Introduction

It is estimated that heart failure (HF) affects 26 million people worldwide and approximately two million in Brazil. Moreover, mortality caused by HF is constantly associated with individual, social, economic and health service indicators. (¹,²,³) This context reflects the complexity of the disease, which is expressed in the multiple dimensions of the person affected.

Limitations, symptoms and multimorbidity caused by the disease are a concern for both the people experiencing situations of potential vulnerability,(⁴) and their support network. The perception of the impact of HF on daily life has repercussions on social relationships,(⁵) which, in turn, influence decisions and behaviors linked to the condition of health vulnerability. Thinking about HF from the perspective of vulnerability means understanding the person immersed in processes of strengthening and weakening their human condition in the social fabric. (⁶)

It is known that several individual aspects have already been elucidated in the literature regarding disease progression, hospitalization and death. Furthermore, there is already a validated and published questionnaire that contemplates aspects of a human person from the perspective of social vulnerability.(⁷) However, social aspects still need to be extensively studied. There is evidence that regions with people living in vulnerable situations related to social issues have higher premature mortality due to HF. In addition to this, readmissions are a measure of the social conditions in which people live, which supports higher levels of health vulnerability.(⁸)

The social aspects of these movements involve everything from bonds and support from family and friends to the person’s relationship with professionals and health systems, committed to care.(⁹,¹⁰) This overview poses the challenge of measuring the social aspects of vulnerability in people with HF.

Social aspects in HF are assessed by scales that measure family functioning(¹¹) and perceived social support.(¹²,¹³) In Brazilian studies, social support in this population has been measured using the Social Support Inventory for People Who Are HIV Positive or Have AIDS.(¹⁴,¹⁵) After reviewing the literature on health vulnerability, it was observed that studies link social processes to family support, and there are no records of instruments that assess social aspects as predictors of vulnerability, indicating a gap in knowledge.(³) In this regard, based on a bank of items on vulnerability in HF, the Health Vulnerability Questionnaire in Heart Failure, Social Version (QVSIC-Social Questionário de Vulnerabilidade em Saúde na Insuficiência Cardíaca, Versão Social) was developed.

The theoretical basis for constructing the item bank was based on a broad review of national and international studies and on solidified references in the areas of health vulnerability and HF.(³,⁵,⁸,¹¹) The 44 items, distributed across two theoretical dimensions (Copresence and Care), were validated with excellent total content validity indices (0.99; p>0.05).(⁹) However, it is necessary to observe evidence of internal structure validity, as it can indicate the degree of relationship between the questionnaire items and the construct on which the questionnaire score interpretations are based.(¹⁶)

Assessing internal structure through robust analyses is essential for gathering the greatest amount of evidence of validity and reliability to measure health vulnerability from a social perspective. Thus, the proposed estimate is predictive for directing nursing and multidisciplinary care actions, with the aim of reducing vulnerability conditions. By providing a reliable instrument, nurses can use it as a complementary part of nursing history and, later, in assessment, confirming that care must be comprehensive and encompass social aspects, and, thus, overcome the emphasis on attention to acute conditions. Furthermore, by identifying social conditions that increase vulnerability, space is opened for interdisciplinary and interprofessional work. In view of this, the research aimed to assess evidence of internal structure validity and reliability of QVSIC-Social.

Methods

This is a methodological study to test evidence of structural validity and reliability(¹⁶) of QVSICSocial, carried out in a tertiary health institution that is a reference in the care of people with cardiopulmonary diseases, located in the city of Fortaleza (CE, Brazil). The institution offers highly complex procedures and has its own outpatient clinics, inpatient unit and rehabilitation sector for patients with HF. The study was developed between June 2019 and September 2021.

In planning the research, the 110 original items were applied.(⁹) However, for the present study, evidence of validity was analyzed only for the social dimension of the instrument (44 items), in order to construct an instrument to assess social vulnerability in this population. Thus, since it is an extensive item bank, and with the intention of preserving heterogeneity, 1,100 participants were included. After losses (92 participants due to incomplete responses to the items), the sample was 1,008 participants.

When related to the 44 items of the social dimension, the mean number of participants was 22.9 per item, a quantity appropriate for the sample size of psychometric studies, estimated based on the number of items, which demonstrate proportions of 10:1 or more.(¹³)

Patients with a medical diagnosis of HF, aged ≥ 18 years, who were followed up or admitted to the institution’s wards, outpatient clinics, cardiac rehabilitation and emergency departments were selected, and those who did not communicate verbally were excluded. A non-probabilistic convenience sampling technique was used, consisting of those who were willing to participate in the investigation.

Data collection planning involved the selection of researchers (research coordinator and two undergraduate nursing students who received Scientific Initiation scholarships) and training. Training was conducted by the research coordinator and took place in three meetings: the first, in a private room, where research objectives were explained and the item bank was presented; the second and third meetings took place at the health institution, in which, in the second, students observed the coordinator applying the items; and, in the third, there was a meeting to clarify doubts.

Data collection took place from June 2019 to January 2020. Patients were approached in the transplant and HF unit and inpatient units and in the cardiac rehabilitation and emergency departments of the institution. After the researcher presented the study objectives and importance, each patient was invited to participate in the research. Those who agreed signed the Informed Consent Form, provided in two copies, and were directed to a private room to guarantee privacy.

Two instruments were used to collect data: a questionnaire on sociodemographic data (sex, age, region of residence, race, education, family income, number of people in the same household, whether the person was engaged in paid work and whether the person received benefits related to the disease), clinical data (etiology of HF, functional class and number of hospitalizations in the last year) and behavioral data (smoking, alcohol consumption and whether the person performed physical activity); and a bank with 44 items to be answered with the help of researchers, relating to the social aspect of the health vulnerability of people with HF.(⁹) Of the 44 items, 12 had a dichotomous response pattern (yes/no) and 32 were polytomous, with a five-point Likert-type scale, from 1 to 5 (never to always).

The data were organized in a spreadsheet in Microsoft Office Excel® and exported to Factor (version 11.05.01) and R (version 3.6.2). To characterize the sociodemographic, clinical and behavioral data of research participants, measures of central tendency and dispersion of quantitative variables and simple frequency and percentage of qualitative variables were calculated. Normality was determined by the Kolmogorov-Smirnov test.

To validate the internal structure of QVSICSocial, exploratory and confirmatory factor analysis were implemented. Exploratory factor analysis requires compliance with the following stages: data inspection techniques and factor analysis method, retention and rotation technique and quality indexes of factor loadings.(¹⁷) Sampling adequacy measures were established using the Kaiser-MeyerOlkin test (>0.60) and Bartlett’s sphericity test.

Dimensionality was verified using the Parallel Analysis Optimal Implementation technique.(¹⁸) Factor extraction was performed using the Robust Diagonally Weighted Least Squares estimation method with polychoric correlation, suitable for polytomous data and reduction of matrix residues,(¹⁹) and Robust Promin rotation.(²⁰) Test robustness was determined from the bootstrap association with extrapolation of the sample to 5,000. To complement the testing of the number of factors, the Unidimensional Congruence (UniCo), Explained Common Variance (ECV) and Mean of Item Residual Absolute Loadings (MIREAL) techniques were applied, which indicate the instrument unidimensionality if UniCo>0.95, ECV>0.85 and MIREAL<0.30.(²¹)

To maintain or remove items from the model, the following were considered: convergence of the polychoric matrix; percentage of covariance destroyed for each item; correlation above 0.2 with two other items (those below were eliminated); kurtosis and asymmetry; commonalities (h2>0.40); and factor loading values – those >0.30 were maintained and heyday cases and those with double saturation were excluded. It is reiterated that the estimated factor solutions were assessed based on theoretical reasonableness and the interpretation of factors in light of theoretical assumptions.(⁹)

In confirmatory factor analysis, the model quality and adequacy were analyzed by the NonNormed Fit Index (NNFI>0.90), Comparative Fit Index (CFI>0.94), Goodness of Fit Index (GFI>0.95), Adjusted Goodness of Fit Index (AGFI>0.93), Root Mean Square Error of Approximation (RMSEA<0.07) and Root Mean Square of Residuals (RMSR<0.08).(¹⁷) The factor scores were assessed for accuracy (Overall Reliability of Fully-Informative Prior Oblique AND AP Scores – ORION>0.70, ability to provide similar results in repeated measurements under the same conditions), representativeness of latent trait (the items represent the construct), effectiveness of factor estimation (Factor Determinacy Index (FDI)>0.80; instrument’s ability to assess what it proposes)(²²) and replicability (Generalized G-H Index>0.80, which suggests a well-defined and stable latent variable across different studies).(²¹)

The reliability of factors was verified by item-total correlation (>0.30) and Cronbach’s alpha coefficient, using the R Psych statistical package,(²³) composite reliability (calculated by the Composite Reliability Calculator, based on standardized factor loadings and error variances: www.thestatisticalmind.com) and McDonald’s Omega (ω). Values ≥ 0.70 are considered satisfactory in exploratory studies.(²⁴,²⁵,²⁶)

The study was developed as recommended by Resolution 466/2012 of the Brazilian National Health Council, with approval by the Research Ethics Committee of the research institution, Certificate of Presentation for Ethical Consideration (Certificado de Apresentação para Apreciação Ética) 20225419.0.0000.5039 (Opinion 3.563.547).

Results

The majority of participants were male (62.9%; 634), brown (85.5%; 862) and from northeastern Brazil (98.5%; 993). Age ranged from 19 to 93 years, with a mean of 56 years (±13.9), with the age group between 51 and 65 years corresponding to 61.3% (⁶¹⁸) of participants. There was a predominance of people with incomplete elementary education, representing 30.7% (³⁰⁹) of the sample and the most prevalent social class was D (income between R$ 1,874.01 and R$ 3,748.00), covering 61% (⁶¹⁵) of participants. Of the total sample, 59.7% (⁶⁰²) of participants lived with someone; 82% (⁸²⁰) were still employed; 72.4% (⁷³⁰) were retired and 70.4% (⁷¹⁰) received government benefits due to sick leave.

The initial assessment of the questionnaire indicated a negative correlation matrix and a percentage of destroyed covariance of 56.5%, explained by the presence of ten dichotomous items (3, 4, 5, 6, 7, 8, 15, 16, 17, 18, 21 and 25). The items were removed and a new analysis was performed. In this analysis, eight items presented high kurtosis (>4.0), and eight had low commonalities (<0.40), reasons that justified their elimination. In the third analysis, among the 18 remaining items, six saturated in more than one factor. Thus, of the 44 items of the original instrument, 12 items were retained to be forwarded to robust analysis.

The sample adequacy measures, established by Bartlett’s sphericity test (8,797.8; degree of freedom of 66 and p<0.001) and by the Kaiser-Meyer-Olkin test (0.653; 95%CI 0.576-0.773), suggested interpretability of the item correlation matrix. Parallel analysis indicated three factors representative of items (Figure 1), represented by the solid line. The scree plot demonstrated three factors responsible for the greatest explained data variance (empirical) and superior to the explained variance of random data (simulated). The multidimensionality of the instrument was confirmed by the UniCo (0.37; 95%CI 0.24-0.43), ECV (0.54; 95%CI 0.48-0.72) and MIREAL (0.30; 95%CI 0.18-0.31) techniques.

The saturation of the item loadings of Factor 1 ranged from 0.841 to 0.882, respectively, for items 26, 27 and 28; Factor 2 saturated items with loadings between 0.526 and 0.759 (30, 31, 32, 33 and 34); and Factor 3, between 0.822 and 0.944 (10, 12, 13 and 14). All items in the model saturated substantially in a single dimension; commonalities were adequate and there was no violation of the limit of the factor loadings (Heywood cases: -1 + 1). Collinearity was observed in items 10, 12, 13, 27 and 28 (>0.85), which may indicate item redundancy and data distribution problems (Table 1). However, the items were maintained, as the model presented adequate indicators.

The analysis of items according to the theoretical framework indicated that each factor comprised social elements of health vulnerability. It was observed that the items saturated in the factors were coherent and reflected the social network of a person with HF, which allowed the naming: Health System (F1), Hospital Discharge (F2) and Social Support (F3). The model psychometric parameters were satisfactory regarding accuracy (ORION: F1=0.910; F2=0.830 and F3=0.955), construct representativeness (FDI: F1=0.954; F2=0.911 and F3=0.977) and replicability by latent G-H index (F1=0.830; F2=0.955 and F=0.910). However, the scores for G-H observed in F1 and F3 were low (0.500 and 0.433, respectively). All model quality and adequacy indicators presented excellent values, which indicated good theoretical fit (NNFI=0.986, 95%CI=0.953-0.993; CFI=0.993, 95%CI=0.977-0.996; GFI=0.990, 95%CI=0.948-0.994; AGFI=0.981, 95%CI=0.896-0.988; RMSEA=0.042, 95%CI=0.041-0.059 and RMSR=0.045, 95%CI=0.044-0.070). Table 2 shows the findings related to item and factor reliability. The item-total correlation coefficients ranged from 0.303 to 0.842, and were considered ideal (above 0.3). The instrument demonstrated a Cronbach’s alpha of 0.60, i.e., below the cut-off point. However, all factors showed satisfactory composite reliability (F1=0.94; F2=0.90 and F3=0.79) and McDonald’s omega of 0.79.

The Exploratory Factor Analysis indexes pointed to a model composed of three factors, with satisfactory factor loadings and good levels of adequacy, quality, reliability and accuracy, with evidence of a consistent internal structure for measuring the social vulnerability construct in HF.

Discussion

Developing and validating an instrument that represents the social dimensions of health vulnerability of people with HF constitute two important strategies in detecting weaknesses and can support the adequacy, significance and usefulness of decisions and inferences made by healthcare professionals, in addition to assisting in decision-making.

Validity based on internal structure is a stage in which we seek to understand whether the instrument factorial structure corresponds to what would be theoretically expected. Therefore, studying the QVSIC-Social factorial structure is essential to achieve all the criteria for construct validity and, thus, understand the components of its dimensions.

First, it was necessary to adapt the initial model. The model adjustment process involved the exclusion of items to correct the correlation matrix that presented six negative eigenvalues. The definition of the positive matrix of an instrument is related to the number of participants per item, its completion and the absence of multicollinearity.(²⁷) Therefore, refining the items in a preliminary manner, based on the theoretical contribution, and excluding those that had accumulated responses in one of the item response poles were ways of correcting this problem.

The model revealed a three-factor structure, which composes and protects two theoretical dimensions: Co-presence (support from family and friends) and Care (support from the health service and hospital discharge). Care can be understood as an act or possibilities and includes the positive way of caring – individual, dynamic and unfinished.(⁹) Among the items that make up the Hospital discharge and Health services (Care) factors are those related to the health institution regarding scheduling appointments, referrals and the possibility of attending unscheduled appointments; beliefs in the health system; planning and team participation in hospital discharge; scheduling of the return date and guidance from professionals at the time of discharge.

Among the social elements of health vulnerability, Co-presences are justified by the essential correlation of the person with others and involve everything they deal with in their daily lives.(⁹) In the Copresence dimension, the items covered support in times of crisis or not due to the disease and acceptance and understanding of family and friends regarding the disease.

The nurse is reinforced as a professional focused on caring for the person, the family and the community, in their context and life circumstances, assuming a social and ethical commitment to the plurality of people in vulnerable situations.(²⁸) Understanding HF from the perspective of vulnerability goes beyond clinical, physiological, psychological and behavioral data and requires the incorporation of social attributes(⁷) to guarantee successful practices in the field of nursing that effectively contribute to care.

In this sense, within the dimensions identified, the nurse ensures and assists care actions; strengthens autonomy and enhances health outcomes. In family support, the nurse engages family members, friends and caregivers in the health-disease process by providing information about HF and its nuances and encouraging them to participate in care. This inclusion is relevant, as support networks favor affective-emotional, instrumental and help support, in addition to adherence to treatment and self-care.(²⁹)

Regarding hospital discharge and health services, it is reiterated that continuity of care is fundamental to the quality of health care and is related to improving patient satisfaction, reducing costs and reducing avoidable hospital admissions.(³⁰) It is a complex and multifaceted concept that, in people with HF, has a close relationship with access to health services and social support (family and friends).

Questionnaires developed on social support restrict their content to subscales regarding the assessment of family functioning and health and support from friends for people with HF.(¹¹,¹³) In Brazil, the scale used to measure social support in this population consists of an adaptation of the Support Inventory for People Who Are HIV Positive or Have AIDS, which covers two dimensions of social support: instrumental and emotional.(³¹)

Although there are validated instruments on social support in HF,(¹¹,¹²,¹³,¹⁴,¹⁵) none assess their effects on vulnerability levels. The processes that emerge from vulnerability are not essentially individual; the social character involves continuous movements of forces that design the paths by which a person is recognized.(⁶) In this context, restricting social aspects to support and support from family and friends, excluding care, represents a reduction of the term, with the loss of fundamental components that also express and permeate a person with HF.

The psychometric parameters of the exploratory stages of available and validated questionnaires on social support in HF were limited to the KaiserMeyer-Olkin test, the Bartlett test, component analyses, item-total correlation and reliability using Cronbach’s alpha.(¹¹,¹²,¹³,¹⁴,¹⁵) In this research, Exploratory Factor Analysis included a series of indicators not commonly used in psychometric studies, with extensive recent techniques, such as accuracy, representativeness, replicability and model adequacy (NNFI, CFI, GFI, RMSEA and RMSR), available for confirmatory analyses, but possible to apply in exploratory analyses using the Factor software.(³²)

The adoption of multiple indicators arises from the need to attest, through various techniques, the instrument validity. It is important to highlight that, despite the exclusion of 24 items, the QVSICSocial demonstrated theoretical consistency and adequate parameters that attest to its quality, through sizing techniques and model adjustment indices.(¹⁶) The use of modern parameters is justified to implement adequate estimates of credibility and reliability of the findings.

In all these parameters, the model revealed satisfactory values, except for the G-H observed in two factors. The application of G-H index assesses how well defined the latent variable is from the instrument’s items, i.e., the viability of a measurement model given by a set of items. Such analyses allow us to assess the probability of the model being stable across studies, populations or subpopulations.(²¹,³³)

When drawing a parallel between instruments already developed on the subject, the Multidimensional Scale of Perceived Social Support,(¹³) a three-factor instrument (Friend, Family and Significant Others) that measures perceived social support in patients with chronic obstructive pulmonary disease and HF, demonstrated excellent internal consistency, with Cronbach’s alpha above 0.90. The Finnish Family Functioning, Family Health and Social Support questionnaire(¹²) consists of three total scales and presented Cronbach’s alpha between 0.73 and 0.95 and significant, although weak, correlations between most factors. However, none of the scales showed good model adjustment according to the Goodness of Fit Index.

Usually, the factor reliability/trustworthiness scores is estimated by calculating Cronbach’s alpha. However, this coefficient is criticized as an indicator of reliability of scores from multidimensional models, due to tau-equivalence: Cronbach’s alpha considers that all items have equivalent loadings.(²⁶) In order to overcome this limitation, composite reliability was calculated, which demonstrated values greater than 0.70 in all factors. Composite reliability is an index that allows a more precise estimate of the score of each domain of the instrument by respecting the factor loading of items, which allows a better assessment of the quality of the structural model of psychometric instruments.(³⁴)

To elucidate reliability, item-total correlations showed homogeneity and interconnection between items of moderate magnitude (r between 0.30 and 0.50) to strong (above 0.50). No item presented a weak correlation (r<0.30). The strong correlation indicates that the items contribute to increasing the variance and help in discrimination between people.(³⁵) Furthermore, modern indices, such as ORION and FDI, have improved item interpretation and reliability by improving the accuracy of estimates(¹⁷) and, therefore, the extent to which factor scores truly represent the construct.

The contemporary and robust statistical techniques(¹⁷,¹⁹,²⁰,²¹,²³,²⁵,²⁶) analyses performed in this research followed rigorous methodological frameworks and enabled understanding the model of the social version of vulnerability in the context of HF. Nursing may benefit from using the instrument to assess social aspects that influence the level of vulnerability and, also, plan care actions by considering a person and their social contexts.

As a limitation to be considered in this study, we highlighted the homogeneity of sociodemographic, economic and clinical profile of research participants, selected from a single health institution. However, it is justified to carry out the collection in a single institution because it is a reference for the entire state and receives patients from the North and Northeast regions. It is recommended that new studies be carried out with patients from other health institutions to consolidate the findings of this research.

Conclusion

The QVSIC-Social indicated evidence of validity of an adequate internal structure to measure social aspects of health vulnerability of people with HF and demonstrated good reliability. The instrument presented 12 items, distributed in three factors, which support the dimensions of Copresence and Care.

References

Edited by

Publication Dates

History