Living with Chronic Illness Scale: International validation through the classic test theory and Rasch analysis among Spanish‐speaking populations with long‐term conditions

Abstract Background The Living with Chronic Illness (LW‐CI) Scale is a comprehensive patient‐reported outcome measure that evaluates the complex process of living with long‐term conditions. Objective This study aimed to analyse the psychometric properties of the LW‐CI scale according to the classic test theory and the Rasch model among individuals living with different long‐term conditions. Design This was an observational, international and cross‐sectional study. Methods A total of 2753 people from six Spanish‐speaking countries living with type 2 diabetes mellitus, chronic obstructive pulmonary disease, chronic heart failure, Parkinson's disease, hypertension and osteoarthritis were included. The acceptability, internal consistency and validity of the LW‐CI scale were analysed using the classical test theory, and fit to the model, unidimensionality, person separation index, item local independency and differential item functioning were analysed using the Rasch model. Results Cronbach's α for the LW‐CI scale was .91, and correlation values for all domains of the LW‐CI scale ranged from .62 to .68, except for Domain 1, which showed correlation coefficients less than .30. The LW‐CI domains showed a good fit to the Rasch model, with unidimensionality, item local independency and moderate reliability providing scores in a true interval scale. Except for two items, the LW‐CI scale was free from bias by long‐term condition type. Discussion After some adjustments, the LW‐CI scale is a reliable and valid measure showing a good fit to the Rasch model and is ready for use in research and clinical practice. Future implementation studies are suggested. Patient and Public Contribution Patient and public involvement was conducted before this validation study ‐ in the pilot study phase.


| INTRODUCTION
Nowadays, long-term conditions (LTCs) are the leading causes of disability and costs worldwide. 1 It is estimated that, by 2030, LTCs will account for three quarters of deaths globally, with considerable social and economic impact due to the exorbitant costs to the system. 2,3 In particular, cardiovascular and respiratory diseases, as well as diabetes, are the LTCs that account for the most deaths woldwide. 1 Based on previous studies, living with an LTC is defined as a complex, dynamic, cyclical and multidimensional process that involves five different attributes, namely, acceptance, coping, self-management, integration and adjustment. 4 Living with an LTC is influenced by social and personal factors; therefore, living with an LTC is a unique, individual and unrepeatable process. 4,5 Health and social professionals need to have an indeep understanding of what it means to live with an LTC from the person's perspective. They should focus on the psychosocial and spiritual areas of the person and not just on the condition as has been done thus far. 1,4,6 In this sense, use of patient-reported outcome measures (PROMs) is paramount to evaluate how the person is living with his/her condition. 7,8 Consequently, an interdisciplinary team could design individualized and comprehensive care pathways and recommendations to achieve positive living with an LTC as a final desired target. 7,8 Currently, most relevant available PROMs in clinical practice and research evaluate specific aspects of the LTC (e.g., stage, symptoms or severity) or outcomes of the process of living with LTCs (e.g., quality of life, satisfaction with life or well-being). 9 The Long-Term Conditions Questionnaire 10 is the only PROM that potentially evaluates how a person is living with/managing an LTC. However, this scale seems to evaluate quality-of-life outcomes and does not tackle items related to acceptance, which is an essential attribute when living with an LTC. 4 Inevitably, this left a gap regarding personcentred measures that can evaluate the process of living with an LTC. To fill this gap and based on previous empirical studies 11,12 and conceptual works, 4,13 the Living with Chronic Illness (LW-CI) Scale was designed. 14 The LW-CI scale is a comprehensive PROM that evaluates the complex process of living with an LTC. 14,15 It is a 26-item self-reported scale with direct applicability in people living with LTCs. It was originally designed for Spanish-speaking populations living with an LTC. 15 The LW-CI scale has been previously published for Spanish-speaking populations in a pilot study carried out among patients living with different LTCs. 15 To date, the LW-CI scale has been tested separately in several LTCs, namely, Parkinson's disease (PD), 16,17 osteoarthritis, 18 chronic heart failure, 19 type 2 diabetes mellitus (T2DM) 20 and chronic obstructive pulmonary disease (COPD). 21 The results from each validation study showed that the LW-CI scale is a feasible, reliable and valid measure to evaluate separately the process of living with an LTC in a Spanishspeaking population. According to those validation studies, 16-21 potential modifications were proposed to achieve a better version of the LW-CI scale for each LTC in particular, such as simplifying the response scale, deleting some items or redesigning the domains. However, could those previous specific-disease validation results be extended across LTC populations? There is an opportunity to test the psychometric properties of the LW-CI scale in a sample representing diverse chronic health conditions to validate the measurement properties across persons living with different LTCs.
All LW-CI scale validation studies were conducted using a classical test theory (CTT) approach to evaluate reliability, validity and sensitivity to change along with acceptability and other parameters, mostly based on correlations and mean difference analyses. Only the LW-CI for people with PD was additionally tested using Rash measurement analysis. 17 The application of the Rasch model, 22 one of the most used applications of the item response theory, combined with the classic test theory approach is recommended for evaluating PROM. [23][24][25] The Rasch model 22 completes the information provided by the CTT because it provides additional and relevant information about the measurement properties of a scale such as unidimensionality and differential item functioning (DIF) by individual groups including LTC type. In addition, it also allows for the calculation of scores on a linear scale. 26 Therefore, the aim of this study is to analyse the psychometric properties of the LW-CI scale according to the classic test theory and the Rasch model among people living with different LTCs. Suggestions for modification of the LW-CI scale are presented accordingly.

| Design
This was an international and cross-sectional study. This study is part of a multicentric and multidisciplinary research programme led by nurses (ReNACE Programme; https://www.unav.edu/web/programa-renace/ proyectos) aimed at achieving an in-depth understanding of the complex process of living with an LTC from the persons' and family/carers' perspectives through the development of individualized interventions and comprehensive PROM. 11,12 In particular, this study has the general aim of achieving a unique and standardized international Spanish-speaking selfreported scale to evaluate how the person is living with his/her LTC in several Spanish-speaking countries.

| Setting
Participants were recruited from different healthcare systems from six Spanish-speaking countries, namely, Spain, Cuba, Argentina, Ecuador, Mexico and Colombia. More concretely, private and public primary and specialized healthcare service-attending outpatients with LTCs were included. Additionally, individuals living with LTCs from the community were also approached, mainly from LTC organisations in Spain, such as the Parkinson's Disease Association.

| Participants
Sampling of consecutive cases 27,28 was applied to select participants.
The sample was composed of individuals living with different LTCs, namely, T2DM, COPD, chronic heart failure (HF), PD, hypertension and osteoarthritis. The following criteria were applied: (1) diagnosed with an LTC by a general practitioner or consultant (T2DM, COPD, HF, PD, hypertension and osteoarthritis); (2) adult patient (≥18 years); (3) able to read, understand and answer written questionnaires; (4) native Spanish-speaking person; and (5) able to provide written informed consent. The applied exclusion criteria were as follows: (1) presence of cognitive deterioration or psychiatric disorders, or any other disorder that could interfere with or impede the appropriate development of the study (e.g., blindness); (2) hospitalized patients; and (3) patients not fulfilling the established inclusion criteria.
According to international criteria, the sample size was estimated to fulfil the rule of 10 participants per item, 29 which exceeds the minimum of 100 subjects required for CTT. Therefore, considering that the LW-CI scale is a 26-item scale, the minimum sample size estimated was 260 participants per condition, aiming for a total of 1560 participants.

| Assessments
Sociodemographic data, such as age, gender, marital status, employment situation and educational level, were collected. In addition, LTC-related information was collected, namely, age at disease and disease duration. In addition to sociodemographic and LTC-related data, the Spanish version of the following self-reported validated PROMs was used: 1. LW-CI scale 15 : The LW-CI scale is a self-reported scale that evaluates the complex process of living with an LTC. The LW-CI scale is a 26-item scale grouped into five domains: Domain 1: Acceptance (4 items); Domain 2: Coping (7 items); Domain 3: Self-management (4 items); Domain 4: Integration (5 items); and Domain 5: Adjustment (6 items). 15 All items are answered using a 5-point Likert scale from never or nothing (0) to always or a lot (4), except for Domain 1: Acceptance, which is reversely scored from never or nothing (4) to always or a lot (0). The LW-CI scale has a total score value ranging from 0 points, indicating negative living with the condition, to 104 points, reflecting positive living with the condition. The DUFSS is a self-reported measure that comprises 11 items evaluating diverse dimensions of social support such as confidant, affective and instrumental support. The score for each item varies from 1 (much less than I would like) to 5 (as much as I would like).
The total score ranges from 11 (lowest level of support 'much less than I would like') to 55 (highest level of support 'as much as I like').
According to the Spanish validation study, 31 the DUFSS presented adequate psychometric properties, showing a Cronbach's α value of .9 and satisfactory construct validity. 30,31 3. Modified version of the Satisfaction with Life Scale (SLS-6): Originally, this is a 7-item self-reported questionnaire. 32 For this study, the modified version of the SLS-6 was used because the original version and, in particular, one of the items were specific for a student population. 33 In this way, a modified version with a 6-item scale was used to evaluate the degree of overall satisfac-  35 It is a global index that may be used to assess self-perception of disease severity of the person living with a disease. The PGIS is a 6-point Likert scale, ranging from 0 (not ill at all) to 5 (extremely ill). It has excellent construct validity and has been widely used in studies of chronic diseases. 36

| Data collection
For people living with PD, data were collected between January and June 2015, and for people living with T2DM, COPD, HF, hypertension and osteoarthritis, data were collected between November 2018 and May 2019 among the different healthcare centres and community settings of six Spanish-speaking countries. Although data collection was performed at different times, a detailed and standardized protocol was used to ensure data collection homogeneity and reduce possible errors during the process. 37 According to the established protocol, potential participants were approached through a health professional (nurse or physician) or member of the research group.
An invitation letter and a participant information sheet as well as verbal information were provided about the study. Participants who agreed to participate signed the consent form and were asked to complete the sociodemographic and LTC-related data as well as PROM when they agreed with his/her health professional or researcher. Hence, participants completed the scales during consultations with the GP, specialized clinician, nurse specialist or primary care nurse. Participants who required help to complete the scales (e.g., due to vision problems) were assisted by a researcher. Once the participants finished answering the scales, a researcher reviewed the answers to identify possible missing data. Hence, no missing data were expected.

| Data analysis
Descriptive statistics, namely, central tendency measures and proportions, were used to determine the sociodemographic and LTC characteristics of the participants. The main data were ordinal or did not fit a normal distribution. Therefore, nonparametric statistics were used.
According to the CTT, the following psychometric attributes were analysed: 1. Feasibility and acceptability: The quality of the data was considered satisfactory if 95% of the data were computable. The limit for missing data was less than 5%, 38 and the mean, median and standard deviation (SD) were estimated to be roughly equivalent (≤10% maximum punctuation). 39 Floor and ceiling effects were deemed acceptable if they were less than 15%, 39  and unidimensionality. In addition, DIF was analysed by the following factors: sex, age groups (70 years or younger vs. older than 70 years), disease duration (categorized by the median; up to 6 vs. 7+ years) and LTC type; DIF by country was not possible due to an unequal distribution of data from different countries, precluding comparisons.
According to the Rasch model, the response to a certain item is a function of the person's ability (or experienced level of the construct) and the item's difficulty (or the measured level of construct by that item), expressed in logits. 22 Fit to the Rasch model is observed when there is a nonsignificant χ 2 difference between the data and the Rasch model, with Bonferroni correction by number of items. 26 In addition, residuals should follow a normal distribution (mean of 0; SD of 1) and fall within the ±2.5 range. Large sample sizes might result in a high statistical power to detect small model deviations and unnecessary modifications. Therefore, for Rasch analysis, we included a random sample of 300 individuals. 49 Reliability is expressed by the PSI, interpreted similarly to Cronbach's α. A threshold is the point of equal response probability between two adjacent categories. Its order is analysed and in the case of disordered thresholds, adjacent categories are collapsed. For items to be locally independent, we expect low correlations (<.30 of the average correlation) between item residuals. 50 Unidimensionality is measured through a principal component of the residuals, and person estimates are compared using t-tests. A lower bound of the associated binomial 95% confidence interval (CI) less than 5% indicates unidimensionality. 51,52 For DIF, analyses of variance (ANOVA) with Bonferroni correlation are conducted for all factors. 53 When several items present significant DIF by a certain factor, a top-down purification procedure is followed by creating two groups of items, with or without DIF, and comparing the estimates in an ANOVA to see if DIF remains. 54 Model modifications were evaluated iteratively.

| RESULTS
A total of 2753 people living with different LTCs were included in the sample. Osteoarthritis presented the lowest sample size (n = 291), and T2DM presented the highest sample size (n = 582). Demographic information is shown in Table 1. The age range was 20-98 years, with a mean age of 68.21 years (SD: 12.21 years). More than half of the sample was female (n = 1441, 52.3%), was married (n = 1555, 56.5%) and had basic/primary education levels (n = 1596, 58.1%). The employment status of the sample was mainly distributed between been retired (n = 934, 34%) and working as housekeeper (n = 827, 30.1%).
All of the participants had at least one LTC, with a duration of 9.80 years (SD: 8.65; range: 0-67 years) and a mean age at diagnosis of 58.37 years (SD: ±13; range: 3-91 years).

| CTT analysis
The results related to feasibility and acceptability showed that the LW-CI scale was fully completed by 2738 participants, with 99.46% of the data computable. Levels of missing data were low and broadly uniform across domains, ranging from 0 missing data (Domain 1: Acceptance) to 6 (Domain 5: Adjustment). The floor effect was absent in all domains and in the total score, whereas Domains 1: Acceptance and 4: Integration showed ceiling effects (19.1% and 15.6%, respectively). For the five domains and the LW-CI total scale, the difference between the mean and the median was less than 10%, the theoretical and observed ranges were coincident and the skewness values were between −1 and +1. Regarding convergent validity (Table 3)

| Rasch analysis
The first Rasch analysis, with all items, showed a significant model deviation: χ 2 (234) = 560, p < .00001. Because the total scale was multidimensional (19% significant t tests, 95% CI: 0.165-0.215), each domain was analysed separately. Table 5 presents the goodness of fit of the LW-CI domains, and Table 6 presents the individual item fit.

| Feasibility and acceptability
In general, all acceptability parameters fulfilled the standard criteria.
Floor and ceiling effects could be explored using both CTT and Rasch analyses. The two domains with the highest ceiling effect at the item level, Acceptance and Self-management, also showed a ceiling effect at the threshold level in the Rasch analysis. This could be attributed to the fact that individuals involved in this study participated voluntarily, potentially implying some degree of acceptance as well as good management of the LTC and hence, in some cases, also a positive living with the disease. A large percentage of individuals tend to score in the highest levels of these domains, which might prevent observation of changes after an intervention in people with initial high scores. To verify this result, group comparisons in the Acceptance and Self-management domains should be performed with caution.
The quality of the data was satisfactory, exceeding 95% of computable data, due to the close supervision performed by researchers during data collection and the standardized protocol established for the data collection.

| Reliability
Reliability was also explored using both CTT and Rasch approaches.

| Internal validity
The internal validity for LW-CI scale domains was excellent, except for Domain 1: Acceptance, with correlation coefficients under .30 with the rest of the domains. This result is consistent with previous validation studies 16 and conceptual work 4 showing that Acceptance is always the first domain to achieve a positive living. Only when the person has accepted and assumed his/her illness can he or she move on to the other domains, such as Coping or Self-management.
Therefore, according to the poor correlation that Acceptance showed with other dimensions of the LW-CI scale and based on the aforementioned conceptual framework, these findings were expected because Acceptance is considered an internal, illness-independent, process through which the person recognizes and accepts the reality.
Interval validity was also supported by Rasch analysis, showing that each domain is unidimensional, and providing support for the calculation of domain scores.

| Known groups' validity
The LW-CI scale demonstrated satisfactory known groups' validity, yielding significantly different scores between men and women, patient-based global impression of disease severity and LTC type.
The result related to differences among types of LTC is understandable, as each condition has its particular symptoms and evolution. For example, PD is defined as a complex and disabling disorder characterized by being a neurodegenerative and progressive disorder, while hypertension is a cardiovascular condition that could be managed with healthy lifestyle patterns. In this sense, the individual may have a better or worst process based on the LTC characteristics.
However, further studies are suggested to verify this result. Regarding gender differences, existing evidence also justifies the identified results. For example, Crispino et al. 56 stated that in general, women with PD showed more positive disease outcomes than men.
Besides, other studies performed in individuals with T2DM 57 showed that women with T2DM are at greater risk than men of psychosocial  59 and full validation study is ongoing, which would allow health and social care professionals to implement personcentred care pathways and referral processes.
This validation study is novel because the LW-CI scale is the first validated rating scale for assessing the phenomenon of living with an LTC. Moreover, this scale has been used for the first time to assess several highly prevalent LTCs in different Spanish-speaking countries using two complementary analytical approaches to ascertain its psychometric properties. Therefore, considering the results, the LW-CI scale could be used in clinical practice to evaluate the degree of living with several highly prevalent LTCs in different Spanishspeaking countries.

| Limitations
We acknowledge that the limitations of the study are mainly related to the convenience sample, with a heterogeneous representation of LTCs and countries. Additionally, although country-based samples were not large enough to produce any country-based (cultural) F I G U R E 1 Person-item threshold distributions for the Living with Chronic Illness domains analysis, further international studies are needed to test item bias by country as well as cultural differences within countries. Besides, considering the sample diversity, including people from six different countries, sociodemographic data related to ethnicity were not collected. Additionally, for this validation study patient and public involvement (PPI) was not conducted because the aim of this study was to statistically analyze the psychometric properties of the LW-CI scale. However, PPI was a key aspect during the development of the scale and piloting phase before the main validation studies. Finally, data collection was performed at different times for PD and the rest of the LTCs. However, to ensure homogeneity of the process and avoid errors, a clear and standardized protocol was followed, and data quality was equally valid for the purpose of this study. On the other hand, the strengths of this study are related to the sample diversity, including the highly prevalent and prototypical LTC population as well as the large age range population, which led to the real value of the LW-CI scale psychometric properties for a general LTC population. Finally, the combined application of the classic test theory approach and Rasch analysis is highly recommended for evaluating patient outcome report measures such as the LW-CI scale.