The mediating effect of health anxiety in the relationship between functional somatic symptoms and illness behavior in Chinese inpatients with depression

In this cross-sectional study, 323 inpatients with depression were recruited using convenient sampling from two large hospitals in China from April 2017 to March 2018: The Second Xiang-Ya Hospital, Central South University, and the Second Affiliated Hospital of Guizhou Medical University. All patients enrolled needed to meet the recruitment criteria: 1) diagnosis of major depressive disorder according to the DSM-V by two psychiatrists with more than five years’ clinical experience and 2) ability to understand the contents of all measures. Exclusion criteria were as follows: 1) comorbidity with bipolar disorder, schizophrenia, alcohol/substance use disorders, and other psychotic disorders and 2) major medical abnormalities, including central nervous system diseases or acute, unstable, or life-threatening medical illnesses (e.g., cancer, infections). All participants provided signed, informed consent to participate in this study. Ethical approval was obtained from the Mental Health Institute of Central South University Second Xiang-Ya Hospital.

Following the completion of informed consent forms, patients filled in a series of questionnaires including the Patient Health Questionnaire (PHQ-15), the scale for the Assessment of Illness Behavior (SAIB), and the Whiteley Index-7 (WI-7). Basic information such as gender, age, education, and illness duration was extracted from the patients’ medical records.

Descriptive statistics and Pearson’s correlation analyses of three variables (FSS, HA, and IB) were conducted with SPSS 22.0 software. The mediating role of HA in the relationship between FSSs and IB was tested using structural equation modeling (SEM) in AMOS 21.0 with the maximum likelihood estimation method conducted for model estimation. The goodness of fit of the model was evaluated using the following indices: CMIN/DF (a value between 1 and 5 indicates acceptable fit between hypothetical model and sample data), RMSEA (< 0.1 reflects reasonable model fit) [32], SRMR (< 0.08 indicates acceptable fit) [33], and CFI (> 0.9 indicates acceptable fit) [34]. A multivariate regression analysis of FSS, HA, IB, and age was performed with Hayes’ SPSS-PROCESS program model 59 to examine the adjusted moderating effect model of age [35]. Bootstrapped method was used to estimate the significance of indirect effects by 5000 bootstraps and 95% confidence interval.P-values < 0.05 were considered statistically significant.

A total of 323 inpatients with depression were collected in our study; 153 (47.7%) were female. Age ranged from 14 to 68 years, and the mean ± SD of the total sample was 33.84 ± 12.35 years (34.16 ± 12.91 in male and 33.48 ± 11.72 in female participants). The duration of education was 12.60 ± 3.30 years. The mean duration of illness (according to electronic medical records) was 21.8 ± 47.17 months.

Table 1 shows the results of the correlational analyses. The PHQ-15 score was negatively correlated with the SAIB score (r = − 0.50) and its five subscales (r = − 0.32~ − 0.47), which means that the severer the SSs, the severer the dysfunctional IB. The PHQ-15 score was positively correlated with the WI-7 score (r = 0.53). In addition, WI-7 also negatively correlated with SAIB (r = − 0.53) and its five subscales (r = − 0.21~ − 0.55), which means that people who have higher HA will have severer dysfunctional IB. The above results support the hypothesis of relationship among the three, and we conducted an intermediary model to further validate the hypothesis.

SEM was used to explore the mediating role of HA in the relationship between FSSs and IB. Most of the fit indicators (CMIN/DF = 4.89, IFI = 0.93, NFI = 0.91, CFI = 0.93, SRMR = 0.05) were acceptable, except RMSEA = 0.11. Therefore, we modified the model according to modification indices, revealing correlation between error3 and error4, which were latent variables of IB. After updating the model, the mediation model (see Fig. 1) showed an acceptable fit (CMIN/DF = 3.79, IFI = 0.95, NFI = 0.94, CFI = 0.95, SRMR = 0.04, RMSEA = 0.09), and all path coefficients were significant at the 0.001 level. The significance of the mediating effects of HA was tested by the bootstrap estimation procedure in AMOS (a bootstrap sample of 5000 was specified). As displayed in Table 2, SSs exerted significant indirect effects on IB via HA. HA also had a partial mediating effect in the relationship between SSs and IB; the indirect effects accounted for 42.86% (Indirect effect/Total effect) of the total effect.

The PHQ-15 score was set as an independent variable and SAIB total score as a dependent variable, the WI-7 score had a mediating effect, and age was a moderating variable. Multiple regression analysis was performed to test which path in the mediation model was adjusted. The result showed that the effect of age on the relationship between FSSs and HA was statistically significant. However, there was no statistically significant effect of the moderating in the relationship between HA and IB, and there was also no statistically significant moderating effect on the relationship between FSSs and IB (Table 3).

To further test the moderating effects of age between SSs and HA, subjects who aged above M + SD (33.84 + 12.35) were set as the high age group and below M − SD (33.84–12.35) as the low age group. The results showed that the moderating effect was established in the low age group (effect = 0.48; 95% CI [0.39 ~ 0. 57]), and the moderating effect also existed in the high age group (effect = 0.29; 95% CI [0.20~0.38]). Furthermore, there was a significant difference between two age groups (t = − 2.7585; p = 0.0061). Those indicated a moderate–mediation model. (Fig. 2).

Correlational analyses found an intensifying relationship between any two of FSS, HA, and IB. These findings are consistent with most previous literature suggesting that both elevated HA and severer SSs are significantly associated with more abnormal IB [9, 36‐39] and with prior studies suggesting that severer SSs are linked to elevated HA [15, 17].

More importantly, according to the results of SEM, HA accounted for 42.86% of the total effect of FSSs on IB. In other words, increasing HA is an indirect approach of FSSs to elevate IB. To the best of our knowledge, this is the first study verifying the mediating role of HA between FSSs and IB and clarifying the portion of this mediating effect in the total effect of FSSs. Given that the remaining total effect may consist of a direct effect of FSSs and indirect effects of other variables beyond the present study’s interest, this finding emphasizes that HA is an important intermediate mechanism when we consider how FSSs influence IB.

The sample of the present study was inpatients with depression, most of them with FSSs, which should be noted to understand the results. It is well known that depression is often accompanied by FSSs [1, 3]. Depression and anxiety often accompany each other, and severe HA was suggested to be reclassified as an anxiety disorder in the DSM-V [40]. Hence, we speculate that patients with depression may have higher anxiety qualities than other people. These people are more sensitive and excessively attend to their own specific signs and symptoms, subsequently worrying about their health and attributing these FSSs to organic causes rather than functional ones, resulting in elevated HA. Typically, patients with high HA tend to hold the catastrophic interpretation with FSSs (this headache means a brain tumor) [41]. To seek reassurance about the fear of such symptoms and to alleviate this kind of HA, patients often enact inappropriate IB [36, 42]. Therefore, it seems reasonable that HA would be a potent mediating factor of this association between FSSs and IB.

Another interesting finding was that the age of patients with depression played a moderating role in the relationship between FSSs and HA. Although there were significant positive relationships between FSSs and HA in both of younger and older groups, a more thorough analysis revealed the age effect. Specifically, compared with the younger group, the older patients had higher HA when FSSs were mild. Given that both groups reached the same levels of HA when FSSs were severe, this pattern of age effect suggests that older patients with depression may overreact to mild FSS.

It is necessary to contrast the age effect found in the present study with the findings of previous studies. Previous research showed that older adults (over age 64) report relatively low levels of worrying about health issues than do college students [43]. The current research indicated that level of HA was higher in the high-aged group (33.84 + 12.35) than in the low-aged group (33.84–12.35), but this growth trend gradually slowed with increasing age. The explanation for the inconsistent result may be that our study focused on a depressed population who may have high anxiety characteristics while the prior research focused on the general population. Another hypothesis from other researchers is that HA may peak at a certain age and decrease thereafter [44, 45]. Our study has an age range from 14 to 68 years; therefore, it is necessary to investigate in a larger age range allowing for comparison across the entire lifespan.

The results of the present study have multifaceted value for clinical practice with patients with depression disorder. First, the partially mediating role of HA between FSSs and IB highlights the need to pay more attention to HA (especially in high-aged patients with depression with SSs, even if mild) when treating depression disorder rather than dealing only with the depressive symptoms. Second, the results suggest that HA can be a promising intervention target for reducing depressive patients’ inappropriate IB. Therefore, we recommend a regular assessment of FSSs and HA in patients with depression. For those with mild HA, the doctor should explain the nature of the symptoms to the patient, thereby avoiding more unnecessary HA and IB. For those with severe HA, timely and effective intervention is necessary. In addition, it is recommended to actively evaluate and intervene in HA in elderly patients with depression with FSS (even if mild).

Currently, the effectiveness of antidepressants in treating FSSs is far from ideal; there are often residual physical symptoms [46, 47]. Studies have shown that the treatment response of HA was much better than FSSs [48]. Thus, the HA intervention becomes more important and efficient in reducing the IB of patients with depression. High levels of health anxiety might be a multidimensional trait, where triggering factor (such as physiological processes), cognitive and behavioral strengthen each other. Although avoiding triggers and using safe behaviors (such as more examinations of symptoms) may lead to a reduction in health anxiety in the short term but higher levels of health anxiety in the long run [49]. Fortunately, psychoeducation, exposure and response prevention, antidepressants, and cognitive restructuring techniques might be helpful for patients with severe health anxiety. Specifically, cognitive-behavioral therapy has been reported to be a highly effective treatment for hypochondriasis/HA [18, 50, 51], even better than drug treatment [52]. The strengths of the cognitive behavioral model of severe health anxiety lies in its account of maintaining factors. The maintenance factor of health anxiety lies in the faulty cognitive belief of bodily sensations and external events. Consequently, the hallmark of cognitive therapy is that it entails components aiming to change these faulty beliefs. It can alleviate HA by helping patients recognize and modify false beliefs about the symptoms [41] . A recent study found that cognitive-behavioral therapy can reduce perceived risk of disease, attention to bodily symptoms, and intolerance of uncertainty significantly to improve HA [53].

A number of limitations of the present study should be considered. First, given the cross-sectional design, the current study could not infer the causal nature of the associations among FSS, HA, and IB. Thus, future study should examine the associations of these variables using a longitudinal design or an experimental manipulation method that allows causal relationship. Second, the present study examined the mediating role of HA in the relationship between FSSs and IB in a clinical depressed sample, whether this conclusion can be extended to other mental disorders requires further confirmation. Therefore, future research should utilize a longitudinal design to examine this association in other samples to determine the clinical relevance of the findings. Third, we did not consider other factors that previous studies have reported that may affect IB, such as personality traits [54] and depression [7] . It is worth mentioning that we found no association between SAIB and the depression scale of Patients Health Questionnaire-9 in the present study, which is consistent with the finding of Rief et al. [29]. While this result is in contrast to results of Wilson Barnett and Trimble who using Illness Behavior Questionnaire (IBQ) as assessment tool [55]. The possible reason for this inconsistence may be that IBQ focus on evaluation of emotional aspects and hypochondriacal concerns rather than aspects of illness behavior (SAIB). However, because the subjects recruited are hospitalized patients with severe symptoms of depression in our study, depression may still have certain effects on variables in the mediation model. It would be helpful to validate the mediation model in the general population. Despite these limiting factors, the strength of this study is the emphasis on the mediating effect of health anxiety in clinical major depressed samples. In future, the result can be further validated in a population-based study and non-hospitalized sample in order to control these effects.