Effectiveness of different educational interventions on knowledge, attitude and willingness to undergo cervical cancer screening in rural Indonesia: a quasi-experimental study

Abstract

Objectives

The low uptake of cervical cancer screening (CCS) in rural areas is mainly related to the lack of awareness regarding cervical cancer. Using educational interventions has been shown to substantially improve knowledge and attitude towards cervical cancer, which are important determinants of undergoing CCS. The aim of this study was to analyse the impact of different educational intervention methods on knowledge, attitudes and willingness to undergo CCS among women in rural Indonesia.

Design

A quasi-experimental study with a non-equivalent control group design.

Setting

The study was conducted in North Tobelo, a rural district in North Halmahera Regency.

Participants

Sample size was calculated using Power and Sample Size Calculations software (V.3.1.2). 150 women were recruited via convenience sampling from 12 Integrated Health Service Posts, cluster-randomised into three arms: control (n=50), flipchart (n=50) and video (n=50). A per-protocol approach was applied; all participants completed pre-intervention and post-intervention assessments within a single visit with no loss to follow-up. Continuous outcomes were analysed using the Kruskal-Wallis and Wilcoxon tests; categorical outcomes using χ² or Fisher’s exact tests with Bonferroni correction. Non-parametric analysis of covariance (Quade’s method) controlled for baseline scores, age at first sexual intercourse and household income, with partial eta-squared reported as the effect size.

Interventions

All study arms received identical educational content presented in simple Indonesian language and avoiding the use of medical terminology. The educational content included cervical cancer epidemiology, aetiology, risk factors, clinical symptoms, prevention methods, various CCS methods, visual inspection with acetic acid screening procedures, benefits and side effects of getting human papillomavirus vaccination and CCS, as well as a brief explanation regarding cervical cancer treatment. Knowledge, attitude and willingness were assessed using a structured pre-intervention and post-intervention questionnaire comprising both dichotomous items and Likert-scale questions.

Results

Although knowledge improved significantly in all intervention groups (p<0.001), only the video group showed a significant improvement in attitude after the intervention (p=0.016). The median change scores of knowledge and attitude were only significantly higher in the video group when compared with the control group (p=0.041 and p=0.013, respectively), but not significantly different in other pairwise comparisons. Moreover, the proportion of participants who were willing to undergo CCS after the intervention was also increased significantly in the video group when compared with the control group (52% vs 22%; p=0.006) and flipchart group (52% vs 24%; p=0.012).

Conclusions

In conclusion, the video-based educational intervention demonstrated the greatest improvements in knowledge, attitude, and willngness to undergo CCS compared to other methods.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• Potential confounders and baseline differences were controlled using analysis of covariance, improving internal validity despite the quasi-experimental design.

• The use of a quasi-experimental design and nonrandomised sampling may lead to selection bias, as evidenced by the presence of differing confounding variables between the comparison groups.

• All study arms received identical educational content with predefined topics and time allocation, ensuring that observed differences were attributable to the delivery modality rather than content variation.

• Intervention standardisation was ensured through timed delivery, use of a standardised script, supervision of video playback and delivery by the same medical doctor across intervention arms.

Introduction

Cervical cancer is a preventable disease, yet still a major health problem in Indonesia, with prevalence rates among the highest globally. According to the Global Cancer Observatory in 2022, the prevalence of cervical cancer in Indonesia reached 120 940 in the past 5 years, ranking third worldwide after India and China. It is the second most common cancer in terms of both incidence and mortality in the country.¹ The high mortality rate in Indonesia is attributed to late-stage diagnoses, where curative treatment is no longer effective.^{2 3} Given its long latency period, early detection and intervention are crucial for improving prognosis and reducing mortality of cervical cancer.^{4 5} Despite a national cervical cancer prevention programme since 2007, Indonesia’s cervical cancer screening (CCS) coverage was only 14.6% in 2023.^{6 7} Lack of knowledge regarding cervical cancer and its prevention contributes to low screening coverage in Indonesia.^{6 8} Previous studies in Indonesia showed that only 50%–53% of participants have good knowledge about cervical cancer,^{9 10} which is concerning since knowledge of symptoms, risk factors and preventive measures is a key determinant for people to get screened and vaccinated.⁵

In terms of case distribution, the incidence and prevalence of cervical cancer were found to be higher in rural than urban areas.^10–12 A study in China observed cervical cancer incidence to be three times higher in rural areas than urban areas.¹¹ Moreover, cervical cancer mortality rates are also significantly higher in rural areas and have remained stagnant over the past few decades.¹³ Apart from limited health services and facilities, the lack of awareness, knowledge and positive attitude towards cervical cancer are also described as factors contributing to the high prevalence, incidence and mortality in rural areas.¹¹ Several studies in rural communities showed that only about 28% to 45% of participants had good knowledge about cervical cancer.^{14 15} In fact, a study in rural Ethiopia showed that only 36% of participants had heard of cervical cancer,¹⁶ and nearly 71% of women in rural India did not know that cervical cancer is a female cancer.¹⁷ Consequently, the CCS uptake is significantly lower in rural (26%) than urban areas (60.7%).¹⁸ Disparities in CCS coverage are also found in Indonesia, the coverage rates range from 0.0% to 52.1%, with the majority of the lowest screening rates found in eastern Indonesian provinces which tend to have a large rural population.^{6 7}

Thus, there is an urgent need for intervention programme to increase public awareness and knowledge of cervical cancer, so that screening coverage rate can be increased.^{19 20} Several studies have shown that educational interventions are one of the most effective ways to increase knowledge and develop positive behaviours towards cervical cancer.^20–22 According to a meta-analysis of studies in rural areas, educational interventions can increase CCS uptake by up to 2.77-fold.²¹ The conceptual relationship among knowledge, attitudes and willingness (KAW) is supported by the KAW pathway, which posits that increased knowledge fosters more favourable attitudes, and these attitudes subsequently influence an individual’s willingness to engage in preventive actions such as screening.^20–22 There are various methods of educational intervention, such as lectures with or without visual aids, educational videos, print media and social media. Several studies have shown that video-based or audiovisual educational interventions can improve knowledge, attitudes and screening intentions related to CCS in diverse populations.^{23 24} However, most existing research has examined single intervention arms without directly comparing multiple educational methods. The aim of this study was to analyse the impact of different educational intervention methods on KAW to undergo CCS among women in rural Indonesia.

Materials and methods

Study design and study area

A quasi-experimental study with a non-equivalent control group design was conducted in North Tobelo district, North Moluccas from 4 October to 17 October 2023.

The North Moluccas is a province in Eastern Indonesia, positioned 2483 km from Jakarta, the capital of the country. This province is characterised by a predominantly rural demographic, with approximately 72.9% of its population residing in rural areas (see online supplemental figure 1A).²⁵ In terms of healthcare infrastructure, the region has a physician density of merely 0.373 per 1000 individuals.^{26 27} The study was conducted in North Tobelo, a district in North Halmahera Regency with a population of 11 901, 50.2% of whom are women. There is only a single primary health facility that supervises 12 integrated health service posts (IHSP) in this district.²⁸ IHSP is a national community health programme where health workers from local facilities visit each subdistrict monthly to provide services such as antenatal care, child growth monitoring and vaccination, contraceptive services and elderly healthcare.²⁹

Eligibility criteria, sampling and recruitment

The eligibility criteria for participants were women aged 21–65 years (the recommended age range for CCS^30–32) who resided in North Tobelo district for at least 6 months, had a basic literacy in Indonesian (able to read, speak and understand the Indonesian language), had been sexually active and provided written informed consent to participate in the study. Literacy in Indonesian was required because the educational module, informed consent and outcome questionnaires were all text-based and self-administered. Participants needed to be able to read and understand the materials to ensure accurate comprehension of the intervention and validity of the responses. Participants who had received a hysterectomy, had gynaecologic cancer or were pregnant were excluded.^30–33

The sample size was calculated using the Power and Sample Size Calculations software V.3.1.2 for comparing several groups in experimental studies.^{34 35} We assumed a statistical power of 80% (Zβ=0.84) and 95% CI, a Zα=1.96. According to a previous study in Indonesia, the proportion of patients with good knowledge of cervical cancer was 53% in the pre-intervention group (PA) and 72% in the post-intervention group (PB).³⁶ The τ is the number of comparisons to be made, which is three in this study. The detailed calculation of sample size is described below. Based on calculations using this formula, the required sample size is 131 for three groups. However, a total of 150 participants were recruited in this study to allow for attrition.

$$n=(PA\left(1-PA\right)+PB\left(1-PB\right)){\left(\frac{Z\alpha /\left(2\tau \right)+Z\beta }{PA-PB}\right)}^2$$

The recruitment process was conducted at 12 IHSPs, each of which was then randomly assigned to one of three different interventions, which are flipchart intervention, video intervention and usual care/control (lecture-only intervention) using a computer-generated randomisation process. Each IHSP was given a unique identifier, and allocation was performed using an online random number generator (https://www.randomizer.org) to ensure unbiased assignment. To minimise contamination between groups, each IHSP was assigned to a single intervention group. In addition, the participating IHSPs were located in different subdistricts and the intervention sessions were conducted on separate occasions. Study participants were selected from women attending IHSPs and met the eligibility criteria, using the convenience sampling method. A detailed recruitment process of participants is shown in online supplemental figure 1B.

Although randomisation was conducted at the cluster (IHSP) level, individual participants were not randomised. Because convenience sampling was applied within each IHSP, baseline equivalence between intervention groups could not be fully guaranteed. Therefore, despite using cluster randomisation to reduce contamination, avoid geographic spillover and minimise allocation bias at the facility level, the study design aligns with a quasi-experimental non-equivalent control group framework.

Study instrument and data collection

The questionnaire used in this study was a modified version of a validated questionnaire from previous studies in Indonesia (online supplemental table 1).^{9 37 38} After obtaining approval from two experts in gynaecologic oncology, a preliminary reliability assessment study was conducted on 30 participants who did not participate in the intervention to assess the validity and reliability of the questionnaire, item clarity and internal consistency. In the reliability assessment, the Cronbach’s alphas of the questionnaire for knowledge related to human papillomavirus (HPV) infection and cervical cancer (aspect 1) was 0.626, knowledge related to prevention of HPV infection and cervical cancer (aspect 2) was 0.662, and overall knowledge related to HPV infection and cervical cancer was 0.790. Moreover, the Cronbach’s alpha value of the attitude towards HPV infection and cervical cancer questionnaire was 0.683. These findings suggest that the knowledge and attitude questionnaires are reliable. In the validity assessment, each question on the knowledge questionnaire had a calculated r value ranging from 0.387 to 0.814 (p<0.05). Meanwhile, the calculated r values for each question on the attitude questionnaire ranged from 0.487 to 0.828 (p<0.05). These findings indicated that both the knowledge and attitude questionnaires were valid to use. Based on the findings of the preliminary assessment and expert review, the instrument was revised to improve linguistic clarity and cultural suitability, while the original item structure was maintained to preserve content validity and ensure comprehensive assessment of knowledge regarding HPV infection, cervical cancer, and their prevention. No questionnaire items were removed or added during this revision process.

The questionnaire consisted of six segments: (1) sociodemographic, (2) reproductive health history and awareness, (3) knowledge, (4) attitude, (5) willingness to undergo CCS and (6) reasons underlying the post-intervention CCS decision. In addition, we also evaluated awareness of HPV infection and cervical cancer (yes/no), awareness of any CCS methods (yes/no), CCS history (yes/no) and self-perceived risk of cervical cancer (yes/no).^{37 38} The self-perceived risk of cervical cancer was evaluated by asking ‘Do you think you are at risk for cervical cancer?’ and responded with either ‘yes’ or ‘no’.^{37 39} Two scoring systems were used for knowledge and attitude segments. For yes/no questions, correct answers scored 2, and incorrect answers scored 0. For four-answer questions (strongly disagree to strongly agree), negative answers scored 0, confirming answers scored 1 and highly confirming answers scored 2.⁹ Final scores were calculated by summing the scores, dividing by the maximum possible score, and multiplying by 100%. A 70% cut-off determined participants’ knowledge and attitudes towards HPV and cervical cancer.^{40 41}

Study procedure

All study arms received identical educational content presented in simple Indonesian language, while avoiding the use of medical terminology. The educational content includes cervical cancer epidemiology, aetiology, risk factors, clinical symptoms, prevention methods, various CCS methods, visual inspection with acetic acid (VIA) screening procedures, benefits and side effects of getting HPV vaccination and CCS, as well as a brief explanation regarding cervical cancer treatment. All education interventions consisted of two structured sessions conducted by one medical doctor (GG), who had received prior briefing and standardized training regarding the intervention protocol to ensure consistent delivery across clusters.

Each intervention lasted approximately 15 min. This duration was selected to reflect the routine operational conditions of IHSPs in rural Indonesia, where health education is typically delivered within a limited timeframe alongside multiple other maternal and child health services. The intervention was intentionally designed as a brief, pragmatic educational session that could be feasibly implemented by local healthcare workers within existing workflows, thereby enhancing the applicability and sustainability of the intervention in routine community practice. The first section was a 10-minute educational session delivered directly by the medical doctor in the lecture-only and flipchart groups, and indirectly in the video group, where the doctor remained present but silent during video playback. To ensure consistency across all intervention arms, the duration and content for each topic were standardised using timed delivery: epidemiology (30 s), aetiology (1.5 min), risk factors (1.5 min), symptoms (1 min), preventive measures (3 min), benefits and side effects of HPV vaccination and CCS (2 min), and treatment explanation (30 s). This was followed by a 5-minute discussion session in which participants asked questions about material they found unclear, typically ranging from three to five questions per session. All educational content was delivered using a predefined script and time-controlled structure to ensure consistency. Before the interventions, participants were required to complete a pre-intervention questionnaire. The control group received the usual care, using a lecture-only method without any teaching aids. The flipchart group used a lecture method supported by a flipchart with text and images. The video group was shown an animated video with identical content via a laptop. Afterwards, participants were then asked to complete a post-intervention questionnaire and offered a free on-site CCS examination using the VIA method. The results of the examination were communicated directly to the patient after the examination along with further clinical recommendations. A per-protocol analytical approach was applied, including only participants who completed both the pre-intervention and post-intervention assessments. All study activities including the pretest, intervention and post-test were conducted during a single visit, with no additional follow-up assessments scheduled. All participants completed both the intervention and the post-intervention assessment within a single visit, resulting in no loss to follow-up.

Patient and public involvement

Community and local health stakeholders were involved in the design and conduct of this study. The Head of the North Halmahera Health Department, the Head of the North Tobelo Public Health Center and village leaders provided input on the research questions, study methodology and feasibility of the intervention. They also supported the selection of study areas, facilitated participant recruitment, and aided in disseminating study information through the IHSP. Study findings were shared with community representatives through direct communication, community meetings and IHSP activities.

Study analysis

Data cleaning and statistical analysis were conducted over a 2-week period using comparative methods appropriate for nonequivalent group designs. The data were collected using a paper questionnaire, entered into Microsoft Excel, and then exported to SPSS V.25.0 for analysis.

The primary analyses were defined as the evaluation of changes in knowledge and attitude towards HPV infection and cervical cancer before and after the intervention, as well as comparisons of post-intervention and change scores across the three intervention groups. The secondary analyses focused on the willingness to undergo CCS and the identification of associated determinants using multivariable regression models. Model performance was assessed using receiver operating characteristic (ROC) curves and area under the curve (AUC). χ² or Fisher’s exact tests were used for categorical data, with post hoc pairwise comparisons adjusted using the Bonferroni correction. McNemar’s test was applied for paired categorical data. Data distribution was assessed with the Kolmogorov-Smirnov test. The Wilcoxon test was used for paired continuous data and the Kruskal-Wallis test for unpaired data. Furthermore, a one-way analysis of covariance (ANCOVA) test was also conducted to compare the effectiveness of each educational intervention method without bias by including pre-intervention scores, monthly household income and age at first sexual intercourse as covariates in analysing the post-intervention and change scores. As data were not normally distributed, Quade’s method was applied, with partial eta-squared (η2 p) determining effect size. Fisher’s Least Significant Difference was used for post hoc comparisons. Effect sizes were classified as small (0.01–0.06), moderate (0.06–0.14) or large (≥0.14).⁴²

Bivariate analysis identified potential factors with p values ≤0.25 for inclusion in the regression model.^{43 44} Variables with zero-value cells were excluded. Stepwise regression assessed determinants of knowledge and attitude, while hierarchical regression evaluated factors influencing willingness to undergo CCS.⁴⁵ Model quality was evaluated using the ROC curve and AUC. The AUC values were classified as excellent (0.9–1.0), very good (0.8–0.9), good (0.7–0.8), satisfactory (0.6–0.7) or unsatisfactory (0.5–0.6).⁴⁶ To assess potential effect modification, interaction terms between intervention group and selected baseline sociodemographic variables (age, educational level and monthly household income) were incorporated into the regression models. Interaction effects were considered statistically significant at p<0.05. Sensitivity analyses were performed through covariate-adjusted models to assess the robustness of the intervention effects and to minimise potential confounding.

Results

In total, we recruited 150 participants, who were equally assigned into three study arms. Sociodemographic characteristics between groups were not significantly different, except for monthly household income (p=0.010) and age at first sexual intercourse (p=0.029). The proportion of participants who had a monthly income >US$100 was higher in the video group than in the other two groups (p=0.010). Moreover, most participants in the control group had first sexual intercourse at the age of more than 22 years, which was significantly different from the other two groups where the majority had sexual intercourse at the age of less than 22 years (p=0.029). Most participants were Muslim (73.3%), belonged to the age group of 21–30 years (66%) and had a low level of education (80%). In terms of knowledge, only less than half of the participants (48.7%) were aware of HPV infection and cervical cancer, and only around 3.3% of them had good knowledge related to HPV infection and cervical cancer. Detailed sociodemographic characteristics of participants are shown in table 1.

Table 1. Sociodemographic characteristics of study participants.

Variable	All participant n=150, n (%)	Control n=50, n (%)	Flipchart n=50, n (%)	Video n=50, n (%)	P value
Age
21–30 years	99 (66%)	36 (72%)	30 (60%)	33 (66%)	0.486*
31–40 years	32 (21.3%)	11 (22%)	10 (20%)	11 (22%)
41–50 years	13 (8.7%)	3 (6%)	6 (12%)	4 (8%)
51–60 years	6 (4%)	0 (0%)	4 (8%)	2 (4%)
Median (IQR)	28 (23–34)	27 (24–32)	28 (23–37)	29 (23–33)	0.716†
Age classification
Young (≤40 years)	131 (87.3%)	47 (94%)	40 (80%)	44 (88%)	0.108*
Old (>40 years)	19 (12.7%)	3 (6%)	10 (20%)	6 (12%)
Religion
Muslim	110 (73.3%)	33 (66%)	36 (72%)	41 (82%)	0.188*
Christian	40 (26.7%)	17 (34%)	14 (28%)	9 (18%)
Education
Unschooled	1 (0.7%)	0 (0%)	1 (2%)	0 (0%)	0.062*
Elementary school	11 (7.3%)	6 (12%)	4 (8%)	1 (2%)
Junior high school	22 (14.7%)	8 (16%)	11 (22%)	3 (6%)
Senior high school	86 (57.3%)	24 (48%)	26 (52%)	36 (72%)
College degree (bachelor, master or doctoral)	30 (20%)	12 (24%)	8 (16%)	10 (20%)
Education level
Low (≤SHS)	120 (80%)	38 (76%)	42 (84%)	40 (80%)	0.607*
High (≥college)	30 (20%)	12 (24%)	8 (16%)	10 (20%)
Occupation
Farmer	17 (11.3%)	6 (12%)	10 (20%)	1 (2%)	0.163*
Entrepreneur	5 (3.3%)	1 (2%)	2 (4%)	2 (4%)
Housewife	113 (75.3%)	40 (80%)	33 (66%)	40 (80%)
Non-governmental employee	5 (3.3%)	1 (2%)	2 (4%)	2 (4%)
Governmental employee	2 (1.3%)	0 (0%)	1 (2%)	1 (2%)
Health workers	3 (2%)	0 (0%)	1 (2%)	2 (4%)
Student	3 (2%)	1 (2%)	0 (0%)	2 (4%)
Teacher	2 (1.3%)	1 (2%)	1 (2%)	0 (0%)
Employment status
Not working	116 (75.3%)	37 (74%)	39 (78%)	40 (80%)	0.766*
Employed	34 (24.7%)	13 (26%)	11 (22%)	10 (20%)
Monthly household income
<US$50	86 (57.3%)	31 (62%)	32 (64%)	23 (46%)	0.040*
US$51–US$99	26 (17.3%)	9 (18%)	10 (20%)	7 (14%)
US$100–US$149	18 (12%)	6 (12%)	1 (2%)	11 (22%)
US$150–US$199	2 (1.3%)	0 (0%)	0 (0%)	2 (4%)
≥US$200	18 (12%)	4 (8%)	7 (14%)	7 (14%)
Median (IQR)	33 (27–100)	33 (27–67)	33 (17–60)	67 (27–67)	0.012 †
Monthly household income classification
Low (≤US$100)	120 (80%)	44 (88%)	43 (86%)	33 (66%)	0.010*
High (>US$100)	30 (20%)	6 (12%)	7 (14%)	17 (34%)
Age at first sexual intercourse
<22 years	94 (62.7%)	24 (48%)	34 (68%)	36 (72%)	0.029 *
≥22 years	56 (37.3%)	26 (52%)	16 (32%)	14 (28%)
Median (IQR)	20 (18–23)	22 (19–24)	20 (18–23)	20 (18–23)	0.110†
Marital status
Single	7 (4.7%)	4 (8%)	2 (4%)	1 (2%)	0.458*
Married	140 (93.3%)	46 (92%)	46 (92%)	48 (96%)
Widowed	3 (2%)	0 (0%)	2 (4%)	1 (2%)
Number of lifetime sexual partners
≤1 person	134 (89.3%)	43 (86%)	46 (92%)	45 (90%)	0.613*
>1 person	16 (10.7%)	7 (14%)	4 (8%)	5 (10%)
Median (IQR)	1 (1–1)	1 (1–1)	1 (1–1)	1 (1–1)	0.615†
Parity status
Nulliparous	8 (5.3%)	1 (2%)	5 (10%)	2 (4%)	0.277*
Parous	142 (94.7%)	49 (98%)	45 (90%)	48 (96%)
Frequency of visits to integrated health service post
<6 per year	32 (21.3%)	8 (16%)	14 (28%)	10 (20%)	0.329*
≥6 per year	118 (78.7%)	42 (84%)	36 (72%)	40 (80%)
Family history of gynaecologic cancer
No	147 (98%)	49 (98%)	49 (98%)	49 (98%)	1.00*
Yes	3 (2%)	1 (2%)	1 (2%)	1 (2%)
Awareness of HPV infection and cervical cancer
No	77 (51.3%)	27 (54%)	30 (60%)	20 (40%)	0.121*
Yes	73 (48.7%)	23 (46%)	20 (40%)	30 (60%)
Awareness of cervical cancer screening methods
No	143 (95.3%)	50 (100%)	48 (96%)	45 (90%)	0.070*
Yes	7 (4.7%)	0 (0%)	2 (4%)	5 (10%)
Previous history of cervical cancer screening
No	148 (98.7%)	50 (100%)	50 (100%)	48 (96%)	0.329*
Yes	2 (1.3%)	0 (0%)	0 (0%)	2 (4%)
Self-perceived risk of cervical cancer
No	135 (90%)	46 (92%)	46 (92%)	43 (86%)	0.513*
Yes	15 (10%)	4 (8%)	4 (8%)	7 (14%)
Knowledge on HPV infection and cervical cancer in general (aspect 1)
Poor	142 (94.7%)	49 (98%)	48 (96%)	45 (90%)	0.277*
Good	8 (5.3%)	1 (2%)	2 (4%)	5 (10%)
Knowledge on the prevention of HPV infection and cervical cancer (aspect 2)
Poor	146 (97.3%)	50 (100%)	49 (98%)	47 (94%)	0.324*
Good	4 (2.7%)	0 (0%)	1 (2%)	3 (6%)
Overall knowledge on HPV infection, cervical cancer and its prevention
Poor	145 (96.7%)	49 (98%)	49 (98%)	47 (94%)	0.620*
Good	5 (3.3%)	1 (2%)	1 (2%)	3 (6%)
Attitude on HPV infection, cervical cancer and its prevention
Negative	148 (98.7%)	49 (98%)	50 (100%)	49 (98%)	1.000*
Positive	2 (1.3%)	1 (2%)	0 (0%)	1 (2%)

The percentage value ‘%’ represents the percentage of the column.

Values in bold indicate statistically significant relationships between variables.

^*χ2 test or Fisher’s exact test.

^†Kruskal-Wallis test.

HPV, human papillomavirus.

Effect of educational interventions on knowledge

At baseline, there were no significant differences in median overall knowledge score (p=0.548) and the proportion of participants with good overall knowledge level (p=0.440) between study arms. After intervention, there was a significant difference in the median overall knowledge score (p=0.001) and overall knowledge level (p=0.006) between groups (see table 2). Moreover, there were significant differences in the median change scores between groups for overall knowledge (p=0.037). The video group had a significantly higher median change score compared with the control group (p=0.041), with no differences found in other pairwise comparisons (see online supplemental table 2). As for the post-intervention knowledge score, we further conducted ANCOVA with the pre-intervention score, monthly household income and age at first sexual intercourse as covariates. After controlling for pre-intervention scores, the median overall knowledge score after the intervention between groups remained significantly different (p<0.001; η2 p=0.098) with significantly higher median scores in the flipchart (p=0.021) and video (p<0.001) groups compared with the control group (see online supplemental table 3).

Table 2. Comparison of knowledge, attitude and willingness to undergo cervical screening between three study groups before and after the educational intervention.

Variable		Pre-intervention	Post-intervention	P value (pre-intervention vs post-intervention)
Knowledge aspect 1: HPV infection and cervical cancer
Knowledge score
Control	Median (IQR)	0 (0–0)	80 (60–80)	Z=−5.984; <0.001*
Flipchart	Median (IQR)	0 (0–20)	80 (80–80)	Z=−5.997; <0.001*
Video	Median (IQR)	0 (0–20)	80 (80–80)	Z=−5.845; <0.001*
P value (between group)		0.768†	0.075†
Knowledge level
Control	Poor	49 (98%)	18 (36%)	<0.001‡
	Good	1 (2%)	32 (64%)
Flipchart	Poor	48 (96%)	12 (24%)	<0.001‡
	Good	2 (4%)	38 (76%)
Video	Poor	45 (90%)	11 (22%)	<0.001‡
	Good	5 (10%)	39 (78%)
P value (between group)		0.182§	0.238§
Knowledge aspect 2: The prevention of HPV infection and cervical cancer
Knowledge score
Control	Median (IQR)	0 (0–20)	60 (40–80)	Z=−6.105; <0.001*
Flipchart	Median (IQR)	0 (0–20)	70 (60–80)	Z=−6.078; <0.001*
Video	Median (IQR)	0 (0–20)	80 (60–100)	Z=−6.134; <0.001*
P value (between group)		0.718†	<0.001†
Knowledge level
Control	Poor	50 (100%)	37 (74%)	<0.001‡
	Good	0 (0%)	13 (26%)
Flipchart	Poor	49 (98%)	25 (50%)	<0.001‡
	Good	1 (2%)	25 (50%)
Video	Poor	47 (94%)	20 (40%)	<0.001‡
	Good	3 (6%)	30 (60%)
P value (between group)		0.168§	0.002§
Overall knowledge
Knowledge score
Control	Median (IQR)	10 (0–10)	65 (50–80)	Z=−6.168; <0.001*
Flipchart	Median (IQR)	0 (0–20)	70 (60–80)	Z=−6.173; <0.001*
Video	Median (IQR)	10 (0–20)	80 (70–90)	Z=−6.117; <0.001*
P value (between group)		0.548†	0.001†
Overall knowledge level
Control	Poor	49 (98%)	25 (50%)	<0.001‡
	Good	1 (2%)	25 (50%)
Flipchart	Poor	49 (98%)	16 (32%)	<0.001‡
	Good	1 (2%)	34 (68%)
Video	Poor	49 (98%)	10 (20%)	<0.001‡
	Good	1 (2%)	40 (80%)
P value (between group)		0.440§	0.006§
Attitude
Attitude score
Control	Median (IQR)	20 (0–30)	30 (20–40)	Z=−5.319; <0.001*
Flipchart	Median (IQR)	10 (0–20)	30 (20–50)	Z=−5.773; <0.001*
Video	Median (IQR)	20 (10–30)	40 (30–60)	Z=−5.555; <0.001*
P value (between group)		0.092†	0.014†
Attitude classification
Control	Negative	49 (98%)	47 (94%)	0.500‡
	Positive	1 (2%)	3 (6%)
Flipchart	Negative	50 (100%)	46 (92%)	0.125‡
	Positive	0 (0%)	4 (8%)
Video	Negative	49 (98%)	42 (84%)	0.016‡
	Positive	1 (2%)	8 (16%)
P value (between group)		0.604§	0.213§
Self-perceived risk of cervical cancer
Control	No. n (%)	46 (92%)	36 (72%)	0.002‡
	Yes. n (%)	4 (8%)	14 (28%)
Flipchart	No. n (%)	46 (92%)	33 (66%)	<0.001‡
	Yes. n (%)	4 (8%)	17 (34%)
Video	No. n (%)	43 (86%)	25 (50%)	<0.001‡
	Yes. n (%)	7 (14%)	25 (50%)
P value (between group)		0.064§	0.195§
Willingness to undergo cervical cancer screening
Control	No. n (%)	48 (96%)	39 (78%)	0.004‡
	Yes. n (%)	2 (4%)	11 (22%)
Flipchart	No. n (%)	47 (94%)	38 (76%)	0.004‡
	Yes. n (%)	3 (6%)	12 (24%)
Video	No. n (%)	46 (92%)	24 (48%)	<0.001‡
	Yes. n (%)	4 (8%)	26 (52%)
P value (between group)		0.703§	0.002§

The percentage value ‘%’ represents the percentage of the column.

Values in bold indicate statistically significant relationships between variables.

^*Wilcoxon test.

^†Kruskal-Wallis test.

^‡McNemar test.

^§χ2 test or Fisher’s exact test.

HPV, human papillomavirus.

Multivariate logistic regression analysis demonstrated two statistically significant factors contributing to good overall knowledge, such as the video-assisted education intervention (adjusted OR (AOR) 2.38; 95% CI 1.04 to 5.43; p=0.039) and having a self-perceived risk of cervical cancer after the intervention (AOR 3.18; 95% CI 1.42 to 7.15; p=0.005) (see table 3). In addition, the ROC analysis of this predictive model was also conducted and an AUC of 0.679 (95% CI 0.591 to 0.767; p<0.001) was obtained, indicating that there is a 67.9% chance that this model can predict the level of knowledge correctly (see online supplemental figure 2A).

Table 3. Multivariate logistic regression analysis on factors associated with level of knowledge regarding HPV infection and cervical cancer.

Variables	Knowledge level			Bivariate analysis*		Multivariate analysis
	Good	Poor	Total	P value	Crude OR†	P value	Adjusted OR†
Educational intervention classification
Education without video	59 (59.6%)	41 (80.4%)	100 (66.7%)	Ref.		Ref.
Education with video	40 (40.4%)	10 (19.6%)	50 (33.3%)	0.010	2.78 (1.25 to 6.18)	0.039	2.38 (1.04 to 5.43)
Education level
Low (≤SHS)	76 (76.8%)	44 (86.3%)	120 (80%)	Ref.		Ref.
High (≥College)	23 (23.2%)	7 (13.7%)	30 (20%)	0.168	1.90 (0.76 to 4.79)	0.214	1.84 (0.70 to 4.83)
Frequency of visits to Integrated Health Service Post
<6 per year	18 (18.2%)	14 (27.5%)	32 (21.3%)	Ref.		Ref.
≥6 per year	81 (81.8%)	37 (72.5%)	118 (78.7%)	0.189	1.70 (0.77 to 3.79)	0.236	1.67 (0.72 to 3.88)
Awareness of cervical cancer screening methods
No	92 (92.9%)	51 (100%)	143 (95.3%)	Ref.		Ref.
Yes	7 (7.1%)	0 (0%)	7 (4.7%)	0.096	–	NA‡	–
Self-perceived risk of cervical cancer (pre-intervention)
No	86 (86.9%)	49 (96.1%)	135 (90%)	Ref.		Ref.
Yes	13 (13.1%)	2 (3.9%)	15 (10%)	0.075	3.70 (0.80 to 17.09)	0.628	1.53 (0.27 to 8.52)
Self-perceived risk of cervical cancer (post-intervention)
No	53 (53.5%)	41 (80.4%)	94 (62.7%)	Ref.		Ref.
Yes	46 (46.5%)	10 (19.6%)	56 (37.3%)	0.001	3.56 (1.61 to 7.893)	0.005	3.18 (1.42 to 7.15)
Level of overall knowledge (pre-intervention)
Poor	94 (94.9%)	51 (100%)	145 (96.7%)	Ref.
Good	5 (5.1%)	0 (0%)	5 (3.3%)	0.167†	–	NA‡	–

The percentage value ‘%’ represents the percentage of the column.

Values in bold indicate statistically significant relationships between variables.

^*Bivariate analysis using χ2 or Fisher exact; any related variables with p<0.25 were considered suitable for inclusion into the multivariate analysis model.

^†OR with 95% CI.

^‡Cells have a value of zero and were therefore excluded from multivariate analysis.

HPV, human papillomavirus; NA, not available; SHS, senior high school.

Effect of educational interventions on attitude

At baseline, there were no significant differences in median attitude score (p=0.092) and the proportion of participants with positive attitude (p=0.604) between study arms. However, a significant improvement in attitude after intervention was only observed in the video group (p=0.016) (see table 2). Similarly, the median change score was also only significant between the video group and the control group (20 (IQR: 10–30) vs 10 (IQR: 0–20), respectively; p=0.013), but not in other pairwise comparisons as shown in online supplemental table 2. Furthermore, we conducted ANCOVA to verify this finding. After adjusting for monthly household income, ANCOVA showed no significant differences in median attitude score after intervention between groups (p=0.066; η2 p=0.036) as shown in online supplemental table 3.

Among the 16 variables found significant in the univariate analysis, only employment status (AOR 4.04; 95% CI 1.05 to 15.48; p=0.042) and self-perceived risk of cervical cancer after intervention (AOR 6.99; 95% CI 1.67 to 29.29; p=0.008) were strongly associated with attitude in the multivariate analysis (see table 4). ROC analysis was also performed to assess the ability of these two variables to predict a positive attitude. The result showed that participants who were employed and had a self-perceived risk of cervical cancer after the intervention were positively associated with positive attitudes towards cervical cancer (AUC: 0.756; 95% CI 0.626 to 0.885; p<0.001) (see online supplemental figure 2B).

Table 4. Multivariate logistic regression analysis on factors associated with attitude towards cervical cancer.

Variables	Attitude			Bivariate analysis*		Multivariate analysis
	Positive	Negative	Total	P value	Crude OR†	P value	Adjusted OR†
Educational intervention classification
Education without video	7 (46.7%)	93 (68.9%)	100 (66.7%)	Ref.		Ref.
Education with video	8 (53.3%)	42 (31.1%)	50 (33.3%)	0.083	2.53 (0.86 to 7.44)	0.927	0.94 (0.23 to 3.87)
Religion
Islam	14 (93.3%)	96 (71.1%)	110 (73.3%)	Ref.		Ref.
Christian	1 (6.7%)	39 (28.9%)	40 (26.7%)	0.072	0.17 (0.02 to 1.38)	0.083	0.13 (0.01 to 1.30)
Education level
Low (≤SHS)	9 (60%)	111 (82.2%)	120 (80%)	Ref.		Ref.
High (≥college)	6 (40%)	24 (17.8%)	30 (20%)	0.080	3.08 (1.00 to 9.48)	0.316	0.38 (0.06 to 2.51)
Employment status
Not working	7 (46.7%)	109 (80.7%)	116 (77.3%)	Ref.		Ref.
Employed	8 (53.3%)	26 (19.3%)	34 (22.7%)	0.006	4.79 (1.59 to 14.41)	0.042	4.04 (1.05 to 15.48)
Monthly household income
Low (≤US$100)	7 (46.7%)	113 (83.7%)	120 (80%)	Ref.		Ref.
High (>US$100)	8 (53.3%)	22 (16.3%)	30 (20%)	0.003	5.87 (1.93 to 17.86)	0.081	3.26 (0.87 to 12.32)
Age at first sexual intercourse
< 22 years	7 (46.7%)	87 (64.4%)	94 (62.7%)	Ref.		Ref.
≥22 years	8 (53.3%)	48 (35.6%)	56 (37.3%)	0.177	2.07 (0.71 to 6.06)	0.540	1.60 (0.36 to 7.16)
Awareness of cervical cancer
No	4 (26.7%)	73 (54.1%)	77 (51.3%)	Ref.		Ref.
Yes	11 (73.3%)	62 (45.9%)	73 (48.7%)	0.044	3.24 (0.98 to 10.68)	0.400	2.02 (0.39 to 10.33)
Awareness of cervical cancer screening methods
No	12 (80%)	131 (97%)	143 (95.4%)	Ref.		Ref.
Yes	3 (20%)	4 (3%)	7 (4.6%)	0.022	8.19 (1.64 to 40.94)	0.509	2.60 (0.15 to 44.36)
A history of cervical cancer screening
No	14 (93.3%)	134 (99.3%)	148 (98.6%)	Ref.		Ref.
Yes	1 (6.7%)	1 (0.7%)	2 (1.4%)	0.191	9.57 (0.57 to 161.54)	0.705	0.32 (0.01 to 112.89)
Self-perceived risk of cervical cancer (pre-intervention)
No	10 (66.7%)	125 (92.6%)	135 (90%)	Ref.		Ref.
Yes	5 (33.3%)	10 (7.4%)	15 (10%)	0.008	6.25 (1.79 to 21.86)	0.163	3.38 (0.61 to 18.64)
Self-perceived risk of cervical cancer (post-intervention)
No	5 (33.3%)	89 (65.9%)	94 (62.7%)	Ref.		Ref.
Yes	10 (66.7%)	46 (34.1%)	56 (37.3%)	0.013	3.87 (1.25 to 11.99)	0.008	6.99 (1.67 to 29.29)
Knowledge level for aspect 1 (pre-intervention)
Poor	11 (73.3%)	131 (97%)	142 (94.6%)	Ref.		Ref.
Good	4 (26.7%)	4 (3%)	8 (5.4%)	0.004	11.91 (2.61 to 54.25)	0.072	6.89 (0.84 to 56.53)
Knowledge level for aspect 2 (pre-intervention)
Poor	13 (86.7%)	133 (98.5%)	146 (97.3%)	Ref.		Ref.
Good	2 (13.3%)	2 (1.5%)	4 (2.7%)	0.050	10.23 (1.33 to 78.76)	1.000	NA‡
Overall knowledge level (pre-intervention)
Poor	13 (86.7%)	132 (97.8%)	145 (96.7%)	Ref.		Ref.
Good	2 (13.3%)	3 (2.2%)	5 (3.3%)	0.078	6.77 (1.04 to 44.26)	0.297	0.06 (0.00 to 11.77)
Overall knowledge level (post-intervention)
Poor	3 (20%)	48 (35.6%)	51 (34%)	Ref.		Ref.
Good	12 (80%)	87 (64.4%)	99 (66%)	0.228	2.21 (0.59 to 8.21)	0.845	1.18 (0.23 to 6.20)
Overall attitude (pre-intervention)
Negative	13 (86.7%)	135 (100%)	148 (98.7%)	Ref.
Positive	2 (13.3%)	0 (0%)	2 (1.3%)	0.009	–	NA§	–

The percentage value ‘%’ represents the percentage of the column.

Values in bold indicate statistically significant relationships between variables.

^*Bivariate analysis using χ2 or Fisher exact; any related variables with p<0.25 were considered suitable for inclusion into the multivariate analysis model.

^†OR with 95% CI.

^‡The 95% CI cannot be calculated because one of the cells has a very low value.

^§Cells have a value of zero and were therefore excluded from multivariate analysis.

NA, not available; SHS, senior high school.

Effect of educational interventions on willingness to undergo CCS

Before the educational intervention, only 9 (6%) participants were willing to undergo CCS and there was no significant difference between groups (control: 4%; flipchart: 6%; video: 8%; p=0.703). However, the proportion of participants who were willing to undergo CCS improved significantly after intervention in control (4% to 22%; df: 18%; p=0.004), flipchart (6% to 24%; df: 18%; p=0.004), as well as video group (8% to 52%; df: 44%; p<0.001). After the intervention, there was a significant difference in willingness to undergo CCS between groups (p=0.002) (table 2). The video group showed a higher proportion of participants willing to undergo CCS compared with both the control (OR: 3.84; p=0.006) and flipchart groups (OR: 3.43; p=0.012). No significant difference was found between the flipchart and control groups (p=1.000) (see online supplemental table 4).

Hierarchical logistic regression analysis was performed to identify determinants predicting willingness to undergo CCS after the intervention. Sociodemographic variables were included in model 1. Only monthly household income significantly influenced willingness to undergo CCS after the intervention (OR: 2.47; 95% CI 1.05 to 5.82; p=0.038). When the variables of reproductive health history and awareness were added into model 2, there was a significant improvement in the model (p<0.001). However, only the self-perceived risk of cervical cancer after the intervention was statistically significant (OR: 8.01; 95% CI 3.32 to 19.62; p<0.001). The educational intervention method was added to model 3 and showed a significant improvement in the model (p=0.006). In addition to the self-perceived risk of cervical cancer, the video-based educational intervention proved to be statistically significant in predicting willingness to undergo CCS (OR: 3.49; 95% CI 1.42 to 8.62; p<0.001). However, when knowledge variables were added into model 4, there was no statistically significant improvement in the model (p=0.717). Finally, attitude towards cervical cancer was added into model 5 and showed significant improvement in the model (p=0.001). In this final model, having a perceived risk of cervical cancer (OR: 7.18; 95% CI 2.69 to 19.17; p<0.001), receiving a video-based educational intervention (OR: 3.41; 95% CI 1.29 to 9.03; p=0.013) and having a positive attitude (OR: 14.03; 95% CI 2.40 to 82.11; p=0.003) were predictors of willingness to undergo CCS after the intervention (see table 5). In addition, ROC analysis was also performed to assess the ability of the three variables to predict participants’ willingness to undergo CCS and an AUC of 0.827 (95% CI 0.755 to 0.899; p<0.001) was obtained as shown in online supplemental figure 2C.

Table 5. Hierarchical logistic regression analysis on factors associated with willingness to undergo CCS after the intervention.

Variable included	Model 1	Model 2	Model 3	Model 4	Model 5
	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)	OR (95% CI)
Model 1: sociodemography
Age classification  (Ref: ≤40 years)	2.34 (0.78 to 6.97)	3.26 (1.00 to 10.68)	3.20 (0.99 to 10.39)	2.99 (0.90 to 9.88)	3.24 (0.89 to 11.78)
Employment  (Ref: not working)	1.28 (0.52 to 3.12)	1.15 (0.41 to 3.20)	1.48 (0.52 to 4.17)	1.53 (0.54 to 4.37)	1.17 (0.37 to 3.74)
Monthly household  (Ref: ≤US$100)	2.47 (1.05 to 5.82)*	2.59 (0.96 to 6.98)	1.83 (0.65 to 5.12)	1.93 (0.68 to 5.52)	1.36 (0.44 to 4.20)
Frequency of visits to IHSP  (Ref: ≤6 per year)	2.23 (0.85 to 5.85)	1.95 (0.68 to 5.57)	1.96 (0.68 to 5.68)	1.92 (0.65 to 5.64)	2.36 (0.70 to 7.98)
Model 2: reproductive health history and awareness
Gynaecologic cancer history in family  (Ref: no)		1.04 (0.07 to 15.98)	1.26 (0.08 to 20.02)	1.34 (0.09 to 20.86)	1.69 (0.09 to 31.31)
Age at first sexual intercourse  (Ref: ≤22 years)		1.48 (0.63 to 3.46)	1.89 (0.77 to 4.66)	1.96 (0.79 to 4.85)	1.88 (0.73 to 4.85)
Awareness of CCS methods  (Ref: no)		2.10 (0.29 to 15.24)	1.25 (0.20 to 7.91)	2.78 (0.18 to 44.05)	2.31 (0.11 to 47.85)
Self-perceived risk of cervical cancer pre-intervention  (Ref: no)		0.61 (0.16 to 2.25)	0.54 (0.14 to 2.16)	0.55 (0.14 to 2.18)	0.36 (0.08 to 1.66)
Self-perceived risk of cervical cancer post-intervention  (Ref: no)		8.01 (3.32 to 19.62)**	7.28 (2.92 to 18.12)**	7.55 (2.93 to 19.44)**	7.18 (2.69 to 19.17)**
Model 3: educational intervention methods
Intervention method  (Ref: without video)			3.49 (1.42 to 8.62)**	3.53 (1.41 to 8.80)**	3.41 (1.29 to 9.03)*
Model 4: knowledge regarding HPV infection and cervical cancer
Knowledge aspect pre-intervention  (Ref: poor)				0.23 (0.01 to 9.39)	0.13 (0.00 to 13.80)
Overall knowledge post-intervention  (Ref: poor)				0.92 (0.35 to 2.40)	0.89 (0.33 to 2.41)
Model 5: attitude towards HPV infection and cervical cancer
Attitude post-intervention  (Ref: negative)					14.03 (2.40 to 82.11)**
Change in model (p value)		<0.001**	0.006**	0.717	0.001**
Receiver operating characteristic analysis
AUC (95% CI)	0.579 (0.478 to 0.679)	0.708 (0.616 to 0.799)	0.766 (0.684 to 0.848)	0.766 (0.684 to 0.848)	0.827 (0.755 to 0.899)
P value	0.118	<0.001***	<0.001***	<0.001***	<0.001***
Sensitivity	0.306	0.653	0.837	0.837	0.755
Specificity	0.851	0.762	0.604	0.604	0.762
PPV	0.500	0.571	0.506	0.506	0.607
NPV	0.716	0.819	0.884	0.884	0.865
Accuracy	0.673	0.726	0.680	0.680	0.760

Values in bold indicate statistically significant relationships between variables.

^*p<0.05, **p<0.01, ***p<0.001.

AUC, area under the curve; CCS, cervical cancer screening; HPV, human papilloma virus; IHSP, integrated health service post; NPV, negative predictive value; PPV, positive predictive value.

Of the 49 participants initially willing to undergo screening, 13 (26.5%) did not participate in VIA screening. Reasons included child uncooperativeness (69.2%), partner objection (15.4%) and changing their mind (15.4%). Additionally, 4 (3.9%) participants who were initially unwilling to undergo CCS participated due to waived fees and reassurance from health workers or family. A detailed summary of the reasons underlying participants’ decisions is shown in online supplemental table 5. Interaction analyses were performed to determine if the educational intervention’s effectiveness varied across different participant strata. Subgroup analyses revealed no statistically significant interaction effects between the intervention group and age, educational level or monthly household income (p>0.05), suggesting a consistent effect across these variables.

Discussion

According to health behavioural change theories, providing adequate information enables people to have a better understanding, which will form a positive attitude and intention to perform positive behaviour.^{47 48} In line with these theories and several other studies,^{21 49 50} our study also showed a substantial improvement in knowledge and attitudes towards cervical cancer as well as participants’ willingness to undergo CCS after the educational intervention (online supplemental figures 3-5). However, the magnitude of impact produced by each educational intervention method varies across variables. This might be explained by the fact that different methods of education can have varying effectiveness depending on the specific situation or group.⁵¹

Comparative effectiveness of different educational intervention methods

Most participants in this study had low education levels (≤senior high school) and limited knowledge of cervical cancer and its prevention, which is concerning given that most of them were aged 21–30 years, the recommended age for starting CCS.^{30–32 52} While controversial, starting screening at less than 30 years of age can reduce cervical cancer risk by up to 48%⁵³ and is better at preventing cervical intraepithelial neoplasia (CIN) progression, which is more common in younger women, particularly those who have not received HPV vaccination like the participants in this study.^{54 55}

All educational interventions significantly improved knowledge and attitude scores, but there were notable differences in the magnitude of improvement between groups. Groups using teaching aids such as flipcharts and videos showed greater increases in scores compared with the lecture-only control group. This aligned with previous studies highlighting the benefits of visual aids, particularly for individuals with low health literacy.^{56 57} Our study also found that both audiovisual methods (flipchart and video) produced greater improvements in knowledge and attitude scores compared with the lecture-only group. Moreover, the video group had higher proportions of participants with good knowledge, positive attitudes and willingness to undergo CCS. However, although the video-based intervention showed the largest mean improvements in knowledge and attitude, these gains were not statistically superior to those of the flipchart intervention. Similar studies have shown greater improvements in knowledge, willingness and CCS participation with video-based interventions compared with other methods.^{58 59} Video-based education has proven effective in various settings, particularly in rural areas with low health literacy.^60–62 In this study, the video-based educational intervention proved highly effective, likely due to the shifting trend in media consumption, with Indonesians increasingly engaging in watching behaviours through television, social media and live performances. Approximately 83.48% of rural communities in Indonesia have the habit of watching television regularly with most people (78.84%) having a frequency of watching up to 6 to 7 days per week.⁶³ Additionally, according to the cognitive theory of multimedia learning, videos enhance information processing and memory recall more effectively by activating both visual and auditory channels.⁶⁴

In this study, we also compared the effect of each educational intervention method without bias by controlling for variables considered as covariates, such as pre-intervention scores, monthly household income and age at first sexual intercourse. After adjusting for pre-intervention scores, the effect on overall knowledge remained unchanged, with greater effects observed in the flipchart and video groups compared with the control group. However, after adjusting for monthly household income and age at first sexual intercourse, the effect on attitude and knowledge aspect 2 changed, with no significant differences between interventions. These findings align with previous studies showing correlations between income and attitude and age at first sexual intercourse and cervical cancer knowledge.^{65 66}

Factors associated with KAW to undergo CCS

Stepwise regression analysis revealed that the method of educational intervention was a predictor of good knowledge and willingness to undergo CCS, with video-based interventions being significantly more effective in achieving these outcomes compared with other methods, consistent with prior studies.^{23 58 67} However, the intervention was not a predictor of positive attitude, aligning with previous research.^{68 69} According to the stage of change theory, attitude modification requires repeated and sustained interventions,^{20 70} suggesting that a single educational session may be insufficient.^{67 71} Although educational interventions did not directly influence attitudes, they were associated with improved knowledge,^{9 72} which in turn has been shown to positively influence attitudes. Thus, it is plausible that participants may require more time to assimilate the knowledge acquired and translate it into a positive attitude.⁶⁷

Moreover, our study also identified employment status as a predictor of a positive attitude towards cervical cancer. In contrast, other studies found that employment status was not significantly associated with attitude but was linked to cervical cancer knowledge.^{73 74} These conflicting results may stem from the fact that most employed women in our study had higher monthly incomes, allowing them better access to information about cervical cancer and enabling them to prioritise health-related activities.^{75 76} In addition, having a self-perceived risk of cervical cancer after intervention was found to be one of the predictors of good knowledge, positive attitude and willingness to undergo CCS, similar to previous studies.^{37 77} Self-perceived risk or perceived susceptibility is a person’s belief of the likelihood of acquiring a disease, which can drive them to seek information regarding the preventive measures.⁷⁸ A study in Ethiopia found that those who perceived themselves at risk for cervical cancer were 3.63 times more likely to engage in health-seeking behaviour.⁷⁹ This behaviour has been shown to improve knowledge and attitudes towards cervical cancer^{80 81} and encourage preventive measures, such as undergoing CCS or receiving HPV vaccination earlier.¹⁸ Furthermore, our study also revealed a positive attitude as a determinant for the willingness to undergo CCS. Similarly, Getahun et al also found that participants who had a positive attitude towards cervical cancer were 6.16 times more likely to be willing to undergo CCS compared with participants who had a negative attitude (AOR=6.164; p<0.05).⁸² These findings are in line with the theory of planned behaviour which states that behavioural intentions are determined by attitudes towards the behaviour, perceived behavioural control and subjective norms.⁸³

Strengths and limitations of the study

Despite taking pragmatic measures to address limitations, we acknowledge several constraints in this study. First, the quasi-experimental design and nonrandomised sampling may introduce selection bias, as observed in the differing confounding variables between groups (table 1).^{84 85} However, we controlled for potential confounders in ANCOVA to mitigate bias and better assess the true effects of each educational intervention on knowledge and attitude. Second, the study was conducted in only one rural province in eastern Indonesia, limiting its generalisability to all rural communities. However, North Moluccas provides valuable insights as it has the highest proportion of rural inhabitants in Indonesia and shares cultural similarities with other eastern provinces, which are predominantly where rural communities are located in Indonesia.²⁵ Third, the intervention was delivered only once for approximately 15 min, the timing of the post-intervention and follow-up assessment may have been too soon; therefore these design features limit our ability to evaluate the sustained or cumulative effects of repeated educational sessions. Fourth, the study was conducted at IHSP, where most women came for child growth monitoring and vaccination. This posed a challenge, as the presence of children may have distracted participants. Although we addressed this by asking mothers to leave their children with health cadres before the intervention, we still found approximately 69.2% of those who cancelled CCS participation did so due to their uncooperative children. Last, although the instruments demonstrated acceptable internal consistency in this study, the KAW to undergo CCS and reasons underlying postintervention CCS decision questionnaires included Likert-scale and dichotomous (yes/no) items, which produce ordinal data. The use of summed questionnaire scores for statistical analysis, without transformation to interval-level measures such as logit, may introduce some measurement limitations and potential bias. Nevertheless, this scoring approach is widely used in Knowledge–Attitude–Practice research and was applied consistently across all study groups and both preintervention and postintervention assessments, allowing valid comparisons of relative changes following the educational interventions. Future studies should consider formal psychometric validation using larger and more diverse samples to further strengthen the measurement properties of the instrument.

A major strength of this study is that it is the first to evaluate the effects of different educational intervention methods on improving KAW to undergo CCS among rural communities in Eastern Indonesia. Given the limited data on this topic, our study provides valuable insights into the role of educational interventions in cervical cancer prevention programmes in rural areas. Consistent with other studies,^{58 59 86} our findings indicate that video-based education demonstrated the greatest improvements in knowledge and attitude outcomes. However, its effectiveness was not statistically superior to the flipchart-based intervention. Our findings highlight the potential importance of repeated educational interventions in shaping attitudes toward cervical cancer; however, this recommendation is based on existing theoretical frameworks and prior literature rather than the empirical data from the present study.

Importantly, all study arms received identical educational content with predefined topics and equal time allocation, ensuring that any observed differences were attributable to the mode of delivery rather than variation in content. Intervention standardisation was further strengthened through timed delivery, use of a standardised script, supervision of video playback and delivery by the same medical doctor across all intervention arms, thereby minimising performance bias and enhancing internal validity.

Conclusions

Considering the long latency of cervical cancer, having a good knowledge of preventive measures is crucial for early detection, reducing mortality and improving prognosis. Rural communities, with lower awareness of cervical cancer, exhibit lower CCS participation and higher mortality rates compared with urban areas. Educational interventions are therefore crucial to improve knowledge and attitudes, which in turn influence CCS uptake. While a one-time intervention improves knowledge, repeated interventions are necessary to foster positive attitudes. In this study, all educational interventions improved knowledge, attitudes and willingness to undergo CCS. The video-based intervention produced the largest mean improvements, and was significantly more effective than the lecture-only intervention across all outcomes. Compared with the flipchart intervention, the video-based intervention produced significantly greater willingness to undergo CCS, whereas improvements in knowledge and attitudes were comparable between the two interventions. Based on these findings, video-based educational interventions should be recommended within cervical cancer prevention programmes, especially in rural and resource-limited settings. In the context of limited medical personnel in rural areas, these results also support further exploration of remotely delivered video-based education, such as via digital video discs, downloadable materials or online platforms, to improve awareness and uptake of screening. Overall, this study underscores the critical role of structured educational interventions in strengthening cervical cancer prevention efforts.

Data availability statement

Data are available on reasonable request.