Using the convenience sampling method, we selected inpatients diagnosed with ACS and undergoing PCI in the heart center from June 2022 to March 2023. Inclusion criteria included: (1) age $\ge$18 years old; (2) a diagnosis of ACS undergoing PCI; (3) heart function class I–III, left ventricular ejection fraction (LVEF) $\ge$35%; (4) the patients, their accompanying person(s), will use a smartphone; (5) informed consent was obtained for participation in this study. Exclusion criteria included (1) patients with severe physical diseases; (2) patients who were automatically discharged, transferred, or died in the hospital; (3) patients who were participating in other studies; (4) withdrawal from the study during the follow-up period; (5) death or loss of contact during follow-up. The Corporate Ethics Committee approved this study. All the participants voluntarily agreed to participate in the study and signed the informed consent form.

A general information questionnaire that includes age, gender, marital status, past medical history, medication history, surgery, laboratory data, and other related information.

Data collection is divided into two parts. In the first part, general data, including SEAMS, APGAR, GAD-2, and PHQ-2 scores of ACS patients, were collected on the day after PCI and before discharge. In the second part, we collected the MMAS-8 scores at 1 month (T1), 3 months (T2), 6 months (T3), 9 months (T4), and 12 months (T5) after PCI for each patient.

The R language was used for data analysis. For descriptive statistics, quantitative data that conform to normal distribution were described as the mean, standard deviation and those that do not conform are described as median and quartile, t-test and Mann-Whitney U test were used. Counting data were expressed as frequency (n) and percentage (%), chi-square test, continuous corrected chi-square tests, and fisher’s exact test were used. In modeling the heterogeneous development track, the hlme function in the lcmm package was used to fit the GMM. The number of categories gradually increased with the single-category model until the optimal model (with the smallest Bayesian information criterion (BIC) value) was found. After the track categories were determined, repeated measures of the analysis of variance (ANOVA) were used to compare drug compliance in different categories at different time points. Univariate and multivariate logistic models were used to identify the predictive factors. A value of p 0.05 was defined as statistically significant.

A total of 150 patients undergoing PCI after ACS were included in this study. During the five follow-up visits, a total of 15 cases were lost to follow-up: males (n = 13), aged $\ge$60 years old (n = 5), high school or above educational background (n = 4), smoking history (n = 9), drinking history (n = 4), diabetes (n = 4), and hypertension (n = 6). No statistical difference was observed in the general information of the included cases. The final sample size was 135 cases (Table 1). There were 108 males (80.0%) and 27 females (20.0%); 90 cases (66.7%) were aged $\ge$60 years; the rate of medical insurance coverage was 95.6%; 39 cases (28.9%) had an educational status of high school or above. The percentages of patients with a smoking history, drinking history, diabetes, hypertension, and stroke were 45.9%, 30.4%, 24.4%, 54.1%, and 5.9%, respectively. The BMI was 24.52 $\pm$ 3.45, and the APGAR index was 10.50 $\pm$ 2.94.

GMM was used in this study to analyze the changing trend in DAPT compliance of ACS patients after PCI, and to identify the heterogeneous development track of ACS patients after PCI. Five DAPT compliance indexes of ACS patients after PCI were fitted, and one to four categories of GMM models were gradually established. The results showed that the model was classified into two groups, with the largest loglik value and the smallest BIC value, so two potential categories were selected for this study (Table 2).

We sequentially established GMM models with 1 to 4 classes by fitting dual antiplatelet adherence indices measured at five postoperative time points. The results demonstrated that the two-class model exhibited a higher log-likelihood (LL) value and lower BIC value than the one-class model. In the two-class solution, 7.41% of the population was classified into subgroup 1, while 92.59% comprised subgroup 2. When testing the three-class models, the LL value plateaued with increased BIC, and the third subgroup showed a class probability of 0%. Similarly, the four-class model demonstrated further BIC inflation with two additional zero-probability subgroups. Therefore, both BIC minimization criteria and substantive interpretability supported the optimality of the two-class solution.

According to the two GMM categories, the heterogeneous development track of ACS patients after PCI was plotted (Fig. 1). There were two completely different developmental features of DAPT compliance in patients with ACS after PCI. A continuous downward trend in MMAS-8 was observed in the first category (low compliance–decreased group). A continuously stable trend in MMAS-8 was noted in the second category (persistently high compliance group).

Fig. 1.

Development track of the mixed model of latent variable growth in DAPT compliance in patients with ACS after PCI. TI, T2, T3, T4, and T5 were recorded 1, 3, 6, 9, and 12 months after discharge of the ACS patient after PCI, respectively. MMAS-8, Morisky Medication Adherence Scales-8.

Repeated measures ANOVA showed significant differences in the DAPT compliance of ACS patients after PCI in different groups and at various time points. The F statistic of the interaction terms of time and group was 83.86, and the corresponding p-value was less than 0.001, showing statistical significance (Table 3).

Univariate analysis showed that age, occupational status and APGAR index values were statistically significant between the two groups (p 0.05). No significant differences were observed between the two groups in gender, health insurance status, marital status, whether living alone or not, educational level, daily exercise time, smoking history, diabetes history, diabetes, hypertension, stroke, history of aspirin, history of ticagrelor, history of clopidogrel, history of statins, stent number, LVEF, BMI, SEAMS, GAD-2, and PHQ-2, (p $>$ 0.05) (Table 4).

The development track category of DAPT compliance of ACS patients after PCI was used as the dependent variable (low compliance–decline group = 1; persistently high compliance group = 2). According to the clinical experience, the BMI and the categories displaying statistical significance in the univariate analysis were analyzed by multivariate logistic regression. Independent variable assignment: age: 60 years old = 1, $\ge$60 years old = 2; occupational status: incumbent = 1, retired = 2; the original value was used for the APGAR index and BMI being administered. Multivariate logistic regression analysis showed that age, APGAR index, and BMI affected the potential subgroup of DAPT compliance development track of ACS patients after PCI (Table 5).

This study used GMM to identify two potential DAPT compliance development track categories in ACS patients after PCI: the low compliance–decreased group (7.41%) and the persistently high compliance group (92.59%). Repeated measures ANOVA was used to clarify further the significant differences in DAPT compliance of ACS patients after PCI from different groups at different time points. The statistical analysis showed significant population heterogeneity in the development track of DAPT compliance of ACS patients within 12 months after PCI.

The 7.41% incidence rate of patients calculated in the low compliance–decreased group in this study indicates that some patients may have good adherence over time in the early phase after an event, which reduces over time. A previous study [15] has shown that 4.6% of patients have poor compliance within 12 months, 4% rapidly decline, and 13.9% gradually decline during the post-discharge period following a non-fatal acute coronary syndrome or post-stroke, drug compliance; these data align with those in this study. These findings may be related to the distrust and perception of these medications among patients who experience a critical illness. The failure of patients to understand the benefits of drug treatment in time is a factor in poor compliance with cardiovascular drugs [16]. Therefore, the medical staff must be aware of these changes and initiate intensive compliance interventions for these patients before discharge and during early follow-up meetings. These include changing the perceptions of patients as to which medications are effective and preventing these patients from stopping or reducing the use of cardiovascular drugs.

The 92.59% share of patients in the persistently high compliance group in this study indicates that most patients maintain good compliance throughout the process. This may be related to the severe consequences for patients who have experienced non-fatal ACS and fear of recurrent events. These major medical events have made patients pay more attention to managing their bodies, especially managing drugs related to treating diseases.

The results of this study showed that people aged $\ge$60 years were more likely to enter the persistently high compliance group. Cho et al. [17] found that participants aged $\ge$560 years were generally more represented in clusters with high adherence than younger participants, which was similar to the result of this study. However, Franchi et al. [18] reported a low adherence of older patients with chronic diseases to medication, which might be related to the fact that older patients in the community often suffer from multiple chronic diseases, with numerous concomitant medications, as well as multiple drug categories, dosage forms, and medication methods. Cao et al. [19] showed that younger patients had better medication compliance than older patients, and the memory and self-care ability of patients worsened with age. Rural areas and cities may also have certain differences in ages and the type of diseases, which are very important. These may be different from previous studies. Indeed, older patients are more willing to follow the advice of doctors owing to the serious consequences of the disease, and their fear of recurrent events. The medical staff needs to be aware of these changes. It is very important to identify patients with low drug compliance early and monitor drug use strategies. Moreover, individualized drug education models should be adopted for patients of differing ages and those in different disease stages. In addition, easy-to-understand vocabulary should be used. Internet plus service should be used to raise the awareness of patients regarding the importance of regular medication and the dangers of unauthorized withdrawal, and incorrect or missed drug dosages.

Populations with higher BMIs were likelier to enter the persistently high compliance group. Liu et al. [20] found that a higher BMI was related to good drug compliance, which was consistent with the results of this study. However, another study has also shown that obesity does not affect the drug compliance of locally advanced rectal cancer [21]. The majority of patients with high BMI values are relatively obese, which is potentially because most obese patients have chronic diseases and need to take drugs for a long time. This makes these patients more aware of their drug needs, thus improving compliance. The medical staff needs to pay more attention to these patients with less compliance in administering their medications to better institute health education on the importance of drugs, identify patients with poor compliance with medicines at an early stage, and provide incentive education plans to improve compliance.

This study showed that the population receiving more family care had increased access to the low compliance–decreased group. This contradicts the results of Wu et al. [22], which state that patients with good family functions have better medication compliance. Since most patients are highly valued by their families during hospitalization, the family care after discharge is not as good as during hospitalization, which affects their medication compliance. However, previous research has also revealed that reduced family care has a positive correlation, leading to high compliance of patients with drugs [23]. The high family care may make patients excessively rely on family members and lead to a decline in the ability of the patients to self-care. When a certain amount of assistance is missing, the self-management ability of patients will be improved to a certain extent. The medical staff should guide the family care of patients after discharge, pay attention to the compliance of patients with DAPT for home-based rehabilitation after discharge, and strengthen healthcare education for families to guide patients and drug management outside of hospitalization. Disease perception and emotional expression of familial caregivers are also very important for drug compliance [23]. The medical staff should also provide psychological education for patient family caregivers.