Community preventive behaviour and perception on the severity of COVID-19 disease in Indonesia, 2021-2022: Structural equation modelling

Version Changes

Revised. Amendments from Version 2

There are no major difference between the previous version with the latest. We added a statement in part of HBM conceptual, we added one reference in discussion as suggested, and we added statement on data collection limitation. As additional, as we reviewed the references, we also added several incomplete references (which was marked with blue text before).

Abstract

Background

This study investigated the determinants of community preventive behavior in complying with the Indonesian regulations to prevent COVID-19 local transmission.

Methods

A cross-sectional study used to collect the data via an online cross using a form created from a google questionnaire forms. A total of 1,802 respondents were gathered at a single point in time. The authors used the Health Belief Model (HBM) approach to measure and create a model for the prevention of local transmission of COVID-19.

Results

This study found that more than half of the respondents still had low perceived susceptibility (16%) and severity (43%). There were only 3% respondents with perceived barriers and 19% with strong self-efficacy. The findings showed that self-efficacy and perceived barriers had statistically significant relationships with preventive behavior (p-value <0.05). The goodness of fit index showed that the proposed model was not fit for the data (RMSE<0.080, GFI>0.950, AGFI>0.950, SRMR<0.100), which means that it was not fit to describe the empirical phenomenon under study.

Conclusions

This study found that more than half of the respondents still had low perceived susceptibility (84%) and severity (67%), but more than half had high perceived benefits (54%). Only a few respondents had significant barriers to implementing COVID-19 transmission prevention behaviours (3%). Still, most respondents had low perceived self-efficacy (81%), and only 60% had good behaviours related to COVID-19 prevention. In the context of COVID-19 preventive behaviour, we recommended to improve perceived susceptibility and severity by providing the correct information (which contain information about how people susceptible to the virus and the impact of infected by the virus) with the local cultural context.

Abbreviations

A

Age

AVE

Average Variance Extracted

AGFI

Adjusted Goodness of Fit Index

BEN1-5

Items for perceive Benefit

BAR 1-5

Items for perceive Barrier

BHV1-6

Items for Behaviour

CFI

Comparative Fit Index

E

Education

G

Gender

GFI

Goods of Fit Index

HBM

Health Belief Model

IBM

International Business Machines

IFI

Incremental Fit Index

KAP

Knowledge Attitude Practice

MI

Modification Indices

NFI

Normed Fit Index

O

Occupation

PNFI

Parsimonious Normed Fit Index

PGFI

Parsimonious Goods of Fit Index

RFI

Relative Fir Index

R

Residence

RMSEA

Root Mean Square Error of Approximation

SEM

Structural Equation Model

SPSS

Statistical Package for The Social Sciences

SRMR

Standardized Root Mean Square

S1-5

Items for perceive Severity

SE1-5

Items for perceive Self-Efficacy

UNICEF

United Nations International Children Fund

WHO

World Health Organization

X1-14

Items for perceive Susceptibility

3M

Wearing mask, washing hand, and social distancing

Introduction

The world is currently besieged by the COVID-19 pandemic. ¹ As of February 7, 2022, there were 394,381,395 confirmed cases of COVID-19, with the World Health Organization (WHO) reporting 5,735,179 deaths. ² The COVID-19 pandemic in Indonesia from mid-2021 until the end of 2021 reached its peak (second wave) when it was dominated by the Delta variant with the total confirmed cases was 4.254.443 and 143.766 death (CFR: 3.4). ³

COVID-19 prevention regulations largely depend on community compliance and behavior. While changing behavior is a significant challenge in health interventions. An AC Nielsen survey (2020) in six major cities with a total of 2,000 respondents, in collaboration with the United Nations International Children’s Emergency Fund (UNICEF), found that less than one-third (31.5%) of all respondents practiced all three-preventive activities (wearing masks, washing hands, and social distancing). Over one-third (36%) practiced only two of the three-preventive behaviors, less than one-fourth (23.2%) practiced only one of the three behaviors, and almost one-tenth (9.3%) did nothing at all. ³ A number of studies described the associations between socio-demographic characteristics and people’s levels of perceptions about the severity of COVID-19 with their preventive behavior. ^{3 – 10}

Based on the official government report (see https://covid19.go.id/monitoring-kepatuhan-protokol-kesehatan), some areas of Indonesia showed a compliance rate of only 60% till I January 2022. ¹¹ The aim of this study was to examine changes in community behaviour to prevent local COVID-19 transmissions and changes in community perceptions about the severity level of COVID-19. The following research questions were formulated: 1) What constitutes as the community perceptions of COVID-19? 2) Is there an association between community perception and community preventive behaviour using the Health Belief Model (HBM) approach? And 3) what kind of model of preventive behaviour on COVID-19 can be predicted using the structural equation model (SEM)?

Methods

Study design and samples

This was a cross-sectional study using Google forms with structured survey questionnaires. The questionnaires have been tested with 30 respondents before the survey was conducted. The proportion of a large population was used to figure out the sample size for a variable, which was used to figure out the minimum sample size. ¹² The margin error=5%, p=50%, and Zα=95%. The minimum sample size for this study was 385 respondents. The data collection was conducted over two weeks from July 14–26, 2021, using WhatsApp, Line, and Telegram. To expand the coverage throughout Indonesia, social media influencers were asked to distribute the survey through Twitter and Instagram. The respondents who participated in this study were aged 15–62 and Indonesian citizens. On average, it took 20–25 minutes to complete the form. Respondents signed informed consent forms before completing the survey. The total final number of respondents was 1,802.

Conceptual model

We proposed a comprehensive approach to understanding behaviour change using the Health Belief Model rather than study it partially. The HBM offers a holistic perspective on behaviour change processes. However, when employing HBM as a theoretical framework, there were latent variables that did not directly observed. These latent variables are represented by various indicators/observe variable. Consequently, Structural Equation Modelling (SEM) is the most suitable analytical method for examining relationships between variables (latent variable and measurement/indicator) within the HBM framework, as it allows for the analysis of both latent variables and the overall model as a single, cohesive unit.

The HBM considers several main concepts/theories that predict why people will take action to prevent, including individual characteristics, perceived susceptibility, severity, benefit, and self-efficacy. ¹³ Since this study aims to observe the behavior change especially preventive health behavior on COVID-19, the HBM conceptual is one of the complete and holistic frameworks that can be used to figure out how a behavior occurs and change. Theories and models serve the purpose of explaining behavior and offering strategies for promoting behavioral changes. An explanatory theory, serves to elucidate and understand the root causes of a problem. These theories also forecast behaviors in specific circumstances and assist in identifying factors that can be altered, such as knowledge, attitudes, self-efficacy, social support, and resource availability. Theories and models can be applied to investigate the reasons behind people not adhering to public health or medical advice and neglecting their health. They can aid in pinpointing the essential information required before developing and structuring an intervention program. ¹⁴

Individual characteristics and community perception of COVID-19

Individual characteristics were represented by residence (R), age (A), gender (G), educational level (E), and occupation (O). Perceived susceptibility, severity, benefit, barriers, and self-efficacy are latent variables in this study. All questions in in the questionnaire in each latent variable were answered using a five-point Likert scale: 1 (strongly disagree), 2 (disagree), 3 (neutral), 4 (agree), and 5 (strongly agree). Observed variables X1-14, S1-5, BEN1-5, BAR1-5, and SE1-5 were used to measure perceived susceptibility, severity, benefit, barriers, and self-efficacy, in that order.

Community preventive behavior against COVID-19 transmission

Community preventive behavior was represented by six indicators in Figure 1 which include frequency of hosting (BHV1), frequency of visiting others (BHV2), frequency of work/study from office/school (BHV3), frequency of handshaking (BHV4), frequency travelling to a red zone (BHV5), and frequency of leaving the house when you are not feeling well (BHV6).

Figure 1. Proposed structural model of community preventive behavior against COVID-19. ¹⁵ .

Data analysis

This study is a kind of behavioral sciences that intended to study theoretical construct which is cannot be observed directly- called as latent variables. Thus, we operationally define the latent variable of interest in terms of behavior believed to represent it. In this study, the latent variable includes perceived susceptibility, perceived severity, perceived benefit, perceived barriers, and perceived self-efficacy and for each latent variables, there are observed variables which is based on theory are representing each of latent variables. SEM analysis allowed us to not only analyze observed measurement variable, but it can incorporate both unobserved (latent) and observed variables. One of the specialties of SEM is the hypothesized model can then be tested statistically in a simultaneous analysis of the entire system of variables to determine the extent to which it is consistent with the data. If goodness-of-fit is adequate, the model argues for the plausibility of postulated relations among variables; if it is inadequate, the tenability of such relations is rejected. ¹⁶ The sample size required using SEM analysis in this study is sufficient (n=1802). The minimum sample size for SEM analysis with seven or less constructs and no under identified constructs is 150. ¹⁷

The authors used Lisrel version 8.8 software to construct the covariance-based SEM. SEM analysis can also be done using R. Steps of doing SEM analysis using R as follows ¹⁸ :

• a.Draw model
• b.Input data in the form of covariance or correlation matrix
• c.Identify the model
• d.Assess parameter estimates
• e.Assess fit measure (chi-square, degree of freedom, residual matrix, GFI, RMSEA)
• f.Check the modification indices
• g.Rerun the model till we get the best fit of the data to the model ad theory.

The six latent variables were perceived susceptibility, severity, benefit, barriers, self-efficacy, and preventive behavior. The 45 observed variables were presented in Table 1 include construct that build by all the observed variables.

Table 1. The observed variables.

No	Item	Scale	Description	Construct
1	R	1. City 2. Village 3. Housing area 4. Apartment	Residence	-
2	A		Age	-
3	G	1. Male 2. Female	Gender	-
4	E	1. No education 2. Elementary school 3. Junior high school 4. Senior high school 5. University	Education level	-
5	O	1. Unemployment 2. Labourer/employee 3. Student/college 4. Housewife 5. Entrepreneur 6. Pensionary 7. Others	Occupation	-
6	X1	1-5 Likert strongly disagree – strongly agree	I am at risk when not keeping a distance of at least 1-2 metres when in the work area.	Perceive Susceptibility
7	X2		I am at risk when not keeping a distance of at least 1-2 metres when at school.
8	X3		I am at risk of not keeping a distance of at least 1-2 metres when at the mall/supermarket.
9	X4		I am at risk when not keeping a distance of at least 1-2 metres when at the traditional market.
10	X5		I am at risk when not keeping a distance of at least 1-2 metres at a restaurant/coffee shop/café.
11	X6		I am at risk when not keeping a distance of at least 1-2 metres when at a touristdestination.
12	X7		I am at risk when not keeping a distance of at least 1-2 metres when at a wedding reception.
13	X8		I have a low risk of getting COVID-19 if I avoid using a bus/taxi.
14	X9		I have a low risk of getting COVID-19 if I avoid using a commuter line.
15	X10		I have a low risk of getting COVID-19 if I avoid using a small bus.
16	X11		I have a low risk of getting COVID-19 if I avoid using motorcycle taxis/online.
17	X12		I have a low risk of getting COVID-19 if I practice work/study from home.
18	X13		I have a low risk of getting COVID-19 if I practice avoiding handshaking.
19	X14		I have a low risk of getting COVID-19 if I practice avoiding travel to a red zone (an area with a high positive rate).
20	S1		I may get COVID-19 and cause health impacts if I do not keep a distance ofat least 1-2 metresina public area.	Perceived severity
21	S2		I may get COVID-19 and cause health impacts if I do not avoid using public transportation.
22	S3		I may get COVID-19 and cause health impacts if I do not work/study from home.
23	S4		I may get COVID-19 and cause health impacts if I do not avoid handshaking.
24	S5		I may get COVID-19 and cause health impacts if I do not avoid travel to a red zone.
25	BEN1		I feel safe from COVID-19 if I keep a distance of at least 1-2 metresin a public area.	Perceived benefit
26	BEN2		I feel safe from COVID-19 if I avoid using public transportation.
27	BEN3		I feel safe from COVID-19 if I work/study from home.
28	BEN4		I feel safe from COVID-19 if I avoid handshaking.
29	BEN5		I feel safe from COVID-19 if I avoid travel to a red zone.
30	BAR1		I find it difficult to stay at least 1-2 metres from people in a public area.	Perceived barrier
31	BAR2		I find it difficult to avoid using public transportation.
32	BAR3		I find it difficult to work/study from home.
33	BAR4		I find it difficult to avoid handshaking.
34	BAR5		I find it difficult to avoid travel to a red zone.
35	SE1		It is easy for me to avoid travel to a red zone.	Perceived self-efficacy
36	SE2		It is easy for me to avoid handshaking.
37	SE3		It is easy for me to work/study from home.
38	SE4		It is easy for me to avoid using public transportation.
39	SE5		It is easy for me to keep a distance of at least 1-2 metres when in a public area.
40	BHV1		Frequency of hosting.	Behaviour
41	BHV2		Frequency of visiting others.
42	BHV3		Frequency of work/study from office/school.
43	BHV4		Frequency of handshaking.
44	BHV5		Frequency travelling to a red zone.
45	BHV6		Frequency of leaving the house when you are not feeling well.

Descriptive statistics were presented as numbers and percentages for individual characteristics, and bar charts for perceptions and behavior. Descriptive statistical analysis was performed using IBM Statistical Package for the Social Sciences (SPSS) version 27 and presented using Microsoft Excel. The descriptive analysis simply can also be completed using Microsoft Excel. The SEM analysis was carried out using the following steps:

• 1)Identifying the degree of freedom of the structural model.
• 2)Assessing construct validity.
• 3)Assessing construct reliability.
• 4)Assessing structural model validity. ^{14 , 19}

Perceived was categorized based on the total score. Good/high perceived were those who chose to “strongly agree” on every question. Regarding behavior, good behavior occurs if the respondent states never or rarely. Rarely was included in the category based on the assumption that people might be challenged to practice preventive behavior related to other factors that require them to leave, such as the environment or critical social activities that cannot be abandoned.

Results

The study results are presented in several parts sequentially, starting with the respondent’s characteristics, followed by the descriptive statistics of the independent variable (community perception of the level of severity of COVID-19 disease), the descriptive statistics of the dependent variable (composite variable of community behavior), and the SEM analysis results.

Most of the respondents lived in the city (46.1%) and a housing area (36.1%), and only a few lived-in villages (16.6%). A large majority of the respondents were women (74.5%), and approximately 80.0% were students and workers ( Table 2).

Table 2. Respondents’ characteristics.

	Frequency	%
Resident
City	831	46.1
Village	299	16.6
Housing area	651	36.1
Apartment	21	1.2
Sex
Male	460	25.5
Female	1342	74.5
Education Level
No education	3	0.2
Elementary school	3	0,2
Junior high school	25	1.4
Senior high school	885	49.1
University	886	49.2
Occupation
Unemployment	128	7.1
Laborer/employee	552	30.6
Student/college	958	53.2
Housewife	35	1.9
Entrepreneur	106	5.9
Pensionary	3	0.2
Others	20	1.1

In terms of perceived susceptibility ( Table 3), more than half of respondents (>50%) chose to strongly agree to practice recommended behaviors such as social distancing at least 1-2 meters in public areas; almost half of them strongly agree that they had a low risk of getting COVID-19 when avoiding using public transportation, practicing work/study from home, handshaking, and travelling to a red zone. Most respondents also strongly agree that if they are exposed to COVID-19, it will affect their health if they do not practice recommended behaviors (perceived severity - Table 3). Likewise, concerning perceived benefit ( Table 3), more than 50% of respondents strongly agree that they can avoid getting COVID-19 if they practice these behaviors. In terms of perceived barriers, less than 20% felt that it would be challenging/difficult to implement the recommended behaviors ( Table 3). Regarding self-efficacy, this is defined as the conviction that one can successfully practice a certain behavior, ¹³ the survey showed that less than 50% were confident that they could implement the recommended behavior ( Table 3).

Table 3. Distribution (%) of perceived items (N=1802).

Perceived	1 (%)	2 (%)	3 (%)	4 (%)	5 (%)
Perceived Susceptibility
I am at high risk when I am not keeping a distance of at least 1-2 m when in a public area
Working area	3.5	2.1	6.3	27	61.2
School	3.3	2.6	8.0	28	58.1
Mall/supermarket	3.1	1.7	5.7	22.6	67
Traditional market	2.9	1.6	5.5	20.5	69.5
Restaurant/coffee shop/café	2.7	2.6	7.5	27.1	60
Tourism objects	3.4	3.2	10.5	27.1	55.8
Wedding reception	2.9	1.8	6.4	22.1	66.8
I have a low risk of getting COVID-19 if I avoid using public transportation
Bus/taxi	6	6	11.4	25.9	50.8
Commuter line	6.4	5.6	12.2	23.9	51.9
City transport	6.4	5.1	9.5	23.5	55.5
Offline motorcycle taxi/online	5.1	10.8	28.4	30.7	25
I have a low risk of getting COVID-19 if I practice following behaviours
Work/study from home	3.4	0.9	5.2	19.3	71.2
Avoid handshaking	2.7	0.9	3.4	19	74
Avoid travel to red zone (are with high positive rate)	2.6	1	3.3	16.6	76.5
Perceived Severity
I may get COVID-19 and cause health impacts to myself if I do not follow behaviors
Keep distance at least 1-2 metre when at public area	3.4	1.6	4.9	26.2	63.9
Avoid using public transportation	3.8	2.9	13.5	29.3	50.4
Work/study from home	5.0	2.4	7.3	27.9	57.4
Avoid handshaking	4.3	1.9	4.2	24.5	65.1
Avoid travel to red zone (are with high positive rate)	4.6	1.6	4.2	21.0	68.6
Perceived Benefit
I feel safe and get off COVID-19 if I practice following behaviours
Keep distance at least 1-2 metre when at public area	2.6	1.1	4.8	23	68.5
Avoid using public transportation	2.3	1.6	9.3	23.9	62.8
Work/study from home	2.6	1.3	5.9	22.4	67.8
Avoid handshaking	2.6	1.2	4.6	20	71.8
Avoid travel to red zone (are with high positive rate)	2.5	1.2	4.3	18.2	73.9
Perceived of Barrier
I find it difficult to practice the following behaviours
Keep distance at least 1-2 metre when at public area	23.3	24.4	16.4	23.6	12.4
Avoid using public transportation	34.8	25.2	15.8	14.1	10
Work/study from home	35.2	25.1	15.9	12.4	11.4
Avoid handshaking	38.8	28.9	14.2	10.2	7.9
Avoid travel to red zone (are with high positive rate)	37.0	25.6	15.9	12.2	9.3
Self-Efficacy
It is easy for me to practice following behaviour
Keep distance at least 1-2 metre when at public area	4.5	11.8	17.0	30	36.7
Avoid using public transportation	4.3	9.8	15.5	23.8	46.5
Work/study from home	6.3	8.0	13.0	24.6	48
Avoid handshaking	3.1	5.0	11.2	28.4	52.4
Avoid travel to red zone (are with high positive rate)	5.0	7.2	13.0	23.9	50.9

Notes: 1: strongly disagree; 2: disagree; 3: neutral, 4: agree, 5: strongly agree.

For each type of recommended behavior as mentioned in Table 1 to prevent the transmission of COVID-19, more than 50% of respondents never and rarely hosted, visited others, worked from the office/school, shook hands, travelled to a red zone, and left the house when not feeling well ( Figure 2).

Figure 2. The description of behavior.

Notes: 1: Never; 2: rarely; 3: occasionally, 4: frequently, 5: always.

Based on the composite of each perceived item, it was found that only 16% of the respondents had a high percentage of good perceived susceptibility; perceived severity was 43%, perceived benefit was 54%, perceived barrier was 3%, self-efficacy was 19%, and only 60% of respondents practiced good behaviours.

SEM results

Identify degree of freedom of structural model

The second step in SEM analysis is to run the identification of observed variables. A general requirement for identifying any type of model in SEM are the model’s degrees of freedom which must be a least zero (df _M ≥ 0). Hence, the solution to meet the requirement is to identify whether the model is under identified, just identified, or overidentified. Overidentified is mandatory in order to meet the requirement. An overidentified structural equation model is identified and has more observations than free parameters (df _M>0)(2). In this study, we found the degree of freedom value to be 821, hence it was concluded that the model was over-identified. Thus, the next step of the analysis can be completed.

Construct validity

Construct validity was performed to test whether the instrument or measurement variable could describe the latent variable correctly and precisely. For this, two tests were conducted: validity, which consisted of convergent and discriminant validity, and reliability. ¹⁵ A convergent validity test examined the loading factor value of the measurement variable in each latent variable construct. If the loading factor value was greater than 0.50, the latent variable construct had good convergent validity. ¹⁵ The results of the convergent validity test showed that almost all items in this study had a loading factor value of more than 0.5, except for items X11 and BHV3. These two items have a loading factor of ≤ 0.5, which indicates that they do not meet the criteria for convergent validity. Hair et al. (2019) stated that items with low loading factors that do not meet the limits of convergent validity should be excluded from the measurement of latent variables. Therefore, items X11 and BHV3 in this study were not included in the measurement of the latent variable. The convergent validity test was then carried out for a second time. All of the items had good convergent validity, which was shown by loading factor values of more than 0.5. The discriminant validity test was carried out by comparing the root value of each latent variable’s average variance extracted (AVE) with the correlation of these latent variables with other latent variables. If the root value of the variable AVE was greater than the correlation of the variable with other variables in the model, the indicator/question item had good discriminant validity.

Table 4 shows the AVE root value for each latent variable and the correlation coefficient between the latent variables. The value of the AVE root is shown as the value on the diagonal of the matrix, while the values beside and below the AVE root are the correlation coefficients between two pairs of variables. The results of the evaluation of discriminant validity show that the root value of AVE in each latent variable is greater than the correlation coefficient of the latent variable with other latent variables in the structural model. Thus, it can be stated that the items/instruments in this study have good discriminant validity.

Table 4. Results of the discriminant validity test.

	Suscept	Severity	Benefit	Barrier	Efficacy	Behaviour
Suscept	0.822	-	-	-	-	-
Severity	0.730	0.911	-	-	-	-
Benefit	0.590	0.810	0.936	-	-	-
Barrier	-0.160	-0.210	-0.260	0.737
Efficacy	0.080	0.100	0.120	-0.480	0.725
Behaviour	-0.120	-0.150	-0.160	0.460	-0.320	0.630

Construct reliability

The construct reliability test was done by examining the composite reliability value. If the combined reliability value is greater than 0.7, it can be said that the variables in the study already have reliable indicators/question items. ¹⁹ All latent variables were found to have a composite reliability value of more than 0.7, which means that each variable has a consistent measurement indicator and good internal consistency.

Structural model validity

Two analyses were conducted to evaluate the structural model validity:1) dependency test and 2) assessing the goodness-of-fit of the model.

Dependency test

The dependence relationship test was employed by looking at the path coefficient and its p-value in the structural model. The path coefficient shows the magnitude and direction of the relationship between the two variables. According to Table 5, sex had a statistically significant relationship with perceive susceptibility ( p-value = 0.000); perceived susceptibility had a statistically significant relationship with perceived severity ( p-value = 0.000); age ( p-value = 0.004), sex ( p-value = 0.030), and education (( p-value = 0.050) had significant relationships with perceived benefit; perceived benefit ( p-value = 0.000), resident ( p-value = 0.009), and sex ( p-value = 0.000) had a significant relationship with perceive barrier; and perceive barrier ( p-value = 0.000) and occupancy ( p-value = 0.040) had a significant relationship with perceive self-efficacy. Regarding behavior, only perceived barriers ( p-value = 0.000) and self-efficacy ( p-value = 0.000) had statistically significant relationships with COVID-19 prevention behavior. The final structural model is as follows: behaviour = - 0.018*susceptibility - 0.041*severity - 0.00045*benefit + 0.39*barrier - 0.13*self-efficacy.

Table 5. Results of the SEM statistical test.

Latent variable	Independent	Path coefficient	Std. error	t-Statistics	p-value
Perceived Susceptibility	Resident	0.009	-0.010	0.870	0.273
	Age	0.005	-0.005	1.020	0.237
	Sex	0.180	-0.024	7.540	0.000
	Education	0.005	-0.006	0.860	0.276
	Occupancy	0.045	-0.029	1.520	0.126
Perceived Severity	Perceived Susceptibility	0.720	-0.020	35.810	0.000
	Regional	0.012	-0.007	1.690	0.096
	Age	0.001	-0.003	0.280	0.384
	Sex	0.014	-0.017	0.810	0.287
	Education	-0.008	-0.005	-1.680	0.097
	Occupancy	-0.020	-0.021	-0.960	0.252
Perceived Benefit	Perceived Severity	0.810	-0.018	44.660	0.000
	Resident	-0.010	-0.006	-1.550	0.120
	Age	-0.009	-0.003	-3.030	0.004
	Sex	0.034	-0.015	2.270	0.030
	Education	0.008	-0.004	2.040	0.050
	Occupancy	0.032	-0.018	1.770	0.083
Perceived Barrier	Perceived Benefit	-0.240	-0.026	-9.260	0.000
	Resident	-0.028	-0.010	-2.770	0.009
	Age	-0.009	-0.005	-1.940	0.061
	Sex	-0.130	-0.025	-5.050	0.000
	Education	0.009	-0.007	1.340	0.163
	Occupancy	0.051	-0.031	1.670	0.099
Perceived Self-Efficacy	Perceived Barrier	-0.480	-0.031	-15.860	0.000
	Resident	0.008	-0.010	0.840	0.280
	Age	-0.003	-0.005	-0.550	0.343
	Sex	-0.022	-0.024	-0.940	0.256
	Education	0.001	-0.006	0.081	0.398
	Occupancy	0.063	-0.029	2.150	0.040
Individual Behaviour	Perceived Susceptibility	-0.018	-0.037	-0.500	0.352
	Perceived Severity	-0.041	-0.053	-0.780	0.294
	Perceived Benefit	0.000	-0.045	-0.010	0.399
	Perceived Barrier	0.390	-0.036	10.930	0.000
	Perceived Self-Efficacy	-0.130	-0.032	-4.000	0.000

The goodness-of-fit model assessment results in Table 6. showed that none met the goodness-of-fit criteria. Therefore, the structural model in this study was not fit for the data and was not fit to describe the empirical phenomenon under study.

Table 6. Goodness-of-fit index results.

Goodness-of-fit index	Cut of value	Value	Conclusion
Probability of χ 2	>0.050	0.000	No fit
RMSEA	<0.080	0.140	No fit
GFI	>0.950	0.560	No fit
AGFI	>0.950	0.490	No fit
SRMR	<0.100	0.140	No fit
CFI	>0.950	0.850	No fit
NFI	>0.950	0.850	No fit
PNFI	>0.950	0.770	No fit
PGFI	>0.950	0.480	No fit
IFI	>0.950	0.850	No fit
RFI	>0.950	0.830	No fit

Notes: RMSEA=Root Mean Square Error of Approximation; GFI=Goods of Fit Index; AGFI=Adjusted Goodness of Fit Index; SRMR=Standardized Root Mean Square; CFI=Comparative Fit Index; NFI=Normed Fit Index; PNFI=Parsimonious Normed Fit Index; PGFI=Parsimonious Goods of Fit Index; IFI=Incremental Fit Index; RFI=Relative Fir Index.

Discussion

The results of categorized all perceived items revealed that only a small proportion of respondents (16%) held beliefs about their chances of experiencing COVID-19 (perceived susceptibility), while 43% of respondents believed in the severity of the COVID-19 effects were on their health. More than half of the respondents believed that the recommended behaviors to prevent COVID-19 infection could protect them from getting an infection (perceived benefit). Only 3% of respondents believed some things hindered the practice of recommended behaviors (perceive barrier), and only 19% believed that they could practice the recommended behaviors (perceive self-efficacy). This study’s results are similar to those conducted in India, Sri Lanka, Iran, and Ethiopia. ^{8 , 10 , 17 , 20} A study in Italiy conducted to nine hundred and seventy-eight Italian adolescents found a similar pattern where they had a low perception of COVID-19 risk, as well as perceived comparative susceptibility and perceived seriousness. They think that COVID-19 is not a potentially severe disease for them as many news stated that young people are less vulnerable to the COVID-19 effect. ²¹ The HBM theory holds that people are likely to practice preventive behaviors or actions if: 1) They believe that it will reduce their risks, ¹³ 2) They perceive themselves as susceptible to COVID-19 infection, 3) They believe that this condition would have a potentially serious impact, 4) They acknowledge the benefit of recommended actions/behaviours in reducing the susceptibility or severity of the virus, and 5) They believe that the anticipated benefits of doing preventive behaviours/actions outweigh the barriers.

Perceived susceptibility was not a significant predictor of behavior in this study. This finding is consistent with studies that measured adherence to COVID-19 precautionary measures in China ²² and Korea. ^{23 , 24} A study that used a HBM framework to look at how Saudi Arabian students at Jazan University felt about the COVID-19 vaccine found that perceived susceptibility was not a good predictor of how they felt about getting the vaccine. ²⁵ However, a similar study conducted in Malaysia found that high perceived susceptibility to COVID-19 infection was also associated with the behavior of vaccination intention. ²⁶ A study that measured student behavior in the US related to non-pharmaceutical interventions (hand washing with soap and water, use of hand sanitizer, wearing a face mask in public, and practicing social distancing) found that perceived susceptibility was associated with multiple interventions more frequently. ²⁷

A previous study conducted by Du Min e t al. found that low perceived risk was associated with vaccine hesitancy. ²⁸ As with perceived susceptibility, perceived severity was not a significant predictor of preventive behavior in this study. Several studies regarding behavior change using BHM had similar results. ^{24 , 26 , 27} Perceived severity, on the other hand, was a significant predictor of preventive behaviors. ^{8 , 22 – 24 , 29} Perceived benefit is one predictor that is not a significant predictor of preventive behavior. This finding is inconsistent with several studies that used BHM to predict behavior change, particularly in relation to COVID-19 prevention ^{8 , 22 , 23 , 25 – 29} where it was found that there was no significant association between perceived benefit and preventive behaviour of.

This study demonstrated that the perceived barrier significantly predicted COVID-19 preventive behaviour. This result is contrasts with behaviour study conducted in Sri Lanka and Iran, which established a significant positive relationship between perceived benefit and self-efficacy in COVID-19 prevention behaviour. ^{10 , 30} Nevertheless, these findings were congruent with a study conducted in Ethiopia that employed the HBM theory to assess student eating behaviour the US ³¹ and other behavioural studies in Iran, India, and Hong Kong. ^{8 , 17 , 32} This study also found that perceived self-efficacy was a significant predictor of COVID-19 preventive behaviour. The results show that, with lower self-efficacy, people were likely to practice COVID-19 prevention behaviour. This result is similar to several studies that examined behaviours using the BHM theory. Those studies found that perceived self-efficacy has a significant relationship to behaviour. ^{10 , 17 , 20 , 23 , 30 , 33} In contrast, a study that assessed the student’s behaviour on the non-pharmaceutical intervention of COVID-19 found that perceived self-efficacy was not a significant predictor of behaviour change. In theory, an individual with good self-efficacy tends to practice recommended action/behaviour, ¹³ which is the preventive behaviour of COVID-19. However, this study was unable to confirm these findings.

The findings in this study illustrated that most respondents (97%) had no barriers to practising the recommended behaviour. Still most respondents (81%) were not confident that they could fully implement the recommended prevention behaviours. As many as 60% of respondents practised COVID-19 prevention behaviour well. In knowledge attitude practice (KAP) studies conducted in Indonesia, this finding (percentage of good behaviour) tended to be lower than the other two findings in Indonesia, where the rate of those who performed the correct behaviour for the prevention of COVID-19 was more than 90%. ^{34 , 35} Studies conducted in other countries also found that respondents who practised COVID-19 prevention behaviours were relatively high (>70%), such as in China, Nepal, Malaysia, Vietnam, and India. ^{36 – 40}

Fundamentally, perceived susceptibility and severity affect how a person decides to act. ¹³ However, most respondents (84%) had low perceived susceptibility in this study. This means that most respondents did not believe that they were also at risk of being affected by COVID-19. This perception represented an obstacle for someone to implement recommended behaviour. It was also known that only 43% believed that if they were infected with COVID-19, they would experience harmful consequences for their bodies. Only half of the respondents believed that the recommended behaviour was able to protect them from COVID-19. This might relate to the information they obtained day-to-day. It was possible that most of them did not have a clear idea about the pathophysiology and epidemiology of COVID-19, made worse by unreliable news or hoaxes circulated on social media, which may have increased negative perceptions about COVID-19. ^{8 , 41}

Regarding obstacles to implementing the recommended behaviour, a few respondents said it was extremely difficult to implement the behaviours (3%). This overall perception then leads to a low belief that respondents are able to implement the recommended behaviour. As a consequence, it will affect COVID-19 prevention behaviour practice. Perception is theoretically influenced by many factors, including demographics and level of knowledge. ^{8 , 10 , 13 , 20 , 34 , 35} A study conducted in China found that knowledge was influenced by educational level and domicile. ³⁶ Good knowledge can form a good attitude, which then creates a good perceived. ^{42 , 43} COVID-19 was a pandemic that touched almost every facet of human existence. People had to adjust their daily routines to accommodate local government rules in order to reduce the virus transmission, and these behavioural shifts may last long after the disease has passed. ⁴⁴ Increasing the respondents’ knowledge is essential to narrowing the gap between knowledge and practice, including myths, hoaxes circulating about COVID-19, and facts. Several important factors that may affect perceived self-efficacy are related to social norms and trust ⁴⁵ in the community. Unfortunately, neither of these factors (including knowledge) was investigated in this study.

The structural model of COVID-19 prevention behaviour in this study was not intended to describe the empirical phenomenon of preventive behaviour. Byrne stated that, if possible, researchers are advised to modify the model by using modification indices (MI) in SEM testing. He said that models with an MI score of more than 10 deserve attention for modification. ⁴⁶ Already it has proved that to confront the increase in demand for care, the need for long-term care workforce, and the costs associated with care. ⁴⁷ However, Hair et al. suggested that modifying the model should not change the model’s structure significantly. ¹⁹

These several studies found a gap in knowledge to inform changes in policies on infection-prevention measures in the community, community infection procedures, the frequency of testing, etc. ^{8 , 41 , 48} The unreliable information may increase the negative perception of COVID-19, leading to the community’s obedience to suggested preventive behaviors. ⁴¹ Moreover, people are more worried about their families and economic conditions due to the spreading pandemic than about complying with lockdown or restriction policies. Several factors that might be related were not examined in this study, such as level of knowledge, social norms, or trust.

Study limitations

Due to the social restrictions (the outdoor community activities are limited by the government based on the law announced every two weeks except for emergency purposes) in Indonesia, the data collection was conducted using a Google form without using a selected sampling method. Thus, the total number of respondents in this study was not representative of the total population in Indonesia and the entire territory of Indonesia. In addition, the entire population of Indonesia did not have the same opportunity to be selected as respondents due to limitations related to internet coverage and utilities. Other than that, since this study used google form to collect the data, researcher really relied on the respondent’s honesty and integrity. In order to limit the study respondents, the researchers already put inclusion criteria in the initial explanation section of the questionnaire that only those who are living in Indonesia and over 18 years old are eligible for the study. Therefore, we did not do any special treatment to control double entry and the location of respondents whether they live in Indonesia or outside the country. However, when we found it located outside of the country, we dropped the data from the data set. Because of the conditions described above, this study was vulnerable to information bias and selection bias. Regarding to the data collection process, we did not restrict respondent to just entry the form one time (in the system) so that there was possibility that one person will entry multiple times using different identity. Similar treatment conducted to clean the data, when researcher found double mobile number, we will randomly select one data and deleted the rest. We recommend to limit participants with “one email for one entry” to minimize the bias. As addition, this study can only describe the COVID-19 in just one time shot while the rate and impact of COVID-19 were changing rapidly over time. We recommend to employ longitudinal study for further since it can observe individual behavior change pattern/trend.

Conclusions and recommendation

This study found that more than half of the respondents still had low perceived susceptibility (84%) and severity (67%), but more than half had high perceived benefits (54%). Only a few respondents had significant barriers to implementing COVID-19 transmission prevention behaviours (3%). Still, most respondents had low perceived self-efficacy (81%), and only 60% had good behaviors related to COVID-19 prevention.

It was found that sex had a statistically significant relationship with perceive susceptibility ( p-value = 0.000); perceived susceptibility had a statistically significant relationship with perceived severity ( p-value = 0.000); age ( p-value = 0.004), sex ( p-value = 0.030), and education (( p-value = 0.050) had significant relationships with perceived benefit; perceived benefit ( p-value = 0.000), resident ( p-value = 0.009), and sex ( p-value = 0.000) had a significant relationship with perceive barrier; and perceive barrier ( p-value = 0.000) and occupancy ( p-value = 0.040) had a significant relationship with perceive self-efficacy. Regarding behaviour, only perceived barriers ( p-value = 0.000) and self-efficacy ( p-value = 0.000) had statistically significant relationships with COVID-19 prevention behaviour. However, the structural model in this study was not fit to the data and was not fit to describe the empirical phenomenon under study.

This study can be an input for public health policy development especially those related to behaviour change interventions/programs. By using HBM theory, policy makers or other stakeholders can consider which stages of behaviour change still require more intervention in addition to demographic factors which can also influence it. Based on HBM theory, for behaviour change to occur, individuals must perceive a threat in their current behaviour (perceived susceptibility and severity), believe that the change will bring meaningful or useful outcomes (perceived benefit), and possess the self-confidence to enact the change (self-efficacy). In this case, this research can be a reminder in terms of making evidence-based policies.

In the context of COVID-19 preventive behaviour change, we recommend to improve perceived susceptibility and severity (since this study found low of perceived and severity) by providing the correct information about COVID-19 in the local cultural context. It is expected that by improving perceived susceptibility and severity, there would be an increase in respondents’ knowledge, increasing perceived susceptibility and severity. The results and concept of this study can also be used/implemented for developing prevention policies against many types of diseases that require community behaviour changes which consider stages of behaviour change.

For further study, it is highly recommended to make inclusion and exclusion criteria prior to the data collection, to create a more rigorous data collection template to reduce selection bias effect, to collect the data onsite and if it is possible, to conduct the longitudinal study. The concept of HBM is one of the recommended theories to study about the health behaviour, hence it is replicable even in the different context of disease and area.

Ethics approval and consent to participate

The Commission approved this study for Research Ethics and Public Health Service, Faculty of Public Health, University of Indonesia Number: Ket-436/UN2.F10. D11/PPM.00.02/2021.

Consent for publication

Informed consent was obtained from all subjects involved in the study.

Author contributions

All authors made substantial contributions to this research and approved the final manuscript. TE and TS contributed to every step of the study (research concept, design, data interpretation, writing, and review). SP contributed to the research review.

Funding Statement

This research was funded by Universitas Indonesia under contract number NKB-630/UN2.RST/HKP.05.00/2022

The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

Dyrad. Data for: Community perception and COVID-19 prevention, https://doi.org/10.5061/dryad.pnvx0k6rp. ⁴⁹

This project contains the following underlying data:

• • DATA_FINAL_tio-edit2.csv (Data include all variables in the questionnaire. The data contain 1802 respondents and 89 variables. The variable names of questions 1–23 were given according to the keyword in each question, while the variable names for question number 24–35 were specified according to the question’s number and its answer option. It is recommended to read the variable code definition in sheet 2 and the area code in sheet 3.)

Data are available under the terms of the Creative Commons Zero “No rights reserved” data waiver (CC0 1.0 Public domain dedication).

Figshare. IMPLEMENTATION OF SOCIAL POLICY DISTANCING IN EFFORT PREVENTION COVID-19 IN INDONESIA. https://protect-us.mimecast.com/s/qsa7CkRwomfYPpGvjC2JN1U?domain=doi.org ¹⁵

This project contains the following supplementary material:

• •Kuesioner Penelitian.pdf (the google questionnaire used in this study), https://doi.org/10.6084/m9.figshare.23292686.v2. ⁵⁰

Data are available under the terms of the Creative Commons Attribution 4.0 International license (CC-BY 4.0).