Trends in socioeconomic inequality of periodontal health status among Dutch adults: a repeated cross-sectional analysis over two decades

The data used in this study were retrieved from a database developed by the Netherlands Organization for Applied Scientific Research (Toegepast Natuurwetenschappelijk Onderzoek, TNO). The Dutch National Oral Health Survey series has continuously gathered information on the oral health and oral health behavior of non-institutionalized civilian adult residents since 1995 [18]. The survey series was designed as a repeated cross-sectional study of the oral health status of the Dutch population over time based on a variety of oral health measures. More specifically, the Dutch National Oral Health Survey series used in this study were conducted among people living in the city of ‘s-Hertogenbosch, the Netherlands. This population is considered to be representative of the general Dutch population in terms of demographic distribution [19]. Under the National Health Care Institute's authority, Health Insurance Fund (HIF) companies were asked to provide the names and addresses of their clients. Since the HIF system focused on the relatively lower-income population before 2006 [20], people with low SES could be overrepresented. Therefore, the analyses were performed in the whole population and in the sub-populations stratified by SES. The stratification method facilitates making comparisons and conducting trend analyses about oral hygiene and periodontal status over time, in line with the approach adopted in a previous study on caries experience [13]. Central Committee on Research Involving Human Subjects concluded that no ethical approval was required for data collection for this purpose. Furthermore, it was determined that all of the requirements of the Personal Data Protection Act had been met (No. m1501261).

Data from four TNO survey cycles conducted between 1995 and 2013 were used. In 1995 and 2002, the sample included only persons aged 25–54 years [21, 22]. In 2007 and 2013, the scope was widened to 74 years [23, 24]. To conduct trend analyses in equal groups, we only included data of participants aged 25–54 years in this study. Information on sociodemographic characteristics and clinical examinations were from four surveys (1995, 2002, 2007, and 2013) while excluding those who were lacking dental measurement or SES data. In total, 3083 respondents were included in the present study (see Fig. 1).

Educational attainment was used as a proxy for socioeconomic status [13]. The level of education was divided into “low” (only having completed primary school or having received a lower-level general secondary education) and “high” (having received a higher-level general secondary education or tertiary school) [25, 26]. Demographic variables, including age (25–34, 35–44, 45–54 years), gender and ethnic background, were collected via questionnaire and interview. Ethnicity (native/non-native Dutch) was determined on the participant’s birth country (the Netherlands versus any other country). Oral hygiene behaviors included the time since the last dental visit, frequency of daily teeth brushing, using dental floss, using tooth pick, and mouthwash. We considered these covariates as potential confounders that were controlled in the multivariable regression models.

Oral health examinations were performed by a trained and qualified team of dentists and dental hygienists in a mobile research facility located in the municipalities in which the participants lived. The comprehensive oral health assessment included dentition status assessment, oral hygiene examination, and periodontal screening. Teeth missing (edentulous areas) or replaced by dental implants were recorded as lost teeth. To assess oral hygiene, the simplified oral hygiene index (OHI-S) was used to quantify dental plaque accumulation [27]. OHI-S was scored based on six index teeth: the buccal surfaces of teeth 16, 11, and 26 and the lingual surfaces of teeth 31, 36, and 46. Details concerning the criteria used in the OHI-S scoring can be found in Table 1. Oral hygiene was defined as “good” if all index teeth had OHI-S = 0, “fair” if at least one tooth with OHI-S = 1, and “poor” if at least one tooth with OHI-S = 2–3. In the current study, good oral hygiene (i.e., plaque-free [OHI-S = 0]) was considered as a binary outcome variable.

Periodontal status was assessed by probing pocket depth index (PDI), which is a standard periodontal parameter [18]. Periodontal examination was conducted on each tooth in two randomly selected quadrants (depending on the research ID number: ending with an odd number: we examined the 2nd and 4th quadrants; In case of an even number, we checked 1st and 3rd quadrants). This approach could minimize the burden of examination in large epidemiological studies. As third molars were excluded, a maximum of 14 teeth per individual could be examined. Three sites (buccal sites of maxillary vs. lingual sites of mandibular) were examined for each tooth using a Community Periodontal Index (CPI) probe recommended by the World Health Organization [28]. The highest score was recorded per tooth. The PDI scores were recorded as follows: PDI = 0, pocket depth < 3.5 mm; PDI = 1, pocket depth of 3.5–5.5 mm; and PDI = 2, pocket depth > 5.5 mm (Table 1). We defined those individuals with PDI = 0 (all teeth have probing pocket depth < 3.5 mm) as constituting the periodontal health group [29, 30]. Moreover, we defined PDI = 2 (≥ 1 tooth with probing pocket depth > 5.5 mm) as deep pockets group. These approaches allowed for binary outcome variables [29, 30].

We calculated the means or the percentages of chosen population characteristics by survey year. The differences between survey years were analyzed by Chi-square test and one-way ANOVA. We calculated age-standardized percentages of oral hygiene (OHI-S = 0, 1, 2/3) and periodontal status (PDI = 0, 1, 2) using the direct method [31] with the data from Dutch population of 2006 as the standardized population. The age-standardized percentages with 95% confidence intervals (CIs) were presented for each study period.

We studied trends in socioeconomic inequality of oral hygiene and periodontal status in two steps. First, socioeconomic inequality was tested on the absolute and relative scales in each survey year (1995, 2002, 2007, 2013) [9]. We used multivariable logistic regressions to estimate the association of SES with oral health outcomes (i.e., plaque-free, periodontal health, and deep pockets). The logistic regressions were adjusted for potential confounders, including age, gender, ethnicity, SES, time since the last dental visit, number of teeth present, and oral hygiene behaviors (toothbrushing frequency, floss usage, and toothpick usage). Rate difference and odds ratio derived from the corresponding logistic model were used as the measure of absolute and relative effect [9, 32]. Rate difference was defined as the difference between the predicted probability of the oral health outcome in high-SES group and that in low-SES group. The predicted probabilities for each outcome were calculated using marginal-effect estimation [33]. 95% CIs for the rate difference were estimated using the delta method [34]. As a second step, the adjusted percentages (i.e., predicted probabilities) of oral health outcomes were presented graphically to evaluate temporal trends using pooling data from four waves. We performed interaction analyses for each oral health outcomes in which the survey year (as a continuous variable) and the interaction between survey year and SES were additionally added into the logistic model in step one. A statistically significant interaction term (survey year × SES) indicated an increase or disease in socioeconomic inequalities of oral health outcomes over time [35]. Complete case analysis was used for handling missing data. R Project for Statistical Computing (version 3.6.0) was used to perform all analyses.

Participant characteristics aged 25–54 years by survey year are presented (Table 2). There were differences in the age and age composition of each survey year. The percentage of people with low-SES in 1995 and 2002 were higher than in 2007 and 2013. Compared to 1995 and 2007, the percentages of females in 2002 and 2013 were higher. Interdental hygiene behaviors (i.e., daily usage of floss and toothpick) improved over time. There were significant differences between survey years for these characteristics (P < 0.01).

The age-standardized percentages of the OHI-S and the PDI in the total population can be found in Table 3. Between 1995 and 2013, there was a decrease in poor oral hygiene (OHI-S ≥ 2: from 41.1% [95% CI 37.9–44.4] in 1995 to 10.9% [8.6–13.2] in 2013). Following this finding, the number of individuals with OHI-S = 0 showed an increasing trend over time. In contrast, no evident changes could be identified in the age-standardized percentages of individuals with deep pockets (PDI = 2) in the total population during the last two decades. Most of the changes in individuals with periodontal pockets (PDI = 1) occurred (from 42.0% [38.8–45.3] in 1995 to 34.0% [30.5–37.5] in 2013). This decreasing trend was accompanied by a steady increase (51.4% [48.1–54.7] to 60.6% [57.0–64.1]) in the proportion of individuals who exhibited periodontal health (PDI = 0, Table 3).

Table 4 showed the relative and absolute socioeconomic inequality of plaque-free, periodontal health, and deep pockets in each survey year. For plaque-free, the significant absolute inequalities were observed in 2007 (absolute index = 8.4; 95% CI = 2.1, 14.7) and 2013 (absolute index = 8.6; 95% CI = 1.5, 15.7) while not in 1995 and 2002. When going from low-SES to high-SES, the adjusted percentage of plaque-free increased by 8.4 percentage points in 2007 and 8.6 percentage points in 2013, respectively. In the relative scale, the regression-based relative effect index presented a pattern similar to absolute inequality (1.698 [1.124–2.565] in 2007 and 1.561 [1.080–2.257] in 2013).

For the proportion of individuals having periodontal health, the significant absolute inequalities were seen in 2002 (10.0 [1.6–18.4]) and 2013 (9.6 [1.8–17.4]) (Table 4). We observed a similar pattern on the relative scale: moving from low-SES to high-SES was associated with a 1.5-fold increase in periodontal health's probability in 2002 and 2013. On the other hand, 1995 and 2007 did not show any significant inequality gap on absolute and relative inequality scales.

Compared with low-SES individuals, the prevalence of deep pockets in high-SES individuals was 4.2 percentage points lower in 1995 (Table 4). The relative scale finding was similar to the absolute inequality. The regression-based relative effect index decreased by almost 50% (0.216–0.869) in 1995, indicating that the high-SES individuals had half risk of deep pockets than those in the low-SES group. In contrast, we did not observe any significant absolute and relative inequality of deep pockets in 2002, 2007, and 2013.

Figure 2a illustrates the trends over time in the crude and adjusted percentages of individuals without detected plaque (OHI-S = 0) by SES-groups. In the 1995–2013 period, there was a significant increase in the adjusted percentage of individuals with OHI-S = 0 in both the low- and high-SES groups (P_trend < 0.001, Table 5), after adjusting for potential confounders. The slope of the high-SES group's temporal trend was slightly steeper than the corresponding slope for the low-SES group (Fig. 2a). However, the interaction effect of survey year × SES on plaque-free did not reach a significant level (P_interaction = 0.198, Table 5).

The secular trends in the crude and adjusted percentage of periodontal health (PDI = 0) are shown in Fig. 2b. After adjusting for confounders, the upward trends with respect to periodontal health were identified, with a percentage increase in the high-SES group (P_trend < 0.001, Table 5). In contrast, the percentage of periodontal health among the low-SES participants remained relatively constant (P_trend = 0.083, Table 5). Although SES inequality in periodontal health tends to widen increasingly over the last two decades (Fig. 2b), the SES group-by-time interaction did not reach statistical significance (P_interaction = 0.490, Table 5).

There was little difference in the adjusted percentage of individuals with deep pockets (PDI = 2) during all survey years, either in the high-SES or low-SES group (Fig. 2c). In the multivariable logistic regression models, no significant temporal trends were found for deep pockets between 1995 and 2013 (low-SES: P_trend = 0.893; high-SES group: P_trend = 0.787, Table 5). Additionally, the interaction term survey year × SES was not statistically significant (P_interaction = 0.678, Table 5).