Qual Life Res DOI 10.1007/s11136-015-0972-0
Measurement invariance of English and French Health Education Impact Questionnaire (heiQ) empowerment scales validated for cancer Jennifer Brunet1 • Sophie Lauzier2,3 • H. Sharon Campbell4,5 • Lise Fillion6,10 Richard H. Osborne7 • Elizabeth Maunsell8,9,10
•
Accepted: 21 March 2015 Springer International Publishing Switzerland 2015
Abstract Purpose If measurement invariance (MI) is demonstrated for a scale completed by respondents from two different language groups, it means that the scale measures the same construct in the same way in both groups. We assessed MI of the French- and English-language versions of the five Health Education Impact Questionnaire (heiQ) empowerment scales validated for the cancer setting. Methods Data came from two cross-sectional studies of Canadian cancer survivors (704 English, 520 French). Single-group confirmatory factor analysis (CFA) was used to test whether the hypothesized factor structure of the French-language heiQ empowerment scales fit the data. Multi-group CFAs were conducted to assess different levels of MI conditions (configural, metric, scalar, strict, as well as MI of factor variances, covariances, and latent means) of the French- and English-language heiQ empowerment scales.
Results The correlated five-factor model showed good fit in both language groups (goodness-of-fit indices: CFI C .97; RMSEA B .07). Goodness-of-fit indices and tests of differences in fit between models supported MI of the five-factor model across the two language groups (DCFI B -.010 combined with DRMSEA B .015). Conclusions The French- and English-language heiQ empowerment scales measure the same five dimensions of empowerment in the same way across both language groups. Thus, any observed similarities or differences between French- and English-speaking respondents completing these scales are valid and reflect similarities or differences in empowerment across language groups, not measurement artifact. Consequently, heiQ empowerment data from English- and French-speaking respondents can be directly pooled or contrasted in data analyses.
& Elizabeth Maunsell [email protected]
6
Faculte´ des sciences infirmie`res, Universite´ Laval, Que´bec, Que´bec, QC, Canada
7
Public Health Innovation, Population Health Strategic Research Centre, Deakin University, Burwood Campus, 221 Burwood Highway, Melbourne, VIC 3125, Australia
8
De´partement de me´decine sociale et pre´ventive, Faculte´ de me´decine, Universite´ Laval, 2180 Chemin Sainte-Foy, Que´bec, QC G1K 7P4, Canada
9
Centre des maladies du sein Descheˆnes-Fabia, Hoˆpital du Saint-Sacrement, Que´bec, QC G1S 4L8, Canada
10
Axe Oncologie, Centre de recherche du Centre Hospitalier Universitaire de Que´bec, 1050 chemin Sainte-Foy, Que´bec, QC G1S 4L8, Canada
1
School of Human Kinetics, University of Ottawa, 125 University, Montpetit Hall, Room 339, Ottawa, ON K1N 6N5, Canada
2
Faculty of Pharmacy, Universite´ Laval, Que´bec, QC G1K 7P4, Canada
3
Axe Sante´ des populations et pratiques optimales en sante´, Centre de recherche du Centre Hospitalier Universitaire de Que´bec, Que´bec, QC G1S 4L8, Canada
4
Faculty of Applied Health Sciences, School of Public Health and Health Systems, University of Waterloo, 200 University Avenue West, Waterloo, ON N2L 3G1, Canada
5
Propel Centre for Population Health Impact, University of Waterloo, 200 University Avenue West, Waterloo, ON N2L 3G1, Canada
123
Qual Life Res
Keywords Empowerment Cancer Language invariance Validation studies Outcome assessment (health care) Translation
Introduction In the cancer context, health-related empowerment has been defined as an individual’s feelings of being able to manage the challenges he or she experiences and of having a sense of control over his or her life [1]. An empowered individual with cancer may better understand and participate in their care, mobilize resources, and take actions that can reduce anxiety, increase use of strategies for dealing with cancer, and improve self-management and quality of life [2–5]. Empowering the person with cancer at different stages of the disease trajectory is often a goal of support services, and patient education and self-management programs [6, 7]. Thus, it is important to have valid measures to assess the effectiveness of such services and programs provided at different points in the disease trajectory. However, empowerment has rarely been assessed as an outcome because few validated questionnaires measuring health-related empowerment exist, particularly in the cancer setting. For these reasons, we began a process that resulted in the adaptation and validation of an existing measure for use in cancer research and program evaluations, in both English and French. As previously described [1], this process involved several steps and considerations. Briefly, to identify candidate instruments, we reviewed published literature in Medline, CINHAL, and PsychINFO available up to January 2007 to identify studies reporting on validations of instruments designed to measure generic and disease-specific empowerment and empowerment-related constructs, including self-efficacy, coping, and perceived control. Having adopted the National Cancer Institute’s definition of a cancer survivor which considers the individual ‘‘to be a survivor from the time of diagnosis until the end of life’’ [8], we looked for instruments that could be used in different cancer-related situations, including active curative or palliative treatment, and post-treatment survivorship. We considered it essential that instrument items to be simple, and that the number of items per construct be low, to facilitate use in a variety of evaluation contexts. Candidate instruments also had to fulfill a number of other criteria: good psychometric properties; good face and content validity as judged by cancer experts from the Canadian Cancer Society (CCS) who assessed them for relevance, comprehensiveness, and applicability when assessing survivor support services and programs; and confirmation of their understandability and ease of use based on results of
123
21 debriefings conducted with cancer survivors who had previously completed the candidate instruments [1, 2]. Synthesis of the information from this process led to our identifying and retaining the Health Education Impact Questionnaire (heiQ) [9] as a candidate instrument for measuring empowerment in the cancer setting. The heiQ was developed in Australia, in English, using a validitydriven approach to achieve full representation of the construct of self-management capability among people with chronic diseases and thus comprehensively measure the effects of health education programs on self-management among such populations [10]. Consequently, eight scales, that can be used independently, arose from a grounded consultation with survivors and stakeholders. Since its development, the heiQ has become widely used for quality and monitoring, evaluation, and as an outcomes framework in systematic reviews of health education and empowerment programs across different disease and language groups [11–15]. However, until work by our group [1], the heiQ had not been specifically validated in the cancer setting. We retained five of the eight scales from the original heiQ to assess key dimensions of empowerment in the cancer setting: Social integration and support, Health service navigation, Constructive attitudes and approaches, Skill and technique acquisition, and Emotional distress. From a conceptual viewpoint, two main considerations guided our selection of these scales as key for quantitatively assessing empowerment in the cancer context. First, the scales had obvious face validity with respect to our basic definition of empowerment for the cancer setting— namely feelings of being better able to manage the multifaceted challenges of the cancer experience and of gaining (or re-gaining) some control over this experience [1]. Secondly, each of these scales contained a proportion of items that incorporated explicit reference to the person’s health, health problems, or healthcare team. We thought such references were necessary for measuring health-related empowerment and would also help demonstrate the potential empowering effects of cancer-related programs. Following a few small modifications for the cancer context (which were also adopted for the heiQ version 3, in all contexts), we assessed the reliability and validity of scores on these five scales among English-speaking Canadians with cancer [1] and found evidence that they were robust measures of health-related empowerment in the cancer setting. In collaboration with the heiQ developer (RHO), we also translated these five empowerment scales (as well as the three remaining scales of the original English-language heiQ) into French. These five translated scales were then used in a study conducted with Frenchspeaking cancer survivors living in Canada.
Qual Life Res
The purpose of the present study was to (1) test the factor structure of the five French-language heiQ empowerment scales, and (2) assess the measurement invariance of the English- and French-language heiQ empowerment scales in cancer survivors living in Canada. The psychometric properties for one version of a questionnaire, in this case the English-language heiQ empowerment scales, may not necessarily generalize to other versions. Thus, it is important to provide evidence of validity for scores on the French-language heiQ empowerment scales. Specifically, the hypothesized factor structure of the French-language heiQ empowerment scales needs to be confirmed to ensure that together the items assess the same five dimensions of empowerment as was reported for the same English-language heiQ scales validated in the cancer setting [1]. As well, for contexts where the two language versions of the heiQ empowerment scales are likely to be administered to Englishand French-speaking cancer survivors within the same study or service/program evaluation, it is also important to determine whether scores obtained from both versions can be considered directly comparable, or in other words, whether they demonstrate measurement invariance (see [16], [17], for technical reviews). To the extent that scores on the Englishand French-language heiQ empowerment scales show measurement invariance, researchers can say that the same underlying dimensions of empowerment are being measured in the same way by the English- and French-language versions. If they are, researchers can validly attribute observed similarities or differences by language group to real similarities or differences in empowerment per se, and not to measurement artifact [17]. As an important corollary, demonstrated measurement invariance would also mean that heiQ empowerment data from English- and French-speaking respondents could be pooled for analyses (see [18] for technical review). This is important information for researchers who might want to use these empowerment scales for service and program evaluation in a number of cancer settings, and for government policy makers and funders who might use such research to assess service and program value.
Methods Participants Participants for the assessment of measurement invariance of the English- and French-language heiQ empowerment scales in the cancer setting came from two separate studies. Participants in the validation study of the five Englishlanguage heiQ empowerment scales in a Canadian cancer setting have been described previously [1] ). Briefly, they were English-speaking cancer survivors who were C18 years of age diagnosed with any type of cancer
(excluding in situ cancer, non-melanoma skin cancer, and neurological tumors) during the 3–27 months prior to February 2009. They were recruited through random sampling in a provincial population-based cancer registry, as well as from users of the CCS’s information and peer support programs from April to July 2009. Participants in the validation study for the translated French-language heiQ empowerment scales were French-speaking cancer survivors who were part of a study that aimed to adapt and validate instruments relevant to psychosocial and health service outcomes, including empowerment. Such outcomes had previously been identified as pertinent to the Navigation Framework by the Canadian Partnership against Cancer (CPAC) [19]. All of the French-speaking participants came from hospitals in the Quebec City area. In each participating hospital, the Archives department provided researchers with lists of all new cancer survivors C18 years of age treated for breast, lung, colorectal, or prostate cancer from September 2008 to March 2009. The ethics committees of the University of Waterloo, the University of Manitoba Bannatyne Campus, of four hospitals of the Centre hospitalier universitaire de Que´bec ˆ pital du Saint-Sacrement, Ho ˆ pital Saint-Franc¸ois (Ho ˆ pital Enfant-Je´sus, and Ho ˆ tel-Dieu de Que´d’Assise, Ho bec), and the Centre de sante´ et de services sociaux de la Vieille-Capitale approved the components of this study. All study procedures were in accordance with these institutions’ ethical standards and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants. Procedures Instrument translation We translated all eight heiQ scales into Canadian French using the standard heiQ translation protocol used by Hawkins and Osborne [20], which is a strict protocol meeting or exceeding internationally accepted translation standards [21, 22]. Specifically, professional translators who were native Canadian French speakers (n = 3) first independently translated the English scales into French. Second, a native English-speaking professional translator translated the French scales back into English. Third, the original and backtranslated scales were compared, and differences were discussed by an expert committee of five members in order to achieve consensus on equivalence. The committee members included researchers with extensive experience in psychosocial oncology research (EM, LF, SL), a native English speaker fully bilingual in French (EM), a professional translator who also has a background in psychosocial oncology research, and a bilingual research assistant. The
123
Qual Life Res
expert committee and the professional translators worked with the heiQ developer (RHO) to ensure every item was equivalent across language versions. When an item is equivalent, it means that a French- or English-speaking respondent (using their respective language version) would endorse the same response option if they had exactly the same characteristics related to the construct being measured by that item. The French-language heiQ empowerment scales were administered to six cancer survivors who were debriefed to assess the comprehensibility, relevance, and sensitivity of the items [23]. The translations of two items were slightly modified as a result of this process. The resulting translation of the eight heiQ scales is recognized by the copyright owner, Deakin University, as the official Canadian French translation of all eight heiQ scales. HeiQ scales retained for testing For this paper, we focused on the five heiQ scales that were validated in English [1]. In the English-language validation, we did not retain scales where no items were explicitly linked with the respondent’s health state or problem in some way (Positive and active engagement in life, Health-directed behavior), even if these might reflect health state indirectly, or scales we considered might not always be applicable to all moments in the cancer trajectory (Health-directed behavior). We did not retain the Selfmonitoring and insight scale because some items represented situations one would not necessarily want cancer patients to be self-managing, the example being health problems that might be a sign of recurrence. We also felt that self-monitoring of health and symptoms, while possibly desirable in some chronic conditions, could also reflect fears of recurrence in cancer patients [24]. In other words, in some survivors, high levels of self-monitoring might not necessarily reflect empowerment, but rather disease-related anxiety or anxiety about recurrence. Data collection All participants completed a self-administered questionnaire which included questions about demographic characteristics and the five heiQ empowerment scales, namely Social integration and support (five items), Health service navigation (five items), Constructive attitudes and approaches (five items), Skill and technique acquisition (four items), and Emotional distress (six items). Each heiQ item was scored on a 4-point Likert scale. Items for the Emotional distress scale were reversed scored. The mean score for each empowerment dimension was calculated as the average of its items, with a range of 1–4. Higher scores indicate higher levels of empowerment.
123
Data analysis Participants with 20 % or more of the items missing were excluded from the analyses. Multiple imputation (expectation–maximization algorithm; [25]) was used to estimate and replace missing observations when, overall, less than 20 % of the items were missing on the English- and French-language heiQ empowerment scales, respectively. Estimated missing values were rounded to the nearest discrete scale value to maintain the ordinal nature of the data. Descriptive statistics were then computed for the English- and Frenchlanguage heiQ empowerment scales separately. The different models were compared, and invariance at each step was established based on multiple criteria: (1) change in CFI (DCFI) is B-.010, and change in RMSEA (DRMSEA) is B.015 when each successive model is compared to the previous model [30], (2) Chi-square difference tests using the DIFFTEST function in Mplus to compare WLSMV Chi-square values across nested models are nonsignificant, and (3) CFI and RMSEA values remain C.90 and B.08, respectively [26]. Factor structure The procedures for validating the factor structure of the five English-language heiQ empowerment scales among cancer survivors and results are reported elsewhere [1]. In the present analyses, we validated the factor structure of the French heiQ empowerment scales using single-group confirmatory factor analysis (CFA) in MPlus 7.0 to determine whether the hypothesized latent variables loaded onto their respective items. The hypothesized five-factor model allowed for: (1) the factors to correlate; (2) each latent variable to load only onto its respective items (i.e., no cross-loadings); and (3) did not allow for correlated errors among the items. The factor loading for the first item of each latent variable was set to 1.0 to establish the metric of the latent variable. Owing to the ordered categorical nature of the data, the polychoric correlation matrices using means and variance-adjusted weight least squares (WLSMV) estimation were used since they yield unbiased goodness-offit indices when dealing with non-normal data. To assess model fit, we used multiple robust indices, namely the Comparative Fit Index (CFI) and the Root Mean Square Error of Approximation (RMSEA) with its 90 % confidence interval (CI) [26]. Values of C.90 for the CFI and values B.08 for the RMSEA indicate acceptable fit of the model [26]. The strength of the standardized factor loadings between each indicator and its corresponding latent variable was also examined. To assess internal consistency of scale scores, ordinal composite reliability coefficients [27] were calculated based on the results of the CFA.
Qual Life Res
Measurement invariance We then tested the measurement invariance of the English- and French-language heiQ empowerment scales using a multi-group CFA framework [28]. This consisted of testing different levels of invariance by comparing a series of nested models across the language groups [17]. This testing was guided by Millsap and YunTein’s [29] recommendations for categorical multi-group CFA in Mplus using theta parameterization. Testing measurement invariance involved a number of steps. First, we tested ‘‘configural invariance,’’ the weakest level of measurement invariance and a prerequisite to test for other levels of invariance, by estimating an unconstrained baseline model simultaneously across the language groups (Model 1). Configural invariance implies the same number of factors in both language groups and the same pattern of fixed and free parameters. Second, we tested ‘‘metric invariance’’ by constraining the factor loadings to be equal across the language groups (Model 2). This level of invariance indicates equal factor loadings across language groups which implies that respondents in the two language groups calibrate their measures in the same way. In other words, the values obtained for each item on the heiQ have the same meaning for the English- and French-speaking samples. Third, we tested ‘‘scalar invariance’’ (Model 3) by adding equality constraints on the item intercepts across the two language groups. This level of invariance infers that there is consistency between cross-language differences in means of the manifest and latent variables representing the different scale items and dimensions of empowerment, respectively. Fourth, we tested ‘‘error invariance’’ by adding equality constraints on the uniquenesses (i.e., errors associated with the items) across the languages by setting all uniquenesses to 1.0 across the two language groups (Model 4). This is seen as a strict test of invariance and is akin to testing invariance of the reliability associated with the items within each factor if the latent factor variances are equal. Error invariance implies that for a given item, the items’ unique variances are similar across language groups. Last, we tested additional levels of invariance by imposing equality constraints across the two language groups on factor variances (Model 5), factor covariances (Model 6), and latent factor means (Model 7) [17]. However, it should be noted that it is not necessary to have invariant factor variances, factor covariances, or error variances to conduct mean comparisons across two groups [18].
Results Overall, a small percentage from each group—3.7 % (27/ 731) and 1.3 % (7/527) of individuals from the English- and French-speaking groups, respectively—had missing data for five or more of the 25 items (C20 %) on the heiQ scales and were excluded from the analyses. Thus, results presented are
based on 704 English-speaking and 520 French-speaking adults diagnosed with cancer. The two language groups were generally similarly aged. However, there were more men, fewer employed participants and a substantially greater proportion of prostate cancers and only a very small proportion of breast cancers, in the French-speaking group compared to the English-speaking one (Table 1). Factor structure Based on the goodness-of-fitness indices of the singlegroup CFA, the data fit the five-factor model well for both the French-language [v2 (265) = 898.21, p \ .001, CFI = .975, RMSEA = .068 (90 % CI .063–.073)] and English-language heiQ empowerment scales for cancer [v2 (265) = 1100.28, p \ .001, CFI = .966, RMSEA = .067 (90 % CI .063–.071)]. These results are comparable to those found in the validation study of the five EnglishTable 1 Characteristics of 704 English- and 520 French-speaking participants included in the measurement invariance assessment English speakers N (%)
French speakers N (%)
Gender Men
292 (41.5)
41 (80.6)
Women
408 (58.0)
201 (19.4)
Missing
4 (.6)
Age (years) 18–44
64 (9.1)
6 (1.2)
45–54
123 (17.5)
60 (11.5)
55–64 65–74
221 (31.4) 170 (24.1)
163 (31.3) 196 (37.7)
C75
109 (15.5)
93 (17.9)
17 (2.4)
2 (.4)
High school or less
305 (43.4)
281 (54.0)
Collegial level or university
380 (54.0)
232 (44.6)
19 (2.7)
7 (1.3)
Missing Highest completed education
Missing Employed at the time of study Yes
390 (55.4)
105 (20.2)
No
300 (42.6)
410 (78.8)
14 (2.0)
5 (1.0)
Missing Type of cancer Breast
239 (33.9)
27 (5.2)
Prostate
106 (15.1)
293 (56.3)
Colorectal
87 (12.4)
153 (29.4)
Lung Lymphoma
28 (4.0) 52 (7.4)
47 (9.0) –
Other
160 (22.7)
–
More than one type
18 (2.6)
–
Missing
14 (2.0)
–
123
Qual Life Res Table 2 Standardized factor loadings and latent variable correlations of the five heiQ empowerment scales for the English- and French-language versions Items
Factor I E/F
Factor II E/F
Factor III E/F
Factor IV E/F
Factor V E/F
Factor I: Social integration and support If I need help, I have plenty of people I can rely on
.78/.84
I have enough friends who help me cope with my health… When I feel ill, my family and carers really understand…
.84/.85 .72/.74
Overall, I feel well looked after by friends or family
.89/.86
I get enough chances to talk about my health…
.84/.84
Factor II: Health services navigation I have very positive relationships with my healthcare…
.81/.86
I communicate very confidently with my doctorsa about…
.89/.83
I confidently give healthcare professionals the information…
.80/.82
I get my needs met from available healthcare resources…
.79/.88
I work in a team with my doctorsa and other healthcare…
.71/.85
Factor III: Constructive attitude and approaches I try not to let my health problems stop me from enjoying life
.82/.84
My health problems do not ruin my life
.86/.83
I feel I have a very good life even when I have health problems
.88/.89
I do not let my health problems control my life
.88/.94
If others can cope with problems like mine, I can too Factor IV: Skill and technique acquisition
.80/.85
I have effective ways to prevent my health problemsb…
.78/.81
I have a very good idea of how to manage my health problems
.80/.89
When I have health problemsb, I have skills that help me cope
.81/.93
I have a good understanding of equipment that could make my life easier
.66/.63
Factor V: Emotional distress (reverse-scored) I often worry about my health
.65/.62
My health problems make me very dissatisfied with my life
.90/.72
I often feel angry when I think about my health
.84/.82
I feel hopeless because of my health problems
.88/.86
I get upset when I think about my health
.89/.87
If I think about my health, I get depressed
.89/.94
E/F: English/French values. All values significant at p \ .001 ‘‘…’’ indicates truncated items (complete versions supplied with license) a
Change for cancer setting: ‘‘doctors’’ substituted for ‘‘doctor’’ in heiQ version 3.0
b
Change for cancer setting: ‘‘health problems’’ substituted for ‘‘symptoms’’ in heiQ version 3.0
language heiQ empowerment scales [1]. All standardized factor loadings were C.62 (Table 2). Ordinal composite reliability coefficients were all C.82, and scale mean scores and standard deviations (SD) ranged from 2.87 (SD = .64) to 3.40 (SD = .46) in both language groups (Table 3). Positive correlation coefficients of moderate to high magnitude were observed between the latent factors and were generally similar for English- and French-language heiQ empowerment scales (Table 4). This five-factor model was used in the subsequent analyses.
123
Measurement invariance The assumptions of measurement invariance were supported for the five tested French- and English-language heiQ empowerment scales, namely configural invariance (Model 1), metric invariance (Model 2), scalar invariance (Model 3), error invariance (Model 4), factor variances (Model 5), factor covariances (Model 6), and latent factor means (Model 7) (Table 5). Specifically, CFI and RMSEA values of each sequential model remained C.90 and B.08,
Qual Life Res Table 3 Reliability coefficients, score ranges, means, and SD for heiQ empowerment scales for English- (n = 704) and French-speaking (n = 520) participants Observed score ranges
Reliabilitya
English
French
English
French
English
French
Social integration and support
1–4
1–4
.91
.91
3.10 (.53)
3.26 (.49)
Health service navigation
1–4
1–4
.90
.93
3.17 (.48)
3.40 (.46)
Constructive attitudes and approaches Skill and technique acquisition
1–4 1–4
1–4 1–4
.92 .82
.94 .87
3.23 (.48) 2.96 (.43)
3.32 (.53) 3.13 (.48)
Emotional distress (reverse-scored)
1–4
1–4
.94
.92
2.87 (.64)
3.03 (.67)
Variables
Mean (SD)
Means and SD based on observed scores a
Composite reliability calculated from confirmatory factor analyses
Table 4 Latent variable correlations of the five heiQ empowerment scales comparing the English- and French-language versions Variables
1
2
1. Social integration and support
–
2. Health services navigation
E: .74 (.70–.78)
3
4
–
F: .82 (.78–.85) 3. Constructive attitude and approaches 4. Skill and technique acquisition 5. Emotional distress
E: .76 (.72–.79)
E: .66 (.61–.70)
F: .80 (.77–.84)
F: .75 (.71–.79)
–
E: .71 (.67–.77)
E: .76 (.72–.80)
E: .80 (.76–.84)
F: .77 (.73–.81)
F: .81 (.77–.85)
F: .83 (.79–.86)
E: .47 (.41–.54)
E: .37 (.30–.44)
E: .67 (.62–.72)
E: .49 (.42–.55)
F: .49 (.42–.56)
F: .39 (.32–.47)
F: .66 (.61–.71)
F: .48 (.42–.55)
–
E: Signifies English values. F: Signifies French values All values are significant at p \ .001. Values in brackets present the 95 % CI around the correlation Table 5 Fit indices for the analyses testing invariance of the English- and French-language versions Models
v2
df
Dv2
1. Configural invariance
1996.98**
530
2. Metric invariance
1931.69**
550
17.44
20
.972
.002
3. Scalar invariance
2055.23**
595
191.65*
45
.971
Ddf
–
CFI
DCFI
RMSEA 90 % CI
DRMSEA
.970
–
.067 (.064–.070)
–
.064 (.061–.067)
-.003
-.001
.063 (.060–.066)
-.001
4. Invariant uniquenesses
2194.07**
620
218.28**
25
.968
-.003
.064 (.061–.067)
.001
5. Invariant factor variances
2007.02**
625
15.77*
5
.972
.004
.060 (.057–.063)
-.004
6. Invariant factor covariances
1432.06**
635
32.34**
10
.984
.012
.045 (.042–.048)
-.015
7. Invariant latent means
1633.71**
640
104.96**
5
.980
-.004
.050 (.047–.053)
.005
No statistically significant differences were found between nested models based on both DCFI B -.010 and DRMSEA B .015 v2 Chi square, df degrees of freedom; Dv2 calculated using the DIFFTEST function in MPlus, RMSEA root mean square error of approximation, CFI comparative fit index, CI confidence interval * p \ .05; ** p \ .001
respectively. As well, DCFI and DRMSEA were B-.010 and D B .015, respectively.
Discussion Our findings from single-group confirmatory factor analysis (CFA), conducted using data from Canadian cancer survivors, indicate that the same five heiQ empowerment
scales that we validated for use among English-speaking cancer survivors [1] were also found for our translated French-language heiQ empowerment scales. Practically speaking, this means that the five heiQ empowerment constructs are shared in these two language groups. In addition, our findings from multi-group CFA testing measurement invariance provide strong evidence that in the cancer setting, the five French- and English-language heiQ empowerment scales measure these same five dimensions
123
Qual Life Res
in the same way. In fact, we found evidence of measurement invariance at all recommended levels, even at the strictest levels described by Vandenberg and Lance [17]. Taken together, these findings are fundamentally important because they show that these scales can be validly used to measure these empowerment dimensions among both English- and French-speaking cancer survivors. They also show that quantitative pooling of data or comparisons of heiQ empowerment scores between these two language groups are appropriate and will yield meaningful interpretations. In other words, any observed similarities or differences seen in levels of Social integration and support, Health service navigation, Constructive attitudes and approaches, Skill and technique acquisition, and Emotional distress for French- and English-speaking respondents would reflect true similarities or differences, not measurement artifact stemming from translation or emic cultural differences. We have assessed potential study limitations. First, the English-speaking participants came from a populationbased registry and community support programs, whereas the French-speaking participants were identified within the hospital setting at the time of a clinic visit. In fact, our results support the robustness of the scales even though data were gathered in different study settings and participants differed somewhat in distributions of gender and cancer characteristics. Many validation studies are predicated on this very idea, namely that the instrument will perform similarly in different populations if the underlying constructs have been appropriately captured. Moreover, the heiQ empowerment scales seem to perform well in groups with different diseases and conditions [14]. Second, we also note that the French-language heiQ empowerment scales were used among individuals diagnosed with lung, breast, prostate, or colorectal cancer, whereas the English-language heiQ empowerment scales were used among a group that included these same cancer diagnoses (65 %), as well as small proportions of other unspecified types of cancers, multiple cancers, or with a type of cancer that was not identified. However, the dimensions assessed by the heiQ empowerment scales are probably ones that are universally relevant to people facing cancer, its treatment and posttreatment survivorship, independent of cancer type. We cannot affirm with certainty that the French-language heiQ empowerment scales we tested, which were translated by and validated with Canadian French speakers, will perform as adequately among French speakers in other parts of the world. It is possible that words commonly used or colloquialisms used or interpretation of some heiQ items could vary somewhat in other French-speaking countries. Nonetheless, it is reassuring to note that the Canadian English versions of these scales perform similarly in terms of psychometrics to other English settings and that our
123
English- and French-language versions perform comparably. From a practical viewpoint, we know that it will not always be possible to perform extensive new validations in every language group where one would want to use these scales, so their robustness in our populations is reassuring and provides some general evidence to support using them in other French-speaking populations. Face/content validity assessment and cognitive debriefings in a new language group are both relatively low-cost methods that can provide general information about generalizability to the new group, and about whether further, more labor-intensive validation work is needed. We are unable to judge the degree of measurement invariance of the three other heiQ scales (Positive and active engagement in life, Health-directed activity, and Selfmonitoring and insight) because we did not administer them to our English-speaking group. For reasons described earlier, we purposefully limited our study with Englishspeaking cancer survivors to include only the five heiQ empowerment scales tested here [1]. However, we note that these three heiQ scales were translated into French by our group at the same time as the five empowerment scales using the same rigorous protocol and expert committee. Thus, researchers wanting to collect and pool data from English- and French-speaking cancer survivors on one or more of these other three heiQ scales (scales which have been demonstrated to have excellent psychometric properties in other settings [11, 14, 31]) could examine measurement invariance beforehand, if feasible, to ensure that group means are directly comparable. Otherwise, given the extensive efforts required in collecting data, researchers could use the scales with the five others and assess measurement invariance of these three scales before pooling their data and/or comparing scores across English- and French-speaking respondents.
Conclusion This study adds to the growing body of evidence about the robustness of the heiQ empowerment scales in general and in cancer setting specifically. Using rigorous statistical analyses, our findings provide evidence of the validity of scores on the five French- and English-language heiQ empowerment scales in the cancer setting. Our translation of the eight heiQ scales, including both the five scales measuring key aspects of generic empowerment in the cancer setting that are the object of the analyses presented here, as well as the three other scales not included in this analysis, has been accepted as the official Canadian French translation of the heiQ measurement system. Our findings mean that Canadian researchers, health service providers, and policy makers across Canada can use and reliably interpret these scales
Qual Life Res
assessing key dimensions of empowerment, an important outcome of oncology health services and programs. With growing emphasis on self-management in patient care in many settings and conditions, these empowerment scales will fill an important gap for meaningful evaluation. Acknowledgments This study was supported by grants to HSC from the Canadian Institutes for Health Research (CIHR) [# KAL82607], to EM from the National Cancer Institute of Canada (NCIC) [# 010498, #010499] via the Propel Centre for Population Health Impact, and to LF from the Canadian Partnership against Cancer (Cancer Journey Portfolio 2007–2012). This manuscript was prepared while JB was supported by a Canadian Cancer Society Career Development Award in Prevention, and SL by a Chercheur-boursier Award from the Fonds de recherche du Que´bec—Sante´ in partnership with the l’Institut National d’Excellence en Sante´ et en Services Sociaux (INESSS). This work was conducted while SL was supported by a Post-doctoral Fellowship Award from the CIHR-funded Strategic Training Initiative in Health Research (STIHR) Psychosocial Oncology Research Training program (PORT). RHO is funded in part through a National Health and Medical Research Council Senior Research Fellowship #APP1059122.
10.
11.
12.
13.
14.
Conflict of interest The authors declare that they have no conflicts of interest. 15.
References 16. 1. Maunsell, E., Lauzier, S., Brunet, J., Pelletier, S., Osborne, R. H., & Campbell, H. S. (2014). Health-related empowerment in cancer: Validity of scales from the Health Education Impact Questionnaire. Cancer, 120(20), 3228–3236. 2. McCorkle, R., Ercolano, E., Lazenby, M., Schulman-Green, D., Schilling, L. S., Lorig, K., et al. (2011). Self-management: Enabling and empowering patients living with cancer as a chronic illness. CA: A Cancer Journal for Clinicians, 61(1), 50–62. 3. Lauzier, S., Campbell, H. S., Livingston, P. M., & Maunsell, E. (2014). Indicators for evaluating cancer organizations’ support services: Performance and associations with empowerment. Cancer, 120(20), 3219–3227. 4. Lovell, M. R., Luckett, T., Boyle, F. M., Phillips, J., Agar, M., & Davidson, P. M. (2014). Patient education, coaching, and selfmanagement for cancer pain. Journal of Clinical Oncology, 32(16), 1712–1720. 5. Yun, Y. H., Sim, J. A., Jung, J. Y., Noh, D. Y., Lee, E. S., Kim, Y. W., et al. (2014). The association of self-leadership, health behaviors, and posttraumatic growth with health-related quality of life in patients with cancer. Psycho-Oncology, 23(12), 1423–1430. 6. Brady, T. J., Kruger, J., Helmick, C. G., Callahan, L. F., & Boutaugh, M. L. (2003). Intervention programs for arthritis and other rheumatic diseases. Health Education and Behavior, 30(1), 44–63. 7. Newman, S., Steed, L., & Mulligan, K. (2004). Self-management interventions for chronic illness. The Lancet, 364(9444), 1523–1537. 8. National Cancer Institute. Dictionary of cancer terms. www.can cer.gov/dictionary. Accessed 15 December 2014. 9. Osborne, R. H., Elsworth, G. R., & Whitfield, K. (2007). The Health Education Impact Questionnaire (heiQ): An outcomes and evaluation measure for patient education and self-management
17.
18.
19. 20.
21.
22.
23. 24.
25.
interventions for people with chronic conditions. Patient Education and Counseling, 66(2), 192–201. Buchbinder, R., Batterham, R., Elsworth, G., Dionne, C. E., Irvin, E., & Osborne, R. H. (2011). A validity-driven approach to the understanding of the personal and societal burden of low back pain: Development of a conceptual and measurement model. Arthritis Research and Therapy, 13(5), R152. Epstein, J., Osborne, R. H., Elsworth, G. R., Beaton, D. E., & Guillemin, F. (2015). Cross-cultural adaptation of the heiQ questionnaire: Experimental study showed expert committee, not back-translation, added value. Journal of Clinical Epidemiology, 68(4), 360–369. Kroon, F. P. B., van Der Burg, L. R. A., Buchbinder, R., Osborne, R. H., Johnston, R. V., & Pitt, V. (2014). Self-management education programmes for osteoarthritis (review). Cochrane database of systematic reviews, issue 1. Art. no.: CD008963. doi:10.1002/14651858.CD008963.pub2. Osborne, R. H., Batterham, R., & Livingston, J. (2011). The quality, impact and implementation of chronic disease selfmanagement across settings: The international experience of the Health Education Impact Questionnaire (heiQ) quality monitoring system. Nursing Clinics of North America, 46(2), 255–270. Schuler, M., Musekamp, G., Bengel, J., Nolte, S., Osborne, R. H., & Faller, H. (2014). Measurement invariance across chronic conditions: A systematic review and an empirical investigation of the Health Education Impact Questionnaire (heiQTM). Health and Quality of Life Outcomes, 12(56). doi:10.1186/1477-7525-12-56. Wahl, A. K., Langeland, E., Larsen, M. H., Robinson, H. S., Osborne, R. H., & Krogstad, A. L. (2015). Positive changes in self-management and disease severity following climate therapy in people with psoriasis. Acta Dermato-Venereologica, 95(2), 317–321. Byrne, B. M., & Watkins, D. (2003). The issue of measurement invariance revisited. Journal of Cross-Cultural Psychology, 34(2), 155–175. Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3(1), 4–70. Steenkamp, J. B. E., & Baumgartner, H. (1998). Assessing measurement invariance in cross-national consumer research. Journal of Consumer Research, 25(1), 78–107. Fillion, S., & Cook, S. (2012). Final report: Navigation Project CPAC and CIHR April 2007 to March 2012. Unpublished report. Hawkins, M., & Osborne, R. (2007). Questionnaire translation and cultural adaptation procedure. Version 1.0. University of Melbourne, Australia, p. 9. Hambleton, R. K., & Kanjee, A. (1995). Increasing the validity of cross-cultural assessments: Use of improved methods for test adaptations. European Journal of Psychological Assessment, 11(3), 147–157. Vallerand, R. J. (1989). Vers une me´thodologie de validation trans-culturelle de questionnaires psychologiques: Implications pour la recherche en langue franc¸aise. Psychologie Canadienne, 30(4), 662–680. Willis, G. B. (1999, August). Cognitive interviewing: A ‘‘how to’’ guide. In Meeting of the American statistical association. Kornblith, A. B. (1998). Psychosocial adaptation of cancer survivors. In J. C. Holland, W. Breitbart, P. B. Jacobsen, M. S. Lederberg, M. Loscalzo, M. J. Massie, & R. McCorkle (Eds.), Psycho-oncology (pp. 223–241). New York: Oxford University Press. Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977). Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society: Series B (Methodological), 39(1), 1–38.
123
Qual Life Res 26. Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. 27. Diamantopoulos, A., & Siguaw, J. (2000). Introducing LISREL. London: SAGE Publication Ltd. 28. Byrne, B. M., Shavelson, R. J., & Muthe´n, B. (1989). Testing for the equivalence of factor covariance and mean structures: The issue of partial measurement invariance. Psychological Bulletin, 105(3), 456–466. 29. Millsap, R. E., & Yun-Tein, J. (2004). Assessing factorial invariance in ordered-categorical measures. Multivariate Behavioral Research, 39(3), 479–515.
123
30. Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504. 31. Morita, R., Arakida, M., Osborne, R. H., Nolte, N., Elsworth, G. R., & Mikami, H. (2013). Adaptation and validation of the Japanese version of the Health Education Impact Questionnaire (heiQ) for the evaluation of self-management education interventions. Japan Journal of Nursing Science, 10(2), 255–266.
Measurement invariance of English and French Health Education Impact Questionnaire (heiQ) empowerment scales validated for cancer Jennifer Brunet1 • Sophie Lauzier2,3 • H. Sharon Campbell4,5 • Lise Fillion6,10 Richard H. Osborne7 • Elizabeth Maunsell8,9,10
•
Accepted: 21 March 2015 Springer International Publishing Switzerland 2015
Abstract Purpose If measurement invariance (MI) is demonstrated for a scale completed by respondents from two different language groups, it means that the scale measures the same construct in the same way in both groups. We assessed MI of the French- and English-language versions of the five Health Education Impact Questionnaire (heiQ) empowerment scales validated for the cancer setting. Methods Data came from two cross-sectional studies of Canadian cancer survivors (704 English, 520 French). Single-group confirmatory factor analysis (CFA) was used to test whether the hypothesized factor structure of the French-language heiQ empowerment scales fit the data. Multi-group CFAs were conducted to assess different levels of MI conditions (configural, metric, scalar, strict, as well as MI of factor variances, covariances, and latent means) of the French- and English-language heiQ empowerment scales.
Results The correlated five-factor model showed good fit in both language groups (goodness-of-fit indices: CFI C .97; RMSEA B .07). Goodness-of-fit indices and tests of differences in fit between models supported MI of the five-factor model across the two language groups (DCFI B -.010 combined with DRMSEA B .015). Conclusions The French- and English-language heiQ empowerment scales measure the same five dimensions of empowerment in the same way across both language groups. Thus, any observed similarities or differences between French- and English-speaking respondents completing these scales are valid and reflect similarities or differences in empowerment across language groups, not measurement artifact. Consequently, heiQ empowerment data from English- and French-speaking respondents can be directly pooled or contrasted in data analyses.
& Elizabeth Maunsell [email protected]
6
Faculte´ des sciences infirmie`res, Universite´ Laval, Que´bec, Que´bec, QC, Canada
7
Public Health Innovation, Population Health Strategic Research Centre, Deakin University, Burwood Campus, 221 Burwood Highway, Melbourne, VIC 3125, Australia
8
De´partement de me´decine sociale et pre´ventive, Faculte´ de me´decine, Universite´ Laval, 2180 Chemin Sainte-Foy, Que´bec, QC G1K 7P4, Canada
9
Centre des maladies du sein Descheˆnes-Fabia, Hoˆpital du Saint-Sacrement, Que´bec, QC G1S 4L8, Canada
10
Axe Oncologie, Centre de recherche du Centre Hospitalier Universitaire de Que´bec, 1050 chemin Sainte-Foy, Que´bec, QC G1S 4L8, Canada
1
School of Human Kinetics, University of Ottawa, 125 University, Montpetit Hall, Room 339, Ottawa, ON K1N 6N5, Canada
2
Faculty of Pharmacy, Universite´ Laval, Que´bec, QC G1K 7P4, Canada
3
Axe Sante´ des populations et pratiques optimales en sante´, Centre de recherche du Centre Hospitalier Universitaire de Que´bec, Que´bec, QC G1S 4L8, Canada
4
Faculty of Applied Health Sciences, School of Public Health and Health Systems, University of Waterloo, 200 University Avenue West, Waterloo, ON N2L 3G1, Canada
5
Propel Centre for Population Health Impact, University of Waterloo, 200 University Avenue West, Waterloo, ON N2L 3G1, Canada
123
Qual Life Res
Keywords Empowerment Cancer Language invariance Validation studies Outcome assessment (health care) Translation
Introduction In the cancer context, health-related empowerment has been defined as an individual’s feelings of being able to manage the challenges he or she experiences and of having a sense of control over his or her life [1]. An empowered individual with cancer may better understand and participate in their care, mobilize resources, and take actions that can reduce anxiety, increase use of strategies for dealing with cancer, and improve self-management and quality of life [2–5]. Empowering the person with cancer at different stages of the disease trajectory is often a goal of support services, and patient education and self-management programs [6, 7]. Thus, it is important to have valid measures to assess the effectiveness of such services and programs provided at different points in the disease trajectory. However, empowerment has rarely been assessed as an outcome because few validated questionnaires measuring health-related empowerment exist, particularly in the cancer setting. For these reasons, we began a process that resulted in the adaptation and validation of an existing measure for use in cancer research and program evaluations, in both English and French. As previously described [1], this process involved several steps and considerations. Briefly, to identify candidate instruments, we reviewed published literature in Medline, CINHAL, and PsychINFO available up to January 2007 to identify studies reporting on validations of instruments designed to measure generic and disease-specific empowerment and empowerment-related constructs, including self-efficacy, coping, and perceived control. Having adopted the National Cancer Institute’s definition of a cancer survivor which considers the individual ‘‘to be a survivor from the time of diagnosis until the end of life’’ [8], we looked for instruments that could be used in different cancer-related situations, including active curative or palliative treatment, and post-treatment survivorship. We considered it essential that instrument items to be simple, and that the number of items per construct be low, to facilitate use in a variety of evaluation contexts. Candidate instruments also had to fulfill a number of other criteria: good psychometric properties; good face and content validity as judged by cancer experts from the Canadian Cancer Society (CCS) who assessed them for relevance, comprehensiveness, and applicability when assessing survivor support services and programs; and confirmation of their understandability and ease of use based on results of
123
21 debriefings conducted with cancer survivors who had previously completed the candidate instruments [1, 2]. Synthesis of the information from this process led to our identifying and retaining the Health Education Impact Questionnaire (heiQ) [9] as a candidate instrument for measuring empowerment in the cancer setting. The heiQ was developed in Australia, in English, using a validitydriven approach to achieve full representation of the construct of self-management capability among people with chronic diseases and thus comprehensively measure the effects of health education programs on self-management among such populations [10]. Consequently, eight scales, that can be used independently, arose from a grounded consultation with survivors and stakeholders. Since its development, the heiQ has become widely used for quality and monitoring, evaluation, and as an outcomes framework in systematic reviews of health education and empowerment programs across different disease and language groups [11–15]. However, until work by our group [1], the heiQ had not been specifically validated in the cancer setting. We retained five of the eight scales from the original heiQ to assess key dimensions of empowerment in the cancer setting: Social integration and support, Health service navigation, Constructive attitudes and approaches, Skill and technique acquisition, and Emotional distress. From a conceptual viewpoint, two main considerations guided our selection of these scales as key for quantitatively assessing empowerment in the cancer context. First, the scales had obvious face validity with respect to our basic definition of empowerment for the cancer setting— namely feelings of being better able to manage the multifaceted challenges of the cancer experience and of gaining (or re-gaining) some control over this experience [1]. Secondly, each of these scales contained a proportion of items that incorporated explicit reference to the person’s health, health problems, or healthcare team. We thought such references were necessary for measuring health-related empowerment and would also help demonstrate the potential empowering effects of cancer-related programs. Following a few small modifications for the cancer context (which were also adopted for the heiQ version 3, in all contexts), we assessed the reliability and validity of scores on these five scales among English-speaking Canadians with cancer [1] and found evidence that they were robust measures of health-related empowerment in the cancer setting. In collaboration with the heiQ developer (RHO), we also translated these five empowerment scales (as well as the three remaining scales of the original English-language heiQ) into French. These five translated scales were then used in a study conducted with Frenchspeaking cancer survivors living in Canada.
Qual Life Res
The purpose of the present study was to (1) test the factor structure of the five French-language heiQ empowerment scales, and (2) assess the measurement invariance of the English- and French-language heiQ empowerment scales in cancer survivors living in Canada. The psychometric properties for one version of a questionnaire, in this case the English-language heiQ empowerment scales, may not necessarily generalize to other versions. Thus, it is important to provide evidence of validity for scores on the French-language heiQ empowerment scales. Specifically, the hypothesized factor structure of the French-language heiQ empowerment scales needs to be confirmed to ensure that together the items assess the same five dimensions of empowerment as was reported for the same English-language heiQ scales validated in the cancer setting [1]. As well, for contexts where the two language versions of the heiQ empowerment scales are likely to be administered to Englishand French-speaking cancer survivors within the same study or service/program evaluation, it is also important to determine whether scores obtained from both versions can be considered directly comparable, or in other words, whether they demonstrate measurement invariance (see [16], [17], for technical reviews). To the extent that scores on the Englishand French-language heiQ empowerment scales show measurement invariance, researchers can say that the same underlying dimensions of empowerment are being measured in the same way by the English- and French-language versions. If they are, researchers can validly attribute observed similarities or differences by language group to real similarities or differences in empowerment per se, and not to measurement artifact [17]. As an important corollary, demonstrated measurement invariance would also mean that heiQ empowerment data from English- and French-speaking respondents could be pooled for analyses (see [18] for technical review). This is important information for researchers who might want to use these empowerment scales for service and program evaluation in a number of cancer settings, and for government policy makers and funders who might use such research to assess service and program value.
Methods Participants Participants for the assessment of measurement invariance of the English- and French-language heiQ empowerment scales in the cancer setting came from two separate studies. Participants in the validation study of the five Englishlanguage heiQ empowerment scales in a Canadian cancer setting have been described previously [1] ). Briefly, they were English-speaking cancer survivors who were C18 years of age diagnosed with any type of cancer
(excluding in situ cancer, non-melanoma skin cancer, and neurological tumors) during the 3–27 months prior to February 2009. They were recruited through random sampling in a provincial population-based cancer registry, as well as from users of the CCS’s information and peer support programs from April to July 2009. Participants in the validation study for the translated French-language heiQ empowerment scales were French-speaking cancer survivors who were part of a study that aimed to adapt and validate instruments relevant to psychosocial and health service outcomes, including empowerment. Such outcomes had previously been identified as pertinent to the Navigation Framework by the Canadian Partnership against Cancer (CPAC) [19]. All of the French-speaking participants came from hospitals in the Quebec City area. In each participating hospital, the Archives department provided researchers with lists of all new cancer survivors C18 years of age treated for breast, lung, colorectal, or prostate cancer from September 2008 to March 2009. The ethics committees of the University of Waterloo, the University of Manitoba Bannatyne Campus, of four hospitals of the Centre hospitalier universitaire de Que´bec ˆ pital du Saint-Sacrement, Ho ˆ pital Saint-Franc¸ois (Ho ˆ pital Enfant-Je´sus, and Ho ˆ tel-Dieu de Que´d’Assise, Ho bec), and the Centre de sante´ et de services sociaux de la Vieille-Capitale approved the components of this study. All study procedures were in accordance with these institutions’ ethical standards and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. Informed consent was obtained from all individual participants. Procedures Instrument translation We translated all eight heiQ scales into Canadian French using the standard heiQ translation protocol used by Hawkins and Osborne [20], which is a strict protocol meeting or exceeding internationally accepted translation standards [21, 22]. Specifically, professional translators who were native Canadian French speakers (n = 3) first independently translated the English scales into French. Second, a native English-speaking professional translator translated the French scales back into English. Third, the original and backtranslated scales were compared, and differences were discussed by an expert committee of five members in order to achieve consensus on equivalence. The committee members included researchers with extensive experience in psychosocial oncology research (EM, LF, SL), a native English speaker fully bilingual in French (EM), a professional translator who also has a background in psychosocial oncology research, and a bilingual research assistant. The
123
Qual Life Res
expert committee and the professional translators worked with the heiQ developer (RHO) to ensure every item was equivalent across language versions. When an item is equivalent, it means that a French- or English-speaking respondent (using their respective language version) would endorse the same response option if they had exactly the same characteristics related to the construct being measured by that item. The French-language heiQ empowerment scales were administered to six cancer survivors who were debriefed to assess the comprehensibility, relevance, and sensitivity of the items [23]. The translations of two items were slightly modified as a result of this process. The resulting translation of the eight heiQ scales is recognized by the copyright owner, Deakin University, as the official Canadian French translation of all eight heiQ scales. HeiQ scales retained for testing For this paper, we focused on the five heiQ scales that were validated in English [1]. In the English-language validation, we did not retain scales where no items were explicitly linked with the respondent’s health state or problem in some way (Positive and active engagement in life, Health-directed behavior), even if these might reflect health state indirectly, or scales we considered might not always be applicable to all moments in the cancer trajectory (Health-directed behavior). We did not retain the Selfmonitoring and insight scale because some items represented situations one would not necessarily want cancer patients to be self-managing, the example being health problems that might be a sign of recurrence. We also felt that self-monitoring of health and symptoms, while possibly desirable in some chronic conditions, could also reflect fears of recurrence in cancer patients [24]. In other words, in some survivors, high levels of self-monitoring might not necessarily reflect empowerment, but rather disease-related anxiety or anxiety about recurrence. Data collection All participants completed a self-administered questionnaire which included questions about demographic characteristics and the five heiQ empowerment scales, namely Social integration and support (five items), Health service navigation (five items), Constructive attitudes and approaches (five items), Skill and technique acquisition (four items), and Emotional distress (six items). Each heiQ item was scored on a 4-point Likert scale. Items for the Emotional distress scale were reversed scored. The mean score for each empowerment dimension was calculated as the average of its items, with a range of 1–4. Higher scores indicate higher levels of empowerment.
123
Data analysis Participants with 20 % or more of the items missing were excluded from the analyses. Multiple imputation (expectation–maximization algorithm; [25]) was used to estimate and replace missing observations when, overall, less than 20 % of the items were missing on the English- and French-language heiQ empowerment scales, respectively. Estimated missing values were rounded to the nearest discrete scale value to maintain the ordinal nature of the data. Descriptive statistics were then computed for the English- and Frenchlanguage heiQ empowerment scales separately. The different models were compared, and invariance at each step was established based on multiple criteria: (1) change in CFI (DCFI) is B-.010, and change in RMSEA (DRMSEA) is B.015 when each successive model is compared to the previous model [30], (2) Chi-square difference tests using the DIFFTEST function in Mplus to compare WLSMV Chi-square values across nested models are nonsignificant, and (3) CFI and RMSEA values remain C.90 and B.08, respectively [26]. Factor structure The procedures for validating the factor structure of the five English-language heiQ empowerment scales among cancer survivors and results are reported elsewhere [1]. In the present analyses, we validated the factor structure of the French heiQ empowerment scales using single-group confirmatory factor analysis (CFA) in MPlus 7.0 to determine whether the hypothesized latent variables loaded onto their respective items. The hypothesized five-factor model allowed for: (1) the factors to correlate; (2) each latent variable to load only onto its respective items (i.e., no cross-loadings); and (3) did not allow for correlated errors among the items. The factor loading for the first item of each latent variable was set to 1.0 to establish the metric of the latent variable. Owing to the ordered categorical nature of the data, the polychoric correlation matrices using means and variance-adjusted weight least squares (WLSMV) estimation were used since they yield unbiased goodness-offit indices when dealing with non-normal data. To assess model fit, we used multiple robust indices, namely the Comparative Fit Index (CFI) and the Root Mean Square Error of Approximation (RMSEA) with its 90 % confidence interval (CI) [26]. Values of C.90 for the CFI and values B.08 for the RMSEA indicate acceptable fit of the model [26]. The strength of the standardized factor loadings between each indicator and its corresponding latent variable was also examined. To assess internal consistency of scale scores, ordinal composite reliability coefficients [27] were calculated based on the results of the CFA.
Qual Life Res
Measurement invariance We then tested the measurement invariance of the English- and French-language heiQ empowerment scales using a multi-group CFA framework [28]. This consisted of testing different levels of invariance by comparing a series of nested models across the language groups [17]. This testing was guided by Millsap and YunTein’s [29] recommendations for categorical multi-group CFA in Mplus using theta parameterization. Testing measurement invariance involved a number of steps. First, we tested ‘‘configural invariance,’’ the weakest level of measurement invariance and a prerequisite to test for other levels of invariance, by estimating an unconstrained baseline model simultaneously across the language groups (Model 1). Configural invariance implies the same number of factors in both language groups and the same pattern of fixed and free parameters. Second, we tested ‘‘metric invariance’’ by constraining the factor loadings to be equal across the language groups (Model 2). This level of invariance indicates equal factor loadings across language groups which implies that respondents in the two language groups calibrate their measures in the same way. In other words, the values obtained for each item on the heiQ have the same meaning for the English- and French-speaking samples. Third, we tested ‘‘scalar invariance’’ (Model 3) by adding equality constraints on the item intercepts across the two language groups. This level of invariance infers that there is consistency between cross-language differences in means of the manifest and latent variables representing the different scale items and dimensions of empowerment, respectively. Fourth, we tested ‘‘error invariance’’ by adding equality constraints on the uniquenesses (i.e., errors associated with the items) across the languages by setting all uniquenesses to 1.0 across the two language groups (Model 4). This is seen as a strict test of invariance and is akin to testing invariance of the reliability associated with the items within each factor if the latent factor variances are equal. Error invariance implies that for a given item, the items’ unique variances are similar across language groups. Last, we tested additional levels of invariance by imposing equality constraints across the two language groups on factor variances (Model 5), factor covariances (Model 6), and latent factor means (Model 7) [17]. However, it should be noted that it is not necessary to have invariant factor variances, factor covariances, or error variances to conduct mean comparisons across two groups [18].
Results Overall, a small percentage from each group—3.7 % (27/ 731) and 1.3 % (7/527) of individuals from the English- and French-speaking groups, respectively—had missing data for five or more of the 25 items (C20 %) on the heiQ scales and were excluded from the analyses. Thus, results presented are
based on 704 English-speaking and 520 French-speaking adults diagnosed with cancer. The two language groups were generally similarly aged. However, there were more men, fewer employed participants and a substantially greater proportion of prostate cancers and only a very small proportion of breast cancers, in the French-speaking group compared to the English-speaking one (Table 1). Factor structure Based on the goodness-of-fitness indices of the singlegroup CFA, the data fit the five-factor model well for both the French-language [v2 (265) = 898.21, p \ .001, CFI = .975, RMSEA = .068 (90 % CI .063–.073)] and English-language heiQ empowerment scales for cancer [v2 (265) = 1100.28, p \ .001, CFI = .966, RMSEA = .067 (90 % CI .063–.071)]. These results are comparable to those found in the validation study of the five EnglishTable 1 Characteristics of 704 English- and 520 French-speaking participants included in the measurement invariance assessment English speakers N (%)
French speakers N (%)
Gender Men
292 (41.5)
41 (80.6)
Women
408 (58.0)
201 (19.4)
Missing
4 (.6)
Age (years) 18–44
64 (9.1)
6 (1.2)
45–54
123 (17.5)
60 (11.5)
55–64 65–74
221 (31.4) 170 (24.1)
163 (31.3) 196 (37.7)
C75
109 (15.5)
93 (17.9)
17 (2.4)
2 (.4)
High school or less
305 (43.4)
281 (54.0)
Collegial level or university
380 (54.0)
232 (44.6)
19 (2.7)
7 (1.3)
Missing Highest completed education
Missing Employed at the time of study Yes
390 (55.4)
105 (20.2)
No
300 (42.6)
410 (78.8)
14 (2.0)
5 (1.0)
Missing Type of cancer Breast
239 (33.9)
27 (5.2)
Prostate
106 (15.1)
293 (56.3)
Colorectal
87 (12.4)
153 (29.4)
Lung Lymphoma
28 (4.0) 52 (7.4)
47 (9.0) –
Other
160 (22.7)
–
More than one type
18 (2.6)
–
Missing
14 (2.0)
–
123
Qual Life Res Table 2 Standardized factor loadings and latent variable correlations of the five heiQ empowerment scales for the English- and French-language versions Items
Factor I E/F
Factor II E/F
Factor III E/F
Factor IV E/F
Factor V E/F
Factor I: Social integration and support If I need help, I have plenty of people I can rely on
.78/.84
I have enough friends who help me cope with my health… When I feel ill, my family and carers really understand…
.84/.85 .72/.74
Overall, I feel well looked after by friends or family
.89/.86
I get enough chances to talk about my health…
.84/.84
Factor II: Health services navigation I have very positive relationships with my healthcare…
.81/.86
I communicate very confidently with my doctorsa about…
.89/.83
I confidently give healthcare professionals the information…
.80/.82
I get my needs met from available healthcare resources…
.79/.88
I work in a team with my doctorsa and other healthcare…
.71/.85
Factor III: Constructive attitude and approaches I try not to let my health problems stop me from enjoying life
.82/.84
My health problems do not ruin my life
.86/.83
I feel I have a very good life even when I have health problems
.88/.89
I do not let my health problems control my life
.88/.94
If others can cope with problems like mine, I can too Factor IV: Skill and technique acquisition
.80/.85
I have effective ways to prevent my health problemsb…
.78/.81
I have a very good idea of how to manage my health problems
.80/.89
When I have health problemsb, I have skills that help me cope
.81/.93
I have a good understanding of equipment that could make my life easier
.66/.63
Factor V: Emotional distress (reverse-scored) I often worry about my health
.65/.62
My health problems make me very dissatisfied with my life
.90/.72
I often feel angry when I think about my health
.84/.82
I feel hopeless because of my health problems
.88/.86
I get upset when I think about my health
.89/.87
If I think about my health, I get depressed
.89/.94
E/F: English/French values. All values significant at p \ .001 ‘‘…’’ indicates truncated items (complete versions supplied with license) a
Change for cancer setting: ‘‘doctors’’ substituted for ‘‘doctor’’ in heiQ version 3.0
b
Change for cancer setting: ‘‘health problems’’ substituted for ‘‘symptoms’’ in heiQ version 3.0
language heiQ empowerment scales [1]. All standardized factor loadings were C.62 (Table 2). Ordinal composite reliability coefficients were all C.82, and scale mean scores and standard deviations (SD) ranged from 2.87 (SD = .64) to 3.40 (SD = .46) in both language groups (Table 3). Positive correlation coefficients of moderate to high magnitude were observed between the latent factors and were generally similar for English- and French-language heiQ empowerment scales (Table 4). This five-factor model was used in the subsequent analyses.
123
Measurement invariance The assumptions of measurement invariance were supported for the five tested French- and English-language heiQ empowerment scales, namely configural invariance (Model 1), metric invariance (Model 2), scalar invariance (Model 3), error invariance (Model 4), factor variances (Model 5), factor covariances (Model 6), and latent factor means (Model 7) (Table 5). Specifically, CFI and RMSEA values of each sequential model remained C.90 and B.08,
Qual Life Res Table 3 Reliability coefficients, score ranges, means, and SD for heiQ empowerment scales for English- (n = 704) and French-speaking (n = 520) participants Observed score ranges
Reliabilitya
English
French
English
French
English
French
Social integration and support
1–4
1–4
.91
.91
3.10 (.53)
3.26 (.49)
Health service navigation
1–4
1–4
.90
.93
3.17 (.48)
3.40 (.46)
Constructive attitudes and approaches Skill and technique acquisition
1–4 1–4
1–4 1–4
.92 .82
.94 .87
3.23 (.48) 2.96 (.43)
3.32 (.53) 3.13 (.48)
Emotional distress (reverse-scored)
1–4
1–4
.94
.92
2.87 (.64)
3.03 (.67)
Variables
Mean (SD)
Means and SD based on observed scores a
Composite reliability calculated from confirmatory factor analyses
Table 4 Latent variable correlations of the five heiQ empowerment scales comparing the English- and French-language versions Variables
1
2
1. Social integration and support
–
2. Health services navigation
E: .74 (.70–.78)
3
4
–
F: .82 (.78–.85) 3. Constructive attitude and approaches 4. Skill and technique acquisition 5. Emotional distress
E: .76 (.72–.79)
E: .66 (.61–.70)
F: .80 (.77–.84)
F: .75 (.71–.79)
–
E: .71 (.67–.77)
E: .76 (.72–.80)
E: .80 (.76–.84)
F: .77 (.73–.81)
F: .81 (.77–.85)
F: .83 (.79–.86)
E: .47 (.41–.54)
E: .37 (.30–.44)
E: .67 (.62–.72)
E: .49 (.42–.55)
F: .49 (.42–.56)
F: .39 (.32–.47)
F: .66 (.61–.71)
F: .48 (.42–.55)
–
E: Signifies English values. F: Signifies French values All values are significant at p \ .001. Values in brackets present the 95 % CI around the correlation Table 5 Fit indices for the analyses testing invariance of the English- and French-language versions Models
v2
df
Dv2
1. Configural invariance
1996.98**
530
2. Metric invariance
1931.69**
550
17.44
20
.972
.002
3. Scalar invariance
2055.23**
595
191.65*
45
.971
Ddf
–
CFI
DCFI
RMSEA 90 % CI
DRMSEA
.970
–
.067 (.064–.070)
–
.064 (.061–.067)
-.003
-.001
.063 (.060–.066)
-.001
4. Invariant uniquenesses
2194.07**
620
218.28**
25
.968
-.003
.064 (.061–.067)
.001
5. Invariant factor variances
2007.02**
625
15.77*
5
.972
.004
.060 (.057–.063)
-.004
6. Invariant factor covariances
1432.06**
635
32.34**
10
.984
.012
.045 (.042–.048)
-.015
7. Invariant latent means
1633.71**
640
104.96**
5
.980
-.004
.050 (.047–.053)
.005
No statistically significant differences were found between nested models based on both DCFI B -.010 and DRMSEA B .015 v2 Chi square, df degrees of freedom; Dv2 calculated using the DIFFTEST function in MPlus, RMSEA root mean square error of approximation, CFI comparative fit index, CI confidence interval * p \ .05; ** p \ .001
respectively. As well, DCFI and DRMSEA were B-.010 and D B .015, respectively.
Discussion Our findings from single-group confirmatory factor analysis (CFA), conducted using data from Canadian cancer survivors, indicate that the same five heiQ empowerment
scales that we validated for use among English-speaking cancer survivors [1] were also found for our translated French-language heiQ empowerment scales. Practically speaking, this means that the five heiQ empowerment constructs are shared in these two language groups. In addition, our findings from multi-group CFA testing measurement invariance provide strong evidence that in the cancer setting, the five French- and English-language heiQ empowerment scales measure these same five dimensions
123
Qual Life Res
in the same way. In fact, we found evidence of measurement invariance at all recommended levels, even at the strictest levels described by Vandenberg and Lance [17]. Taken together, these findings are fundamentally important because they show that these scales can be validly used to measure these empowerment dimensions among both English- and French-speaking cancer survivors. They also show that quantitative pooling of data or comparisons of heiQ empowerment scores between these two language groups are appropriate and will yield meaningful interpretations. In other words, any observed similarities or differences seen in levels of Social integration and support, Health service navigation, Constructive attitudes and approaches, Skill and technique acquisition, and Emotional distress for French- and English-speaking respondents would reflect true similarities or differences, not measurement artifact stemming from translation or emic cultural differences. We have assessed potential study limitations. First, the English-speaking participants came from a populationbased registry and community support programs, whereas the French-speaking participants were identified within the hospital setting at the time of a clinic visit. In fact, our results support the robustness of the scales even though data were gathered in different study settings and participants differed somewhat in distributions of gender and cancer characteristics. Many validation studies are predicated on this very idea, namely that the instrument will perform similarly in different populations if the underlying constructs have been appropriately captured. Moreover, the heiQ empowerment scales seem to perform well in groups with different diseases and conditions [14]. Second, we also note that the French-language heiQ empowerment scales were used among individuals diagnosed with lung, breast, prostate, or colorectal cancer, whereas the English-language heiQ empowerment scales were used among a group that included these same cancer diagnoses (65 %), as well as small proportions of other unspecified types of cancers, multiple cancers, or with a type of cancer that was not identified. However, the dimensions assessed by the heiQ empowerment scales are probably ones that are universally relevant to people facing cancer, its treatment and posttreatment survivorship, independent of cancer type. We cannot affirm with certainty that the French-language heiQ empowerment scales we tested, which were translated by and validated with Canadian French speakers, will perform as adequately among French speakers in other parts of the world. It is possible that words commonly used or colloquialisms used or interpretation of some heiQ items could vary somewhat in other French-speaking countries. Nonetheless, it is reassuring to note that the Canadian English versions of these scales perform similarly in terms of psychometrics to other English settings and that our
123
English- and French-language versions perform comparably. From a practical viewpoint, we know that it will not always be possible to perform extensive new validations in every language group where one would want to use these scales, so their robustness in our populations is reassuring and provides some general evidence to support using them in other French-speaking populations. Face/content validity assessment and cognitive debriefings in a new language group are both relatively low-cost methods that can provide general information about generalizability to the new group, and about whether further, more labor-intensive validation work is needed. We are unable to judge the degree of measurement invariance of the three other heiQ scales (Positive and active engagement in life, Health-directed activity, and Selfmonitoring and insight) because we did not administer them to our English-speaking group. For reasons described earlier, we purposefully limited our study with Englishspeaking cancer survivors to include only the five heiQ empowerment scales tested here [1]. However, we note that these three heiQ scales were translated into French by our group at the same time as the five empowerment scales using the same rigorous protocol and expert committee. Thus, researchers wanting to collect and pool data from English- and French-speaking cancer survivors on one or more of these other three heiQ scales (scales which have been demonstrated to have excellent psychometric properties in other settings [11, 14, 31]) could examine measurement invariance beforehand, if feasible, to ensure that group means are directly comparable. Otherwise, given the extensive efforts required in collecting data, researchers could use the scales with the five others and assess measurement invariance of these three scales before pooling their data and/or comparing scores across English- and French-speaking respondents.
Conclusion This study adds to the growing body of evidence about the robustness of the heiQ empowerment scales in general and in cancer setting specifically. Using rigorous statistical analyses, our findings provide evidence of the validity of scores on the five French- and English-language heiQ empowerment scales in the cancer setting. Our translation of the eight heiQ scales, including both the five scales measuring key aspects of generic empowerment in the cancer setting that are the object of the analyses presented here, as well as the three other scales not included in this analysis, has been accepted as the official Canadian French translation of the heiQ measurement system. Our findings mean that Canadian researchers, health service providers, and policy makers across Canada can use and reliably interpret these scales
Qual Life Res
assessing key dimensions of empowerment, an important outcome of oncology health services and programs. With growing emphasis on self-management in patient care in many settings and conditions, these empowerment scales will fill an important gap for meaningful evaluation. Acknowledgments This study was supported by grants to HSC from the Canadian Institutes for Health Research (CIHR) [# KAL82607], to EM from the National Cancer Institute of Canada (NCIC) [# 010498, #010499] via the Propel Centre for Population Health Impact, and to LF from the Canadian Partnership against Cancer (Cancer Journey Portfolio 2007–2012). This manuscript was prepared while JB was supported by a Canadian Cancer Society Career Development Award in Prevention, and SL by a Chercheur-boursier Award from the Fonds de recherche du Que´bec—Sante´ in partnership with the l’Institut National d’Excellence en Sante´ et en Services Sociaux (INESSS). This work was conducted while SL was supported by a Post-doctoral Fellowship Award from the CIHR-funded Strategic Training Initiative in Health Research (STIHR) Psychosocial Oncology Research Training program (PORT). RHO is funded in part through a National Health and Medical Research Council Senior Research Fellowship #APP1059122.
10.
11.
12.
13.
14.
Conflict of interest The authors declare that they have no conflicts of interest. 15.
References 16. 1. Maunsell, E., Lauzier, S., Brunet, J., Pelletier, S., Osborne, R. H., & Campbell, H. S. (2014). Health-related empowerment in cancer: Validity of scales from the Health Education Impact Questionnaire. Cancer, 120(20), 3228–3236. 2. McCorkle, R., Ercolano, E., Lazenby, M., Schulman-Green, D., Schilling, L. S., Lorig, K., et al. (2011). Self-management: Enabling and empowering patients living with cancer as a chronic illness. CA: A Cancer Journal for Clinicians, 61(1), 50–62. 3. Lauzier, S., Campbell, H. S., Livingston, P. M., & Maunsell, E. (2014). Indicators for evaluating cancer organizations’ support services: Performance and associations with empowerment. Cancer, 120(20), 3219–3227. 4. Lovell, M. R., Luckett, T., Boyle, F. M., Phillips, J., Agar, M., & Davidson, P. M. (2014). Patient education, coaching, and selfmanagement for cancer pain. Journal of Clinical Oncology, 32(16), 1712–1720. 5. Yun, Y. H., Sim, J. A., Jung, J. Y., Noh, D. Y., Lee, E. S., Kim, Y. W., et al. (2014). The association of self-leadership, health behaviors, and posttraumatic growth with health-related quality of life in patients with cancer. Psycho-Oncology, 23(12), 1423–1430. 6. Brady, T. J., Kruger, J., Helmick, C. G., Callahan, L. F., & Boutaugh, M. L. (2003). Intervention programs for arthritis and other rheumatic diseases. Health Education and Behavior, 30(1), 44–63. 7. Newman, S., Steed, L., & Mulligan, K. (2004). Self-management interventions for chronic illness. The Lancet, 364(9444), 1523–1537. 8. National Cancer Institute. Dictionary of cancer terms. www.can cer.gov/dictionary. Accessed 15 December 2014. 9. Osborne, R. H., Elsworth, G. R., & Whitfield, K. (2007). The Health Education Impact Questionnaire (heiQ): An outcomes and evaluation measure for patient education and self-management
17.
18.
19. 20.
21.
22.
23. 24.
25.
interventions for people with chronic conditions. Patient Education and Counseling, 66(2), 192–201. Buchbinder, R., Batterham, R., Elsworth, G., Dionne, C. E., Irvin, E., & Osborne, R. H. (2011). A validity-driven approach to the understanding of the personal and societal burden of low back pain: Development of a conceptual and measurement model. Arthritis Research and Therapy, 13(5), R152. Epstein, J., Osborne, R. H., Elsworth, G. R., Beaton, D. E., & Guillemin, F. (2015). Cross-cultural adaptation of the heiQ questionnaire: Experimental study showed expert committee, not back-translation, added value. Journal of Clinical Epidemiology, 68(4), 360–369. Kroon, F. P. B., van Der Burg, L. R. A., Buchbinder, R., Osborne, R. H., Johnston, R. V., & Pitt, V. (2014). Self-management education programmes for osteoarthritis (review). Cochrane database of systematic reviews, issue 1. Art. no.: CD008963. doi:10.1002/14651858.CD008963.pub2. Osborne, R. H., Batterham, R., & Livingston, J. (2011). The quality, impact and implementation of chronic disease selfmanagement across settings: The international experience of the Health Education Impact Questionnaire (heiQ) quality monitoring system. Nursing Clinics of North America, 46(2), 255–270. Schuler, M., Musekamp, G., Bengel, J., Nolte, S., Osborne, R. H., & Faller, H. (2014). Measurement invariance across chronic conditions: A systematic review and an empirical investigation of the Health Education Impact Questionnaire (heiQTM). Health and Quality of Life Outcomes, 12(56). doi:10.1186/1477-7525-12-56. Wahl, A. K., Langeland, E., Larsen, M. H., Robinson, H. S., Osborne, R. H., & Krogstad, A. L. (2015). Positive changes in self-management and disease severity following climate therapy in people with psoriasis. Acta Dermato-Venereologica, 95(2), 317–321. Byrne, B. M., & Watkins, D. (2003). The issue of measurement invariance revisited. Journal of Cross-Cultural Psychology, 34(2), 155–175. Vandenberg, R. J., & Lance, C. E. (2000). A review and synthesis of the measurement invariance literature: Suggestions, practices, and recommendations for organizational research. Organizational Research Methods, 3(1), 4–70. Steenkamp, J. B. E., & Baumgartner, H. (1998). Assessing measurement invariance in cross-national consumer research. Journal of Consumer Research, 25(1), 78–107. Fillion, S., & Cook, S. (2012). Final report: Navigation Project CPAC and CIHR April 2007 to March 2012. Unpublished report. Hawkins, M., & Osborne, R. (2007). Questionnaire translation and cultural adaptation procedure. Version 1.0. University of Melbourne, Australia, p. 9. Hambleton, R. K., & Kanjee, A. (1995). Increasing the validity of cross-cultural assessments: Use of improved methods for test adaptations. European Journal of Psychological Assessment, 11(3), 147–157. Vallerand, R. J. (1989). Vers une me´thodologie de validation trans-culturelle de questionnaires psychologiques: Implications pour la recherche en langue franc¸aise. Psychologie Canadienne, 30(4), 662–680. Willis, G. B. (1999, August). Cognitive interviewing: A ‘‘how to’’ guide. In Meeting of the American statistical association. Kornblith, A. B. (1998). Psychosocial adaptation of cancer survivors. In J. C. Holland, W. Breitbart, P. B. Jacobsen, M. S. Lederberg, M. Loscalzo, M. J. Massie, & R. McCorkle (Eds.), Psycho-oncology (pp. 223–241). New York: Oxford University Press. Dempster, A. P., Laird, N. M., & Rubin, D. B. (1977). Maximum likelihood from incomplete data via the EM algorithm. Journal of the Royal Statistical Society: Series B (Methodological), 39(1), 1–38.
123
Qual Life Res 26. Hu, L. T., & Bentler, P. M. (1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55. 27. Diamantopoulos, A., & Siguaw, J. (2000). Introducing LISREL. London: SAGE Publication Ltd. 28. Byrne, B. M., Shavelson, R. J., & Muthe´n, B. (1989). Testing for the equivalence of factor covariance and mean structures: The issue of partial measurement invariance. Psychological Bulletin, 105(3), 456–466. 29. Millsap, R. E., & Yun-Tein, J. (2004). Assessing factorial invariance in ordered-categorical measures. Multivariate Behavioral Research, 39(3), 479–515.
123
30. Chen, F. F. (2007). Sensitivity of goodness of fit indexes to lack of measurement invariance. Structural Equation Modeling: A Multidisciplinary Journal, 14(3), 464–504. 31. Morita, R., Arakida, M., Osborne, R. H., Nolte, N., Elsworth, G. R., & Mikami, H. (2013). Adaptation and validation of the Japanese version of the Health Education Impact Questionnaire (heiQ) for the evaluation of self-management education interventions. Japan Journal of Nursing Science, 10(2), 255–266.
