Study population
The study was approved by the Institutional Review Board of the School of Dentistry at Seoul National University, Seoul, Republic of Korea (IRB No. S-D20200018). All methods were performed in accordance with the relevant guidelines and regulations by including a statement in the declaration section under the ethical approval and consent for participation section.
Dental appointments were performed from October 7, 2019 to February 29, 2020, at two detention centers in South Korea, including the Seoul Detention Center and the Seoul Eastern Detention Center (Additional file 1: Appendix 1). Dental records were compiled by the authors in March 2020. A retrospective analysis of the dental records of 642 inmates was performed. Informed consent was waived by the intstitutional review board because of the retrospective study design. Separate serial numbers were coded and assigned to the inmates without collecting identifiable information. Only the researchers had access to the collected data, and the data were kept in encrypted files protected by the research manager for three years after the study’s completion.
For comparison, raw data related to oral examinations were similarly obtained from the Korea National Health and Nutrition Examination Survey (KNHANES) VII (2016–2018) conducted by the Korea Disease Control and Prevention Agency (KDCA) [19]. The target population of the survey was all non-institutionalized civilian South Korean individuals one year of age or older. The survey employed stratified multistage probability sampling units based on the geographical area, sex, and age, which were determined based on the household registries of the National Census Registry—the most recent 5 years of the national census in South Korea [20]. Using the census data, 192 primary sampling units (PSU) were selected annually across South Korea [21]. Raw data for 16,489 people surveyed between 2016 and 2018 were released most recently in the KNHANES VII.
According to the South Korean Criminal Act, individuals below age 14 cannot be admitted into correctional institutions. In addition, the growth of permanent teeth (except for wisdom teeth) is usually complete after age 14. Therefore, KNHANES data from 2523 people below age 14 were excluded. In addition, the survey excluded data from 621 people aged 80 or older, as top-coding was used to de-identify the elderly population. Therefore, 13,345 individuals were included in the KNHANES group.
Data collection
For detention center inmates, clinical examination was performed using a dental chart during dental appointments conducted by a public health dentist (I. Hwang; Additional file 1: Appendix 1). Examinations were performed in a dental clinic in the correctional institutions equipped with basic dental equipment such as; a dental chair, artificial light, and intra-oral mirror. Additionally, a brief self-reported questionnaire was applied together with the clinical examination to quickly and comprehensively assess a patient’s oral health condition during the appointment. The questionnaire was not previously validated and was created by the authors. The questionnaire items were the same as those used in the oral examination conducted by the National Health Insurance Service which is used nationwide in health checkups based on Article 52 of the National Health Insurance Act. However, questions on diet, smoking, and the use of fluoride toothpaste were excluded because all participants were expected to provide the same response, as these aspects were controlled at the correctional institution. Unfortunately, there are no such clinical examination data matching this questionnaire for the general population.
At the time of writing, there are no existing protocols for dental triage in this setting; therefore, inmates in this study were treated sequentially based on their order of application. However, an inmate’s appointment was prioritized if an emergency was reported through an employee or a referral was made by a medical doctor.
The determination of epidemiological indices of dental caries such as the Decayed, Missing, and Filled Teeth (DMFT) index, the decayed teeth (DT) index, the missing teeth (MT) index, and the filled teeth (FT) index was based on the criteria of the KNHANES protocol, which are modeled after on the World Health Organization criteria [22]. While counting the DT, cavitated lesions and filled crown with caries were included. According to these guidelines, the assessments were made only through history taking and visual inspection, without the use of radiographs. The prevalence of untreated caries was determined by calculating the proportion of participants with any DT. An example of the dental records form used during the dental appointments is provided in the Additional file 1: Appendix 1.
Since 2007, the oral health of the general population of South Korea has been assessed using KNHANES (KNHANES has been implemented since 1998). Specifically, the KDCA conducts various oral examinations, including three self-reported oral health questions, oral/prosthodontic status tests (e.g., dental caries), treatment requirement tests, periodontal tests, and dental fluorosis tests. Since the DMFS index was used during the KNHANES collection phase, this information was converted to the DMFT index for comparison with inmate data.
Statistical analysis
A contingency table was constructed to summarize the age and sex distribution by population group, and a chi-square test was performed to determine the difference in age and sex distribution between the two groups.
To examine the association between independent variables and the prevalence of untreated caries, a binary logistic regression analysis on a generalized linear model was used.
Assuming that the epidemiological indices of dental caries such as DMFT, DT, MT, and FT have a natural number of data, we can apply Poisson's regression analysis or negative binomial regression analysis on a generalized linear model. Based on the use of the Akaike information criterion, negative binomial regression analysis was considered more suitable. This regression analysis was similarly used in a previous study on dental caries [23].
The self-reported questionnaire in the inmate group was analyzed. To provide a comprehensive evaluation of oral health, the self-reported questionnaire included questions related to general health, perceived oral health, and oral health habits. For general health, a history of diabetes and cardiovascular diseases was considered. For perceived oral health, the following 11 questions were asked: “How long has it been since you last visited the dentist?” “Have you been uncomfortable chewing in the last three months?” “Have you felt pain in the last three months? (history of dental pain),” “Have your gums bled during the last three months? (bleeding of gums),” “What is your perception regarding your mouth condition when you evaluate yourself? (self-rated oral health).” Regarding oral health habits, the following questions were asked: “Have you ever learned how to brush your teeth?” “How many times do you brush your teeth in a day?” “How many times have you brushed your teeth right before going to bed in the last week?” and “How often do you use dental floss?” All variables from the questionnaire were included in the multivariable models based on their association with dental caries, as reported in the literature. Variable selection was not applied from the analysis even if it was insignificant [24]. A generalized linear model was used on all variables including sex, age, and the 11 questions as covariates, and a binary logistic regression analysis for the prevalence of untreated caries and negative binomial regression analysis for DMFT, DT, MT, and FT were performed.
Similarly, the individuals in the KNHANES group also received self-reported questionnaire. This questionnaire comprised two questions: “Have you experienced a toothache in the past year?” and “What is your perception of your mouth condition when evaluating yourself?” Moreover, a generalized linear model was used on all variables including sex, age, and the two questions as covariates, and a binary logistic regression analysis for the prevalence of untreated caries and negative binomial regression analysis for DMFT, DT, MT, and FT were performed.
A comparison of the prevalence of untreated caries and epidemiological indices of dental caries between the two groups was conducted. Among the data of the two groups, a history of dental pain during the past three months was assessed in the inmate group, and during the past one year in the general population group. The duration of interest differed between the two groups; however, we considered the responses comparable since they both addressed whether the patients or respondents had experienced general pain before visiting the clinic. Ultimately, we found four common variables: sex, age, history of dental pain, and self-rated oral health. Variable selection was not applied as the same above, and analysis was performed including all common variables. Thus, to compare the prevalence of untreated caries and epidemiological indices of dental caries between the two groups, a generalized linear model was used by adjusting for sex, age, history of dental pain, and self-rated oral health as covariates, which employed a binary logistic regression analysis for the prevalence of untreated caries and negative binomial regression analysis for DMFT, DT, MT, and FT.
In addition, to separately investigate the interaction effect of sex, age, history of dental pain, or self-rated oral health on the prevalence of untreated caries and epidemiological indices of dental caries by the group, we checked for the interaction effects between “group by sex,” “group by age,” “group by history of dental pain,” and “group by self-rated oral health,” with adjustments for the effects of sex, age, group, history of dental pain, and self-rated oral health. A generalized linear model was used, which employed a binary logistic regression analysis for the prevalence of untreated caries and negative binomial regression analysis for DMFT, DT, MT, and FT were performed.
Statistical analyses were performed using R software version 4.0.2 (R Foundation for Statistical Computing, Vienna, Austria). Statistical significance was set at a p value of < 0.05.