Latent profiles of advance care planning engagement among Chinese older adults

Abstract

Background

Advance Care Planning (ACP) enables older adults to document future care preferences while cognitively competent. In China, ACP adoption remains limited despite aging population pressures, associated with cultural taboos around death, family-centered decision norms, and socioeconomic disparities.

Research Gaps

Existing research predominantly employs variable-centered approaches, potentially masking qualitatively distinct subgroups with different engagement patterns. Person-centered analytical methods are needed to identify heterogeneous ACP engagement profiles and inform targeted interventions.

Objective

This study identified distinct latent profiles of ACP engagement among Chinese older adults using knowledge, attitudes, beliefs, and motivation as LPA indicators within the Knowledge-Attitude-Practice framework and examined sociodemographic and health-related factors associated with profile membership; ACP intention was treated as a distal outcome.

Method

A cross-sectional survey was conducted among 1,776 community-dwelling older adults (aged ≥ 60) across multiple Chinese provinces between January and March 2025. Knowledge, attitudes, beliefs, and motivation were entered as LPA indicators; intention was excluded from profile formation and examined as a distal continuous outcome via BCH analysis. R3STEP multinomial logistic regression examined factors associated with profile membership.

Results

Three distinct profiles emerged: The Disengaged Group (33%), characterized by uniformly low scores across all four indicators; The Motivated Pioneers Group (26%), displaying above-average beliefs and motivation alongside below-average attitudes; and The Broadly Engaged Group (41%), exhibiting the highest scores across all indicators. Medical insurance type, hospitalization frequency, and bereavement experience were significantly associated with profile membership. BCH-adjusted intention differed significantly across profiles (Broadly Engaged: M = 19.44; Motivated Pioneers: M = 18.83; Disengaged: M = 15.66).

Conclusions

Chinese older adults exhibit heterogeneous ACP engagement patterns associated with distinct sociodemographic and experiential factors, supporting profile-tailored interventions: foundational ACP education for The Disengaged Group, decision aids and family communication scripts for the Motivated Pioneers, and system-level facilitation for The Broadly Engaged Group.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12889-026-27528-5.

Introduction

The accelerating demographic shift toward population aging has repositioned end‑of‑life decision‑making as a core health‑systems imperative. In the terminal phase, many older adults experience clinically significant cognitive decline that undermines decisional capacity and complicates time‑critical choices regarding resuscitation and other invasive interventions [1]. Within this context, Advance Care Planning (ACP) functions as a normatively grounded, procedurally specified pathway through which individuals, while cognitively competent, can formulate, communicate, and document preferences for future care, including designation of surrogate decision‑makers and completion of Advance Directives [2]. When implemented with fidelity, ACP has been associated with preserving patient autonomy, alleviating decisional burden for families, and curtailing nonbeneficial treatment, leading international bodies to identify it as integral to high‑quality end‑of‑life care [3, 4].

Comparative evidence nevertheless reveals substantial cross‑national heterogeneity in ACP uptake and routinization. High‑income Western systems have institutionalized ACP through coordinated legislation, reimbursement, clinician training, and public education, yielding comparatively higher documentation rates [5, 6]. Many Asian settings, by contrast, are structured by family‑centered decision practices and physician authority, foregrounding relational autonomy over liberal individualism; coupled with procedural inconsistency and limited standardization, these dynamics may constrain diffusion and everyday use [5, 7]. The resulting divergence underscores the need for culturally responsive implementation strategies that negotiate normative pluralism rather than presume a single autonomy model.

China exemplifies these tensions with particular clarity. Despite rapid health-system reforms, ACP literacy remains low among patients and community-dwelling older adults, including those who report a preference for autonomous decision-making [8]. This knowledge–willingness paradox [9] appears to be associated with filial piety scripts and enduring taboos around death-related talk, which tend to be linked with preferences for life-sustaining interventions and reduced anticipatory dialogue [10]. Structural disparities in resources, educational opportunities, and institutional supports may further limit routinization across regions, potentially contributing to inequities in access to informed end-of-life planning.

To conceptualize these multilevel determinants, this study employs the Knowledge–Attitude–Practice (KAP) model as its primary theoretical framework. KAP articulates a sequential framework in which ACP informational mastery is theorized to shape evaluative orientations that, in turn, are posited to condition behavioral enactment [11]. In the present study, these three components are operationalized as ACP knowledge (Knowledge), psychological readiness and attitudinal beliefs toward ACP (Attitude), and behavioral engagement intention (Practice), respectively. This sequential logic provides a structured lens for examining how cognitive foundations are associated with affective appraisals, which in turn correspond to behavioral readiness. To enrich the interpretation of attitude-intention pathways within this framework, we draw upon constructs from the Theory of Planned Behavior (TPB), which specifies how perceived social norms and self-efficacy dimensions of perceived behavioral control may modulate the translation of attitudes into intentions [12]. Perceived social norms in this study reflect the influence of familial expectations and cultural scripts around death-related conversations, while perceived behavioral control is indexed by individuals’ confidence in their capacity to initiate and complete ACP-related tasks. While KAP organizes the overarching conceptual architecture, TPB-derived constructs serve as auxiliary interpretive resources that help illuminate the social and self-efficacy mechanisms through which attitudes are enacted within family-centered normative environments characteristic of Chinese contexts.

A parallel methodological consideration concerns how heterogeneity in ACP engagement is represented analytically. Throughout this paper, ACP engagement is used as an umbrella term referring to the multidimensional process encompassing knowledge acquisition, psychological readiness, and behavioral intentions; psychological readiness denotes the affective and evaluative orientation toward ACP; and behavioral intention refers specifically to the expressed willingness to undertake concrete ACP actions. The prevailing variable-centered tradition in Chinese ACP research estimates average associations under assumptions of population homogeneity [13]. Such designs are informative about mean trends yet may obscure qualitatively distinct subgroups with fundamentally different engagement profiles. A person-centered alternative using Latent Profile Analysis (LPA) directly addresses this limitation by identifying discrete engagement patterns on the basis of response configurations across KAP-informed indicators. This approach furnishes an empirical basis for tailoring interventions to profile-specific needs rather than assuming uniform responses across all participants.

Guided by this integrative framework, this research addressed the following question: What distinct profiles of ACP engagement exist among Chinese older adults based on their knowledge, psychological readiness, attitudinal beliefs, and behavioral intentions, and what sociodemographic and health-related factors are associated with profile membership and ACP behavioral intention? To address this question, the study first examined sociodemographic and health-related correlates of ACP knowledge, psychological readiness, and behavioral intention; it then employed Latent Profile Analysis as the central person-centered analytical contribution to identify and characterize discrete engagement patterns; and finally tested predictors of latent profile membership while comparing intention levels across profiles. These analyses collectively aim to advance understanding of ACP engagement within China’s heterogeneous healthcare environment and to provide an empirical foundation for developing targeted, equity-oriented interventions.

Methodology

Participants, setting and design

We implemented a sequential, two-phase quantitative design. Stage I comprised a 500-participant pilot to evaluate the psychometric performance of study instruments and calibrate the main-study sample size; Stage II entailed primary data collection using the validated measures. Instrument validation followed Taherdoost’s framework [14], incorporating expert review for content validity, exploratory factor analysis to examine internal structure, and Cronbach’s alpha to assess internal consistency.

Recruitment used multi-stage convenience sampling across 27 Chinese provinces between January and March 2025, with sites concentrated primarily in cities along China’s eastern coastline. We established partnerships with community health centers, senior activity centers, and residential care facilities to support systematic participant identification. Trained liaisons disseminated study materials with survey links and QR codes. Targeted WeChat outreach to older-adult networks restricted access to individuals aged ≥ 60 years.

Eligibility criteria included age ≥ 60 [15], capacity to comprehend and complete the instruments [16], and willingness to provide electronic informed consent [17]. We excluded individuals with diagnosed severe cognitive impairment or psychiatric disorders [16], acute conditions precluding participation, or refusal to consent.

Measures

Instrument selection was theory-driven by the Knowledge-Attitude-Practice framework [11], which conceptualizes health decision-making as an integrated interplay of cognition, attitudes, and behavioral intention. Measures comprised the Advance Care Planning Knowledge Questionnaire (ACPKQ) [18], adapted from the original scale developed by Lai et al. [19], for knowledge assessment (example items: “Have you heard of advance care planning?” “Have you heard of advance palliative care decision-making?“); the Advance Care Planning Readiness Questionnaire (ACPRQ) [20] for psychological readiness evaluation, comprising three subscales (Attitudes [10 items], Beliefs [5 items], and Motivation [7 items]), rated on a 5-point Likert scale (1–5) (example item: “I do not believe that advance care planning is meaningful to me”); and the ACP Intention Scale (ACPIS) [21] for behavioral intention measurement (example item: “I will consider discussing advance care planning in the future”). Negatively worded items within the Attitudes subscale were reverse-scored prior to analysis; all other subscale items required no reversal. Higher Beliefs scores reflect a more ACP-supportive orientation, greater endorsement of personal planning and autonomous decision-making, whereas lower scores indicate stronger adherence to traditional, family-centered approaches to end-of-life care. Across all instruments, higher scores indicate more favorable ACP-related outcomes. Psychometric characteristics appear in Table 1; English translations are provided in Supplement File 1. To preserve validated properties, we retained the original item wording and factor structure of each instrument and implemented reverse scoring at the data-processing stage in accordance with the published scoring guidelines.

Table 1.

Validated questionnaire instruments employed

Instrument	Construct Measured	Description and Scoring	Original Study Psychometric Properties	Current Study Psychometric Properties
General Information Questionnaire (Researcher-developed)	Demographic and background information	A 16-item questionnaire across five dimensions: demographics (6 items), socioeconomic status (3 items), health status and medical experience (4 items), care experience and end-of-life contact (2 items), and ACP participation willingness (1 item). Rated on various scales including a 4-point scale for willingness	Not applicable	Not applicable
Advance Care Planning Knowledge Questionnaire	ACP knowledge	A 6-item subscale adapted from the knowledge dimension was rated on a 3-point Likert scale. While the subscale contained an open-ended question about information sources, this item was used for qualitative assessment only and excluded from scoring. The remaining five items yielded total scores ranging from 5–15, with higher scores indicating better knowledge of ACP concepts. Score interpretation: 5–8 (low knowledge), 9–12 (moderate knowledge), 13–15 (high knowledge).	Cronbach’s α = 0.841; test-retest reliability = 0.823	Cronbach’s α = 0.873; CFA fit indices: χ²/df = 2.156, GFI = 0.943, AGFI = 0.892, CFI = 0.951, TLI = 0.927, RMSEA = 0.068; single-factor structure confirmed
Advance Care Planning Readiness Questionnaire	ACP psychological readiness	A 22-item scale measuring three dimensions: attitudes (10 items), beliefs (5 items), and motivation (7 items). Rated on a 5-point Likert scale (1–5). Total scores range from 22–110, with higher scores indicating greater psychological readiness for ACP participation. Score interpretation: 22–51 (low readiness), 52–81 (moderate readiness), 82–110 (high readiness)	Cronbach’s α: 0.900 (attitudes), 0.880 (beliefs), 0.835 (motivation), 0.923 (overall)	Cronbach’s α = 0.916; CFA fit indices: χ²/df = 2.287, GFI = 0.935, AGFI = 0.883, CFI = 0.944, TLI = 0.919, RMSEA = 0.071; three-factor structure confirmed with factor loadings > 0.60
Advance Care Planning Intention Scale	ACP behavioral intention	A 5-item subscale from the intention dimension, rated on a 5-point Likert scale (1–5). Total scores range from 5–25, with higher scores indicating stronger intention to participate in ACP. Score interpretation: 5–11 (low intention), 12–18 (moderate intention), 19–25 (high intention)	Cronbach’s α = 0.885; split-half reliability = 0.867; test-retest reliability = 0.834	Cronbach’s α = 0.894; CFA fit indices: χ²/df = 2.193, GFI = 0.938, AGFI = 0.877, CFI = 0.947, TLI = 0.923, RMSEA = 0.069; single-factor model demonstrated excellent fit

Sample size and data collection

Pilot estimates informed Stage II sample-size planning using Cochran’s procedure [22]: n = (Z·σ/E)², with Z = 1.96 (95% confidence), σ = 10.274 (maximum pilot standard deviation), and E = 0.05·σ, yielding 1,537. Allowing 10% attrition set a target of 1,691; the realized sample of 1,776 exceeded this threshold and was adequate for planned analyses.

Data collection was administered via Questionnaire Star under a pre-specified data quality plan. The interface was optimized for older adults through simplified navigation, enlarged typography, and streamlined item presentation. Recognizing heterogeneous digital literacy, protocols permitted family assistance under standardized instructions while safeguarding response validity. The online instrument presented study aims, procedures, potential benefits and risks, confidentiality protections, and withdrawal rights; informed consent was documented electronically after review. All materials underwent pre-launch usability testing and cognitive pretesting.

Two independent researchers implemented a sequential screening protocol based on completion time and response pattern quality. Of 2,284 submissions, 208 were excluded for completion time < 240 s [23]. The remaining 2,076 cases underwent further response quality assessment, resulting in the exclusion of an additional 300 cases. Of these, 54 demonstrated highly invariant response patterns consistent with straightlining [24]. The other 246 cases were identified through a composite quality review that combined distributional screening (> 2 SD from the sample mean) with item-level evaluation of within-person response variability; cases were excluded only where both criteria jointly indicated non-substantive engagement. The final analytic sample comprised 1,776 participants (effective response rate = 77.76%). Figure 1 depicts the screening cascade. To evaluate the influence of the 246-case exclusion, a sensitivity analysis repeated the latent profile analysis on the sample retaining these cases (n = 2,022), applying identical model specifications.

Fig. 1.

The Process of Data Exclusion

Data analysis

Data analysis was conducted using IBM SPSS Statistics v.26 (IBM Corp., Armonk, NY, USA) and Mplus v.8.7 (Muthén & Muthén, Los Angeles, CA, USA). Following recommendations by Field [25], we tested univariate and multivariate assumptions to identify potential sources of bias. Descriptive statistics and correlation analyses were used to characterize participants’ demographic features.

For latent profile analysis conducted in Mplus 8.7, subscale total scores from the four instruments (knowledge, attitudes, beliefs, and motivation) were entered as continuous profile indicators, with ACP intention excluded from profile formation to prevent outcome leakage. Models were estimated using the Robust Maximum Likelihood (MLR) estimator, which accounts for non-normality, and missing data were handled using Full Information Maximum Likelihood (FIML). To identify the global maximum likelihood and avoid local maxima, models were estimated using 500 random sets of start values with 20 final stage optimizations, and convergence was verified for all models. Multiple model solutions were evaluated to determine the optimal number of profiles. Model selection was based on established fit criteria [26]: lower AIC, BIC, and aBIC values indicating better fit, entropy values ≥ 0.80 demonstrating adequate classification accuracy, and significant LMR and BLRT tests (p < 0.05) confirming that the k-profile solution provides superior fit compared to the (k-1)-profile alternative. Classification quality was further assessed using average posterior probabilities (AvePP), with values > 0.70 indicating well-separated classes.

To identify sociodemographic factors associated with latent profile membership, we employed the 3-step approach (R3STEP) in Mplus [27]. Covariates were entered as continuous variables; Mplus R3STEP automatically estimates all pairwise latent class comparisons without requiring a pre-specified reference category. Additionally, to examine differences in ACP intention across the identified profiles, the BCH method (Distal Outcome Analysis) was utilized [27]. This approach tests for the equality of means across latent classes while correcting for classification error, treating intention as a distal continuous outcome. Statistical significance was set at p < 0.05.

Ethical approval and consent

This study was conducted in accordance with the Declaration of Helsinki [17] and was approved by the Institutional Review Board of Zhejiang Chinese Medical University (approval number: 20250120-1). Electronic informed consent procedures were implemented consistent with minimal-risk survey research conducted entirely through online platforms.

Results

Common method bias test

Harman’s single-factor test yielded a first unrotated factor accounting for 28.81% of the variance, below the 40% threshold, indicating that common method bias was not a substantial concern.

Demographic data and univariate analysis

Of the 1,776 respondents, most were women and 60–79 years old; the majority were married and living with a spouse or children. Nearly half reported primary schooling or less, and roughly two fifths had monthly income of 1,001–3,000 RMB. ACP knowledge was significantly associated with age, marital status, number of children, living situation, education level, monthly income, type of medical insurance, self-rated health status, self-rated disease status, number of hospitalizations after age 60, experience of caring for friends or family in hospital, and death of friends or family in the past five years (p < 0.05). Psychological readiness was associated with gender, age, marital status, number of children, education level, monthly income, type of medical insurance, self-rated health status, self-rated disease status, and number of hospitalizations after age 60 (p < 0.05). ACP intention was associated with education level, monthly income, type of medical insurance, religious beliefs, and self-rated disease status (p < 0.05). Full estimates are reported in Table 2, with least significant difference (LSD) post hoc contrasts, applied conditional on significant omnibus F-tests, in Supplement File 2.

Table 2.

Demographic data and univariate analysis (n = 1776)

Variable	Frequency (%)	Knowledge	t/F	p	Readiness	t/F	p	Intention	t/F	p
Gender
Male	473 (26.63)	9.86 ± 2.06	-0.38	0.70	71.71 ± 6.52	4.58	<0.01**	18.01 ± 2.26	-0.02	0.98
Female	1303 (73.37)	9.90 ± 2.09			70.20 ± 6.03			18.02 ± 2.19
Age
60–69 years	851 (47.92)	10.14 ± 2.14	11.57	<0.01**	71.07 ± 6.46	10.45	<0.01**	18.13 ± 2.18	2.22	0.11
70–79 years	712 (40.09)	9.68 ± 1.99			69.80 ± 5.85			17.92 ± 2.21
≥ 80 years	213 (11.99)	9.62 ± 2.04			71.42 ± 6.00			17.89 ± 2.32
Marital Status
Married	1324 (74.55)	9.97 ± 2.09	23.93	<0.01**	70.78 ± 6.09	5.59	<0.01**	18.08 ± 2.19	2.55	0.08
Divorced or Widowed	405 (22.80)	9.45 ± 1.90			70.35 ± 6.48			17.87 ± 2.27
Single	47 (2.65)	11.45 ± 2.25			67.83 ± 6.10			17.51 ± 2.32
Number of Children
≤ 1 child	546 (30.74)	10.29 ± 2.15	22.42	<0.01**	71.31 ± 5.87	5.14	<0.01**	18.11 ± 2.18	2.01	0.14
2 children	719 (40.48)	9.90 ± 2.15			70.25 ± 6.31			18.06 ± 2.20
≥ 3 children	511 (28.77)	9.45 ± 1.79			70.35 ± 6.32			17.85 ± 2.26
Living Situation
Living alone	322 (18.13)	9.69 ± 2.08	3.86	0.02*	70.66 ± 7.27	2.01	0.14	18.13 ± 2.31	0.55	0.58
Living with spouse or children	1400 (78.83)	9.91 ± 2.07			70.53 ± 5.93			17.99 ± 2.18
Living with others	54 (3.04)	10.50 ± 2.14			72.24 ± 6.00			18.04 ± 2.27
Education Level
Primary school or below	865 (48.70)	9.41 ± 1.83	37.31	<0.01**	69.78 ± 5.92	12.95	<0.01**	17.82 ± 2.19	4.62	<0.01**
Middle school	436 (24.55)	10.13 ± 2.14			70.80 ± 6.44			18.27 ± 2.19
High school/Technical secondary school	221 (12.44)	10.26 ± 2.37			72.22 ± 6.59			18.17 ± 2.20
Associate degree or above	254 (14.30)	10.78 ± 2.09			71.66 ± 5.93			18.09 ± 2.25
Monthly Income
≤ 1000 RMB	336 (18.92)	9.21 ± 1.69	16.79	<0.01**	70.04 ± 6.55	4.36	<0.01**	17.55 ± 2.28	3.92	<0.01**
1001 ~ 3000 RMB	749 (42.17)	9.77 ± 2.04			70.10 ± 5.67			18.13 ± 2.17
3001 ~ 6000 RMB	440 (24.77)	10.40 ± 2.11			71.35 ± 6.65			18.13 ± 2.16
6001 ~ 9000 RMB	130 (7.32)	10.08 ± 2.39			71.55 ± 5.87			18.18 ± 2.19
9001 ~ 12,000 RMB	61 (3.43)	10.10 ± 2.02			70.72 ± 6.79			17.84 ± 2.21
>12,000 RMB	60 (3.38)	10.85 ± 2.32			72.32 ± 6.34			18.17 ± 2.41
Type of Medical Insurance
Urban Employee Medical Insurance	418 (23.54)	10.19 ± 2.17	4.74	<0.01**	72.27 ± 6.12	10.95	<0.01**	18.22 ± 2.08	6.783	<0.01**
Urban Resident Medical Insurance	1115 (62.78)	9.84 ± 2.07			70.15 ± 5.97			18.08 ± 2.21
Commercial Medical Insurance	23 (1.29)	10.57 ± 2.35			70.74 ± 6.48			17.74 ± 2.88
Self-funded	172 (9.68)	9.48 ± 1.79			69.41 ± 7.05			17.26 ± 2.22
Public-funded	48 (2.70)	9.67 ± 2.14			70.85 ± 6.09			17.60 ± 2.45
Religious Beliefs
Buddhism	410 (23.08)	9.85 ± 2.01	0.84	0.50	70.45 ± 5.73	0.42	0.80	18.15 ± 2.12	2.80	0.03*
Catholicism	47 (2.65)	9.72 ± 1.47			69.87 ± 6.82			17.74 ± 2.08
Islam	3 (0.17)	10.67 ± 2.89			69.67 ± 4.04			14.33 ± 1.16
Taoism and Others	31 (1.75)	10.48 ± 1.77			71.45 ± 5.27			17.74 ± 2.27
None	1285 (72.35)	9.89 ± 2.12			70.66 ± 6.34			18.00 ± 2.23
Self-assessment of health status
Not ideal	119 (6.70)	9.66 ± 1.91	5.88	<0.01**	71.37 ± 6.53	2.91	0.03*	17.81 ± 2.39	1.86	0.14
Less ideal	309 (17.40)	9.53 ± 1.92			71.14 ± 6.53			17.83 ± 2.17
More ideal	1166 (65.65)	9.95 ± 2.10			70.29 ± 5.83			18.10 ± 2.19
Ideal	182 (10.25)	10.26 ± 2.25			71.19 ± 7.49			17.90 ± 2.24
Self-assessment of disease status
Mild	823 (46.34)	9.86 ± 2.08	7.01	<0.01**	70.32 ± 5.61	3.20	0.02*	18.14 ± 2.14	3.42	0.02*
Moderate	370 (20.83)	9.63 ± 1.87			71.08 ± 6.43			17.70 ± 2.33
Severe	65 (3.66)	9.34 ± 1.95			72.40 ± 6.62			18.14 ± 2.11
None	518 (29.17)	10.19 ± 2.20			70.48 ± 6.80			18.02 ± 2.23
Number of hospitalizations after age 60
0	680 (38.29)	10.04 ± 2.12	5.812	<0.01**	70.79 ± 6.33	4.459	<0.01**	18.13 ± 2.14	1.459	0.21
1	444 (25.00)	10.03 ± 2.12			69.65 ± 6.01			17.97 ± 2.26
2–3	477 (26.86)	9.77 ± 2.01			71.12 ± 6.07			17.93 ± 2.26
4–5	103 (5.80)	9.26 ± 1.78			71.55 ± 6.71			18.14 ± 2.01
≥ 6	72 (4.05)	9.28 ± 2.00			69.93 ± 5.54			17.60 ± 2.48
Rescue experience after age 60
Yes	156 (8.78)	9.90 ± 2.04	0.09	0.93	71.27 ± 6.85	1.41	0.16	18.10 ± 2.18	0.52	0.60
No	1620 (91.22)	9.89 ± 2.08			70.54 ± 6.13			18.01 ± 2.21
Experience of caring for friends or family in hospital in the past five years
Yes	781 (43.98)	10.00 ± 2.08	1.94	0.05*	70.83 ± 6.30	1.36	0.18	18.06 ± 2.20	0.77	0.44
No	995 (56.02)	9.81 ± 2.08			70.43 ± 6.12			17.98 ± 2.22
Death of friends or family in the past five years
Yes	1055 (59.40)	9.79 ± 2.05	-2.41	0.02*	70.67 ± 6.16	0.58	0.57	18.07 ± 2.18	1.28	0.20
No	721 (40.60)	10.03 ± 2.11			70.50 ± 6.25			17.93 ± 2.25

*p<0.05,**p<0.01. Values are presented as mean ± SD. Negatively worded items were reverse coded so that higher scores indicate more favorable ACP knowledge, readiness, and intention. Group differences were examined using independent-samples t tests or one-way ANOVA, with least significant difference (LSD) post hoc tests for variables with more than two categories. Knowledge was measured by Advance Care Planning Knowledge Questionnaire (ACPKQ), Readiness was measured by Advance Care Planning Readiness Questionnaire (ACPRQ), and Intention was measured by Advance Care Planning Intention Scale (ACPIS)

Descriptive statistics and correlations analysis

Scale scores indicated moderate ACP knowledge, generally favorable psychological readiness, and moderate intention; internal consistency coefficients were acceptable to high. All pairwise correlations among the five study dimensions were positive and significant at p < 0.01 (Supplement File 3). Knowledge showed small but significant positive associations with attitudes, beliefs, motivation, and intention. Among the readiness-related dimensions, motivation demonstrated the strongest association with intention (r = 0.77), followed by beliefs (r = 0.58) and attitudes (r = 0.29).

Latent profile analysis

Fit indices (AIC, BIC, aBIC, LMR, BLRT) and class size were evaluated across one- through five-class solutions (Table 3). AIC, BIC, and aBIC decrements leveled off after the three-class solution; four- and five-class models, while statistically significant on LMR and BLRT (p < 0.01), yielded classes comprising less than 10% and less than 5% of the sample, respectively. The three-profile solution was retained, with entropy = 0.82, minimum AvePP = 0.85, and class proportions of 26%–41%. The sensitivity analysis (n = 2,022) identified three profiles as optimal under the same specifications, with membership proportions and indicator patterns closely paralleling the primary solution.

Table 3.

Latent profile analysis (n = 1776)

Model	AIC	BIC	aBIC	Entropy	Min AvePP	P value		Categorical Probability
						LMR	BLRT
1	35747.77	35791.63	35766.21
2	33683.35	33754.62	33713.32	0.93	0.97	< 0.01**	< 0.01**	0.35/0.65
3	33405.11	33503.78	33446.60	0.82	0.85	< 0.01**	< 0.01**	0.33/0.26/0.41
4	33138.42	33264.51	33191.44	0.83	0.80	< 0.01**	< 0.01**	0.31/0.38/0.09/0.22
5	32880.24	33033.74	32944.78	0.92	0.83	< 0.01**	< 0.01**	0.28/0.52/0.08/0.10/0.02

**p<0.01

AIC Akaike information criterion, BIC Bayesian information criterion, aBIC sample-size–adjusted BIC, Entropy classification precision index, Min AvePP Minimum average posterior class-membership probability, LMR Lo–Mendell–Rubin likelihood ratio test, BLRT Bootstrap likelihood ratio test, Categorical Probability average posterior probability for each latent class

To facilitate cross-profile comparison, class-specific means were standardized against the full-sample mean and standard deviation for each indicator (Fig. 2). For the beliefs subscale, higher scores reflect a more ACP-supportive orientation favoring autonomous planning, whereas lower scores indicate stronger adherence to traditional family-centered decision-making. Class 1, The Disengaged Group (33%), exhibited negative Z-scores across all four indicators (knowledge Z = -0.17, attitudes Z = -0.35, beliefs Z = -1.02, motivation Z = -1.25), reflecting uniformly low engagement with the most pronounced deficits in beliefs and motivation. Class 3, The Broadly Engaged Group (41%), showed the highest standardized scores across all four dimensions (knowledge Z = 0.19, attitudes Z = 0.90, beliefs Z = 0.59, motivation Z = 0.73); as the largest subgroup, this profile reflects a broadly favorable cognitive-affective orientation toward ACP, yet favorable dispositions may still require institutional facilitation to translate into concrete planning actions. Class 2, The Motivated Pioneers Group (26%), presented a distinctive cross-dimensional configuration: below-average attitudes (Z = -0.96) coexisted with above-average beliefs (Z = 0.40) and motivation (Z = 0.47), while knowledge remained near the sample mean (Z = -0.07), suggesting a pragmatically oriented subgroup whose engagement is driven by perceived benefits and behavioral readiness rather than by global attitudinal favorability.

Fig. 2.

Latent profile analysis folding chart (n = 1776)

Profile Validation

R3STEP auxiliary analysis

Among Class 1 (The Disengaged Group) vs. Class 3 (The Broadly Engaged Group) comparisons, type of medical insurance, number of hospitalizations after age 60, and death of friends or family in the past five years were significantly associated with profile membership. For Class 2 (The Motivated Pioneers Group) vs. Class 3 (The Broadly Engaged Group), gender, self-assessment of disease status, experience of caring for friends or family in hospital, and death of friends or family in the past five years were significant. For Class 1 vs. Class 2, gender, type of medical insurance, and rescue experience after age 60 were significant (Table 4).

Table 4.

Multinomial logistic R3STEP analysis

Items	B	Standard Error	OR	95% Confidence Interval	p Value
#C1 vs. C3
Type of Medical Insurance	0.21	0.07	1.24	[1.08, 1.41]	< 0.01**
Number of hospitalizations after age 60	0.17	0.07	1.19	[1.04, 1.35]	0.01*
Death of friends or family in the past five years	0.33	0.13	1.39	[1.08, 1.80]	0.01*
#C2 vs. C3
Gender	-0.61	0.18	0.55	[0.38, 0.78]	< 0.01**
Self-assessment of disease status	0.17	0.06	1.18	[1.04, 1.34]	0.01*
Experience of caring for friends or family in hospital in the past five years	-0.41	0.16	0.67	[0.48, 0.92]	0.01*
Death of friends or family in the past five years	0.38	0.17	1.46	[1.05, 2.01]	0.02*
※C1 vs. C2
Gender	-0.36	0.17	0.70	[0.50, 0.96]	0.03*
Type of Medical Insurance	-0.20	0.08	0.82	[0.70, 0.97]	0.02*
Rescue experience after age 60	-0.69	0.28	0.50	[0.29, 0.86]	0.01*

C1 is The Disengaged Group, C2 is The Motivated Pioneers Group, C3 is The Broadly Engaged Group. #C3 as the reference group; ※C1 as the reference group. OR Odds Ratio, CI Confidence Interval. 95% CIs are presented on the odds ratio scale, computed from the Confidence Intervals of Odds Ratios section of the Mplus R3STEP output. Results are derived from R3STEP multinomial logistic regression. *p < 0.05; **p < 0.01

BCH analysis of distal outcomes

BCH-adjusted ACP intention differed significantly across profiles (Wald χ² = 1961.8, p < 0.01). Class 3 (The Broadly Engaged Group) showed the highest adjusted mean (19.44), followed by Class 2 (The Motivated Pioneers Group; 18.83) and Class 1 (The Disengaged Group; 15.66), with all pairwise differences significant (p < 0.01; Tables 5 and 6).

Table 5.

BCH-adjusted means of ACP intention

Latent Profile	n (%)	ACP Intention Mean (SE)	95% CI	Knowledge (M ± SD)	Attitudes (M ± SD)	Beliefs (M ± SD)	Motivation (M ± SD)	Intention (M ± SD)
Class 1	596 (33%)	15.66 (0.07)	[15.52, 15.80]	9.53 ± 1.98	30.79 ± 3.35	15.30 ± 1.45	21.64 ± 1.41	15.77 ± 1.69
Class 2	458 (26%)	18.83 (0.10)	[18.64, 19.02]	9.75 ± 2.06	27.55 ± 3.49	18.67 ± 1.70	26.96 ± 1.44	18.81 ± 1.70
Class 3	722 (41%)	19.44 (0.05)	[19.34, 19.54]	10.28 ± 2.12	37.31 ± 3.08	19.11 ± 1.70	27.79 ± 1.37	19.38 ± 1.18

C1 is The Disengaged Group, C2 is The Motivated Pioneers Group, C3 is The Broadly Engaged Group. SE Standard Error, M Mean, SD Standard Deviation, CI Confidence Interval. ACP Intention Means are adjusted using the BCH method. Negatively worded items in the Attitude dimension were reverse-scored prior to calculation

Table 6.

BCH comparisons of ACP intention

Comparison	Statistic / mean diff. (SE)	p Value
Overall Test (χ²)	1961.8	< 0.01**
C2 vs. C1	3.17 (0.12)	< 0.01**
C3 vs. C1	3.78 (0.09)	< 0.01**
C3 vs. C2	0.61 (0.12)	< 0.01**

C1 is The Disengaged Group, C2 is The Motivated Pioneers Group, C3 is The Broadly Engaged Group. SE Standard Error. The Overall Test statistic represents the Wald Chi-Square value (χ²) testing equality of means across all profiles. Pairwise comparisons report the mean difference between profiles. *p < 0.05; **p < 0.01

Discussion

This study not only characterizes the overall level of advance care planning (ACP) engagement intention among Chinese older adults but, more importantly, employs latent profile analysis (LPA) to reveal the heterogeneity underlying ACP participation. The three profile solution is consistent with the view that ACP engagement may not be a monolithic individual choice and may instead reflect social location and cultural norms [9]. Educational attainment, income level, and medical insurance type are significantly associated with ACP knowledge, psychological readiness, and behavioral intention, which suggests potential pathways through which social structural factors may be related to health literacy and the capacity to navigate complex healthcare systems in ways that may shape the motivational foundations of planning engagement [28]. The positive correlations observed among knowledge, attitudes, beliefs, motivation, and intention suggest an integrated cognitive affective system in which favorable evaluations and perceived benefits of ACP are associated with stronger willingness to participate [29]. These findings highlight the potential importance of adopting an equity oriented and culturally responsive approach to ACP implementation in the stratified healthcare landscape in China.

Regarding social structural determinants, the univariate analyses reveal consistent associations between socioeconomic status (SES) indicators and ACP engagement dimensions. Higher educational attainment is associated with greater ACP knowledge, more favorable psychological readiness and stronger behavioral intention. Within the Chinese healthcare context, education may be associated with enhanced health literacy that in turn may facilitate comprehension of the ethical and legal implications of ACP and may reduce cognitive barriers to deliberating sensitive end of life scenarios [28]. Higher monthly income is linked to elevated knowledge and readiness scores, which may reflect greater access to informational resources, continuity of care, and reduced financial anxiety that together may lower the perceived risk of engaging in anticipatory planning [28]. Medical insurance type emerges as a significant correlate across all three dimensions, and holders of Urban Employee Medical Insurance demonstrate the highest scores. In the absence of comprehensive reimbursement mechanisms for palliative care, the financial protection conferred by insurance may serve as a psychological foundation that is associated with a greater willingness among older adults to contemplate end of life planning, not only as a payment mechanism but also as a source of institutional trust and perceived system navigability that may help normalize participation in structured decision processes [30].

The correlation analyses identify positive associations among attitudes, beliefs, motivation, and intention, which warrants careful interpretation within the Knowledge Attitude Practice (KAP) framework [11]. Higher attitude scores are correlated with stronger ACP intention, in line with the proposition of the Theory of Planned Behavior (TPB) that evaluative orientations toward a behavior are proximally associated with behavioral intentions [12]. The beliefs dimension, which captures respondents’ perceived benefits of ACP, is also positively associated with intention. Higher belief scores suggest that older adults who perceive greater value in ACP, for example in terms of reducing family burden, ensuring respect for personal wishes, and maintaining family harmony, tend to exhibit stronger willingness to engage [31]. This pattern invites a cultural reinterpretation. Within Chinese contexts that are informed by Confucian traditions, elevated belief scores may reflect an altruistic framing of ACP in which planning is understood as an act of filial consideration that protects one’s children from decisional burden and preserves family harmony [32], rather than as an assertion of individual autonomy grounded in Western notions of the self. The association between beliefs and intention therefore can be interpreted as being consistent with the idea that traditional filial piety (xiao) may be undergoing a contemporary transformation from a norm that discourages discussion of death toward a norm that frames planning as a way to protect family members. This reconceptualized form of filial piety may represent a potentially important cultural consideration in efforts to promote ACP adoption in China, in that it aligns advance planning with deeply held relational values rather than positioning it in tension with them.

The Broadly Engaged Group (Class 3, 41%) exhibited the highest standardized scores across all four indicators, with attitudes (Z = 0.90) and motivation (Z = 0.73) particularly elevated. As the largest subgroup, this profile represents a cognitively and affectively favorable orientation toward ACP. The BCH-adjusted intention means confirm this pattern, with The Broadly Engaged Group demonstrating the highest scores (M = 19.44), followed by the Motivated Pioneers (M = 18.83) and the Disengaged (M = 15.66). ACPIS scores reflect expressed willingness to engage in ACP-related activities; cross-profile intention differentials are therefore best interpreted as indicators of motivational orientation toward planning rather than as evidence of enacted planning behavior. Nevertheless, favorable dispositions do not automatically translate into completed advance directives. Within the Chinese context, the cultural discourse of avoiding open discussion of death (yin hui wen hua) [33] and the absence of standardized ACP communication protocols [34] may constrain this group’s translation of readiness into action, suggesting that system-level facilitation rather than further educational intervention represents the priority for this profile. This ordering of intention scores is consistent with the importance of affective and evaluative dimensions in shaping willingness to plan.

The Disengaged Group (Class 1, 33%) exhibits uniformly low scores across all dimensions and can be characterized as representing a dual deficit in cognitive understanding and motivational readiness. The R3STEP analysis shows that there are associations in which less comprehensive medical insurance coverage and greater hospitalization frequency are associated with membership in this disengaged profile, relative to The Broadly Engaged Group, which suggests that accumulated negative healthcare experiences and institutional precarity are associated with disengagement from anticipatory planning [35]. The Motivated Pioneers Group (Class 2, 26%), by contrast, displayed above-average beliefs (Z = 0.40) and motivation (Z = 0.47) alongside below-average attitudes (Z = -0.96), representing a subgroup whose engagement is driven by perceived practical benefits of ACP and behavioral readiness rather than by global attitudinal favorability. Analyses using the R3STEP procedure further identify that male gender and recent experience of the death of a family member are associated with higher odds of membership in this pragmatically oriented profile relative to The Broadly Engaged Group. Women may instead gravitate toward broadly engaged orientations, potentially reflecting caregiving experiences that foster multidimensional ACP readiness [36], while bereavement may heighten planning awareness following proximate encounters with mortality [37]. The tripartite profile structure poses challenges for intervention approaches that assume that a single undifferentiated strategy is sufficient and suggests that effective promotion of ACP may require differentiated strategies that are calibrated to each subgroup’s particular configuration of cognitive, attitudinal, and motivational characteristics.

Taken together, these findings refine understanding of ACP engagement among Chinese older adults by suggesting that participation patterns may be jointly shaped by structural position and cultural meaning making. The consistent positive associations among KAP dimensions, particularly the facilitating role of beliefs framed as perceived benefits, imply that ACP implementation in China may be more acceptable when it is aligned with traditional family oriented values. The observed associations between medical insurance, perceived feasibility of end of life planning, and engagement further point to the potential relevance of institutional arrangements as part of the broader context in which planning decisions are considered.

Practical implications

This study provides empirical evidence for developing stratified, person-centered ACP implementation strategies among Chinese older adults. A dual-pathway approach is warranted: structural interventions addressing systemic inequalities in education, medical insurance, and healthcare access, combined with differentiated behavioral interventions tailored to the three engagement profiles identified through latent profile analysis. Community health workers should receive training to initiate ACP discussions using simplified decision aids and the “Teach-Back” method during routine chronic disease management visits [38]. ACP conversations should be systematically embedded into insurance-covered annual health examinations to enhance accessibility while reducing financial barriers [39]. These implementation strategies align with the UN Decade of Healthy Ageing framework’s vision of person-centered integrated care for older populations [40].

The disengaged group (33%)

This profile, characterized by the lowest scores across knowledge, attitudes, beliefs, and motivation, likely includes older adults with limited education and weaker social support networks. Interventions should prioritize foundational ACP education and cultural desensitization rather than direct ACP decision-making discussions. Trusted community health workers or family physicians can introduce ACP within the broader context of “healthy aging” rather than explicitly framing it as end-of-life planning. Visual aids and plain-language materials should be employed to address basic misconceptions and alleviate fear [41]. Cultural reframing initiatives should position ACP as a mechanism for reducing family decision-making burden and fulfilling filial responsibility, rather than emphasizing individual autonomy [42]. Measurable outcomes include improved mean scores on ACP knowledge assessments and increased participation rates in community-based educational activities.

The motivated pioneers group (26%)

This profile displayed above-average beliefs and motivation alongside below-average attitudes, indicating a pragmatically oriented subgroup whose main barrier lies in attitudinal readiness. Interventions should focus on bridging this attitudinal gap through structured empowerment strategies. Healthcare providers should integrate multimedia educational materials and decision aids into routine medical encounters to accommodate varying health literacy levels [43]. To address the cultural barrier of initiating end-of-life conversations within Chinese families, communication scripts and family facilitation sessions can be provided [44], teaching older adults how to frame ACP as “reducing the burden on children”—thereby leveraging Confucian values to promote rather than impede dialogue. Measurable outcomes include elevated ACPIS scores and increased rates of formal healthcare proxy designation.

The Broadly Engaged Group (41%)

This profile exhibited the highest scores across all four indicators, indicating substantial cognitive and affective readiness for ACP engagement. For this group, the intervention priority shifts from education to action facilitation. Healthcare systems should establish dedicated ACP consultation services or streamlined pathways within community health centers to assist with advance directive completion. Electronic health record systems should integrate standardized ACP preference documentation to ensure accessibility during critical care transitions and emergency situations, thereby enhancing trust in the healthcare system and reinforcing self-efficacy [45]. Measurable outcomes include increased percentages of completed and formally documented ACP preferences within medical record systems.

Limitations and prospects

Several limitations should be considered when interpreting these findings: (1) Reliance on self-report measures may introduce common method variance, though Harman’s single-factor test suggested this risk was limited; (2) The online sampling format may have systematically underrepresented individuals with limited digital literacy, those in the oldest age cohorts, socioeconomically disadvantaged groups, and rural populations with restricted digital access; (3) Recruitment spanning 27 provinces introduces potential clustering effects that were not modeled in the primary analysis, as within-province homogeneity may influence standard error estimates; (4) The cross-sectional design precludes causal inferences regarding the relationships among study variables. (5) The LSD post hoc procedure, although applied conditional on significant omnibus F-tests, is a comparatively liberal correction method, and pairwise comparisons should be interpreted with appropriate caution; additionally, nominal-level covariates were entered as numerically coded continuous variables in the R3STEP analysis, implicitly assuming ordinality, and future studies should consider dummy-coded specifications to relax this assumption. Future research prospects include: (1) implementing longitudinal designs to capture the dynamic nature of ACP attitudes and behaviors across the aging trajectory and examine causal relationships between identified determinants; (2) employing mixed-methods approaches to deepen understanding of cultural nuances and family dynamics influencing ACP decision-making in Chinese contexts.

Conclusion

This study identified three heterogeneous ACP engagement profiles among community-dwelling Chinese older adults: The Disengaged Group (33%), The Motivated Pioneers Group (26%), and The Broadly Engaged Group (41%). Knowledge demonstrated small but significant positive associations with readiness dimensions, and motivation showed the strongest association with intention, suggesting an integrated yet dimensionally distinct motivational system. Profile membership was associated with medical insurance type, hospitalization frequency, and bereavement experiences. ACP intention differed significantly across profiles, with The Broadly Engaged Group demonstrating the highest BCH-adjusted means; these observed differences may inform profile-tailored interventions: ACP education and cultural reframing for The Disengaged Group, decision aids and family communication tools for the Motivated Pioneers, and system-level facilitation for The Broadly Engaged Group. Future longitudinal research is needed to examine dynamic ACP engagement trajectories and the potential mechanisms through which structural determinants may relate to planning behavior in diverse Chinese populations.

Authors’ contributions

Wenyu Yue: Conceptualization, Methodology, Software, Formal analysis, Investigation, Data curation, Writing - original draft; Yihui Lin: Conceptualization, Methodology, Investigation, Data curation, Writing - original draft; Qing Chu: Investigation, Data curation, Validation; Xiaoqin Ma: Conceptualization, Supervision, Project administration, Writing - review & editing.

Funding

This study was supported by (1) 2025 General Projects under the Soft Science Research Program of the Department of Science and Techology of Zhejiang Province (Grant No. 2025C35037); (2) 2025 Zhejiang Chinese Medical University Cultivation Plan for Top Innovative Talents of Postgraduates (No. 2025YJSBJ005).

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

All procedures involving human participants were conducted in accordance with the Declaration of Helsinki. This study received ethical approval from the Ethics Committee of Zhejiang Chinese Medical University (Approval number: 20250120-1).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.