Introduction
Chronic heart failure (CHF) presents mainly with dyspnea, edema, and fatigue.1 The condition has grown steadily more common over recent decades: more than 64 million people are now affected worldwide,2 and approximately 12.1 million of them live in China.3
For patients, CHF means repeated exacerbations and prolonged physical suffering. Each readmission adds to caregiver burden and household financial strain, and the accumulated pressure takes a real psychological toll.4 Helplessness is a common response to declining physical function,5 and it tends to open the door to fear, anxiety, and depression.6 Qualitative work suggests that this fear deepens as the disease advances.7 Fear of disease progression (FoP) denotes the apprehension a patient feels about what further deterioration or recurrence of the illness might bring.8 The problem is circular: adverse psychological states erode treatment confidence and worsen symptoms, and worse symptoms in turn feed the psychological burden and negative emotions.9 One factor consistently associated with such emotions is time perspective10—the characteristic way a person frames past, present, and future, which colors cognition, emotion, and behavior.11 The construct comprises five dimensions: past negative (a tendency to dwell on past events in a negative light), past positive (a warm regard for one’s past), present fatalistic (a helpless, resigned stance toward life), present hedonistic (a focus on immediate gratification), and future (planning and striving toward goals). Higher past negative and present fatalistic scores have been linked to a greater likelihood of adverse emotions such as depression.12
FoP overlaps with, but is distinct from, anxiety, depression, fear of recurrence, and illness perception. Anxiety and depression describe generalized affective states, whereas FoP is the apprehension that the illness will worsen;8 fear of recurrence is the corresponding concept in cancer survivors, and illness perception refers to a patient’s beliefs about the causes, consequences, and controllability of the illness, which may shape FoP but are not themselves an emotional response. The construct and the instruments used to measure it were developed mainly in oncology, so most FoP questionnaires and cut-offs, including those used here, originate in cancer populations.8 Evidence on FoP in heart failure is comparatively recent and limited.5
Time perspective is also distinct from these constructs and from other psychological correlates of FoP: coping styles describe how a person responds to a stressor, and social support describes available resources; time perspective, in contrast, describes a habitual cognitive orientation toward the past, present, and future that shapes how experiences are interpreted.11 A patient with a strongly negative orientation toward the past may read a new symptom as a return of earlier suffering, and a patient with a fatalistic orientation toward the present may regard deterioration as unavoidable; either pattern could heighten FoP regardless of anxiety or depression levels. To our knowledge, how individual time perspective dimensions relate to FoP in patients with CHF has barely been examined. We therefore examined all five dimensions as primary exposures and tested whether each was associated with FoP after accounting for demographic and clinical factors.
Because past negative and present fatalistic orientations have been linked with adverse emotions,12 these two were of particular interest, and the five coefficients were treated as a single family for multiplicity adjustment.
Methods
Participants
We conducted a single-center cross-sectional survey in the Department of Cardiovascular Medicine of the Affiliated Hospital of Nantong University, a tertiary hospital in Nantong, between December 2025 and August 2026. Participants were recruited by convenience sampling from patients admitted to the ward during this period.
Eligible patients were aged ≥18 years, had CHF diagnosed according to the Chinese guidelines for the diagnosis and treatment of heart failure 202413 with New York Heart Association (NYHA) functional class II–IV, were clinically stable (stable vital signs and a condition that, in the judgment of the attending physician, permitted completion of the questionnaire), were able to communicate, and took part voluntarily after giving written informed consent. No minimum disease duration was required for entry, so the cohort includes patients close to their initial diagnosis. We excluded patients whose condition was unstable or deteriorating rapidly, those with severe mental disorders or cognitive dysfunction, those with serious comorbidities (such as malignant tumors or severe dysfunction of other vital organs), those enrolled in other interventional studies, and those facing language or cultural barriers.
By the time the first 205 questionnaires had been returned, 211 eligible patients had been approached and 6 had declined; their reasons for declining were not recorded. For the remainder of the recruitment period, the numbers of patients approached and declining were not recorded, so from that point the recruitment flow can be reported only from the returned questionnaires onward (see Results and Limitations).
Sample Size
The sample size was planned with the rule of thumb of 5–10 participants per study variable that is commonly applied in questionnaire-based nursing research.14 With 25 variables considered, a minimum of 125 participants was required; allowing for a 20% rate of invalid responses, the minimum target was 157. This rule is a heuristic and not a formal a priori power calculation. Recruitment continued until the end of the planned data-collection window on 17 August 2026.
In total, 266 questionnaires were returned and 254 valid questionnaires were analyzed (see Results); the regression was fitted to the 237 participants with complete data on every model term, so the minimum target was exceeded. A preliminary analysis had been performed on the first 205 questionnaires while recruitment was still under way. It was not presented at any meeting, posted as a preprint, registered, or published, and it did not influence the hypotheses, the variables collected or analyzed, the planned sample size, or the decision to stop recruitment.
The present report is based on every valid questionnaire collected over the study period and supersedes that preliminary analysis. As a sensitivity check and not as a substitute for an a priori justification, with 237 complete cases and 19 predictor terms the study had 85% power at α = 0.05 to detect an overall effect of f2 = 0.10 and 80% power to detect an individual predictor contributing f2 = 0.036.
Data Collection
Before any data were collected, we explained the study aims to each participant, made clear that taking part was voluntary, and obtained documented consent. Trained researchers then administered the questionnaires face to face under uniform instructions during the hospital stay, once the patient’s condition permitted, that is, when vital signs were stable and the patient was able to communicate, as judged by the attending physician; the day after admission on which the questionnaire was completed was not fixed and was not recorded. For patients with limited literacy, visual impairment, or physical constraints, the researchers read each item aloud and explained it where needed.
Completed questionnaires were checked on the spot for omissions and logically inconsistent answers; nevertheless, 11 questionnaires were returned with both scales left blank and were excluded (see Results), and a small number of clinical and demographic items remained blank and were treated as missing (see Statistical Analysis).
Measurements
General Information Questionnaire
The demographic characteristics included sex, age, marital status, educational level, place of residence, living status, main caregivers, occupational status, monthly household income, main source of income, medical payment methods, and smoking and drinking status; Disease-related information included number of complications, course of CHF, number of hospitalizations within the last year, NYHA class, left ventricular ejection fraction (LVEF), number of medications taken, and family history.
The Chinese Version of the Fear of Progression Questionnaire-Short Form (FoP-Q-SF)
We used the Chinese version of the FoP-Q-SF,15 which Wu et al translated from Mehnert’s original instrument.16 The scale contains two domains—physical health and social-family—with 6 items each. Items are rated on a 5-point Likert scale, giving a total score between 12 and 60; a score of 34 or higher is taken to indicate FoP beyond the normal range, and higher scores denote greater fear. The instrument and this cut-off were developed and validated in cancer patients, and the cut-off was derived by a median-split procedure; it is a pragmatic rather than a diagnostic threshold and has not been calibrated in heart failure.
Except for the 11-item sensitivity analysis described below, every analysis used the 12-item total score; neither subscale score entered any analysis. In the present sample, internal consistency of the total scale was Cronbach’s α = 0.675 and McDonald’s ω = 0.706, against α = 0.883 in the original Chinese validation study.15 The two-domain structure was not reproduced here. Sampling adequacy was modest (Kaiser–Meyer–Olkin = 0.683; Bartlett’s test of sphericity χ2(66) = 737.28, P < 0.001).
Exploratory factor analysis (principal component extraction on the 12-item correlation matrix) returned five eigenvalues greater than 1, the first two accounting for 40.5% of the variance, and in a forced two-factor solution (varimax rotation) the items did not separate into the original physical health and social-family domains (Supplementary Table S1). Internal consistency was 0.640 for the physical health items and 0.234 for the social-family items. Item-level examination traced most of this shortfall to a single item (item 7), whose corrected item-total correlation was −0.053 and whose correlations with the other items were weak and mixed in sign (−0.27 to +0.14); removing it raised the total-scale α to 0.722.
To rule out a scoring error, we examined item 7 directly: reversing its scoring raised the total-scale α only from 0.675 to 0.693, so the problem is not a simple reverse-coding error. A sensitivity analysis based on the remaining 11 items is reported below.
The Chinese Version of the Zimbardo Time Perspective Inventory (ZTPI-C)
The ZTPI dates from 1999; we adopted the Chinese revision by Li et al,17 which retains 25 items across five domains: past negative, past positive, present fatalistic, present hedonistic (labeled present impulsive in the Chinese revision), and future. The five dimensions comprise 7, 6, 3, 4, and 5 items respectively, giving theoretical ranges of 7–35, 6–30, 3–15, 4–20, and 5–25. Items are answered on a 5-point Likert scale, and each dimension is scored as the sum of its items, so a higher score reflects a stronger orientation toward that dimension; because the dimensions differ in length, sum scores are not directly comparable across dimensions.
Cronbach’s α coefficients in our study were 0.849, 0.753, 0.704, 0.804, and 0.786, respectively.
Statistical Analysis
Analyses were run in SPSS 27.0. Continuous variables are presented as means with standard deviations (SD). Distributional form was examined with the Shapiro–Wilk test together with skewness and kurtosis; FoP scores were left-skewed (W = 0.930, P < 0.001), and parametric tests were retained because t and F-tests are robust to moderate departures from normality at this sample size.
Group differences in FoP across demographic and clinical categories were tested with independent samples t-tests or one-way analysis of variance (ANOVA); homogeneity of variance was checked with Levene’s test, and Welch’s t-test or Welch’s ANOVA was substituted wherever it was violated. Because the medical payment methods other than urban worker and rural cooperative insurance had very few participants (n = 4, 3, and 2), they were combined into one category before testing. Correlations among the study variables were examined with Pearson correlation coefficients, and Spearman rank correlations were computed as a sensitivity analysis because FoP scores were not normally distributed.
All variables listed in Table 1, together with the five time perspective dimensions, served as candidates; those reaching P < 0.05 in the bivariate analyses entered a multiple linear regression with FoP as the dependent variable. Ordinal variables (age group, NYHA class, and LVEF category) entered the model as continuous scores to test for linear trends; two-level variables (smoking, drinking, and course of CHF) entered as indicators, and the remaining multi-category variables were dummy-coded against a reference category (the coding scheme is given with the regression results). Monthly household income (P = 0.118), the number of hospitalizations in the past year (P = 0.068) and the number of medications taken (P = 0.093) did not meet the screening criterion and were not entered.
To quantify the contribution of time perspective, a hierarchical model was also fitted, entering the demographic and clinical predictors in block 1 and the five time perspective dimensions in block 2. Because a large number of significance tests was performed, multiplicity was addressed explicitly. Two test families were specified: the 25 bivariate screening tests, and the five time perspective coefficients in the final model, which carry the primary hypothesis of the study.
Within each family, P values were adjusted with the Holm step-down procedure, which controls the family-wise error rate, and with the Benjamini–Hochberg procedure, which controls the false discovery rate. Unadjusted P values are reported throughout, with the adjusted values given alongside them where they bear on interpretation. Because selecting covariates by bivariate P value can omit clinically relevant confounders and yield unstable estimates, and because ordinal scores assume equal spacing between categories, three further specifications were fitted post hoc, after the primary model had been estimated.
First, in a clinically specified model, covariates were chosen on clinical grounds irrespective of their bivariate P values (age group, sex, educational level, living status, NYHA class, LVEF category, course of CHF, number of complications, number of hospitalizations in the past year, and number of medications taken; age group, educational level, living status, NYHA class and LVEF category were dummy-coded), together with the five time perspective dimensions. The covariate set was fixed before this model was fitted and was not modified afterward. Second, the primary model was refitted with age group, NYHA class and LVEF entered as categorical variables instead of ordinal scores, and the linear-trend assumption was tested with nested F-tests.
Third, because the residuals were not normally distributed, the primary model was re-estimated with heteroscedasticity-consistent (HC3) standard errors and with 5000 bootstrap resamples (percentile 95% confidence intervals). Holm and Benjamini–Hochberg adjustments were applied to the five time perspective coefficients within each specification.Table 1Demographic and Clinical Characteristics of the Participants and Their Association with FoP (N = 254)CharacteristicCategoryn (%)FoP Score, Mean (SD)t/FPEffect SizeAge, yearsa18–5942 (16.8)36.29 (3.20)26.112< 0.001η2 = 0.17560–7499 (39.6)32.99 (4.44)≥75109 (43.6)30.65 (4.68)SexMale153 (60.2)32.67 (5.11)0.5070.613d = 0.07 (−0.19 to 0.32)Female101 (39.8)32.36 (4.21)EducationPrimary school or below153 (60.5)32.16 (4.92)1.8790.134η2 = 0.022Junior middle school58 (22.9)33.31 (4.55)High school/technical secondary school26 (10.3)32.15 (4.64)Junior college or above16 (6.3)34.56 (3.29)Marital statusHave a spouse218 (86.2)32.37 (4.77)−1.2920.198d = 0.24 (−0.12 to 0.59)No spouse35 (13.8)33.49 (4.64)Place of residenceCity97 (38.2)31.95 (5.06)1.4890.228η2 = 0.012Town22 (8.7)33.59 (4.91)Rural135 (53.1)32.80 (4.50)Living statusaLive with a spouse142 (56.6)32.61 (4.65)6.504< 0.001η2 = 0.073Live with children39 (15.5)30.33 (4.75)Live with spouse and children48 (19.1)32.54 (5.01)Live alone22 (8.8)35.73 (2.68)Main caregiversaSpouse115 (45.3)33.43 (4.45)9.651< 0.001η2 = 0.071Parents/children111 (43.7)31.14 (4.97)Others28 (11.0)34.46 (3.64)Occupational statusbEmployed24 (9.4)36.46 (3.50)10.075< 0.001η2 = 0.108Self-employed/Farmer143 (56.3)32.36 (4.50)Retired82 (32.3)31.39 (4.91)Unemployed5 (2.0)37.80 (2.17)Monthly household income, RMB< 300061 (24.0)33.49 (4.40)2.171c0.118η2 = 0.0203000–5000126 (49.6)32.59 (4.14)> 500067 (26.4)31.60 (5.93)Main source of incomeaLabor income61 (24.0)35.41 (3.07)22.347c< 0.001η2 = 0.162Retirement pension77 (30.3)30.90 (4.71)Financial support from children107 (42.1)31.75 (4.84)Others9 (3.5)36.67 (3.28)Medical payment methodsMedical insurance for urban workers94 (37.0)32.37 (4.83)0.131c0.878η2 = 0.001New rural cooperative medical system151 (59.4)32.68 (4.54)Other (urban residents’ insurance, commercial insurance, or self-pay)9 (3.5)32.11 (7.72)SmokingaNo208 (81.9)32.19 (4.76)−2.5230.012d = 0.41 (0.09 to 0.73)Yes46 (18.1)34.13 (4.49)DrinkingaNo191 (75.2)32.16 (4.81)−2.2360.026d = 0.32 (0.04 to 0.61)Yes63 (24.8)33.70 (4.47)Course of CHF, yearsa< 2134 (53.6)33.76 (4.11)4.418< 0.001d = 0.56 (0.31 to 0.81)≥2116 (46.4)31.17 (5.15)Number of hospitalizations within the last year, times< 3216 (85.0)32.35 (4.89)−1.860c0.068d = 0.28 (−0.07 to 0.62)≥338 (15.0)33.66 (3.83)NYHA classaII35 (13.8)30.91 (5.83)4.3130.014η2 = 0.033III194 (76.4)32.58 (4.46)IV25 (9.8)34.52 (4.87)LVEFa< 40%98 (39.7)34.50 (3.67)18.081c< 0.001η2 = 0.14040%–49%61 (24.7)32.67 (3.74)≥50%88 (35.6)30.47 (5.36)Number of complications< 3106 (41.7)32.89 (4.65)0.9720.332d = 0.12 (−0.13 to 0.37)≥3148 (58.3)32.30 (4.85)Number of medications taken< 557 (22.7)33.53 (3.45)1.6860.093d = 0.25 (−0.04 to 0.55)≥5194 (77.3)32.34 (4.97)Family historyNo181 (71.3)32.46 (4.78)−0.4460.656d = 0.06 (−0.21 to 0.33)Yes73 (28.7)32.75 (4.75)Notes: Numbers do not always sum to 254 because of missing values (age 4, educational level 1, marital status 1, living status 3, course of CHF 4, LVEF 7, and number of medications taken 3). aEntered the primary regression model as a covariate; bmet the screening criterion but was not entered because of collinearity with main source of income (see Methods); all other variables are descriptive only. P values are unadjusted screening tests (t-test or one-way ANOVA); cWelch’s test was used where Levene’s test indicated unequal variances (monthly household income, main source of income, number of hospitalizations in the past year, LVEF, and the combined medical payment categories).
Effect sizes are Cohen’s d (95% CI) for two-group comparisons and η2 for comparisons of three or more groups. Medical payment methods other than urban worker insurance and the rural cooperative system (urban residents’ insurance, n = 4; commercial insurance, n = 3; self-pay, n = 2) were combined because of the small numbers. After Holm adjustment across the 25 screening tests, age, living status, main caregivers, occupational status, main source of income, course of CHF, and LVEF remained significant.Abbreviations: SD, standard deviation; CHF, chronic heart failure; NYHA, New York Heart Association; LVEF, left ventricular ejection fraction.
Multicollinearity was assessed with variance inflation factors. In an initial specification containing both occupational status and main source of income, the “retired” category reached a variance inflation factor of 10.80: 75 of the 82 retired patients (91.5%) named a retirement pension as their main source of income, and 75 of the 77 patients living on a pension (97.4%) were retired, so the two variables carried almost the same information in this sample. Occupational status was therefore dropped and main source of income retained, the latter being the more direct index of economic circumstance and the more strongly associated with FoP in the bivariate analysis (Welch F = 22.347 vs F = 10.075; Table 1).
Every variance inflation factor in the final model was ≤ 2.47. Dropping main source of income instead of occupational status (19 terms, maximum variance inflation factor 4.73, R2 = 0.495) yielded the same six significant predictors, as did dropping both variables (16 terms, maximum variance inflation factor 1.73, R2 = 0.489). For 20 cells in 19 questionnaires (age 4, course of CHF 4, LVEF 7, number of medications taken 3, marital status 1, and educational level 1) the item was left blank on the paper form and a value was subsequently entered in the coded dataset by the investigators.
The source of these values was not documented; the investigators could not confirm that they had been retrieved from the medical records and believe that they were probably inferred, so they were not used in any analysis. In addition, three participants described their living arrangement as “other”, which could not be assigned to any of the four categories, and were treated as missing on that variable. These cells were treated as missing in all analyses: descriptive and bivariate analyses used all available cases for each variable, and the regression was fitted to the 237 participants with complete data on every model term.
A further sensitivity analysis re-estimated the same model with the FoP-Q-SF total recomputed from the 11 items remaining after the poorly performing item was dropped.
Results
Demographic Characteristics and Comparison of the Study Variables
A total of 266 questionnaires were returned over the study period (recruitment details are given in Methods). Eleven were excluded because the FoP-Q-SF and the ZTPI had not been completed (only the general information section had been filled in), and one because most of its clinical items had been left blank, leaving 254 responses for analysis. Of these, 17 lacked data on at least one regression term (age 4, living status 3, course of CHF 4, and LVEF 7; one participant lacked both age and course of CHF), so the regression included 237 participants.
The sample comprised 153 men and 101 women, and most patients were aged ≥60 years (83.2%). Educational attainment was generally low—60.5% had elementary schooling or less—while 86.2% were married and 53.1% lived in rural areas. Clinically, NYHA class III predominated (76.4%), and 46.4% of the patients had been ill for at least two years.
Table 1 summarizes the details.
FoP scores differed significantly across age, living status, main caregivers, occupational status, main source of income, smoking, drinking, course of CHF, NYHA class, and LVEF (all unadjusted P < 0.05). After Holm adjustment across the 25 screening tests (20 demographic and clinical variables and 5 time perspective correlations), age, living status, main caregivers, occupational status, main source of income, course of CHF, and LVEF remained significant, whereas smoking, drinking, and NYHA class did not (Holm-adjusted P = 0.172, 0.315, and 0.187, respectively); all three were nevertheless entered into the model under the screening rule described in Methods.
Correlation Analysis
The mean FoP total score was 32.54 (4.77) and the median 34.0 (interquartile range 30.0–35.8), with observed scores ranging from 13 to 43; 130 patients (51.18%) scored 34 or higher. This threshold was derived in cancer patients and has not been validated in heart failure, so the proportion is a descriptive indicator and not the prevalence of clinically significant FoP; it is also sensitive to the cut-off (33.86% at ≥35). For the ZTPI dimensions, mean (SD) scores were 18.23 (4.03) for the past negative, 17.83 (2.74) for the past positive, 9.38 (1.98) for the present fatalistic, 11.90 (2.64) for the present hedonistic, and 14.74 (2.66) for the future dimension.
All five dimensions were significantly correlated with FoP (unadjusted P < 0.05): past positive and future time perspectives showed inverse correlations, whereas past negative, present fatalistic, and present hedonistic time perspectives correlated positively with it. Spearman coefficients led to the same conclusions (Supplementary Table S2). Table 2 presents the Pearson coefficients.Table 2Pearson Correlations Among FoP and the Five Time Perspective Dimensions (N = 254)VariablesFoPPast NegativePast PositivePresent FatalisticPresent HedonisticFoP1Past negative0.260**1Past positive−0.197**0.0041Present fatalistic0.210**0.468**−0.0611Present hedonistic0.139*0.174**−0.0390.1071Future−0.290**0.0660.252**−0.081−0.158*Notes: *P < 0.05, **P < 0.01 (unadjusted).
Spearman rank correlations, computed as a sensitivity analysis, are given in Supplementary Table S2.Abbreviation: FoP, fear of disease progression.
Multiple Linear Regression Analysis
With FoP as the dependent variable, predictors showing significant bivariate associations entered the multivariable model; the dummy coding is given in Table 3. Age, living status (living alone), course of CHF, and LVEF each retained an independent association with FoP, whereas NYHA class did not once the remaining variables were held constant (P = 0.150). Among the psychological variables, past negative and present fatalistic time perspectives remained independently associated as well (Table 4, which shows all model terms).
After adjustment for multiple comparisons within the five time perspective coefficients, the past negative association remained significant under both procedures (Holm-adjusted P = 0.008; Benjamini–Hochberg q = 0.008), whereas the present fatalistic association did not (Holm-adjusted P = 0.116; Benjamini–Hochberg q = 0.073) and is therefore reported as suggestive rather than confirmed. The coefficients for all 19 model terms and their 95% confidence intervals are displayed in Figure 1; the significance symbols in Figure 1 refer to unadjusted P values.Figure 1Forest plot of the multiple linear regression of fear of disease progression on demographic, clinical and time perspective variables (n = 237). Filled squares denote terms with unadjusted P < 0.05 and open circles terms with unadjusted P ≥ 0.05; the symbols therefore do not reflect adjustment for multiple comparisons.
For the five time perspective coefficients, the Holm-adjusted P value is printed beside each estimate and † marks Holm-adjusted P < 0.05. Horizontal bars are 95% confidence intervals and the dashed line marks the null value of zero. Coefficients are unstandardized, so the units differ across terms: ordinal predictors are expressed per category, indicator variables per contrast with the reference category shown in Table 3, and the five time perspective dimensions per one-point increase in the corresponding dimension score.
Coefficients are those reported in Table 4.A forest plot shows unstandardized regression coefficients with 95 percent confidence intervals for multiple predictors, with a dashed vertical reference line at 0. A legend indicates a filled square for, Unadjusted P less than 0.05 and an open circle for, Unadjusted P greater than or equal to 0.05. The x-axis label is, Unstandardized regression coefficient (95 percent CI).
No unit is shown. The x-axis shows tick labels at minus 2, 0, 2 and 4. The y-axis lists predictors: Age (per category); Living status: with children; Living status: with spouse and children; Living status: living alone; Main caregiver: parents children; Main caregiver: others; Income source: retirement pension; Income source: support from children; Income source: others; Smoking (yes vs no); Drinking (yes vs no); Course of CHF (greater than or equal to 2 vs less than 2 years); NYHA class (per class); LVEF (per category); Past negative time perspective; Past positive time perspective; Present fatalistic time perspective; Present hedonistic time perspective; Future time perspective.
Plotted points and intervals along the x-axis: Age (per category) is a filled square at approximately minus 1.6 with a 95 percent confidence interval from approximately minus 2.4 to approximately minus 0.8. Living status: with children is an open circle at approximately minus 0.2 with an interval from approximately minus 1.8 to approximately 1.3. Living status: with spouse and children is an open circle at approximately minus 0.1 with an interval from approximately minus 1.3 to approximately 1.2.
Living status: living alone is a filled square at approximately 2.9 with an interval from approximately 1.0 to approximately 4.9. Main caregiver: parents children is an open circle at approximately minus 0.6 with an interval from approximately minus 1.9 to approximately 0.6. Main caregiver: others is an open circle at approximately minus 0.5 with an interval from approximately minus 2.3 to approximately 1.4.
Income source: retirement pension is an open circle at approximately minus 1.2 with an interval from approximately minus 2.6 to approximately 0.3. Income source: support from children is an open circle at approximately minus 1.0 with an interval from approximately minus 2.4 to approximately 0.4. Income source: others is an open circle at approximately 0 with an interval from approximately minus 2.8 to approximately 2.7.
Smoking (yes vs no) is an open circle at approximately 0 with an interval from approximately minus 1.5 to approximately 1.5. Drinking (yes vs no) is an open circle at approximately 0.1 with an interval from approximately minus 1.3 to approximately 1.4. Course of CHF (greater than or equal to 2 vs less than 2 years) is a filled square at approximately minus 2.4 with an interval from approximately minus 3.4 to approximately minus 1.5.
NYHA class (per class) is an open circle at approximately 0.7 with an interval from approximately minus 0.3 to approximately 1.7. LVEF (per category) is a filled square at approximately minus 1.2 with an interval from approximately minus 1.8 to approximately minus 0.7. For the five time perspective rows, text at the far right reads: Holm P equals 0.008 followed by a dagger symbol for Past negative time perspective; Holm P equals 0.349 for Past positive time perspective; Holm P equals 0.116 for Present fatalistic time perspective; Holm P equals 0.665 for Present hedonistic time perspective; and Holm P equals 0.311 for Future time perspective.
Past negative time perspective is a filled square at approximately 0.2 with an interval from approximately 0.08 to approximately 0.35. Past positive time perspective is an open circle at approximately minus 0.1 with an interval from approximately minus 0.30 to approximately 0.06. Present fatalistic time perspective is a filled square at approximately 0.3 with an interval from approximately 0.03 to approximately 0.57.
Present hedonistic time perspective is an open circle at approximately 0.04 with an interval from approximately minus 0.15 to approximately 0.23. Future time perspective is an open circle at approximately minus 0.16 with an interval from approximately minus 0.36 to approximately 0.03.A forest plot of regression coefficients for demographic, clinical and time perspective predictors.Table 3Coding of the Variables Entered in the Multivariable Regression ModelVariablesCoding MethodAge, years18–59 = 1, 60–74 = 2, ≥ 75 = 3Living statusLive with a spouse (Z1 = 0, Z2 = 0, Z3 = 0);Live with children (Z1 = 1, Z2 = 0, Z3 = 0);Live with spouse and children (Z1 = 0, Z2 = 1, Z3 = 0);Live alone (Z1 = 0, Z2 = 0, Z3 = 1)Main caregiversSpouse (Z1 = 0, Z2 = 0);Parents/children (Z1 = 1, Z2 = 0);Others (Z1 = 0, Z2 = 1)Main source of incomeLabor income (Z1 = 0, Z2 = 0, Z3 = 0);Retirement pension (Z1 = 1, Z2 = 0, Z3 = 0);Financial support from children (Z1 = 0, Z2 = 1, Z3 = 0);Others (Z1 = 0, Z2 = 0, Z3 = 1)SmokingNo = 1, Yes = 2DrinkingNo = 1, Yes = 2Course of CHF, years< 2 = 1, ≥ 2 = 2NYHA classII = 1, III = 2, IV = 3LVEF< 40% = 1, 40%–49% = 2, ≥ 50% = 3Notes: Z terms are dummy codes; a variable with k levels requires k − 1 dummies. The reference category (all Z = 0) is “live with a spouse” for living status, “spouse” for main caregivers, and “labor income” for main source of income.
Age, smoking, drinking, course of CHF, NYHA class and LVEF were entered with the ordinal codes shown above, and the five time perspective dimensions as continuous scores; together with the 8 dummy terms this gives the 19 predictor terms reported in Table 4.Abbreviations: CHF, chronic heart failure; NYHA, New York Heart Association; LVEF, left ventricular ejection fraction.Table 4Multiple Linear Regression of FoP on Demographic, Clinical and Time Perspective Variables (n = 237)VariablesBSEβtP95% CIHolm PaConstant38.8903.020—12.877< 0.00132.937 to 44.842Age (per category: 18–59/60–74/≥75)−1.5960.423−0.244−3.775< 0.001−2.429 to −0.763Living status: live with children−0.2320.785−0.017−0.2960.768−1.779 to 1.314Living status: live with spouse and children−0.0750.634−0.006−0.1180.906−1.325 to 1.176Living status: live alone2.9190.9800.1802.9780.0030.987 to 4.851Main caregiver: parents/children−0.6480.620−0.068−1.0450.297−1.871 to 0.574Main caregiver: others−0.4510.920−0.030−0.4900.625−2.265 to 1.363Main source of income: retirement pension−1.1800.728−0.116−1.6210.106−2.615 to 0.255Main source of income: support from children−1.0170.725−0.106−1.4030.162−2.446 to 0.411Main source of income: others−0.0561.376−0.002−0.0400.968−2.768 to 2.657Smoking (yes vs no)0.0280.7640.0020.0360.971−1.479 to 1.534Drinking (yes vs no)0.0960.6830.0090.1410.888−1.250 to 1.442Course of CHF (≥2 vs < 2 years)−2.4430.499−0.259−4.893< 0.001−3.427 to −1.459NYHA class (per class)0.7190.4980.0751.4430.150−0.263 to 1.700LVEF (per category)−1.2220.287−0.225−4.262< 0.001−1.787 to −0.657Past negative time perspective0.2170.0680.1843.2130.0020.084 to 0.3500.008Past positive time perspective−0.1240.091−0.070−1.3630.174−0.303 to 0.0550.349Present fatalistic time perspective0.3020.1370.1282.1980.0290.031 to 0.5730.116Present hedonistic time perspective0.0410.0950.0220.4340.665−0.146 to 0.2280.665Future time perspective−0.1640.100−0.093−1.6350.104−0.362 to 0.0340.311Notes: R2 = 0.496, adjusted R2 = 0.452, F (19, 217) = 11.239, P < 0.001. All 19 predictor terms that met the bivariate screening criterion (dummy-coded as in Table 3) are shown; the model was fitted to the 237 participants with complete data on every term. Adding the five time perspective dimensions to the demographic and clinical block increased the explained variance by ΔR2 = 0.081 (ΔF(5, 217) = 6.960, P < 0.001). aHolm adjustment across the five time perspective coefficients (primary hypothesis family); the Benjamini–Hochberg q values are 0.008, 0.218, 0.073, 0.665, and 0.173 for the past negative, past positive, present fatalistic, present hedonistic, and future dimensions, respectively.
P values in the P column are unadjusted.Abbreviations: B, unstandardized regression coefficient; SE, standard error; β, standardized regression coefficient; CI, confidence interval; CHF, chronic heart failure; NYHA, New York Heart Association; LVEF, left ventricular ejection fraction.
Sensitivity analyses are summarized in Supplementary Table S3 (full coefficients of the clinically specified model are in Supplementary Table S4). When covariates were chosen on clinical grounds and not by bivariate screening (n = 233, 22 predictor terms, R2 = 0.541), the past negative association was similar (B = 0.179, 95% CI 0.052 to 0.306, P = 0.006; Holm-adjusted P = 0.024). The present fatalistic association was larger in this model (B = 0.429, 95% CI 0.165 to 0.693, P = 0.002) and remained significant after Holm correction (adjusted P = 0.008), and the future dimension reached nominal significance (B = −0.222, P = 0.021; Holm-adjusted P = 0.062; Benjamini–Hochberg q = 0.035).
When age, NYHA class and LVEF were entered as categories (n = 237, 22 terms, R2 = 0.501), there was no evidence against the linear-trend assumption (nested F-tests: age P = 0.353, NYHA class P = 0.401, LVEF P = 0.518; jointly P = 0.561), and the time perspective coefficients were similar (past negative B = 0.209, P = 0.002, Holm-adjusted P = 0.012; present fatalistic B = 0.299, P = 0.032, Holm-adjusted P = 0.128). With HC3 robust standard errors the two associations were unchanged (past negative P = 0.001; present fatalistic P = 0.026), and the bootstrap 95% confidence intervals excluded zero for both (past negative 0.095 to 0.342; present fatalistic 0.037 to 0.556). The past negative association was therefore significant after multiplicity adjustment in the primary model and in both alternative specifications, whereas the multiplicity-adjusted status of the present fatalistic association depended on the covariate set; we continue to regard the latter as suggestive and not confirmed.
Discussion
In this cross-sectional study we examined how time perspective relates to FoP in patients with CHF. The mean FoP-Q-SF total score was 32.54 (4.77), and 51.18% of the patients scored ≥34, compared with 23.4% reported in older Chinese patients with CHF by Tian et al;18 the samples differ, so the figures are not directly comparable. Because the ≥34 threshold was derived in cancer patients, has not been calibrated in heart failure, and fell exactly at the sample median here, the proportion is sensitive to where the threshold is placed (33.86% at ≥35) and should not be read as the prevalence of clinically significant FoP in CHF.
Given how chronic, progressive, and unpredictable heart failure is, a heightened fear of exacerbation is understandable, and time perspective is one candidate psychological factor that may accompany it.
Four demographic and clinical factors retained an independent association with FoP: age, living status, disease duration, and LVEF. Younger patients scored higher, echoing earlier reports;19 illness at this stage of life disrupts career building and family formation, tasks that weigh most heavily on the young.20 Patients who lived alone also reported more fear than those living with family. Chronically ill people who live alone often lack day-to-day emotional and practical support, and the resulting loneliness feeds negative emotional states.21 Shorter disease duration, somewhat counterintuitively, went with higher FoP—a pattern Zhen et al also observed;22 patients close to diagnosis may still be struggling to accept it, with all the emotional volatility that entails.
Higher NYHA class was associated with greater fear in the bivariate comparison, though it no longer reached significance once the other variables were held constant; NYHA class indexes functional limitation, and curtailed exercise tolerance and independence open a path from bodily decline to psychological distress.23 Lower LVEF, by contrast, retained an independent association: it marks more severe systolic dysfunction and has been linked to depressive symptoms in cardiac patients.24 The number of admissions in the past year did not differ significantly in the bivariate comparison (P = 0.068) and therefore did not enter the primary model, although it was associated with higher FoP in the clinically specified model (≥3 admissions, B = 1.918, P = 0.004; Supplementary Table S4), and recurrent hospitalization remains a recognized source of physical suffering and cumulative financial strain in this population.25 These cross-sectional associations are consistent with earlier work and may help generate hypotheses about which patients are more vulnerable to FoP, but they do not by themselves establish rules for risk stratification.
The central finding is that a past negative time perspective remained independently associated with FoP after demographic and clinical confounders were controlled. The association survived Holm and Benjamini–Hochberg correction within the five time perspective coefficients and was retained in every sensitivity analysis (clinically specified covariates, categorical coding, robust standard errors, bootstrap confidence intervals, and the 11-item FoP score), so it is the more robust of the two time perspective findings. The past negative dimension also correlated positively with FoP (r = 0.260, P < 0.01), in line with observations in other chronic disease populations.26,27 One possible explanation, offered as a hypothesis and not as a finding of this study, is that recurrent acute exacerbations and the distress of symptomatic episodes in CHF leave persistent fearful memories, so that patients with a strong past negative orientation read their current state through the lens of earlier suffering.
We did not measure illness memories or symptom appraisal, and the design is cross-sectional, so the data cannot distinguish this account from reverse causation, in which more intense fear shapes how the past is recalled.
The present fatalistic dimension was positively associated with FoP in the primary model (r = 0.210, P < 0.01; β = 0.128, P = 0.029), but this association did not survive Holm or Benjamini–Hochberg correction in that model, and its adjusted significance depended on the covariate set: it remained significant after Holm correction when covariates were chosen on clinical grounds (adjusted P = 0.008) but not with categorical coding of age, NYHA class and LVEF (adjusted P = 0.128). We therefore regard it as suggestive and in need of replication, not as established on the same evidential footing as the past negative association. Chronic illness often breeds a sense of uncontrollability,28 and fatalistic orientations predict anxiety and depression29 and drain the motivation for self-care, shading into resignation.30 Thapa et al30 found that fatalism mediates the link between symptom burden and self-care management in heart failure.
It is conceivable that less self-care, worse clinical status, and more fear of progression reinforce one another, but our design cannot test that sequence.
Past positive, future, and present hedonistic time perspectives correlated with FoP in the bivariate analysis, but none was significantly associated with FoP in the primary model. The future dimension reached nominal significance in the clinically specified model (B = −0.222, P = 0.021; Holm-adjusted P = 0.062) and in the 11-item sensitivity model (P = 0.022), but not in the primary model, with categorical coding, or with robust standard errors, so we draw no conclusion about a protective role of future-oriented thinking. Part of the bivariate associations may run through sociodemographic factors—age, living arrangements, and disease severity—that are related to both temporal orientation and FoP.
Several approaches that act on temporal cognition and emotion regulation, including time perspective therapy, mindfulness-based cognitive training, and emotion regulation training, have been studied in other populations.31–33 This study did not evaluate any intervention, and its cross-sectional associations cannot show that modifying time perspective would reduce FoP, improve adherence, or change quality of life or prognosis in patients with CHF; these approaches can be offered only as hypotheses for future trials. Self-management and patient engagement are central to heart failure care, and digital health tools are increasingly used to support them.34 Whether such support interacts with patients’ temporal orientations is not known and was not examined here.
Limitations
This study has several limitations. First, and most important, the primary outcome measure performed poorly in this cohort. Internal consistency of the FoP-Q-SF was lower than in its original validation study (α = 0.675 and ω = 0.706 vs α = 0.883), the two-domain structure was not reproduced, and the shortfall was most marked in the social-family subscale (α = 0.234) though also present in the physical health subscale (α = 0.640); 7 of the 12 items had corrected item-total correlations below 0.30 and one was negative.
Because that one item accounted for much of the shortfall, we repeated the analysis with the total recomputed from the remaining 11 items, which raised α to 0.722. The two totals correlated at r = 0.977, model fit was marginally better rather than worse (R2 = 0.506 vs 0.496), and all six independent predictors retained significance, including past negative (B = 0.196, P = 0.003) and present fatalistic (B = 0.278, P = 0.040) time perspective. This does not establish the construct validity of either the original or the modified score in heart failure; the validity of the primary outcome is not guaranteed, so every FoP finding, and the classification based on the ≥34 cut-off in particular, should be treated as exploratory until the instrument has been validated in patients with heart failure.
Second, the cross-sectional design cannot tell us whether maladaptive time perspectives precede the fear or whether intense fear entrenches a negative, fatalistic outlook—only longitudinal and interventional work can settle the direction. Third, the observed effect sizes (r values) were small to moderate, so the clinical weight of time perspective should not be overstated; unmeasured factors—most importantly depressive and anxiety symptoms, which were not assessed in this study and are known to covary with both time perspective and FoP—as well as social support, coping styles, and personality traits, likely account for much of the variance, so the present findings should not be read as isolating an effect of time perspective. Fourth, multiplicity and model specification are further concerns.
Of the five time perspective coefficients, only the past negative association survived correction in the primary model, while the present fatalistic association did not (Holm-adjusted P = 0.116; Benjamini–Hochberg q = 0.073) and survived Holm correction only in the clinically specified model; it should be treated as suggestive and in need of replication. The primary model selected covariates by bivariate P value, which can omit clinically relevant confounders and give unstable estimates; the clinically specified model, categorical coding, and robust and bootstrap estimates were added to examine this, but they were specified after the primary model had been fitted, the clinically specified model did not include socioeconomic indicators other than educational level, and residual confounding cannot be excluded. Fifth, the 5–10 participants-per-variable rule is a heuristic and the power figures are sensitivity statements and not a substitute for an a priori sample-size calculation; a preliminary analysis of the first 205 questionnaires preceded the final analysis, as described in Methods.
Sixth, for 20 cells in 19 questionnaires the item was left blank on the paper form and a value was later entered by the investigators without a documented source; these values were probably inferred and were therefore treated as missing in all analyses, with available-case and complete-case analysis rather than imputation, which reduced the regression sample to 237 participants. Seventh, occupational status and main source of income were near-redundant in this sample, and occupational status was dropped from the final model on that ground; because the choice between two collinear variables is not settled by the data alone, the coefficients of the retained income-source dummies should not be given a causal reading. The standardized residuals also departed from normality (skewness −1.33, excess kurtosis 4.15), driven by a small group of patients with very low FoP scores, which is why robust and bootstrap estimates were added.
Eighth, the number of medications taken did not meet the bivariate screening criterion (P = 0.093) and was therefore not entered into the primary regression, although it was included in the clinically specified model; we draw no inference from it. Ninth, a single-center convenience sample limits generalizability; the numbers of patients approached and declining were recorded only for the first 211 patients approached (6 declined, reasons not recorded), so an overall response rate cannot be reported, and 11 patients who returned the questionnaire without completing the two scales were excluded; bias arising from non-participation and incomplete response therefore cannot be excluded, and multi-center studies with larger samples are needed. Finally, all measures were self-reported, so response bias cannot be excluded; future work should add more objective assessment methods.
Conclusion
In this single-center cross-sectional sample of patients with CHF, a past negative time perspective was independently associated with greater FoP, and the association remained after correction for multiple comparisons and across alternative model specifications. A present fatalistic orientation showed a weaker, exploratory association whose significance after multiplicity correction depended on the model specification. Because the design was cross-sectional, the FoP-Q-SF showed limited reliability in this sample, and its cut-off has not been validated in heart failure, these findings do not establish that time perspective influences FoP, nor that interventions targeting time perspective would improve adherence, quality of life, or prognosis.
Validation of the FoP instrument in heart failure, longitudinal studies to clarify directionality, and interventional studies are needed before implications for clinical practice are drawn.