Dental Care Use, Edentulism, and Systemic Health among Older Adults

Abstract

Past research suggests there are systematic associations between oral health and chronic illness among older adults. Although causality has not yet been credibly established, periodontitis has been found to be associated with higher risk of both heart disease and stroke. We advance this literature by estimating the direct association between dental care use and systemic health using multiple waves of the 1992 to 2016 Health and Retirement Study. Through the inclusion of individual fixed effects in our regression models, we account for unobservable time-invariant characteristics of individuals that might otherwise bias estimates of the association between dental care use and health. We find statistically significant negative associations between dental care use and the number of health conditions, self-reported overall health, the incidence of heart disease, and the incidence of stroke. In particular, the use of dental care within the past 2 y is associated with a 2.7% reduction in the likelihood of being diagnosed with a heart condition and a reduction in the likelihood of a stroke diagnosis of between 5.3% and 11.6%. We also find large positive correlations between edentulism and the measures of chronic illness. Associations from models estimated separately for men and women are qualitatively similar to one another. These findings provide additional motivation for the consideration of a Medicare dental benefit.

Introduction

Connections between oral disease and comorbid chronic illnesses in elderly individuals have long been suspected (United States Department of Health and Human Services [US DHHS] 2000). Retrospective studies provide evidence of an association between periodontitis and several illnesses, including coronary heart disease (CHD) and stroke. Consider, for example, evidence on the association between oral health and stroke. Grau et al. (2004) found that patients with severe periodontitis had a 4.3 times higher risk of acute or transient ischemic stroke than those with good oral health. Likewise, a study among patients with an initial ischemic attack in the past 5 d found that a probing pocket depth of >4.5 mm was the largest risk factor for stroke, with an odds ratio of 8.5 (CI 1.1–68.2; Pradeep et al. 2010). Lee et al. (2013) found that those with treated periodontitis had a lower instantaneous probability of suffering from stroke than the control group did, whereas those with untreated periodontitis had a higher probability of suffering from stroke. All of these studies are consistent with a systematic review by Pillai et al. (2018), who concluded that periodontitis and tooth loss were independently associated with stroke.

Studies on a potential relationship between periodontitis and CHD predate studies on stroke, but weaknesses in the design of early investigations made it difficult to identify a causal relationship (Beck et al. 1998). Subsequent research has targeted chronic inflammation from periodontitis as a mechanism to draw a more direct link to CHD (Spahr et al. 2006), and a systematic review and meta-analysis found periodontal disease increases the risk of CHD independent of traditional risk factors (Humphrey et al. 2008). Recently, a workshop jointly organized by the European Federation of Periodontology and World Heart Federation detailed the significant body of epidemiological evidence and studies of mechanistic links between periodontitis and cardiovascular disease more broadly (Sanz et al. 2020). Indicative of the relationship between periodontitis and both CHD and stroke, studies have identified a link between periodontitis and mortality, both directly and indirectly, through edentulism (Paganini-Hill et al. 2011; Kim et al. 2013).

In addition to cardiovascular and cerebrovascular diseases, periodontitis has been linked to other chronic conditions, such as diabetes (Preshaw et al. 2012, Borgnakke et al. 2013; Lee et al. 2015; Graziani et al. 2018), Alzheimer’s disease (Teixeira et al. 2017), and cancer (Corbella et al. 2018), although the evidence for many of these conditions is not conclusive. Studies also found reductions in health care costs associated with heart disease and diabetes following periodontal therapy (Paganini-Hill et al. 2011; Kim et al. 2013; Jeffcoat et al. 2014; Nasseh et al. 2017).

The hypothesized mechanism through which periodontitis, or poor oral health more generally, is assumed to affect chronic illness is persistent inflammation (Bartold and Van Dyke 2013; Holmstrup et al. 2017; Sanz et al. 2020). Specifically, it is thought that the translocation of oral microbiota could directly or indirectly lead to inflammation that contributes to atherothrombosis, causing myocardial infarction, angina, or cerebral ischemia (Reyes 2013; Schenkein and Loos 2013; Sanz et al. 2020).

At least 60% of seniors in the United States have moderate or severe gum disease, which is a cause for concern and provides an incentive to better understand how to reduce the impact of poor oral health on chronic illness (Eke et al. 2012). Despite numerous studies on the link between oral health and chronic disease, there is much less empirical evidence on the effectiveness of dental care in reducing chronic illness. In this article, we extend the literature on the relationship between poor oral health and chronic illness by estimating the direct association between dental care use and systemic health. We use longitudinal data from the 1992 to 2016 waves of the Health and Retirement Study (HRS), which affords several advantages in our analysis. First, the HRS contains general measures of overall physical health as well as indicators for 2 chronic conditions that have been shown to be consistently associated with poor oral health (heart disease and stroke) and 2 conditions for which there is more limited evidence of an association (cancer and diabetes). The data also contain a measure of edentulism, which past research has shown to be correlated with chronic illness. Second, because the HRS contains multiple observations for each respondent over time, we are able to conduct a within-person (i.e., fixed effects) analysis and control for unobserved time-invariant characteristics of individuals that could confound relationships between dental care use and health. In past studies that use cross-sectional data, unobservable individual characteristics could bias estimates of the association between oral health and chronic illness.

Methods

Data Source

The HRS is a nationally representative longitudinal household survey in the United States that collects self-reported data from interviews with individuals older than 50 y and their spouses every 2 y. We used the 1992 to 2016 RAND HRS longitudinal file containing 42,053 elderly individuals. ¹ After removing individuals with zero person weights, our analysis used data on 38,702 respondents who, on average, were observed in 5.4 survey waves (over approximately 11 y). ² However, because of the nonresponse for some dependent and key independent variables, or because some questions were asked in only a limited number of survey waves, some models were estimated using smaller sample sizes.

From the self-reported HRS data on medical conditions, we created 2 measures of overall health and 6 indicators for specific chronic conditions. The measures of overall health included an index of self-reported health ranging from excellent (1) to poor (5) and the total number of health conditions that the respondent had been told she has by a doctor. The measures of specific chronic conditions included indicators for cancer, diabetes, heart disease, and stroke. For each condition, we created a variable capturing whether the respondent had been diagnosed with the condition since the last survey wave (approximately 2 y prior) and a variable measuring whether a doctor had ever told the respondent that she has the condition in the current survey wave or in any prior wave. The former variables were coded as 1 in the wave in which the respondent first reported the diagnosis of the health condition and as 0 in all other waves, whereas the latter variables were coded as 1 in the wave when the respondent first reported the diagnosis as well as in all subsequent waves. ³

The HRS contains 1 measure of dental care use in each wave, specifically, whether the respondent visited a dentist at least once over the previous 2 y. There is also a measure of oral health in waves 8 to 13 in the form of a question asking respondents whether they have lost all of their upper and lower permanent teeth. We created indicator variables for past dental visits and being edentulous. In addition, we created a separate indicator variable to capture whether the respondent failed to answer the question about lost teeth.

We generated several control variables from the core HRS to account for sociodemographic factors that are correlated with physical health and chronic disease. These included controls for age, race/ethnicity, gender, census division of residence, family structure (marital status, household composition), labor market participation, household income, health insurance coverage (Medicare, Medicaid, private insurance, other insurance), and the number of difficulties with activities of daily living (ADLs) or instrumental activities of daily living (IADLs).

Table 1 contains the definition, mean, standard deviation, and sample size of all our dependent variables measuring health status and chronic conditions and our independent variables. The mean number of health conditions reported by an HRS respondent was 1.9, and self-reported health was “good,” on average. In the preceding 2 y, 2.3% and 1.8% of individuals were diagnosed with cancer and stroke, respectively, whereas 3.5% and 2.6% of individuals were diagnosed with a heart problem and diabetes, respectively. Of the elderly individuals, 23% and 19% had been told in the past by a doctor that they had a heart problem and diabetes, respectively. A total of 13% and 8.3% of individuals had previously been told that they had cancer and stroke, respectively. Of the respondents, 62% had visited the dentist within the past 2 y, and 15% were edentulous.

Table 1.

Definitions and Summary Statistics of Variables.

Variable	Definition	Mean	SD	NT
Dependent variables
Self-reported health	Self-report of health: excellent (1); very good (2); good (3); fair (4); poor (5)	2.8886	1.1247	210,239
Number of conditions	Number of health conditions doctor has ever told the respondent he or she had	1.9263	1.4632	210,356
Cancer_last 2 y	Respondent had cancer since last IW (waves 2–13)	0.0228	0.1491	183,309
Cancer_ever	Doctor ever told respondent he or she had cancer or a malignant tumor	0.1314	0.3378	209,978
Diabetes_last 2 y	Respondent had diabetes since last IW (waves 2–13)	0.0264	0.1602	183,410
Diabetes_ever	Doctor ever told respondent he or she had diabetes	0.1896	0.3920	210,088
Heart_last 2 y	Respondent had heart problems since last IW (waves 2–13)	0.0346	0.1829	183,456
Heart_ever	Doctor ever told respondent he or she had heart problems	0.2288	0.4201	210,117
Stroke_last 2 y	Respondent had stroke since last IW (waves 2–13)	0.0181	0.1335	183,540
Stroke_ever	Doctor ever told respondent he or she had stroke or transient ischemic attack	0.0830	0.2760	210,200
Independent variables
Dental visit	At least 1 dental visit in past 2 y	0.6152	0.4865	210,356
Lost teeth	Lost all upper and lower natural permanent teeth (waves 8–13)	0.1528	0.3598	109,820
Lost teeth_missing a	Lost teeth infomation missing (waves 8–13)	0.3034	0.4597	109,820
Age <65 y	Age <65 y	0.4404	0.4964	210,356
Age 65–69 y	65 ≤ Age ≤ 69 y	0.1420	0.3491	210,356
Age 70–74 y	70 ≤ Age ≤ 74 y	0.1411	0.3482	210,356
Age 75–79 y	75 ≤ Age ≤ 79 y	0.1188	0.3235	210,356
Age 80–84 y	80 ≤ Age ≤ 84 y	0.0841	0.2775	210,356
Hispanic	Race/ethnicity: Hispanic	0.1030	0.3039	210,356
Black	Race/ethnicity: Black, non-Hispanic	0.1579	0.3647	210,356
Other race	Race/ethnicity: other, non-Hispanic	0.0262	0.1598	210,356
Race_missing	Race/ethnicity: missing	0.0016	0.0405	210,356
Female	Gender is female	0.5843	0.4928	210,356
Married	Current marital status: married	0.6139	0.4869	210,356
Single	Current marital status: single	0.3499	0.4769	210,356
Partnered	Current marital status: partnered	0.0354	0.1848	210,356
Residents	Number of residents in the household	2.2456	1.2231	210,356
Children	Number of living children	3.1680	2.1572	210,356
Couple	Whether household is a couple	0.6493	0.4772	210,356
Retired	Respondent not in the labor force and is fully retired	0.4766	0.4995	210,356
Partly retired	Respondent is partly retired	0.1598	0.3664	210,356
Full-time	Respondent works full-time	0.2863	0.4520	210,356
Unemployed	Respondent is unemployed	0.0114	0.1061	210,356
New England	Census division: New England	0.0373	0.1894	210,356
Mid Atlantic	Census division: Mid Atlantic	0.1199	0.3249	210,356
WN Central	Census division: West North Central	0.0773	0.2671	210,356
S Atlantic	Census division: South Atlantic	0.2372	0.4254	210,356
ES Central	Census division: East South Central	0.0583	0.2343	210,356
WS Central	Census division: West South Central	0.1078	0.3102	210,356
Mountain	Census division: Mountain	0.0549	0.2277	210,356
Pacific	Census division: Pacific	0.1286	0.3348	210,356
Not US	Census division: not US or US territory	0.0015	0.0389	210,356
Census_missing	Census division is missing	0.0194	0.1378	210,356
Income	Real household income/person (2016 USD)	37,046	171,503	210,356
Medicare	Respondent enrolled in Medicare	0.5820	0.4932	210,356
Medicaid	Respondent enrolled in Medicaid	0.0932	0.2908	210,356
Private insurance	Respondent enrolled in commercial insurance	0.4759	0.4994	210,356
Other insurance	Respondent enrolled in another insurance plan	0.1900	0.3923	210,356
No. of insur. plans	No. of private insurance plans respondent enrolled in	0.6766	0.5988	210,356
No ADL	Respondent has no ADLs (waves 2–13)	0.8298	0.3758	210,356
No IADL	Respondent has no IADLs (waves 2–13)	0.8396	0.3670	210,356
Number of ADLs	Number of ADL (waves 2–13)	0.3378	0.8786	210,356
Number of IADLs	Number of IADL (waves 2–13)	0.3480	0.9729	210,356

Means are weighted to be nationally representative. ADL, activity of daily living; IADL, instrumental activity of daily living; IW, interview wave; NT, (number of respondents)*(number of time periods) is the total number of observations.

^aWe coded 395 respondents (1,189 observations) who gave conflicting answers in different rounds to the question about lost teeth as missing. Other missing values are due to nonresponse or our inability to determine edentulism for reinterviews in waves 9, 10, 12, and 13.

Empirical Models

Our empirical analysis is based on the following linear regression model:

$$Y_{it}=\alpha +\beta \cdot Denta{l}_{it}+{\gamma }^{\prime }\cdot X_{it}+c_i+w_t+{\epsilon }_{it}$$ (1)

The outcome variable Y_it is either an index measuring self-reported health or the total number of physical health conditions, or it is a binary variable measuring the chronic condition incidence during the past 2 y or at any point in the past for person i in year (wave) t. Dental_it is either the indicator for visiting a dentist in the past 2 y or the absence of all permanent teeth, X_it is a vector of sociodemographic control variables, c_i is an indicator variable for person (i.e., an individual fixed effect), w_t is an indicator variable for year (i.e., a time fixed effect), ε_it is a white noise error term, and α, β, and γ are parameters to be estimated. When Y_it is a binary variable, β measures an association in percentage points. The association can be converted to a percentage by dividing the coefficient by the mean of the dependent variable and multiplying by 100, which we report in the text to facilitate comparisons across models.

The inclusion of individual fixed effects (c_i) in the model removes the confounding influence of time-invariant unobservable factors, such as persistent health behaviors or attitudes toward risk or medical/dental care. Likewise, the inclusion of time fixed effects (w_t) captures any unobserved aggregate-level factors that vary over time but are fixed across individuals, such as macroeconomic shocks. In addition to estimating fixed effect models, we also estimate conventional ordinary least squares (OLS) regressions without individual fixed effects for comparison purposes. In all models, we use the HRS survey weights to make the estimates nationally representative and adjust the standard errors of estimates for the complex design of the HRS by accounting for clustering on primary sampling units and stratification.

Results

Table 2 contains ordinary OLS estimates without individual fixed effects (denoted OLS) and fixed effect estimates (denoted FE) of the effect of having a dental visit in the past 2 y and having lost all permanent teeth on the health measures. In panel A, we report estimates from models in which the dependent variable is either the number of health conditions or a specific chronic condition that was diagnosed at any time before the survey date. The ordinary OLS correlations between having a dental visit and the chronic condition measures were negative and statistically significant, with the exception of cancer, which had a small positive correlation with dental care. After we controlled for time-invariant unobservable factors using the fixed-effect model, all correlations were negative, as expected, and the magnitudes of the effects were smaller. In particular, visiting a dentist at least once in the past 2 y was associated with 0.03 fewer health conditions, a 0.62 percentage point lower likelihood of ever being diagnosed with a heart condition (a 2.7% reduction relative to the overall sample mean of the dependent variable), and a 0.44 percentage point (5.3%) lower likelihood of a stroke diagnosis. The full estimation results for these ordinary OLS and fixed effect models are reported in Appendix Tables 1 and 2.

Table 2.

Impact of Having a Dental Visit and the Loss of Teeth on Health Conditions.

	Dental Visit			Lost Teeth
	OLS	FE	NT	OLS	FE	NT
A. Person diagnosed at any time  prior to survey wave
Number of conditions	−0.1352*** (0.0151)	−0.0330*** (0.0074)	210,356	0.4263*** (0.0304)	0.5941*** (0.0276)	109,975
Cancer	0.0076** (0.0034)	−0.0030 (0.0020)	209,978	0.0136* (0.0076)	0.0548*** (0.0084)	109,728
Diabetes	−0.0277*** (0.0035)	−0.0034 (0.0022)	210,088	0.0625*** (0.0067)	0.0660*** (0.0074)	109,797
Heart disease	−0.0221*** (0.0036)	−0.0062*** (0.0019)	210,117	0.0691*** (0.0085)	0.0955*** (0.0118)	109,804
Stroke	−0.0102*** (0.0023)	−0.0044*** (0.0014)	210,200	0.0218*** (0.0063)	0.0263*** (0.0064)	109,863
B. Person diagnosed within the  past 2 y
Self-reported health	−0.2684*** (0.0096)	−0.0270*** (0.0074)	210,239	0.2663*** (0.0169)	0.1403***(0.0244)	109,900
Cancer	−0.0009 (0.0009)	−0.0007 (0.0015)	183,309	0.0006 (0.0017)	0.0021 (0.0061)	99,674
Diabetes	−0.0027*** (0.0008)	0.0014 (0.0014)	183,410	0.0020 (0.0020)	−0.0088 (0.0060)	99,738
Heart disease	−0.0035*** (0.0011)	−0.0001 (0.0015)	183,456	0.0072*** (0.0021)	0.0154** (0.0076)	99,752
Stroke	−0.0025*** (0.0007)	−0.0021* (0.0011)	183,540	0.0024* (0.0013)	−0.0066 (0.0053)	99,816

Standard errors in parentheses are adjusted for the complex design of the Health and Retirement Study. All regressions include the covariates listed in Table 1, with the exception of Lost Teeth_Missing, which is included only in the lost teeth regressions. NT, (number of respondents)*(number of time periods) is the total number of observations; OLS, Ordinary Least Squares; FE, Fixed Effect Estimates.

^***P < 0.01, **P < 0.05, *P < 0.1.

The ordinary OLS and fixed effect correlations between the loss of all permanent teeth and the chronic conditions are all positive and statistically significant. The fixed effect correlations are larger than those from the ordinary OLS models and suggest a strong association between edentulism and poor health. For example, we estimate that edentulism is associated with a 5.5 percentage point (41.7%) increase in a cancer diagnosis, 6.6 percentage point (34.8%) increase in diabetes, 9.6 percentage point (41.7%) increase in heart disease, and a 2.6 percentage point (31.7%) greater likelihood of stroke.

Panel B of Table 2 contains estimates from models in which the dependent variable is either the index of self-reported health or a specific chronic condition that was diagnosed within the past 2 y. The OLS correlations of having visited a dentist at least once in the past 2 y and the outcome variables are all negative, and correlations with self-reported health, diabetes, heart disease, and stroke are statistically significant. In the fixed effect models, the magnitudes of the estimates are small, but the correlations for self-reported health and stroke remain statistically significant. Specifically, having visited a dentist is associated with a small improvement in self-reported health of 0.03 index points (0.9%) and a 0.2 percentage point reduction (11.6%) in the incidence of stroke in the previous 2 y. The OLS correlations between having lost all permanent teeth and self-reported health, heart disease, and stroke are all positive and statistically significant, but only correlations with self-reported health and heart disease are statistically significant in the fixed effect model. The latter indicate that the loss of permanent teeth is associated with a reduction in self-reported health of 0.14 index points (4.9%) and a 1.5 percentage point (44.5%) higher likelihood of a recent heart disease diagnosis.

In Table 3, we report estimated fixed effect correlations between outcome variables measuring the number of conditions and the diagnosis of specific conditions at any time prior to the survey date with having visited a dentist in the past 2 y and with edentulism, separately by gender. All correlations in panel A are statistically significant, except for the association between visiting a dentist and cancer in men and the associations between visiting a dentist and diabetes in both men and women. Overall, the fixed effect correlations are similar for men and women. Relative to the respective sample mean, the associations between visiting a dentist and both the number of medical conditions and a cancer diagnosis are larger for women, but the associations with heart disease and stroke are larger for men. Associations in panel B between edentulism and both the number of health conditions and the incidence of stroke are also similar for men and women. However, when measured relative to sample means, there is a larger association between the loss of all permanent teeth and both cancer and diabetes among women and a larger association with heart disease among men.

Table 3.

Impact of Having a Dental Visit and Edentulism on the Diagnosis of Conditions at Any Time Prior to Survey Wave from Fixed Effect Model, by Gender.

	Men	NT	Women	NT
A. Dental visit
Number of conditions	−0.0288** (0.0111)	87,435	−0.0354*** (0.0091)	122,921
Cancer	0.0006 (0.0030)	87,260	−0.0055** (0.0022)	122,718
Diabetes	−0.0050 (0.0035)	87,326	−0.0019 (0.0026)	122,762
Heart disease	−0.0078** (0.0032)	87,341	−0.0048* (0.0028)	122,776
Stroke	−0.0055*** (0.0021)	87,361	−0.0034* (0.0020)	122,839
B. Lost teeth
Number of conditions	0.5869*** (0.0450)	45,993	0.5939*** (0.0407)	63,982
Cancer	0.0455*** (0.0112)	45,889	0.0612*** (0.0116)	63,839
Diabetes	0.0619*** (0.0130)	45,928	0.0686*** (0.0088)	63,869
Heart disease	0.1162*** (0.0203)	45,928	0.0773*** (0.0113)	63,876
Stroke	0.0272** (0.0115)	45,942	0.0251*** (0.0065)	63,921

Standard errors in parentheses are adjusted for the complex design of the Health and Retirement Study. All regressions include the covariates listed in Table 1, with the exception of Lost Teeth_Missing in panel A. NT, (number of respondents)*(number of time periods) is the total number of observations.

^***P < 0.01, **P < 0.05, *P < 0.1.

Discussion

Both our OLS and fixed effect estimates indicate that dental care during the previous 2 y is associated with better health and a lower probability of suffering from heart disease or stroke. However, the magnitudes of the fixed effect estimates are smaller and somewhat less precise than the OLS estimates are. For example, lack of dental care use is associated with a 9.7% greater likelihood of ever being diagnosed with heart disease in the OLS model but only a 2.7% greater likelihood of diagnosis in the corresponding fixed effect model. Furthermore, none of the fixed effect associations of dental care use with cancer or diabetes are statistically significant. This is consistent with the conclusions from systematic reviews and meta-analyses that find evidence of credible associations between periodontitis and both CHD and stroke but inconclusive evidence of an association with cancer (Humphrey et al. 2008; Corbella et al. 2018; Pillai et al. 2018; Sanz et al. 2020). In the case of diabetes, cohort studies in Japan and Taiwan reported an elevated risk of type 2 diabetes among individuals with periodontitis, but evidence is lacking for other populations (Graziani et al. 2018). ⁴

The qualitative conclusions from models measuring the diagnosis at any time prior to the survey wave and those measuring a diagnosis in the past 2 y are very similar. The only fixed effect associations that are statistically significant across both sets of models are the correlation between visiting the dentist and the incidence of stroke and the correlation between edentulism and heart disease. In percentage terms, the magnitude of the former correlation is approximately twice as large when the dependent variable measures a stroke diagnosis in the past 2 y, rather than at any time in the past. Because dental care use is also measured over the past 2 y, the larger association with a stroke diagnosis in the past 2 y likely reflects a stronger correlation between contemporaneously measured variables. ⁵

One limitation of the fixed effects approach is the inability to account for unobserved characteristics of individuals that vary over time. We expect the primary unobserved factors that confound the relationship between dental care use and health to be low aversion to risk, myopia, and poor access to care, which are relatively constant over time and well-controlled using individual fixed effects. Edentulism is likely to be correlated with a broader array of unobservables, including unobserved comorbidities that are also correlated with indicators for specific chronic conditions. As a result, the potential for bias is greater in models that include the loss of all teeth than in those that include prior dental care use. ⁶

Another limitation is that we did not observe the specific dental services respondents received. Among U.S. Medicare beneficiaries, 73% of individuals with any dental care use had at least 1 cleaning, and among those who received preventive dental care, only 29% of visits were for services other than cleanings, exams, and X-rays (Moeller et al. 2010). However, the inclusion of individuals in our sample who received only dental services that were unrelated to systemic health attenuated our estimates. In addition, because of the length of our data set and modeling approach, our estimates capture only the short-term impact of dental care use on chronic illness.

Our analysis helps to establish initial estimates of the potential for greater use of dental care to reduce the incidence of chronic disease among elderly individuals. Further research is needed to establish causality in the relationship between dental care use and chronic illness, such as analyses based on natural experiments that result in changes in dental care use or oral health status that are plausibly unrelated to other determinants of chronic illness (Craig et al. 2017). Nonetheless, our finding that dental care use is associated with an improvement in overall health and reductions in the incidence of heart disease and stroke of between 2.7% and 11.6% helps inform the debate over whether to add a dental care benefit to the Medicare program. Previous research suggests that such a benefit would increase the use and lower the financial burden of dental care, and this analysis suggests there could be an additional health benefit through lower rates of chronic disease (Kreider et al. 2015; Manski et al. 2015, Aravamudhan et al. 2019; Meyerhoefer et al. 2019).

Better information about the association between dental care use, oral health problems, and chronic disease among the elderly could also inform clinical practice guidelines. However, the development of guidelines must be based on a solid scientific foundation, and although progress is being made on the mechanistic role of oral bacteria and systemic inflammation in the relationship between poor oral health and both cardiovascular and cerebrovascular diseases, further study is necessary to confirm the precise causal links (Patrakka et al. 2019; Zardawi et al. 2021). Such research could inform the development of targeted dental interventions as well as systemic treatments for dental and other diseases.

Author Contributions

C.D. Meyerhoefer, contributed to conception, design, data acquisition, analysis, and interpretation, drafted and critically revised the manuscript; J.V. Pepper, contributed to conception, design, and data interpretation, drafted and critically revised the manuscript; R.J. Manski, J.F. Moeller, contributed to conception, design, data acquisition and interpretation, and critically revised the manuscript. All authors gave final approval and agree to be accountable for all aspects of the work.