Latent Profiles of Early Personal Mastery and Their Association with Out-of-Hospital Rehabilitation Exercise Adherence One Month After Hip Arthroplasty: A Single-Centre Prospective Observational Study

Abstract

Purpose

In this single-centre prospective observational study, we aimed to identify latent profiles of early personal mastery in patients after hip arthroplasty, examine factors associated with profile membership, and explore differences in out-of-hospital rehabilitation exercise adherence across profiles.

Patients and Methods

A total of 350 patients who underwent primary hip arthroplasty at a tertiary hospital in Urumqi, Xinjiang, from August to December 2024, were recruited by convenience sampling; 17 were lost to follow-up, and 333 completed the 1-month assessment and were included in the analysis. Personal mastery was assessed on postoperative day 1–2 with the Personal Mastery Scale, together with a general information questionnaire; adherence to out-of-hospital rehabilitation exercise was assessed at 1 month postoperatively with the Functional Exercise Adherence Scale for Orthopedic Patients. Latent profile analysis (LPA) was used to identify profiles of personal mastery, and multinomial logistic regression to examine factors associated with profile membership. Differences in adherence across profiles were analysed using the Bolck–Croon–Hagenaars (BCH) approach to account for classification uncertainty, and were additionally examined in linear models adjusted for demographic and clinical covariates.

Results

Of the 350 patients recruited, 17 were lost to follow-up, giving a 1-month follow-up retention rate of 95.14%. LPA identified three profiles of early personal mastery: low (31.53%), intermediate (32.13%) and high (36.34%) personal mastery. In the multinomial logistic regression, age, gender, education level, employment status and chronic comorbidities were independently associated with profile membership (p < 0.05), whereas place of residence was not. Adherence at 1 month differed significantly across profiles (χ² = 81.398, p < 0.001), and profile membership remained independently associated with adherence after adjustment for demographic and clinical covariates (total adherence score F(2, 307) = 23.23, p < 0.001). The intermediate and high personal mastery profiles did not differ significantly from each other on the learning dimension, the physical dimension or the total adherence score.

Conclusion

Early personal mastery in hip arthroplasty patients is heterogeneous and is associated with out-of-hospital rehabilitation exercise adherence at 1 month. These findings suggest that interventions tailored to patients’ mastery profiles warrant investigation, although causal relationships remain to be established.

Introduction

Artificial hip arthroplasty involves the use of specialised artificial materials to replace a damaged or diseased hip joint and improve hip function.^1,2 By 2030, the number of primary total hip arthroplasties performed in the United States is projected to reach 635,000,³ and comparable growth is projected in other countries,⁴ reflecting the ageing of the population and the rising burden of hip disease. Favourable functional outcomes have been reported up to 1 year after surgery for hip fracture.⁵ However, the success of surgery does not mark the end of treatment: patients need to continue rehabilitation exercises after surgery to promote recovery.⁶ During postoperative rehabilitation, patients often experience psychological distress such as anxiety and depression, limited self-care capacity, and uncertainty about functional recovery. These factors negatively affect health-promoting behaviours and rehabilitation outcomes.⁷

Social Cognitive Theory (SCT) provides a foundational framework for understanding health behaviour. According to this theory, motivation and behaviour are regulated by forethought and by self-efficacy beliefs. Self-efficacy is a task-specific cognitive judgement referring to an individual’s confidence in their ability to perform a particular behaviour, such as adhering to rehabilitation exercises.^8,9 Personal mastery, by contrast, is a higher-order and relatively stable generalised belief reflecting an individual’s overarching cognitive framework regarding their capacity to influence the outcomes of life events.^10,11

Personal mastery can be conceptualised as a distal psychological resource that provides a general orientation for how individuals interpret and respond to health challenges,¹¹ whereas self-efficacy functions as a proximal mechanism directly involved in the initiation and maintenance of specific behaviours.⁹ Within this framework, personal mastery may be associated with behavioural outcomes through its influence on self-efficacy beliefs in specific contexts.

Compared with task-specific self-efficacy, personal mastery—the focus of this study—captures more stable differences in patients’ psychological resources after surgery, and is therefore well suited to identifying subgroups of patients who share a similar psychological profile. Although the Personal Mastery Scale is unidimensional, person-centred methods such as latent profile analysis can identify subgroups of patients who share a similar configuration of item responses and, potentially, similar support needs, which a single total score does not convey. On this basis, we examined the patterns of association between distinct personal mastery profiles and rehabilitation adherence, providing a higher-order psychological framework for understanding individual differences. Evidence on latent profiles of early personal mastery in relation to out-of-hospital rehabilitation exercise adherence after hip arthroplasty remains limited, and the present prospective study is intended to enrich that evidence base.

In this study, latent profile analysis (LPA) was used to identify subgroups of patients with similar levels of early personal mastery, and the Bolck–Croon–Hagenaars (BCH) approach was used to compare out-of-hospital rehabilitation exercise adherence across these subgroups while accounting for classification uncertainty. Clarifying the relationship between early personal mastery and rehabilitation adherence may help inform future supportive strategies aimed at strengthening patients’ sense of control during recovery.

Aims and Hypotheses

This study aimed to: (1) identify the latent profiles of early personal mastery in patients after hip arthroplasty; (2) analyse the sociodemographic and clinical factors associated with profile membership; and (3) compare out-of-hospital rehabilitation exercise adherence across these profiles.

We hypothesised that: (H1) early personal mastery in patients after hip arthroplasty would manifest as distinct latent profiles; (H2) these profiles would be associated with sociodemographic and clinical characteristics such as gender, education level and place of residence; and (H3) the profiles would be associated with different levels of out-of-hospital rehabilitation exercise adherence at 1 month postoperatively. These hypotheses were exploratory.

Materials and Methods

Sample Size

The sample size was estimated using Kendall’s rule of thumb, which recommends 10–20 participants per candidate variable. Eighteen variables were considered: the 14 sociodemographic and clinical variables listed in Table 1, the Personal Mastery Scale score, and the three dimension scores of the Functional Exercise Adherence Scale for Orthopedic Patients. Using the more conservative ratio of 10 participants per variable and allowing for an anticipated 20% attrition, the minimum required sample size was 18 × 10 ÷ (1 − 0.20) = 225 participants. A total of 333 patients were ultimately included, which exceeded this minimum. For the latent profile analysis, all three retained profiles contained more than 100 participants (105, 107 and 121) and the smallest accounted for 31.53% of the sample, well above the 5% threshold prespecified for profile stability.

Table 1.

General Information of Respondents (n=333)

Item	Category	n	%
Age	≤60 years	150	45.05
	>60 years	183	54.95
Gender	Male	144	43.24
	Female	189	56.76
Education level	Primary school and below	110	33.03
	Junior high school	108	32.43
	High school and technical secondary school	44	13.21
	College degree and above	71	21.32
Place of residence	Urban	258	77.48
	Rural	75	22.52
Marital status	Married	265	79.58
	Single	19	5.71
	Divorced/widowed	49	14.71
Employment status	Employed	136	40.84
	Unemployed	197	59.16
Chronic comorbidities	Yes	267	80.18
	No	66	19.82
Living arrangement	Living alone	46	13.81
	Living with family	287	86.19
Per capita monthly family income (CNY)	<1000	41	12.31
	1000–2999	71	21.32
	3000–4999	123	36.94
	>5000	98	29.43
Payment method	Medical insurance	321	96.40
	Out-of-pocket	12	3.60
Anaesthesia grade	Level I	6	1.80
	Level II	230	69.07
	Level III	97	29.13
Machine-assisted surgery	Yes	73	21.92
	No	260	78.08
Preoperative diagnosis	Avascular necrosis of femoral head	155	46.55
	Femoral fracture	112	33.63
	Arthritis	17	5.11
	Developmental dysplasia of the hip	33	9.91
	Other primary hip disease	16	4.80
Surgical procedure	Total hip arthroplasty	264	79.28
	Hemiarthroplasty	69	20.72

Notes: 1. Anaesthesia grade was recorded using the American Society of Anesthesiologists (ASA) physical status classification (the “anesthesia grade” field of the inpatient medical-record front page), captured as grades I–III here (I = normal healthy; II = mild systemic disease; III = severe systemic disease that is not incapacitating). 2. Machine-assisted surgery was defined as the intraoperative use of computer navigation or a robotic-assisted system, as distinct from conventional freehand technique.

Participants

A total of 350 patients who underwent primary hip arthroplasty at a tertiary hospital in Urumqi, Xinjiang, from August to December 2024 were recruited by convenience sampling, of whom 333 completed the 1-month assessment and were included in the analysis.

Inclusion criteria

(1) underwent primary hip arthroplasty for avascular necrosis of the femoral head, femoral fracture, hip osteoarthritis, developmental dysplasia of the hip or other primary hip disease; (2) age ≥ 18 years; (3) clear consciousness and normal language expression ability; (4) agreed to participate and provided written informed consent.

Exclusion criteria

(1) severe cerebrovascular accident (stroke) or malignant tumour, or severe cognitive impairment preventing completion of the survey; (2) hearing impairment affecting communication. Withdrawal of consent during follow-up, and inability to contact a participant after three consecutive telephone attempts, were treated as loss to follow-up.

Patients received the usual postoperative care of the ward. Mobilisation began on postoperative day 1–2, assisted by the responsible nurse and a rehabilitation therapist. Exercise in bed began with ankle pumps and isometric contraction of the quadriceps and gluteal muscles, and progressed to active movement of the hip and knee within the limits of pain; walking was undertaken with a frame. The exercises, their dose and the advice on weight bearing and positioning were set by the operating surgeon and the therapist according to each patient’s stage of recovery. Before discharge, the nurse and the therapist demonstrated each movement, and patients were given a QR code linking to a recorded version of the exercise programme. Patients were discharged once the incision was dry and they could transfer and walk safely with the frame, and outpatient review was arranged at 1, 3 and 6 months after surgery. Between reviews, patients could contact the ward through a WeChat mini-program or by telephone and could attend the rehabilitation outpatient clinic. The WeChat and telephone contacts made by the research team served only to arrange and complete the 1-month outcome assessment; they provided no exercise instruction and did not alter the advice given at discharge, so the adherence reported in this study is adherence to the programme each patient had been prescribed.

Measures

General Information Questionnaire

A self-designed questionnaire collected demographic and clinical information, including age, gender, education level, place of residence, marital status, employment status, living arrangement, per capita monthly family income, payment method, chronic comorbidities, anaesthesia grade, machine-assisted surgery, preoperative diagnosis and surgical procedure.

Personal Mastery Scale (PMS)

This scale was developed by Pearlin and Schooler¹² on the basis of stress coping theory and is used to measure individuals’ feelings of control over the outcomes of life events. The scale has seven items and is unidimensional. Each item is scored on a 5-point Likert scale from 1 (“not at all”) to 5 (“very much so”), and items 1 to 5 are reverse-scored. The reverse-scored items were recoded before analysis, so that after recoding all seven items are oriented in the same direction and a higher score on every item indicates a stronger sense of personal mastery. The total score ranges from 7 to 35, with higher scores indicating a higher level of personal mastery. The Cronbach’s α coefficient in this study was 0.891.

Functional Exercise Adherence Scale for Orthopedic Patients (FECS-OP)

This scale was developed by Tan et al¹³ on the basis of functional exercise and the adherence concept defined by the World Health Organization in 2003; the instrument was published as the compliance scale of functional exercise for orthopaedic patients, and adherence is the term used throughout the present manuscript. It is used to assess adherence to rehabilitation exercise in orthopaedic patients. The scale comprises three dimensions: psychological adherence (8 items), active learning adherence (4 items) and physical adherence (3 items); these are referred to throughout this manuscript as the psychological, learning and physical dimensions. Each item is scored on a 5-point Likert scale from 1 (“cannot do it at all”) to 5 (“can do it completely”), giving a total score of 15–75, where higher scores indicate better adherence. The scale has a reported content validity index of 0.936 and a Cronbach’s α of 0.930; in this study the Cronbach’s α was 0.918.

Research Design

This study used a prospective observational design with two measurement time points. Baseline data, including personal mastery and demographic and clinical variables, were collected on postoperative day 1–2, and adherence to out-of-hospital rehabilitation exercise was assessed at 1 month postoperatively. Personal mastery was measured at this early time point because it was used as a predictor of later rehabilitation behaviour and therefore had to precede, in time, the behaviour it was intended to predict. Postoperative day 1–2 also falls before discharge, when patients move from supervised in-hospital exercise to unsupervised exercise at home and when the demands of the home rehabilitation programme first become salient. Because all participants met the inclusion criteria regarding consciousness and language ability, they were able to complete the questionnaires at this point.

Data Collection

Prior to the survey, investigators underwent standardised training to ensure consistent data collection. During recruitment, investigators explained the objectives and procedures and obtained written informed consent. Participants provided WeChat and telephone contact details for follow-up. Baseline questionnaires were collected face-to-face in the ward before discharge. Follow-up adherence data were collected via WeChat or by telephone at 1 month postoperatively. After data collection, two researchers checked completeness and entered the data into the database.

Statistical Analysis

Missing data. Missing data involved only the age variable: 5 of the 333 retained patients (1.5%) had missing age information. The main study variables, including personal mastery and rehabilitation exercise adherence, were complete. Age was recorded as a continuous variable in the original dataset, and the 5 missing values were imputed on this continuous variable using multiple imputation by chained equations (MICE; mice package, R version 4.2.0) with predictive mean matching, generating five imputed datasets; the imputation model included all demographic and clinical variables entered in the subsequent analyses. After the five imputations had been completed, the imputed continuous age values were converted into the binary age category (≤60 vs >60 years) used in the analyses, yielding the single completed analytic dataset on which all subsequent analyses were run.

Participant flow and attrition analysis. Of the 350 patients initially recruited, 17 (4.86%) were lost to follow-up because their 1-month adherence outcome was unavailable; these 17 patients were therefore excluded, and the analytic sample comprised 333 patients (1-month follow-up retention rate 95.14%). Patients lost to follow-up were compared with those retained across all 14 baseline variables listed in Table 1; the comparison is reported here rather than in a separate table. Because several cells contained small expected frequencies, exact permutation tests (B = 20,000) were used in addition to the chi-square test. The two groups were comparable on age (8 [47.1%] vs 150 [45.0%] aged ≤60 years, p = 1.000), gender (male 8 [47.1%] vs 144 [43.2%], p = 0.804), education level (p = 0.081), place of residence (urban 11 [64.7%] vs 258 [77.5%], p = 0.235), marital status (p = 1.000), employment status (p = 1.000), living arrangement (p = 0.489), per capita monthly family income (p = 0.677), payment method (p = 0.655), anaesthesia grade (p = 0.570), machine-assisted surgery (p = 0.386), preoperative diagnosis (p = 0.915) and surgical procedure (p = 1.000). The only baseline variable that differed significantly was chronic comorbidities, which were less common among patients lost to follow-up than among those retained (8 [47.1%] vs 267 [80.2%], p = 0.004). Attrition was therefore low and, apart from chronic comorbidities, the retained sample was comparable with the initially recruited cohort; this difference is addressed in the Limitations.

SPSS version 26.0 was used for descriptive statistics and univariate analyses, R version 4.2.0 for multiple imputation and for the covariate-adjusted linear models, and Mplus version 8.3 for latent profile analysis and the BCH analysis. Continuous variables are presented as mean ± standard deviation and categorical variables as frequency and percentage. Statistical significance was set at two-sided p < 0.05. Exact p values are reported to three decimal places, and to four decimal places where p < 0.01, so that the Bonferroni-adjusted values can be reproduced.

Latent profile analysis (LPA) was conducted in Mplus version 8.3 using the seven item scores of the Personal Mastery Scale as continuous indicators. Models were estimated using robust maximum likelihood (MLR). To reduce the risk of convergence to local maxima, 2000 random sets of starting values were generated, of which 500 were retained for final-stage optimisation, with 100 initial-stage iterations.

Models with one to four profiles were compared, and a five-profile solution was additionally estimated as a sensitivity analysis to assess the stability of the selected solution and the possibility of profile overextraction. Model convergence and solution stability were evaluated on the basis of normal termination of model estimation and replication of the best log-likelihood value across random starts.

Model selection was based on the Akaike information criterion (AIC), Bayesian information criterion (BIC), sample-size-adjusted BIC (aBIC), entropy, the Lo–Mendell–Rubin adjusted likelihood ratio test (LMR-LRT), the bootstrap likelihood ratio test (BLRT), profile size, substantive interpretability and parsimony. Lower AIC, BIC and aBIC values indicate better relative fit, whereas entropy values closer to 1 indicate better classification accuracy. Profiles comprising less than 5% of the sample were considered potentially unstable.

After the optimal class solution had been selected, factors associated with profile membership were examined using multinomial logistic regression, with the reference class specified in the table notes.

Finally, the Bolck–Croon–Hagenaars (BCH) approach in Mplus was used to compare rehabilitation exercise adherence across latent profiles while accounting for classification uncertainty.¹⁴ Bonferroni correction was applied to the three pairwise comparisons, with the adjusted significance threshold set at α = 0.05/3 = 0.017; in the analysis, exact and Bonferroni-adjusted p values are reported for every comparison.

To examine whether profile membership was independently associated with adherence, each of the four adherence outcomes was additionally regressed on profile membership in a linear model adjusted for all 14 demographic and clinical characteristics listed in Table 1, including preoperative diagnosis and surgical procedure, with C3 as the reference category. Adjusted group means were computed as average predictive margins and pairwise differences between margins were tested with Bonferroni correction for three comparisons. In a further set of models, each of the four adherence outcomes was regressed on the Personal Mastery total score and on a squared term for that score, with the same 14 covariates, to examine whether the association between personal mastery and adherence was linear.

Results

Basic Information and Personal Mastery Scores

A total of 350 patients were recruited. Seventeen were lost to follow-up and 333 completed the 1-month assessment, giving a follow-up retention rate of 95.14%. The personal mastery total score ranged from 10 to 35, with a mean of 23.00 ± 6.46 and a mean item score of 3.29 ± 0.92. Patient demographic and clinical characteristics are shown in Table 1.

Latent Profiles of Personal Mastery

All candidate models terminated normally, and the best log-likelihood value was replicated across random starts, supporting stable model estimation. The model-fit indices are presented in Table 2.

Table 2.

Comparison of Model-Fit Indices for Latent Profile Models with One to Five Profiles (n = 333)

Number of Profiles	Log-Likelihood	AIC	BIC	aBIC	Entropy	LMR-LRT p	BLRT p	Smallest Profile, n (%)	Best log-Likelihood Replicated
1	−3691.232	7410.464	7463.778	7419.369	—	—	—	—	—
2	−3184.019	6412.037	6495.816	6426.031	0.910	<0.001	<0.001	150 (45.05)	Yes
3	−3080.406	6220.812	6335.056	6239.894	0.860	0.0004	<0.001	105 (31.53)	Yes
4	−3050.500	6177.000	6321.710	6201.171	0.852	0.3995	<0.001	30 (9.01)	Yes
5	−3023.236	6138.471	6313.646	6167.731	0.884	0.1545	<0.001	8 (2.40)	Yes

Notes: Models were estimated in Mplus 8.3 by robust maximum likelihood (MLR) using the seven Personal Mastery Scale item scores as continuous indicators, with 2000 random sets of starting values of which 500 were retained for final-stage optimisation. “Best log-likelihood replicated” indicates whether the best log-likelihood value was reproduced across random starts, ie whether the solution is free of local maxima. For the retained three-profile solution the estimated profile proportions were 0.315/0.321/0.363 (n = 105/107/121). The four-profile solution was not retained because its LMR-LRT was not significant (p = 0.3995) and its additional profile was a further subdivision of the lower-mastery participants rather than a substantively distinct profile. The five-profile solution was estimated as a sensitivity analysis: its LMR-LRT was not significant over the four-profile solution (p = 0.1545) and it produced a profile of only 8 participants (2.40%), below the prespecified 5% threshold for profile stability. The three-profile model was therefore retained.

Abbreviations: AIC, Akaike information criterion; BIC, Bayesian information criterion; aBIC, sample-size-adjusted BIC; LMR-LRT p, Lo–Mendell–Rubin adjusted likelihood ratio test comparing the k-profile solution with the (k−1)-profile solution; BLRT p, bootstrap likelihood ratio test.

The three-profile solution showed good classification accuracy (entropy = 0.860), with relatively balanced profile sizes. The LMR-LRT indicated that the three-profile model provided a statistically significant improvement over the two-profile model (p = 0.0004).

Although the four-profile solution yielded lower AIC, BIC, and aBIC values than the three-profile solution, the LMR-LRT comparing the four-profile with the three-profile model was not statistically significant (p = 0.3995). Inspection of the item-level profiles showed that the intermediate- and high-personal-mastery profiles remained essentially unchanged, whereas the additional fourth profile primarily represented a further subdivision of the lower-mastery participants rather than a substantively distinct profile.

As a sensitivity analysis, a five-profile solution was additionally estimated. The LMR-LRT did not indicate a statistically significant improvement over the four-profile solution (p = 0.1545), and the five-profile model generated a very small profile comprising only 8 participants (2.40% of the sample). This profile size was below the prespecified 5% threshold and suggested potential overextraction.¹⁵ Considering statistical fit, classification quality, solution stability, profile size, parsimony and substantive interpretability, the three-profile model was retained.

Profile Characteristics of Personal Mastery

The LPA identified three distinct profiles of early personal mastery. The average posterior probabilities for the most likely profile membership were 0.966, 0.914 and 0.936 for C1, C2 and C3 respectively, indicating good classification quality. The profile patterns are presented in Figure 1.

Figure 1.

Mean item scores on the seven personal mastery Scale (PMS) items across the three latent profiles of early personal mastery in patients after hip arthroplasty. Profiles were labelled according to the overall level of personal mastery as C1 (low personal mastery, n = 105; 31.53%), C2 (intermediate personal mastery, n = 107; 32.13%), and C3 (high personal mastery, n = 121; 36.34%). Personal mastery was assessed on postoperative day 1–2 using the seven-item unidimensional PMS. Each point represents the mean score of a single PMS item (range 1–5) within a profile; after recoding of the reverse-scored items, all seven PMS items are oriented in a consistent direction, with higher values indicating a greater sense of personal mastery. The three profiles displayed consistently low, intermediate, and high mean item scores across all seven items, without evidence of crossover.

C1: Low personal mastery (n = 105, 31.53%). Participants in this profile had the lowest mean scores across all seven Personal Mastery Scale items.

C2: Intermediate personal mastery (n = 107, 32.13%). Participants in this profile had intermediate mean scores across the seven items.

C3: High personal mastery (n = 121, 36.34%). Participants in this profile had the highest mean scores across all seven items.

Univariate Analysis of Factors Associated with Personal Mastery Profiles

Univariate analysis showed significant differences across profiles in age (χ² = 7.332, p = 0.026), gender (χ² = 7.866, p = 0.020), education level (χ² = 19.179, p = 0.004), place of residence (χ² = 6.279, p = 0.043), employment status (χ² = 8.670, p = 0.013), chronic comorbidities (χ² = 13.566, p = 0.001), machine-assisted surgery (χ² = 7.353, p = 0.025), preoperative diagnosis (χ² = 24.344, p = 0.002) and surgical procedure (χ² = 21.375, p < 0.001), as presented in Table 3. Marital status, living arrangement, per capita monthly family income, payment method and anaesthesia grade did not differ significantly across profiles (all p > 0.05).

Table 3.

General Data and Univariate Analysis of Latent Profiles of Personal Mastery (n=333)

Item	Category	n	C1 (n=105)	C2 (n=107)	C3 (n=121)	χ2	p
Age	≤60 years	150	36 (34.3)	55 (51.4)	59 (48.8)	7.332	0.026
	>60 years	183	69 (65.7)	52 (48.6)	62 (51.2)
Gender	Male	144	34 (32.4)	49 (45.8)	61 (50.4)	7.866	0.020
	Female	189	71 (67.6)	58 (54.2)	60 (49.6)
Education level	Primary school and below	110	43 (41.0)	31 (29.0)	36 (29.8)	19.179	0.004
	Junior high school	108	41 (39.0)	34 (31.8)	33 (27.3)
	High school and technical secondary school	44	5 (4.8)	21 (19.6)	18 (14.9)
	College degree and above	71	16 (15.2)	21 (19.6)	34 (28.1)
Place of residence	Urban	258	86 (81.9)	74 (69.2)	98 (81.0)	6.279	0.043
	Rural	75	19 (18.1)	33 (30.8)	23 (19.0)
Marital status	Married	265	87 (82.9)	83 (77.6)	95 (78.5)	2.488	0.647
	Single	19	3 (2.9)	8 (7.5)	8 (6.6)
	Divorced/widowed	49	15 (14.3)	16 (15.0)	18 (14.9)
Employment status	Employed	136	31 (29.5)	52 (48.6)	53 (43.8)	8.670	0.013
	Unemployed	197	74 (70.5)	55 (51.4)	68 (56.2)
Chronic comorbidities	Yes	267	94 (89.5)	88 (82.2)	85 (70.2)	13.566	0.001
	No	66	11 (10.5)	19 (17.8)	36 (29.8)
Living arrangement	Living alone	46	16 (15.2)	14 (13.1)	16 (13.2)	0.262	0.877
	Living with family	287	89 (84.8)	93 (86.9)	105 (86.8)
Per capita monthly family income	<1000	41	18 (17.1)	11 (10.3)	12 (9.9)	7.194	0.303
	1000–2999	71	19 (18.1)	26 (24.3)	26 (21.5)
	3000–4999	123	42 (40.0)	41 (38.3)	40 (33.1)
	>5000	98	26 (24.8)	29 (27.1)	43 (35.5)
Payment method	Medical insurance	321	103 (98.1)	104 (97.2)	114 (94.2)	2.727	0.256
	Out-of-pocket	12	2 (1.9)	3 (2.8)	7 (5.8)
Anaesthesia grade	Level I	6	1 (1.0)	2 (1.9)	3 (2.5)	0.747	0.945
	Level II	230	73 (69.5)	74 (69.2)	83 (68.6)
	Level III	97	31 (29.5)	31 (29.0)	35 (28.9)
Machine-assisted surgery	Yes	73	14 (13.3)	25 (23.4)	34 (28.1)	7.353	0.025
	No	260	91 (86.7)	82 (76.6)	87 (71.9)
Preoperative diagnosis	Avascular necrosis of femoral head	155	37 (35.2)	54 (50.5)	64 (52.9)	24.344	0.002
	Femoral fracture	112	53 (50.5)	31 (29.0)	28 (23.1)
	Arthritis	17	4 (3.8)	7 (6.5)	6 (5.0)
	Developmental dysplasia of the hip	33	7 (6.7)	8 (7.5)	18 (14.9)
	Other primary hip disease	16	4 (3.8)	7 (6.5)	5 (4.1)
Surgical procedure	Total hip arthroplasty	264	68 (64.8)	88 (82.2)	108 (89.3)	21.375	<0.001
	Hemiarthroplasty	69	37 (35.2)	19 (17.8)	13 (10.7)

Notes: χ² = Pearson chi-square statistic, with df = 2 × (number of categories − 1) for the three-profile comparison, ie df = 2 for binary variables, 4 for marital status and anaesthesia grade, 6 for education level and per capita monthly family income, and 8 for preoperative diagnosis; P values are exact to three decimal places. C1 = low personal mastery; C2 = intermediate personal mastery; C3 = high personal mastery. Percentages are column percentages within each profile.

Multivariate Analysis of Factors Associated with Personal Mastery Profile Membership

Using personal mastery profile membership as the dependent variable, all variables that reached p < 0.05 in univariate analysis were entered into a multinomial logistic regression model, together with potential confounders reported in the literature.^16–18 The model therefore included nine variables: age (≤60 vs >60 years), gender (male vs female), education level (primary school and below, junior high school, high school and technical secondary school, college degree and above), place of residence (urban vs rural), employment status (unemployed vs employed), chronic comorbidities (no vs yes), machine-assisted surgery (no vs yes), preoperative diagnosis (avascular necrosis of the femoral head as reference; femoral fracture, arthritis, developmental dysplasia of the hip, other primary hip disease) and surgical procedure (total hip arthroplasty vs hemiarthroplasty). Before model construction, multicollinearity among the candidate predictors was assessed using the variance inflation factor (VIF); all VIF values were below 10 and all tolerance values above 0.2, indicating no serious multicollinearity¹⁹ (Table 4). The multinomial logistic regression showed that age, gender, education level, employment status and chronic comorbidities were independently associated with profile membership (p < 0.05), whereas place of residence, machine-assisted surgery, preoperative diagnosis and surgical procedure were not (Table 5).

Table 4.

Multicollinearity Diagnostics for the Candidate Predictors of Personal Mastery Profile Membership

Item	Tolerance	VIF
Age	0.561	1.781
Gender	0.844	1.185
Education level	0.458	2.185
Place of residence	0.842	1.188
Employment status	0.786	1.272
Chronic comorbidities	0.853	1.173
Machine-assisted surgery	0.816	1.226
Preoperative diagnosis	0.440	2.275
Surgical procedure	0.510	1.961

Notes: Tolerance = 1/VIF. All VIF values were below 10 and all tolerance values above 0.2, indicating no serious multicollinearity.

Abbreviation: VIF, variance inflation factor.

Table 5.

Multinomial Logistic Regression of Latent Profiles of Personal Mastery (Reference: C1; n = 333)

Panel A. C2 (intermediate personal mastery) vs C1 (low personal mastery)
Variable	β	SE	Wald χ2	p	OR	95% CI for OR
Constant	0.920	0.643	2.049	0.152	2.509	0.712–8.839
Age >60 years (ref. ≤60)	0.293	0.406	0.522	0.470	1.341	0.605–2.969
Female (ref. male)	−0.308	0.324	0.901	0.343	0.735	0.389–1.388
Primary school and below (ref. college and above)	−0.635	0.468	1.844	0.174	0.530	0.212–1.325
Junior high school (ref. college and above)	−0.472	0.429	1.209	0.272	0.624	0.269–1.447
High school and technical secondary school (ref. college and above)	1.227	0.615	3.978	0.046	3.409	1.021–11.380
Urban (ref. rural)	−0.601	0.374	2.584	0.108	0.548	0.264–1.141
Employed (ref. unemployed)	0.668	0.334	4.005	0.045	1.950	1.014–3.751
Chronic comorbidities (ref. no)	−0.295	0.449	0.433	0.511	0.744	0.309–1.795
Machine-assisted surgery (ref. no)	0.332	0.422	0.618	0.432	1.393	0.609–3.185
Femoral fracture (ref. avascular necrosis)	−0.506	0.462	1.197	0.274	0.603	0.244–1.493
Arthritis (ref. avascular necrosis)	0.290	0.707	0.169	0.681	1.337	0.334–5.348
Developmental dysplasia of the hip (ref. avascular necrosis)	−0.417	0.617	0.456	0.499	0.659	0.197–2.209
Other primary hip disease (ref. avascular necrosis)	0.514	0.697	0.545	0.460	1.673	0.427–6.556
Hemiarthroplasty (ref. total hip arthroplasty)	−0.164	0.479	0.117	0.732	0.849	0.332–2.172
Panel B. C3 (high personal mastery) vs C1 (low personal mastery)
Variable	β	SE	Wald χ2	p	OR	95% CI for OR
Constant	1.108	0.622	3.169	0.075	3.029	0.894–10.260
Age >60 years (ref. ≤60)	0.811	0.395	4.220	0.040	2.251	1.038–4.882
Female (ref. male)	−0.764	0.322	5.618	0.018	0.466	0.248–0.876
Primary school and below (ref. college and above)	−0.335	0.449	0.558	0.455	0.715	0.297–1.723
Junior high school (ref. college and above)	−0.839	0.417	4.047	0.044	0.432	0.191–0.979
High school and technical secondary school (ref. college and above)	0.672	0.617	1.185	0.276	1.958	0.584–6.565
Urban (ref. rural)	0.363	0.401	0.817	0.366	1.437	0.654–3.157
Employed (ref. unemployed)	0.466	0.335	1.930	0.165	1.593	0.826–3.073
Chronic comorbidities (ref. no)	−0.986	0.423	5.424	0.020	0.373	0.163–0.855
Machine-assisted surgery (ref. no)	0.372	0.406	0.838	0.360	1.451	0.654–3.217
Femoral fracture (ref. avascular necrosis)	−0.643	0.446	2.082	0.149	0.526	0.219–1.259
Arthritis (ref. avascular necrosis)	−0.054	0.738	0.005	0.942	0.947	0.223–4.021
Developmental dysplasia of the hip (ref. avascular necrosis)	0.498	0.563	0.784	0.376	1.646	0.546–4.958
Other primary hip disease (ref. avascular necrosis)	0.070	0.743	0.009	0.925	1.072	0.250–4.597
Hemiarthroplasty (ref. total hip arthroplasty)	−0.949	0.490	3.747	0.053	0.387	0.148–1.012

Notes: Reference outcome category is C1 (low personal mastery). Reference categories for the predictors are given in the first column. Wald χ² has 1 df; OR = exp(β). McFadden pseudo-R² = 0.100. N = 333.

Differences in Functional Exercise Adherence Across Personal Mastery Profiles

The BCH approach was used to compare adherence to out-of-hospital rehabilitation exercise across the three profiles of early personal mastery. Patients in the high personal mastery profile had the highest total adherence score. For the total score and for the psychological and learning dimensions, scores increased in the order C1 < C2 < C3, and the overall differences across profiles were significant (psychological χ² = 93.835, learning χ² = 43.984, total χ² = 81.398; all p < 0.001). In the pairwise comparisons, C1 differed significantly from both C2 and C3 on all four outcomes after Bonferroni correction. The C2–C3 difference was significant only for the psychological dimension (χ² = 9.210, p = 0.0024; Bonferroni-adjusted p = 0.007). It was not significant for the learning dimension (χ² = 2.632, p = 0.105; adjusted p = 0.314), the physical dimension (χ² = 0.053, p = 0.818; adjusted p = 1.000) or the total score (χ² = 4.625, p = 0.032; adjusted p = 0.095). For the physical dimension, both the intermediate and the high personal mastery profiles scored higher than the low personal mastery profile, whereas the two higher profiles were virtually indistinguishable (d = 0.03). These results are presented in Table 6.

Table 6.

Differences in Out-Of-Hospital Rehabilitation Exercise Adherence Across Profiles of Early Personal Mastery (BCH Analysis; n = 333)

Outcome	C1 (n = 105)	C2 (n = 107)	C3 (n = 121)	Overall BCH Wald χ2 (df = 2), p	C1 vs C2: χ2 (df = 1) (p; p adj)	C1 vs C3: χ2 (df = 1) (p; p adj)	C2 vs C3: χ2 (df = 1) (p; p adj)
Psychological dimension	28.060 ± 0.612	32.706 ± 0.583	34.918 ± 0.363	93.835, <0.001	28.055 (<0.001; <0.001)	93.507 (<0.001; <0.001)	9.210 (0.0024; 0.007)
Learning dimension	10.938 ± 0.364	13.308 ± 0.394	14.224 ± 0.359	43.984, <0.001	18.126 (<0.001; <0.001)	41.513 (<0.001; <0.001)	2.632 (0.105; 0.314)
Physical dimension	7.356 ± 0.329	9.459 ± 0.325	9.348 ± 0.317	26.016, <0.001	19.188 (<0.001; <0.001)	19.149 (<0.001; <0.001)	0.053 (0.818; 1.000)
Total adherence score	46.354 ± 1.084	55.472 ± 1.036	58.491 ± 0.824	81.398, <0.001	34.359 (<0.001; <0.001)	79.952 (<0.001; <0.001)	4.625 (0.032; 0.095)

Note: C1 = low personal mastery; C2 = intermediate personal mastery; C3 = high personal mastery. Values for C1, C2 and C3 are BCH-weighted means ± standard errors. χ² values are from the BCH Wald test (overall test, df = 2; pairwise tests, df = 1). P = unadjusted two-sided P value, reported to three decimal places and to four decimal places where P < 0.01 so that the adjusted value can be reproduced; P adj = Bonferroni-adjusted P value for three pairwise comparisons (P × 3), against an adjusted significance threshold of α = 0.05/3 = 0.017. Pairwise differences that remain significant after Bonferroni correction are: C1 vs C2 and C1 vs C3 for all four outcomes, and C2 vs C3 for the psychological dimension only. The C2–C3 differences for the learning dimension, the physical dimension and the total adherence score were not statistically significant.

To move beyond inference based on p values alone, Cohen’s d effect sizes were calculated for all pairwise comparisons across the four outcomes (psychological dimension, learning dimension, physical dimension and total score). Cohen’s d was computed as the difference between the two BCH-weighted profile means divided by the pooled standard deviation, so that classification uncertainty was carried through the BCH weighting; the derivation is set out in the note to Table 7. Effect sizes are reported in Table 7 and should be read together with the significance tests in Table 6.

Table 7.

Effect Sizes (Cohen’s D) for Differences in Out-Of-Hospital Rehabilitation Exercise Adherence Across Profiles of Early Personal Mastery

Outcome	C1 vs C2	C1 vs C3	C2 vs C3
Psychological dimension	0.76	1.32	0.44
Learning dimension	0.61	0.85	0.23
Physical dimension	0.62	0.58	0.03
Total adherence score	0.84	1.21	0.31

Notes: C1 = low personal mastery; C2 = intermediate personal mastery; C3 = high personal mastery. Cohen’s d was computed as the difference between the two BCH-weighted profile means divided by the pooled standard deviation, using the BCH weights from the latent profile model; the within-profile standard deviations were recovered from the BCH-weighted standard errors reported in Table 6 (SD = SE × √n), and the pooled standard deviation was calculated with the standard formula √[((n₁−1)SD₁² + (n₂−1)SD₂)² / (n₁ + n₂ − 2)]. Because classification uncertainty is incorporated through the BCH weighting, the effect sizes are based on the weighted estimates rather than on modal class assignment. Conventional interpretation thresholds: 0.20 = small, 0.50 = medium, 0.80 = large.

Psychological Dimension

For the psychological dimension, a moderate effect size was observed between C1 and C2 (d = 0.76) and a large effect size between C1 and C3 (d = 1.32), indicating a substantial difference in psychological adherence between the low personal mastery profile and the other two profiles. The difference between C2 and C3 was also statistically significant (p = 0.0024; Bonferroni-adjusted p = 0.007) but corresponded to a small-to-moderate effect size (d = 0.44), suggesting limited practical differentiation between these two profiles on this dimension.

Learning Dimension

In the learning dimension, a moderate effect was observed between C1 and C2 (d = 0.61) and a large effect between C1 and C3 (d = 0.85), indicating a substantial gap in learning-related adherence between the low personal mastery profile and the other two profiles. The difference between C2 and C3 was not statistically significant (χ² = 2.632, p = 0.105; Bonferroni-adjusted p = 0.314) and corresponded to a small effect size (d = 0.23), indicating that learning-related adherence did not differ appreciably between the intermediate and high personal mastery profiles.

Physical Dimension

For the physical dimension, moderate effect sizes were observed for both C1 vs C2 (d = 0.62) and C1 vs C3 (d = 0.58), indicating a substantial difference in physical exercise adherence between the low personal mastery profile and the other two profiles. The difference between C2 and C3 was not statistically significant (χ² = 0.053, p = 0.818; Bonferroni-adjusted p = 1.000) and the effect size was negligible (d = 0.03). Physical exercise adherence was therefore very similar in the intermediate and high personal mastery profiles.

Total Score

For the total adherence score, a large effect size was observed between C1 and C2 (d = 0.84) and a large effect size between C1 and C3 (d = 1.21), indicating substantially lower overall adherence in the low personal mastery profile than in the other two profiles. The difference between C2 and C3 was not statistically significant after Bonferroni correction (χ² = 4.625, p = 0.032; adjusted p = 0.095) and corresponded to a small effect size (d = 0.31), indicating limited differentiation between the intermediate and high personal mastery profiles in overall adherence.

Covariate-Adjusted Analysis

To determine whether profile membership was independently associated with adherence, the four adherence outcomes were additionally regressed on profile membership with adjustment for all 14 demographic and clinical characteristics listed in Table 1, including preoperative diagnosis and surgical procedure (Table 8). Profile membership remained significantly associated with the psychological dimension (F(2, 307) = 27.56, p < 0.001, R² = 0.355), the learning dimension (F(2, 307) = 9.00, p < 0.001, R² = 0.275), the physical dimension (F(2, 307) = 6.83, p = 0.001, R² = 0.162) and the total adherence score (F(2, 307) = 23.23, p < 0.001, R² = 0.344). After Bonferroni correction, C1 differed significantly from both C2 and C3 on the psychological dimension, the physical dimension and the total adherence score, and C2 differed significantly from C3 on the psychological dimension (adjusted p = 0.009). The remaining comparisons did not reach the corrected threshold: C1 vs C2 on the learning dimension (adjusted p = 0.050), and C2 vs C3 on the learning dimension (adjusted p = 0.154), the physical dimension (adjusted p = 1.000) and the total adherence score (adjusted p = 0.033). Adding preoperative diagnosis and surgical procedure improved model fit for the psychological dimension (p < 0.001), the learning dimension (p = 0.014) and the total adherence score (p = 0.005), but not for the physical dimension (p = 0.152).

Table 8.

Covariate-Adjusted Differences in Out-of-Hospital Rehabilitation Exercise Adherence Across Profiles of Early Personal Mastery (n = 333)

Outcome	Adj Mean C1	Adj Mean C2	Adj Mean C3	F (2, 307)	P	R2	C1 vs C2: Difference (95% CI), adj P	C1 vs C3: Difference (95% CI), adj P	C2 vs C3: Difference (95% CI), adj P
Psychological dimension	29.04	32.31	34.31	27.56	<0.001	0.355	−3.28 (−4.68, −1.88) <0.001*	−5.27 (−6.67, −3.87) <0.001*	−1.99 (−3.31, −0.68) 0.009*
Learning dimension	11.69	12.91	13.85	9.00	<0.001	0.275	−1.22 (−2.22, −0.22) 0.050	−2.15 (−3.15, −1.15) <0.001*	−0.93 (−1.87, 0.01) 0.154
Physical dimension	7.72	9.06	9.35	6.83	0.001	0.162	−1.35 (−2.26, −0.43) 0.012*	−1.63 (−2.55, −0.72) 0.001*	−0.29 (−1.15, 0.57) 1.000
Total adherence score	48.45	54.29	57.50	23.23	<0.001	0.344	−5.84 (−8.47, −3.22) <0.001*	−9.05 (−11.68, −6.43) <0.001*	−3.21 (−5.68, −0.74) 0.033

Notes: Multivariable linear models adjusted for all 14 demographic and clinical characteristics listed in Table 1, including preoperative diagnosis and surgical procedure. Profile membership was entered as a categorical predictor using the most likely class assignment from the retained three-profile model, with C3 (high personal mastery) as the reference category; treatment contrasts were used. All 14 covariates were entered as categorical factors (23 indicator variables), giving 26 estimated parameters including the intercept and the two profile contrasts and hence 307 residual degrees of freedom. Classification uncertainty was accounted for in the BCH analysis (Table 6) and is not carried into these linear models, which therefore complement rather than replace the BCH results. Adjusted means are average predictive margins. Pairwise cells report the difference between adjusted means with its 95% confidence interval and the Bonferroni-adjusted P value for three comparisons (P × 3); * denotes Bonferroni-adjusted P < 0.017. Added-variable tests for preoperative diagnosis plus surgical procedure: psychological dimension P < 0.001, learning dimension P = 0.014, total adherence score P = 0.005, physical dimension P = 0.152.

Linearity of the association with the continuous mastery score. In the models containing the total score and its square, the squared term was negative and statistically significant for the total adherence score (b = −0.0347, SE = 0.0135, t = −2.566, p = 0.010), the psychological dimension (b = −0.0170, SE = 0.0076, t = −2.242, p = 0.025) and the physical dimension (b = −0.0121, SE = 0.0046, t = −2.636, p = 0.009), and was not significant for the learning dimension (b = −0.0056, SE = 0.0051, t = −1.100, p = 0.272); the linear term was positive in all four models. Adherence therefore increased with the personal mastery score at a diminishing rate rather than in proportion to it.

Discussion

Differences in Personal Mastery Profiles After Hip Arthroplasty

Latent profile analysis identified three profiles of personal mastery in patients after hip arthroplasty. The mean personal mastery score in this study was 23.00 ± 6.46, slightly higher than that reported for patients with chronic heart failure by Xiang et al.¹⁶ This difference may reflect the longer disease course, recurrent and severe symptoms and poorer prognosis associated with heart failure, compared with the relatively abrupt, single-event trajectory of joint replacement. (1) The low personal mastery profile accounted for 31.53% of the sample and was characterised by the lowest mean item scores across the seven Personal Mastery Scale items. Previous work has shown that sense of control is closely related to patients’ clinical situation;^20,21 patients in this profile may therefore be more likely to experience reduced perceived control and lower decision-making initiative when facing postoperative challenges. (2) The intermediate personal mastery profile accounted for 32.13% of the sample and showed intermediate mean scores across the seven items. Because personal mastery was assessed at a single time point, this profile describes a level of personal mastery rather than a fluctuation over time. For patients in the first two profiles, psychological support such as counselling or cognitive behavioural approaches may help to strengthen positive beliefs and support engagement in functional exercise. These suggestions are offered as hypotheses to be tested rather than as interventions demonstrated to be effective in the present study. (3) The high personal mastery profile accounted for 36.34% of the sample and was characterised by the highest mean item scores. These patients reported a stronger general sense of control, yet their physical state may still limit engagement in daily activities. Close monitoring and timely support may therefore help to sustain their recovery.

Factors Associated with Personal Mastery Profile Membership

Multinomial logistic regression identified several factors independently associated with personal mastery profile membership after hip arthroplasty: age, gender, education level, employment status and chronic comorbidities. Place of residence, machine-assisted surgery, preoperative diagnosis and surgical procedure were not independently associated with profile membership.

• (1)Age: Age was also associated with personal mastery profile membership. Patients older than 60 years were more likely to belong to the high personal mastery profile rather than the low personal mastery profile (C3 vs C1: OR = 2.251, 95% CI 1.038–4.882, p = 0.040), whereas the comparison between the intermediate and low personal mastery profiles was not significant (C2 vs C1: OR = 1.341, 95% CI 0.605–2.969, p = 0.470). Previous population-based studies have reported age-related differences in perceived control and personal mastery, although the direction of these associations varies across age groups and populations.^22,23 Older adults may maintain higher levels of mastery through accumulated life experience and adaptive coping strategies, whereas age-related functional decline and increased disease burden may negatively affect perceived control.²⁴ Therefore, the higher likelihood of older patients belonging to the high personal mastery profile in this study may reflect differences in psychological adaptation rather than an effect of age itself.
• (2)Gender: Compared with the C3 group, female patients were more likely to be classified into the C1 group (C3 vs C1: OR = 0.466, 95% CI 0.248–0.876, p = 0.018); the corresponding C2 vs C1 comparison was not significant (OR = 0.735, 95% CI 0.389–1.388, p = 0.343). Differences in perceived control between men and women have been reported in large population studies,²⁴ and anxiety and depression are closely linked to lower personal mastery;²⁵ women facing stressful events may therefore be at higher risk of negative emotions and persistent psychological distress, which could be associated with a reduced sense of personal mastery. Healthcare professionals may nevertheless need to attend closely to the emotional status of these patients and provide individualised guidance and support.
• (3)Education level: Educational attainment was also associated with profile membership. Compared with patients with a college degree or above, those with a high school or technical secondary school education were more likely to be classified into the intermediate profile rather than the low profile (C2 vs C1: OR = 3.409, 95% CI 1.021–11.380, p = 0.046), whereas patients with a junior high school education were less likely to be classified into the high profile rather than the low profile (C3 vs C1: OR = 0.432, 95% CI 0.191–0.979, p = 0.044). Previous research has consistently linked educational attainment with perceived control and mastery.^24,26,27 Education may enhance individuals’ capacity to access, understand and integrate health information, and may provide repeated opportunities to develop problem-solving and decision-making skills; these experiences can strengthen confidence in managing challenging situations and contribute to a greater sense of control.^27,28 In the postoperative setting, a better understanding of the disease, the surgical procedure and the rehabilitation requirements may enable patients to form clearer expectations and participate more actively in recovery. Healthcare professionals may therefore need to take patients’ educational background and health-information comprehension into account when delivering rehabilitation education, for example by supplementing verbal explanations with illustrated materials or videos for patients who may have difficulty understanding rehabilitation information.
• (4)Place of residence: Place of residence was associated with profile membership in univariate analysis (χ² = 6.279, p = 0.043), with the highest proportion of rural residents in C2 (33/107, 30.8%) compared with C1 (19/105, 18.1%) and C3 (23/121, 19.0%), but it was not independently associated with profile membership after adjustment for the other variables (C2 vs C1: OR = 0.548, 95% CI 0.264–1.141, p = 0.108; C3 vs C1: OR = 1.437, 95% CI 0.654–3.157, p = 0.366). Rural residents may face greater barriers to obtaining post-discharge care and information, as has been described for rural older patients and their caregivers in other care contexts;²⁹ such barriers could attenuate the association between residence and mastery level once education, employment status and comorbidity are taken into account. Access to rehabilitation services was not measured in this cohort, so this explanation is speculative.
• (5)Employment status: Unemployed patients were more likely to be classified into the C1 group than into the C2 group (C2 vs C1: OR = 1.950, 95% CI 1.014–3.751, p = 0.045 for employed compared with unemployed). The C1 group consisted primarily of retired older patients and of patients who were unemployed because of illness. For retired patients, changes in social roles and the narrowing of social circles within the family may be associated with a reduced sense of personal value, a possibility consistent with reported changes in the form of intergenerational relationships in Chinese families,³⁰ although neither social role nor sense of value was measured here. Patients who are unable to work because of illness are more likely to develop feelings of powerlessness, stemming from physical and financial limitations as well as concerns about their future careers, and lower personal mastery is closely associated with anxiety and depression.²⁵ These findings suggest that healthcare providers should attend to the psychological needs of unemployed patients and, where appropriate, encourage retired patients to establish new life goals and take part in suitable social activities. For patients unable to work because of illness, vocational rehabilitation information and psychological support may help them cope with concurrent life pressures.
• (6)Chronic comorbidities: Chronic comorbidity was also associated with personal mastery profile membership. Patients with chronic comorbidities were less likely to belong to the high personal mastery profile than to the low personal mastery profile (C3 vs C1: OR = 0.373, 95% CI 0.163–0.855, p = 0.020), whereas the comparison between the intermediate and low personal mastery profiles was not significant (C2 vs C1: OR = 0.744, 95% CI 0.309–1.795, p = 0.511). Previous studies have suggested that managing multiple chronic conditions increases the complexity of self-management and is associated with lower perceived control and mastery beliefs.^23,31 Patients with multimorbidity may also experience greater symptom burden and functional limitations, which may affect psychological adaptation during recovery.³² Therefore, the lower likelihood of patients with chronic comorbidities belonging to the high personal mastery profile may reflect differences in perceived control and adaptation capacity among patients with different health burdens.

Association Between Personal Mastery and Functional Exercise Adherence

Patients are often discharged within 5–10 days after hip arthroplasty,³³ and functional exercise is therefore performed mainly after discharge. Postoperative functional exercise plays an important part in recovery and in the prevention of complications,⁶ yet adherence to postoperative rehabilitation exercise may be suboptimal³⁴ and rehabilitative guidance after total knee arthroplasty remains heterogeneous.³⁵ In this study, adherence at 1 month differed across the early mastery profiles (BCH-weighted mean ± SE: C1 46.35 ± 1.08, C2 55.47 ± 1.04, C3 58.49 ± 0.82; Table 6). Because this study was observational, these results indicate an association rather than a causal relationship.

This study therefore identified differences in adherence to out-of-hospital rehabilitation exercise at 1 month across profiles of early personal mastery, using the BCH approach to account for classification uncertainty. Adherence scores were higher in C2 and in C3 than in C1 for the total score and for all three subdimension scores, so that higher early personal mastery was associated with better adherence to out-of-hospital rehabilitation exercise; C2 and C3 differed significantly from each other only on the psychological dimension. This is broadly consistent with the expectations of Social Cognitive Theory regarding the role of generalised psychological resources in health behaviour:⁸ an individual’s belief in their ability to perform a specific behaviour is a key cognitive factor in the adoption and maintenance of health behaviour. Personal mastery, as a positive expectation of controllability over future events, can be regarded as a generalised manifestation of such beliefs in the rehabilitation context.

Participants in the high personal mastery profile reported both higher perceived personal control and higher rehabilitation exercise adherence. Personal mastery is associated with disease adaptation, and higher levels tend to correlate with greater life satisfaction and treatment adherence,^36,37 which may in turn be linked to more proactive engagement in rehabilitation exercise. The early postoperative period may therefore be a relevant time point for out-of-hospital rehabilitation exercise adherence after hip arthroplasty, and the hospitalisation period may represent a window for supporting personal mastery.

Future intervention studies could examine whether patients with low or intermediate personal mastery benefit from additional psychological or rehabilitation support in the early postoperative period, and whether gender, education level, place of residence, employment status and chronic comorbidities should be taken into account when designing such support.

The effect size analysis adds two observations. Across all four outcomes, C1 showed moderate-to-large effects when compared with C2 and C3 (d = 0.58–1.32), which identifies patients with low personal mastery as a priority group for future intervention research aimed at enhancing adherence to functional exercise. By contrast, the differences between C2 and C3 were consistently small or negligible—largest for the psychological dimension (d = 0.44) and smallest for the physical dimension (d = 0.03), where the two profiles were virtually indistinguishable. Patients with intermediate personal mastery may therefore already have reached a level of behavioural adherence comparable to that of the high personal mastery profile, particularly in the physical domain, and may not require intensive reinforcement there; the psychological and learning dimensions (d = 0.44 and d = 0.23) are where more room for improvement remains. Taken together, these results are consistent with a stratified approach to postoperative rehabilitation in future intervention research: patients with low personal mastery may need comprehensive support across psychological, learning and behavioural domains, whereas those with intermediate mastery may need targeted cognitive and psychological support with less emphasis on reinforcing physical exercise behaviour.

Contribution of the Latent Profile Approach

Analysing personal mastery through latent profiles rather than as a continuous score yielded three results that the total score alone did not. First, the size of the group at risk was made explicit: patients in the low personal mastery profile accounted for 31.53% of the sample, a proportion that the mean score of 23.00 ± 6.46 does not convey. Second, the association between mastery and adherence was not linear. In the models containing the total score and its square, the squared term was significant and negative for the total adherence score, the psychological dimension and the physical dimension, so adherence rose with mastery at a diminishing rate. The profiles identify where this increase levels off: the intermediate and high personal mastery profiles were statistically indistinguishable on the learning dimension (d = 0.23), the physical dimension (d = 0.03) and the total adherence score (d = 0.31), whereas both differed substantially from the low personal mastery profile (d = 0.58–1.32). A single slope for a continuous score would average the large difference between the low profile and the two higher profiles together with the absence of any detectable difference between those two profiles, and would be read as a steady gain in adherence with increasing mastery. Third, the comparison was made with the Bolck–Croon–Hagenaars approach, which carries classification uncertainty into the model of the distal outcome, and classification was accurate (entropy = 0.860; average posterior probabilities of 0.914–0.966), so the differences between profiles are not an artefact of cut points applied to the total score. In practice, the seven-item scale takes a few minutes to complete at the bedside, so the group in which poor adherence is concentrated can be identified before discharge rather than inferred from the mean score of the whole sample. Whether support organised around these levels improves adherence is a question for intervention research.

The profile approach also has limits that should be weighed against these advantages. Because the Personal Mastery Scale is unidimensional, the profiles and the total score rank patients in the same order, and the two approaches differ in what they make visible rather than in how patients are ranked. Profile membership is probabilistic; although the Bolck–Croon–Hagenaars weighting carries that uncertainty into the comparisons, the profile assigned to any individual patient remains an estimate. The proportions observed here are specific to this sample and to a single measurement occasion. Before mastery levels of this kind could be used to decide who receives additional support after discharge, the same three levels and the boundaries between them would need to be reproduced in other populations and at other points in recovery. The two ways of analysing mastery are complementary rather than competing: the continuous score describes how strongly mastery and adherence move together, whereas the profiles describe how mastery is distributed in this population and where poor adherence is concentrated.

Limitations

Several limitations should be acknowledged. First, this was a single-centre study conducted at one tertiary hospital in Urumqi using convenience sampling, without stratification or random selection. The findings may therefore not generalise to other hospitals, regions or populations. Second, because the profiles were derived from a single measurement occasion, their temporal stability and any change over the recovery trajectory could not be evaluated, and the extent to which the profiles reflect enduring dispositions rather than a transient response to surgery remains unclear. Responses given on postoperative day 1–2 may also have been affected by transient postoperative symptoms such as pain and fatigue. Third, rehabilitation exercise adherence was self-reported at 1 month, which is a short follow-up for a procedure whose functional benefits accrue over several months; self-report is susceptible to recall and social-desirability bias, and no objective measure of exercise performance was available. Adherence at 3 and 6 months was not assessed, so medium- and long-term behavioural patterns could not be described. Fourth, the sample was clinically heterogeneous with respect to surgical indication and procedure type, and several potentially relevant variables—including postoperative pain, analgesic use, complications, caregiver support and access to community rehabilitation resources—were not measured, so residual confounding cannot be excluded. Fifth, 17 of the 350 recruited patients (4.86%) were lost to follow-up; the retained and lost groups were comparable on all baseline characteristics except chronic comorbidities, but some selection bias remains possible. Sixth, personal mastery was assessed with the seven-item unidimensional Personal Mastery Scale, and latent profiles of a unidimensional indicator tend to differ mainly in level rather than in pattern; the profiles should therefore be interpreted as a categorisation of mastery level. Future multicentre studies with repeated measurement of personal mastery, longer follow-up, objective adherence monitoring and explicitly measured mediating variables are warranted to clarify the longitudinal and causal relationships suggested here.

Conclusion

This study identified heterogeneity in early personal mastery among patients after hip arthroplasty, with three distinct profiles: C1 low personal mastery, C2 intermediate personal mastery and C3 high personal mastery. The profiles differed significantly in out-of-hospital rehabilitation exercise adherence at 1 month postoperatively, and the differences remained significant after adjustment for demographic and clinical covariates, with higher personal mastery associated with better adherence. The intermediate and high personal mastery profiles did not differ significantly from each other on the learning dimension, the physical dimension or the total adherence score, although they did differ on the psychological dimension (Bonferroni-adjusted p = 0.007). Because this was a single-centre observational study with a 1-month follow-up and self-reported adherence outcomes, the findings support an association rather than a causal relationship. Considering mastery heterogeneity during hospitalisation may inform the design of rehabilitation support strategies, although interventional studies are needed to establish whether tailored approaches improve outcomes.

Funding Statement

The authors declare financial support was received for the research, authorship, and/or publication of this article. This work was supported by School-level Discipline Cultivation Construction Project of Nursing, Xinjiang Second Medical College (No. XKP202503) and Xinjiang “Tianshan Talents” Medical and Health High-level Talent Training Programme (No. TSYC202301B091).

Data Sharing Statement

The data supporting the findings of this study are available from the corresponding author, Lingyun Shi, upon reasonable request.

Ethics Statement

The study involving human participants was approved by the Medical Ethics Committee of the First Affiliated Hospital of Xinjiang Medical University under the ethics number K202412-09. The study was conducted in accordance with the Declaration of Helsinki and relevant institutional regulations. The participants provided their written informed consent to participate in this study.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors report support for the manuscript from School-level Discipline Cultivation Construction Project of Nursing, Xinjiang Second Medical College (No: XKP202503), during the conduct of the study.The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.