Abstract
Objective
This study aimed to identify distinct in-hospital cardiac rehabilitation (CR) adherence profiles and explore their associated clinical and sociodemographic factors among patients following percutaneous coronary intervention (PCI).
Methods
A cross-sectional survey was conducted among patients undergoing Phase I cardiac rehabilitation following percutaneous coronary intervention (PCI) who were hospitalized in the cardiology department between June and July 2025 (n=384). Data were collected using a general information questionnaire and a treatment adherence questionnaire (Since the study population consisted of inpatients undergoing PCI followed by phase I cardiac rehabilitation, the dimension of follow-up compliance was excluded). LPA, a person-centered method that identifies unobserved subgroups (profiles) based on response patterns, was prespecified to classify CR adherence profiles. Multinomial logistic regression was performed to examine factors associated with profile membership. Clinical indicators (number of diseased vessels, LVEF, LDL-C, and serum creatinine) were included as candidate predictors; after LASSO selection, LDL-C and number of diseased vessels were retained and entered the final multinomial logistic regression model as continuous variables (original values).
Results
Three distinct CR adherence profiles were identified: Low CR Adherence (125/384, 32.55%), Medium CR Adherence (169/384, 44.01%), and High CR Adherence (90/384, 23.44%). Profile membership was significantly associated with gender, living situation, family monthly income, residential distance, smartphone use/proficiency and LDL-C (P < 0.05).
Conclusion
CR adherence among post-PCI patients was overall moderate-to-low, with substantial heterogeneity across adherence patterns. The associated sociodemographic and contextual factors may help inform profile-based, tailored support to improve CR adherence after PCI. Given the cross-sectional design, these associations are non-causal and should be validated in future multicenter longitudinal and intervention studies.
Keywords: PCI, latent profile analysis, cardiac rehabilitation adherence, associated factors
Introduction
Percutaneous Coronary Intervention (PCI) is a primary treatment for coronary heart disease, offering advantages such as minimal trauma and rapid recovery.1 Nevertheless, the risks of restenosis and recurrent cardiovascular events persist after PCI, making secondary prevention essential. Cardiac rehabilitation (CR) is recommended by contemporary cardiovascular prevention guidelines as an effective, comprehensive strategy to improve risk-factor control, functional capacity, and long-term outcomes.2–4CR is a multidisciplinary treatment method involving five major prescriptions:2,5 medication, exercise, nutrition, psychological (including sleep management), and patient education (risk factor management and smoking cessation). It can help patients develop healthy lifestyles, reduce cardiovascular disease risk, decrease recurrence and disability, and improve quality of life.
Despite these benefits, CR participation and adherence remain suboptimal worldwide. Specifically, the initial stage following PCI—Phase I (in-hospital) CR—is critical for establishing recovery behaviors and psychological readiness. Improving adherence during this acute window is essential for ensuring a successful transition to long-term outpatient rehabilitation.Multiple barriers at the patient, provider, and health-system levels limit engagement, and adherence after revascularization procedures is often poor.6 Evidence from China also indicates particularly low CR adherence among post-PCI patients, with a reported non-adherence rate of 94%.7 Importantly, adherence is heterogeneous rather than uniform: patients may exhibit distinct patterns across medication, exercise, and lifestyle behaviors. Moreover, prior person-centered studies in CR have often focused on single dimensions of rehabilitation, leaving uncertainty about multi-domain adherence patterns after PCI and their real-world correlates.8 Clarifying these patterns is increasingly relevant as home-based and digitally supported CR models expand, yet digital access and skills may differentially shape adherence behaviors.8,9
To address this gap, a “person-centered” approach is required to identify unobserved subgroups with distinct adherence phenotypes. Latent Profile Analysis (LPA) is a sophisticated statistical technique that allows for the identification of these hidden subpopulations based on specific behavioral patterns. Unlike traditional methods, LPA can capture the nuances of how different dimensions of adherence—such as medication, diet, and early mobilization—coalesce within individual patients. Identifying these specific profiles during Phase I CR is a prerequisite for developing tailored interventions that address the unique needs of different patient clusters.
The present study focuses strictly on the Phase I (in-hospital) period (1–7 days post-PCI) to capture the earliest stage of rehabilitation behavior. By employing LPA, we aim to: (1) identify distinct adherence profiles within this acute window; and (2) examine the clinical and sociodemographic factors associated with these profiles. Given the cross-sectional nature of this study and its specific focus on the inpatient setting, follow-up components were excluded to ensure the content validity of our assessment. The findings will provide a scientific basis for clinicians to implement “nudge” strategies and personalized support before patients are discharged.
Materials and Methods
Study Design
This study was conducted in the cardiovascular ward of a tertiary hospital in Lishui City. It focused on inpatient (Phase I) cardiac rehabilitation. The research subjects were patients who were hospitalized in the cardiology department of this hospital from June to July 2025 and received elective PCI treatment for stable coronary heart disease.
Participants
A convenience sampling method was used to recruit patients who had successfully undergone PCI for coronary heart disease.
Inclusion criteria: (1) Patients meeting cardiac rehabilitation criteria and participating in Phase I cardiac rehabilitation; (2) Cardiac function classified as New York Heart Association (NYHA) Class I or II (NYHA class I–II refers to patients with no (class I) or only mild (class II) limitation of physical activity, with symptoms occurring only during ordinary activity (class II) but not at rest.The cardiac function class I–II criterion was applied to ensure safety and comparability of CR recommendations; patients with more severe functional limitation (eg., NYHA class III–IV) often require closely supervised, individualized rehabilitation and may have higher exercise risk); (3) Age ≥18 years, with good comprehension and communication skills; (4) Clear consciousness and normal expression ability, ensuring smooth communication;
Exclusion criteria: (1) Co-morbid severe impairment of other major organs (eg., liver, lung, kidney) or other serious chronic diseases (eg., malignant tumors); (2) Major perioperative complications or acute clinical instability, such as cardiogenic shock, malignant arrhythmias, or acute heart failure (Killip Class III–IV); (3) Significant life-stress events during hospitalization (eg., loss of a family member) that could severely bias self-reported psychological and behavioral scores.; (4) PCI performed for acute coronary syndromes (ACS; eg., STEMI, NSTEMI, or unstable angina) or emergency PCI, and any revascularization other than PCI (eg., CABG).Therefore, the study population represented post-PCI patients with stable CAD only, minimizing confounding due to differences between elective and ACS-related PCI.
The Definition and Delivery of Phase I (Inpatient) Cardiac Rehabilitation (CR)
Phase I (inpatient) CR was delivered as a standardized clinical pathway for patients after elective PCI. After hemodynamic stability was confirmed by the treating team, bedside guidance and education were initiated within 24 hours. Importantly, this Phase I program consisted of individualized bedside counseling and supervised mobilization during routine care.
The Phase I CR pathway included three core components:
Early mobilization: a stepwise protocol progressed from bedside sitting to assisted standing and corridor ambulation (Levels 1–4) as tolerated. Intensity was monitored to maintain a heart-rate increase <20 bpm above resting and a Borg rating of perceived exertion (RPE) ≤11.
Standardized education: one-to-one sessions covered risk-factor modification, adherence to dual antiplatelet therapy (DAPT), medication safety, and recognition of warning symptoms (eg., chest pain or dyspnea).
Discharge planning: before discharge, patients received individualized home-based activity/exercise advice and, when appropriate, referral information for Phase II outpatient rehabilitation.
During the index hospitalization (typically 3–5 days), the intended minimum exposure included at least two mobilization contacts per day and at least three structured education contacts, documented in routine nursing/medical records, to promote consistent delivery of early rehabilitation guidance across participants.
Ethical Approval Statement
The study was conducted in accordance with the Declaration of Helsinki and approved by the Medical Ethics Committee of Lishui People’s Hospital (Approval No.: 2025 Research (023–01); Approved on June 10, 2025). All eligible participants were informed of the study’s objectives, procedures, potential risks, and their right to withdraw at any stage without prejudice to their subsequent medical treatment. Written informed consent was obtained from all patients prior to the commencement of data collection.Given that the study population included older adults and individuals with varying levels of health literacy, specific measures were implemented to safeguard voluntariness and privacy:Privacy Protection: Surveys were conducted in a private, quiet setting within the cardiovascular ward to prevent interference from hospital staff or other patients.Researcher Assistance: For patients with limited reading ability or digital literacy, researchers provided objective, non-leading assistance in clarifying terminology to ensure comprehension while minimizing social desirability bias.Confidentiality: All collected data were anonymized and accessible only to the primary research team.
Instruments
General Information Questionnaire
A self-made questionnaire was used to collect data on the subjects’ gender, age, education, place of residence, marital status, average family monthly income, health insurance type, complication type, etc.
Cardiac Rehabilitation Adherence Survey
The Treatment Adherence Questionnaire for post-PCI patients compiled by Liu Yan10 was used to measure patient CR adherence, including three dimensions: medication adherence, follow-up adherence, and lifestyle adherence. As this was a cross-sectional survey during hospitalization, follow-up adherence was not included. Medication adherence included 4 items, scored 1–4 from “cannot do at all” to “completely do,” with a total score of 4–16. A score of 16 was considered complete adherence, and <16 was partial adherence. Lifestyle adherence included exercise adherence and non-exercise lifestyle adherence. Exercise adherence included 1 item: exercising 3–5 times a week was complete adherence, <3 times was partial adherence, and no regular exercise was non-adherence. Non-exercise lifestyle adherence included 8 items, scored 1–2 from “No” to “Yes,” with a total score of 8–16. A score of 16 was complete adherence, and <16 was partial adherence. Based on the adherence level of each dimension, scores were reassigned: non-adherence (1 point), partial adherence (2 points), and complete adherence (3 points). The scores for medication adherence, exercise adherence, and non-exercise lifestyle adherence were summed to obtain the total CR adherence score(Table S1). The Cronbach’s α coefficient for this questionnaire was 0.875. In this study, the Cronbach’s α of the questionnaire was 0.860, indicating good internal consistency. Among them, the Cronbach’s α for medication compliance was 0.881, and the Cronbach’s α for lifestyle compliance was 0.813.
Data Collection
Data collection was conducted during the patient hospitalization period (days 1–7 post-PCI) by a team of trained clinical researchers. To ensure data quality and the high recovery rate observed, the following standardized procedures were implemented: Researchers screened the cardiovascular ward daily for eligible patients based on the predefined inclusion and exclusion criteria. Once identified, patients were approached during a period of clinical stability (typically 48 hours post-procedure) to explain the study’s objectives and provide written informed consent. The survey was administered using a paper-based format in a private, quiet setting within the ward to minimize external interference and protect participant privacy. To control for social desirability bias, participants were repeatedly assured that their responses would remain anonymous and would not influence their subsequent medical care or relationship with the clinical staff. Upon completion, researchers immediately checked the questionnaires for missing items; if any were identified, patients were asked if they wished to provide a response, thereby eliminating data incompleteness at the source. All data were double-entered into a secure database by two independent researchers to ensure accuracy. During recruitment, 410 eligible patients were screened; 26 declined participation after being informed about the study and signing was not obtained, leaving 384 participants who provided written informed consent and completed the questionnaire on site (effective response rate among enrolled participants: 100%.
Statistical Analysis
Using the 13 items of the treatment adherence questionnaire as manifest indicators (recoded to ordered categorical scores as described above), Mplus 8.3 software was used to fit K latent profile models, starting from a single profile. LPA is a person-centered finite mixture modeling approach and was selected to capture unobserved heterogeneity in CR adherence behaviors and to identify clinically meaningful adherence profiles that may not be apparent from overall adherence scores. The Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), sample-size adjusted BIC (aBIC), Entropy, and Lo-Mendell-Rubin likelihood ratio test (LMRT) were used to evaluate the model’s classification effectiveness. Lower AIC and BIC values indicate a better model fit; an Entropy value closer to 1 indicates more precise classification; an LMRT P<0.05 indicates that a K-profile model is superior to a (K-1)-profile model.11 Additionally, each profile should have some differentiation on the classification indicators, and the proportion of people in each class should be >5%.12
In the primary mixture model, the 13 adherence indicators were treated as ordered categorical variables (three-level or binary after recoding) and analyzed in Mplus using a finite mixture model (TYPE = MIXTURE) with the robust weighted least squares estimator (WLSMV). Thresholds were freely estimated across classes, and within-class residual associations were fixed to zero to satisfy local independence. For missing indicator responses, Mplus used all available data under WLSMV (pairwise present) assuming missing at random. As a sensitivity analysis, we refit the model treating the recoded indicators as continuous and estimated parameters using robust maximum likelihood (MLR) to assess whether class enumeration and profile patterns were robust to indicator scaling.
SPSS 27.0 was used for statistical analysis, with LASSO variable selection and AUC estimation conducted in R (eg., glmnet and pROC). Categorical variables were summarized as n (%) and compared using the χ2-test (or Fisher’s exact test when appropriate). Ordinal variables were analyzed using the Mann–Whitney U-test or Kruskal–Wallis H-test. Continuous variables were assessed for normality using the Shapiro–Wilk test and Q–Q plots; normally distributed variables were described as mean ± standard deviation (
) and compared using one-way ANOVA, whereas non-normally distributed variables were described as median (interquartile range) and compared using the Kruskal–Wallis H-test. To identify factors associated with adherence, a multinomial logistic regression model was fitted. Candidate predictors were first screened using least absolute shrinkage and selection operator (LASSO) with k-fold cross-validation, and variables retained by LASSO were entered into the final multinomial logistic regression mo model fit was evaluated using the Akaike information criterion (AIC). Model discrimination was assessed by receiver operating characteristic (ROC) analysis, reporting one-vs-rest area under the curve (AUC) values for each adherence category and a macro-averaged AUC. A two-sided P value < 0.05 was considered statistically significant. Potential effect modification was explored by testing a priori plausible interaction terms in the multinomial model, including age × smartphone proficiency, age × residential distance, income × residential distance, education × smartphone proficiency, gender × smartphone proficiency, and living status × smartphone proficiency. Where sparse cells occurred, living status was collapsed (living alone vs living with others) for interaction testing. Interactions were evaluated using likelihood-ratio tests comparing models with and without the interaction term, and findings were treated as exploratory/sensitivity analyses to minimize overfitting.
Results
General Clinical and Sociodemographic Characteristics of the Participants
A total of 384 inpatients who underwent PCI and participated in Phase I CR were included in this study (Table 1).
Table 1.
Baseline Characteristics by CR Adherence Profile (n = 384)
| Characteristic | Overall (n=384) | Low (n=125) | Medium (n=169) | High (n=90) | P value |
|---|---|---|---|---|---|
| Gender | <0.001 | ||||
| Male | 246 (64.06%) | 90 (72.00%) | 117 (69.23%) | 39 (43.33%) | |
| Female | 138 (35.94%) | 35 (28.00%) | 52 (30.77%) | 51 (56.67%) | |
| Age group (years) | 0.820 | ||||
| <60 | 136 (35.42%) | 46 (36.80%) | 62 (36.69%) | 28 (31.11%) | |
| 60–69 | 130 (33.85%) | 41 (32.80%) | 53 (31.36%) | 36 (40.00%) | |
| 70–74 | 55 (14.32%) | 16 (12.80%) | 25 (14.79%) | 14 (15.56%) | |
| ≥75 | 62 (16.15%) | 22 (17.60%) | 28 (16.57%) | 12 (13.33%) | |
| Education level | <0.001 | ||||
| Primary or below | 208 (54.17%) | 73 (58.40%) | 91 (53.85%) | 44 (48.89%) | |
| Junior high | 105 (27.34%) | 39 (31.20%) | 49 (28.99%) | 17 (18.89%) | |
| High school/Jr. college | 43 (11.20%) | 6 (4.80%) | 22 (13.02%) | 15 (16.67%) | |
| College or above | 28 (7.29%) | 7 (5.60%) | 7 (4.14%) | 14 (15.56%) | |
| Living situation | <0.001 | ||||
| Living alone | 43 (11.20%) | 27 (21.60%) | 11 (6.51%) | 5 (5.56%) | |
| With friends/colleagues | 32 (8.33%) | 9 (7.20%) | 20 (11.83%) | 3 (3.33%) | |
| With family | 309 (80.47%) | 89 (71.20%) | 138 (81.66%) | 82 (91.11%) | |
| Family monthly income per capita | <0.001 | ||||
| <2000 yuan | 43 (11.20%) | 24 (19.20%) | 14 (8.28%) | 5 (5.56%) | |
| 2000–5000 yuan | 222 (57.81%) | 75 (60.00%) | 107 (63.31%) | 40 (44.44%) | |
| 5000–10,000 yuan | 93 (24.22%) | 23 (18.40%) | 43 (25.44%) | 27 (30.00%) | |
| ≥10,000 yuan | 26 (6.77%) | 3 (2.40%) | 5 (2.96%) | 18 (20.00%) | |
| Residential distance to tertiary hospital (km) | 0.123 | ||||
| ≤3 | 252 (65.62%) | 76 (60.80%) | 110 (65.09%) | 66 (73.33%) | |
| 3–5 | 132 (34.11%) | 49 (39.20%) | 59 (34.91%) | 24 (25.56%) | |
| 5–10 | 0 (0.00%) | 0 (0.00%) | 0 (0.00%) | 0 (0.00%) | |
| ≥10 | 0 (0.00%) | 0 (0.00%) | 0 (0.00%) | 0 (0.00%) | |
| Transportation to hospital | 0.066 | ||||
| Driving | 227 (59.11%) | 64 (51.20%) | 99 (58.58%) | 64 (71.11%) | |
| Public transport | 97 (25.26%) | 37 (29.60%) | 46 (27.22%) | 14 (15.56%) | |
| E-bike | 37 (9.64%) | 17 (13.60%) | 16 (9.47%) | 4 (4.44%) | |
| Walking | 13 (3.39%) | 4 (3.20%) | 5 (2.96%) | 4 (4.44%) | |
| Other | 10 (2.60%) | 3 (2.40%) | 3 (1.78%) | 4 (4.44%) | |
| Payment type | 0.002 | ||||
| Medical insurance | 360 (93.75%) | 109 (87.20%) | 165 (97.63%) | 86 (95.56%) | |
| Self-pay | 17 (4.43%) | 13 (10.40%) | 2 (1.18%) | 2 (2.22%) | |
| Other | 7 (1.82%) | 3 (2.40%) | 2 (1.18%) | 2 (2.22%) | |
| Preferred rehabilitation mode | <0.001 | ||||
| Home-based | 287 (74.74%) | 108 (86.40%) | 129 (76.33%) | 50 (55.56%) | |
| In-hospital | 97 (25.26%) | 17 (13.60%) | 40 (23.67%) | 40 (44.44%) | |
| Smartphone proficiency | 0.030 | ||||
| Very proficient | 98 (25.52%) | 26 (20.80%) | 43 (25.44%) | 29 (32.22%) | |
| Proficient | 108 (28.12%) | 31 (24.80%) | 45 (26.63%) | 32 (35.56%) | |
| Not very proficient | 109 (28.39%) | 46 (36.80%) | 44 (26.04%) | 19 (21.11%) | |
| Not proficient at all | 69 (17.97%) | 22 (17.60%) | 37 (21.89%) | 10 (11.11%) | |
| Procedure-related variables | |||||
| Number of diseased vessels | 1.00 (1.00, 1.00) | 1.00 (1.00, 1.00) | 1.00 (1.00, 1.00) | 1.00 (1.00, 1.00) | 0.738 |
| Laboratory indicators | |||||
| LVEF (%) | 51.75 (49.38, 53.70) | 51.80 (48.70, 53.70) | 51.50 (49.50, 53.20) | 52.10 (49.38, 53.90) | 0.599 |
| LDL-C (mmol/L) | 1.40 (1.20, 1.60) | 1.40 (1.30, 1.60) | 1.40 (1.20, 1.60) | 1.50 (1.30, 1.70) | 0.285 |
| Serum creatinine, Scr (µmol/L) | 93.35 (81.22, 148.97) | 95.70 (81.30, 148.70) | 90.10 (80.60, 135.90) | 115.35 (82.50, 164.73) | 0.227 |
| Past history: Hypertension | 0.447 | ||||
| Yes | 240 (62.50%) | 78 (62.40%) | 101 (59.76%) | 61 (67.78%) | |
| No | 144 (37.50%) | 47 (37.60%) | 68 (40.24%) | 29 (32.22%) | |
| Past history: Diabetes mellitus | 0.352 | ||||
| Yes | 138 (35.94%) | 45 (36.00%) | 66 (39.05%) | 27 (30.00%) | |
| No | 246 (64.06%) | 80 (64.00%) | 103 (60.95%) | 63 (70.00%) | |
| Past history: Hyperlipidemia | 0.338 | ||||
| Yes | 14 (3.65%) | 7 (5.60%) | 4 (2.37%) | 3 (3.33%) | |
| No | 370 (96.35%) | 118 (94.40%) | 165 (97.63%) | 87 (96.67%) | |
| Past history: Current smoking | <0.001 | ||||
| Yes | 220 (57.29%) | 86 (68.80%) | 107 (63.31%) | 27 (30.00%) | |
| No | 164 (42.71%) | 39 (31.20%) | 62 (36.69%) | 63 (70.00%) |
Note: Values are n (%) for categorical variables and mean ± SD or median (IQR) for continuous variables, depending on normality (Shapiro–Wilk test). P values were derived from χ2-tests, one-way ANOVA, or Kruskal–Wallis tests.
Sociodemographic Profile
Overall, 246 participants were male (64.06%). The age distribution was: <60 years 136 (35.42%), 60–69 years 130 (33.85%), 70–74 years 55 (14.32%), and ≥75 years 62 (16.15%). Regarding education, 208 (54.17%) had primary school or below, 105 (27.34%) junior high, 43 (11.20%) high school/junior college, and 28 (7.29%) college or above.In terms of smartphone proficiency, 98 (25.52%) were very proficient, 108 (28.12%) proficient, 109 (28.39%) not very proficient, and 69 (17.97%) not proficient at all.
Clinical Characteristics
Procedure-related data showed a median number of diseased vessels of 1.00 (1.00, 1.00). Laboratory indicators were as follows: median LVEF 51.75% (49.38, 53.70), median LDL-C 1.40 mmol/L (1.20, 1.60), and median serum creatinine (Scr) 93.35 µmol/L (81.22, 148.97).Common cardiovascular risk factors were prevalent: hypertension in 240 (62.50%), diabetes mellitus in 138 (35.94%), hyperlipidemia in 14 (3.65%), and current smoking in 220 (57.29%).
CR Adherence in Post-PCI Patients
The results of this study showed that the medication adherence score of post-PCI patients was (13.93±2.28), the lifestyle adherence score was (12.91±1.17), and the exercise adherence score was (1.72±0.71). Scores for each item are shown in Table 2 and the mean scores for each adherence domain across the three profiles are visualized in Figure 1.
Table 2.
Scores for CR Adherence Items in Post-PCI Patients (X±s)
| Item | Content | Dimension | Score |
|---|---|---|---|
| 1 | Can you take your medication according to the frequency prescribed by your doctor? | Medication Adherence | 3.50±0.66 |
| 2 | Can you take your medication according to the dosage prescribed by your doctor? | Medication Adherence | 3.64±0.62 |
| 3 | Can you take your medication at the times prescribed by your doctor? | Medication Adherence | 3.32±0.65 |
| 4 | After intervention, can you adhere to long-term medication as prescribed without interruption? | Medication Adherence | 3.48±0.65 |
| 5 | Do you smoke? | Non-exercise Lifestyle Adherence | 1.44±0.50 |
| 6 | Have you quit smoking for 6 months or more? | Non-exercise Lifestyle Adherence | 1.72±0.45 |
| 7 | Do you drink more than 1 liang (50g) of white wine or spirits daily? | Non-exercise Lifestyle Adherence | 1.25±0.43 |
| 8 | Have you already quit drinking? | Non-exercise Lifestyle Adherence | 1.85±0.35 |
| 9 | Eat no more than 100g (2 liang) of lean meat and other meat products daily. | Non-exercise Lifestyle Adherence | 1.78±0.41 |
| 10 | Eat no more than 25g (0.5 liang) of cooking oil daily. | Non-exercise Lifestyle Adherence | 1.77±0.42 |
| 11 | Consume 250g (5 liang) of dairy products daily. | Non-exercise Lifestyle Adherence | 1.30±0.46 |
| 12 | Eat 400–500g (8 liang - 1 jin) of fresh vegetables and 100g (2 liang) of fruit daily. | Non-exercise Lifestyle Adherence | 1.79±0.41 |
| 13 | Do you adhere to suitable exercise? (Req: ≥3 times/week, ≥20 min/time) | Lifestyle Adherence | 1.72±0.71 |
Note: Values are presented as mean ± standard deviation (x±s).
Figure 1.
Profile Plots of CR Adherence Groups.
Notes: The x-axis represents the 13 specific items of the cardiac rehabilitation adherence questionnaire (including medication, exercise, and lifestyle adherence domains). The y-axis represents the mean score or probability corresponding to each item across the three latent profiles.
Latent Profile Analysis of Post-PCI CR Adherence
The number of latent profiles (n) for In-hospital CR Adherence was set from 1 to 5. LPA model fitting showed that AIC, BIC, and aBIC decreased as the number of classes increased. When n = 3, Entropy = 0.987, and the P-values for both LMRT and Bootstrap Likelihood Ratio Test (BLRT) were <0.05. When n = 4, Entropy = 0.905, but the P-values for LMR and BLRT were both >0.05, indicating that selecting 3 latent classes for post-PCI CR adherence provided the optimal model fit. See Table 3. Beyond statistical fit, we considered clinical interpretability and class stability. Although AIC/BIC/aBIC continued to decrease with additional classes, the 4-class solution did not yield a clinically distinct adherence pattern and mainly reflected a minor split of an existing profile, as supported by the domain-level mean score patterns (Table S2) and the model-fit trends shown in the Supplementary Material (Figures S1–S4). Therefore, the 3-class model was retained as the most parsimonious and interpretable solution. C1 had 125 cases (32.55%), C2 had 169 cases (44.01%), and C3 had 90 cases (23.44%). C1 patients had the lowest scores on all CR adherence items among the 3 classes and were named the “Low CR Adherence Type”; C2 patients had medium scores and were named the “Medium CR Adherence Type”; C3 patients had the highest scores and were named the “High CR Adherence Type”.
Table 3.
LPA Results for Post-PCI CR Adherence (n = 384)
| Model | AIC | BIC | aBIC | Entropy | P | Class Probability (%) | |
|---|---|---|---|---|---|---|---|
| LMRT | BLRT | ||||||
| 1 | −644.844 | −621.141 | −640.178 | ||||
| 2 | −1126.407 | −1086.900 | −1118.629 | 1.000 | 0.057 | <0.001 | 49.48/50.52 |
| 3 | −1181.702 | −1126.393 | −1170.813 | 0.987 | <0.001 | <0.001 | 50.52/30.99/18.49 |
| 4 | −1313.055 | −1241.944 | −1299.055 | 0.905 | 0.094 | 0.097 | 22.40/17.19/28.13/32.29 |
| 5 | −1412.702 | −1319.788 | −1322.591 | 0.887 | 0.512 | 0.704 | 20.40/16.49/35.60/17.08/10.43 |
Note: Class probability (%) indicates the estimated proportion of participants in each latent class.
Univariate Analysis of Different Post-PCI CR Adherence Classifications (Table 4)
Table 4.
Univariate Analysis of Different Post-PCI CR Adherence Profiles (n = 384)
| Item | Category | Overall (n=384) | Low (n=125) | Medium (n=169) | High (n=90) | Statistic | P value |
|---|---|---|---|---|---|---|---|
| Gender | Male | 246 (64.06%) | 90 (72.00%) | 117 (69.23%) | 39 (43.33%) | χ2 =22.179 | <0.001 |
| Female | 138 (35.94%) | 35 (28.00%) | 52 (30.77%) | 51 (56.67%) | χ2 =22.179 | <0.001 | |
| Age Group | <60years | 130 (33.85%) | 47 (37.60%) | 63 (37.28%) | 20 (22.22%) | χ2 =17.915 | 0.006 |
| 60-69years | 116 (30.21%) | 41 (32.80%) | 53 (31.36%) | 22 (24.44%) | χ2 =17.915 | 0.006 | |
| 70-74years | 68 (17.71%) | 16 (12.80%) | 25 (14.79%) | 27 (30.00%) | χ2 =17.915 | 0.006 | |
| ≥75 years | 70 (18.23%) | 21 (16.80%) | 28 (16.57%) | 21 (23.33%) | χ2 =17.915 | 0.006 | |
| Education | Primary or below | 208 (54.17%) | 73 (58.40%) | 91 (53.85%) | 44 (48.89%) | χ2 =22.772 | <0.001 |
| Junior high | 105 (27.34%) | 39 (31.20%) | 49 (28.99%) | 17 (18.89%) | χ2 =22.772 | <0.001 | |
| High school / Jr. college | 43 (11.20%) | 6 (4.80%) | 22 (13.02%) | 15 (16.67%) | χ2 =22.772 | <0.001 | |
| College or above | 28 (7.29%) | 7 (5.60%) | 7 (4.14%) | 14 (15.56%) | χ2 =22.772 | <0.001 | |
| Living Situation | Living alone | 43 (11.20%) | 27 (21.60%) | 11 (6.51%) | 5 (5.56%) | χ2 =25.965 | <0.001 |
| With friends/colleagues | 32 (8.33%) | 9 (7.20%) | 20 (11.83%) | 3 (3.33%) | χ2 =25.965 | <0.001 | |
| With family | 309 (80.47%) | 89 (71.20%) | 138 (81.66%) | 82 (91.11%) | χ2 =25.965 | <0.001 | |
| Avg. Monthly Income | <2000 yuan | 43 (11.20%) | 24 (19.20%) | 14 (8.28%) | 5 (5.56%) | χ2=48.270 | <0.001 |
| 2000-5000 (excl.) | 222 (57.81%) | 75 (60.00%) | 107 (63.31%) | 40 (44.44%) | χ2=48.270 | <0.001 | |
| 5000-10k (excl.) | 96 (25.00%) | 23 (18.40%) | 43 (25.44%) | 27 (30.00%) | χ2=48.270 | <0.001 | |
| ≥10k yuan | 26 (6.77%) | 3 (2.40%) | 5 (2.96%) | 18 (20.00%) | χ2=48.270 | <0.001 | |
| CR Service in Area | Yes | 190 (49.48%) | 56 (44.80%) | 78 (46.15%) | 56 (62.22%) | χ2 =7.689 | 0.021 |
| No | 194 (50.52%) | 69 (55.20%) | 91 (53.85%) | 34 (37.78%) | χ2 =7.689 | 0.021 | |
| Distance to Hospital | ≤3 km | 90 (23.44%) | 21 (16.80%) | 40 (23.67%) | 29 (32.22%) | χ2 =14.822 | 0.022 |
| 3-5 km (excl.) | 92 (23.96%) | 38 (30.40%) | 38 (22.49%) | 16 (17.78%) | χ2 =14.822 | 0.022 | |
| 5-10 km (excl.) | 70 (18.23%) | 17 (13.60%) | 32 (18.93%) | 21 (23.33%) | χ2 =14.822 | 0.022 | |
| ≥10 km | 131 (34.11%) | 49 (39.20%) | 59 (34.91%) | 23 (25.56%) | χ2 =14.822 | 0.022 | |
| Transportation | Driving | 227 (59.11%) | 64 (51.20%) | 99 (58.58%) | 64 (71.11%) | χ2 =14.667 | 0.047 |
| Public transport | 97 (25.26%) | 37 (29.60%) | 46 (27.22%) | 14 (15.56%) | χ2 =14.667 | 0.047 | |
| E-bike | 37 (9.64%) | 17 (13.60%) | 16 (9.47%) | 4 (4.44%) | χ2 =14.667 | 0.047 | |
| Walking | 13 (3.39%) | 4 (3.20%) | 5 (2.96%) | 4 (4.44%) | χ2 =14.667 | 0.047 | |
| Other | 10 (2.60%) | 3 (2.40%) | 3 (1.78%) | 4 (4.44%) | χ2 =14.667 | 0.047 | |
| Payment Type | Medical insurance | 360 (93.75%) | 109 (87.20%) | 165 (97.63%) | 86 (95.56%) | χ2 =16.639 | 0.002 |
| Self-pay | 17 (4.43%) | 13 (10.40%) | 2 (1.18%) | 2 (2.22%) | χ2 =16.639 | 0.002 | |
| Other | 7 (1.82%) | 3 (2.40%) | 2 (1.18%) | 2 (2.22%) | χ2 =16.639 | 0.002 | |
| Rehab Method | Home-based | 287 (74.74%) | 108 (86.40%) | 129 (76.33%) | 50 (55.56%) | χ2 =26.773 | <0.001 |
| In-hospital | 97 (25.26%) | 17 (13.60%) | 40 (23.67%) | 40 (44.44%) | χ2 =26.773 | <0.001 | |
| Smartphone Proficiency | Very proficient | 98 (25.52%) | 26 (20.80%) | 43 (25.44%) | 29 (32.22%) | χ2 =14.007 | 0.030 |
| Proficient | 108 (28.12%) | 31 (24.80%) | 45 (26.63%) | 32 (35.56%) | χ2 =14.007 | 0.030 | |
| Not very proficient | 109 (28.39%) | 46 (36.80%) | 44 (26.04%) | 19 (21.11%) | χ2 =14.007 | 0.030 | |
| Not proficient at all | 69 (17.97%) | 22 (17.60%) | 37 (21.89%) | 10 (11.11%) | χ2 =14.007 | 0.030 | |
| Number of diseased vessels | 1.18±0.42 | 1.19±0.49 | 1.15±0.36 | 1.20±0.43 | F=0.467 | 0.627 | |
| LVEF (%) | 51.43±2.68 | 51.35±2.79 | 51.37±2.62 | 51.63±2.65 | F=0.333 | 0.717 | |
| LDL-C (mmol/L) | 1.44±0.24 | 1.43±0.24 | 1.43±0.23 | 1.48±0.23 | F=1.291 | 0.276 | |
| Serum creatinine (umol/L) | 122.26±59.49 | 121.23±58.15 | 119.20±60.47 | 129.45±59.52 | F=0.898 | 0.408 | |
| Adherence Score | 2.83±0.78 | 2.98±0.80 | 3.68±0.50 | F=38.361 | <0.001 | ||
| Healthy Diet Score | 1.36±0.36 | 1.45±0.38 | 1.66±0.38 | F=16.802 | <0.001 |
Note: Values are n (categorical variables) or mean ± standard deviation (continuous variables). P < 0.05 indicates statistical significance.
Univariate comparisons across adherence profiles were performed to describe between-profile differences and to provide context for subsequent model building.Because multivariable modeling was the primary analytical approach, these univariate results are interpreted as descriptive screening rather than independent evidence of association(Table 4).
Multinomial Logistic Regression Analysis of Factors Associated with CR Adherence Profiles
With post-PCI CR adherence profile membership as the dependent variable (High Adherence Type as the reference group), candidate predictors were screened using LASSO with cross-validation, and variables retained by LASSO were entered as independent variables in the multinomial logistic regression model.Independent variable coding is shown in Table 5. The results showed that gender, living situation, average family monthly income, residential distance, smartphone proficiency and LDL-C were significantly associated with post-PCI CR adherence profile membership (P < 0.05), see Table 6. The discriminative ability of the multinomial logistic regression model was evaluated using one-vs-rest (OVR) receiver operating characteristic (ROC) analysis. The AUCs were 0.653 for the Low-adherence profile, 0.613 for the Medium-adherence profile, and 0.758 for the High-adherence profile, with a macro-average AUC of 0.675 (weighted-average AUC: 0.660). Overall, the final model incorporating sociodemographic factors and selected clinical indicators (retained after LASSO) demonstrated moderate discrimination in distinguishing adherence profiles during the Phase I inpatient recovery period (Table S3).Exploratory interaction analyses indicated that the interaction between living status (alone vs living with others) and smartphone proficiency improved model fit (likelihood-ratio P = 0.039), whereas other tested interactions (age × smartphone proficiency, age × distance, income × distance, education × smartphone proficiency, and gender × smartphone proficiency) were not statistically significant (all P > 0.05). Given the exploratory nature of interaction testing and to reduce overfitting, the primary results are presented from the main-effects mo the interaction finding is reported as a sensitivity analysis (Table S4).
Table 5.
Independent Variable Assignments
| Variable | Assignment |
|---|---|
| Gender | Male=1, Female=2 |
| Age | Reference: >75 years. Dummy variables: <60 years (1,0,0), 60–69 years (0,1,0), 70–74 years (0,0,1) |
| Education | Reference: College or above. Dummy variables: Primary or below (1,0,0), Junior high (0,1,0), High school/Jr. college (0,0,1) |
| Living Situation | Reference: With family. Dummy variables: Living alone (1,0), With friends/colleagues (0,1) |
| Avg. Monthly Income | Reference: ≥10k yuan. Dummy variables: <2000 yuan (1,0,0), 2000–5000 (excl.) (0,1,0), 5000–10k (excl.) (0,0,1) |
| CR Service in Area | Yes=1, No=2 |
| Residential Distance | Reference: ≥10 km. Dummy variables: ≤3 km (1,0,0), 3–5 km (excl.) (0,1,0), 5–10 km (excl.) (0,0,1) |
| Transportation | Reference: Other. Dummy variables: Driving (1,0,0,0), Public transport (0,1,0,0), E-bike (0,0,1,0), Walking (0,0,0,1) |
| Payment Type | Reference: Other. Dummy variables: Medical insurance (1,0), Self-pay (0,1) |
| Rehab Method | Home-based=1, In-hospital=2 |
| Smartphone Proficiency | Reference: Not proficient at all. Dummy variables: Very proficient (1,0,0), Proficient (0,1,0), Not very proficient (0,0,1) |
| Number of diseased vessels | Entered as original value |
| LDL-C (mmol/L) | Entered as original value |
| Serum creatinine (umol/L) | Entered as original value |
Abbreviations: Ref, reference category in regression models; Excl, excluding.
Table 6.
Multinomial Logistic Regression Analysis of Factors Associated with Post-PCI CR Adherence Profiles
| Predictor | Comparison (ref=High) | B | SE | Wald χ2 | OR | 95% CI (Lower) | 95% CI (Upper) | P |
|---|---|---|---|---|---|---|---|---|
| Gender | Low vs High | −1.186 | 0.320 | 13.740 | 0.305 | 0.163 | 0.572 | <0.001 |
| Living situation | Low vs High | 0.601 | 0.304 | 3.903 | 1.824 | 1.005 | 3.312 | 0.048 |
| Family monthly Income | Low vs High | −1.170 | 0.228 | 26.317 | 0.310 | 0.198 | 0.485 | <0.001 |
| Residential distance (km) | Low vs High | 0.593 | 0.338 | 3.076 | 1.809 | 0.933 | 3.507 | 0.079 |
| Smartphone use | Low vs High | 0.398 | 0.151 | 6.916 | 1.489 | 1.107 | 2.003 | 0.009 |
| Diseased vessels | Low vs High | −0.040 | 0.340 | 0.014 | 0.961 | 0.494 | 1.871 | 0.907 |
| LDL_C(mmolL) | Low vs High | −1.473 | 0.694 | 4.508 | 0.229 | 0.059 | 0.893 | 0.034 |
| Gender | Medium vs High | −1.057 | 0.290 | 13.275 | 0.348 | 0.197 | 0.614 | <0.001 |
| Living situation | Medium vs High | 0.478 | 0.293 | 2.658 | 1.614 | 0.908 | 2.868 | 0.103 |
| Family monthly Income | Medium vs High | −0.770 | 0.200 | 14.837 | 0.463 | 0.313 | 0.685 | <0.001 |
| Residential distance (km) | Medium vs High | 0.481 | 0.317 | 2.304 | 1.618 | 0.869 | 3.010 | 0.129 |
| Smartphone use | Medium vs High | 0.348 | 0.141 | 6.126 | 1.416 | 1.075 | 1.866 | 0.013 |
| Diseased vessels | Medium vs High | −0.296 | 0.326 | 0.824 | 0.744 | 0.393 | 1.409 | 0.364 |
| LDL_C(mmolL) | Medium vs High | −1.383 | 0.642 | 4.642 | 0.251 | 0.071 | 0.883 | 0.031 |
Note: High-adherence profile was the reference group. Predictors were selected using LASSO with 10-fold cross-validation and entered into the final multinomial model. Candidate predictors were screened using LASSO; clinical variables (No. diseased vessels, and LDL-C) were retained and entered as continuous predictors in the final model.
Abbreviations: B, regression coefficient; SE, standard error; Wald χ2 = (B/SE)2; OR, odds ratio; CI, confidence interval.
Discussion
Current Status of CR Adherence in Post-PCI Patients
The present study identified three distinct adherence phenotypes among patients during the acute Phase I (in-hospital) recovery period following PCI. Our findings reveal that over half of the participants belonged to the Low Adherence profile, suggesting that suboptimal engagement with cardiac rehabilitation begins as early as the inpatient stage. This result is consistent with previous research indicating that the immediate post-procedural period is a window of significant behavioral transition and psychological stress, which can hinder the proactive adoption of rehabilitation protocols.7,9,13
CR involves multiple aspects such as medication management, exercise training, diet, and psychological adjustment. Its comprehensive and long-term nature makes it difficult for patients to adhere. This study found that medication adherence was relatively high, while exercise adherence was significantly low. One possible explanation is that some patients may have an insufficient understanding of the benefits of exercise or may fear exercise-induced cardiac events, which could be linked to avoidance behaviors. This aligns with the Health Belief Model.14 Accordingly, clinical education may prioritize building patients’ confidence and perceived safety regarding exercise through individualized, gradually progressive plans and appropriate counseling; however, the effectiveness of such strategies may be verified in future intervention studies. Notably, among the added clinical indicators, LDL-C remained independently associated with adherence profile membership, suggesting that residual lipid risk and overall cardiovascular risk management may be linked to patients’ rehabilitation behaviors. In contrast, the number of diseased vessels showed no statistically significant association in the final model, indicating that angiographic burden alone may be insufficient to explain adherence heterogeneity during the Phase I recovery period. LVEF and serum creatinine were evaluated as candidate variables but were not retained after LASSO selection in this dataset, which may reflect the relatively preserved functional status of the included patients and limited between-profile separation in these measures.
Furthermore, although medication adherence is ideal during hospitalization, long-term maintenance remains difficult,15 suggesting the potential value of structured post-discharge follow-up and support. In terms of lifestyle, patients performed better in smoking and alcohol cessation, but dietary adherence was generally low.A large-scale US cohort study further showed that diet control is the weakest link in lifestyle interventions,16 while a European study pointed out that patients’ lack of awareness of the delayed effects of dietary intervention may be a key reason.17 Thus, future programs could consider strengthening dietary self-efficacy via tailored guidance and behavior-support components.
Latent Profiles of Post-PCI CR Adherence
Based on LPA, this study classified post-PCI patient CR adherence into three types: “Low CR Adherence Type” (50.52%), “Medium CR Adherence Type” (30.99%), and “High CR Adherence Type” (18.49%). This distribution pattern is similar to the findings of Wang Yuxiu et al18 in elderly patients with chronic diseases and is also consistent with multinational studies including the US, Canada, and Australia.8 This further indicates that patient adherence generally presents a “low-adherence dominant” characteristic. This phenomenon may stem from difficulties in long-term behavior maintenance, fluctuating motivation, and insufficient external support. Accordingly, these profiles may be useful for designing and testing stratified, profile-informed support—particularly for the low-adherence group—in future implementation and intervention studies.
Multidimensional Factors Associated with Adherence Profiles
This study indicates that the rehabilitation compliance of patients after PCI is associated with multiple factors spanning individual characteristics, social context, and care delivery models. Lower educational level may restrict the acquisition and understanding of health information, thereby reducing compliance, which is consistent with existing research.19,20 Older age was also associated with poorer adherence, potentially reflecting declines in cognition and physical capacity; however, age effects may be confounded by comorbidity and functional status. Notably, in the present study, education level and age were not retained in the final LASSO-selected multinomial model, suggesting that their independent contributions may be attenuated after accounting for other contextual and digital-access factors. In terms of digital health literacy, this study reveals that patients with higher digital skills exhibit better compliance, which aligns with the current development trend of digital health management. This finding supports the potential role of remote monitoring, rehabilitation apps, and online education, but it does not by itself establish that digital interventions may improve adherence in all groups.21 For patients with limited digital literacy, simplified tools, training, and technical support may help reduce barriers and narrow the digital divide.22,23 This pattern suggests that smartphone proficiency may act as a key enabler of engagement with digitally supported rehabilitation, shaping how patients access education, reminders, symptom monitoring, and feedback during phase I and after discharge. Accordingly, the design of digital or hybrid CR pathways may benefit from stratifying patients by digital capability: for example, offering simplified interfaces, step-by-step onboarding, caregiver-assisted use, and alternative low-tech options (telephone follow-up or printed action plans) for patients with limited skills, while providing more interactive functions (self-monitoring dashboards, automated reminders, and bidirectional messaging) for digitally capable patients. These implications are hypothesis-generating and may be tested in future intervention studies, but they provide a clinically actionable framework for tailoring CR delivery to reduce digital barriers and improve adherence.22 In addition, evidence from a randomized clinical trial suggests that combining early case management with modest financial incentives can increase CR completion in socioeconomically disadvantaged patients.24
Moreover, smartphone proficiency may interact with contextual barriers such as geographic distance or limited family support: when access to center-based services is constrained, digital tools could partially compensate by enabling remote education and follow-up; conversely, limited digital skills may amplify the negative impact of distance or social isolation on adherence. Exploring such effect modification can help clarify adherence mechanisms and inform targeted resource allocation in digital CR programs.25–27 In particular, for patients who live alone and have limited smartphone proficiency, a pragmatic telephone-based CR pathway (Tele-CR)—including scheduled nurse calls, simplified paper-based action plans/exercise logs, and symptom-triggered escalation—may help maintain engagement while minimizing digital barriers.
At the social support level, “living alone” is identified as an associated factor. This may be due to the lack of daily reminders, emotional support, and behavioral guidance from family members, resulting in the interruption or lax implementation of the rehabilitation plan. Meanwhile, studies in the United States28 indicated that social support is positively correlated with CR compliance. However, a qualitative study in China29 reported mixed findings, suggesting cultural and structural factors may moderate the relationship between perceived support and adherence. Future research could more precisely measure support quality (not only its presence) and examine potential moderators.
In addition, low income and geographic distance were also associated with poorer adherence, plausibly because financial constraints limit transportation, medications, or rehabilitation resources, while longer distance increases time costs for center-based rehabilitation.These associations suggest that improving access (eg., community-based services and scalable home-based/remote options) may be beneficial.30
In terms of clinical intervention, the “in-hospital rehabilitation” model contributes to improving patient compliance due to its professional on-site guidance and supervision. However, considering the limited medical resources, it is challenging to widely implement the long-term in-hospital rehabilitation model. Two studies from the United States31 indicates that remote follow - up and digital support may enhance patients’ rehabilitation adherence and potentially reduce their reliance on traditional in - hospital rehabilitation. Meanwhile, leveraging digital platforms, such as follow - up interventions based on WeChat or dedicated apps, can help improve the compliance of home - based patients and ensure the continuity and effectiveness of the rehabilitation process. Therefore, when resources permit, in - hospital rehabilitation may be organically combined with home - based/remote rehabilitation: implement in - hospital rehabilitation for patients who require intensive supervision, promote user - friendly digital follow - up and community support for patients who can be managed remotely, and improve the rehabilitation accessibility of patients in low - income and remote areas through policy and financial support to promote their long - term health. These delivery options should be framed as hypothesis-generating implications and ideally linked to feasibility planning and prospective evaluation.
Strengths and Limitations
Strengths
This study applies a person-centered approach (Latent Profile Analysis) to characterize adherence phenotypes specifically during the Phase I (in-hospital) cardiac rehabilitation stage. By identifying unobserved subgroups within the acute post-PCI window, we provide a more granular understanding of early behavioral inertia, which is often masked in traditional aggregate analyses. Furthermore, the inclusion of clinical indicators like LVEF and PCI indications enhances the clinical relevance of our findings for cardiovascular nursing and medicine.
Limitations
Despite its strengths, several limitations must be acknowledged:First, Given the cross-sectional nature of the data, the identified associations between digital literacy, clinical factors, and adherence profiles are correlational rather than causal. Our findings should be interpreted as hypothesis-generating rather than definitive evidence of causation.Second, CR adherence was measured using a locally developed Chinese questionnaire (Liu Yan’s Treatment Adherence Questionnaire) rather than a widely used international instrument, which may limit cross-study comparability and external generalizability; further validation against internationally recognized measures and objective indicators is needed. Third, this research was restricted to the 1–7 day post-PCI window. Therefore, the identified profiles represent in-hospital adherence intentions and early behaviors, which may not fully predict long-term Phase II or Phase III outpatient adherence.Fourth, data were primarily obtained through self-reported scales, which may be subject to recall bias and social desirability; future research could incorporate objective and multi-source measures (eg., exercise records from smart bracelets, medication data from electronic pillboxes) to strengthen verification. Additionally, the three-level recoding strategy (especially the wide “partially adherent” interval, eg., 5–15 for medication totals) may have grouped participants with meaningfully different behaviors into the same category, potentially attenuating item/domain variance and affecting profile separation (eg., entropy) and clinical sensitivity. Fifth, as a single-center study, the generalizability of these profiles to different healthcare settings or cultural contexts requires further multi-center validation.In addition, during recruitment, although questionnaires were only distributed to consenting patients, selection bias due to non-participation remains possible.
Conclusions
This study employed latent profile analysis to identify three distinct cardiac rehabilitation adherence profiles—low adherence, moderate adherence, and high adherence—among patients following percutaneous coronary intervention (PCI), highlighting significant heterogeneity in post-PCI rehabilitation behaviors. Profile assignment showed significant associations with sociodemographic and contextual characteristics, including gender, living arrangements, household monthly income, distance to facilities, and smartphone usage. These findings suggest that a one-size-fits-all approach may be inadequate, necessitating stratified, profile-based personalized management in clinical practice. In line with our objective, these results emphasize early (phase I) adherence heterogeneity and the associated contextual correlates within an inpatient setting.
Clinical Implications
Practical interventions can prioritize risk characteristics: Patients with low adherence traits (especially those living alone and/or with low income) should receive priority for home-based and community-linked cardiac rehabilitation (eg., caregiver involvement, community nurse follow-ups, structured health monitoring). Patients facing access barriers due to distant residences may benefit from home-based or community rehabilitation, supplemented by telephone follow-ups for those with limited digital literacy. Conversely, patients proficient in smartphone operation may prioritize digitally supported CR (eg., app/WeChat-based education, reminders, symptom reporting, and remote feedback) to enhance post-discharge treatment continuity. It is particularly important to note that these recommendations aim to guide practical resource allocation based on risk characteristics. They require evaluation through prospective intervention studies. Future multicenter longitudinal intervention studies are needed to validate the stability of these characteristics over time, elucidate underlying mechanisms, and assess the effectiveness and cost-effectiveness of personalized strategies based on these characteristics in improving CR adherence after PCI.
Funding Statement
This research was funded by the Lishui City Science and Technology Project (Key Research and Development Project) (2023zdyf21).
AI Statement
The language editing was carried out using ChatGpt 4.0.
Abbreviations
AIC, Akaike Information Criterion; aBIC, sample-size adjusted Bayesian Information Criterion; BIC, Bayesian Information Criterion; CR, cardiac rehabilitation; LMRT, Lo–Mendell–Rubin likelihood ratio test; LPA, latent profile analysis; PCI, percutaneous coronary intervention.
Data Sharing Statement
The data and materials that support the findings of this study are available from Miss Xingrong He upon reasonable request.
Author Contributions
All authors made a significant contribution to the work reported, whether that is in study design, 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
Xingrong He is the first author, while Jing Wang is the co-first author for this study. The authors declare no conflict of interest in this work.
References
- 1.Han YL. New advances in interventional therapy for coronary heart disease. Clin J Med Off. 2017;45(6):551–15. doi: 10.16680/j.1671-3826.2017.06.01 [DOI] [Google Scholar]
- 2.Liu MB, He XY, Yang XH, et al. Interpretation of the key points of China Cardiovascular Health and Disease Report 2024. Chin J Cardiovasc Med. 2025;30(4):384–399. [Google Scholar]
- 3.Visseren FLJ, Mach F, Smulders YM, et al. 2021 ESC Guidelines on cardiovascular disease prevention in clinical practice. Eur Heart J. 2021;42(34):3227–3337. doi: 10.1093/eurheartj/ehab484 [DOI] [PubMed] [Google Scholar]
- 4.Virani SS, Newby LK, Arnold SV, et al. 2023 AHA/ACC/ACCP/ASPC/NLA/PCNA Guideline for the Management of Patients With Chronic Coronary Disease: executive Summary: a Report of the American Heart Association/American College of Cardiology Joint Committee on Clinical Practice Guidelines. Circulation. 2023;148(9):e9–e119. doi: 10.1161/CIR.0000000000001168 [DOI] [PubMed] [Google Scholar]
- 5.Brown TM, Pack QR, Aberegg E, et al. Core components of cardiac rehabilitation programs: 2024 update: a scientific statement from the American Heart Association and the American Association of Cardiovascular and Pulmonary Rehabilitation. Circulation. 2024;150:e328–e347. doi: 10.1161/CIR.0000000000001289 [DOI] [PubMed] [Google Scholar]
- 6.Ades PA, Keteyian SJ, Balady GJ, et al. Enhancing participation in cardiac rehabilitation: systematic strategies to improve uptake and adherence. Prog Cardiovasc Dis. 2022;70:102–110. doi: 10.1016/j.pcad.2022.01.003 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Du RF, Wang PP, Tian L, et al. Research progress on compliance of cardiac rehabilitation in patients with coronary heart disease. Chin J Nurs. 2017;52(10):1246–1251. [Google Scholar]
- 8.Chai Z, Fan Y, Gong X, et al. Adherence to phase I cardiac rehabilitation in post-PCI patients: a latent class analysis. Front Cardiovasc Med. 2025;12:1460855. doi: 10.3389/fcvm.2025.1460855 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Xia C, Zheng Y, Ji L, et al. Comparative effectiveness of different interventions on adherence to exercise-based CR among patients after percutaneous coronary intervention: a network meta-analysis of randomized controlled trials. BMC Nurs. 2024;23:897. doi: 10.1186/s12912-024-02561-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Liu Y. Correlation Analysis Between Compliance and Social Support in Patients After Percutaneous Coronary Intervention. Beijing: Peking Union Medical College; 2006. [Google Scholar]
- 11.Wenzl X, Xie JY, Wang HH. Principles, procedures and programs of latent class models. J East China Norm Univ. 2023;41(01):1–15. doi: 10.16382/j.cnki.1000-5560.2023.01.001 [DOI] [Google Scholar]
- 12.Tein J-Y, Coxe S, Cham H. Statistical power to detect the correct number of classes in latent profile analysis. Struct Equ Model Multidiscip J. 2013;20:640–657. doi: 10.1080/10705511.2013.824781 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Saarikoski EO, Roiha ETM, Kiviniemi AM, et al. Adherence to exercise after an acute coronary syndrome: a 6-month randomized controlled trial. Front Physiol. 2024;15:1319907. doi: 10.3389/fphys.2024.1319907 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Su YH, Huang Y, Liu LL, et al. The mediating effect of kinesiophobia between health literacy and cardiac rehabilitation barriers in patients with coronary heart disease after PCI. Pract J Card Cereb Pneum Vasc Dis. 2025;33(02):124–129. [Google Scholar]
- 15.Bilicki DJ, Reeves MJ. Outpatient follow-up visits to reduce 30-day all-cause readmissions for heart failure, COPD, myocardial infarction, and stroke: a systematic review and meta-analysis. Prev Chronic Dis. 2024;21E74. doi: 10.5888/pcd21.240138 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Inoue-Choi M, Ramirez Y, Fukunaga A, Matthews CE, Freedman ND. Association of adherence to healthy lifestyle recommendations with all-cause and cause-specific mortality among former smokers. JAMA Network Open. 2022;5(9):e2232778. doi: 10.1001/jamanetworkopen.2022.32778 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Cacau LT, Hanley-Cook GT, Vandevijvere S, et al. Association between adherence to the EAT-Lancet sustainable reference diet and cardiovascular health among European adolescents: the HELENA study. Eur J Clin Nutr. 2024;78:202–208. doi: 10.1038/s41430-023-01379-4 [DOI] [PubMed] [Google Scholar]
- 18.Wang YX, Yang Y, Wang Y, et al. A latent profile analysis of medication adherence to Chinese herbal decoctions among elderly patients with chronic diseases. Evid Based Nurs. 2025;11(6):1096–1103. [Google Scholar]
- 19.Simpson RM, Knowles E, O’Cathain A. Health literacy levels of British adults: a cross-sectional survey using two domains of the Health Literacy Questionnaire (HLQ). BMC Public Health. 2020;20:1819. doi: 10.1186/s12889-020-09727-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Michaud T, Puga T, Archer R, et al. Enhancing text message support with media literacy and financial incentives for vaping cessation in young adults: protocol for a pilot randomized controlled trial. JMIR Res Protoc. 2025;14:e60527. doi: 10.2196/60527 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Braver WJ, Jackson L, Klabbers RE, et al. Digitally enabled cardiac rehabilitation for coronary heart disease: a systematic review and meta-analysis. Eur Heart J Digit Health. 2025;6(4):688–700. doi: 10.1093/ehjdh/ztaf050 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Cruz-Cobo C, Bernal-Jiménez MÁ, Calle G, et al. Efficacy of a Mobile Health App (eMOTIVA) regarding compliance with cardiac rehabilitation guidelines after a coronary event: randomized controlled trial. JMIR mHealth uHealth. 2024;12:e55421. doi: 10.2196/55421 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Gallegos-Rejas VM, Ashcroft J, Taylor RS, et al. Key features in telehealth-delivered cardiac rehabilitation that improve secondary prevention outcomes: a systematic review protocol. Eur Heart J Digit Health. 2024;5(3):208–216. doi: 10.1093/ehjdh/ztad084 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Gaalema DE, Khadanga S, Savage PD, et al. Improving cardiac rehabilitation adherence in patients with lower socioeconomic status: a randomized clinical trial. JAMA Intern Med. 2024;184(9):1095–1104. doi: 10.1001/jamainternmed.2024.3338 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Anderson L, Sharp GA, Norton RJ, et al. Home-based versus centre-based cardiac rehabilitation. Cochrane Database Syst Rev. 2023;10:CD007130. doi: 10.1002/14651858.CD007130.pub5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Golbus JR, Tieu L, Li S, et al. A randomized trial of a mobile health intervention to augment physical activity in cardiac rehabilitation. NPJ Digit Med. 2023;6. doi: 10.1038/s41746-023-00921-9 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Arias López MP, Ong BA, Borrat Frigola X, et al. Digital literacy as a new determinant of health: a scoping review. PLOS Digit Health. 2023;2:e0000279. doi: 10.1371/journal.pdig.0000279 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Mathews L, Mok Y, Ding N, et al. Social support, social isolation, and outpatient cardiac rehabilitation participation among older adults in the community: the ARIC Study. J Am Heart Assoc. 2024:13e033171. doi: 10.1161/JAHA.123.033171 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Zhang Z, Liu Y, Gao Z, et al. Factors influencing cardiovascular patients to complete phase II cardiac rehabilitation – a qualitative study. BMC Cardiovasc Disord. 2025;25:688. doi: 10.1186/s12872-025-05174-w [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.American Heart Association. Home-based cardiac rehabilitation: a scientific statement From the American Association of Cardiovascular and Pulmonary Rehabilitation, the American Heart Association, and the American College of Cardiology. Circulation. 2019;140:e69–e89. doi: 10.1161/CIR.0000000000000663 [DOI] [PubMed] [Google Scholar]
- 31.Li R, Wang M, Chen S, et al. Comparative efficacy and adherence of telehealth cardiac rehabilitation interventions for patients with cardiovascular disease: a systematic review and network meta-analysis. Int J Nurs Stud. 2024;158:104845. doi: 10.1016/j.ijnurstu.2024.104845 [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
The data and materials that support the findings of this study are available from Miss Xingrong He upon reasonable request.