Abstract
Social participation prevents social isolation and loneliness among older adults while having numerous positive effects on their health and well-being in rapidly aging societies. We aimed to estimate the effect of retaining more natural teeth on social participation among older adults in Japan. The analysis used longitudinal data from 24,872 participants in the Japan Gerontological Evaluation Study (2010, 2013, and 2016). We employed a longitudinal modified treatment policy approach to determine the effect of several hypothetical scenarios (preventive scenarios and tooth loss scenarios) on frequent social participation (1 = at least once a week/0 = less than once a week) after a 6-y follow-up. The corresponding statistical parameters were estimated using targeted minimum loss-based estimation (TMLE) method. Number of teeth category (edentate/1–9/10–19/≥20) was treated as a time-varying exposure, and the outcome estimates were adjusted for time-varying (income, self-rated health, marital status, instrumental activities of daily living, vision loss, hearing loss, major comorbidities, and number of household members) and time-invariant covariates (age, sex, education, baseline social participation). Less frequent social participation was associated with older age, male sex, lower income, low educational attainment, and poor self-rated health at the baseline. Social participation improved when tooth loss prevention scenarios were emulated. The best preventive scenario (i.e., maintaining ≥20 teeth among each participant) improved social participation by 8% (risk ratio [RR] = 1.08; 95% confidence interval [CI], 1.05–1.11). Emulated tooth loss scenarios gradually decreased social participation. A hypothetical scenario in which all the participants were edentate throughout the follow-up period resulted in a 11% (RR = 0.89; 95% CI, 0.84–0.94) reduction in social participation. Subsequent tooth loss scenarios showed 8% (RR = 0.92; 95% CI, 0.88–0.95), 6% (RR = 0.94; 95% CI, 0.91–0.97), and 4% (RR = 0.96; 95% CI, 0.93–0.98) reductions, respectively. Thus, among Japanese older adults, retaining a higher number of teeth positively affects their social participation, whereas being edentate or having a relatively lower number of teeth negatively affects their social participation.
Introduction
The term social participation refers to an individual’s involvement in activities that allow them to interact with others in their community or society in general (Levasseur et al. 2010). Social participation among older adults is an essential component of healthy aging because it has numerous positive effects on both individuals and society (Golinowska et al. 2016). With increasing population aging, older adults’ participation in social activities is becoming a key element to prevent social isolation (World Health Organization [WHO] 2018). Hence, it is important to consider when formulating aging-friendly policies. Previous studies have linked higher levels of social participation to higher life expectancy, lower cognitive decline, well-being, and functioning of older adults (Hikichi et al. 2016; Wanchai and Phrompayak 2018). Community-level health promotion and prevention activities such as physical activity, smoking, and alcohol interventions could also be facilitated through social engagement (Saito et al. 2019). A wide range of determinants, including health-related factors, influence older adults’ level of social participation (Cornwell and Waite 2009; Zhang et al. 2020).
Teeth are important in different aspects of daily life, such as eating, speaking, smiling, and making facial expressions, all of which are essential for positive social interactions. Tooth loss is highly prevalent among older adults due primarily to a lifelong accumulation of chronic dental conditions such as dental caries and periodontal diseases (Bernabe et al. 2020). Previous studies have consistently linked social and neighborhood-related factors such as social capital and social participation to oral health–related outcomes among older adults (Aida et al. 2011; Takeuchi et al. 2013; Rouxel et al. 2015). Much less is known about the effect of oral health on participation in social activities.
The WHO recommends that older adults should have at least 20 occluding teeth, also known as “minimal functioning dentition,” to maintain proper oral function (WHO 2013). This criterion can be used as a benchmark to assess the effect of teeth on social participation among older adults. One way to achieve this is by contrasting the amount of observed social participation against the social participation estimates when participants reach closer to or move away from the minimal functional dentition standard. The longitudinal modified treatment policy (LMTP), a novel nonparametric causal inference approach, can be adapted to obtain such contrasts by emulating multiple hypothetical number of teeth scenarios that mimic tooth loss prevention (i.e., reaching toward the benchmark level) and tooth loss (i.e., moving away from the benchmark level) (Díaz et al. 2021; Ikeda et al. 2022).
The literature for causal inference based on binary exposures is extensive (Höfler 2005). However, dichotomization of the exposure using an arbitrary cutoff point leads to loss of information on the exposure and hinders the ability to observe any “dose–response” effect on the outcome. LMTP, on the other hand, allows us to quantify the effect of a treatment that changes the observed level of exposure in each individual to a new level (Díaz et al. 2021). In other words, this framework can be adapted to quantify counterfactual outcomes for questions such as, “What would happen to the prevalence of social participation if everyone in the study population increased or decreased their number of teeth by a certain amount?” Furthermore, the corresponding statistical parameters for LMTP can be estimated using doubly robust statistical estimators, such as the targeted minimum loss-based estimation (TMLE), avoiding strict parametric modeling assumptions (Schuler and Rose 2016; van der Laan and Rose 2018).
The study aimed to determine the impact of the number of remaining teeth on the social participation of older Japanese adults over a 6-y period, while accounting for the time-dependent nature of the variable associations.
Methods
Data
Data from the Japan Gerontological Evaluation Study (JAGES) were used in this analysis (Kondo et al. 2018). JAGES is an ongoing cohort study for over-65 community-dwelling older adults living across 24 urban and suburban municipalities in Japan. In the baseline (2010), 95,827 postal survey questionnaires were randomly distributed within participating municipalities, and 62,418 people responded (response rate: 65.1%). Out of them, only 52,053 participants were functionally dependent and had valid information about their follow-up status. For this analysis, data from 24,872 individuals who participated in the baseline and 2 subsequent surveys (2013 and 2016) were included. During the 6 y of follow-up, 4,611 people died, 8,099 became functionally dependent, and 14,471 were lost to follow-up due to other reasons (study flowchart in Fig. 1).
Study flowchart indicating the selection of analytical sample over the 6-y follow-up period.
Outcome Variable
The outcome of this study was social participation in 2016. JAGES recorded the frequency of social participation (“nearly every day,” “twice or thrice a week,” “once a week,” “once or twice a month,” “a few times/year,” “never”) for various social activities. We assessed the frequency of participation in any of the following activities: hobby groups, sports clubs, senior citizens’ clubs, residence groups, or volunteer groups. Participation in any of the aforementioned activities once a week or more frequently was defined as indicative of frequent social participation (1 = participation, 0 = nonparticipation). To assess the robustness of the cutoff, a sensitivity analysis was conducted (Appendix Table 1) using once a month or more frequently as the cutoff to define the outcome (Shiba et al. 2021).
Exposure
The number of remaining natural teeth at the time of the surveys in 2010 and 2013 was used as a time-varying exposure. The self-reported number of teeth category was recorded using the response to the question, “How many natural teeth do you currently have?” The response options were 20 or more, 10 to 19, 1 to 9, no teeth. The “20 or more teeth” category was indicative of having a minimal functioning dentition. The 10 to 19 teeth category was intended to represent the initial stages of losing minimal functioning dentition, and the 1 to 9 teeth and edentate categories indicate stages of severe tooth loss (Kassebaum et al. 2014).
Covariates
Because the number of teeth was evaluated as a time-varying exposure in this study, both time-invariant and time-variant covariates were taken into account. Age at the baseline (65–99 y), sex, years of formal education (6/6–9/10–12/13 or more), and social participation in 2010 (outcome at baseline) were adjusted for as time-invariant covariates. Equalized annual household income (million yen), self-rated health (very good/good/fair/poor), instrumental activities of daily living (IADL) score (0–13), vision loss (yes/no), hearing loss (yes/no), cancer (yes/no), heart disease (yes/no), stroke (yes/no), number of household members (1/2/3/4/5/6 or more), and marital status (married/single, widowed, or divorced) were included as time-varying covariates.
Statistical Analysis
Hypothesized associations between variables are shown in the directed acyclic graph in Appendix Figure 1. A descriptive analysis was performed to identify the characteristics of participants stratified by outcome. Then, to specify the impact of number of teeth on social participation, the observed number of teeth of each individual at each time point was shifted to new levels to emulate 4 tooth loss prevention scenarios and 4 tooth loss scenarios using the LMTP framework. Specifically, the following hypothetical scenarios were evaluated. Scenarios for the prevention of tooth loss (S1–S4; see Fig. 2):
1. “What if edentate participants had retained at least 1–9 teeth?”
2. “What if edentate retained 1–9 teeth and participants with 1–9 retained 10–19 teeth?”
3. “What if edentate retained 1–9 teeth and participants with 1–9 retained 10–19 teeth and participants with 10–19 teeth retained ≥20 teeth?”
4. “What if all participants had retained ≥20 teeth?”
Tooth loss scenarios (S5–S8; see Fig. 2):
5. “What if participants with ≥20 teeth became 10–19 teeth?”
6. “What if participants with ≥20 teeth became 10–19 teeth and participants with 10–19 teeth became 1–9 teeth?”
7. “What if participants with ≥20 teeth became 10–19 teeth and participants with 10–19 teeth became 1–9 teeth and participants with 1–9 became edentate?”
8. “What if all participants became edentate?”
Figure 2 illustrates how the observed level of exposure was shifted to emulate the above exposure scenarios. Furthermore, Appendix Figure 2 shows how these scenarios were emulated in a longitudinal setting.
Illustration of how the observed level of exposure was shifted to emulate multiple exposure scenarios. Numbers refer to the number of teeth. Eden, edentulous.
TMLE was used to estimate the level of social participation with the shifted and the observed exposures (Díaz et al. 2021). In TMLE, the probabilities of the exposure conditional on the covariates (exposure model) and the conditional probabilities of the outcome given the exposure and covariates (g-computation/outcome model) were estimated. Then, to obtain unbiased estimation of the counterfactual outcomes, g-computation estimates were updated using negative inverse probability weights derived from the propensity score model (Schuler and Rose 2016). Therefore, if either the exposure model or the outcome model was consistently estimated, unbiased estimates could be obtained (van der Laan and Gruber 2012). To increase the likelihood of robust specification of exposure and outcome models, Super Learner algorithms was used (Schomaker et al. 2019). Within the Super Learner, generalized linear models (glm), generalized additive models (gam), and extreme gradient boosting models (xgboost) were used. Additional information regarding the usage of Super Learner in this analysis is provided in the Appendix Text Box 1.
Finally, the estimates of each emulated hypothetical scenario were contrasted against the outcome estimate under the observed exposure to calculate risk ratios (RRs) and 95% confidence intervals (CIs) for each respective scenario. All estimates were appropriately controlled for abovementioned time-variant and time-invariant covariates. In addition, estimates were accounted for attrition of the study population (Lendle et al. 2017). A comparison of baseline characteristics by participants’ follow-up status is reported in Appendix Table 2. Furthermore, corresponding E values were calculated for each RR estimate to report the potential impact of unmeasured confounders (VanderWeele and Ding 2017). Finally, a supplementary logistic regression analysis using baseline exposure and covariates was conducted to assess the difference in estimates using the traditional method and the counterfactual-based LMTP approach (Appendix Table 3).
Random forest–based multivariate imputation by chained equations (MICE) was used to impute missing data (Van Buuren and Groothuis-Oudshoorn 2011). Random forest MICE has been shown to produce less biased parameter estimates compared to parametric MICE (Shah et al. 2014). Analyses were performed using 5 imputed data sets, and the estimates were pooled using Rubin’s rules (Rubin 2004). Percentages of missingness for each covariate are shown in Appendix Figures 3 and 4. The lmtp R package was used to emulate exposure scenarios and to compute TMLE estimates (Williams and Díaz 2020). All R codes used to generate our results can be found at https://github.com/upulcooray/social_participation. All the analyses were performed using R version 4.1.2 for x86_64,linux-gnu. Reporting of this study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.
Results
Baseline characteristics of participants stratified by the outcome variable are presented in Table 1. In the 2016 follow-up, 12,079 (52.4%) reported a social participation frequency of less than at least once a week. Compared to the baseline, a total of 1,817 (7.9%) participants reported a lower number of teeth category in 2013. Baseline characteristics associated with less frequent social participation in 2016 were older age, male sex, low income, low educational attainment, poor self-reported health, and lower frequency of social participation at baseline.
Baseline Characteristics of Participants Stratified by the Outcome.
IADL, instrumental activities of daily living.
Mean (SD) for continuous variables; frequency (%) for categorical variables.
Table 2 provides risk ratios related to preventive and tooth loss scenarios after adjusting for the covariates and the censoring during the follow-up. The results showed that the prevention of tooth loss had a positive effect on social participation. The largest improvement (8%) in social participation was observed with the scenario that retained ≥20 teeth among all older adults at each time point during the follow-up (scenario 4: RR = 1.08; 95% CI, 1.05–1.11). The intervention that prevented individuals with 1 to 9 teeth becoming edentate (scenario 1) and the intervention that prevented tooth loss in individuals with only 1 to 9 teeth and 10 to 19 teeth (scenario 2) did not significantly improve social participation. On the other hand, all emulated tooth loss scenarios (scenarios 5–8) reduced social participation. The hypothetical scenario in which all participants became edentate (Fig. 2; scenario 8) resulted in a 11% reduction in the likelihood of frequent social participation among participants (RR = 0.89; 95% CI, 0.84–0.94). The rest of the tooth loss scenarios resulted in 8% (scenario 7: RR = 0.92; 95% CI, 0.88–0.95), 6% (RR = 0.94; 95% CI, 0.91–0.97), and 4% (RR = 0.96; 95% CI, 0.93–0.98), respectively, in the descending order of severity of tooth loss. The RR plot in Figure 3 indicates the presence of dose–response relationship between tooth loss and social participation among older adults.
Risk Ratios (RRs) and 95% Confidence Intervals (CIs) Calculated by Contrasting Emulated Scenarios against the Observed Outcome Estimate.
Scenario 1: What if edentate were 1–9? Scenario 2: What if edentate were 1–9 and 1–9 were 10–19? Scenario 3: What if edentate were 1–9, 1–9 were 10–19, and 10–19 were ≥20? Scenario 4: What if everyone were ≥20? Scenario 5: What if ≥20 were 10–19? Scenario 6: What if ≥20 were 10–19 and 10–19 were 1–9? Scenario 7: What if ≥20 were 10–19, 10–19 were 1–9, and 1–9 were edentate? Scenario 8: What if everyone were edentate?
Risk ratio plot to indicate changes in social participation in response to number of teeth scenarios.
Discussion
Our findings show that retaining more teeth during the follow-up period had a positive effect on social participation among older adults, whereas a decrease in the number of teeth during the follow-up had a negative effect on social participation among study participants. These findings support our hypothesis and are consistent with previous related research. Previous studies, however, used the number of teeth as the outcome variable (Aida et al. 2011; Takeuchi et al. 2013). Using longitudinal data and a robust causal inference method, this study added evidence related to the importance of maintaining an adequate number of teeth for frequent social participation among older adults. Given the consistent evidence that social participation improves older adults’ health and well-being, mechanisms that lead to increased levels of social participation should be promoted and encouraged. In this context, our findings emphasize the importance of older adults retaining a greater number of teeth, not only for obvious benefits on oral functions, such as mastication and speech, but also to have better social relationships and thus reap the benefits associated with social participation.
The mechanism that explains our findings is straightforward and intuitive. Teeth play an important role in social interactions such as smiling, speaking, eating, and maintaining facial aesthetics (Steele et al. 2004). As a result, tooth loss would naturally lead to a reluctance to engage in social activities. A recent cross-sectional study by Koyama et al. (2021) examined the association between the number of teeth and social isolation among older adults using data from Japan and England. They found that having fewer teeth was significantly associated with being socially isolated in both countries. Abbas et al. (2022) also found a similar association between teeth and dental prosthesis and being socially isolated in a longitudinal study. Although Koyama et al. and Abbas et al. investigated a different outcome, the mechanism between the number of teeth and social isolation may be similar to that of the current study.
The analytical approach used in this study allowed us to obtain counterfactual estimates without needing to dichotomize the exposure variable (number of teeth), thus enabling the detection of gradual changes in social participation. Traditional methods contrast counterfactual outcomes only at the extremes of the exposure (i.e., “what if everyone is exposed vs. everyone is not exposed”). For example, with the same data, a traditional method would only allow us to estimate the difference between the counterfactual outcomes of being edentate versus having teeth or having ≥20 teeth versus having <20 teeth, which can be an unrealistic contrast (Rudolph et al. 2021). Furthermore, the LMPT approach minimized the positivity assumption violations (Petersen et al. 2010) (i.e., all had a nonzero probability of obtaining a given exposure level) as the counterfactual exposure levels (shifted levels) were assigned based on individuals’ observed number of teeth level at a particular time point (Appendix Fig. 2). Also, by using TMLE to estimate corresponding statistical parameters, we were able to minimize parametric modeling assumptions regarding variables (Rose and Rizopoulos 2019; Díaz et al. 2021). Considering that the estimates in this study were contrasted against the natural outcome, we believe that our estimates are conservative. Therefore, estimated effect sizes were relatively small (e.g., retaining ≥20 teeth improved social participation only by 8%). It might be unrealistic for a single exposure such as number of teeth alone to have a large effect on a complex behavioral outcome such as social participation. Thus, our estimates are grounded in the epidemiological reality that outcomes related to well-being need multisectoral effort (Amri et al. 2022), and oral health may be a small yet important part of it. Although it is difficult to quantify how a given percentage increase in social participation translates into meaningful and desirable health or quality-of-life outcomes, any improvement in social participation due to retention of more teeth should be considered beneficial among the older population.
Even though we used hypothetical scenarios to estimate our research question, these scenarios are embedded in any real-world oral health promotion activities or interventions aimed at achieving at least minimal functional dentition in older adult populations. Realistically, to ensure at least a minimally functional dentition in old age, oral disease prevention should be an integral part of one’s life course (Heilmann et al. 2015). Suboptimal emulated scenarios (e.g., preventing 10–19 teeth from becoming 1–9 teeth) in this study might be more in line with targeted interventions toward older adults, such as improving access to dental care by providing financial assistance (Cooray et al. 2020), orienting services to be aging-friendly, and collaborating with other geriatric health care providers to identify vulnerable groups for early interventions.
We note several limitations of our analysis and the data that may cause the estimates to be biased. First, the variables in this study were self-reported, which are prone to measurement and classification errors. Previous studies in Japan, however, have shown the validity of the self-reported number of teeth measure (Matsui et al. 2016). Second, causal inference with time-varying exposure necessitates no unmeasured confounding assumption at each time point (conditional exchangeability assumption) (Hernan 2006). Therefore, despite adjusting for multiple time-varying and time-invariant confounders, the possibility of unmeasured confounding cannot be ruled out. We reported E values for estimates to reflect the potential effect of unmeasured confounding (VanderWeele and Ding 2017). Third, a large attrition of the sample population within 6 y (n = 52,053 at baseline to n = 24,872 at 2016 follow-up) was unavoidable as we used panel data with older adult participants who took part in all 3 waves of the JAGES. To minimize the bias due to this attrition, censoring status of all individuals was modeled into our analysis, obtaining estimates accounted for censoring (Lendle et al. 2017). In addition, we examined the baseline characteristics associated with censoring. Censoring was associated with a lower number of teeth at baseline. Having fewer teeth had a negative impact on social participation in our analyses. Fourth, in this study, only the organized social activities were captured. However, it might be useful to include informal social interactions as well. Fifth, the incidence of tooth loss in the observed data was only 7.6%; given the large number of covariates considered in this study, the possibility of positivity violation is higher when emulating counterfactual scenarios. Finally, we had no information about the locations of missing teeth in our data. Missing anterior teeth have a greater impact on facial aesthetics and speech, whereas missing posterior teeth would have a greater impact on masticatory functions. As a result, the location of missing teeth would have had a different effect on social participation.
Despite these limitations, our findings provide robust evidence that retaining more teeth is positively associated with frequent social participation among Japanese older adults, whereas tooth loss negatively affects their social participation. This emphasizes the importance of incorporating tooth loss prevention into interventions aimed at increasing social participation among older adults.
Conclusion
Hypothetical scenarios for tooth loss prevention improved social participation among Japanese older adults, whereas emulated tooth loss had negative effects. This suggests that retaining more natural teeth has a positive impact on social participation among older adults in Japan.
Author Contributions
U. Cooray, contributed to conception and design, data analysis and interpretation, drafted the manuscript; G. Tsakos, J. Aida, contributed to conception and design, data interpretation, critically revised the manuscript; A. Heilmann, R.G. Watt, contributed to data interpretation, critically revised the manuscript; K. Takeuchi, contributed to data design, critically revised the manuscript; K. Kondo, contributed to data acquisition, critically revised the manuscript; K. Osaka, contributed to conception, data acquisition, critically revised the manuscript. All authors gave their final approval and agree to be accountable for all aspects of the work.
Supplemental Material
sj-docx-1-jdr-10.1177_00220345231164106 – Supplemental material for Impact of Teeth on Social Participation: Modified Treatment Policy Approach
Supplemental material, sj-docx-1-jdr-10.1177_00220345231164106 for Impact of Teeth on Social Participation: Modified Treatment Policy Approach by U. Cooray, G. Tsakos, A. Heilmann, R.G. Watt, K. Takeuchi, K. Kondo, K. Osaka and J. Aida in Journal of Dental Research
Footnotes
Acknowledgements
The authors gratefully acknowledge all participants of the JAGES survey. JAGES was supported by MEXT (Ministry of Education, Culture, Sports, Science, and Technology–Japan) –Supported Program for the Strategic Research Foundation at Private Universities (2009–2013); JSPS (Japan Society for the Promotion of Science); KAKENHI grants (JP15H01972, JP15H04781, JP15H05059, JP15K03417, JP15K03982, JP15K16181, JP15K 17232, JP15K18174, JP15K19241, JP15K21266, JP15KT0007, JP15KT0097, JP16H05556, JP16K09122, JP16K00913, JP16K 02025, JP16K12964, JP16K13443, JP16K16295, JP16K16595, JP16K16633, JP16K17256, JP16K19247, JP16K19267, JP16K 21461, JP16K21465, JP16KT0014, JP18KK0057, JP19K24060, JP19H03860, JP18390200, JP22330172, JP22390400, JP2324 3070, JP23590786, JP23790710, JP24390469, JP24530698, JP24 683018, JP25253052, JP25870573, JP25713027, JP25870881, JP26285138, JP26460828, JP26780328, JP26882010, 23243070, 22390400, 24390469, 15H01972, 20H00557, 22H03299); the Japanese Ministry of Health, Labour, and Welfare, Health and Labour Sciences research grants (H22-Choju-Shitei-008,H24-Junkanki [Seishu]-Ippan-007, H24-Chikyukibo-Ippan-009, H24-Choju-Wakate-009, H25-Kenki-Wakate-015, H25-Choju-Ippan-003, H26-Irryo-Shitei-003 [Fukkou], H26-Choju-Ippan-006, H26-Choju-Ippan-006, H27-Ninchisyou-Ippan-001, H28-Choju-Ippan-002, H28- Ninchisyou-Ippan-002, H30-Kenki-Ippan-006 and H30-Junkankitou-Ippan-004, 19FA2001, 19FA1012); AMED (the Japan Agency for Medical Research and Development) grants (16dk0110017h0002, 16ls0110002h0001, JP17dk0110017, JP18dk0110027, JP18ls0110002, JP18le0110009, JP19dk0110034, JP20dk0110034); the Japanese National Center for Geriatrics and Gerontology, Research Funding for Longevity Sciences grants (20-19, 24-17, 24-23, 29-42, 30-22); the World Health Organization Centre for Health Development (WHO Kobe Centre) grant (WHOAPW 2017/713981); and JST (Japan Science and Technology Agency) OPERA: JPMJOP1831.
A supplemental appendix to this article is available online.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
