GENERALIZED CONFIRMATORY FACTOR ANALYSIS FOR KNOWING IMPACT OF KNOWLEDGE, ATTITUDES, AND BEHAVIORAL FACTORS HIV/AIDS IN INDONESIA

ABSTRACT


INTRODUCTION
HIV (Human Immunodeficiency Virus) is a type of virus that infects white blood cells and causes a decrease in the human immune system. AIDS, or Acquired Immune Deficiency Syndrome, is a collection of symptoms of diseases that arise due to decreased immunity caused by infection with HIV. A decrease in the immune system causes a person to be easily exposed to various infectious diseases that are often fatal. People with HIV need treatment with Antiretroviral (ARV) to reduce the amount of HIV virus in the body so it doesn't enter the AIDS stage, while people with AIDS require ARV treatment to prevent opportunistic infections with various complications [1]. Sexuality has long been a taboo subject, but it has undergone sweeping changes over centuries. On top of that, the emergence of HIV-AIDS has overwhelmed the entire world. The specific issue of AIDS is definitely a problem that has no short-term solution [2]. The long-term social consequences of HIV/AIDS would certainly add to the gravity of the current health problem. HIV/AIDS contributes to the first trajectory involved in state collapse in that care for sufferers adds to the resource burden countries face and intensifies competition over scarce resources [3].
In Indonesia, the cumulative number of HIV/AIDS cases detected in the January -March 2021 period was 9,327, consisting of 7,650 HIV and 1,677 AIDS reported by 498 districts and cities from 514 districts and cities. The Ministry of Health reported that the number of HIV cases decreased 16.5%, on the other hand, AIDS cases increased by 22.78% [4]. In Indonesia, the main transmission of HIV-AIDS in Indonesia is the use of unsafe injecting needles among users of narcotics and illegal drugs (drugs), followed by heterosexual intercourse, and transmission from mother to fetus during pregnancy, childbirth, or breastfeeding [5].
Previous research has shown that the clinical condition of HIV/AIDS patients was significant to predisposing factors, supporting factors, and reinforcing factors, with each loading factor value is 0.342, 0.544, and 0.143, respectively. Predisposing factors include people's knowledge and attitudes towards health, traditions, and public trust in matters relating to health, the value system adopted by the community, education level, socio-economic level, and so on [6]. In this predisposing factor is measured through indicators of knowledge, attitudes, and self-concept. Enabling factors include the availability of facilities and infrastructure or public health facilities. These facilities basically support or enable the realization of Health Behaviors [6]. These contributing factors include ARV therapy. Reinforcing factors includes the attitude and behavior factor of community leaders, religious leaders, and officials' Health [7].
Meanwhile, research by [8] shows that the five dimensions of collaboration significantly describe the function of collaboration in the prevention and control of HIV/AIDS. The governance dimension is the most important contribution dimension, followed by the organizational dimensions of autonomy, administration, and cohesion. The dimension norm is the smallest dimension contribution. This can be said as an external factor. The study by [9] highlights several misconceptions about HIV transmission, intolerance, stigma and discrimination against PLHIV, and risky sexual practices, which need to be addressed. HIV/AIDS related education programs should include specific interventions to change practices, together with knowledge and attitudes.
Internal factors such as knowledge, attitudes, and behavior also influence the transmission of HIV/AIDS. Previous research [10] shows that adequate knowledge of HIV/AIDS enables a person to make an early diagnosis and protect himself from the risk of transmission. Knowledge affects the formation of attitudes and behavior because, from experience and research, it turns out that behavior based on knowledge will last longer than behavior that is not based on knowledge [10]. Although students are knowledgeable about HIV/AIDS, they have little personal concern about becoming infected and do not take appropriate safe sex precautions. The findings of the present study show that gender, ethnic background, and knowing someone infected by HIV/AIDS influence students' level of concern about infection [11]. The 2017 IDHS report presents data and information related to knowledge, perceptions, and behaviors about HIV-AIDS at the national and provincial levels according to background characteristics such as demographic, social, and economic aspects [1].
Through the CFA (Confirmatory Factor Analysis) method, confirmation of the theory is carried out to measure the accuracy of the parameters. In measuring the latent variable, the t-test statistic is used as the significance of the indicator. This is because the standardized estimate regression in the CFA is the loading factor ( ) [12]. The CFA approach is used to determine the indicators that have the greatest influence on the latent variables of knowledge, attitudes, behavior about HIV/AIDS in Indonesia. Previous research that conducted by Gusti [7] used FIMIX-PLS with several factors including predisposing factors with indicators in the form of knowledge, attitudes and self-concept which gave the result that the clinical picture of HIV/AIDS patients was significant to these factors.
The purpose of this research is to support the National Research Master Plan (RIRN) 2017-2045 and the National Research Priority (PRN) 2020 -2024, which stipulates national research products for the health sector [4], namely infectious diseases that are still dominant. In addition, as an illustration of which indicators are the most influential in increasing the rate of recovery. The results of the study are expected to contribute to statistical science in the health sector, especially in supporting government programs, namely the achievement of three zero HIV/AIDS elimination by 2030, namely no more transmission of new HIV infections, no more deaths from AIDS, and no more stigma and discrimination against people living with HIV/AIDS (ODHA).
Compared to previous studies, this study uses Generalized Confirmatory Factor Analysis, which is suitable for IDHS survey data with a nominal data scale. From this analysis, it can also be developed into GSEM (Generalized Structural Equation Modeling) analysis which is a combination of SEM and GLM [13].

Data
The data used is secondary data from the Indonesia Demographic and Health Survey (IDHS) 2017 jointly carried out by the Central Statistics Agency (BPS), the National Population and Family Planning Agency (BKKBN), and the Ministry of Health (Kemenkes). Data collection took place from 24 July to 30 September 2017 in all regions of Indonesia. The implementation of the IDHS uses 4 (four) types of questionnaires, namely household questionnaires, women of childbearing age, married men, and teenage boys, where women of childbearing aged 15-49, never married aged 15-24. Of the 1,650 households selected by the IDHS, 1,607 households were found, and of these 1,594 or 99.2% of the households were successfully interviewed. Good data can be obtained from the measurement process using research instruments, generally in the form of a questionnaire [14].

Confirmatory Factor Analysis (CFA)
CFA is an analysis to determine whether some indicator variables represent a construct. The purpose of the CFA is to confirm the existing theory in measuring the accuracy of the parameters. CFA defines the relationship between measured and latent variables in loading factors ( ) [15]. Therefore, the significance of the indicator in measuring the latent variable was determined using the t-test statistic [16]. The hypothesis used is as follows.

=̂(̂)
(1) Decision: If │ │ > /2, , then 0 is rejected, and it can be concluded that is significant in measuring the latent variable. The value of construct reliability (pc) is used to determine the reliability of the latent variable which is calculated by Equation (4).
If pc ≥ 0.7, then the latent variable is said to be reliable [17].

Goodness of Fit
Goodness of fit models are taken from 4 measures, namely Comparative Fit Index (CFI), Tucker-Lewis Index (TLI), Root Mean Square Error of Approximation (RMSEA), and Standardized Root Mean Residual (RMSR). The explanation regarding 4 measures can be seen in Table 1  The model is said to be appropriate if it has a CFI value of more than 0.90, a Tucker-Lewis Index value of more than 0.90, an RMSEA value of less than 0.08, and an RMSR value of less than 0.09. The model is said to be inappropriate if it has a model suitability test value that does not match the criteria.

Research Variables
The variables of knowledge, attitude, and behavior about HIV/AIDS are the variables of this study as described in Table 2.

Descriptive Statistics
Characteristics of respondents can be known through several variables shown in the following pictures Based on Figure 1, it can be seen that the largest distribution of provincial origins came from West Java Province, which was 5,090 people (10.26%). A total of 3,729 people (7.51%) were respondents from East Java Province. A total of 3,414 people (6.88%) were respondents from Central Java Province. Based on Figure 2, it can be seen that the distribution of the age range of respondents is 7,936 people (15.99%) are respondents aged 15-19 years. A total of 6,830 people (13.76%) were respondents aged 20 -24 years. A total of 6,785 people (13.67%) were respondents aged 25-29 years. Based on Figure 3, it can be seen that the type of rural settlement is 23,202 people (46.8%) and as many as 26,425 people (53.2%) are respondents who live in urban areas. Based on Figure 4, it can be seen that the most recent education taken is 27,340 people (55.09%) who are respondents with the highest education in secondary schools. Based on Figure 5, it can be seen that the literacy level of most research respondents is 46,185 people (93.06%) are respondents who can read all sentences.

Confirmatory Factor Analysis
Before performing the confirmatory factor analysis, the goodness of fit was tested. If the model meets the criteria of goodness, then the model is said to be good. The results of the model goodness test are presented in Table 3.  Table 3 shows the results of the goodness of fit test. The latent variables knowledge and attitude have CFI and TLI values less than 0.90. The RMSEA and RMSR values are more than 0.08 or do not meet the criteria for the goodness of the model, so it can be said that the model does not fit.
The latent variable Behavior has a CFI and TLI value less than 0.90. The RMSEA value is more than 0.08, while the RMSR is less than 0.08. Only the RMSR criteria meet the goodness of the model. The conclusion can be said that the model does not fit.

CFA of Knowledge Latent Variables
The loading factors value of each indicator on three variables can be seen in Table 4. As seen in Table 4, indicator P2.14 has the highest loading factor value of 0.915, which means that the influence of indicator P2.14 on latent variable Sources is 91,5% compared to 13 other indicators. This shows that the statement P2.14, namely "Source for AIDS knowledge: other," has the greatest impact and is important to "How did people know about the HIV-AIDS virus." On the latent variable mother and child, indicator P8.1 has the highest loading factor value of 0.837. This means that the influence of indicator P8.1 on the latent variable mother and child is 83.7% compared to 2 other indicators. This shows that the statement P8.1, namely "HIV transmitted during pregnancy," has the greatest impact and is important to "Can the virus that causes HIV-AIDS be transmitted from a mother to her child: a) pregnancy, b) birth, c) breastfeeding." The indicator on latent variable infection that has the highest loading factor value of 0.790 is P11.5, which means that the influence of indicator P11.5 on latent variable infection is 79% compared to 8 other indicators. This shows that the statement P11.5, namely "Know STI: Chlamydia," has the greatest impact and is important to "What infections do you know of?" Indicator P10 has the highest loading factor value of 0.979, which means that the influence of indicator P10 on latent variable risk is 97.9% compared to 9 other indicators. This shows that the statement P10, namely "Ever heard of a Sexually Transmitted Infection (STI)," has the greatest impact and is important to latent variable risk.
Furthermore, it will be seen among several variables, source, mother and child, infection, and risk, which has the greatest influence on the latent variable knowledge. Variable infection has the highest loading factor value of 0.613, which means that the influence of infection on latent variable knowledge is 62.3% compared to 3 others variables. This shows that the statement Infection "What infections do you know of?" has the greatest impact and is important to latent variable knowledge.

Source
Mother and child Infection Risk   As seen in Table 6, the indicator S1.4 has the largest loading factor value of 15,849, which means that the influence of the indicator S1.4 on the latent variable identification is 1584,9% compared to the other 4 indicators. This shows that statement S1.4, namely "Identifying someone with HIV-AIDS using other methods," has the most influence big and important way to identify someone with HIV-AIDS. While the value of loading factors is less than 0.5 which causes the variables S1.1, S1.2, S1.3, and S1.5 to have no significant effect. This shows that S1.1 (identify through physical changes), S1.2 (identify through changes in behavior), S1.3 (identify through blood tests), and S1.5 (don't know how to identify) have no significant effect on latent variable identification.

Knowledge
Indicator S4 has the highest loading factor value of 0.524, which means that the influence of the S4 indicator on the latent variable habit is 52.4% compared to the other 8 indicators. This shows that the statement S4, namely "Would want HIV infection in family to remain secret", has the greatest and most important influence on the latent variable habit. While the loading factor value is less than 0.5 which causes the variables S2, S3, S6, S7, S8, S9, S10 to not have a significant effect on the latent variable habit.
Furthermore, it will be seen among several variables, identification and habit, which has the greatest influence on the latent variable knowledge.  Table 8. Loading factor variables information, symptom, symptom (men), and knowing As seen in Table 8, indicator R7.6 has the highest loading factor value, which is 0.922, which means that the influence of indicator R7.6 on the latent variable information is 92.2% compared to the other 12 indicators. This shows that statement R7.6, namely "Source for STI knowledge: religious institutions," has the greatest and most important influence on "Where did you get information about sexually transmitted infections (STI)?"

Loading
On latent variable symptom, indicator R8.15 has the highest loading factor value of 0.937, which means that the influence of indicator R8.15 on latent variable symptoms is 93.7% compared to 15 other indicators. This shows that the statement R8.15, namely "Male STI symptoms: no symptoms: has the greatest impact and is important to If a man contracts a sexually transmitted infection (STI), what are the symptoms?" The indicator that has the highest loading factor value of 0.951 is R9.15, which means that the influence of indicator R9.15 on latent variable symptoms (men) is 95.1% compared to 15 other indicators. This shows that the statement R9.15, namely "Female STI symptoms: no symptoms", has the greatest influence and is important to "If a woman contracts a sexually transmitted infection (STI), what are the symptoms?" The R5 indicator has the highest loading factor value of 0.746, which means that the R5 indicator has a large influence on the latent variable knowing, which is 74.6% compared to the other 5 indicators. This shows that R5's statement, namely "People talk badly about people with or believed to have HIV", has the biggest and most important influence on latent variable knowing.
Furthermore, it will be seen among several variables, information, symptoms, symptoms (men), knowing, which has the greatest influence on the latent variable Behavior. Based on the results of these calculations, the CR value for the variable knowledge is obtained of 0.999048 or more than 0.7. It can be concluded that the variable Knowledge has already reliable and consistent. In the same way, the CR value of the variable behavior is equal to 0.999606. The two latent variables describe a fairly high consistency or are already reliable. Variable consistency shows that significant indicators are indicators that are robust, aligned and corresponding to the formation of latent variables.

CONCLUSIONS
The conclusion obtained from this study is that the distribution of the most respondents from the province of West Java is 5,090 people (10.26%). The top three age distributions of the respondents were 7936 people (16%) aged 15-19, 7611 people (15.3%) aged 35-39, and 7190 people (14.5%) aged 30-34. Most respondents (55.1%) were at the secondary level for the highest education level, living in urban areas (53.2%), having the poorest wealth index (22.2%), and being able to read the whole sentence (93.1%). The CFA results show that the variable P11 (about known infections) has the largest loading factor value, which is 0.613 on the knowledge latent variable. In the latent variable attitude, the variable S1 (about the identification of how respondents know someone is infected with HIV-AIDS) has the largest loading factor value of 0.514. While the behavioral latent variable, the variable R8 (if a man has ever been infected with a sexually transmitted disease (STI), what are the symptoms) has the largest loading factor value, which is 0.954. The largest loading factor value shows the most influential indicator. Suggestions for further research are to determine the relationship between latent variables using the Generalized Structural Equation Modeling (GSEM) method, so that the direction of the relationship between the three latent variables in the aggregate is known.