Development and Validation of Questionnaire to Measure Parents’ Knowledge, Attitude, and Practice on Self-Medication of Children with Antibiotics in Bandung, Indonesia

Introduction

The emergence of resistance among the most important bacterial pathogens is recognized as a major public health threat affecting humans on a global scale and should be considered a high priority by all healthcare workers and institutions. The irrational use of antibiotics, such as the wrong antibiotic or dose leads to a significant increase in the infection rate of multidrug-resistant bacteria.¹ This infection pertains to infectious diseases instigated by bacteria possessing a specific resistance to antimicrobial drugs. After the onset of drug-resistant bacterial infection, it substantially prolongs the patient’s duration of hospitalization and exerts an influence on the case fatality rate and overall medical expenditures.^2,3

The incidence of antibiotic resistance (AMR) is also a public health threat relating to children. According to WHO data, infections caused by multidrug-resistant (MDR) bacteria result in 700,000 deaths across all ages, of which around 200,000 are newborns.⁴ In Europe, MDR infections in pediatric patients may represent up to 30% of the total cases.⁵ In Southeast Asia, including Indonesia, 83% of children are affected by E. coli resistant to first-line antibiotics.⁶ This is because AMR is more difficult to treat and is related to a more severe and prolonged illness leading to longer hospitalization times, with a 20% increase in length of stay and a poorer outcome, raising mortality by up to 40% in MDR hospital-acquired infections.^7–10

Lack of knowledge among parents regarding the judicious use of antibiotics in managing common childhood illnesses is one of the reasons for increased AMR.¹¹ According to Karuniawati et al, 63% of 575 parents assumed that antibiotics could be used to treat viral infections and reduce fever.¹² In addition, some stated that antibiotics could be stopped when the disease was cured and reused in the future.¹² In another previous study, there were significant gaps in parents’ understanding of antibiotics. There is a belief held by some parents that antibiotics should be administered whenever a child encounters a fever, and shorter courses of antibiotics are more beneficial for their well-being.^13,14 Therefore, the development of psychometric properties to assess the knowledge, attitude, and practice (KAP) of parents regarding this issue is urgently needed.

This study aimed to develop and validate a questionnaire to measure the knowledge, attitude, and practice of parents regarding antibiotic use in children using structural equation modeling (SEM) analysis. This is the first analysis to investigate the psychometric properties of the Indonesian-parent KAP in the use of antibiotics in children. The questionnaire is used as a standard that can provide specifications, guidelines, and characteristics that can be used consistently. A validated standard is important to ensure accurate data, provide clear data, and ensure the legitimacy of research findings. The results provided data for stakeholders to better strategize interventional plans and implement necessary measures to reduce the impact of antibiotic misuse on children.

Methods

Questionnaire Design

Literature Review and Item Generation

The initial step in developing the study instrument was to identify the most representative variables in four domains as presented in Table 1. These were identified and selected based on a literature review of articles published in international journals from India,¹⁵ the United Arab Emirates,¹⁶ and Saudi Arabia¹¹ as well as according to the government guidelines regarding the Handbook of Antibiotic use in Indonesia.¹⁷ Based on these variables, a draft of the instrument for the parents’ assessment of self-medication of children with antibiotics consisting of 48 items grouped in five blocks. The first 9-item set was intended for social and demographic identification and the rest of them was a Likert scale. Scores were encrypted from 1 to 5 (“strongly disagree”, “disagree”, “doubtful”, “agree”, and “strongly agree”) to prevent intermediate values from being positive or negative on each item.

Table 1 The Critical Ratios of the Trajectories Between the Items of the 4 Latent Variables

Internal Review and Refinement

Upon identification of the conceptual concerns, a panel comprising seasoned experts from one clinical and four academic backgrounds engaged in a collaborative brainstorming session. The objective of this session was to formulate components derived from the item pool that align with the purpose of this study. Furthermore, the items were reviewed and the wording accurately captured the intended concept for gathering relevant information. The measures of readability, length of items, reversed items, and the format of measurement were also considered in the refinement process. The knowledge, attitude, practice, and facilitator assessments contained 14, 8, 12, and 4 items, respectively.

Structural Model Analysis

The statistical analyses were performed using the SPSS Version 26.0 software package for Windows (SPSS, Inc., Chicago, IL, USA), and AMOS version 26.0 was employed for confirmatory factor analysis (CFA) through SEM. A CFA and SEM were conducted using AMOS 26.0, to determine the constructing validity of the instrument. SEM was widely used in empirical study to investigate relationships among variables, while maximum likelihood estimation (MLE) robust extraction method was utilized to estimate parameters of an assumed probability distribution, given some observed data. To evaluate the global fitting quality, seven indices were used to assess model fit: 1) chi-square (Chisq) with the acceptance level of p>0.05, 2) ratio of chi-square and degrees of freedom (CMIN/DF): when the value of the equation was equal or smaller than two, lower than five, and greater than five, the fit was perfect, and unacceptable, 3) goodness of fit index (GFI) with recommended values greater than 0.9 were considered to represent good fits, 4) root mean square error of approximation (RMSEA) recommended values under 0.08 indicated perfect fits,¹⁸ 5) Tucker-Lewis Index (TLI) and Comparative fit Index (CFI) recommended values equal or greater than 0.95 represented good fits.¹⁹ Meanwhile, Normal fit index (NFI) had an acceptance level equal to or greater than 0.80.²⁰

Reliability and Validity of Measurement Models

The reliability and validity of reflective measurement models were analysed, and the process entailed evaluation of indicator reliability, internal consistency reliability, discriminant validity, and convergent validity. Composite reliability (CR) was assessed to measure internal consistency and indicator reliability of the reflective constructs.^21,22 The CR appraised the degree to which a set of individual variables in a particular construct offered their estimation of the construct. The recommended value of CR achieved by the construct was 0.6,^23–25 and AVE was used to assess the convergent validity. The result determined the level at which a group of variables represented the same construct. Hair et al recommended a value of 0.5, where both construct and individual variables had high validity.²⁴

Population Samples

The study population consisted of parents from Arcamanik District, Bandung, Indonesia. The inclusion criteria were as follows: parents with children aged 0–11 years old, ability to use and understand the Google Form platform, can read well and are willing to fill out informed consent for approval. Convenience sampling was applied to obtain a diverse pool of participants, and 3 minimum sample sizes for 4 latent variables were sufficient for SEM analysis, producing good output for estimating item location.²⁶

Data Collections

Data were collected with a questionnaire for two weeks, from January 5 to 19, 2023, through an on-site interview. The instrument was distributed to the parents in four sub-districts of Arcamanik, namely Sukamiskin, Cisaranten Endah, Cisaranten Bina Harapan, and Cisaranten Kulon through Posyandu (Indonesian community activity which integrated essential health services for mothers and children). Furthermore, respondents’ information from the Google Form and paper-based questionnaire was extracted, coded, and entered into SPSS version 26.0, then exported into Analysis of Moment Structures (AMOS) version 26.0 for confirmatory analysis through SEM.

Ethical Approval

The ethical approval was granted by the Health Research Ethics Committee of Universitas Harapan Bangsa No. B.LPPM-UHB/1807/05/2023. The respondents provided written informed consent for participation in the survey. This research was conducted ethically in accordance with the World Medical Association Declaration of Helsinki.

Results

Factor Analysis: Construct Validity

Construct validity was estimated using confirmatory factor analysis (CFA). The path diagrams used for the model were composed of circles or ellipses that represented unobserved variables. The single-headed and double-headed arrows represented the impact and correlation of one variable on another. Table 1 shows the critical ratios of the trajectories between the items of the 4 latent variables.

After the measurement model was assessed to examine reliability and validity, the suitability was also analyzed. The good fit indices implied that the result achieved an acceptable model fit according to normal practice.²⁰ In this study, all the fit indices indicated a good fit, with chi-square = 0.0004, CFI = 0.977, RMSEA = 0.044, CMIN/DF = 1.162, AGFI = 0.651, TLI = 0.973, and NFI = 0.860. However, the GFI parameter did not fit with the output value of 0.718 (the cut-off value is GFI > 0.95) (Table 2). At least four indices should be presented between the model and the incremental fit groups.²⁷ Overall, the result of the proposed model showed acceptable fits and the items were the best fit to the data, within a desirable range, as shown in Figure 1.

Table 2 Index Category and Levels of Acceptance

Figure 1 Representation of the model selected by CFA analysis.

Measurement Models’ Reliability and Validity

The reliability and validity of the reflective measurement model were also conducted, and the process entailed evaluation of indicator reliability, internal consistency, discriminant validity, and convergent validity. In this study, composite reliability (CR) was utilized to measure internal consistency and indicator reliability of the reflective constructs, which were found as follows: knowledge = 0.950, practice = 0.96, attitude = 0.961; and facilitator = 0.793, with a recommended CR value of 0.6 and above. The attainment of reliability was accomplished for the variables, characterized as heterogeneous but possessing similarities within each construct. Furthermore, 36 observed variable indicators in variables passed the item reliability test with loading factors > 0.50 (Table 3) with a cut-off value of 0.4.²⁴

Table 3 Satisfaction Constructs’ Reliability and Validity

Convergent and discriminant validities were used to examine the instruments. The assessment assured the variables were associated with the latent construct measured with the average variance extracted (AVE).^28,29 It was reported that all the dimensions in variables had an AVE value of 0.671, 0.733, 0.793, and 0.693 for knowledge, practice, attitude, and facilitator, as shown in Table 3. To attain validity, the value of the AVE must be greater than or equal to 0.5.³⁰ Therefore, it can be concluded that all the variables met the requirements of the validity and reliability tests.

Discussion

The global impact of antibiotic resistance, specifically in children, pervades all corners of the world, and when immediate and appropriate measures are not taken on a global scale, severe consequences in terms of fatalities and financial losses are anticipated. The knowledge, attitudes, and practices of parents constitute a crucial determinant among various factors contributing to the improper usage of antibiotics. Extensive study has consistently demonstrated that a significant proportion maintains the belief that antibiotics are efficacious in treating common colds in children and expediting their recovery from associated symptoms.^31–33 Therefore, lowering the level of antibiotic resistance depends on parents’ knowledge, attitudes, and practice toward the self-medication of children with antibiotics. The availability of a validated tool that can fit their populations should serve as a study instrument and enable program adaptation, as well as comparison across different settings.

This is the first study that investigates the psychometric characteristics of the Indonesian-parent KAP concerning antibiotic usage in children. Therefore, the psychometric characteristics of a KAP questionnaire were developed and validated to close this gap. The content validity was also confirmed by the scale and content indices. The confirmatory factor analysis (CFA) and test-retest reliability demonstrated the questionnaire’s construct validity and reliability. The high response rate and low percentage of unanswered questions were important indicators that the public embraced the questionnaire specifically the parent population.

Knowledge and attitudes were stable and unlikely to change quickly.³⁴ According to the results, the construct validity evaluation showed an acceptable fit for the proposed model, with chi-square = 0.0004, CFI = 0.977, RMSEA = 0.044, CMIN/DF = 1.162, AGFI = 0.651, TLI = 0.973, and NFI = 0.860. The items had a favorable range and provided the best fit to the data. Additionally, 36 observed indicator values in variables passed the item reliability test with loading factors > 0.50 and a cutoff value of 0.4. The results showed that the 36-item questionnaire used to gauge Indonesian-parent KAP on the use of antibiotics on children was reliable. The AVE values for all the dimensions of variables were 0.671, 0.733, 0.793, and 0.693 for knowledge, practice, attitude, and facilitator, respectively. According to these findings, the AVE value was greater than 0.5, meaning all the variables conformed with the validity test requirements.

Based on previous results, the questionnaire was regarded as a valid and trustworthy instrument. It facilitated the assessment of a parent’s knowledge, attitudes, and practice of self-medication with antibiotics for children. Even though the 36-item questionnaire was only utilized for this study, it was designed with the possibility of being used in other settings. The questionnaire also assisted in gathering information from both antibiotic and non-users, which was important for epidemiological studies on Knowledge, Attitudes, and Practices in antibiotic usage investigation.

The findings were also related to previous study conducted by Mallah et al, who validated the knowledge, attitude, and practice questionnaires for individual use of antibiotics in Spain.³⁵ The results led to the general adult population overwhelmingly approving of the questionnaire’s final version. This current study specifically validated the knowledge, attitudes, and practices questionnaires for parents of children who utilized antibiotics. The distinction should be noted since previous results focused on validating similar questionnaires solely for adults.

This study has several limitations, first, the findings could not accurately represent the entire population of Indonesian parents with children utilizing antibiotics due to small sample size. Therefore, further investigation of the questionnaire in various settings and with bigger sample sizes is required. Since the study involved the use of self-administered questionnaires, there is a potential for recall bias concerning antibiotic usage. Subjectivity and varying interpretations of the questions also affected the accuracy of parents’ reported knowledge, attitudes, and practices. To address these limitations, future studies should consider incorporating specific questions that restrict the duration of antibiotic use, aiming to minimize memory-related biases when assessing the actual antibiotic usage in children.

Conclusion

This study developed and validated a questionnaire for parents on the self-medication of children with antibiotics. The scales showed appropriate reliability and validity, including internal consistency, discriminant validity, and convergent validity. This scale was used to assess the knowledge, practice, attitude, and facilitators of parents in the use of antibiotics. The content validity of the questionnaire was also confirmed by the scale and content indices. The confirmatory factor analysis (CFA) and test-retest reliability demonstrated the questionnaire’s construct validity and reliability. The high response rate and low percentage of unanswered questions were important indicators that the public embraced the questionnaire. Even though the questionnaire is indeed a valid and reliable tool, a more extensive sample size of parents who have strictly adhered to a limited duration of antibiotic use should be gathered to mitigate any potential biases.