Bifactor SEM and MIRT Structure of a 12-Item Human Immunodeficiency Virus Stigma Scale in Peruvian Adults

Abstract

Background:

The stigma associated with human immunodeficiency virus (HIV) can lead to prejudice and discrimination against people who have been infected by this virus, consequently, it is important to have a validated tool to measure this phenomenon. However, there is only 1 national precedent that has validated the scores of this instrument in its 21-item version. Therefore, this study examined the bifactor structural equation method (SEM) and multidimensional item response theory (MIRT) structure of a 12-item human immunodeficiency virus stigma scale in Peruvian adults.

Methods:

We evaluated 342 patients (57.6% female and 42.45% male) diagnosed with HIV receiving highly active antiretroviral therapy (HAART) from a hospital located in East Lima, aged 18 to 45 years (M = 31.4, SD = 9.79). A SEM was used to test 2 measurement models, a 4-factor correlated oblique model and a bifactor model due to high interfactor relationships.

Results:

Acceptable fit indices were identified for the oblique model (χ²/df = 1.26, SRMR = 0.044, RMSEA [90% CI] = 0.028 [0.000-0.047], CFI = 0.996, TLI = 0.994). In the same way, similar results were evident for the bifactor model (χ²/df = 1.14, SRMR = 0.039, RMSEA [90% CI] = 0.020 [0.000-0.044], CFI = 0.998, TLI = 0.997), however, in the latter it showed a greater explanation for the unidimensional model (H = 0.87, PUC = 0.82, LCA = 0.70), which was also evidenced by the bifactor MIRT analysis.

Conclusion:

It is concluded that the 12-item HIV Stigma Scale meets the psychometric properties of internal structure and unifactorial reliability.

Keywords

stigma HIV psychometric properties factorial invariance structural equations Peru

Introduction

The human immunodeficiency virus (HIV) epidemic is a public health problem, affecting approximately 38.4 million people around the world today, indicating an increase of 1.5 million new cases over the past year. In addition, it is estimated that 0.7% of people between 15 and 49 years of age are infected with HIV, being more prevalent in men than in women.¹ Although there is still no cure for this virus, most countries have implemented treatment plans through antiretroviral therapy (ART). In fact, 28.7 million people with HIV are receiving ART, which is approximately 75% coverage.² However, this implies a significant gap in the number of people who do not receive ART. This situation is also reflected in Peru. In fact, the latest figures show that there are 87 000 people with HIV (80% receive ART), where 31.8% are transgender women, 10% gay men, 1.8% of the indigenous population of the Amazon (Awajún), and 0.3% of the general adult population.³

One of the important factors is non-adherence to treatment or non-testing for HIV, which may be due to stigmatizing aspects associated with the disease,⁴ since infected people do not know they have HIV, which generates a greater spread of the virus, a late diagnosis that leads to a greater risk of contracting opportunistic infections.^5,6 In addition, stigma causes the person to abandon treatment, as well as to hide his or her health condition from family, partners, friends, or other health professionals. This results in not receiving adequate support in the care of the disease.^7
-9 Stigma associated with HIV, therefore, refers to a negative label, attributes, and attitude, which produces prejudice and discrimination toward those who have been infected with HIV.^10,11

Several instruments have been developed to measure this phenomenon; however, one of the most widely used is the HIV Stigma Scale,^6,12 whose original version presents 40 items grouped in 4 dimensions: (1) personal stigma (experiences of rejection or fear of rejection for having HIV), (2) concern about disclosure (controlling who knows their HIV status or fear of disclosure), (3) negative self-image (guilt or shame toward oneself for having HIV), and (4) concern about the attitude of others about HIV (concern or fear that others will discriminate against them or that they may lose opportunities because they have HIV).¹³ On the other hand, although the HIV Stigma Scale has been adapted and translated into different languages such as Spanish,¹⁴ Portuguese,^15,16 Swedish,¹⁷ Urdu,¹⁸ Xhosa,¹⁹ Tamil,²⁰ Japanese,²¹ or Thai.²² In a meta-analytical study, it is recognized that this instrument exists in multiple versions that vary in the number of items, including 7, 8, 10, 12, 13, 13, 17, 18, 21, 23, 25, 32, and 39 items, and its internal structure differs even among models with the same number of items, as between 2 and 5 dimensions and various internal structures have been identified.⁶

Regarding the initial version of the instrument, difficulties have been observed in its application due to its large number of items, which can generate fatigue in the participants and difficulties in its use in environments with a large number of patients.²³ In previous research, adequate psychometric properties have been found in more concise versions of the instrument, as reported by Mlouki et al,²⁴ who identified a 12-item model with 3 dimensions in patients from Tunisia, Africa. Similarly, De los Santos²⁵ examined the 2-factor structure using 11 items in a US population living with HIV.

The most widely used abbreviated version of the HIV Stigma Scale is that developed by Reinius et al,⁴ since it maintains the 4 dimensions originally proposed by the author through 12 items and has proven to have solid psychometric properties in various contexts.^15,16,23,26 However, in these investigations, the exhaustive evaluation of its structure through more rigorous analyses, such as the Item Response Theory model, which would allow isolating the characteristics of the participants from the psychometric findings, is lacking.²⁷

The stigma associated with HIV has a significant negative impact on the quality of life of people living with this condition, and can result in social isolation, as well as contribute to the development of depression and anxiety in these individuals.²⁸ By assessing and addressing stigma in a brief manner, rapid interventions can be implemented to improve the quality of life of this population. Therefore, more instruments related to HIV are needed to learn more about its problems, such as the negative attitudes of health professionals or the social support perceived and received by this population.^29,30

In this context, the main objective of the research is to examine the psychometric properties of the 12-item version of the HIV Stigma Scale in the Peruvian population. This is justified due to the scarcity of previous psychometric studies. In fact, there is only 1 national precedent that has validated the scores of this instrument, but in its 21-item version.¹⁴ Therefore, having an abbreviated version is more convenient due to its brevity and ease of administration, which makes it a good option for measuring this psychological measure. In addition, the availability of this tool allows for studies that contribute to the understanding of stigma toward people with HIV in Peru. This is particularly as stigma is one of the most influential factors in the non-testing of the virus, which, in turn, increases late diagnosis and, therefore, an increased risk of human death and transmission.⁵

Materials and Methods

The research was conducted under an instrumental design, since it sought to report the psychometric properties of a psychological measurement instrument.³¹ In addition, it was non-experimental because no manipulation of variables was performed and cross-sectional since the information was collected within a single moment of time.³²

Participants

The study participants were 342 HIV-diagnosed patients receiving highly active antiretroviral therapy (HAART) from a national hospital located in Lima, Peru. Participants ranged in age from 18 to 45 years (M = 31.4, SD = 9.79). Approximately 57.6% (197 evaluated) of patients are male and 42.4% (149 evaluated) are female, while 70.8% (242 evaluated) maintained a heterosexual sexual orientation and 29.2% (100 evaluated) homosexual type, and bisexual. In addition, 75.7% (259 respondents) did not have a partner at the time of the evaluation and 24.3% (83 respondents) did have a partner relationship. Sampling was non-probabilistic and purposive given that inclusion and exclusion criteria were met.³³ The inclusion criteria were: to be of legal age, continuing patients receiving HAART treatment for at least 3 months, and to complete the informed consent form. Participants who were minors and those who did not respond correctly to the evaluation instruments were excluded. A priori, a statistical power calculator for the structural equation model was used to calculate the minimum sample size required (n = 207) in the confirmatory factor analysis (CFA), with the use of observed (12) and latent (4) variables in the model, a minimum predicted effect size of 0.30, probability level of 0.05 and statistical power of 0.95 according to previous studies of psychometric properties.^34
-36

Instrument

The Berger HIV Stigma Scale,¹³ in the short 12-item version of Reinius et al.⁴ For this study, the items adapted to Spanish from the Peruvian version of Franke et al,¹⁴ which can be found in Supplemental Material 1, were taken into account. The instrument used consists of 4 dimensions: (1) personalized stigma (eg, “Las personas evitan tocarme si saben que tengo VIH”/“People avoid touching me if they know I have HIV”), (2) disclosure concerns (eg, “Decir a alguien que tengo VIH es riesgoso”/“Telling someone I have HIV is risky”), (3) concerns about public attitudes (eg, “La mayoría de gente se incomodan con personas con VIH”/“Most people are uncomfortable around someone with HIV”), and (4) negative self-image (eg, “Las actitudes de la gente me hacen sentir peor conmigo mismo”/“People’s attitudes make me feel worse about myself”). Each item represents a statement that people can answer by means of a Likert-type scale of 4 alternatives (strongly disagree, disagree, agree, and strongly agree). In the present study, reliability indices were calculated by the omega coefficient for the total scale (ω = .90) and for the 4 factors: personalized stigma (ω = .75), disclosure concerns (ω = .71), concerns about public attitudes (ω = .79), and negative self-image (ω = .71). The content of the items can be found in the appendix section.

Procedure

In the first instance, permissions were requested from the original author who adapted the instrument in its Spanish version. Subsequently, permission was requested from the directors of a national hospital in the city of Lima to carry out the application of the instrument in patients who have a white code and who are receiving highly active antiretroviral therapy (HAART). In turn, the scale was reviewed by 3 expert judges to verify its conceptual content, who rated each of the items as acceptable. The application of the questionnaire was carried out in person by a team of 4 researchers who are part of the study. This occurred during one of the psychological sessions scheduled for patients as part of their comprehensive health care following their visit to the infectious disease department to receive HAART. Each evaluation took approximately 10 min per patient. Additionally, an informed consent form was provided making it clear that participation in the study was completely voluntary and anonymous. In this way, compliance with the ethical and deontological standards established by the College of Psychologists of Peru, as well as the principles set forth in the Declaration of Helsinki, was ensured. This study was approved by the ethics committee of the Universidad César Vallejo.

Data Analysis

The data analysis was performed in different stages, in the first stage the descriptive analysis of each of the items was demonstrated, such as mean, standard deviation, skewness, and kurtosis. The latter 2 must have values within the range of ±1.5 to rule out the existence of extreme measurements.³⁷

In the second stage, the CFA was determined, where the assumptions of normality (<70) or non-normality (>70) were tested by means of Mardia’s multivariate coefficient.³⁸ In addition, a robust Unweighted Least Squares (ULS) estimator was used since the variables were ordinal.³⁹ Two measurement models were evaluated, the first consisted of a 4-factor correlated oblique model (M1) and the second of a bifactor model (M2). Each model was checked by the following fit indices: Chi-square over degrees of freedom (χ²/df < 2),⁴⁰ root mean square standardized residual root (SRMR < 0.08), root mean square error of approximation (RMSEA < 0.06),⁴¹ comparative index (CFI > 0.95), and the Tucker-Lewis index (TLI > 0.90).⁴²

Additionally, within the bifactor model, the measures of the hierarchical omega coefficient for the specific factor (ωh) and for the overall factor (ωH) were tested,⁴³ as well as the H-value for the overall and specific factors,⁴⁴ which allow to consider the maximum reliability of the specific factors with a control of the influence of the overall factor. A value of ωh > 0.30 demonstrates substantiality for the factors by excluding the shared variability of the overall factor,⁴⁵ a ωh > 0.80 allows considering the unidimensionality of the model⁴⁶ and an H-index >0.70 determines the latent variable as well defined.⁴⁷ The percentage of reliability variance of the overall factor over the specific factors was recognized with the Percentage of Reliable Variance (PRV), with acceptable values when exceeding 50% (PRV > 50).⁴⁸ In addition, the common variance explained with the Explained Common Variance was calculated for the overall factor (ECV) and for the individual items (ECV-I), as well as the percentage of uncontaminated correlations with the Percentage of Uncontaminated Correlations (PUC).^49,50 It can be evidenced that an ECV > 0.70 and a PUC > 0.70 confirm the existence of a unifactorial model. Similarly, finding a PUC > 0.80, an ECV > 0.60, and ωH > 0.70 allows us to accept the unidimensionality of the model (Reise et al⁴⁹). While an ECV-I > 0.85 suggests the influence of the overall factor on the variance of the items.⁵¹ Reliability was determined by means of the hierarchical omega coefficient⁵² and H-indexes,⁴⁶ where values above .70 indicate acceptable reliability.⁵³

Finally, using the multidimensional item response theory (MIRT) approach, a bifactor MIRT model based on the Metropolis-Hastings Robbins-Monro (MH-RM) method was performed.⁵⁴ Given the polytomous nature of the data, the graded response model (GRM) and a 2-parameter logistic extension (ML2p) were used. The first parameter refers to discrimination (a), which measures the ability of the item to differentiate subjects according to their level on the latent trait. In the MIRT bifactor model, the first discrimination parameter (a1) corresponds to the general factor, while the following ones correspond to specific factors, for example: personalized stigma (a2), disclosure concerns (a3), concerns with public attitudes (a4), and negative self-image (a5), where values were considered low between 0.35 to 0.64, moderate 0.65 to 1.34, high 1.35 to 1.69, and very high greater than 1.70.⁵⁵ In addition, each item consists of 3 difficulty parameters (b) in line with the 4-point Likert scale of the measurement scale. Each threshold indicates the level of construct measurement at which participants have a 50% probability of endorsing one or the other option of the response categories.⁵⁶ For adequate model fit, the M2* test and fit indices of CFI > 0.95, RMSEA ≤ 0.08, and SRMSR ≤ 0.05 were used.⁵⁷ Additionally, a 4-factor multidimensional IRT model was performed to compare with the bifactor IRT, considering other indicators for the choice of the best model, such as: Akaike information criterion (AIC), Bayesian information criterion (BIC), sample-adjusted BIC (SABIC), and log-likelihood (LL). The model with the lowest values for the information indicators was considered optimal.

For the analysis of most of the results, the JAMOVI v.2.2.6 program was used, where the descriptive data of the items, the results of the CFA, the reliability, and the MIMIC model could be recognized. However, for the analysis of the additional bifactor model indices (ωh, ωH, H, ECV, ECV-I, and PUC), the online calculator was used to evaluate the Bifactor model indices⁵⁸ and for the TRI bifactor analysis the MIRT package in the free software Rstudio.⁵⁴

Results

In the descriptive results, a higher score is evident within item 11 (M = 2.52, SD = 1.049) and item 10 (M = 2.46, SD = 1.037). In addition, all items were within ±1.5 for skewness and kurtosis (Bandalos & Finney, 2010), with a multivariate coefficient of mardia demonstrating a lack of normality due to exceeding the minimum range of 70 (g2 = 206.967, P < .001)³⁸ (Table 1).

Table 1.

Descriptive Analysis of the Items.

Variables	M	SD	g1	g2
Item 1	2.08	0.952	0.437	−0.818
Item 2	2.21	0.951	0.336	−0.816
Item 3	2.21	1.008	0.301	−1.034
Item 4	1.98	0.885	0.591	−0.402
Item 5	2.18	1.003	0.350	−0.984
Item 6	1.98	0.966	0.625	−0.654
Item 7	1.91	0.906	0.703	−0.385
Item 8	2.08	1.004	0.482	−0.906
Item 9	2.00	0.938	0.579	−0.620
Item 10	2.46	1.037	−0.017	−1.166
Item 11	2.52	1.049	−0.024	−1.189
Item 12	2.16	1.025	0.306	−1.030
Mardia’s kurtosis	206.967**

Abbreviations: g1, asymmetry; g2, kurtosis; M, mean; SD, standard deviation.

P < .01.

To determine the internal structure-based validity of the Berger HIV Stigma Scale, the CFA was used, where adequate fit indices were observed for a 4-factor oblique model (χ²/df = 1.26, SRMR = 0.044, RMSEA [90% CI] = 0.028 [0.000-0.047], CFI = 0.996, TLI = 0.994), which has factor loadings ranging from 0.431 to 0.789 for each of the items.

On the other hand, for the bifactor model (Figure 1) it was observed that the fit indices were better compared to the oblique model (χ²/df = 1.14, SRMR = 0.039, RMSEA [90% CI] = 0.020 [0.000-0.044], CFI = 0.998, TLI = 0.997), with factor loadings between 0.339 and 0.697. Regarding the additional fit indices, a ωH equal to 0.52 was found, which determines that the specific factors are the main source of variance compared to the overall factor, however, the ωh values showed that factor 1 and factor 2 obtained low values (0.19 and 0.00 respectively), factor 3 moderate (0.28) and the last factor was substantial (0.32) when excluding the shared variability of the overall factor.⁴⁵ On the other hand, 92% of the reliable variance is explained by the overall factor (PRV = 0.92) and between 26% and 45% by the specific factors. The H coefficient provides greater evidence for the overall factor (H = 0.87) relative to the specific factors (H < 0.70) except for factor 4 (H = 0.73). It was evident that the items were influenced to a greater degree by the specific factors, given that the common variance explained by the overall factor had values below what was expected in comparison with the variance of the items (ECV = 0.70, average ECV-I = 0.73), in addition to a PUC of 0.82. For this reason, the preference for a unidimensional model is recognized.

Figure 1.

Bifactor model.

Finally, in the bifactor MIRT analysis, all items presented a higher discrimination parameter in the overall factor (HIV stigma) compared to the individual factors (Table 2). For example, item 1 presented an a parameter of 2.106 on the general factor and an a parameter of 0.512 on the personalized stigma factor, suggesting that item 1 provides more information on the general factor compared to the individual factor (in this case, personalized stigma factor). Similarly, the a parameters in all individual factors were lower than the overall factor, with the highest a parameter in item E4 (a1 = 2.640) and the lowest in item E6 (a1 = 0.650). Regarding the thresholds, according to b1 values, it is necessary to have a moderately high level of the measurement trait (eg, >1.0 for 9 of the 12 items) to move from a 1 (strongly disagree) to a 2 (disagree) response, but not necessarily a very high level of HIV stigma (b2 and b3 < 0, across all 12 items) to move from a 3 (agree) to a 4 (strongly agree) response.

Table 2.

Discrimination and Difficulty Parameters for the Bifactor MIRT Model.

Items	a1	a2	a3	a4	a5	b1	b2	b3
E1	2.106	0.512				1.216	−1.264	−3.757
E2	1.436	−0.047				1.385	−0.846	−2.768
E3	2.482	−1.030				0.846	−1.943	−4.259
E4	2.640		1.943			1.762	−1.165	−4.166
E5	2.190		1.521			1.353	−1.152	−3.601
E6	0.650		−0.020			1.304	−0.010	−1.626
E7	2.162			0.855		1.434	−0.910	−3.307
E8	2.446			0.910		1.280	−2.022	−4.648
E9	1.717			0.468		1.201	−0.881	−2.865
E10	2.425				1.880	1.189	−2.105	−4.885
E11	2.395				1.902	0.841	−2.461	−5.197
E12	1.538				1.063	0.873	−1.169	−3.119

Abbreviations: a1, general factor; a2, personalized stigma; a3, disclosure concerns; a4, concerns with public attitudes; a5, negative self-image.

Regarding the comparison of the confirmatory and bifactorial MIRT, it was found that the bifactorial MIRT presented a better fit (Table 3). The values of the model fit indicators were in favor of the bifactor model, being this model the one that presented the lowest values in the M2 test, the AIC, SABIC, BICC, and LL information criteria, as well as adequate values in the CFI, SRMSR, and RMSEA.

Table 3.

Fit Indicators for the Confirmatory and Bifactor MIRT Models.

Model fit indicators	MIRT of 4 dimensions (first-order model)	Bifactor MIRT
SRMSR	0.352	0.054
RMSEA	0.252	0.081
M2	1160.569	135.62
Akaike information criteria	9649.095	9195.138
Bayesian information criterion	9844.671	9425.226
BIC adjusted to sample	9682.887	9234.893
Log-likelihood	−4773.548	−4537.569

Abbreviations: M2, bifactor model; MIRT, multidimensional item response theory; RMSEA, root mean squared error of approximation; SRMSR, standardized root mean square residual.

Discussion

People living with HIV are exposed to greater discrimination and rejection by other subjects, including prejudices, labels, and negative attitudes toward themselves,^7,8,10 this interferes with adherence and timely treatment, as well as early detection to reduce its spread.^5,6 The study aimed to examine the psychometric properties of the 12-item version of the Berger et al.¹³ HIV Stigma Scale in a Peruvian population receiving HIV treatment.

It was found that the model with 4 correlated factors (M1) had optimal adjustment indexes; however, since there was a strong correlation between each of the dimensions, it was decided to test a bifactor model (M2). Other studies, such as the one conducted by Gonçalves et al¹⁵ with HIV patients in 2 hospitals in Portugal showed that the original 4-factor model for the 12-item Portuguese version had better fit indices than the unidimensional model. Similarly, Luz et al¹⁶ applied a virtually brief version of the instrument with 12 items adapted to Brazilian Portuguese in 3 samples of HIV-positive transgender and cisgender patients, where they confirmed the existence of a 4-factor model for the 3 groups. This model is also consistent with the original 40-item model proposed by Berger et al,¹³ who found 4 correlated dimensions in a sample of people with HIV in the United States. Subsequently, in a review study by Wanjala et al⁶ of this instrument, 14 studies were found that determined this tetrafactorial model. However, given the interfactor correlations in our study, a bifactor model was considered, which is reported in the second model (M2) proposed in the present study.

The results of the bifactor model showed higher fit indices than the 4-factor correlated model, which is to be expected in the use of structural equation models because traditional indices (eg, RMSEA, SRMR, TLI, CFI, among others) are usually higher for the bifactor model and may even report false positives.^59,60 Therefore, additional adjustment indices were reported to identify the degree of influence of the overall factor on the specific factors and on each of the items. Under this perspective, it is recognized that the average explained variance of the items was higher in relation to the overall factor, while the latter proved to be the main source of variance with respect to the specific factors, which determined greater evidence for the representation of a unifactorial model. These findings are consistent with the original model of Berger et al,¹³ the authors examined a bifactor structure with 4 dimensions through an exploratory factor analysis and evidenced the existence of a general construct for HIV stigma.

Regarding reliability, the results showed an adequate index to conclude in favor of unidimensionality, given that the H coefficient had values above 0.80 for the overall factor, which is compatible with previous findings, where total reliability obtained high values.^6,13 Therefore, the model has adequate internal consistency. In a systematic review study, there were 24 articles reporting the reliability of the Berger HIV Stigma Scale, where values for overall reliabilities ranging from 0.73 to 0.96 were presented, most considered only reporting Cronbach’s alpha values, except for 2 studies that reported the omega coefficient value and one study that added the ordinal alpha; all values exceeded .70, except for one study that validated a 4-item version.⁶ Our research reported an omega coefficient value of .90 for the 12-item model, which contrasts with previous studies by exceeding the minimum value of .70 and demonstrating adequate internal consistency reliability.

Regarding the 2-factor MIRT analysis, the evidence provides minimal support for the usual 4-factor structure.^15,16 Rather, the a parameters in the overall factor were greater than the individual factors, while when compared to the first-order model, the indices of the bifactor model supported the fact that the overall factor explained more construct variance than the 4 dimensions. This means that the general factor elicited differences in HIV stigma of clinical patients better than the dimensions, confirming a unidimensional model. In this sense, a unidimensional structure was empirically adequate for the study population. This indicates that a single latent trait best explains HIV stigma in HIV-diagnosed patients on highly active antiretroviral therapy. In addition, the item with the highest discrimination was item E4 (“Decirle a alguien que tengo VIH es riesgoso”/“Telling someone I have HIV is risky”) and its content is linked to shame and concealment of their clinical diagnosis, given that the item does not refer to a specific group. This would involve a large environment of people involved in an individual’s life (eg, family members, friends, co-workers, health professionals, among others).¹³ Therefore, this item could be useful in principle to identify whether an adult diagnosed with HIV experiences stigma due to his or her medical condition and, on the other hand, could be helpful in psychological intervention processes for clinical treatment adherence, family support, and social support.^4,7
-9 Finally, this research contributes to the cross-cultural discussion on the usefulness and use of the HIV stigma scale through a mixed approach of classical test theory and MIRT. Furthermore, the study has methodological implications on the sources of evidence of the structural validity of the HIV stigma scale in a clinical population belonging to a middle-income developing country, based on the concept that HIV stigma represents diverse societies and values constructed by culture, education, and customs.^10,11

Conclusion

In conclusion, the 12-item version of the HIV Stigma Scale showed a better fit in the bifactor model and high reliability. Therefore, the scores of the brief version of the HIV Stigma Scale in Peru are valid and reliable.

Limitations and Strengths

Among the limitations of the study, we can mention that the sample was obtained in a non-probabilistic manner, which is why the results could not be generalized to the Peruvian population. In addition, since it was a cross-sectional study, it was not possible to demonstrate the temporal stability of the internal consistency or of the scores obtained. Therefore, it is recommended that future research should follow longitudinal designs with probability sampling in order to be replicable.

However, among its strengths, the fact that it is the first Peruvian study to examine the psychometric properties of the HIV Stigma Scale using a classical test theory and MIRT approach¹³ in its abbreviated version stands out, which will allow health professionals to use this instrument in a simple manner and in a short time. In addition, having this scale available allows for the development of research to gain more knowledge about HIV stigma; this is important because it can contribute to the prevention and reduction of the transmission of the disease, which is relevant considering that it is one of the factors that most affects the lack of adherence to HIV treatment, the performance of screening tests, and treatment abandonment; this generates a worse prognosis for those who have been infected and a greater spread of the virus^4
-6 and in the concealment of the state of health before family, friends, or close people, who in the absence of social support present a deterioration of psychological well-being when suffering from the disease.^7
-9

Supplemental Material

sj-pdf-1-jpc-10.1177_21501319231197589 – Supplemental material for Bifactor SEM and MIRT Structure of a 12-Item Human Immunodeficiency Virus Stigma Scale in Peruvian Adults

Supplemental material, sj-pdf-1-jpc-10.1177_21501319231197589 for Bifactor SEM and MIRT Structure of a 12-Item Human Immunodeficiency Virus Stigma Scale in Peruvian Adults by Cristian Ramos-Vera, Miguel Basauri-Delgado, Misael Diaz Peña, Jose Tinoco Alberto, Karen Perez Arroyo, Betsabel Herrera Mamani, Andy Sánchez-Villena and Jacksaint Saintila in Journal of Primary Care & Community Health

Footnotes

Acknowledgements

We acknowledge the contribution of the participants and co-researchers. Additionally, the authors would like to thank Dr. Varisier Noel during the manuscript writing process.

Author Contributions

CR-V and MAB designed and wrote the study profile. MDP, JTB, and KPA collected the data. CR-V, AS-V, and BHM performed statistical analysis. CR-V, MAB, and JS wrote the first draft of the manuscript. All authors contributed to the intellectual content of this manuscript and approved the final manuscript.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Consent to Participate

All participants provided written informed consent.

ORCID iD

Jacksaint Saintila

Data Availability Statement

Data supporting the conclusions of this research will be made available in coordination with the corresponding author.

Supplemental Material

Supplemental material for this article is available online.

References

World Health Organization. HIV. Global Health Observatory. 2022. Accessed October 14, 2022. https://www.who.int/data/gho/data/themes/hiv-aids

World Health Organization. Infección por el VIH. 2022. Accessed October 14, 2022. https://www.who.int/es/news-room/fact-sheets/detail/hiv-aids

Ministerio de Salud. Gobierno del Perú. VIH y SIDA. Prevención y control de VIH-SIDA, enfermedades de transmisión sexual (ETS) y hepatitis. Minsa. 2022. Accessed October 16, 2022. https://www.gob.pe/16439

Reinius

Wettergren

Wiklander

, et al. Development of a 12-item short version of the HIV stigma scale. Health Qual Life Outcomes. 2017;15:1-9.

Muelas Fernandez

Rojas Lievano

Perez Vidal

, et al. Prevalencia de diagnóstico tardío en infección por VIH. Med Clin. 2020;155:388-391.

Wanjala

Too

Luchters

, et al. Psychometric properties of the Berger HIV Stigma Scale: a systematic review. Int J Environ Res Public Health. 2021;18:13074. doi:10.3390/IJERPH182413074

Ayieko

Brown

Anthierens

, et al. ‘Hurdles on the path to 90-90-90 and beyond’: qualitative analysis of barriers to engagement in HIV care among individuals in rural East Africa in the context of test-and-treat. PLoS One.2018;13: e0202990.

Bilardi

Hulme-Chambers

Chen

, et al. The role of stigma in the acceptance and disclosure of HIV among recently diagnosed men who have sex with men in Australia: a qualitative study. PLoS One. 2019;14:e0224616.

Layer

Kennedy

Beckham

, et al. Multi-level factors affecting entry into and engagement in the HIV continuum of care in Iringa, Tanzania. PLoS One.2014;9:e104961.

10.

Earnshaw

Chaudoir

SR.

From conceptualizing to measuring HIV stigma: a review of HIV stigma mechanism measures. AIDS Behav. 2009;13:1160.

11.

Mahajan

Sayles

Patel

, et al. Stigma in the HIV/AIDS epidemic: a review of the literature and recommendations for the way forward. AIDS. 2008;22:S67.

12.

Gavan

Hartog

Koppenol-Gonzalez

, et al. Assessing stigma in low- and middle-income countries: a systematic review of scales used with children and adolescents. Soc Sci Med. 2022;307:115121.

13.

Berger

Ferrans

Lashley

FR.

Measuring stigma in people with HIV: psychometric assessment of the HIV stigma scale. Res Nurs Health. 2001;24:518-529.

14.

Franke

Muñoz

Finnegan

, et al. Validation and abbreviation of an HIV stigma scale in an adult spanish-speaking population in urban Peru. AIDS Behav. 2010;14:189-199.

15.

Gonçalves

Morgado

Sousa

, et al. HIV Stigma Scale: translation and validation of the short version to the Portuguese population. Psicologia. 2022;36:39-50.

16.

Luz

Torres

Almeida-Brasil

, et al. Translation and validation of the Short HIV stigma scale in Brazilian Portuguese. Health Qual Life Outcomes. 2020;18:1-12.

17.

Lindberg

Wettergren

Wiklander

, et al. Psychometric evaluation of the HIV stigma scale in a Swedish context. PLoS One. 2014;9:e114867. doi:10.1371/JOURNAL.PONE.0114867

18.

Bint-E-Saif

Shahzad

Urdu translation and adaptation of the HIV stigma scale in Pakistani injectable drug users with HIV. J Pak Med Assoc. 2020;70:505-510.

19.

Boyes

Mason

Cluver

LD.

Validation of a brief stigma-by-association scale for use with HIV/AIDS-affected youth in South Africa. AIDS Care. 2013;25:215-222.

20.

Jeyaseelan

Kumar

Mohanraj

, et al. Assessing HIV/AIDS stigma in south India: validation and abridgement of the Berger HIV Stigma scale. AIDS Behav. 2013;17:434-443.

21.

Kagiura

Fujii

Kihana

, et al. Brief HIV stigma scale for Japanese people living with HIV: validation and restructuring using questionnaire survey data. AIDS Care. 2020;32:1-9.

22.

Rongkavilit

Wright

Chen

, et al. HIV stigma, disclosure and psychosocial distress among Thai youth living with HIV. Int J STD AIDS. 2010;21:126-132.

23.

Yendewa

Kpaka

Sellu

, et al. Perceived and enacted HIV-related stigma in Eastern and Southern Sierra Leone: a psychometric study of an HIV Stigma Scale. medRxiv. Published online July 12, 2023. doi: 10.1101/2023.07.07.23292152

24.

Mlouki

Zammit

Ghammem

, et al. Validity and reliability of a modified short version of a stigma scale for use among Tunisian COVID-19 patients after quarantine: a cross-sectional study. Health Sci Rep. 2022;5:e520. doi:10.1002/hsr2.520

25.

De los Santos

JAA

. Development and psychometric evaluation of the HIV-Related Stigma in Health Facilities (HIV-SHF) Scale. J Health Sci Med Res. 2023;41:e2023922.

26.

Marbaniang

Borse

Sangle

, et al. Development of shortened HIV-related stigma scales for young people living with HIV and young people affected by HIV in India. Health Qual Life Outcomes. 2022;20:119.

27.

Lin

Imani

Griffiths

, et al. Validity of the Yale Food Addiction Scale for Children (YFAS-C): classical test theory and item response theory of the Persian YFAS-C. Eat Weight Disord. 2021;26:1457-1466.

28.

Armoon

Fleury

Bayat

, et al. HIV related stigma associated with social support, alcohol use disorders, depression, anxiety, and suicidal ideation among people living with HIV: a systematic review and meta-analysis. Int J Ment Health Syst. 2022;16:17. doi:10.1186/S13033-022-00527-W

29.

Wallace

Pack

Uhrig Castonguay

, et al. Validity of social support scales utilized among HIV-infected and HIV-affected populations: a systematic review. AIDS Behav. 2019;23:2155-2175.

30.

Gázquez-López

García-García

González-García

, et al. Validation of the attitudes towards people living with HIV/AIDS scale in nursing students. BMC Nurs. 2023;22: 1-11.

31.

Ato

López

Benavente

A classification system for research designs in psychology. An Psicol. 2013;29: 1038-1059.

32.

Bernal

Metodología de la Investigación. Administración. Economía. Humanidades y Ciencias Sociales. 3rd ed. Pearson Educación; 2010.

33.

Otzen

Manterola

Sampling techniques on a population study. Int J Morphol. 2017;35:227-232.

34.

Ramos-Vera

CA.

Un método de cálculo de tamaño muestral de análisis de potencia a priori en modelos de ecuaciones estructurales. Rev del Cuerpo Médico Hosp Nac Almanzor Aguinaga Asenjo. 2021;14:104-105.

35.

Ramos-Vera

Dávila-Campos

Castañeda-Barros

, et al. Estructura e invarianza factorial de una escala de perfeccionismo infantil adaptada para adolescentes. Rev Psicopatol Psicol Clín. 2021;26:141-154.

36.

Soper

. A-priori Sample Size Calculator for Structural Equation Models [Software].

37.

Hancock

Stapleton

Mueller

RO.

The Reviewer’s Guide to Quantitative Methods in the Social Sciences. Routledge; 2018.

38.

Mardia

K V.

Measures of multivariate Skewness and Kurtosis with applications. Biometrika. 1970;57:519.

39.

Schumacker

Beyerlein

ST.

Confirmatory factor analysis with different correlation types and estimation methods. Struct Equal Model. 2000;7:629-636.

40.

Tabachnick

Fidell

LS.

Using Multivariate Statistics. 7th ed. Pearson; 2019.

41.

Marsh

Hau

K-T

Wen

In Search of golden rules: comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler’s (1999) findings. Struct Equal Model. 2004;11: 320-341.

42.

Byrne

BM.

Testing for multigroup equivalence of a measuring instrument: a walk through the process. Psicothema. 2008;20:872-882.

43.

Zinbarg

Revelle

Yovel

, et al. Cronbach’s α, Revelle’s β, and Mcdonald’s ω H : their relations with each other and two alternative conceptualizations of reliability. Psychometrika. 2005;70:123-133.

44.

Raykov

Hancock

GR.

Examining change in maximal reliability for multiple-component measuring instruments. Br J Math Stat Psychol. 2005;58:65-82.

45.

Smits

IAM

Timmerman

Barelds

DPH

, et al. The Dutch Symptom Checklist-90-revised. Eur J Psychol Assess. 2015;31:263-271.

46.

Rodriguez

Reise

Haviland

MG.

Applying bifactor statistical indices in the evaluation of psychological measures. J Pers Assess. 2016;98:223-237.

47.

Dominguez-Lara

SA.

Evaluación de la confiabilidad del constructo mediante el Coeficiente H: breve revisión conceptual y aplicaciones . Psychol Av la Discip. 2016;10:87-94.

48.

Hammer

McDermott

Levant

, et al. Dimensionality, reliability, and validity of the Gender-Role Conflict Scale-Short Form (GRCS-SF). Psychol Men Masculinity. 2018;19:570-583.

49.

Reise

Bonifay

Haviland

MG.

Scoring and modeling psychological measures in the presence of multidimensionality. J Pers Assess. 2013;95:129-140.

50.

Rios

Wells

Validity evidence based on internal structure. Psicothema. 2014;26:108-116.

51.

Stucky

Edelen

. Using hierarchical IRT models to create unidimensional measures from multidimensional data. In: Reise SP, Revicki DA (eds). Handbook of Item Response Theory Modeling. Routledge; 2014: 201-224.

52.

McDonald

RP.

Test Theory: A Unified Treatment. Lawrence Erlbaum Associates Publishers. 1999. Accessed October 14, 2022. https://psycnet.apa.org/record/1999-02770-000

53.

Dunn

Baguley

Brunsden

From alpha to omega: a practical solution to the pervasive problem of internal consistency estimation. Br J Psychol. 2014;105:399-412.

54.

Chalmers

RP.

Mirt: a multidimensional item response theory package for the R environment. J Stat Softw. 2012;48:1-29.

55.

Baker

Kim

S-H.

The Basics of Item Response Theory Using R. 1st ed. Springer; 2017.

56.

Muñiz

Introducción a la Psicometría : teoría clásica y TRI. Pirámide; 2018.

57.

Maydeu-Olivares

Joe

Assessing approximate fit in categorical data analysis. Multivariate Behav Res. 2014;49: 305-328.

58.

Dueber

Bifactor indices calculator: a microsoft excel-based tool to calculate various indices relevant to bifactor CFA models. Educ Sch Couns Psychol Res Tools. 2016;1: 1-5. Accessed October 14, 2022. http://sites.education.uky.edu/apslab

59.

Dominguez-Lara

Rodriguez

Índices estadísticos de modelos bifactor. Interacciones Rev Av en Psicol. 2017;3:59-65.

60.

Gignac

GE.

The higher-order model imposes a proportionality constraint: that is why the bifactor model tends to fit better. Intelligence. 2016;55:57-68.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.39 MB