Abstract
Objectives
To identify behavioral segments of integrative care use among older adults with knee osteoarthritis in a rural district in Northern Thailand and to compare these segments according to PRECEDE-related factors and baseline characteristics.
Methods
This analytical cross-sectional study included 302 community-dwelling older adults with primary knee osteoarthritis in Thoen District, Lampang Province, Northern Thailand. Data were collected through face-to-face interviews using a structured questionnaire that assessed demographic and health characteristics, PRECEDE-related factors, and behaviors across 3 care domains: professional medical care, traditional medicine, and community-based medicine. Standardized behavioral domain scores were analyzed using hierarchical cluster analysis, followed by k-means clustering. Differences among segments were examined using chi-square tests, one-way analysis of variance, or Kruskal-Wallis tests, as appropriate.
Results
Three behavioral segments were identified: high engagement (20.9%), intermediate engagement (48.0%), and low engagement (31.1%). The 3 behavioral domains differed significantly across segments (all p<0.001; η2=0.400–0.535). Baseline demographic and health characteristics did not differ significantly among segments. In contrast, significant differences among segments were observed for the total enabling score, access, decision-making, total reinforcing score, social support, and attitude, whereas perceived information showed no significant difference.
Conclusions
Older adults with knee osteoarthritis were not behaviorally homogeneous in their use of professional, traditional, and community-based care. Segment differences were more pronounced for enabling and reinforcing conditions than for perceived information alone. These findings support segment-informed health promotion strategies in primary care and community settings.
Keywords: Osteoarthritis, Aged, Health behavior, Integrative medicine, Primary health care, Thailand
INTRODUCTION
Knee osteoarthritis is a common source of pain and functional limitation in later life. As populations age, increasing numbers of people are living with chronic conditions that restrict mobility, reduce independence, and impair quality of life [1]. Musculoskeletal disorders contribute substantially to disability worldwide [2], and knee osteoarthritis is a condition that limits everyday functioning in older adults [3,4]. In Thailand, symptomatic knee osteoarthritis has been associated with female sex, older age, and obesity [5]. Accordingly, the condition warrants attention not only in clinical care, but also in health promotion and long-term care planning.
Recommended care for knee osteoarthritis includes exercise, self-management support, weight control, and regular follow-up within a multimodal management approach [6,7]. In practice, however, patterns of care use may be considerably more heterogeneous. Older adults may obtain services from hospitals or primary care units while also relying on traditional medicine, family assistance, or community resources [8]. These combinations are common in health systems where formal services remain closely connected with family and community support, as is the case in Thailand [9].
The local burden of knee osteoarthritis further underscores the need for primary care responses tailored to the study area. In Health Region 1, knee osteoarthritis was identified in 44 608 of 806 960 screened older adults (5.53%). The corresponding rates were 7.81% in Lampang Province and 9.56% in Thoen District, both of which exceeded the regional rate [10]. In this context, primary care must support treatment, follow-up, self-management, referral, health promotion, and linkage to community resources [11]. Community-level integrated care is also particularly relevant for chronic conditions in older adults, especially in settings where family members, local health personnel, and community networks contribute to day-to-day support [12,13].
Health promotion programs often approach older adults with knee osteoarthritis as a single target population. In rural communities, however, this approach may not reflect how care is actually used. Some individuals may rely primarily on professional services, whereas others may combine formal care with traditional or community-based support. Still others may have limited engagement with all available care options. Behavioral segmentation can identify subgroups with shared patterns of engagement [14] and help local programs allocate limited resources more appropriately [15]. The PRECEDE framework adds an explanatory perspective by examining predisposing, enabling, and reinforcing conditions that may be related to behavior [16]. Together, these approaches can inform communication strategies, support planning, and intervention design [17].
Much of the osteoarthritis literature has examined factors associated with behavior or service use as separate variables. Less is known about how older adults combine professional, traditional, and community-based care in everyday life, particularly in rural Thai settings where these options coexist. This study used the PRECEDE framework to examine differences in predisposing, enabling, and reinforcing conditions across patterns of integrative care use. The objectives were to identify behavioral segments among older adults with knee osteoarthritis in Thoen District, Lampang Province, Northern Thailand, and to compare these segments according to PRECEDE-related factors and baseline demographic and health characteristics.
METHODS
Study Design and Participants
This analytical cross-sectional study was conducted in Thoen District, Lampang Province, Northern Thailand, from August 2023 to December 2023. Thoen is a rural district comprising 8 subdistricts. The source population consisted of 1122 older adults with knee osteoarthritis recorded in local primary care registers. Eligible participants were community-dwelling adults aged 60 years or older who had primary knee osteoarthritis, had lived in the study area for at least 6 months, and were able to complete a face-to-face interview. Individuals with cognitive impairment that precluded meaningful participation were excluded. Primary knee osteoarthritis referred to previously diagnosed cases documented in routine primary care service registers. Case identification was based on diagnoses made by health personnel in the local primary care system; no additional research-based diagnostic confirmation was performed.
The sample size was calculated during study planning for the planned analysis of associations between PRECEDE-related factors and integrative care use. Based on an anticipated correlation coefficient of 0.193, an alpha level of 0.05, and statistical power of 0.90, the minimum required sample size was 274. After allowing 10% for incomplete data, the target sample was 302 participants. No separate a priori sample-size calculation was performed specifically for cluster analysis. For the segmentation analysis, sample adequacy was considered in relation to the total sample size of 302, the use of 3 clustering variables, the final segment sizes of 63, 145, and 94 participants, and the broadly similar 3-profile structure obtained under alternative k-means starting conditions.
Data Collection and Measures
Data were collected through face-to-face interviews using a structured questionnaire developed from relevant literature and the PRECEDE framework [16]. The instrument included 5 domains: demographic and health characteristics, predisposing factors, enabling factors, reinforcing factors, and integrative care use. Demographic and health variables included age, sex, body mass index, education, occupation, income, and selected symptom-related variables relevant to knee osteoarthritis, including self-reported symptom duration and knee pain score.
The primary outcome was integrative care use, which was assessed with a 25-item instrument using a 3-point response scale. The instrument comprised 3 domains: professional medical care (10 items; score range, 10–30), traditional medicine (9 items; score range, 9–27), and community-based medicine (6 items; score range, 6–18). Higher scores indicated greater engagement in the corresponding domain. Predisposing factors were represented by attitude toward integrative health promotion, which was assessed with 4 items (score range, 4–12). Enabling factors were measured using 14 items covering access, distance, time burden, and decision-making, whereas reinforcing factors were measured using 5 items covering social support and perceived information. Higher scores indicated more favorable conditions. For descriptive interpretation, total scores could be classified as low, moderate, or high according to the original instrument framework; however, these categories were not used to generate the behavioral segments.
The questionnaire was designed to capture the content of each domain rather than a single clinical symptom dimension. Professional medical care items reflected formal health-promoting practices and services, including avoidance of harmful knee-loading behaviors, weight control, diet related to joint health, exercise for the knee and surrounding muscles, and receipt of formal Thai traditional medicine or acupuncture when relevant. Traditional medicine items covered commonly used folk-sector approaches, including massage, herbal compresses or poultices, and herbal remedies. Community-based medicine items reflected self-care and family-linked or community-linked help-seeking.
The attitude items assessed views on the value of integrative health promotion. The enabling domain addressed access, travel distance, time burden, and decision-making, including perceived credibility, safety, effectiveness, and cost-effectiveness of available care options. The reinforcing domain assessed social support and perceived information related to integrative care and knee osteoarthritis.
Content validity was assessed by 3 experts, and items with item-objective congruence values of at least 0.50 were retained. The final questionnaire had an overall item-objective congruence value of 0.95. Reliability was tested in 50 older adults from a nearby district who were not included in the main study. Overall Cronbach’s alpha was 0.77, with subscale values of 0.72 for attitude, 0.72 for enabling factors, 0.74 for reinforcing factors, and 0.71 for integrative care use.
Behavioral Segmentation
In this study, behavioral segmentation referred to empirical grouping of participants according to relatively similar patterns of engagement across professional medical care, traditional medicine, and community-based medicine. Segment formation was based solely on the 3 behavioral domain scores. PRECEDE-related factors and baseline characteristics were not included in cluster formation; instead, they were used only to describe and compare the final segments. Before clustering, the 3 domain scores were standardized as z-scores to ensure comparability across variables and to preserve score variability.
Statistical Analysis
A 2-stage clustering approach was used [18]. First, hierarchical cluster analysis using Ward’s method and squared Euclidean distance was conducted to explore candidate solutions. Second, k-means clustering was used to refine membership and derive the final non-overlapping solution. Two-cluster, 3-cluster, and 4-cluster solutions were compared according to separation across behavioral domains, interpretability, subgroup size, and parsimony. The 3-cluster solution was retained because it preserved a clear graded pattern across the 3 care domains while avoiding both the loss of an intermediate profile in the 2-cluster solution and the additional fragmentation produced by the 4-cluster solution. As a sensitivity check, the clustering procedure was rerun under alternative k-means starting conditions. These reruns yielded a broadly similar 3-profile structure, supporting retention of the 3-cluster solution, although the model should still be interpreted as an internally derived, sample-specific solution.
After final cluster membership was assigned, participant characteristics and PRECEDE-related factors were compared across segments. Categorical variables were analyzed using the chi-square or Fisher exact test, as appropriate. Continuous variables were analyzed using one-way analysis of variance or Kruskal-Wallis tests, depending on data distribution. Tukey post hoc testing was used for significant one-way analysis of variance results. Effect sizes were reported as eta squared for continuous variables and Cramér’s V for categorical variables. All tests were 2-sided, and p-value <0.05 was considered statistically significant. No missing values were observed in the variables included in the present analysis. All analyses were performed using SPSS version 20 (IBM Corp., Armonk, NY, USA).
Ethics Statement
This study was approved by the Human Research Ethics Committee of Thammasat University (project code: 66PU054; approved on 16 July 2023). All participants provided informed consent before data collection.
RESULTS
A total of 302 older adults with primary knee osteoarthritis were included in the analysis. The mean age was 69.93±6.55 years, the mean body mass index was 22.89±3.48 kg/m2, and the mean knee pain score was 1.97±0.66. Participants reported a mean symptom duration of 65.35±41.46 months (median, 60; interquartile range, 36–84). Most participants were female (72.8%), had primary education (86.4%), and reported a monthly household income of 1001–5000 Thai baht (52.6%).
Baseline Characteristics by Segment
Table 1 presents baseline demographic and health characteristics across the 3 behavioral segments. No statistically significant differences were observed among segments for age (p=0.063, η2=0.018), body mass index (p=0.134, η2=0.013), or knee pain score (p=0.207, η2=0.010). Similarly, no significant differences were observed for sex (p=0.592, V=0.059), education (p=0.278, V=0.127), occupation (p=0.083, V=0.166), monthly household income (p=0.562, V=0.090), or daily posture or activity related to knee loading (p=0.297, V=0.140). Overall, the 3 segments were broadly comparable in baseline demographic and health characteristics.
Table 1.
Baseline demographic and health characteristics of participants by behavioral segment
| Characteristics | Total (n=302) | Segment 1: High-engagement group (n=63) | Segment 2: Intermediate-engagement group (n=145) | Segment 3: Low-engagement group (n=94) | p-value1 | Effect size2 |
|---|---|---|---|---|---|---|
| Age (y) | 69.93±6.55 | 70.24±5.76 | 70.63±6.72 | 68.63±6.66 | 0.063 | η2=0.018 |
| Body mass index (kg/m2) | 22.89±3.48 | 22.26±3.41 | 22.84±3.51 | 23.39±3.46 | 0.134 | η2=0.013 |
| Knee pain score | 1.97±0.66 | 1.87±0.68 | 2.03±0.64 | 1.93±0.68 | 0.207 | η2=0.010 |
| Sex | 0.592 | V=0.059 | ||||
| Male | 82 (27.2) | 17 (27.0) | 36 (24.8) | 29 (30.9) | ||
| Female | 220 (72.8) | 46 (73.0) | 109 (75.2) | 65 (69.1) | ||
| Education level | 0.278 | V=0.127 | ||||
| No formal education | 2 (0.7) | 0 (0) | 2 (1.4) | 0 (0) | ||
| Primary education | 261 (86.4) | 52 (82.5) | 127 (87.6) | 82 (87.2) | ||
| Secondary education | 27 (8.9) | 6 (9.5) | 10 (6.9) | 11 (11.7) | ||
| Bachelor’s degree or higher | 12 (4.0) | 5 (7.9) | 6 (4.1) | 1 (1.1) | ||
| Occupation | 0.083 | V=0.166 | ||||
| Farmer | 131 (43.4) | 21 (33.3) | 65 (44.8) | 45 (47.9) | ||
| Daily wage worker | 20 (6.6) | 4 (6.3) | 10 (6.9) | 6 (6.4) | ||
| Merchant/trader | 34 (11.3) | 7 (11.1) | 15 (10.3) | 12 (12.8) | ||
| Retired government officer | 9 (3.0) | 5 (7.9) | 3 (2.1) | 1 (1.1) | ||
| Unemployed/not working | 104 (34.4) | 23 (36.5) | 51 (35.2) | 30 (31.9) | ||
| Other | 4 (1.3) | 3 (4.8) | 1 (0.7) | 0 (0) | ||
| Monthly household income (THB) | 0.562 | V=0.090 | ||||
| ≤1000 | 58 (19.2) | 10 (15.9) | 30 (20.7) | 18 (19.1) | ||
| 1001–5000 | 159 (52.6) | 33 (52.4) | 82 (56.6) | 44 (46.8) | ||
| 5001–10 000 | 59 (19.5) | 13 (20.6) | 24 (16.6) | 22 (23.4) | ||
| >10 000 | 26 (8.6) | 7 (11.1) | 9 (6.2) | 10 (10.6) | ||
| Daily posture or activity related to knee loading | 0.297 | V=0.140 | ||||
| Sitting on the floor with folded legs >1 hr/day | 36 (11.9) | 8 (12.7) | 13 (9.0) | 15 (16.0) | ||
| Squatting >1 hr/day | 63 (20.9) | 14 (22.2) | 29 (20.0) | 20 (21.3) | ||
| Standing >1 hr/day | 88 (29.1) | 16 (25.4) | 41 (28.3) | 31 (33.0) | ||
| Frequent stair climbing or walking on slopes | 46 (15.2) | 9 (14.3) | 29 (20.0) | 8 (8.5) | ||
| Frequent heavy lifting | 29 (9.6) | 8 (12.7) | 16 (11.0) | 5 (5.3) | ||
| Other | 40 (13.2) | 8 (12.7) | 17 (11.7) | 15 (16.0) |
Values are presented as mean±standard deviation for continuous variables and number (%) for categorical variables; Percentages may not total 100 due to rounding.
THB, Thai baht.
For continuous variables were obtained using one-way analysis of variance; For categorical variables were obtained using the chi-square test or Fisher exact test, as appropriate.
Effect sizes are reported as eta squared (η2) for continuous variables and Cramér’s V for categorical variables.
Behavioral Domain Scores by Segment
After the 2-cluster, 3-cluster, and 4-cluster solutions were compared, the 3-cluster model was retained. The 2-cluster model separated participants into broad higher-engagement and lower-engagement groups but did not capture an intermediate profile. The 4-cluster model generated additional subgroups but offered limited additional interpretive value for primary care and community health planning. The 3-cluster model provided a clearer graded profile across the 3 behavioral domains while maintaining adequate subgroup sizes. Reruns using alternative starting conditions showed a broadly similar structure.
Table 2 presents the final 3 behavioral segments. Segment 1 included 63 participants (20.9%), Segment 2 included 145 participants (48.0%), and Segment 3 included 94 participants (31.1%). Scores differed significantly across segments in all 3 domains (all p<0.001), with moderate-to-large effect sizes. Mean professional medical care scores were 24.30±1.60, 20.75±1.79, and 19.12±1.73 for Segments 1, 2, and 3, respectively (η2=0.535). The corresponding scores were 18.87±2.97, 17.67±2.07, and 13.18±1.60 for traditional medicine (η2=0.531) and 14.44±1.45, 12.59±1.50, and 10.94±1.64 for community-based medicine (η2=0.400). Tukey post hoc tests confirmed a graded pattern across all domains, with Segment 1 scoring highest, followed by Segment 2 and Segment 3.
Table 2.
Integrative care use domain scores by behavioral segment1
| Behavioral domains | Total (n=302) | Segment 1: High-engagement group (n=63) | Segment 2: Intermediate-engagement group (n=145) | Segment 3: Low-engagement group (n=94) | p-value | Effect size (η2) | Post hoc |
|---|---|---|---|---|---|---|---|
| Professional medical care | 21.20±2.71 | 24.30±1.60a | 20.75±1.79b | 19.12±1.73c | <0.001 | 0.535 | S1>S2>S3 |
| Traditional medicine | 16.48±3.17 | 18.87±2.97a | 17.67±2.07b | 13.18±1.60c | <0.001 | 0.531 | S1>S2>S3 |
| Community-based medicine | 12.43±1.92 | 14.44±1.45a | 12.59±1.50b | 10.94±1.64c | <0.001 | 0.400 | S1>S2>S3 |
Values are presented as mean±standard deviation; Cluster derivation was based on standardized domain scores.
Superscript letters indicate the results of Tukey post hoc comparisons. Within each row, means sharing at least one superscript letter are not significantly different, whereas means with no superscript letter in common differ significantly at p<0.05.
PRECEDE-related Factors by Segment
Table 3 compares PRECEDE-related factors across segments. Among predisposing factors, attitude differed significantly across segments but had a small effect size (p=0.017, η2=0.027); Segment 1 had the highest mean score, and Segment 3 had the lowest. In contrast, perceived information did not differ significantly across segments (p=0.581, η2=0.004).
Table 3.
PRECEDE-related factor scores by behavioral segment1
| PRECEDE-related factors | Total (n=302) | Segment 1: High-engagement group (n=63) | Segment 2: Intermediate-engagement group (n=145) | Segment 3: Low-engagement group (n=94) | p-value2 | Effect size (η2)3 | Post hoc |
|---|---|---|---|---|---|---|---|
| Predisposing factors | |||||||
| Attitude | 10.07±1.14 | 10.37±1.10a | 10.10±1.10a,b | 9.84±1.19b | 0.017 | 0.027 | S1>S3 |
| Enabling factors | |||||||
| Total enabling score | 25.79±2.67 | 27.25±2.02a | 25.74±2.28b | 24.77±2.94c | <0.001 | 0.115 | S1>S2>S3 |
| Access | 11.43±1.88 | 12.56±1.42a | 11.48±1.64b | 10.70±2.15c | <0.001 | 0.121 | S1>S2>S3 |
| Decision-making | 14.31±1.58 | 14.70±1.35a | 14.26±1.51a,b | 14.06±1.79b | 0.046 | 0.020 | S1>S3 |
| Reinforcing factors | |||||||
| Total reinforcing score | 11.80±1.51 | 12.33±1.27a | 11.80±1.47b | 11.47±1.67b | 0.002 | 0.041 | S1>S2=S3 |
| Social support | 7.60±1.08 | 8.03±0.97a | 7.59±1.05b | 7.30±1.05b | <0.001 | 0.057 | S1>S2=S3 |
| Perceived information | 4.22±0.79 | 4.30±0.73 | 4.21±0.73 | 4.17±0.90 | 0.581 | 0.004 | NS |
Values are presented as mean±standard deviation.
NS, not significant.
Superscript letters (a, b, c) indicate the results of Tukey post hoc comparisons. Within each row, means sharing at least one superscript letter are not significantly different, whereas means with no superscript letter in common differ significantly at p<0.05.
Using one-way analysis of variance.
Effect sizes are reported as eta squared (η2).
More pronounced differences were observed for enabling factors. The total enabling score differed significantly across segments (p<0.001, η2=0.115), with a graded pattern from Segment 1 to Segment 3. Access showed a similar gradient and had 1 of the largest effect sizes among the PRECEDE-related variables (p<0.001, η2=0.121). Decision-making also differed significantly, although the effect size was small (p=0.046, η2= 0.020).
Reinforcing factors showed a similar pattern. The total reinforcing score differed significantly across segments (p=0.002, η2=0.041), and social support also differed significantly (p<0.001, η2=0.057), with Segment 1 scoring highest and Segment 3 lowest. Overall, differences among segments were more pronounced for enabling and reinforcing factors, particularly access and social support, than for perceived information alone.
DISCUSSION
This study identified 3 distinct behavioral segments among older adults with knee osteoarthritis in a rural Thai community. Older adults in this population did not use available care in the same way; instead, engagement differed clearly across professional medical care, traditional medicine, and community-based medicine. These findings suggest that audience segmentation may be useful in public health practice, especially when a population is not behaviorally uniform [14]. This approach is particularly relevant in local settings with limited resources, where intervention priorities must often be carefully selected [15]. The findings also align with the PRECEDE framework, which conceptualizes health behavior as being shaped by predisposing, enabling, and reinforcing conditions [16]. In practical terms, recognizing these segment differences may help guide communication strategies and intervention planning [17].
In the study setting, the 3 care domains should be interpreted within the local rural service ecology rather than as isolated categories. Professional medical care mainly refers to services delivered through the district hospital, primary care units, and formal Thai traditional medicine services within the public health system. Traditional medicine and folk-sector care may include locally rooted practices, such as herbal use, massage, and other community-based healing approaches. Community-based care includes self-care, family support, older adults’ groups, and support embedded in everyday community life. This interpretation is consistent with the broader concept of community-level integrated care for older adults [12].
Another important finding was that segment differences were more evident for enabling and reinforcing conditions than for perceived information. In this sample, perceived information did not differ significantly across segments, whereas attitude, total enabling score, access, decision-making, total reinforcing score, and social support did differ significantly. This pattern suggests that practical and social conditions distinguished segment membership more clearly than perceived information alone. Although previous evidence suggests that self-management education can improve self-efficacy in knee osteoarthritis [19] and that social support is related to self-management behaviors [20], the present cross-sectional findings should be interpreted cautiously as differences in associated conditions, not as evidence that access or social support directly shaped care behavior over time.
This interpretation is relevant to primary care because uptake of recommended osteoarthritis care is influenced by service accessibility and care organization [21]. Primary care-based models of care also emphasize continuity, follow-up, and support for long-term engagement [22]. Structured service models, including the PARTNER model, have been developed to improve osteoarthritis care delivery in primary care [23]. Trial evidence further suggests that redesigned service delivery can improve care processes and some outcomes [24]. Recent synthesis also indicates that implementation strategies may promote uptake of osteoarthritis practice guidelines in routine care [25].
These findings are particularly relevant in Thailand, where community-integrated care and village health volunteers contribute to long-term care for older adults [13]. Qualitative evidence from Thailand has also emphasized the importance of culturally appropriate and context-sensitive service design in community care [26]. In addition, village health volunteers contribute to home care, social support, community surveillance, and chronic disease management in rural Thailand [27–29]. These observations suggest that segment-informed support could be incorporated into existing primary care and community health structures rather than requiring entirely new programs. Because the study area is located in Northern Thailand, local cultural beliefs and regional healing traditions may also shape care-related behavior. Previous work has shown that Lanna local wisdom remains relevant to health promotion and well-being among older adults in Northern Thailand [30]. However, the present study did not operationalize a Lanna health culture model as a formal analytic framework; therefore, this concept should be treated as contextual background rather than as a directly measured explanatory model.
From a practical perspective, the 3 identified segments may help guide tailored support strategies. The high-engagement group may benefit most from support that preserves continuity and reinforces ongoing care use. The intermediate-engagement group may benefit from more structured support to improve consistency and strengthen confidence in care choices [31]. The low-engagement group may require more proactive, low-threshold, and outreach-oriented approaches. Evidence from digital self-management [32], exercise-based interventions and patient beliefs [33], person-based digital tools [34], and community pharmacist-led support [35] suggests that engagement may be influenced by service design, usability, beliefs, and practical support. As summarized in Table 4, these implications should be interpreted as theory-informed, practice-oriented applications of the observed segment profiles rather than as tested intervention effects.
Table 4.
Theoretical interpretation and practice implications of the identified behavioral segments
| Segments | Behavioral profile | Theoretical interpretation | Priority for support | Practice implication for primary care and community health |
|---|---|---|---|---|
| Segment 1: High-engagement group | High engagement across professional medical care, traditional medicine, and community-based medicine | This segment appears to represent a maintenance-oriented profile in which sustained engagement is supported by relatively favorable enabling and reinforcing conditions [16] | Preserve continuity | This group is likely to benefit most from support that preserves continuity rather than from initial activation; This interpretation is consistent with implementation evidence showing that recommended osteoarthritis care requires active support to be sustained in routine practice [25] |
| Segment 2: Intermediate-engagement group | Moderate engagement across all 3 care domains | This segment appears to reflect a transitional profile in which engagement is present but not yet fully stable or consolidated [14] | Strengthen consistency and navigation | This group may benefit from structured support that improves consistency of participation, strengthens confidence in care choices, and helps patients navigate available services; This interpretation is supported by qualitative primary care evidence showing that implementation of structured osteoarthritis care depends on context-sensitive delivery and service organization [31] |
| Segment 3: Low-engagement group | Low engagement across professional medical care, traditional medicine, and community-based medicine | This segment appears to represent an access-constrained, low-engagement profile in which limited participation is more likely to reflect weaker enabling and reinforcing conditions than lack of information alone [16] | Reduce barriers and support entry into care | This group may require more proactive, low-threshold, and outreach-oriented support; Evidence from digital self-management suggests that structured support can help sustain engagement over time [32], while mixed-method evidence indicates that patient beliefs influence participation in exercise-based osteoarthritis care [33]; Community-linked educational and medication-review support also shows potential to improve pain-related outcomes outside specialist settings [35] |
The interpretations presented in this table are theory-informed and practice-oriented rather than intervention-tested for the current sample; They are intended to support interpretation of segment profiles and to guide tailored support planning in primary care and community settings [34].
This study has several strengths. It moved beyond a variable-centered approach by identifying empirically derived behavioral segments and was conducted in a real-world Thai community where multiple forms of care coexist. Several limitations should also be considered. The cross-sectional design precludes causal inference, the segments may be sample-dependent, and the study was conducted in a single district. Cases were identified from routine primary care records rather than through a study-specific diagnostic protocol, and the detailed diagnostic criteria used in routine service practice were not independently reverified. More detailed clinical information, such as radiographic severity, laterality, and functional grading, was not available. Data on ethnic composition and local geographic access conditions were also unavailable. Finally, cluster stability was not tested in an independent external sample, and temporal ordering between PRECEDE-related factors and segment membership could not be established.
Older adults with knee osteoarthritis in this rural Thai community were not behaviorally homogeneous in their use of professional medical care, traditional medicine, and community-based care. The identified segments differed more clearly in enabling and reinforcing conditions, particularly access and social support, than in perceived information. Behavioral segmentation may provide a practical basis for planning more tailored and context-sensitive support within primary care and community health systems. Future studies should evaluate whether segment-informed approaches improve service engagement, self-management, and health outcomes across different settings.
Footnotes
Data Availability
The datasets generated and/or analyzed during the current study are not publicly available because of participant confidentiality but are available from the corresponding author on reasonable request.
Conflict of Interest
The authors have no conflicts of interest associated with the material presented in this paper.
Funding
None.
Acknowledgements
The authors thank the participants and local health personnel in Thoen District, Lampang Province, for their cooperation and support during the study.
Author Contributions
Conceptualization: Jiraratsatit K. Data curation: Ubolkam W. Formal analysis: Jiraratsatit K. Funding acquisition: None. Methodology: Jiraratsatit K. Writing – original draft: Jiraratsatit K. Writing – review & editing: Jiraratsatit K, Ubolkam W, Sumpowthong K.
REFERENCES
- 1.World Health Organization Ageing and health. 2025. [cited 2026 Mar 1]. Available from: https://www.who.int/news-room/fact-sheets/detail/ageing-and-health.
- 2.World Health Organization Musculoskeletal health. 2022. [cited 2026 Mar 1]. Available from: https://www.who.int/news-room/fact-sheets/detail/musculoskeletal-conditions.
- 3.GBD 2021 Osteoarthritis Collaborators Global, regional, and national burden of osteoarthritis, 1990–2020 and projections to 2050: a systematic analysis for the Global Burden of Disease Study 2021. Lancet Rheumatol. 2023;5(9):e508–e522. doi: 10.1016/S2665-9913(23)00163-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.World Health Organization Osteoarthritis. 2023. [cited 2026 Mar 1]. Available from: https://www.who.int/news-room/fact-sheets/detail/osteoarthritis.
- 5.Vanitcharoenkul E, Adulkasem N, Unnanuntana A, Chotiyarnwong P. Symptomatic knee osteoarthritis in older Thai adults: prevalence, risk factors, and associations with osteoporosis, sarcopenia, and fall risk-a nationwide cross-sectional study. BMC Public Health. 2026;26(1):1603. doi: 10.1186/s12889-026-27158-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Kolasinski SL, Neogi T, Hochberg MC, Oatis C, Guyatt G, Block J, et al. 2019 American College of Rheumatology/Arthritis Foundation guideline for the management of osteoarthritis of the hand, hip, and knee. Arthritis Care Res (Hoboken) 2020;72(2):149–162. doi: 10.1002/acr.24131. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Bannuru RR, Osani MC, Vaysbrot EE, Arden NK, Bennell K, Bierma-Zeinstra SM, et al. OARSI guidelines for the non-surgical management of knee, hip, and polyarticular osteoarthritis. Osteoarthritis Cartilage. 2019;27(11):1578–1589. doi: 10.1016/j.joca.2019.06.011. [DOI] [PubMed] [Google Scholar]
- 8.Nik Shafii NA, Yaacob LH, Ishak A, Kadir AA. Traditional and complementary medicine use in knee osteoarthritis and its associated factors among patients in Northeast Peninsular Malaysia. Oman Med J. 2018;33(2):148–153. doi: 10.5001/omj.2018.27. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.World Health Organization Regional Office for the Western Pacific Thailand health system review. 2024. [cited 2026 Mar 1]. Available from: https://iris.who.int/handle/10665/376912.
- 10.Lampang Provincial Public Health Office . Inspection and supervision report, Health Region 1, Lampang Province, round 2, fiscal year 2022. Lampang: Lampang Provincial Public Health Office; 2022. (Thai) [Google Scholar]
- 11.World Health Organization Primary health care. 2025. [cited 2026 Mar 1]. Available from: https://www.who.int/news-room/fact-sheets/detail/primary-health-care.
- 12.World Health Organization Integrated care for older people: guidelines on community-level interventions to manage declines in intrinsic capacity. 2017. [cited 2026 Mar 1]. Available from: https://www.who.int/publications/i/item/9789241550109. [PubMed]
- 13.Aung MN, Moolphate S, Aung TN, Koyanagi Y, Kurusattra A, Chantaraksa S, et al. Effectiveness of a community-integrated intermediary care (CIIC) service model to enhance family-based long-term care for Thai older adults in Chiang Mai, Thailand: a cluster-randomized controlled trial TCTR2019041. Health Res Policy Syst. 2004;2022;20(Suppl 1):110. doi: 10.1186/s12961-022-00911-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Slater MD. Theory and method in health audience segmentation. J Health Commun. 1996;1(3):267–283. doi: 10.1080/108107396128059. [DOI] [PubMed] [Google Scholar]
- 15.Slater MD, Kelly KJ, Thackeray R. Segmentation on a shoestring: health audience segmentation in limited-budget and local social marketing interventions. Health Promot Pract. 2006;7(2):170–173. doi: 10.1177/1524839906286616. [DOI] [PubMed] [Google Scholar]
- 16.Kim J, Jang J, Kim B, Lee KH. Effect of the PRECEDE-PROCEED model on health programs: a systematic review and meta-analysis. Syst Rev. 2022;11(1):213. doi: 10.1186/s13643-022-02092-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Noar SM. An audience-channel-message-evaluation (ACME) framework for health communication campaigns. Health Promot Pract. 2012;13(4):481–488. doi: 10.1177/1524839910386901. [DOI] [PubMed] [Google Scholar]
- 18.Hair JF, Jr, Black WC, Babin BJ, Anderson RE. Multivariate data analysis. 8th ed. Andover: Cengage Learning; 2019. pp. 216–218. [Google Scholar]
- 19.Uritani D, Koda H, Sugita S. Effects of self-management education programmes on self-efficacy for osteoarthritis of the knee: a systematic review of randomised controlled trials. BMC Musculoskelet Disord. 2021;22(1):515. doi: 10.1186/s12891-021-04399-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Chen YY, Weng LC, Li YT, Huang HL. Mediating effect of self-efficacy on the relationship between social support and self-management behaviors among patients with knee osteoarthritis: a cross-sectional study. BMC Geriatr. 2022;22(1):635. doi: 10.1186/s12877-022-03331-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Wallis JA, Ackerman IN, Brusco NK, Kemp JL, Sherwood J, Young K, et al. Barriers and enablers to uptake of a contemporary guideline-based management program for hip and knee osteoarthritis: a qualitative study. Osteoarthr Cartil Open. 2020;2(4):100095. doi: 10.1016/j.ocarto.2020.100095. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Cunningham J, Doyle F, Ryan JM, Clyne B, Cadogan C, Cottrell E, et al. Primary care-based models of care for osteoarthritis; a scoping review. Semin Arthritis Rheum. 2023;61:152221. doi: 10.1016/j.semarthrit.2023.152221. [DOI] [PubMed] [Google Scholar]
- 23.Egerton T, Hinman RS, Hunter DJ, Bowden JL, Nicolson PJ, Atkins L, et al. PARTNER: a service delivery model to implement optimal primary care management of people with knee osteoarthritis: description of development. BMJ Open. 2020;10(10) doi: 10.1136/bmjopen-2020-040423.e040423 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Hunter DJ, Bowden JL, Hinman RS, Egerton T, Briggs AM, Bunker SJ, et al. Effectiveness of a new service delivery model for management of knee osteoarthritis in primary care: a cluster randomized controlled trial. Arthritis Care Res (Hoboken) 2023;75(6):1320–1332. doi: 10.1002/acr.25037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Ramirez MM, Fillipo R, Allen KD, Nelson AE, Skalla LA, Drake CD, et al. Use of implementation strategies to promote the uptake of knee osteoarthritis practice guidelines and improve patient outcomes: a systematic review. Arthritis Care Res (Hoboken) 2024;76(9):1246–1259. doi: 10.1002/acr.25353. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Aung TN, Lorga T, Moolphate S, Koyanagi Y, Angkurawaranon C, Supakankunti S, et al. Towards cultural adequacy of experience-based design: a qualitative evaluation of community-integrated intermediary care to enhance the family-based long-term care for Thai older adults. Healthcare (Basel) 2023;11(15):2217. doi: 10.3390/healthcare11152217. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Pagaiya N, Noree T, Hongthong P, Gongkulawat K, Padungson P, Setheetham D. From village health volunteers to paid care givers: the optimal mix for a multidisciplinary home health care workforce in rural Thailand. Hum Resour Health. 2021;19(1):2. doi: 10.1186/s12960-020-00542-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Kaweenuttayanon N, Pattanarattanamolee R, Sorncha N, Nakahara S. Community surveillance of COVID-19 by village health volunteers, Thailand. Bull World Health Organ. 2021;99(5):393–397. doi: 10.2471/BLT.20.274308. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Theeranut A, Methakanjanasak N, Lertsinudom S, Surit P, Panyaek N, Leeladapattarakul S, et al. Integrated care model by the village health volunteers to prevent and slow down progression of chronic kidney disease in a rural community, Thailand. J Prim Care Community Health. 2024;15:21501319241240355. doi: 10.1177/21501319241240355. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Chanprasit C, Lirtmunlikaporn S, Kaewthummanukul T, Jongrungrotsakul W, Sawasdisingha P. Lanna local wisdom in promoting the elderly well-being: a pilot study. Nurs J CMU. 2020;47(1):185–197. (Thai. [Google Scholar]
- 31.Frost H, Tooman T, Cowie J, Gillespie N, Ackerman P, Krievs E, et al. Advanced practice physiotherapists and the implementation of the JIGSAW-E model for the management of osteoarthritis in Scottish primary care settings: a qualitative case study. Physiotherapy. 2022;117:81–88. doi: 10.1016/j.physio.2022.08.007. [DOI] [PubMed] [Google Scholar]
- 32.Safari R, Jackson J, Sheffield D. Digital self-management interventions for people with osteoarthritis: systematic review with meta-analysis. J Med Internet Res. 2020;22(7):e15365. doi: 10.2196/15365. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Hurley M, Dickson K, Hallett R, Grant R, Hauari H, Walsh N, et al. Exercise interventions and patient beliefs for people with hip, knee or hip and knee osteoarthritis: a mixed methods review. Cochrane Database Syst Rev. 2018;4(4):CD010842. doi: 10.1002/14651858.CD010842.pub2. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Stevenson RD, Chowdhury EA, Inza VB, Western MJ, Walsh NE, Jones SL, et al. Development of the intelligent knee osteoarthritis lifestyle app: a person-based approach. BMC Musculoskelet Disord. 2024;25(1):189. doi: 10.1186/s12891-024-07313-4. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.Thapa P, Kc B, Gyawali S, Leong SL, Mohamed Ibrahim MI, Lee SW. Effectiveness of community pharmacist-led interventions in osteoarthritis pain management: a cluster-randomized trial. Res Social Adm Pharm. 2024;20(2):149–156. doi: 10.1016/j.sapharm.2023.10.012. [DOI] [PubMed] [Google Scholar]