Acute and Chronic Diabetes Complications Associated With Self-reported Oral Health

A Retrospective Cohort Study

Kamini Kaura Parbhakar; Laura C. Rosella; Sonica Singhal; Carlos R. Quiñonez


BMC Oral Health. 2020;20(66) 

In This Article


Study Design

A retrospective cohort study was designed to explore the odds of acute and chronic diabetes complications among Ontario residents self-reporting oral health status. Study participants were selected from a combined pool of respondents interviewed from the 2003 and 2007–08 Canadian Community Health Survey (CCHS).[27] Briefly, the CCHS is a national survey that utilizes a multi-stage, stratified, clustered-probability survey sampling design and is representative of 98% of the Canadian population.[28] The survey is administered by Statistics Canada and collects self-reported health information from Canadians over the age of 12, excluding those living on indigenous reserves, residing in institutions and full-time members of the Canadian armed forces.[28] Details regarding the CCHS methodology are documented elsewhere.[28]

The study cohort was restricted to Ontario residents over the age of 40 with a validated diabetes diagnosis. As all residents of the province of Ontario are covered by a single payer insurance system, the Ontario Health Insurance Plan (OHIP), health system encounters among these participants can be followed. With de-identified health card numbers, health encounters in electronic medical records held by the Institute for Clinical Evaluative Sciences (ICES) were followed from the initial CCHS interview date, until March 31, 2016. A validated diabetes diagnosis was confirmed using the Ontario Diabetes Database (ODD), an ICES-derived disease registry containing physician diagnosed cases of diabetes in Ontario.[29]

The final study sample was composed of individuals over the age of 40, who participated in the oral health component of the CCHS and had an ODD confirmed diagnosis of diabetes. Individuals were excluded if they could not be linked to electronic medical records, were OHIP-ineligible during the follow-up period, or if they did not report oral health status during survey administration. The final analytical sample consisted of 5183 participants, which represents a weighted sample of 1.31 million Ontario residents. Ethical approval for this study was obtained from the University of Toronto Research Ethics Board and was followed by ICES approval for data creation and access at Sunnybrook Hospital in Toronto, Ontario (protocol reference no. 34553).

Oral Health Status

CCHS cycles 2003 and 2007–08 were selected for the availability of oral health data for the province of Ontario.[28] Self-reported oral health status was assessed through the question "would you say the health of your teeth and mouth is: excellent, very good, good, fair or poor?". According to the distribution of self-reported oral health status among study participants and to determine the differences among positively and negatively inclined oral health responses, the exposure variable was re-categorized into two values representing "good to excellent" and "poor to fair" oral health categories. Further description of oral health content from the CCHS is described elsewhere.[28]

Diabetes Complications

The primary outcome of this study was the first diabetes complication experienced by participants after the CCHS interview date. Complications were extracted from hospitalization and emergency department records. International Classification of Disease (ICD-9) codes, were used to extract diabetes-specific complications from these data.[30] Acute complications included non-specific hypoglycemia and hyperglycemia; chronic complications included myocardial infarction, stroke, skin infections, amputation, dialysis, and retinopathy. Following the CCHS interview, participants were classified in three categories comprising of those who did not experience any complication, those who experienced an acute complication and those who experienced a chronic complication.


CCHS participants' demographic characteristics, health behaviours and medical histories, were included in the study as covariates, according to their association with oral health and diabetes outcomes. Covariates included the following: age, sex, income, education, rural/urban index, race, physical activity, smoking status, alcohol consumption, dental visits, basal metabolic index (BMI), prior co-morbidity, duration of diabetes prior to interview date, stress, community sense of belonging, health care visits prior to the complication and self-reported overall health.

Age, BMI and duration of diabetes were continuous measures while all other covariates were categorical. The duration of diabetes prior to interview date was used to consider the impact of a dated or early diabetes diagnosis. The Rurality Index of Ontario (RIO) was used to classify individuals residing in rural or urban areas. Participants with RIO scores < 39 were categorized as urban residents and scores ≥40 were categorized as rural residents.[31] Co-morbidity at the interview date was assessed from CCHS questions regarding chronic diseases and participants were classified into two categories comprising of participants without any co-morbidity and participants with any of the following conditions: arthritis, chronic obstructive pulmonary disease (COPD), heart disease and stroke. Health care visits was assessed by extracting OHIP codes for visits to a general practitioner or specialist. Visits were classified into three categories, comprising of participants who visited their general practitioner, specialist or both general and specialist practitioners for diabetes management.

Statistical Analysis

Starting with baseline characteristics, all variables were assessed for their association with self-reported oral health. T-tests were used to express the means and standard deviations of continuous variables and Chi-squared tests were used to express the cross tabulated frequencies of categorical variables. Bivariate analysis was conducted by the multi-categorical outcome. Furthermore, variables with a p-value ≤0.25 following the baseline and bivariate analysis were included and the most parsimonious model was built.[32] Variables that did not reach a p-value of ≤0.25 but were associated with the exposure or outcome in the literature were also included in the model. Multinomial logistic regression models were used to determine the odds of experiencing an acute or chronic outcome following the interview date among participants reporting "poor to fair" oral health. Multinomial odds ratios with confidence intervals were estimated for diabetes complications associated with oral health, where individuals reporting "good to excellent" oral health status, whom did not experience any diabetes complications following the interview date, represented the reference group. Oral health was the explanatory variable and covariates were included in the model if they were clinically significant or were associated with diabetes outcomes in existing literature.

Bootstrapping sample weights provided by Statistics Canada were applied to the analysis to adjust for the complex nature of the CCHS survey design. This generated inferable estimates for the Ontario population. All statistical analyses were performed using SAS version 9.4 and data was accessed at Sunnybrook Hospital ICES Central in Toronto, Canada. PROC SURVEY MEANS, PROC SURVEYFREQ and PROC SURVEYLOGISTIC were used in the analysis.[33]