.4 Physical activity in older adults in South Korea has been gradually decreasing since 2008: Among those aged 70 years or older, only 30.4% and 45.5% of women and men, respectively, participate in moderate-intensity
Conflict of interest: None of the authors have any conflicting interests. * Corresponding author. Department of Nursing Science, College of Health Sciences, EwhaWomans University, 309, Hellen hall, 11-1 Daehyun-dong, Seodaemoon-gu, Seoul, Korea. Tel.: þ82 2 3277 6693; fax: þ82 2 3277 2850. E-mail address: [email protected] (D. Jung). 0197-4572/$ e see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.gerinurse.2015.03.005
physical activity.5 Indeed, men and women aged 70 years or older appear to be the age group with the lowest level of participation in moderate-intensity physical activity in South Korea.6 The above examples indicate that older adults’ rate of participation in physical activity tends to be lower than that of other age groups, which is problematic because physical activity is expected to help reduce the burden on health and social care by enabling healthy aging7; it is difficult for older adults to expect this benefit unless they regularly take part in physical activity. The factors related to decreased physical activity in older adults include medical problems, physical frailty, negative experience, fear of activityrelated injury or falling, having a sedentary lifestyle in the past, insufficient understanding of physical activity, living in an unsafe neighborhood, and lack of company.8,9 In contrast, people with higher self-efficacy for exercise are known to be more likely to initiate and maintain physical activity. In this regard, self-efficacy has been referred to as an important factor influencing physical activity.10e12 Self-efficacy theory originated in social cognitive theory,13 which has been used to explain the factors affecting exercise behavior among older individuals. The primary assumption of selfefficacy theory is that behavioral change and maintenance of that change rely predominantly on an individual’s belief in their ability to perform a certain behavior (i.e., their self-efficacy).13 Self-efficacy plays an important role in maintaining older adults’ participation in
302
M. Choi et al. / Geriatric Nursing 36 (2015) 301e305
functional activities and exercises.14 Several interventions have been developed and tested to strengthen self-efficacy expectations related to exercises.12,15,16 There have been many efforts to measures people’s confidence in maintaining physical activity when confronting obstacles to exercise. One such measure, the Self-Efficacy for Exercise scale (SEE), was developed and verified as a valid and reliable tool in older populations,17 and further applied to African American and Latino older adults18 in the United States. The SEE has also been translated into Chinese19 and Swedish,20 and these two translated versions were evaluated for the reliability and validity in samples of Taiwanese and Swedish population. We found no scale in Korean for self-efficacy for exercise for older adults available; thus, we deemed the SEE scale a suitable choice for measuring self-efficacy expectation because it is developed specifically for older adults and easily applicable to them, owing to its short length. With the aim of wide use of the SEE scale in the geriatric care and physical activity fields for Korean older adults, we evaluated the psychometric properties of the Korean version of the SEE (SEE-K) by using confirmatory factor analysis and Rasch analysis. Material and methods Participants and procedures Study participants were recruited from three community welfare centers for older adults in two districts of Seoul, South Korea, and one located close to Seoul. The inclusion criteria were being aged 65 years or older, community dwelling, and independent in activities of daily living. With the cooperation of personnel from the three community welfare centers, we visited the centers and announced this research study to older adults who were at the community welfare centers in August, 2013. Trained research assistants approached about 240 older adults, of which 215 consented to participate in the study. A signed informed consent form was procured from each participant before data collection. The questionnaires were completed either by participants themselves or, for participants unable to complete the questionnaires by hand, by research assistants using interviews. After completion of the questionnaire, a small gift was given to the participants. After removing incomplete questionnaires, 212 surveys were included in the final data analysis. This study was approved by the institutional review board of the College of Nursing at Yonsei University in Seoul, South Korea (No. 2013-0024). Measurements We collected information on participants’ age, sex, education, height, weight, perceived health status, and number of diseases using the questionnaire. Then, body mass index (BMI) was calculated from participants’ reported height and weight. The other measurements used in this study are described below. Self-Efficacy for Exercise The original English version of the SEE contains nine items measuring confidence in ability to exercise under certain conditions, for example, bad weather, emotions, and busy or stressful situations. Participants rated their confidence about maintaining exercise three times per week for 20 min for each condition, ranging from 0 (not confident) to 10 (very confident). A higher score represents better self-efficacy for exercise.17 After permission was received from the original developer of the SEE, the English version was translated into Korean (forward translation). Then, two nursing professors in Korea reviewed the translation accuracy. Finally, this version was back-translated into
English by a Korean-American, who is proficient in both Korean and English. Outcome Expectations for Exercise-2 The Korean version of the Outcome Expectations for Exercise-2 scale (OEE-2-K)21 was also used. This scale was derived from the OEE-2, which consists of 13 items, measuring nine positive and four negative outcome expectations for exercise.22 The scale uses a 5-point Likert scale, ranging from 1 (not at all) to 5 (absolutely agree). A higher score indicates higher expectations for the outcomes of exercise. The Cronbach’s alpha in the present study was 0.834. Quality of life The EuroQol group’s 3-level version of the EQ-5D was used to measure quality of life (QoL).23 This tool was developed as a generic core index of health for use in health care evaluation worldwide, including in South Korea. Using a questionnaire format, this tool requires each participant to rate five dimensions of QoLdmobility (M), self-care (SC), usual activities (UA), pain/discomfort (PD), and anxiety/depression (AD)don a three-point scale of severity, as follows: no problem (level 1), some or moderate problem (level 2), and serious problem (level 3). The EQ-5D health states can be converted into a single summary index by applying a formula that essentially attaches values (i.e., weights) to each of the levels in each dimension. The following formula has been used in the Korean population to calculate the EQ-5D-3L index24:
EQ 5D index ¼ 1 ð0:050 þ 0:096 M2 þ 0:418 M3 þ 0:046 SC2 þ 0:136 SC3 þ 0:051 UA2 þ 0:208 UA3 þ 0:037 PD2 þ 0:151 PD3 þ 0:043 AD2 þ 0:158 AD3 þ 0:050 N3Þ For example, “M2” represents someone who gave a level 2 (some or moderate problem) for mobility and “N3” indicates whether there is any dimension rated on a level 3. Thus, the EQ-5D index value for an individual who gave ratings of 3, 2, 3, 2, and 2 for the five dimensions would be calculated as follows: 1 (0.050 þ 0.418 þ 0.046 þ 0.208 þ 0.037 þ 0.043 þ 0.050) ¼ 0.148.24 Physical activity To assess physical activity, the International Physical Activity Questionnaire-short form (IPAQ) was used. This consists of assessments of vigorous- and moderate-intensity activities and walking. We converted minutes and number of days spent engaging in those three types of physical activity into total MET minutes per week. Then, we categorized participants into three groupsdlow, moderate, and high physical activitydbased on the total MET minutes per week. The IPAQ data processing guideline25 was used for calculation of MET minutes per week and grouping. Data analysis Stata version 12 was used for descriptive and inferential statistics and confirmatory factor analysis, and Winsteps 3.81 was used for Rasch model testing. Validity Construct validity was tested using confirmatory factor analysis with the maximum likelihood estimation method. To estimate model fit, the chi-square statistic, chi-square/degrees of freedom ratio, comparative fit index (CFI), TuckereLewis index (TLI), and
M. Choi et al. / Geriatric Nursing 36 (2015) 301e305
root mean-square error of approximation (RMSEA) were used. Although a p-value of greater than 0.05 on the chi-square goodness-of-fit test indicates a good fit, the chi-square test tends be very sensitive to sample size, and for large samples, the chi-square will almost always be significant. Thus, the chi-square/degrees of freedom ratio was also calculated; a value of less than 2 indicates a good model fit.26 The CFI, TLI, and RMSEA are less affected by sample size, and CFI and TLI values close to 0.95 and RMSEA values close to 0.06 are considered to indicate good fit.27 The concurrent validity, a subtype of criterion-related validity, was determined by testing whether SEE-K scores were correlated with OEE-2-K and EQ-5D Index scores and physical activity levels. Previous studies identified positive correlations between the SEE and OEE12,28 and another study revealed positive correlations between SEE and QoL and between SEE and physical activity.29 Thus, we hypothesized that the OEE-2-K, EQ-5D, and the intensity of physical activity were positively related to the SEE-K. We tested the relations of the OEE-2-K and EQ-5D with the SEE-K using Pearson’s correlation coefficients. For intensity of physical activity, we used analysis of variance (ANOVA) to compare the physical activity groups to determine if SEE-K scores increase with physical activity. Rasch analysis, which is based on item response theory, was performed to test the validity of the SEE-K by considering it as a unidimensional model. The model fit of the Rasch measurement model was tested using INFIT and OUTFIT item fit statistics. Acceptable limits of INFIT and OUTFIT statistics are between 0.6 and 1.4. INFIT and OUTFIT values of less than 0.6 indicate that there are likely redundant items, leading to overfitting. In contrast, INFIT and OUTFIT values of greater than 1.4 indicate that items may define the construct of interest differently from the rest of the items in the scale, implying underfitting.30,31 We also built a person-item map to display the item difficulty distribution of the SEE-K against individuals’ ability level; the item difficulty is displayed on the right side of the vertical line and individuals’ ability level on the left side on the same scale, which is known as a logit scale.31 When persons are placed outside of the most difficult or easiest items on the map, additional items may be needed to differentiate those who have higher or lower scores.31 Further details on Rasch analysis can be found elsewhere, both in relation to theoretical aspects31 and the nursing field.21,32,33 Reliability The internal consistency of the SEE-K was determined using Cronbach’s alpha, where 0.7 or greater was considered indicative of good internal consistency.34 From the confirmatory factor analysis, a coefficient of determination (R2) of 0.5 for each item and a composite R2 of 0.6 for the overall instrument are considered adequate indicators of reliability.35 Finally, a person separation index with reliability and an item separation index with reliability were also reported from the Rasch analysis. The person separation index is used to classify high and low performers. Likewise, the item separation index is used to discriminate items from one another according to the item difficulty hierarchy.31 Person and item separation indices of 1.5 are considered to have an acceptable level of separation, with 2.0 being a good level and 3.0 being an excellent level. The person or item reliability can be interpreted as Cronbach’s alphas.36 Results The general characteristics of the study participants are presented in Table 1. Of the 212 participants, the majority were women (n ¼ 144, 67.9%). The mean age was 76.02 5.80 years, ranging from 65 to 97 years. About two-thirds of participants were in the
303
Table 1 Participants’ general characteristics, SEE-K, OEE-2-K, EQ-5D, and IPAQ (n ¼ 212). Characteristics
Categories
n (%)
Sex
Male Female 65e74 75 25: Overweight or obese Good Fair Bad Elementary school or less Middle school High school College or higher
68 144 90 122 4 135 60 93 85 34 71 43 59 37
Age (yr) a
2
BMI (kg/m )
Perceived health status Educationa
SEE-K OEE-2-K EQ-5D Indexb IPAQc
(32.1) (67.9) (42.5) (57.5) (2.0) (67.8) (30.2) (43.9) (40.1) (16.0) (33.8) (20.5) (28.1) (17.6)
Mean SD
76.02 5.80
52.50 20.12 54.81 6.86 0.86 0.10 Low Moderate High
49 (23.1) 152 (71.7) 11 (5.2)
SEE-K: the Korean version of the Self-Efficacy for Exercise; OEE-2-K: the Korean version of the Outcome Expectations for Exercise-2; EQ-5D: EuroQoL-5 Dimensions; IPAQ: International Physical Activity Questionnaire-short form. a Missing cases excluded. b Calculated from the EQ-5D Index formula.24 c Categorized based on the “Guidelines for Data processing and Analysis of the International Physical Activity Questionnaire.”26
normal BMI range, 18.5e25 kg/m2, and most participants perceived their health status as fair or good. The means of the total scores on the SEE-K and OEE-2-K were 52.50 20.12 of a possible 90 and 54.81 6.86 of a possible 65, respectively, and the mean of EQ-5D index was 0.86 0.10. About 70% of older adults had participated in moderate level of physical activity. Validity The construct validity was tested with confirmatory factor analysis. Table 2 shows the path coefficients and R2 estimates for the SEE-K items. Path coefficients for all nine items were greater than 0.5 and statistically significant. The chi-square statistic was statistically significant (c2 ¼ 48.87, df ¼ 26, p ¼ 0.004). However, the chi-square/degrees of freedom ratio was 1.88, indicating a fair model fit. The CFI, TLI, and RMSEA demonstrated acceptable model fit with values of 0.959, 0.944, and 0.064, respectively. Table 2 Path coefficient, and coefficient of determination of each item of SEE-K from confirmatory factor analysis (n ¼ 212). Question items How confident are you right now that you could exercise three times per week for 20 min if 1 .the weather was bothering you? 2 .you were bored by the program or activity? 3 .you felt pain when exercising? 4 .you had to exercise alone? 5 .you did not enjoy it? 6 .you were too busy with other activities? 7 .you felt tired? 8 .you felt stressed? 9 .you felt depressed? Total
Path coefficients (l)
Reliability estimates (R2)
0.57 0.59
0.33 0.35
0.61 0.53 0.62 0.53
0.37 0.28 0.38 0.28
0.71 0.57 0.59 e
0.50 0.33 0.35 0.82
SEE-K: the Korean version of the Self-Efficacy for Exercise.
304
M. Choi et al. / Geriatric Nursing 36 (2015) 301e305
Table 3 Correlations among SEE-K, OEE-2-K, and EQ-5D (n ¼ 212). Variables
Table 5 Rasch model testing fit statistics and item difficulty.
EQ-5D Index
OEE-2-K
r (p) SEE-K EQ-5D Index
0.169 (0.014)*
0.384 (
Contents lists available at ScienceDirect
Geriatric Nursing journal homepage: www.gnjournal.com
Feature Article
Psychometric evaluation of the Korean Version of the Self-Efficacy for Exercise Scale for older adults Mona Choi, PhD, RN a, Sangwoo Ahn, BSN, RN b, Dukyoo Jung, PhD, RN c, * a
Nursing Policy Research Institute, College of Nursing, Yonsei University, Seoul, Korea College of Nursing, Yonsei University, Seoul, Korea c Division of Nursing, College of Health Sciences, Ewha Womans University, Seoul, Korea b
a r t i c l e i n f o
a b s t r a c t
Article history: Received 26 November 2014 Received in revised form 21 March 2015 Accepted 30 March 2015 Available online 29 April 2015
We evaluated the psychometric properties of the Korean version of the Self-Efficacy for Exercise Scale (SEE-K). The SEE-K consists of nine items and was translated into Korean using the forward-backward translation method. We administered it to 212 community-dwelling older adults along with measures of outcome expectation for exercise, quality of life, and physical activity. The validity was determined using confirmatory factor analysis and Rasch analysis with INFIT and OUTFIT statistics, which showed acceptable model fit. The concurrent validity was confirmed according to positive correlations between the SEE-K, outcome expectation for exercise, and quality of life. Furthermore, the high physical activity group had higher SEE-K scores. Finally, the reliability of the SEE-K was deemed acceptable based on Cronbach’s alpha, coefficients of determination, and person and item separation indices with reliability. Thus, the SEE-K appears to have satisfactory validity and reliability among older adults in South Korea. Ó 2015 Elsevier Inc. All rights reserved.
Keywords: Aged Exercise Psychometrics Questionnaires Self-efficacy Rasch analysis
Introduction Physical activity helps to promote and maintain older adults’ health. Regardless of sex, people with high levels of physical activity are likely to have a longer life expectancy than people who are sedentary.1 Furthermore, physical activity can help reduce depression, particularly in older adults.2 Despite these facts, older adults appear to participate less in physical activity compared with adults younger than 65 years old. For instance, in the United States, 11.9% of older adults aged 65 and over in 2012 met the recommended guidelines for physical activity, contrary to the 17.2% of adults aged 45e64.3 In the United Kingdom, the number of people who do not participate in at least moderateintensity activity increases proportionally with age, with an especially steep rate among people aged 65 or older.4 Physical activity in older adults in South Korea has been gradually decreasing since 2008: Among those aged 70 years or older, only 30.4% and 45.5% of women and men, respectively, participate in moderate-intensity
Conflict of interest: None of the authors have any conflicting interests. * Corresponding author. Department of Nursing Science, College of Health Sciences, EwhaWomans University, 309, Hellen hall, 11-1 Daehyun-dong, Seodaemoon-gu, Seoul, Korea. Tel.: þ82 2 3277 6693; fax: þ82 2 3277 2850. E-mail address: [email protected] (D. Jung). 0197-4572/$ e see front matter Ó 2015 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.gerinurse.2015.03.005
physical activity.5 Indeed, men and women aged 70 years or older appear to be the age group with the lowest level of participation in moderate-intensity physical activity in South Korea.6 The above examples indicate that older adults’ rate of participation in physical activity tends to be lower than that of other age groups, which is problematic because physical activity is expected to help reduce the burden on health and social care by enabling healthy aging7; it is difficult for older adults to expect this benefit unless they regularly take part in physical activity. The factors related to decreased physical activity in older adults include medical problems, physical frailty, negative experience, fear of activityrelated injury or falling, having a sedentary lifestyle in the past, insufficient understanding of physical activity, living in an unsafe neighborhood, and lack of company.8,9 In contrast, people with higher self-efficacy for exercise are known to be more likely to initiate and maintain physical activity. In this regard, self-efficacy has been referred to as an important factor influencing physical activity.10e12 Self-efficacy theory originated in social cognitive theory,13 which has been used to explain the factors affecting exercise behavior among older individuals. The primary assumption of selfefficacy theory is that behavioral change and maintenance of that change rely predominantly on an individual’s belief in their ability to perform a certain behavior (i.e., their self-efficacy).13 Self-efficacy plays an important role in maintaining older adults’ participation in
302
M. Choi et al. / Geriatric Nursing 36 (2015) 301e305
functional activities and exercises.14 Several interventions have been developed and tested to strengthen self-efficacy expectations related to exercises.12,15,16 There have been many efforts to measures people’s confidence in maintaining physical activity when confronting obstacles to exercise. One such measure, the Self-Efficacy for Exercise scale (SEE), was developed and verified as a valid and reliable tool in older populations,17 and further applied to African American and Latino older adults18 in the United States. The SEE has also been translated into Chinese19 and Swedish,20 and these two translated versions were evaluated for the reliability and validity in samples of Taiwanese and Swedish population. We found no scale in Korean for self-efficacy for exercise for older adults available; thus, we deemed the SEE scale a suitable choice for measuring self-efficacy expectation because it is developed specifically for older adults and easily applicable to them, owing to its short length. With the aim of wide use of the SEE scale in the geriatric care and physical activity fields for Korean older adults, we evaluated the psychometric properties of the Korean version of the SEE (SEE-K) by using confirmatory factor analysis and Rasch analysis. Material and methods Participants and procedures Study participants were recruited from three community welfare centers for older adults in two districts of Seoul, South Korea, and one located close to Seoul. The inclusion criteria were being aged 65 years or older, community dwelling, and independent in activities of daily living. With the cooperation of personnel from the three community welfare centers, we visited the centers and announced this research study to older adults who were at the community welfare centers in August, 2013. Trained research assistants approached about 240 older adults, of which 215 consented to participate in the study. A signed informed consent form was procured from each participant before data collection. The questionnaires were completed either by participants themselves or, for participants unable to complete the questionnaires by hand, by research assistants using interviews. After completion of the questionnaire, a small gift was given to the participants. After removing incomplete questionnaires, 212 surveys were included in the final data analysis. This study was approved by the institutional review board of the College of Nursing at Yonsei University in Seoul, South Korea (No. 2013-0024). Measurements We collected information on participants’ age, sex, education, height, weight, perceived health status, and number of diseases using the questionnaire. Then, body mass index (BMI) was calculated from participants’ reported height and weight. The other measurements used in this study are described below. Self-Efficacy for Exercise The original English version of the SEE contains nine items measuring confidence in ability to exercise under certain conditions, for example, bad weather, emotions, and busy or stressful situations. Participants rated their confidence about maintaining exercise three times per week for 20 min for each condition, ranging from 0 (not confident) to 10 (very confident). A higher score represents better self-efficacy for exercise.17 After permission was received from the original developer of the SEE, the English version was translated into Korean (forward translation). Then, two nursing professors in Korea reviewed the translation accuracy. Finally, this version was back-translated into
English by a Korean-American, who is proficient in both Korean and English. Outcome Expectations for Exercise-2 The Korean version of the Outcome Expectations for Exercise-2 scale (OEE-2-K)21 was also used. This scale was derived from the OEE-2, which consists of 13 items, measuring nine positive and four negative outcome expectations for exercise.22 The scale uses a 5-point Likert scale, ranging from 1 (not at all) to 5 (absolutely agree). A higher score indicates higher expectations for the outcomes of exercise. The Cronbach’s alpha in the present study was 0.834. Quality of life The EuroQol group’s 3-level version of the EQ-5D was used to measure quality of life (QoL).23 This tool was developed as a generic core index of health for use in health care evaluation worldwide, including in South Korea. Using a questionnaire format, this tool requires each participant to rate five dimensions of QoLdmobility (M), self-care (SC), usual activities (UA), pain/discomfort (PD), and anxiety/depression (AD)don a three-point scale of severity, as follows: no problem (level 1), some or moderate problem (level 2), and serious problem (level 3). The EQ-5D health states can be converted into a single summary index by applying a formula that essentially attaches values (i.e., weights) to each of the levels in each dimension. The following formula has been used in the Korean population to calculate the EQ-5D-3L index24:
EQ 5D index ¼ 1 ð0:050 þ 0:096 M2 þ 0:418 M3 þ 0:046 SC2 þ 0:136 SC3 þ 0:051 UA2 þ 0:208 UA3 þ 0:037 PD2 þ 0:151 PD3 þ 0:043 AD2 þ 0:158 AD3 þ 0:050 N3Þ For example, “M2” represents someone who gave a level 2 (some or moderate problem) for mobility and “N3” indicates whether there is any dimension rated on a level 3. Thus, the EQ-5D index value for an individual who gave ratings of 3, 2, 3, 2, and 2 for the five dimensions would be calculated as follows: 1 (0.050 þ 0.418 þ 0.046 þ 0.208 þ 0.037 þ 0.043 þ 0.050) ¼ 0.148.24 Physical activity To assess physical activity, the International Physical Activity Questionnaire-short form (IPAQ) was used. This consists of assessments of vigorous- and moderate-intensity activities and walking. We converted minutes and number of days spent engaging in those three types of physical activity into total MET minutes per week. Then, we categorized participants into three groupsdlow, moderate, and high physical activitydbased on the total MET minutes per week. The IPAQ data processing guideline25 was used for calculation of MET minutes per week and grouping. Data analysis Stata version 12 was used for descriptive and inferential statistics and confirmatory factor analysis, and Winsteps 3.81 was used for Rasch model testing. Validity Construct validity was tested using confirmatory factor analysis with the maximum likelihood estimation method. To estimate model fit, the chi-square statistic, chi-square/degrees of freedom ratio, comparative fit index (CFI), TuckereLewis index (TLI), and
M. Choi et al. / Geriatric Nursing 36 (2015) 301e305
root mean-square error of approximation (RMSEA) were used. Although a p-value of greater than 0.05 on the chi-square goodness-of-fit test indicates a good fit, the chi-square test tends be very sensitive to sample size, and for large samples, the chi-square will almost always be significant. Thus, the chi-square/degrees of freedom ratio was also calculated; a value of less than 2 indicates a good model fit.26 The CFI, TLI, and RMSEA are less affected by sample size, and CFI and TLI values close to 0.95 and RMSEA values close to 0.06 are considered to indicate good fit.27 The concurrent validity, a subtype of criterion-related validity, was determined by testing whether SEE-K scores were correlated with OEE-2-K and EQ-5D Index scores and physical activity levels. Previous studies identified positive correlations between the SEE and OEE12,28 and another study revealed positive correlations between SEE and QoL and between SEE and physical activity.29 Thus, we hypothesized that the OEE-2-K, EQ-5D, and the intensity of physical activity were positively related to the SEE-K. We tested the relations of the OEE-2-K and EQ-5D with the SEE-K using Pearson’s correlation coefficients. For intensity of physical activity, we used analysis of variance (ANOVA) to compare the physical activity groups to determine if SEE-K scores increase with physical activity. Rasch analysis, which is based on item response theory, was performed to test the validity of the SEE-K by considering it as a unidimensional model. The model fit of the Rasch measurement model was tested using INFIT and OUTFIT item fit statistics. Acceptable limits of INFIT and OUTFIT statistics are between 0.6 and 1.4. INFIT and OUTFIT values of less than 0.6 indicate that there are likely redundant items, leading to overfitting. In contrast, INFIT and OUTFIT values of greater than 1.4 indicate that items may define the construct of interest differently from the rest of the items in the scale, implying underfitting.30,31 We also built a person-item map to display the item difficulty distribution of the SEE-K against individuals’ ability level; the item difficulty is displayed on the right side of the vertical line and individuals’ ability level on the left side on the same scale, which is known as a logit scale.31 When persons are placed outside of the most difficult or easiest items on the map, additional items may be needed to differentiate those who have higher or lower scores.31 Further details on Rasch analysis can be found elsewhere, both in relation to theoretical aspects31 and the nursing field.21,32,33 Reliability The internal consistency of the SEE-K was determined using Cronbach’s alpha, where 0.7 or greater was considered indicative of good internal consistency.34 From the confirmatory factor analysis, a coefficient of determination (R2) of 0.5 for each item and a composite R2 of 0.6 for the overall instrument are considered adequate indicators of reliability.35 Finally, a person separation index with reliability and an item separation index with reliability were also reported from the Rasch analysis. The person separation index is used to classify high and low performers. Likewise, the item separation index is used to discriminate items from one another according to the item difficulty hierarchy.31 Person and item separation indices of 1.5 are considered to have an acceptable level of separation, with 2.0 being a good level and 3.0 being an excellent level. The person or item reliability can be interpreted as Cronbach’s alphas.36 Results The general characteristics of the study participants are presented in Table 1. Of the 212 participants, the majority were women (n ¼ 144, 67.9%). The mean age was 76.02 5.80 years, ranging from 65 to 97 years. About two-thirds of participants were in the
303
Table 1 Participants’ general characteristics, SEE-K, OEE-2-K, EQ-5D, and IPAQ (n ¼ 212). Characteristics
Categories
n (%)
Sex
Male Female 65e74 75 25: Overweight or obese Good Fair Bad Elementary school or less Middle school High school College or higher
68 144 90 122 4 135 60 93 85 34 71 43 59 37
Age (yr) a
2
BMI (kg/m )
Perceived health status Educationa
SEE-K OEE-2-K EQ-5D Indexb IPAQc
(32.1) (67.9) (42.5) (57.5) (2.0) (67.8) (30.2) (43.9) (40.1) (16.0) (33.8) (20.5) (28.1) (17.6)
Mean SD
76.02 5.80
52.50 20.12 54.81 6.86 0.86 0.10 Low Moderate High
49 (23.1) 152 (71.7) 11 (5.2)
SEE-K: the Korean version of the Self-Efficacy for Exercise; OEE-2-K: the Korean version of the Outcome Expectations for Exercise-2; EQ-5D: EuroQoL-5 Dimensions; IPAQ: International Physical Activity Questionnaire-short form. a Missing cases excluded. b Calculated from the EQ-5D Index formula.24 c Categorized based on the “Guidelines for Data processing and Analysis of the International Physical Activity Questionnaire.”26
normal BMI range, 18.5e25 kg/m2, and most participants perceived their health status as fair or good. The means of the total scores on the SEE-K and OEE-2-K were 52.50 20.12 of a possible 90 and 54.81 6.86 of a possible 65, respectively, and the mean of EQ-5D index was 0.86 0.10. About 70% of older adults had participated in moderate level of physical activity. Validity The construct validity was tested with confirmatory factor analysis. Table 2 shows the path coefficients and R2 estimates for the SEE-K items. Path coefficients for all nine items were greater than 0.5 and statistically significant. The chi-square statistic was statistically significant (c2 ¼ 48.87, df ¼ 26, p ¼ 0.004). However, the chi-square/degrees of freedom ratio was 1.88, indicating a fair model fit. The CFI, TLI, and RMSEA demonstrated acceptable model fit with values of 0.959, 0.944, and 0.064, respectively. Table 2 Path coefficient, and coefficient of determination of each item of SEE-K from confirmatory factor analysis (n ¼ 212). Question items How confident are you right now that you could exercise three times per week for 20 min if 1 .the weather was bothering you? 2 .you were bored by the program or activity? 3 .you felt pain when exercising? 4 .you had to exercise alone? 5 .you did not enjoy it? 6 .you were too busy with other activities? 7 .you felt tired? 8 .you felt stressed? 9 .you felt depressed? Total
Path coefficients (l)
Reliability estimates (R2)
0.57 0.59
0.33 0.35
0.61 0.53 0.62 0.53
0.37 0.28 0.38 0.28
0.71 0.57 0.59 e
0.50 0.33 0.35 0.82
SEE-K: the Korean version of the Self-Efficacy for Exercise.
304
M. Choi et al. / Geriatric Nursing 36 (2015) 301e305
Table 3 Correlations among SEE-K, OEE-2-K, and EQ-5D (n ¼ 212). Variables
Table 5 Rasch model testing fit statistics and item difficulty.
EQ-5D Index
OEE-2-K
r (p) SEE-K EQ-5D Index
0.169 (0.014)*
0.384 (
