A cross-sectional study was conducted involving community-dwelling participants aged 65 years and over in Kaohsiung, Taiwan from May 2018 to January 2019. This study adopted a multistage, stratified cluster sampling method to select subjects from elderly population. There are 39 districts in Kaohsiung City in three geographic classifications (urban, rural, and mountainous areas). In the first stage of sampling, 25 districts were randomly selected from the 39 districts using probability proportional to size sampling method. In the second stage, community care centers in those 25 districts were randomly selected. In the final stage, older adults aged 65 years or older were recruited from each community care centers selected. There were total of 1180 participants recruited from urban areas (n=609), rural area (n=555), and mountainous areas (n=16).
Participants were excluded if they self-reported any of the following: (a) mental disorders, (b) expressive language disorders, (c) moderate or severe cognitive impairment, which was determined by a short portable mental status questionnaire (SPMSQ) (15), or (d) high-to-total dependence with respect to performance of Activities of Daily Living (ADL) (16). The final analysis included 1,100 participants (response rate: 93.2%).
Based on the sample size criteria for SEM analysis, a large sample size was required because of the highly complex model with relatively many free parameters that we intended to use. The ideal sample size is generally considered to be 20 participants per free parameter (17). In this study, 220 was considered the minimum sample size because there were 11 free parameters to be estimated.
Instrument
A structured questionnaire was developed to collect data including demographics (i.e., age, sex, education level), mental depression, oral function (i.e., xerostomia and dysphagia), physical function (i.e., frailty, sarcopenia), and OHRQoL. All instruments were translated from English to Chinese and back-translated to English by bilingual research staff and then verified for accuracy by two senior researchers. Items were reviewed by a panel of experts to assess content validity. The content validity index was 0.89-1.00. To ensure that the study participants understood the content, the questionnaires were pilot tested on 30 seniors. The reliability of each scale was assessed by examining internal consistency (Cronbach’s α coefficient). Color-changeable chewing gum was used to assess masticatory performance (Xylitol, 3.0 g; Lotte, Saitama, Japan).
Dental Examination
Dental examinations were performed by seven dentists according to World Health Organization criteria. The kappa coefficient of tooth decay was 0.77 for interrater agreement. The Kendall’s W coefficient for the plaque index was 0.87 for interrater agreement. The dental status and oral hygiene (i.e. plaque index and tongue coating) were recorded.
Oral Health-related Quality of Life
OHRQoL was measured using the Geriatric Oral Health Assessment Index (GOHAI), which is a 12-item instrument with three different domains: physical function (eating, speech, and swallowing), psychosocial function (worry or concern regarding oral health, dissatisfaction with appearance, self-consciousness regarding oral health, and avoidance of social interaction because of oral problems), and pain/discomfort (using medication to relieve pain or discomfort in the mouth). The English version of GOHAI was translated into Chinese for the participants (GOHAI-T)(18). Each item in GOHAI was rated on a 5-point Likert scale ranging from, 1 (always) to 5 (never). The total score could range from 12 to 60 points, with a higher score indicating a more favorable OHRQoL. The Cronbach’s alpha indicated an internal consistency of 0.75 for the scales.
Xerostomia
A condenced version of Xerostomia Inventory questionnaire was used to identify and classify mouth dryness. Participants respond to five items: (a) My mouth feels dry when eating a meal, (b) My mouth feels dry, (c) I have difficulty eating dry foods, (d) I have difficulty swallowing certain foods, and (e) My lips feel dry. Each item was scored as 1 (never), 2 (occasionally), or 3 (often) (19). The sum score could range from 5 to 15 points, with a higher score reflecting a drier mouth feel. Participants with sums ≥10 points represented the top 50% of xerostomia sum scores (20). The Cronbach’s alpha indicated an internal consistency of 0.88 for the scale.
Dysphagia
The dysphagia variable was measured using 15 question items in the Ohkuma questionnaire to quickly screen the community members; examples of questions include “Do you ever have difficulty when you swallow,” “Do you ever have difficulty with coughing up phlegm during or after a meal,” “Does it take you longer to eat a meal than before,” “Do you feel that it is becoming difficult to eat solid foods,” and “Do you ever have difficulty sleeping because of coughing during the night.” (21). Responses could be “obviously” (numerous times), “slightly” (sometimes), or “no” (never). The seniors who had at least one severe symptom were classified as having dysphagia. The Cronbach’s alpha was 0.85, indicating satisfactory internal consistency.
Masticatory Performance
Masticatory performance was evaluated using color-changeable chewing gum. As chewing progresses, the gum changes to red because the yellow and blue dyes seep into saliva, and red appears because of citric acid elution (22). Participants were asked to chew for 2 minutes. The observer assessed the color of the gum subsequently, using a color chart with five color gradations ranging from 1 (very good) to 5 (very poor). The masticatory performance scoring was simplified into three categories, 1–3 = good, 4 = moderate, 5 = poor.
Occlusal Condition
Occlusal condition was measured using the Eichner index (23), which is used to measure posterior occlusal contacts of existing functional teeth in the premolar and molar regions—these are called supported zones. The posterior regions are divided into four supporting zones on both sides and are classified as Eichner’s A, B, or C category. Class A contacts 4 support regions; class B contacts 1-3 regions or the anterior region only; and class C does not have contact with any support region, although a few teeth may remain.
Dental Status
Variables related to dental status were the numbers of remaining natural teeth, functional teeth, fixed artificial teeth, and complete and removable partial dentures. The number of functional teeth was defined as the number of natural teeth, excluding teeth with grade III mobility and residual roots and including fixed artificial teeth (i.e., abutment teeth, bridge, implant-supported prostheses).
Oral Hygiene Status
Plaque index recorded plaque scores for four surfaces (buccal, lingual, mesial, and distal) of tooth numbers 12, 16, 24, 32, 36, and 44; scores ranged from 0 to 3. Tongue coating was assessed using the Winkel tongue coating index (24). The tongue was divided into six areas (three posterior, three anterior), and coating was scored as 0 = no coating, 1 = light coating, and 2 = severe coating; scores could range from 0 to 12 points.
Covariates
Physiological and psychologic/psychiatric state may be potential affected by oral health status. Hence, we examined age, sex, educational level, depression level, frailty and sarcopenia as our covariates. Moreover, since the data collection was designed to be self-reported, participants with cognitive impairment (moderate or severe cognitive impairment) were excluded from the study. Level of depression was assessed using the Geriatric Depression Scale (GDS) (25). Frailty was assessed using the Study of Osteoporotic Fractures index (SOF) (26). Sarcopenia was assessed using the SARC-F self-administered questionnaire (27). Finally, all the covirates collected during the study were retained in the final model of pathway analysis.
Data Collection
The data were collected by well-trained interviewers during face-to-face interviews, following a standard protocol. To prevent information bias from occurring during interviewing, each interviewer was asked to attend a 1-hour training course regarding the standard process and criteria of data collection. The data collection process included three stops. First, a dentist performed the dental examination and a dental hygienist recorded the dental status, plaque index score, and tongue coating score. Second, the structured questionnaire was administered by an interviewer in Mandarin or Taiwanese. The entire interview process took approximately 30−45 minutes. Finally, the research staff collected data regarding masticatory performance by asking participants to chew color-changeable chewing gum and data regarding physical function and frailty.
Statistical Analysis
Stata 13.1 (StataCorp LP, College Station, Texas, USA) was used for statistical analysis. Participants were divided into two groups by xerostomia status. The demographic characteristics, dental status, oral function, and physical function were compared between two groups using chi-square tests and two-sample t tests. A multiple linear regression model was adjusted using the hierarchical method to obtain a predictive model of OHRQoL. To adjust the final model, changes in the adjusted R2 and F values were considered for each new independent variable added, and the variance inflation factor (VIF) was used to assess multicollinearity among all the independent variables. Path analysis was then used to test a model exploring the relationship between xerostomia, dental status, oral function, and OHRQoL and to identify both direct and indirect relationships in the model. Path analysis is a specific tool of SEM analysis that analyzes the assumed relations of the multivariate data simultaneously. SEM is one of the most favored statistical techniques in the social sciences and has become popular in dental science (28, 29). In the present study, the following criteria were used to define the goodness of model fit: χ2/df <3.00, Root Mean Square Error of Approximation (RMSEA) <0.08, and Comparative Fit Index (CFI) ≥0.90. To obtain an acceptable model, initial fit, modification (using MI method in STATA), and rafting steps were performed (30). After estimating the full model, nonsignificant direct paths were removed to generate a statistically parsimonious model, which was re-estimated and compared with the full model using a chi-square test. Finally, all the steps were repeated until the model was a good fit. Significance was set at p < 0.05 for all statistical tests.