1. Background

Cardiovascular diseases (CVDs) include coronary artery disease (CAD) and cerebrovascular disease [1]. The incidence of CVD is rapidly increasing worldwide, with approximately 17.9 million people having died from CVD in 2019, accounting for 32% of the global death toll [1]. The Global Burden of Diseases, Injuries, and Risk Factors Study 2019 announced that the prevalence of CVD from 1990 to 2019 had approximately doubled from 271 million to 523 million, and mortalities due to CVD had also increased steadily from 12.1 million to 18.6 million [2]. In Korea, the mortality rate from CVD reached 61.5 per 100,000 people in 2021 [3]. The increase in the mortality rate from CVD among the population aged 25~64 years in the United States [4] has necessitated the implementation of health management programs to reduce the mortality rate.

CAD, which is the most rapidly increasing CVD, is caused by an inability to provide oxygen to the heart muscles due to the development of cortical sclerosis in the coronary arteries and is a typical condition of angina and myocardial infarction [5]. Percutaneous coronary intervention (PCI) is a typical treatment method used to prevent the progression of ischemia and infarction in CAD [6]. The number of PCIs conducted in Korean medical institutions in 2019 has steadily increased with the prevalence of CAD [7]. Patients who underwent PCI had a 31% reduction in CVD mortality and fewer cardiovascular events than those who underwent drug treatment [6]. Despite successful PCI intervention, restenosis and heart disease occur in 15%~50% of patients [8]; therefore, active and consistent healthy behavior involving lifestyle correction is important, even after patients with CAD receive PCI [9].

Patients with CAD may be uncertain about symptoms or post-treatment outcomes for the diagnosis of myocardial infarction or angina and the implementation of PCI [10]. Mishel [11] discussed the ongoing uncertainty among those with chronic diseases that are likely to recur and require continuous care. Mishel’s uncertainty in illness theory (UIT) [12] suggests that uncertainty is a meaningful factor in life. Mishel [11] argued that uncertainty management is important for disease adaptation, and that the primary purpose is to manage uncertainty in chronic diseases through providing a new perspective on life in which uncertainty is normalized.

For patients with CAD who face threats to their lives, uncertainty concerning their survival or prognosis acts as a serious psychosocial stressor. Over 30% of CAD patients experience anxiety and depression [13], which further exacerbates uncertainty [14]. Therefore, it is crucial to address the significant psychological issues caused by uncertainty. The latest tests or treatments for events occurring in the course of transdermal coronary interventions and medications can all be perceived as threats [10]. Moreover, uncertainly negatively affect the performance of health-related activities, such as physical activity and treatment instruction implementation [15]. Therefore, an intervention program should be developed to effectively maintain health through identifying the level of uncertainty in patients with CAD, thereby providing a theoretical basis for reducing or resolving uncertainty and promoting healthy behaviors.

UIT has been applied to patients with chronic diseases such as CAD [10], systemic lupus erythematosus [16], and cancer [17], as well as those undergoing hemodialysis [18]. Most studies have shown a relationship between uncertainty and several factors, such as social support and quality of life. Studies concerning the application of UIT to patients with CAD include a comparative study of psychological factors such as uncertainty levels, depression, and anxiety in patients undergoing PCI and coronary artery bypass graft surgery (CABG) [19] and a correlation study of uncertainty and coping [20].

A recent study investigated the predictive variables related to uncertainty in patients with CABG [10]; another employed a structural equation model to examine the mediating role of uncertainty in illness and depressive symptoms between symptom distress and health-related quality of life in patients with heart failure [14]. However, few studies have been conducted in relation to factors required to develop an intervention strategy to manage the uncertainty experienced by patients undergoing PCI following a diagnosis of CAD. Factors involved in well-performing health behaviors to reduce mortality and complications, and ultimately to improve quality of life in the presence of uncertainty, particularly when evaluating uncertainty as an opportunity or danger, need to be examined. Specifically, the health behavior of patients with CAD is directly related to their prognosis, and these health behaviors significantly impact the quality of life of patients [9, 15]. Therefore, it is important to investigate predictive variables that influence the health behaviors of patients experiencing uncertainty and to identify a simple yet optimal path through which uncertainty indirectly affects health behavior via path analysis. This will provide evidence for enhancing health behaviors among CAD patients experiencing uncertainty.

This study aimed to examine the optimal paths from uncertainty to health behavior based on reconceptualized Mishel’s UIT [11] through verifying the effects of perceived benefit and intention, perceived risk, social support, and self-efficacy on the path among patients with CAD who underwent PCI.

2. Theoretical foundation and hypothetical model

This study’s theoretical foundation and hypothetical model were derived from Mishel’s [11] reconceptualized UIT (Figure 1). uncertainty arises from the inability to accurately define disease-related events and situations and the lack of sufficient information to predict treatment outcomes [12]. The UIT explains how patients perceive uncertainty and structure meaning about disease-related events [21]. The original model of UIT focuses on the uncertainty experienced by acutely ill patients, while the reconceptualized UIT addresses the processes and outcomes experienced by patients living with uncertainty in chronic diseases or with the potential for recurrence. Therefore, this study emphasized the process and outcomes following uncertainty in patients with CAD based on the reconceptualized UIT.

Figure 1
Hypothesized model.
No. = Number; PCI = Percutaneous coronary intervention.

• Download Figure

The appraisal of uncertainty involves evaluating uncertainty as an opportunity or a threat [12]. After evaluating uncertainty, the next step is coping strategies, which were categorized into affect-control and buffering strategies [22]. While this study did not explicitly differentiate affect-control and buffering strategies, affect-control involves recognizing one’s emotions and selecting and regulating thoughts or actions that emphasize positive emotions or alleviate negative emotions, so variables such as self-efficacy [15, 23], perceived risk [24], perceived intention, and benefit [10] were set. Buffering strategy involves utilizing external support systems and social support systems, hence social support [10] was set as a variable. These coping strategy-related variables are crucial for coping with uncertainty and positively impacting health behaviors [10, 15, 23, 24].

Patients living with chronic uncertainty due to CAD require careful self-management, and health behavior is the core of self-management [9]. Therefore, in this study, the adaptation state of CAD patients was set as performing health behaviors. Even if uncertainty is initially evaluated as a threat rather than an opportunity in the reconceptualized model, over time, the evaluation of uncertainty may shift from risk to opportunity, leading to a reconstruction of perspectives on life and adaptation [11]. Therefore, it is deemed essential to identify the optimal path through which uncertainty affects adaptation (health behavior) through coping strategies, regardless of whether uncertainty is appraised as an opportunity or a threat [11]. Based on prior literature [25], this study attempted to explore the optimal path by adding a new path as uncertainty would be directly related to self-efficacy. This study aims to validate hypotheses related to this premise. Therefore, the hypothetical model of this study is as follows:

1. Study design

In this study, based on Mishel’s reconceptualization of the UIT model [11], we established a hypothetical model that included influencing factors explaining the adaptation process of patients with CAD. Cross-sectional data were collected from patients with CAD. This study presents a path model construction to verify the suitability and hypotheses of the path model.

2. Setting and sample

This study comprised patients with CAD, such as angina pectoris or myocardial infarction, who had been diagnosed by an internal medicine specialist, who then received follow-up care from the outpatient department of Cardiology Medicine at Pusan National University Hospital in Busan Metropolitan City. The inclusion criteria were as follows: (1) patients aged ≥ 18 years who had been diagnosed with CAD at a medical institution and (2) those who had undergone PCI within the past five years. The exclusion criteria were as follows: (1) patients with dementia or a history of cognitive impairment or mental illness; (2) patients with serious diseases (e.g., end-stage renal disease, cancer, stroke) that would affect their quality of life within one year; and (3) patients who had difficulty understanding and answering the questionnaire owing to cognitive issues. All participants provided informed consent prior to participation.

This study required a sample size of five times the free parameters, including all error variables and path coefficients [26]. We used maximum likelihood estimation, which is the most common estimation method in structural equation model analysis, and the number of samples was calculated based on a sample size of 100~150 [27]. A total of 44 free parameters were included in the hypothetical model. A questionnaire was administered to 242 individuals accordingly, an increase of 10% from the minimum 220 individuals required. Questionnaires with more than 10% missing values for each measurement tool were excluded from the analysis, and missing values were replaced with the mean. As a result, a total of 228 questionnaires were included in the analysis.

3. Measurements

4. Data collection

The data were collected between September 2020 and June 2021. The project investigator explained the purpose of the study to three cardiologists and obtained their cooperation in data collection. One researcher and one research assistant surveyed the patients, and another researcher collected the patients’ disease-related characteristic data from electronic medical records.

The researcher and research assistant explained the purpose of the study and the questionnaire response method to patients who visited the cardiology outpatient department and then distributed the questionnaire to the participants who provided written consent. The participants undertook the questionnaire in the outpatient clinic of the Department of Cardiology, which took approximately 25~30 minutes to complete.

5. Data analysis

Data were analyzed using SPSS win 25.0 and AMOS 25.0 software (IBM Co.). Descriptive statistics were used for participants’ characteristics and major variables, a t-test and ANOVA were used for differences in health behavior according to participants’ characteristics, and the Scheffé test was used for post-hoc testing. Covariate structural analysis was performed to test the hypothetical model. The normality of the measured variables and multicollinearity between variables were tested through calculating the correlation coefficients, skewness and kurtosis, variance expansion factors and tolerance limits, and state indices. The Durbin–Watson index was used for the autocorrelation of the dependent variables [37].

The reliability of each measurement tool was calculated using Cronbach’s α, and CFA was performed to verify its validity. To verify the model’s fit, χ², normed χ² (χ² / df), GFI, SRMR, CFI, and RMSEA were used, as recommended by Kline [37]. The model was evaluated as acceptable when χ² was small and p > .05, χ² / df was ≤ 3, SRMR and RMSEA were ≤ .08, and GFI and CFI were > .90. The significance of the model path was confirmed using the regression coefficient, standard error, threshold, and a p-value. Statistical significance was set at a critical ratio (CR) of ≥ 1.965 and p < .05. Two-tailed significance was used, and the significance of the direct and total effects was verified at p < .05. For the indirect effect, a 95% confidence interval (CI) was obtained using the bootstrap method, and the number of bootstrap samples was set to 5,000. When there are multiple mediating variables in a multiple-mediator model, phantom variables are used to verify the differences between specific indirect effect paths. In this study, path analysis was conducted step by step for Model 1 and Model 2. Model 1, as suggested in the theory, tested a model consisting of a sequential path from uncertainty, uncertainty evaluation, various coping strategies, and health behavior. Model 2 tested the model that omitted the uncertainty assessment step in Model 1, establishing a direct path from uncertainty to self-efficacy [11, 25]. If control variables are used in structural equation models, the model better reflects reality in as far as possible and can be more readily applied in practice. Moreover, the influence of independent variables on dependent variables in a research model needs to be measured as accurately as possible [38]. In this study, therefore, based on a literature review, age, gender, smoking status, and the number of PCIs [2, 6, 10] were set as control variables that may affect the health behavior of patients with CAD. Gender and PCI frequency were changed to dummy variables and analyzed.

6. Ethical consideration

This study was approved by the Institutional Review Board of Pusan National University Hospital (approval no. H-2007-025-093). All aspects of the study were explained to potential participants. Those who agreed to participate signed a written consent form prior to data collection.

1. Study participant characteristics

Among the participants, 171 (75.0%) were men, the average age was 64.6 years (range, 36.0~88.0), and the average BMI was 25.34 kg/m2 (range, 16.0~49.6). Regarding medical history, 137 (60.1%) patients had hypertension, and 145 (63.6%) were diagnosed with myocardial infarction. The mean LVEF was 53.7% (range, 28.0~72.0), and 101 patients (44.3%) had one diseased vessel. Concerning disease duration, 124 (54.4%) patients had a disease duration > 1 year, and 189 (82.9%) participants had undergone PCI for the first time (Table 1).

Table 1
Differences in Health Behavior in Relation to Patient Characteristics (N = 228)

• Download Table

As a result of examining the differences in participants’ characteristics according to health behavior, the only statistically significant difference in health behavior was current smoking status. The post-test results showed that the health behavior scores of former smokers and non-smokers were higher than those of current smokers.

2. Descriptive statistics, and correlation among the main variables

Table 2 presents the descriptive statistics and correlations among the variables. In this study, the multivariate kurtosis index was 1.23, which was less than the threshold of 23.69 at df = 14, and α = .05, and multivariate normality was satisfied. Furthermore, there was no issue with multicollinearity because all values of tolerance limits among the measurement variables were .47~.86 (> .10), and all variance inflation factors were 1.16~2.13 (< 10).

Table 2
Descriptive Statistics and Correlation among the Measurement Variables (N = 228)

• Download Table

The correlation coefficient between all variables was statistically significant, except for that between uncertainty opportunity appraisal and social support. The variables highly correlated with health behavior were uncertainty (r = -.42, p < .001), self-efficacy (r = .42, p < .001), and perceived benefit and intention (r = .54, p < .001).

3. Model fit of the hypothetical model

As a result of testing Model 1, all fit indices were found to be unacceptable, so the hypothesized model was modified (χ² = 147.45, χ² / df = 3.88, GFI = .89, SRMR = .09, CFI = .79, RMSEA = .11). The decision criteria for modifying the hypothetical model were based on the significance of fit indices, path coefficients, and modification indices. Sequentially deleting paths with low coefficients resulted in removing five paths: from danger to social support and self-efficacy, from perceived risk to health behavior, and from gender and number of PCIs to health behavior. Consequently, the fit indices indicated with some fit indices were not acceptable (χ² = 162.55, χ² / df = 3.78, GFI = .90, SRMR = .09, CFI = .77, RMSEA = .11).

We tested Model 2 (Figure 2) to identify the optimal path and found that all goodness-of-fit indices were acceptable (χ² = 79.19, χ² / df = 2.83, SRMR = .07, GFI = .94, CFI = .90, RMSEA = .09). The χ² difference test was performed to evaluate whether there was an improvement in the model fit [37], χ² statistic decreased from 147.45 to 79.19, and df decreased by 10 (38~28), which was significant at α = .05 level. Additionally, the remaining fit indices generally satisfied the hypothesis model; therefore, Model 2 was selected as the optimal model.

Figure 2
Standardized estimates of the final model.
e = Error term of each indicator; PCI = Percutaneous coronary intervention.

^*p < .01, ^**p < .001.

• Download Figure

4. Path analysis of the model

In Model 2, the analysis of individual paths between variables, excluding covariance, showed that 11 of the 12 paths were significant (Figure 2, Table 3). Uncertainty had a negative direct effect on opportunity (β = -.33, p < .001) and a positive direct effect on danger (β = .55, p < .001). Unlike the hypothetical path, uncertainty had a negative direct effect on self-efficacy (β = -.47, p < .001). Opportunity appraisal had a direct effect on social support (β = .32, p = .001), perceived risk (β = -.18, p = .002), self-efficacy (β = .14, p = .001), and perceived benefit and intention (β = .29, p < .001). Danger appraisal had a direct effect on perceived risk (β = .35, p < .001). Social support (β = .15, p = .005), self-efficacy (β = .30, p < .001), and perceived benefit and intention (β = .43, p < .001) had direct effects on health behavior.

Table 3
Estimates, Direct, Indirect, and Total Effects of the Modified Model (N = 228)

• Download Table

As a result of analyzing the indirect effects, the largest indirect effects on health behavior were in relation to perceived benefit and intention (β = .43, p < .001), self-efficacy (β = .30, p < .001), and uncertainty (β = -.25, p < .001).

In the path model of this study, the relationship between uncertainty and health behavior can be explained by several indirect paths. The total indirect effect of uncertainty on health behavior was significant (β = -.25, 95% CI = -.33~-.17). Statistical differences in the four specific indirect effect paths were significant (Table 3). According to the results, the simple mediating effect through self-efficacy (β = -.14, 95% CI = .21~-.08) was most effective.

The results of validating the hypotheses of this study are as follows: