Impact of Oral Health Chatbot With and Without Toothbrushing Training on Childhood Caries

Abstract

Objectives

Standalone oral health chatbots targeting young children's oral health are rare. The aim of this research was to compare the effectiveness of a standalone chatbot and a combination chatbot with in-person toothbrushing training for caregivers in improving young children's oral health.

Methods

A randomised, parallel, 2-group pretest–posttest design was employed with 320 caregiver–child pairs (aged 6–42 months). Group I (160 pairs) used the 21-Day FunDee (modified) chatbot along with in-person toothbrushing training, whilst Group II (160 pairs) used only the 21-Day FunDee Plus chatbot. Oral examination assessed plaque levels and caries, whilst a self-administered questionnaire evaluated oral hygiene care, dietary practices, and oral health perceptions based on the protection motivation theory (PMT). Data were analysed using 2-way repeated-measures analysis of variance, a t test, and chi-square measures for group comparisons.

Results

The majority of caregivers were Muslim mothers. No significant differences were observed between groups at the baseline, 3-month, and 6-month follow-ups in mean dmft (Group I: 4.16, 4.64, and 5.30 vs Group II: 4.30, 5.54, and 5.82), mean plaque scores (Group I: 0.72, 0.53, and 0.55 vs Group II (0.84, 0.52, and 0.59), and most dietary habits. However, significant improvements were found within groups from baseline to follow-ups in plaque reduction, toothbrushing practices, overall knowledge score, PMT perceptions, proper tooth brushing, fluoride toothpaste usage, and dietary behaviours (frequency of bottle feeding, frequency of nocturnal bottle feeding, proportion of children who went to bed without consuming anything after cleaning their teeth before bedtime). The significant differences between groups were found in self-efficacy at all time points, but only at the 6-month evaluation for percentage of fluoride toothpaste and overall PMT perceptions.

Conclusions

Both interventions were comparable in preventing caries, reducing plaque, improving feeding practices, increasing parental involvement in tooth brushing, and enhancing knowledge. The standalone chatbot 21-Day FunDee Plus presents a viable alternative for promoting oral health in young children.

Introduction

Early childhood caries (ECC) is one or more decaying, missing, or filled tooth in the primary dentition of children younger than 6.5 years¹ that negatively impacts quality of life for the affected children and their families.^2,3 In Thailand, dental caries was prevalent amongst 3-year-olds at a rate of 52.9% in 2017, with Southern Thailand having a high rate of 57%.⁴ This figure surpassed the Southeast Asia regional average of 43.7% and the global average of 42.7% for 2019.⁵ ECC prevention strategies include improving oral hygiene, restricting dietary sugars, fluoride application, and early dental examination.⁶ Reviews recommend promoting oral hygiene through both in-person toothbrushing training via parental supervised brushing as well as theory-based behavioural modification and barrier-focussed oral health education with additional support.^7,8 Additionally, improving parental knowledge, establishing early dental care, and individualised caries management plans are essential.⁹ However, implementing customised interventions across populations requires substantial resources.¹⁰ Furthermore, barriers to accessing oral health care must also be addressed.^11,12

Recently, the 2020–2025 World Health Organization Global Strategy for Digital Health encourages individuals in the adoption of digital technologies to enhance health service delivery through trust-based, patient-centred care.¹³ Chatbots are computer programmes enabling remote human communication on devices via text, voice, or images. In health care, they are emerging as scalable, personalised tools to deliver on-demand health education.^14,15 Compared to traditional methods, chatbots enable 24/7 access, individualisation, interactive dialogue for improved guidance, and inclusive reach across populations.16, 17, 18 Systematic reviews have found promise in the use of chatbots in public health to manage chronic diseases and promote health behaviours.¹⁹ Consistent with a comprehensive review and meta-analysis, chatbot interventions have demonstrated efficacy for improving sleep, physical activity, and diet across age groups. Benefits were seen for both short- and long-term implementations, either as standalone tools or part of a multicomponent approach.²⁰

Chatbot applications in oral health education remain scarce.²¹ Our prior comparative study²² explored an oral health chatbot's potential to serve as an alternative to in-person caregiver toothbrushing instruction for improving the oral hygiene of young children. Whilst plaque reduction results were promising, longer evaluation periods, robust study designs, and incorporation of dietary counselling are required to establish efficacy for caries prevention. It remains unclear whether chatbot oral health education alone can improve long-term caries outcomes. This study compared chatbot interventions with and without in-person toothbrushing training in reducing dental caries rates throughout a 6-month follow-up amongst children aged 6 to 42 months. These chatbots were also designed using the protection motivation theory (PMT)^23,24 and Zhang's artificial intelligence chatbot behaviour change model²⁵ for chatbot design, namely 21-Day FunDee, to promote behaviour change, as shown elsewhere.²² The chatbots were adjusted from a previous version by employing new Thai platforms, Botnoi and Wowbot. The enhanced chatbots, 21-Day FunDee (modified) and 21-Day FunDee Plus, include additional content emphasising the reduction of sugary foods and drinks and the improvement of feeding habits. The modified version is utilised with in-person toothbrushing training. Meanwhile, the standalone chatbot, 21-Day FunDee Plus, builds upon the modified one by providing additional guidance on toothbrushing techniques for young children and integrating motivational application programming interfaces (APIs) to replace in-person toothbrushing training.

Methods

Study setting and participants

This study was conducted as a part of the Wowbot project, which aimed to assess the efficacy of oral health care chatbots developed using the Wowbot platform, and approved by the Faculty of Dentistry, Prince of Songkla University Institutional Review Board (EC6408-055), and registered with the Clinical Trials Registry (TCTR20210927010).

Two oral health care chatbots—the 21-Day FunDee (modified) and 21-Day FunDee Plus—were developed and pilot-tested in 30 participants. A single-blinded, parallel-arm, randomised clinical control trial was conducted in 320 pairs of caregivers and children between January and September 2023 at 2 study sites: Maikan district in Pattani and Kong-Ra district in Phatthalung province, Thailand. Potential child participants were identified using each study site's health registry. Subsequently, their caregivers were contacted and invited to participate in this study.

The sample size was determined based on the primary outcome, the dmft caries scores, measured at baseline, 3 months, and 6 months. A repeated-measures analysis of variance (ANOVA) was utilised, assuming a small to moderate effect size of 0.17, corresponding to Cohen's f, which was considered a small to medium effect. To achieve 80% power at a 5% significance level and to detect a clinically meaningful difference between Group I and Group II, the initial sample size required was calculated to be 137 participants per group. Considering a potential dropout rate of 20% over the follow-up period, the final sample size was increased to 160 participants per group. Calculations were performed using G*Power software (version 3.1.9.7), ensuring adequate power to detect significant differences between groups across the time points.

Participants eligible to participate in this study were children between the ages of 6 and 42 months, accompanied by their caregivers who attended a health-promoting hospital for a vaccination or a health checkup. Eligible caregivers must communicate fluently in Thai, be Facebook Messenger users, and have a smartphone with internet access throughout the inclusion period. The child must have at least 1 tooth and not have any serious medical condition or intellectual disability that could impact dental health or cooperation, for example, clefts or Down syndrome. Participants were excluded if they declined to provide consent to the study or if the child was uncooperative during the dental examination. After providing written informed consent, participants were randomly allocated to Group I (n = 160) or Group II (n = 160). Randomisation was achieved using a computer-generated sequence to assign participants to each group. These assignment numbers were then placed in sealed, opaque envelopes to ensure allocation concealment. Dental hygienists selected an envelope, which determined the participant's group assignment, maintaining the integrity of the randomisation process.

Group I participants received in-person toothbrushing training along with a subscription to the 21-Day FunDee (modified) chatbot programme, whilst Group II participants received the 21-Day FunDee Plus chatbot programme. Interviews and dental examinations were conducted at study baseline and at 3- and 6-month follow-up (Figure 1).

Fig. 1.

Chatbot initialisation interface on Facebook Messenger.

Interventions

Comparison of Group I [in-person toothbrushing training and the 21-Day FunDee (modified) chatbot programme] and Group II (the 21-Day FunDee Plus chatbot programme) interventions are presented in the Appendix (Table 6).

Chatbots

Concerning the foundational aspects of both chatbots, 21 daily dialogues encompassed diverse elements such as games, infographics, animations, and animated songs and disseminating general knowledge on oral health care for young children. Additionally, recommendations regarding feeding practices and control of sugar consumption were provided. Notably, greeting cards and moral support infographics at the conclusion of daily dialogues were provided every day, addressing themes related to oral health care and maternal experience.

As an expansion of the modified 21-Day FunDee, the 21-Day FunDee Plus chatbot in Group II concentrated on providing young children with additional instruction regarding toothbrushing techniques. Moreover, it incorporated motivational APIs to facilitate the comparison of plaque levels in 3 periods, offering API frames as motivational rewards as a substitute for in-person toothbrushing training. Group II was presented with the optional 4-day dialogues for the 21-Day FunDee Plus chatbot, which encompassed discussions on the significance of deciduous teeth, methods for preparing children for dental visits, the necessity of decreasing sugar intake, and requesting caregivers' wishes for their children. Notably, all dialogues were constructed according to the PMT^23,24 and Zhang model.²⁵

The chatbot initialisation interface on Facebook Messenger is depicted in Figure 1. Figure 2 presents a 1-day chatbot conversation aimed at promoting oral health care for young children. The dialogue incorporates various elements of PMT including a greeting daily objective, infographics, multimedia content, and interactive games and concludes with an inspirational or humorous child care message.

Fig. 2.

Example of daily dialogues in a 1-day chatbot.

In-person toothbrushing training

Each child in this study was provided with a toothbrush. However, only the caregivers in Group I were given in-person toothbrushing training, which lasted 10 to 15 minutes. Caregivers were given the objective of toothbrushing for children and guidance on the appropriate amount of toothpaste to use and the brushing technique by calibrated dental hygienists. The need to clean the cervical area was highlighted, ensuring that all tooth surfaces were thoroughly cleansed. Gauze or a clean diaper was used for tongue cleaning and removal of residual toothpaste. Caregivers were instructed to brush their children's teeth independently whilst being supervised by dental hygienists. Nevertheless, the dental hygienist may demonstrate on the child if necessary. Last, the dental hygienist provided positive feedback on their practice.

Data collection and outcome assessment

Comprehensive training and calibration protocols were administered to 8 dental hygienists for interview data collection and 5 dentists for clinical data collection. The primary outcomes were caries incidence based on the International Caries Detection and Assessment System (ICDAS) II merged code (0 = sound, A = initial stage decay, B = moderate decay, and C = extensive decay) and the average score of the visible plaque deposits (0 = no visible plaque, 1 = plaque detected by probing, 2 = plaque visible to the naked eye). A dental caries examination was carried out for all surfaces of the erupted teeth. Plaque examination was assessed on the buccal surface using a blunt probe and mouth mirror under natural light.

The standardised examination for interexaminer calibration, conducted with an expert on a sample of 10 children, yielded a kappa value ranging from 0.82 to 0.92 for dental caries assessment. However, the assessment of dental plaque deposits was standardised through a training process rather than direct calibration on children, as that was not feasible. The training involved detailed sessions in which the expert and the dental hygienists worked together to establish a consensus on the assessment criteria.

The children who participated in the pilot session and those involved in the outcome assessment training were not included in the main study. These groups were separate and solely used for preparatory purposes, ensuring that there was no overlap or potential bias affecting the main trial's results.

A self-administered structured questionnaire was designed to collect sociodemographic characteristics at baseline, 3 months, and 6 months for the following information: oral health knowledge (11 items), oral health perceptions based on PMT (40 items), oral health practices, and dietary behaviours (10 items). A chatbot satisfaction survey (0 = very dissatisfied, 5 = very satisfied) was administered at 3-month follow-up. Three experts (a community dentist, a pedodontist, and a public hospital dentist) assessed the content and construct validity of the questionnaires. The questionnaire's content validity was confirmed with an index of item–objective congruence of 0.85, whilst the index of item–objective congruence of the oral health perception based on PMT was 1.0. Cronbach reliability was 0.74. After that, face validity was determined via a questionnaire pilot test with 25 participants.

Data analysis

Descriptive and analytical statistics were used to analyse the data. Missing answers for the main outcomes led to exclusion from the analysis. Using the ICDAS II code system, the number of teeth decayed (d), missing (m), and with fillings (f) were computed for both surface (dmfs) and tooth levels (dmft) on a per-person basis. The prevalence of caries at baseline, 3 months, and 6 months was calculated. The overall scores and scores for each domain were computed for knowledge and oral health care perceptions (a correct or positive answer = 1 point, an incorrect or negative answer = 0 points). The percentage of individuals responsible for administering care determined oral hygiene practices and dietary habits, such as using fluoride toothpaste, using bottle feeding, and sleeping with a milk bottle. The baseline characteristics of the study samples in both groups were summarised and compared using either an independent t test or a chi-square test. Repeated-measures ANOVA was used to examine the intervention's effectiveness across time for the primary outcomes, number of dental caries (dmft and dmfs), and plaque scores as well as knowledge, PMT-based oral health care perception, and oral health practice.

Responses to chatbot satisfaction questions were grouped into 7 domains: usefulness, perception of building relational capacity, health care quality perspective, conversational quality, content features, AI perspective, and user experiences. The average scores for each domain were computed. The overall satisfaction score and scores within each domain between the 2 groups were compared using an independent t test. IBM SPSS version 29.0.0.0(241) was used for all analyses. A statistical significance was defined as a P value less than .05.

Results

The study sample consisted of 303 participants divided into Group I (n = 152) and Group II (n = 151). All participants underwent oral examinations, whilst 152 in Group I and 148 in Group II also completed questionnaires. Participation details are shown in Figure 3.

Fig. 3.

CONSORT flow of the project.

The participants in both groups were mostly mothers of children. The mean age of the studied children was around 24 months. There were no significant differences amongst caregiver's age, education levels, family income status, occupations of the parents, religions, and receipt of oral health care advice (Table 1).

Table 1.

Comparison of participants’ general characteristics.

	Group I (n = 152)		Group II (n = 148)		P value
Caregiver age, mean ± SD, y	32.4 ± 7.4		32.0 ± 8.0		.843
Child age, mean ± SD, mo	23.4 ± 9.9		24.0 ± 10.6		.602
Main caregiver, No. (%)
Mother	128	84.2	125	85.1	.824
Other relatives	24	15.8	23	14.9
Caregiver education level, No. (%)
Elementary school or lower	8	5.3	16	10.9	.083
High school	76	50.7	84	57.1
Diploma/bachelor's degree	63	42.0	46	31.3
Higher than bachelor's degree	3	2.0	1	0.7
Economic status, No. (%)
Not enough	38	25.0	40	27.0	.163
Enough	87	57.2	93	62.8
Enough and savings	27	17.8	15	10.1
Occupation, No. (%)
Stay-at-home parent	57	37.5	79	53.4	.090
Civil servant	33	21.7	19	12.8
Private-sector employee	36	23.7	31	20.9
Agriculturist	11	7.2	9	6.1
Merchant	14	9.2	10	6.8
Others	1	0.7	-	-
Religion, No. (%)
Muslim	136	89.5	126	85.1	.259
Buddhist	16	10.5	22	14.9
Received advice about oral health care for young children, No. (%)
Ever	128	84.2	115	77.7	.151
Never	24	15.8	33	22.3

An independent t test was applied to compare mean values between groups, and a chi-square test was applied to compare proportions between groups.

After the experiment, the samples in both groups were reduced to 152 and 151 people, respectively (the percentages of sample loss was 5.0% and 5.6%, respectively).

Repeated-measures ANOVA revealed no significant between-group differences in mean dmfs, dmft, and plaque scores at 3 and 6 months. However, within-group comparisons showed significantly different mean dmfs, dmft, and plaque scores from baseline to 3 and 6 months in both groups (Table 2). Additionally, dental caries prevalence increased by 6.6% in Group I and 11.9% in Group II.

Table 2.

Comparison of oral health examination results before and after the study.

Status	Group I (n = 152), mean (SD)			Group II (n = 151), mean (SD)			P value time	P value group
	BL	3 mo	6 mo	BL	3 mo	6 mo
dmfs	7.98 (12.66)	8.92 (13.00)	10.74 (14.64)	8.12 (11.49)	9.71 (12.50)	12.22 (16.30)	<.001*†‡	.593
dmft	4.16 (5.22)	4.64 (5.35)	5.30 (5.72)	4.30 (4.97)	5.44 (5.38)	5.82 (5.73)	<.001*†‡	.415
Pl	0.72 (0.65)	0.53 (0.59)	0.55 (0.57)	0.84 (0.63)	0.52 (0.55)	0.59 (0.60)	<.001*†	.381

BL, baseline; Pl, ________.

^⁎Significant difference between baseline and 3 months.

^†Significant difference between baseline and 6 months.

^‡Significant difference between 3 months and 6 months.

The evaluation phase revealed that appropriate behaviours were parental toothbrushing or children brushing followed by parental re-brushing. The proportions of these behaviours increased significantly from the baseline in both groups, with no between-group difference. However, the proportion using fluoride toothpaste increased significantly more in Group II vs Group I at 6 months (Table 3).

Table 3.

Comparison of dietary behaviour and oral hygiene care between groups.

Behaviours	Group I			P value*	Group II			P value*	P value between groups
	BL	3 mo	6 mo		BL	3 mo	6 mo		BL	3 mo	6 mo
Oral hygiene practices
Parental toothbrushing, %	61.4	80.3	78.3	<.001 .001	57.5	79.1	71.6	<.001 .002	.504	.707	.182
Frequency of child's toothbrushing (times per week), mean (SD)	4.4 (2.8)	4.7 (2.7)	4.6 (2.9)	.043 .253	4.0 (2.9)	4.4 (2.9)	4.3 (3.0)	.065 .031	.289	.336	.531
Usage of fluoride toothpaste, %	78.6	92.5	87.4	.021 .096	73.4	96.5	96.1	<.001 <.001	.441	.188	.021
Dietary behaviour
Frequency of bottle feeding (No. of times per day), mean (SD)	3.6 (2.8)	3.4 (3.1)	2.9 (2.9)	.484 .033	3.5 (3.1)	2.9 (2.8)	2.7 (2.7)	.017 .003	.851	.200	.448
Frequency of sleeping with a milk bottle (No. of times per week), mean (SD)	2.1 (2.9)	1.2 (2.0)	0.9 (1.8)	.001 <.001	1.8 (2.8)	1.7 (2.7)	1.5 (2.6)	.610 .165	.449	.080	.030
Frequency of child waking to drink from a bottle or breastfeed at night (No. of times per week), mean (SD)	3.1 (3.2)	2.4 (2.9)	1.9 (2.9)	.016 <.001	3.4 (3.7)	2.5 (3.1)	1.9 (2.8)	.006 <.001	.587	.843	.978
After last time brushing before bed, child did not consume milk, snacks, or any other food, %	35.2	52.8	61.3	.110 <.001	27.4	46.1	54.7	.004 .003	.268	.466	.314
Did not add sugar to food or drinks for the child, %	75.2	74.2	73.0	.627 .522	78.1	75.7	76.4	.871 1.000	.558	.764	.508
Frequency of consuming sweet foods (No. of times per day), mean (SD)	2.6 (1.7)	2.6 (1.8)	2.5 (1.5)	.753 .603	2.3 (1.6)	2.5 (1.7)	2.6 (1.7)	.109 .006	.121	.509	.488

^⁎Initial and second P values represent the comparison between baseline and 3 months and baseline and 6 months, respectively.

The repeated-measures ANOVA revealed significantly higher mean knowledge scores across all aspects at 3 and 6 months compared with baseline, with no between-group differences or group-by-time interaction effects (Table 4).

Table 4.

Comparison of knowledge on child oral health care before and after intervention.

Knowledge (total score)	Group I (n = 143), mean (SD)			Group II (n = 141), mean (SD)			P value time	P value group
	BL	3 mo	6 mo	BL	3 mo	6 mo
The importance of primary teeth and effect of dental caries (4)	2.44 (0.89)	2.76 (0.83)	2.78 (0.81)	2.25 (0.90)	2.74 (0.83)	2.67 (0.72)	<.001*†	.148
Frequency of tooth brushing (1)	0.66 (0.48)	0.80 (0.40)	0.75 (0.44)	0.66 (0.48)	0.77 (0.42)	0.71 (0.46)	<.001*†	.600
Tooth brushing method (4)	1.99 (1.03)	2.31 (1.02)	2.22 (1.00)	1.87 (0.90)	2.24 (1.01)	2.11 (0.97)	<.001*†	.289
Use of fluoride toothpaste (2)	1.09 (0.66)	1.38 (0.67)	1.36 (0.59)	1.03 (0.65)	1.42 (0.66)	1.34 (0.64)	<.001*†	.849
Overall knowledge (11)	6.18 (2.16)	7.24 (2.09)	7.11 (2.11)	5.80 (1.91)	7.17 (2.13)	6.83 (1.95)	<.001*†	.228

BL, baseline.

No statistical significant effect of time and group.

^⁎Significant difference between baseline and 3 months.

^†Significant difference between baseline and 6 months.

The repeated-measures ANOVA revealed significantly higher mean perception scores across all domains and overall at 3 and 6 months compared with baseline. No between-group differences were observed for perceived severity, perceived vulnerability, or response efficacy, whilst self-efficacy perception and overall perception differed significantly between groups. No interaction effect was identified between the study group and duration (Table 5).

Table 5.

Comparison of perceptions in child oral health care before and after intervention.

Perceptions (total score)	Group I (n = 145), mean (SD)			Group II (n = 143), mean (SD)			P value time	P value group
	BL	3 mo	6 mo	BL	3 mo	6 mo
Perceived severity (25)	18.1 (3.8)	20.0 (3.6)	19.9 (4.0)	17.6 (3.6)	19.4 (3.7)	19.0 (3.4)	<.001*†	.060
Perceived vulnerability (10)	6.6 (1.8)	7.5 (1.8)	7.5 (1.9)	6.4 (1.8)	7.2 (1.8)	7.4 (1.9)	<.001*†	.233
Self-efficacy (30)	20.3 (2.8)	21.4 (3.1)	21.9 (3.7)	19.5 (2.6)	21.1 (3.1)	20.9 (3.0)	<.001*†	.013
Response efficacy (5)	3.6 (1.1)	3.8 (1.0)	3.8 (1.1)	3.5 (1.0)	3.8 (1.0)	3.7 (1.1)	.001*†	.177
Overall perception (70)	48.7 (7.5)	52.8 (7.6)	53.1 (8.8)	46.9 (6.2)	51.5 (7.3)	50.9 (8.1)	<.001*†	.023

BL, baseline.

No statistical significant effect of time and group.

^⁎Significant difference between baseline and 3 months.

^†Significant difference between baseline and 6 months.

Participants in Group I had an average engagement with the chatbot of 13.8 ± 8.2 days out of 21 days, and participants in Group II had an average usage rate of the chatbot of 15.4 ± 7.5 days out of 21 days. The overall average satisfaction levels with using the chatbots were good in both groups (Table 6).

Table 6.

Satisfaction with using the chatbots.

	Group I (n = 149), mean (SD)	Group II (n = 143), mean (SD)	P value
The average of satisfaction levels in all aspects	4.0 (0.5)	4.0 (0.6)	.631
Overall satisfaction	4.2 (0.7)	4.2 (0.8)	.496
Usefulness	4.2 (0.7)	4.2 (0.7)	.543
Perception of relational and capacity	4.2 (0.6)	4.2 (0.7)	.446
Health care quality perspective	4.1 (0.6)	4.1 (0.6)	.475
Conversational quality	4.1 (0.6)	4.1 (0.7)	.537
Content features	4.0 (0.8)	4.1 (0.8)	.866
AI perspective	3.9 (0.7)	3.8 (0.7)	.497
User experiences	3.6 (0.7)	3.6 (0.7)	.758

Discussion

This study represents the initial exploration of employing an exclusive chatbot for oral health education to enhance caries control amongst young children. The caries prevalence in children increased by 6.6% in Group I and 11.9% in Group II, highlighting that both interventions were effective in controlling new carious lesions.

Our study aligns with a 3-arm investigation (A: motivation education including discussion with PowerPoint presentation, pamphlets, and 2 reinforced phone calls; B: traditional health education; C: control) designed to prevent ECC in infants aged 6 to 18 months with healthy dentition. At the 6-month follow-up, 9%, 14%, and 23% of children with decayed teeth were detected in groups A, B, and C, respectively.²⁶ Similarly, a 6-month comprehensive education intervention incorporating individual and group discussions, biweekly SMS reminders, and pamphlets showed 13% caries incidence vs 35% with standard well-baby care alone amongst children aged 12 to 36 months.²⁷ Noteworthy and comparable to the context of our research, a 9-month cohort study in Southern Thailand²⁸ found that the prevalence of dental caries in infants aged 9 to 18 months increased by 66.1% (from 2% to 68.1%), suggesting that our interventions delayed caries progression relative to no intervention.

Compared to a 1-year Thai study for children aged 6 to 19 months²⁹ showing an increment of 3.5 ± 3.4 and 3.2 ± 3.5 in tooth caries in an intervention group using a participatory approach vs a control group using national guidelines, respectively, our 6-month study found a lower net increment of 1.2 ± 2.5 and 1.5 ± 2.6) teeth in Group I and Group II, respectively.

The caries increment was greater than in a study by Hasadisevee et al,³⁰ which implemented one doll demonstration with an in-person toothbrushing training session, 2 caregiver group discussions for seeking oral health solutions, and 4 volunteer home visits over 10 to 12 months with 0- to 2-year-old children (mean age, 17.1 months) and dmfs at baseline of 2.74. They reported an 8.8% annual caries increment by surfaces. In comparison, our study observed 6-month caries surface increment rates of 7.9% and 9.2% for Group I and Group II, respectively. The higher incidence here may relate to a lack of ongoing follow-up activities that promote continuity of care. Additionally, the older age (23–24 months) and higher baseline caries experience (dmfs: 7.98–8.12) of our sample could further increase susceptibility.

It is noteworthy that our study sample had much higher caries risk and experience compared to the 8th National Oral Health Survey in Thailand (2017).⁴ Despite the mean age being around 2 years, the sample had 4.2 and 4.3 dmft per child in Group I and Group II, respectively, whereas 3-year-olds nationally showed only 2.8 dmft. A 2019 systematic review by Hummel et al³¹ found children with high caries experience have higher rates of new carious lesions in permanent teeth compared to those with moderate or low baseline caries experience.

Contributing greatly to ECC are feeding practices.^32,33 Both groups improved practices like bottle feeding frequency, sleep bottle feeding, abstinence from food and liquids following toothbrushing before bedtime, and nocturnal bottle feeding at 3 and 6 months. This is consistent with the finding of the systematic review, which stated that mobile health (mHealth) in Africa can enhance child-feeding practices, but only with regard to breastfeeding.³⁴ In contrast, an integrated review revealed that efforts to improve feeding behaviour produced inconsistent outcomes.³⁵ By applying the PMT, our chatbots incorporated dialogues on disease severity with prolonged bottle use alongside weaning options and motivational messages empowering caregivers to cease bottle feeding. However, considering other high-risk dietary behaviours such as sweet foods and drinks, no improvements were found in either group. This is consistent with the findings of Weber-Gasparoni et al³⁶ and Manchanda et al,²⁶ who observed no significant improvement after 6 to 8 months of follow-up. It is possible that chatbots could facilitate a restricted number of dialogues pertaining to these subjects; furthermore, they were introduced in the third week, which may avoid certain users, and the topics may not effectively address their obstacles. The meta-analysis recommended that in order to decrease the consumption of sweets and sugary drink, greater emphasis should be placed on health-promoting activities that engage a broader spectrum of stakeholders and environments.³⁷

Usage of fluoride toothpaste increased in both groups, with significantly greater improvement observed in Group II (22.7%) compared to Group I (8.8%) at the 6-month follow-up. It is noteworthy that during evaluation, 87.4% of Group I and 96.1% of Group II samples contained fluoride toothpaste, exceeding typical rates (70-90%) in comparable studies. These findings align with or exceed results from other oral health interventions in young children, including a two-year multi-component program for 3-year-olds (83.4% usage),³⁸ a Thai study with 6-19 month-olds reported 87% usage,²⁹ while a 2-year education program for 8-23 month-olds achieved 70-74% usage.³⁹ This may correspond to the chatbots' repeated involvement with animations, games, and dialogues that continuously promoted the use of fluoride toothpaste, especially in Group II. Early use of fluoride toothpaste is recommended once the first primary tooth erupts in young children.⁶ Additionally, the caries-preventive effect increases in those with high baseline caries levels,⁴⁰ which was consistent with the characteristics of our study sample.

Reduced plaque accumulation likely results from improved toothbrushing behaviours. Both groups showed similar plaque reduction of around 23.6% to 26.4% in Group I and 29.8% to 38.1% in Group II, consistent with increased caregiver-reported toothbrushing frequency in Group I of 19.7% to 21.7% and in Group II of 14.8% to 22.3% after the interventions. A meta-analysis supports the notion that mHealth can serve as a supplementary element in the management of gingivitis, the acquisition of oral health knowledge, and the enhancement of oral hygiene.⁴¹ This result is in line with the outcomes of a 6-month study in which mothers of infants aged 12 to 49 months participated in a video intervention, revealing proportions of plaque presence in children comparable to those observed from brochure distribution. Nonetheless, discernible data regarding enhancements in plaque reduction were not evident.³⁶ Likewise, knowledge of positioning and toothbrushing technique as well as self-efficacy in toothbrushing improved in both groups. It is noteworthy that a comparative analysis of 2822 children aged 3.5 to 4.5 from an international cross-sectional study revealed a 30% reduction in cavity rates between those with optimal and suboptimal parental supervised brushing.⁴²

It was also noteworthy that the group receiving combined in-person and chatbot training had significantly higher self-efficacy and overall PMT perceptions at all time points compared to the chatbot alone, possibly reflecting the impact of in-person toothbrushing instruction. Integrated education and hands-on training are known to augment skills like self-assurance and behaviour change, leading to better oral hygiene.^30,43,44 However, the chatbot-only group (Group II) in this study still showed promising plaque reduction despite a lack of in-person training. This chatbot may substitute for in-person toothbrushing instruction by illustrating proper brushing techniques via supplementary videos, infographics, animations, and games. Furthermore, immediate rewards were added on days 4, 6, and 20 based on real-time plaque feedback via APIs. This follows reinforcement principles, which emphasise cognitive and affective development and require repetition until the actions become habits.⁴⁵ The implementation of reinforcement strategies to reduce ECC is crucial and has been suggested for either in-home visits or using technology.^7,46, 47, 48, 49

Habit formation theory states that behaviours require ongoing reinforcement through repeated performance to become ingrained habits.^7,50 On average, habit establishment takes 66 days, ranging from 18 to 254 days depending on behaviour complexity.⁵⁰ Our study, conducted over 21 days, revealed consistent changes in toothbrushing habits with fluoride toothpaste and feeding patterns at the 3-month and 6-month evaluations. This finding aligns with several studies that have demonstrated the important impact of parental behaviour on habit development that promotes oral health.51, 52, 53

Knowledge scores increased significantly in both groups on all topics at 3 and 6 months. No significant between-group differences were observed for breastfeeding, dental caries, brushing frequency or posture, or fluoride toothpaste. The chatbot-only group showed comparable knowledge gains to the combined chatbot with brushing training group. These findings align with a systematic review, showing that using exclusively mobile application interventions can improve parental oral health knowledge within 3 months.⁵⁴ Notably, the continuance of oral health behaviours can be inferred from the evaluation of behaviour changes at 3 and 6 months; this may indicate habit formation.

Significant improvements were observed in our study's process (including knowledge, attitudes, and practices), output (plaque reduction), and outcome evaluations (caries control) in both groups. This aligns with the results of several studies that indicate an inverse relationship between the level of caries in young children and the oral health behaviours exhibited by parents or caregivers towards their children.^51,53,55 Aligning with existing research indicating that teledentistry lowers plaque, gingival inflammation, and white spot lesion incidence, this technology could address limitations in traditional oral health care delivery models.⁵⁶

Caries prevalence (dmft, dmfs) showed no significant differences between groups at 3-and 6-month follow-ups, indicating similar caries control efficacy. Decay dominated these indices, highlighting the impact of oral health behaviours on caries progression. Consequently, both interventions demonstrate potential utility in mitigating caries advancement and warrant consideration for clinical application.

It is important to highlight that the majority of studies utilised control groups that received either no intervention or conventional oral health education activities (eg, pamphlet distribution or verbal suggestions) with or without in-person instruction on toothbrushing.^{29,43,44,57,58} In contrast, our study incorporated both intensive in-person toothbrushing training and the innovative chatbot utilising behaviour modification theory, which has been shown to improve young children's oral hygiene.²² This demonstrated encouraging benefits, as the intervention group required fewer resources than the control group, as evidenced by comparable outcomes in both groups.

Interestingly, comprehensive individualised oral health education in controlling ECC has utilised various interactive media, motivational and participatory approaches, and reinforcement via home visits, phone calls, SMS, or other technologies to improve knowledge, attitudes, and practices.^35,56,59,60 Chatbots may integrate these functions through various AI-enabled, theory-based strategies and API technologies to better meet the needs of participants.^14,61 Further assessment of the cost-effectiveness and long-term caries prevention of chatbot interventions compared to alternative methods like home visits and technologies is recommended.

Our chatbots are likely able to provide scalability with diminished human and time resources. Recent interest has focussed on involving nondental providers in promoting oral health for pregnant women and mothers of young children.⁶² Chatbots may aid such efforts, as they can be efficiently scaled to serve large populations without substantially greater resource requirements. Incorporating chatbots into health care could expand population reach and improve health care availability.⁶³ If cost-effective, chatbots could enable customised, population-level ECC prevention education.¹⁰

Parental education predicts ECC⁶⁴ and may affect chatbot adoption. Our study found that 72% to 74% of caregivers had at least high school education, indicating potential for such interventions. In Thailand, high smartphone/internet penetration (>90%) amongst those with primary school education or higher may mitigate educational barriers.⁶⁵ However, lower usage rates amongst those aged 60+ (62.1% internet, 83.6% smartphone), who may care for young children, present a challenge. A rapid review indicates that whilst health care chatbots show promise for improving efficiency and quality, their integration requires careful consideration to ensure equitable access, particularly for underprivileged groups.⁶⁶

Recent studies have found limited evidence of oral health disparities between Muslim and non-Muslim countries, despite lower caries prevalence in Muslim nations.⁶⁷ Islamic teachings emphasise oral hygiene as part of religious practice, encouraging mouth-rinsing and teeth cleaning with miswak before prayers and social interaction and after sleep. These practices may reduce plaque accumulation and oral disease risk.^68,69 Despite the predominance of Muslim mothers as caregivers in our study, who initially showed higher dmft/dmfs of their children than non-Muslim caregivers, postintervention caries increment did not differ significantly between religious groups. This suggested that the interventions effectively modified oral health behaviours regardless of religious affiliation.

Strengths

The randomised allocation of participants into groups ensured balanced distribution of confounding factors, minimising their influence on outcomes. This study had a remarkably low attrition rate of 5.3%. The characteristics of those lost to follow-up closely resembled the remaining sample, suggesting that attrition did not substantially impact results. The use of a blinded examiner unaware of group assignments and self-administered behaviour questionnaires reduced interviewer bias. Standardised examination procedures enhanced data credibility. The application of the ICDAS criteria enabled the early detection of carious lesions. The chatbot's content and dialogue, guided by qualitative research into the reasons, barriers, and solutions to these challenges¹² and underpinned with PMT, are important factors in improving caregivers' oral health behaviour, leading to better oral health.⁴⁴

Limitations

A possible limitation is that the Hawthorne effect could incentivise caregivers to improve their oral hygiene practices for young children during a 3-month evaluation that includes a dental plaque examination and brief questionnaires on oral health behaviour. However, the likelihood of a sustained impact on long-term evaluation is diminished due to the fact that these evaluations were performed on both groups, and the 6-month follow-up occurred, thus far, after this motivational period. Sustained long-term caries prevention through behaviour modification requires age-appropriate ongoing motivational support and the establishment of a healthy environment.

In conclusion, the 3- and 6-month evaluations indicate that the 21-Day FunDee (modified) chatbot, incorporating in-person toothbrushing instruction (Group I), achieved comparable results to the 21-Day FunDee Plus chatbot (Group II) in preventing carious lesions, reducing plaque buildup, improving feeding practices, increasing parental involvement in brushing young children's teeth, and enhancing overall knowledge scores on oral health care. The results of this study indicate that the exclusive chatbot 21-Day FunDee Plus could be effective as an alternative method for promoting young children's oral health.

Author contributions

JH, SP, SN, KP, WT, and PW were involved in all phases of the research methodology, with a particular emphasis on literature review, research design, data collection, intervention training, and the provision of quality control for intervention and data collection. JH and KP were the primary developers of chatbots, which were subsequently improved by SP, SN, WT, and PW. SP, JH and SN were responsible for the conceptualisation and design of the data presentation. SN, JH, and SP addressed the initial draft of the paper. All authors revised the article and to produce the final manuscript.

Conflict of interest

None disclosed.