The validity and reliability of the Indonesian version of the Summated Xerostomia Inventory

Abstract Objective To validate and determine the reliability of the Indonesian version of the Summated Xerostomia Inventory (SXI‐ID) questionnaire. Background Xerostomia is a common problem in older people, and the SXI is increasingly being used to measure it. Materials and methods The SXI questionnaire was cross‐culturally adapted to create an Indonesian version (SXI‐ID), which was assessed for validity and reliability in a cross‐sectional study of older adults living in nursing homes in Jakarta, Indonesia. Each participant signed an informed consent and was interviewed with the SXI‐ID. A subset of participants was interviewed again after two weeks. A standard question was used to check criterion‐related validity, by plotting the mean SXI‐ID scale scores against the ordinal response categories of the standard question. The reliability check included Cronbach's alpha, total and inter‐item correlation, and intraclass correlation for internal consistency, along with test‐retest reliability. Results A total of 110 older adults participated. Cronbach's alpha value for the SXI‐ID was .85, and the intraclass correlation coefficient value for test and retest in 15 participants was .9. The SXI‐ID total score showed a strong positive correlation (.87) with the global question. There was a consistent ascending gradient in mean SXI‐ID scores across the ordinal response categories of the global item. Conclusion The SXI‐ID is psychometrically valid and reliable for measuring xerostomia in the Indonesian population.

a single-item question ("global item") or with multiple item approaches such as batteries of items or summated rating scales. 1 In 1999, Thomson et al 10 described the Xerostomia Inventory (XI), which comprised 11 questions that could detect and measure the severity of xerostomia. Later, the XI was modified to a short-form version, termed the Summated XI (SXI). 11 The SXI uses fewer items and a simpler response format which is more suitable for frail older people. Items in the XI which were not directly related to dry mouth (such as those on dryness of the eyes, nose and facial skin), and the behavioural consequences of dry mouth were not included in the SXI questionnaire. 11 The SXI questionnaire has been used to assess xerostomia in several different countries, and it has been adapted and validated in Dutch, English, Chinese, Portuguese and Japanese. 12 Indonesia is the fourth most populous country, with ever-increasing numbers of older people. 13 To date, there has been no validation of the Indonesian version of this questionnaire. Accordingly, this study aimed to validate an Indonesian version of the SXI questionnaire to measure xerostomia in the older Indonesian population.

| Study design
This was a cross-sectional study which aimed to assess the validity and reliability of the Indonesian version of the Summated Xerostomia Inventory (SXI-ID) questionnaire among residents of government nursing homes in Jakarta, Indonesia. DKI Jakarta Province has 5 nursing homes in different areas across the province.

| Data collection
This study was conducted in all people aged ≥60 years who had been living in a government nursing home in Jakarta for at least 1 month, who, understood Indonesian, were able to communicate well without cognitive impairment, had a Mini-mental State Examination (MMSE) value of >24, and who provided written informed consent. Excluded were people with apraxia, terminal illness, fever or Sjögren's syndrome, and those who had undergone (or were currently undergoing) head and neck radiotherapy. This study was reviewed and approved by the Faculty of Dentistry Universitas Indonesia Research Ethics Committee (No. 82/Ethical Approval/FKGUI/X/2018).
After the study was explained to the participants, they were interviewed by a researcher who administered the MMSE questionnaire to assess the cognitive function of the participant.
Another researcher then conducted the SXI-ID interview. The interview method was adjusted considering the limitations of older people (such as low educational background and weak eyesight).
Accordingly, the declarative sentences in the questionnaire were amended to become interrogative sentences. Each participant was asked to answer the SXI-ID and was told that the answer should be the one that immediately came to their mind. The participants for the test-retest analysis were selected from one nursing home that we visited for the first time. The participants in the retest were randomly chosen from those who had been previously assessed.
Before testing for validity and reliability, the SXI-ID was cross-culturally adapted. The questionnaire was translated into Indonesian using the guidelines for cross-cultural adaptation. 14,15 The adaptation process was undertaken according to a previously published method consisting of the 6 stages of translation, synthesis, back-translation, expert review, testing the

| Statistical analysis
Data were analysed using the Statistical Package for Social Sciences (SPSS) program, version 22. The SXI-ID score was computed by summing the item response scores. We further adapted the 5-response format to the 3-response format, as follows: 1 = 1; 2 through 3 = 2; and 4 through 5 = 3 for statistical analysis. In the 3-point Likert scale, 1 = never, 2 = occasionally and 3 = often. The level of statistical significance was set at P = .05.
An item-total correlation procedure was conducted to determine the correlation of each SXI-ID item with the total score; each correlation value should be above 0.3. Reliability was evaluated using Cronbach's alpha. An alpha above 0.7 indicates acceptable reliability, while an alpha above 0.8 indicates good reliability. 16 The procedures were repeated for 15 participants after 2 weeks in order to evaluate the test-retest reliability of the SXI-ID. The reliability was assessed using the intraclass correlation coefficient (ICC), computed using a one-way random-effects model. 17

| Participants
Sociodemographic information on the 110 participants is sum-

| Responses to the SXI-ID
The responses of the study participants to each SXI-ID item using are presented in Table 3. There were no missing SXI-ID data. The mean total SXI-ID score was 6.6 (SD, 2.3).

| Validity and reliability of SXI-ID
The correlations among SXI-ID items are shown in Table 4. The correlation between the SXI-ID summary score and the standard question was 0.87 (95% CI 0.82 to 0.91). Cronbach's alpha was .85, indicating very good reliability. The inter-item correlation values ranged from .40 to 1.00.
To assess the reliability of the SXI-ID questionnaire, it was repeated with 15 participants after 2 weeks. The mean total score of the SXI-ID for test and retest was both 7.0 (SD, 3.2). The ICC for test-retest reliability was 0.9. The intraclass correlation coefficient (ICC) and item-total correlation (ITC) values are shown in Table 5.
These were found to be acceptable.
In order to examine the criterion-related validity, the mean SXI-ID scale scores were compared against responses to the standard question (Table 6). There was a gradient in mean SXI-ID score observed across the categories of the standard question, with the highest mean SXI-ID score seen in those responding "Always," and the lowest seen in the "Never" group. The gradient in mean scores across the categories of the standard questions was statistically significant. This was an initial validation study, conducted with older residents of a government nursing home in Jakarta. A limitation of this study is that the findings may not be generalisable. Future use of the questionnaire in more diverse older Indonesian samples would be useful and informative. We also did not conduct a formal sample size calculation prior to conducting the study, but we did undertake a post hoc power calculation (using G*Power version 3.1.9.4) based on the collected data. This showed that our N was more than sufficient, with, for example, 10 people required to demonstrate even the difference in mean SXI score between those responding "Never" and those responding "Occasionally" to the standard question (given the observed effect size of 1.0 and the power of 0.95 to demonstrate a difference).

| D ISCUSS I ON
Despite the limited number and diversity of participants, a strength of this study was the cognitive function screening of the participants using the MMSE. This ensured that participants were cognitively able to respond to the SXI-ID, and so there were no missing data. Only older adults with an MMSE score >24 were included.
We further formatted the responses into a 3-point Likert scale format for consistency with previous studies. 19 The construct and discriminant validity of the SXI-ID questionnaire was tested through associations and comparisons between the scores of the SXI-ID and the standard question. The mean Cronbach's alpha value of .90 suggests that the questions in the SXI-ID are measuring same construct and have excellent internal consistency. The data indicate that the SXI-ID questionnaire is valid for use as a measuring instrument in Indonesia. The mode of questionnaire administration may have influenced the validity of the findings. In this study, the data collection from participants used a standard interview approach with the SXI-ID items as interrogative statements without adding any information to the interview. There was only one interviewer, to eliminate inter-interviewer bias. In this study, interviewing was chosen as the method of administration over the self-administered method in order to maximise participation and minimise cognitive burden bias. 20 The original SXI validation 11 used six different geriatric population samples; half of those studies used interviews, and half used written questionnaires. Comparison of the psychometric and criterion validation data shows no systematic differences (tables II and III, and figure 1 in that paper), suggesting that our choice of administration method in the current study is unlikely to have unduly affected the findings.
The correlations observed in this study were well above the recommended threshold and were similar to those obtained using Chinese and Portuguese versions of the SXI. 18,19,21 The current study also showed the expected gradient in mean SXI-ID scores across the response categories of the standard question, suggesting that the SXI-ID meets the criteria for independent validation. Exploration of its association with objective measures of salivary flow rate is needed, although of course the SXI-ID is not measuring salivary flow. Further investigation of the SXI-ID in studies using sialometry would be informative.

| CON CLUS ION
The SXI-ID has excellent psychometric properties and is a valid and reliable tool for measuring xerostomia in the Indonesian population, much like its parent English version.

ACK N OWLED G EM ENT
The authors would like to acknowledge the support given by the Government of Special Region of Jakarta, Indonesia, and Dr Patoni for his support during the coordination of data collection. This study was supported by Universitas Indonesia 2019.

AUTH O R CO NTR I B UTI O N S
YSW planned and assisted the survey, supervised the study, drafting and finalising the manuscript. FR dealt with local bureaucracy, obtained study permits, acted as a liaison to obtain ethical clearance, and supervised the study and drafting of the manuscript. DAM contributed to statistical analysis and data interpretation. WM contributed to data collection, data analysis and initial data report. WMT revised the manuscript critically for important intellectual content and contributed to the drafting and finalising of the manuscript. All authors have approved the final version and agreed to be accountable for all aspects of the work.