Cross-cultural adaptation and psychometric validation of the physical oral health index in Chinese older inpatients

Abstract

Purpose

Oral health is a key component of healthy aging. According to the dual-axis conceptual model, oral health consists of a physical axis and a psychosocial axis. While the psychosocial axis is well covered by validated patient-reported instruments with established Chinese versions, no comprehensive, theoretically grounded instrument for assessing the physical dimension of oral health is currently available in China. The Physical Oral Health Index (PHOX), grounded in a dual-axis theoretical model, is a comprehensive instrument specifically designed to assess the physical dimension of oral health, including teeth, soft tissues, jaw function, and pain perception. Therefore, this study aimed to develop a Chinese version of the PHOX and evaluate its psychometric properties among older inpatients in China.

Patients and methods

The PHOX was cross-culturally adapted using Cruchinho’s eight-step methodology. A total of 1,365 older inpatients (aged ≥ 65 years) from five departments of a tertiary hospital in Zhejiang Province, China, were enrolled. The psychometric properties of the Chinese version were evaluated in terms of content validity, construct validity, internal consistency, and test–retest reliability.

Results

The Chinese version retained the original 14-item, five-factor structure. Following revision, both the item-level content validity index and the scale-level content validity index reached 1.00. The scale demonstrated excellent internal consistency (Cronbach’s α = 0.905; CR = 0.898–0.930) and test–retest reliability (ICC_total = 0.923; ICC_dimension = 0.858–0.919). Exploratory factor analysis supported a five-factor structure, explaining 79.2% of the total variance, with factor loadings ranging from 0.71 to 0.95. Confirmatory factor analysis conducted using the calibration, validation, and total samples demonstrated a robust model fit (RMSEA = 0.065, CFI = 0.975, and SRMR = 0.027). Convergent properties (AVE = 0.742–0.834) and discriminant validities were confirmed. Weight equivalence analysis further supported the applicability of the original scoring weights in the Chinese population.

Conclusion

The PHOX-Chinese demonstrated strong psychometric properties and good cultural adaptability, supporting its use as a comprehensive tool for assessing physical oral health among older inpatients. Its application in clinical practice and research may facilitate the early identification of oral health impairments, inform individualized oral care planning, and support targeted nursing interventions that may contribute to improved health outcomes in aging populations.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12903-026-09578-8.

Introduction

Global population aging poses significant challenges for healthcare systems and professionals. Population aging, a defining demographic shift of the twenty-first century, continues to accelerate worldwide, with the proportion of individuals aged 60 years and older projected to nearly double from 12% to 22% by 2050 [1]. China, already at an advanced stage of this demographic transition, was home to 310.31 million people aged ≥ 60 years and 220.23 million aged ≥ 65 years at the end of 2024, representing 22.0% and 15.6% of its total population, respectively [2]. Despite being a critical indicator of health among older adults, oral health remains a neglected dimension of healthy aging [3]. Recent policy initiatives in China, including the Healthy Oral Health Action Plan (2019–2025) and China’s 15th Five-Year Plan (2026–2030), have increasingly prioritized oral health promotion among older adults [4, 5]. The World Health Organization defines oral health as a multidimensional state of well-being in which the mouth and related structures function effectively to support essential physiological needs while also contributing to psychological well-being, self-confidence, and full social participation [6]. Aging directly and indirectly affects both hard and soft oral tissues, increasing the risk of oral health problems among older adults.

Oral and systemic health are bidirectionally related [7]. Chronic conditions such as diabetes, cardiovascular disease, and hypertension not only increase the risk of periodontal disease and tooth loss but may also be exacerbated by severe oral infections through heightened systemic inflammation [8, 9]. Furthermore, oral diseases and many systemic conditions share common risk factors, including unhealthy diet, smoking, and alcohol consumption [10, 11]. Older adults often neglect their oral health, resulting in poor oral conditions that further disrupt this relationship. Oral diseases frequently manifest as pain and discomfort, leading to functional limitations such as impaired mastication and dysphagia, which compromise nutritional intake and increase the risk of malnutrition and related adverse health outcomes [12, 13]. Additionally, poor dental appearance, tooth loss, and halitosis may diminish self-esteem, promote social withdrawal, and negatively affect psychosocial well-being among older adults [14]. Oral health problems in this population arise from the combined effects of age-related physiological changes, including degeneration of oral tissues, and an increased susceptibility to chronic diseases. This vulnerability is further compounded by geriatric syndromes such as physical frailty, functional dependence, and cognitive impairment, creating a cumulative burden that accelerates oral health deterioration [15].

In the specific context of hospitalized older adults, poor oral health increases the risk of aspirating oral bacteria into the lungs, leading to hospital-acquired infections such as pneumonia and negatively affecting mortality rates, length of stay, and healthcare costs [16]. Maeda et al. further reported that poor oral health status at admission independently predicted in-hospital mortality among acutely hospitalized older patients [17]. Therefore, oral health assessment in older inpatients is essential for identifying oral conditions, evaluating the risk of systemic complications, and supporting comprehensive health management.

Currently, the most commonly used instruments for assessing oral health in older adults are the Brief Oral Health Status Examination (BOHSE) and its revised version, the Oral Health Assessment Tool (OHAT) [18, 19]. Although these instruments have been translated into multiple languages and validated across various settings, they primarily focus on screening observable clinical manifestations of oral disease [20–22]. For example, the OHAT assesses eight domains—lips, gums, oral mucosa, tongue, natural teeth, dentures, mastication, and swallowing—through visual inspection of anatomical integrity and basic oral function [19]. Similarly, the BOHSE evaluates observable characteristics such as oral cleanliness, mucosal color, and gingival swelling [18]. However, neither instrument assesses important dimensions such as the qualitative characteristics of orofacial pain (e.g., pain frequency), sensory abnormalities, temporomandibular joint function, or occlusal support. These dimensions are not merely supplementary but are integral to understanding a patient’s oral health status [23]. For instance, the presence and frequency of orofacial pain may severely restrict mastication and nutritional intake regardless of the number of remaining teeth. Likewise, temporomandibular joint dysfunction can impair chewing efficiency, while sensory abnormalities may increase the risk silent aspiration. Therefore, the BOHSE and OHAT have important limitations in comprehensively evaluating oral health, particularly the physiological and perceptual dimensions that are highly relevant to older hospitalized patients.

Oral health encompasses both physical and psychosocial dimensions. The psychosocial dimension, operationalized as oral health-related quality of life, has been extensively assessed using well-validated patient-reported outcome measures, such as the Oral Health Impact Profile (OHIP), which has been translated and validated in Chinese populations [24]. In contrast, a significant gap remains in the comprehensive, standardized assessment of the physical dimension, particularly for hospitalized older adults at imminent risk of aspiration pneumonia, malnutrition, and functional decline, all of which are directly linked to physiological impairments [16, 17]. Existing physical oral health indices are either condition-specific (e.g., periodontal and caries indices) and lack a common metric, or they combine objective measures with subjective self-reports without a clear conceptual foundation. To date, no single instrument integrates structural integrity, functional capacity, and pain perception into a unified physical oral health score for the general older population [23].

In contrast, the Physical Oral Health Index (PHOX), developed by Reissmann et al. [25], offers a more comprehensive and theoretically grounded approach to assessing the physical dimension of oral health. The conceptual framework underlying the PHOX is based on a dual-axis model that explicitly distinguishes physical oral health from psychosocial oral health [25]. Physical oral health is defined as the objectively measurable condition and function of all anatomical oral structures, with pain incorporated as a key perceptual component [25]. Accordingly, the PHOX operationalizes this construct across multiple domains, including teeth and supporting structures, intraoral and extraoral soft tissues, jaw function, and pain/sensation, enabling a holistic evaluation of the physical aspects of oral health using a single, standardized index.

Therefore, this study aimed to translate and cross-culturally adapt the PHOX into Chinese and evaluate its validity and reliability among older hospitalized patients. By focusing specifically on the physical axis of oral health, this work seeks to fill the critical gap in standardized physiological assessment, providing clinicians with a tool to identify functional impairments that directly affect patient safety and nutritional status, thereby informing targeted nursing interventions that are distinct from psychosocial screening.

Methods

Study design

We followed the eight-step practical guide for cross-cultural adaptation developed by Cruchinho et al. [26]. This guide provides a systematic framework for adapting measurement instruments across cultures while minimizing methodological bias arising from researcher subjectivity or inexperience during the translation and adaptation process. Prior to translation, permission to adapt the PHOX was obtained from its original developer via email (Supplementary Material S1).

The Original PHOX

The original PHOX, developed by Reissmann et al. [25], is a comprehensive instrument for assessing physical oral health. It comprises 14 items organized into five domains: teeth and supporting structures, intraoral soft tissues, extraoral soft tissues and the jaw, function, and perception. Each item is scored on a 0–4 ordinal scale, and the weighted item scores are summed to generate a composite score ranging from 0 to 100, with lower scores indicating poorer oral health. The instrument demonstrated acceptable criterion validity, with correlation coefficients of 0.43 and 0.55 against patient- and examiner-based assessments, respectively. It also demonstrated high test–retest reliability (intraclass correlation coefficient [ICC] = 0.87), supporting its stability over time.

The original PHOX employs an item-specific weighting system (weights: 1–3) to derive a composite summary score ranging from 0 (worst oral health) to 100 (optimal oral health). According to Reissmann et al., this weighting scheme serves two purposes: (i) to reflect the relative clinical importance of different oral anatomical components in overall physical oral health, as determined by an expert panel using a modified Delphi technique and hierarchical regression analyses; and (ii) to standardize the total score on an intuitive 0–100 scale for clinical and research use. The highest weights were assigned to the “Teeth/supporting structures” domain (44% of the total weight), followed by “Function” and “Intraoral soft tissues” (16% each), reflecting their stronger associations with global oral health ratings and oral health-related quality of life (OHRQoL) in the original pilot study.

Translation, cross-cultural adaptation, and psychometric testing

Step 1: forward translation

Two native Chinese speakers proficient in English independently translated the original English version into Chinese (versions C1 and C2). The translators were instructed to ensure conceptual equivalence between the source and target versions while using clear and concise language. One translator, who did not have a medical background, focused on semantic equivalence, whereas the other, a healthcare professional with expertise in the instrument content, focused on conceptual equivalence. Both translators were unfamiliar with the questionnaire and unaware that it would subsequently undergo back-translation. They documented their queries and comments according to the format provided by the research team.

Step 2: forward translation synthesis

At this stage, a committee meeting was held to discuss discrepancies between the two forward translations. The committee comprised the two forward translators, two master’s-prepared nurses, one geriatrician, one geriatric nurse, one dentist, and one dental nurse. Committee members were provided with a comparison table containing the original instrument, the two translated versions (C1 and C2), and the queries and comments documented during Step 1. The committee resolved ambiguities and discrepancies between the forward translations and reached consensus on the most appropriate wording for each item, thereby producing the reconciled version (C3).

Step 3: back translation

Back-translation is a rigorous, standardized quality assurance procedure that is integral to the cross-cultural adaptation of psychometric instruments [27]. The synthesized Chinese version (C3) was independently translated back into English (BC1 and BC2) by two additional bilingual translators who were completely blinded to the original instrument. They were not involved in Step 1 to ensure the independence and blinding of the back-translation process. Each translator submitted a written report accompanying the completed translated version. Subsequently, the research team compared the two back-translated versions to identify discrepancies in wording, meaning, and conceptual equivalence. These discrepancies, together with both back-translated versions and the translators’ written reports, were documented and forwarded to the expert committee established in Step 4 for harmonization and resolution.

Step 4: harmonization

The two translators involved in Step 3 were invited to join the expert committee established in Step 2. Building on the discrepancies identified during Step 3, the committee systematically compared the original instrument, the reconciled Chinese version, and the two English back-translated versions. The evaluation focused on identifying significant semantic, conceptual, and contextual differences between the versions rather than achieving literal correspondence. Through this process, potential ambiguities and culturally incongruent items were identified and discussed. Accordingly, the Chinese version was revised to maximize conceptual equivalence within the Chinese cultural context. Finally, a pre-test version of the Chinese PHOX (PHOX-Chinese) was developed.

Step 5: pre-testing

The pre-test aimed to identify issues that might affect the reliability and validity of the translated instrument and to resolve any remaining semantic and conceptual discrepancies. Because the PHOX contains specialized dental terminology, this study employed an expert panel to conduct pre-testing and assess content validity, thereby ensuring the clarity, relevance, and cultural appropriateness of the Chinese version. The use of an expert panel offers distinct advantages in the cross-cultural adaptation of measurement instruments involving highly specialized terminology [26]. Because an excessively large number of experts may reduce the consistency of opinions, a panel of 5–10 experts is recommended for such evaluations [28]. In this study, the expert panel comprised eight members: one dentist, one dental nurse, one geriatrician, one geriatric nurse, one senior nursing researcher, two master’s students in nursing, and one psychology expert.

First, the experts independently reviewed the instrument and provided quantitative ratings for each item based on relevance and clarity using a 4-point Likert scale (1 = not relevant/clear; 4 = highly relevant/clear). Experts were invited to provide specific qualitative recommendations for items they considered problematic, and their ratings and written comments were subsequently collected by the research team. Expert ratings were dichotomized (ratings of 3–4 = 1, relevant/clear; 1–2 = 0, not relevant/clear) to calculate the item-level and scale-level content validity indices (I-CVI and S-CVI/Ave) separately for relevance and clarity. Excellent content validity was defined as an I-CVI ≥ 0.78 and an S-CVI/Ave ≥ 0.90 for both dimensions [29].

In addition to the expert panel evaluation, we conducted a clinical pre-test to assess the feasibility and practicality of the PHOX-Chinese in real-world clinical settings. A convenience sample of 30 older inpatients (aged 65–89 years) admitted to the geriatric department, who met the same inclusion criteria as those in the main study, was recruited for the pre-test. Three trained assessors (two nurses and one dentist) administered the full PHOX-Chinese to these participants according to the standardized protocol. For each assessment, the assessor recorded the total time required to complete the evaluation, documented any items that were difficult to administer or interpret, and noted patients’ spontaneous questions or expressions of confusion during the verbal questioning section. The assessors also provided written feedback regarding the clarity of the instructions, the ease of scoring each item, and the overall flow of the assessment. Patient responses during the verbal questioning were additionally used to evaluate item comprehension. All feedback from both the assessors and patients was systematically collated, summarized, and presented to the expert panel for discussion alongside the experts’ quantitative ratings.

Based on this comprehensive review, which included the expert panel’s quantitative ratings and feedback from the clinical pre-test, the expert panel convened to conduct structured discussions on items with low ratings or divergent opinions, focusing on resolving issues related to semantic ambiguity, cultural relevance, and the standardization of professional terminology. Following extensive discussion, the panel reached consensus and performed the final revisions to the instrument’s items, response options, and instructions, thereby producing the final version of the PHOX-Chinese for subsequent large-scale psychometric testing.

Step 6: field testing

The primary task in this step was to finalize the pre-final version of the measurement instrument, including determining whether reverse scoring was required for negatively worded items and calculating the minimum sample size for psychometric validation. Because the PHOX-Chinese primarily assesses oral health in older adults, no items were negatively worded; therefore, reverse scoring was not required. A minimum sample size of 5–10 participants per item is commonly recommended for factor analysis to ensure a robust factor structure [26, 30]. Accordingly, this study adopted a ratio of 10 participants per items. Given that the PHOX-Chinese contains 14 items and assuming a 20% attrition rate, the minimum required sample size was 175.

Step 7: psychometric validation

A cross-sectional study design was employed to evaluate the psychometric properties of the PHOX-Chinese. The psychometric evaluation included assessments of internal consistency reliability, test–retest reliability, and structural validity using both exploratory factor analysis (EFA) and confirmatory factor analysis (CFA).

Participants

Participants were recruited using a convenience sampling approach based on their availability and willingness to participate. The study was conducted between September 2025 and January 2026 in the departments of geriatrics, general practice, respiratory medicine, cardiovascular medicine, and neurology at a tertiary hospital in Zhejiang Province, China. The inclusion criteria were as follows: (1) age ≥ 65 years; (2) provision of informed consent to participate voluntarily. Older adults with severe cognitive impairment or other psychiatric disorders that could preclude reliable self-reporting or cooperation during the oral examination were excluded. Ultimately, 1,365 older inpatients were recruited.

Sample size

Park et al. [31] suggested that the sample size for assessing test–retest reliability should be at least eight times the number of instrument items to obtain robust ICC estimates. Given the 14 items in the PHOX-Chinese and assuming a 20% attrition rate, a minimum of 140 participants was considered appropriate. From the total study sample, 140 participants were selected using systematic random sampling for the test–retest reliability assessment.

Assessment tools

Demographic and clinical data were collected from participants, including age, sex, body mass index (BMI), comorbidity (≥ 2 chronic diseases: yes/no), polypharmacy (≥ 5 medications: yes/no), alcohol consumption (never, former, or current), smoking history (never, former, or current), education level (< 5 years or ≥ 5 years), and place of residence (urban or rural).

The PHOX-Chinese consists of 14 items across five domains, each rated on a 0–4 ordinal scale. Weighted item scores (raw score × weight) were calculated and summed to generate a total score ranging from 0 to 100, with lower scores indicating poorer oral health.

Data collection

Data were collected by a team of four nurses and two dentists. All assessors completed a 2-hour training session supervised by a dental professional and were required to pass a competency assessment before data collection commenced. The training covered oral anatomy, common oral diseases, the use of examination instruments (e.g., periodontal probes and dental mirrors), examination procedures (including oral percussion and palpation), and the scoring criteria for each PHOX-Chinese item.

The oral examination kit included tongue depressors, gauze, handheld lights, disposable gloves, a standard periodontal probe, and a dental mirror. To ensure standardized positioning and optimal visualization of the oral cavity, all participants were examined in either a recumbent or semi-recumbent position in their hospital beds. Any disagreements in scoring during the oral health assessment were resolved through discussion between the two dentists.

A short retest interval may increase the likelihood of participants recalling their previous responses, whereas a longer interval may allow genuine clinical changes to occur. Therefore, a retest interval of 1–2 weeks is generally considered appropriate [31, 32]. In this study, the retest assessment was conducted two weeks after the initial evaluation. Participants were reassessed by the same trained assessors using the same instruments and evaluation criteria. Participants who had been discharged were contacted by telephone to arrange a convenient time and location for the follow-up assessment. They were informed that the assessment was part of a routine follow-up and were not told that the results would be compared with those from the initial assessment, thereby minimizing potential response bias.

Statistical analysis

SPSS was used to randomly divide the total sample into calibration and validation datasets, whereas RStudio was employed to calculate Cronbach’s α coefficients and perform EFA and CFA. Descriptive statistics included frequencies and percentages for categorical variables, means ± standard deviations for normally distributed continuous variables, and medians with interquartile ranges for non-normally distributed continuous variables. Differences between the calibration and validation datasets were assessed using chi-square tests, independent-samples t tests, or non-parametric tests, as appropriate.

Internal consistency was evaluated using Cronbach’s α. For measurement instruments adapted and validated across different cultural contexts, a Cronbach’s α > 0.7 is generally considered acceptable [33]. Composite reliability (CR) was calculated from the CFA results, with values > 0.7 indicating good reliability [34]. Test–retest reliability was assessed using ICC, with values interpreted as moderate (0.50–0.75), good (0.75–0.90), and excellent (> 0.90) [35].

Sampling adequacy and factorability were assessed using the Kaiser–Meyer–Olkin (KMO) measure (> 0.8) and Bartlett’s test of sphericity (p < 0.05) [36]. Principal axis factoring with promax rotation was used for factor extraction. The number of factors was determined primarily by parallel analysis and was supplemented by examination of the scree plot and eigenvalue > 1 criterion [37]. Items were retained if their primary factor loading was ≥ 0.40 and their cross-loadings were < 0.20.

A cross-validation approach was employed. EFA was first conducted using the entire dataset. The sample was then randomly divided into calibration (n = 838) and validation (n = 527) samples in a 6:4 ratio. This ratio was selected because the calibration phase requires a larger sample to ensure stable parameter estimation, while the validation sample remained well above the minimum size required for reliable CFA. A 6:4 split is widely used in psychometric studies when the total sample is sufficiently large, as it maximizes statistical power for model development while preserving adequate power for independent validation [38, 39]. CFA was subsequently performed on the calibration sample using maximum likelihood estimation, followed by validation in the independent validation sample and, finally, in the total sample (N = 1,365). Model fit was evaluated using the following criteria: χ²/df < 5; standardized root means square residual (SRMR) ≤ 0.05; root mean square error of approximation (RMSEA) ≤ 0.08; goodness-of-fit index (GFI) > 0.90; comparative fit index (CFI) > 0.90; and Tucker–Lewis index (TLI) > 0.90 [40].

To examine whether the original German weighting scheme remained applicable to the Chinese older inpatient population, we performed a weight equivalence analysis using a CFA-based proportional weighting approach [41]. Standardized factor loadings obtained from the CFA were used to recalculate item weights within each domain while preserving the original domain-level weight totals. Total scores were then calculated using both the original and the CFA-derived weighting schemes. Agreement between the two scoring methods was assessed using Pearson’s product–moment correlation coefficient and Bland–Altman analysis, with the 95% limits of agreement (LoA) calculated as the mean difference ± 1.96 × the standard deviation of the differences [42]. As a sensitivity analysis, two additional data-driven weighting methods (Criteria Importance Through Intercriteria Correlation and entropy weighting) were applied to further examine the robustness of the findings.

The quality of the measurement model was evaluated using composite reliability (CR; acceptable: >0.70) and average variance extracted (AVE; acceptable: >0.50), which reflect the internal consistency and convergent validity of the factor structure [34]. Discriminant validity was assessed using the Fornell–Larcker criterion, which compares the square root of the AVE for each latent construct with its correlations with all other latent constructs. Adequate discriminant validity is demonstrated when the square root of the AVE exceeds the construct’s correlations with all other constructs in the model [43].

Step 8: analysis of psychometric properties

In accordance with Cruchinho et al., the final psychometric findings were reviewed to verify that the PHOX-Chinese demonstrated satisfactory reliability, validity, and overall consistency. Based on the results of the content validity, construct validity, and reliability analyses, the PHOX-Chinese was confirmed as an appropriate instrument for assessing physical oral health among older Chinese inpatients.

Results

Translation and cross-cultural adaptation

During the forward translation process, both translators indicated that the technical term “Ramfjord teeth” in Item 3 was difficult to understand and raised concerns regarding the translation of the terms “familiar pain” and “unfamiliar pain” in Item 8, as well as “supporting zones” in Item 12. During the forward translation synthesis, the committee decided to add an explanatory note for “Ramfjord teeth.” “Familiar pain” was translated as “recurrent, habitual pain,” “unfamiliar pain” as “first-time pain,” and “supporting zones” as “dental occlusal support zones.” Subsequently, through the back-translation and harmonization processes, a pre-test version of the PHOX-Chinese was developed. The synthesized version (C3) was highly similar to the pre-test version, with only minor revisions made to improve language conciseness, clarity, and readability.

Pilot testing of the PHOX-Chinese demonstrated that the instrument was feasible and well accepted among older hospitalized patients. The mean completion time was 12.60 ± 3.61 min (range: 8–18 min). All items were administered smoothly by the trained nurses and dentists, and no operational difficulties were reported. For the verbally administered items (e.g., Items 8, 13, and 14), 27 of the 30 patients (90.0%) provided coherent and consistent responses without requiring further clarification. The remaining three patients (10.0%) expressed confusion regarding Item 8 (“familiar pain”), prompting the addition of illustrative examples to the user guide to standardize interpretation across assessors. The assessors also reported that the standardized scoring criteria and supplementary user guide facilitated smooth implementation and ensured consistent interpretation among professionals from different disciplinary backgrounds. No substantial modifications were made to the item content after the pre-test. However, based on the aggregated feedback from patients and assessors, minor adjustments were made to the wording of Item 8, and the user guide was expanded to further enhance clarity and standardization.

During the pre-test, adjustments were made to improve the relevance and clarity of the pre-test version based on expert feedback, ensuring that the adapted instrument was culturally appropriate and readily understandable in the Chinese context. The expert committee also considered it necessary to develop a user guide for the PHOX-Chinese (Supplementary Material S2). Consequently, the final 14-item version of the PHOX-Chinese was established (Supplementary Material S3).

Baseline characteristics of participants

A total of 1,365 older inpatients participated in this study. Their mean age was 76.54 ± 7.83 years (range: 65–103 years). Men accounted for 53.11% of the sample, whereas women comprised 46.89%. The participant characteristics are presented in Table 1.

Table 1.

Demographic and clinical characteristics of the participants (N = 1,365)

Variable	Total sample (N = 1,365)	Calibration sample (n = 838)	Validation sample (n = 527)	p-value
Sex				0.378
Male	725	453	272
Female	640	385	255
Age (y)	76.54 ± 7.83	76.51 ± 7.83	76.59 ± 7.82	0.845
BMI (kg/m²)	23.65 ± 4.25	23.66 ± 4.24	23.64 ± 4.28	0.945
Comorbidity				0.391
Yes	613	384	229
No	752	454	298
Polypharmacy				0.489
Yes	565	353	212
No	800	485	315
Education				0.287
< 5 years	1034	643	391
≥ 5 years	331	195	136
Residence				0.276
Urban	783	471	312
Rural	582	367	215
Alcohol consumption				0.199
Never	1103	672	431
Past	79	56	23
Current	183	110	73
Smoking history				0.894
Never	1121	685	436
Past	93	58	35
Current	151	95	56

Reliability

Internal consistency

The PHOX-Chinese demonstrated an overall Cronbach’s α of 0.905. Cronbach’s α coefficients for the five domains were 0.912, 0.925, 0.889, 0.897, and 0.903, respectively, all exceeding 0.80, indicating excellent internal consistency.

A total of 132 participants completed both assessments in the test–retest reliability study, yielding a valid response rate of 94.3%. The ICC for the total scale was 0.923 (95% confidence interval [CI]: 0.595–0.971). The ICC values for the five domains were as follows: Domain 1 = 0.918, 95% CI [0.818, 0.956]; Domain 2 = 0.919, 95% CI [0.882, 0.944]; Domain 3 = 0.858, 95% CI [0.499, 0.940]; Domain 4 = 0.876, 95% CI [0.812, 0.917]; and Domain 5 = 0.888, 95% CI [0.845, 0.919]. The mean scores for each domain and the total scale did not differ significantly between the two assessments (p > 0.05), indicating the absence of systematic measurement bias. These findings demonstrate excellent temporal stability over a two-week interval.

Validity

Content validity

The I-CVI for Item 9 was 0.75 in the first round of assessment, which was below the acceptable threshold of 0.78. Based on the experts’ recommendations, this item was revised. In the second round of evaluation, the I-CVI for all items and the S-CVI/Ave both reached 1.00, indicating excellent content validity.

Regarding language clarity, the clarity index for Item 10 was 0.75 in the first round of assessment, which was below the acceptable threshold (≥ 0.78). The experts’ written feedback indicated that the wording of this item was difficult to understand. The research team therefore made targeted revisions based on the experts’ recommendations. In the second assessment round, the clarity indices for all 14 items and the overall S-CVI/Ave both reached 1.00. Consequently, all items were retained, and the instrument was considered clear, understandable, and linguistically appropriate.

Construct validity

EFA was conducted using data from 1,365 participants. The KMO measure of sampling adequacy was 0.859, and Bartlett’s test of sphericity was significant (χ² = 15657.7, p < 0.001), indicating that the data were well suited for factor analysis. Based on the results of the parallel analysis, together with the eigenvalues, scree plot, and factor interpretability, five common factors were extracted. The parallel-analysis scree plot is presented in Fig. 1. The five-factor solution explained 79.2% of the total variance. Factor loadings ranged from 0.71 to 0.95, and no item exhibited substantial cross-loadings (difference < 0.2). The factor-loading matrix is presented in Fig. 2.

Fig. 1.

Parallel analysis scree plot

Fig. 2.

Factor loading matrix of exploratory factor analysis

The final PHOX-Chinese comprised 14 items across five domains: Domain 1 (teeth/supporting structures), Items 1–4; Domain 2 (intraoral soft tissue), Items 5–7; Domain 3 (soft tissues and jaw), Items 8–10; Domain 4 (function), Items 11–12; and Domain 5 (perception), Items 13–14. Following expert panel discussions, the original scoring system was retained in the Chinese version (see Supplementary Material S3).

Sociodemographic characteristics were compared between the calibration and validation samples to assess their comparability for CFA. No significant differences were observed (p > 0.05), indicating successful sample splitting and good comparability between the two samples. Detailed results are presented in Table 1.

The five-factor structure identified through EFA was used to specify the initial CFA model in the calibration sample (n = 838). All model-fit indices were acceptable or good (Table 2). The stability of the model was subsequently evaluated using the independent validation sample (n = 527), which also demonstrated good model fit (see Table 2), supporting the stability of the factor structure across samples and indicating no evidence of overfitting. To further confirm the structural validity of the PHOX-Chinese, the model was then fitted using the entire sample (N = 1,365). The model demonstrated acceptable to good fit (Table 2: χ²/df = 6.73, RMSEA = 0.065, CFI = 0.975, TLI = 0.967, and SRMR = 0.027).

Table 2.

Confirmatory factor analysis model fit indices for the chinese version of the PHOX

	χ²/df (< 5)	RMSEA ≤ 0.08 (95% CI)	PNFI > 0.5	SRMR ≤ 0.05	GFI > 0.9	CFI > 0.9	TLI > 0.9
Calibration sample(n = 838)	4.12	0.061 (0.054–0.069)	0.716	0.026	0.992	0.979	0.971
Validation sample (n = 527)	3.59	0.070 (0.061–0.080)	0.706	0.032	0.988	0.970	0.960
Total sample (N = 1,365)	6.73	0.065 (0.059–0.071)	0.715	0.027	0.992	0.975	0.967

Abbreviations: SRMR Standardized root mean square residual, RMSEA Root mean square error of approximation, GFI Goodness-of-fit index, CFI Comparative fit index, TLI Tucker–Lewis index, PNFI Parsimonious normed fit index

The standardized factor loadings in the total-sample CFA model ranged from 0.78 to 0.961 (all p < 0.01; Table 3). The CR values (0.898–0.930) and AVE values (0.742–0.834) exceeded the recommended thresholds, indicating good internal consistency convergent validity of the five-factor measurement model (Table 3). As shown in Table 4, the square roots of the AVE for each latent construct (0.861–0.913; diagonal values) exceeded the correlations between that construct and all other constructs (0.106–0.627), indicating good discriminant validity. The item correlation matrix is presented in Fig. 3.

Table 3.

Confirmatory factor analysis results for the Chinese version of the PHOX: standardized factor loadings, composite reliability, and average variance extracted

Dimension	Item	Standardized Factor Loadings	Composite Reliability	Average Variance Extracted
Dimension 1	Item 1	0.862*	0.920	0.742
	Item 2	0.923*
	Item 3	0.875*
	Item 4	0.78*
Dimension 2	Item 5	0.934*	0.930	0.817
	Item 6	0.954*
	Item 7	0.818*
Dimension 3	Item 8	0.812*	0.902	0.755
	Item 9	0.854*
	Item 10	0.936*
Dimension 4	Item 11	0.961*	0.909	0.834
	Item 12	0.863*
Dimension 5	Item 13	0.924*	0.898	0.815
	Item 14	0.881*

* p < 0.05

Table 4.

Discriminant validity of the dimensions of the Chinese version of the PHOX

	Dimension 1	Dimension 2	Dimension 3	Dimension 4	Dimension 5
Dimension 1	0.861	0.318	0.627	0.572	0.597
Dimension 2		0.904	0.160	0.106	0.186
Dimension 3			0.869	0.574	0.485
Dimension 4				0.913	0.566
Dimension 5					0.903

Fig. 3.

Heatmap of correlation matrix among items of the Chinese version of the PHOX

Weight equivalence analysis

To examine weight equivalence, item weights were recalculated using standardized CFA factor loadings while preserving the original domain-level weight totals. The CFA-derived weights showed excellent agreement with the original weights (Fig. 4). Total scores calculated using the CFA-derived weights were almost perfectly correlated with those calculated using the original weights (Pearson’s r = 0.999, p < 0.001). Bland-Altman analysis demonstrated a mean difference of 0.49 points, with 95% limits of agreement ranging from − 1.01 to 1.99 on the 0–100 scale (Fig. 5). The maximum absolute difference between the two scoring methods was 3.00 points. Sensitivity analyses using the CRITIC and entropy weighting methods also yielded total scores that were significantly correlated with those generated using the original weighting scheme (r ≥ 0.985, all p < 0.001). See Figure S1 and Figure S2.

Fig. 4.

Weight equivalence test: original vs. CFA-based weights

Fig. 5.

Bland-Altman plot: agreement between original and CFA-based weights

Discussion

This study successfully translated and cross-culturally adapted the PHOX into Chinese using Cruchinho’s eight-step methodology and validated its psychometric properties in a large sample of older inpatients in China. The PHOX-Chinese demonstrated excellent reliability, strong structural validity, and good feasibility for clinical use. The following sections discuss the key findings from four perspectives: the cross-cultural adaptation process, psychometric properties, dimensional structure, and clinical applicability.

The cross-cultural adaptation process of the PHOX-Chinese achieved a balance between fidelity and practicality

The cross-cultural adaptation of the PHOX into Chinese followed Cruchinho’s eight-step practical guide [26], which provides a systematic framework for maximizing semantic, conceptual, and cultural equivalence between the original and target versions. Several key challenges encountered during this process, together with their corresponding resolutions, warrant discussion.

First, the translation of dental-specific terminology required careful consideration. The term “Ramfjord teeth”—which refers to specific index teeth used in periodontal assessment—presented a conceptual challenge because this technical concept is unfamiliar to many Chinese clinicians outside specialized dental practice. Rather than adopting a literal translation that could obscure its meaning, the expert committee decided to retain the original term and provide an explanatory note in the user guide, enabling clinicians to accurately identify the designated teeth during examination. This approach balanced technical accuracy with clinical usability, reflecting a key principle of cross-cultural adaptation: when a source-language concept lacks a direct linguistic equivalent, providing supplementary contextual information may be preferable to a forced literal translation [27].

Second, the distinction between “familiar pain” and “unfamiliar pain” in Item 8 required conceptual refinement. The committee addressed this by translating “familiar pain” as “recurrent, habitual pain” and “unfamiliar pain” as “first-time pain,” thereby framing the distinction in terms of the temporal pattern of pain rather than emotional familiarity. This adaptation preserved the intended clinical meaning while improving cultural relevance. It also illustrates that when direct conceptual equivalence cannot be achieved, functional equivalence—maintaining the same clinical intent through culturally appropriate wording—represents a valid and methodologically sound alternative [44].

Third, the wording of Item 10, originally phrased as “size ratio of jaws (tooth position),” was revised to “occlusal relationship” based on expert feedback. The original wording was considered potentially misleading because the item assesses abnormalities in occlusal relationships rather than the actual dimensional proportions of the jaws. This revision enhanced clarity and reduced ambiguity, ensuring that clinicians would interpret the item as intended. The adaptation process also involved a deliberate decision to retain the original scope of Item 9 (integrity of the jawbone, palate, and tongue). Several experts initially proposed shifting the focus to alveolar ridge height because of its clinical relevance to denture retention among older Chinese adults. However, following consultation with the original author, the committee decided to retain the original wording. The developer explained that alveolar bone atrophy is largely an age-related physiological change rather than a pathological condition and that incorporating it would deviate from the original conceptual framework of the PHOX, which is intended to assess pathological conditions and their consequences rather than normal physiological processes [25]. This decision underscores an important principle of cross-cultural adaptation: the theoretical constructs underlying the original instrument should be preserved in the adapted version rather than modified for local clinical practice or convenience. Although this principle is fundamental to rigorous instrument adaptation, it is not consistently observed in translation studies [45, 46].

Collectively, these adaptation decisions reflect the complex considerations inherent in cross-cultural instrument development, including balancing linguistic accuracy, conceptual equivalence, clinical feasibility, and theoretical fidelity. The successful resolution of these challenges is further supported by the excellent content validity of the PHOX-Chinese, with both the I-CVI and S-CVI/Ave reaching 1.00 in the second round of expert review.

The PHOX-Chinese demonstrated robust psychometric properties

Internal consistency and test–retest reliability

The PHOX-Chinese demonstrated excellent internal consistency, with Cronbach’s alpha coefficients exceeding 0.88 for both the total scale and all five dimensions. Several factors may explain this finding. First, the conceptual clarity achieved during the translation and cross-cultural adaptation process minimized item ambiguity, ensuring that each item consistently measured its intended construct across participants. Second, the relative homogeneity of the study population—older inpatients with similar health profiles—likely reduced extraneous variability in responses. Third, the five-factor structure identified through EFA and confirmed by CFA provided a coherent theoretical framework in which items within each dimension measured a common underlying construct, as reflected by the high composite reliability (CR) values (0.898–0.930), all of which exceeded the recommended threshold of 0.70 [47].

Test–retest reliability over the two-week interval was also excellent (ICC = 0.923), exceeding that reported for the original PHOX (ICC = 0.87). Several factors may explain this finding. First, the assessors underwent rigorous standardized training, which likely minimized inter-rater variability during data collection. Second, high temporal stability is clinically meaningful in the inpatient setting, where substantial changes in oral health are relatively uncommon over a short period. These findings indicate high temporal stability of the PHOX-Chinese; whether it can detect clinically meaningful changes over time warrants further investigation through responsiveness analysis.

Construct validity

The construct validity of the PHOX-Chinese was rigorously evaluated using EFA and CFA within a cross-validation framework, thereby enhancing the robustness of the findings [37]. Guided by parallel analysis, the EFA identified a five-factor structure that was consistent with the original instrument and explained 79.2% of the total variance [48]. Parallel analysis was employed to determine the number of factors to retain because its accuracy and robustness have consistently been shown to exceed those of Bartlett’s test, the eigenvalue > 1 criterion, and the scree test [37, 48]. Although the eigenvalue of the fifth factor was 0.85, which was slightly below the conventional threshold of 1.0, the parallel-analysis scree plot showed that the corresponding eigenvalue from the observed data exceeded that generated from random data [36]. Furthermore, the two items comprising this factor (Items 13 and 14) conceptually represented a distinct dimension and demonstrated excellent internal consistency (α = 0.897). Therefore, retaining the five-factor structure was both statistically and conceptually justified.

The subsequent CFA provided further support for the proposed factor structure. Although the χ²/df value for the total sample (6.73) exceeded the conventional threshold of < 5, this index is well known to be highly sensitive to large sample sizes and should therefore be interpreted alongside other model-fit indices rather than in isolation [49, 50]. With a sample of 1,365 participants, even minor model misspecifications can inflate the χ² statistic, whereas alternative indices such as the RMSEA and SRMR are considerably less influenced by sample size [51]. The consistently high GFI, CFI, and TLI vales (> 0.96) across all samples, together with the low and stable SRMR values, provides strong evidence that the hypothesized five-factor model adequately represents the internal structure of the PHOX-Chinese.

The high AVE and CR values further demonstrated strong convergent validity within each dimension, confirming that the items coherently measured their respective latent constructs [47]. Discriminant validity was also supported by the Fornell–Larcker criterion. Notably, the most significant correlation was observed between Domain 1 and Domain 3 (r = 0.627), which is clinically plausible because the structural integrity of the teeth and supporting bone directly influences the condition of the surrounding soft tissues and jaw relationships. Despite this expected association, each dimension remained empirically distinct, confirming the instrument’s ability to differentiate the multiple components of physical oral health into interpretable and clinically meaningful domains.

Weight equivalence analysis

An important consideration in cross-cultural instrument adaptation is whether the original scoring weights, which reflect the developers’ conceptualization of the relative importance of each domain, remain applicable in the target population. The original PHOX weights were established through a combination of expert consensus and empirical associations with global oral health ratings and OHRQoL in a German sample [25]. Although the present study investigated a markedly different population—Chinese older inpatients rather than German general dental patients—the excellent agreement and narrow Bland-Altman limits observed in the weight equivalence analysis suggest that the relative importance of the oral health domains remains stable across these populations. These findings provide preliminary evidence supporting the cross-cultural applicability of the original PHOX weighting scheme. Nevertheless, future studies should evaluate weight equivalence in other populations and healthcare settings to further establish the generalizability of the original scoring system.

The PHOX-Chinese maintained structural fidelity to the original five-factor model

The five-factor structure identified through EFA corresponded precisely to the dimensions proposed in the original PHOX. No item exhibited substantial cross-loadings, and the proportion of variance explained was high, indicating that the translated items measured the same latent constructs as the original instrument. This structural consistency provides further support for the cross-cultural applicability of the dual-axis theoretical model underpinning the PHOX.

It is also important to situate the PHOX-Chinese within the broader landscape of geriatric oral health assessment tools. The OHAT and BOHSE were primarily developed as rapid screening instruments for use by non-dental healthcare professionals. Their items predominantly assess observable pathological signs—such as lip condition, gingival swelling, oral cleanliness, and dry mouth—and include only a basic evaluation of swallowing and chewing difficulties [52]. By integrating the structural, functional, and perceptual dimensions of oral health, the PHOX-Chinese extends the scope of existing screening tools. This finding reinforces the theoretical robustness of the scale beyond its original German context.

The PHOX-Chinese shows promising potential for clinical application

The PHOX-Chinese possesses several features that facilitate its integration into routine geriatric care. The average administration time of approximately 12 min, together with the detailed user guide, enables nurses and non-dental healthcare professionals to complete standardized assessments after brief structured training. This is particularly relevant in Chinese tertiary hospitals, where dental professionals are not always readily available in general medical wards and oral care is often delegated to bedside nurses. By providing a standardized assessment protocol and explicit scoring criteria, the instrument has the potential to improve the quality and consistency of nursing assessment and documentation.

Unlike basic screening tools such as the OHAT, which focus primarily on visible pathology, the PHOX-Chinese systematically assesses masticatory function, occlusal support, and pain perception. This enables clinicians to identify early functional impairments that may compromise nutritional intake or increase the risk of aspiration. Early identification facilitates timely referral to dental specialists or dietitians, potentially preventing malnutrition and hospital-acquired pneumonia. Furthermore, by identifying older inpatients with the poorest oral health, the PHOX-Chinese may assist hospitals in prioritizing high-risk patients for comprehensive dental evaluation or specialized nutritional support. Such targeted interventions could optimize the use of limited healthcare resources and potentially reduce hospital length of stay and readmission associated with oral health-related complications.

The reproducibility of assessment results across different raters is a critical determinant of an instrument’s scalability in clinical practice. Although the present study did not formally evaluate inter-rater reliability, several methodological safeguards were implemented to minimize inter-rater variability. First, all six assessors—four nurses and two dentists—completed a standardized two-hour training program led by a dental professional, covering oral anatomy, common oral pathologies, instrument use, examination techniques, and item-specific scoring criteria. Each assessor was required to pass a competency assessment before participating in data collection. Second, we developed a comprehensive user guide that provides explicit operational definitions, illustrative examples, and decision algorithms for ambiguous cases. Third, during the main study, any scoring disagreements between the two dentists were resolved immediately through consensus discussion, thereby maintaining scoring consistency in real time. These measures are consistent with best-practice recommendations for rater-dependent clinical instruments, which emphasize that standardized training and operational protocols are fundamental to ensuring reproducibility. Nevertheless, formal evaluation of inter-rater reliability across different clinical settings and professional groups remains an important direction for future research to fully establish the scalability of the PHOX-Chinese.

Limitations

This study has several limitations that should be considered when interpreting the findings.

First, the generalizability of the findings is constrained by the sampling strategy. We employed convenience sampling at a single tertiary hospital in Zhejiang Province, China. Consequently, the sample may not be representative of the broader older adult population across different geographic regions, healthcare settings, or health conditions. The applicability of the PHOX-Chinese in community, rural, and long-term care settings, as well as its measurement invariance across demographic and clinical subgroups, remains to be established. Future studies should adopt multicenter, stratified sampling designs to evaluate its generalizability and examine measurement invariance.

Second, we did not assess criterion validity or convergent validity against external measures because no universally accepted gold standard exists for multidimensional physical oral health assessment in older adults. Commonly used tools such as the BOHSE and OHAT primarily assess observable pathology rather than the functional and perceptual domains captured by the PHOX, making direct comparisons conceptually challenging. Nonetheless, the strong evidence for content validity, construct validity, and reliability supports the overall validity of the PHOX-Chinese. Future studies should examine criterion and convergent validity using conceptually related instruments and investigate predictive validity against clinically relevant outcomes such as aspiration pneumonia, nutritional decline, and mortality.

Third, the exclusion of older inpatients with cognitive impairment or psychiatric disorders limits the applicability of the PHOX-Chinese to some of the most vulnerable patient groups. While this criterion was necessary to ensure reliable self-report and standardized data collection, the comprehensibility and measurement performance of the instrument in these populations remain unknown. Future research should validate the PHOX-Chinese among individuals with cognitive impairment using observational methods, proxy reporting, or modified administration procedures.

Fourth, although rigorous assessor training and a standardized user guide were implemented, we did not formally evaluate inter-rater reliability. Although the high test–retest reliability and standardized assessment procedures indirectly support measurement consistency, they cannot substitute for direct evidence of agreement between raters. Future studies should formally assess inter-rater reliability across diverse clinical settings and professional groups to quantify agreement and identify items that may be susceptible to variation in assessor interpretation.

Conclusion

This study systematically translated and cross-culturally adapted the PHOX into Chinese and evaluated its psychometric properties among hospitalized older adults in China. The PHOX-Chinese demonstrated excellent reliability, strong validity, and good feasibility for clinical use. It provides a comprehensive and standardized approach to assessing the physical dimensions of oral health, complementing existing assessment tools by integrating structural, functional, and perceptual domains. The PHOX-Chinese has the potential to facilitate the early identification of oral health problems and inform individualized nursing care planning; its impact on clinical outcomes remains to be established in future interventional studies. Future research should further validate the PHOX-Chinese across diverse populations and healthcare settings and evaluate its predictive validity and clinical utility.

Abbreviations

BOHSE

Brief Oral Health Status Examination

OHAT

Oral Health Assessment Tool

PHOX

Physical Oral Health Index

I-CVI

Item-level Content Validity Index

S-CVI/Ave

Scale-level Content Validity Index / Average

EFA

exploratory factor analysis

CFA

confirmatory factor analysis

ICC

intraclass correlation coefficient

CR

composite reliability

SRMR

standardized root mean square residual

RMSEA

root mean square error of approximation

GFI

goodness-of-fit index

CFI

comparative fit index

TLI

Tucker-Lewis index

AVE

Average variance extracted

Authors’ contributions

H.Y., J.X.: Conceptualization, Data curation, Funding acquisition, Formal analysis, Investigation, Methodology, Project administration, Software, Supervision, Validation, Visualization, Writing – original draft, Writing – review & editing. L.W., C.H., and Y.C.: Investigation, Project administration, Data curation, Supervision, Validation, Visualization, Writing – review & editing. D.R.R.: Methodology, Project administration, Resources, Writing – original draft. All authors read and approved the final manuscript.

Funding

This work was supported by the Zhejiang Provincial Medical and Health Science and Technology Plan [Grant No. 2025KY1857].

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

The study protocol was reviewed and approved by the Hospital Ethics Committee (Approval No. 2025-KY101-01). All procedures were conducted in accordance with the ethical standards of the 1964 Declaration of Helsinki and its subsequent amendments. All participants provided written informed consent prior to enrolment. Before providing consent, they received comprehensive information about the study objectives, procedures, potential risks and benefits, and were informed that they could withdraw from the study at any time without affecting their routine care.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.