Persistent physical symptoms after COVID-19 infection and the risk of Somatic Symptom Disorder

Abstract

Objective

Evidence shows that many patients with COVID-19 present persistent symptoms after the acute infection. Some patients may be at a high risk of developing Somatic Symptom Disorder (SSD), in which persistent symptoms are accompanied by excessive and disproportionate health-related thoughts, feelings and behaviors regarding these symptoms. This study assessed the frequency of persistent physical symptoms and SSD and their associated factors in patients with confirmed COVID-19.

Methods

We conducted a longitudinal retrospective study after the first two French lockdowns at the Lille University Hospital (France), including all patients with confirmed COVID-19. Persistent physical symptoms and excessive preoccupations for these symptoms were measured 8 to 10 months after the onset of COVID-19. The combination of the Patient Health Questionnaire-15 and the Somatic Symptom Disorder-B Criteria Scale was used to identify the individuals likely to present with SSD. Two linear regression models were performed to identify sociodemographic and medical risk factors of SSD.

Results

Among the 377 patients with a laboratory-confirmed diagnosis, 220 (58.4%) completed the questionnaires. Sixty-five percent of the 220 included patients required hospitalization, 53.6% presented at least one persistent physical symptom and 10.4% were considered to present SSD. Female sex, older age, infection during the second wave and having probable PTSD were significantly associated with the severity of SSD and SSD was associated with a significantly higher healthcare use.

Conclusions

The identification of SSD should encourage clinicians to move beyond the artificial somatic/psychiatric dualism and contribute to a better alliance based on multi-disciplinary care.

1. Introduction

Many patients with COVID-19 present with persistent symptoms or even develop new ones after the acute infection phase [[1], [2], [3]]. Some of these patients may experience a detrimental impact of these persistent symptoms on their quality of life, thus suffering from a “post-COVID condition”, frequently referred to as “long COVID” [[4], [5], [6]]. The most frequently observed symptoms include fatigue, headache, dyspnea, cognitive impairment, sleep disorders, muscle pain, concentration problems, and digestive symptoms [7]. The UK Office for National Statistics (ONS) estimates that 21% of patients with COVID-19 display at least one symptom at 5 weeks and 9.9% at 12 weeks post-infection (ONS 2020).

Despite increasing research regarding risk factors and sequelae of persistent physical symptoms after COVID-19, the pathophysiology of long COVID remains poorly understood [3]. Importantly, persistent symptoms are not only related to the severity of the acute infection, as evidence shows that it may concern patients with severe forms but also those with very mild acute disease [1,8,9]. A study including 26,823 adults from a large population-based French cohort even found that persistent physical symptoms may be associated more with self-reported COVID-19 than with having laboratory-confirmed COVID-19 infection [10], suggesting that mechanisms accounting for long COVID may not be specific to SARS-CoV-2.

As in other poorly understood conditions such as fibromyalgia or chronic fatigue syndrome, patients with long COVID often feel dismissed or overlooked [11], and the fear associated with the lack of knowledge about the disease and its long-term consequences often make it difficult to reassure them regarding their symptoms. The current COVID-19 pandemic mirrors past pandemics and infectious diseases that have been related to a high risk of developing Somatic Symptom Disorder (SSD) [12,13]. SSD, previously known as somatoform disorders, has been defined in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition [14] as persistent and clinically significant somatic symptoms (criterion A) accompanied by excessive and disproportionate health-related thoughts, feelings and behaviors (criterion B) regarding the symptoms, which typically persist at least for 6 months (criterion C). SSD can result in long-term negative consequences including impairment of patients' quality of life and level of functioning [15] as well as high levels of health care use, with repeated investigations and overtreatment [16].

SSD has been proposed as a perceptual disorder [17] in which adverse events, dysfunctional cognitions, negative affectivity, or maladaptive behaviors influence the perception, perpetuation, and deterioration of somatic symptoms [18,19]. The COVID-19 pandemic context regroups many factors which are likely to influence the perception of symptoms and their intensity and to contribute to the development of SSD, such as social isolation, adverse media exposure, and limited access to health services [13]. The psychological distress induced by the disease may also be a critical factor. Indeed, evidence shows that some patients with COVID-19 develop a post-traumatic stress disorder (PTSD) in relation to the disease [20] and that these patients present with a higher prevalence of persistent COVID-related physical symptoms than COVID-19 patients without PTSD [21]. Nevertheless, to date, no study has assessed SSD in the context of COVID 19.

The current study thus assessed the prevalence of persistent symptoms and SSD in patients with laboratory-confirmed COVID-19. It also aimed at identifying predictive factors of developing SSD in this population.

2. Materials and methods

2.1. Study design and population

We conducted a longitudinal retrospective study after the first two French lockdowns between March 17 and May 11, 2020 (wave 1), and between October 29 and December 15, 2020 (wave 2) at the Lille University Hospital (France). In this center, patients could be tested if they had symptoms, if they had been in contact with patients with COVID-19, or if they came back from a country impacted by the COVID-19 pandemic. During the first lockdowns, clinicians of the unit of consultation-liaison psychiatry had to modify and adapt their activity. They thus developed a systematic assessment of the mental health of all patients in COVID-19 units, to detect potential symptoms or disorders and propose adapted care. Patients were also proposed to be recontacted a few months later for a new assessment. All medical and demographic data are recorded in the patient medical records. This was realized through interviews and phone calls, with systematic recontacts.

Patients were eligible for inclusion if they had a laboratory-confirmed diagnosis of COVID-19 if they volunteered to participate and were aged 18 or older. Patients were excluded when they presented communication problems that prevented their ability to respond to questionnaires (e.g., deafness, dementia, mental retardation). Written informed consent was obtained from all patients.

Patients answered an online questionnaire 8 to 10 months after they had a confirmed diagnosis of COVID-19. In case of no response after 2 weeks, a reminder was sent by phone. The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008. In France, confidentiality approval from the Commission Nationale de l'Informatique et des Libertés (CNIL) is sufficient for non-interventional observational studies that does not require the collection of additional information compared to standard patient management; ethical approval is not mandatory (Law n° 2004–800 on bioethics, Aug. 6, 2004). This study was approved by the CNIL and the Local Data Protection Service (DEC20–155).

2.2. Collected data

2.2.1. Somatic symptom disorder assessment

Patients were asked to complete the Patient Health Questionnaire-15 (PHQ-15) and the Somatic Symptom Disorder-B Criteria Scale (SSD-12). They were asked to assess the symptoms that had developed after the COVID-19 infection onset.

We used the PHQ-15 score to assess the presence and severity of persistent physical symptoms (criteria A of SSD) related to COVID-19 [[22], [23], [24]]. The PHQ-15 is one of the most frequently used instruments to identify people with elevated symptom burden. It assesses the presence and severity of common physical symptoms, such as fatigue, gastrointestinal, musculoskeletal, pain, and cardiopulmonary symptoms within the last 4 weeks using 15 items related to 15 somatic symptoms. Each symptom is scored from 0 (“not bothered at all”) to 2 (“bothered a lot”), with a total score ranging from 0 to 30. Cronbach's α for the PHQ-15 generally reaches around 0.80 [25].

The SSD-12 is a brief 12-item self-report questionnaire that has been developed to assess the criteria B of DSM-5 somatic symptom disorder, i.e., the cognitive, affective, and behavioral aspects of SSD [26]. Each of the three psychological sub-criteria is measured by four items on a scale from 0 (“never”) to 4 (“very often”), with a total score ranging from 0 to 48. The SSD-12 has good item characteristics and excellent reliability (Cronbach's α = 0.95) [19,26].

Since, PHQ-15 and SSD-12 have not been validated in French, we translated the English version following a translation and back-translation procedure (supplementary file 1). Cronbach's alpha and confidence interval was 0.91 [0.88–0.93] for PHQ-15 and 0.95 [0.93–0.96] for SSD-12.

The combination of the PHQ-15 and the SSD-12 has been shown as a reliable and valid self-report measure allowing the identification of individuals likely to present with SSD [19]. Optimally combined cut-offs are ≥9 for the PHQ-15, and ≥ 23 for the SSD-12 (sensitivity and specificity reach 78% and 56% for an SSD-12 ≥ 23, and 69% and 70% for both an SSD-12 ≥ 23 and a PHQ-15 ≥ 9) [19,26].

2.2.2. Covariates

In addition to the SSD-12 and the PHQ-15, patients were also asked to declare if they had consulted their general practitioner or a specialist since they had contracted the COVID-19, and to complete the PTSD Checklist for DSM-5 (PCL-5), a 20-item scale exploring PTSD symptom severity over the past month, in which COVID-19 infection was presented as the stressful experience. A score greater than or equal to 33 at the PCL-5 indicates a probable diagnosis of PTSD [27].

Medical and demographic data, retrospectively extracted from the patient medical records, included age, sex (male, female), living situation (alone or not), being a healthcare worker (yes, no), history of psychiatric disorders (yes, no), pre-existing physical conditions (including obesity, cardiovascular disease, diabetes, pulmonary disease, and immunodepression; yes, no), type of care related to the COVID-19 (ambulatory, conventional hospitalization, or hospitalization including a stay in an intensive care unit (ICU)), impact of COVID-19 on relatives (none, affected relative or death of a relative).

2.3. Statistical analysis

First, we described the sample using means and standard deviations or medians with interquartile ranges (IQR) for quantitative variables that were normally distributed or not, respectively. Numbers and percentages were used for qualitative variables. We compared the baseline characteristics of patients who completed the questionnaire with those who did not, using t-tests for quantitative variables and Chi-square tests (or Fisher tests, the non-parametric counterpart) for qualitative variables. Median SSD-12 and PHQ-15 scores were described, as well as the proportions of participants reporting a score above the cut-offs.

The prevalence of SSD, and its 95% confidence interval (95% CI), as defined by the combination of the two scores above the cut-offs, was calculated.

Due to an insufficient number of cases to perform a logistic regression model explaining the presence of SSD (as defined by the combination of the two scores), we chose to perform two linear regression models to identify the risk factors for SSD: one explaining the PHQ-15 score and one explaining the SSD-12 score. All covariates were included in the models. Associations between risk factors and outcomes were presented as the difference (d) in mean PHQ-15 and SSD-12 scores and their 95% confidence intervals (95% CI).

The conditions of application of the models have been checked. There was no collinearity in the models insofar as the variance inflation factors were all <5 [28]. But, due to the non-normal distribution of the PHQ-15 and SSD-12 scores, sensitivity analysis explaining the log-transformed variables were performed (log(Y + 1) transformation to deal with zero values). The results were similar. Furthermore, given the response rate, weighted models have also been carried out, considering the inverse of the inverse of a propensity score obtained through a logistic regression model predicting response to the questionnaire. The variables available for respondents and non-respondents were included in the model: recruitment period, sex, age, healthcare worker, pre-existing physical condition, living alone, relatives, type of care, and psychiatric history. Results are provided in supplementary file 2.

Data analysis was performed using R 3.6.1. The significance level was set at α = 0.05, and all tests were 2-tailed.

3. Results

3.1. Comparison between respondents and non-respondents

Among the 377 patients with a laboratory-confirmed diagnosis, 220 (58.4%) completed the questionnaires. The groups of respondents and non-respondents were comparable on all except two variables: the period of recruitment, and the presence of physical comorbidities (see Table 1 ). People infected during wave 2 and people with pre-existing physical comorbidities were under-represented among the respondents, compared to the non-respondents (54.1% vs 77.7%, p < 0.001, concerning wave 2, and 43.2% vs 54.8%, p = 0.034, concerning pre-existing health conditions).

Table 1.

Comparison between respondents and non-respondents.

	Non-respondents	Respondents	p
	N = 157	N = 220
Recruitment period, n (%)
Wave 1	35 (22.3)	101 (45.9)	<0.001
Wave 2	122 (77.7)	119 (54.1)
Sex, n (%)
Male	89 (56.7)	122 (55.5)	0.895
Female	68 (43.3)	98 (44.5)
Age, m (sd)	52.9 (15.9)	52.9 (15.4)	0.991
Healthcare worker, n (%)
No	97 (61.8)	148 (67.3)	0.500
Yes	51 (32.5)	65 (29.5)
NA	9 (5.7)	7 (3.2)
Pre-existing physical condition, n (%)
No	71 (45.2)	125 (56.8)	0.034
Yes	86 (54.8)	95 (43.2)
Living alone, n (%)
No	128 (81.5)	180 (81.8)	1.000
Yes	27 (17.2)	37 (16.8)
NA	2 (1.3)	3 (1.4)
Relatives, n (%)
Deceased	4 (2.5)	11 (5.0)	0.132
Infected	96 (61.1)	107 (48.6)
Non infected	53 (33.8)	83 (37.7)
NA	4 (2.5)	19 (8.6)
Type of care, n (%)
ICU	44 (28.0)	53 (24.1)	0.248
Hospitalization	51 (32.5)	90 (40.9)
Ambulatory	62 (39.5)	77 (35.0)
Psychiatric history, n (%)
No	140 (89.2)	196 (89.1)	0.886
Yes	16 (10.2)	20 (9.1)
NA	1 (0.6)	4 (1.8)

3.2. Sample characteristics

A total of 220 patients were analyzed (see Table 1). Almost half fo the sample (54.1%) was recruited during the 2nd wave. The sample was composed of 98 men (55.5%) and 65 of the 220 participants (29.5%) were health-care workers. The mean age was 53 (±15) years, and 37 patients (16.8%) declared they lived alone. Regarding medical history, 20 patients (9.1%) had a history of a psychiatric disorder and 95 (43.2%) had pre-existing physical comorbidity. The medical management of COVID-19 required hospitalization for 143 patients (65%), including a stay in an ICU for 53 patients (24.1%). Among the 220 participants, 118 (53.6%) reported having a relative also affected by COVID-19 and among them, 11 (5.0% of the global sample) had lost a relative due to the disease. Finally, the median PCL-5 score [IQR] was 4 [[1], [2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12], [13], [14]], and 16 patients (7.3%) presented a score greater than or equal to 33.

3.3. Somatic symptom disorder assessments

The median PHQ-15 score [IQR] was 6 [[2], [3], [4], [5], [6], [7], [8], [9], [10], [11], [12]], and 86 (39.1%) reported a score greater than or equal to 9 (see supplementary file 2 for item-by-item responses). Most participants reported at least one persistent symptom at the time of the survey: 53.7% [46.8–60.3], 36.9% [30.5–43.6], and 25.5% [19.9–31.8] reported at least one, two or three persistent symptoms, respectively. The median SSD-12 score [IQR] was 4 [0–14], and 27 patients (12.3%) presented a score greater than or equal to 23. By considering patients whose SSD-12 and PHQ-15 scores exceed the cut-offs as likely to present with SSD, we identified 23 participants with SSD, i.e., a prevalence [95%CI] of 10.4% [6.9–15.4] (see Fig. 1 ).

Fig. 1.

Proportions of participants with a score above the cut-off on the PHQ-15 scale or/and on the SSD-12 scale.

SSD-12 score increased with the number of persistent symptoms (p < 0.001) (see Fig. 2 ).

Fig. 2.

SSD-12 score according to the number of persistent symptoms rated 2 (“bothered a lot”) on the PHQ-15 scale.

3.4. Factors associated with SSD

Bivariate analyzes showed significant associations between gender and SSD-12 and PHQ-15 scores: female gender was associated with higher scores (d = 4.12 [1.36–6.88], p = 0.004, and 2.32 [0.59–4.05], p = 0.009, respectively). Compared to participants with a PCL-5 score under 32, those with a higher score also presented higher SSD-12 and PHQ-15 scores (d = 22.51 [17.95–27.06], p < 0.001, and 9.44 [6.30–12.59], p < 0.001, respectively). Participants recruited during the 2nd wave also presented higher SSD-12 and PHQ-15 scores (d = 4.19 [1.44; 6.94], p = 0.003, and 6.23 [4.69; 7.78], p < 0.001, respectively). Finally, older participants had a significantly higher SSD-12 score (d = 0.09 [0.00–0.18], p = 0.045) (see Table 2 ).

Table 2.

Factors associated with SSD-12 and PHQ-15 scores: results of the bivariate analyzes.

	SSD-12 score		PHQ-15 score
	d⁎ [95%CI]	p	d⁎ [95%CI]	p
Sex
Male	Ref		Ref
Female	4.12 [1.36; 6.88]	0.004	2.32 [0.59; 4.05]	0.009
Age	0.09 [0.00; 0.18]	0.045	0.05 [−0.01; 0.11]	0.074
Recruitment period
Wave 1	Ref		Ref
Wave 2	4.19 [1.44; 6.94]	0.003	6.23 [4.69; 7.78]	<0.001
Healthcare worker
No	Ref		Ref
Yes	−2.26 [−5.37; 0.84]	0.152	0.16 [−1.78; 2.10]	0.870
Pre-existing physical condition
No	Ref		Ref
Yes	1.62 [−1.19; 4.44]	0.256	1.16 [−0.60; 2.92]	0.196
Living alone
No	Ref		Ref
Yes	−2.77 [−6.11; 0.57]	0.104	1.43 [−0.91; 3.77]	0.230
Relatives
Deceased	Ref		Ref
Infected	2.24 [−4.47; 8.96]	0.511	1.86 [−2.25; 5.98]	0.373
Non infected	2.60 [−4.20; 9.41]	0.451	1.47 [−2.70; 5.65]	0.487
Type of care
ICU	Ref		Ref
Hospitalization	0.86 [−2.73; 4.44]	0.638	1.40 [−0.84; 3.65]	0.218
Ambulatory	−1.10 [−4.80; 2.60]	0.558	0.43 [−1.88; 2.75]	0.712
Psychiatric history
No	Ref		Ref
Yes	2.87 [−1.98; 7.73]	0.245	2.62 [−0.40; 5.65]	0.089
PCL score > 32
No	Ref		Ref
Yes	22.51 [17.95; 27.06]	<0.001	9.44 [6.30; 12.59]	<0.001

^⁎d = difference in mean scores.

After adjustment, female gender and PCL-5 score (as a binary variable: >32 or not) were significantly associated with both higher SSD-12 score and higher PHQ-15 score. Compared to men, women had a 4.59 [1.66; 7.51] point higher SSD-12 score (p = 0.002) and a 2.73 [0.89; 4.56] point higher PHQ-15 score (p = 0.004). Participants whose PCL-5 score was >32 reported a 23.48 [18.62; 28.34] point higher SSD-12 score (p < 0.001) and a 9.09 [6.03; 12.14] point higher PHQ-15 score (p < 0.001). The SSD-12 and PHQ-15 scores were also associated with age (d [95%CI]: 0.11 [0.00; 0.23], p = 0.039, and 0.10 [0.03; 0.17], p = 0.007) (see Table 3 ). Finally, participants recruited during the second wave had significantly higher SSD-12 (3.10 [0.11; 6.09], p = 0.042) and PHQ-15 (6.28 [4.40; 8.16], p < 0.001) scores.

Table 3.

Results of the linear regression model explaining SSD-12 and PHQ-15 scores.

	SSD-12 score		PHQ-15 score
	d⁎ [95%CI]	p	d⁎ [95%CI]	p
Sex
Male	Ref		Ref
Female	4.59 [1.66; 7.51]	0.002	2.73 [0.89; 4.56]	0.004
Age	0.11 [0.00; 0.23]	0.039	0.10 [0.03; 0.17]	0.007
Recruitment period
Wave 1	Ref		Ref
Wave 2	3.10 [0.11; 6.09]	0.042	6.28 [4.40; 8.16]	<0.001
Healthcare worker
No	Ref		Ref
Yes	−2.01 [−5.90; 1.89]	0.310	0.31 [−2.14; 2.76]	0.803
Pre-existing physical condition
No	Ref		Ref
Yes	−1.53 [−4.45; 1.39]	0.301	−0.70 [−2.53; 1.13]	0.453
Living alone
No	Ref		Ref
Yes	−2.75 [−6.06; 0.55]	0.102	0.73 [−1.35; 2.81]	0.490
Relatives
Deceased	Ref		Ref
Infected	0.49 [−5.16; 6.15]	0.863	0.50 [−3.05; 4.06]	0.780
Non infected	1.29 [−4.54; 7.13]	0.662	0.45 [−3.21; 4.12]	0.807
Type of care
ICU	Ref		Ref
Hospitalization	−0.79 [−4.02; 2.44]	0.630	−0.09 [−2.13; 1.94]	0.926
Ambulatory	0.23 [−4.54; 5.00]	0.924	−0.89 [−3.89; 2.11]	0.559
Psychiatric history
No	Ref		Ref
Yes	−2.16 [−6.80; 2.48]	0.359	1.43 [−1.48; 4.34]	0.334
PCL score > 32
No	Ref		Ref
Yes	23.48 [18.62; 28.34]	<0.001	9.09 [6.03; 12.14]	<0.001

N = 181 (39 observations deleted due to missingness).

^⁎d = difference in mean scores.

3.5. Healthcare use and SSD

Among the 23 participants with a probable SSD, healthcare use (consultation with a general practitioner or a specialist) was significantly higher than in the other participants. They were 22 (95.6%) participants with a probable SSD to declare having consulted their general practitioner, against 154 (78.2%) in the other group (p = 0.029). The consultation with the specialist concerned 20 (86.9%) participants with probable SSD, and 118 (59.9%) participants without SSD (p = 0.014).

4. Discussion

This study assessed the frequency of persistent physical symptoms and SSD as well as their associated factors in patients with confirmed COVID-19. In our sample that included many patients with a severe form of COVID-19, we showed that 53.6% of the 220 included patients presented at least one persistent physical symptom. Regarding SSD, 10.4% of the patients were considered to present SSD (PHQ-15 ≥ 9 and SSD-12 ≥ 23). Female sex, older age, having been infected during the second wave and having probable PTSD (PCL-5 > 32) were significantly associated with the severity of SSD, and SSD was associated with a significantly higher healthcare use.

These results indicate that about half of the patients exhibited physical symptoms >6 months after their infection, which confirms the important risk of persistent symptoms after COVID-19. Indeed, most previous studies showed that after acute COVID-19, >30% of patients present persistent symptoms and that the majority of patients presented at least one symptom several weeks after COVID-19 infection [4,9,28]. It should be noted, however, that the frequency of persistent symptoms after COVID-19 depends on the delay of their evaluation after the acute infection and the type and severity of symptoms [1]. What is more interesting is the frequency of SSD, which concerned one in ten patients of our sample. This prevalence was higher than the prevalence of SSD in the general population, usually estimated at around 4–5% [17,29]. It tends to confirm data from previous studies that have highlighted infection as a precipitating factor of SSD symptoms, suggesting that immunological mechanisms contribute to their development [30,31]. Nevertheless, in this pandemic context, other factors have likely contributed to the development of SSD, such as the social isolation, limited access to health service and fear associated with the lack of knowledge about the disease.

The risk factors of developing persistent symptoms (SSD criterion A) were also found to be risk factors of developing excessive preoccupations for these symptoms (SSD criterion B). This suggests that the intensity of physical symptoms may conduct to dysfunctional cognitions and behaviors, but also that the intensity of physical symptoms experienced by the patient and the severity of the psychological impact may rely on common factors. The main risk factor was the severity of PTSD symptoms. Having a probable PTSD (PCL-5 > 32) was associated with a substantial increase of 24 points on the SSD-12 scale (total score ranging from 0 to 48; cut-off =23). This result is in line with data showing that patients with COVID-related PTSD have a higher burden of persistent physical symptoms and are less likely to feel fully recovered after their COVID-19 illness [21,32]. The role of PTSD in the development of SSD has been previously described, with a meta-analysis indicating that individuals who suffer from PTSD are almost 3 times more likely to have SSD [33]. Several factors may explain why patients with PTSD are at high- risk of developing physical symptoms. PTSD is associated with autonomic nervous system and hypothalamic-pituitary-adrenal axis dysregulation which affects neuroendocrine and immune functions: it can therefore result in a wide range of somatic symptoms often classified as ill-defined or medically unexplained [34]. PTSD is also associated with limbic instability, particularly of the anterior insula which is involved in the perception of internal body states, and likely plays a role in the pathogenesis of physical symptoms observed in PTSD [34,35]. COVID-19-related symptoms may cause heightened psychological distress, dysfunctional cognitions, and maladaptive behaviors, which may in turn influence the perception and perpetuation of somatic symptoms and finally lead to SSD [18,19,21].

Being a woman was also associated with more severe SSD symptoms, which is a consistent finding in the SSD literature [29,36]. The link between age and the severity of physical symptoms may be explained by the increased likelihood of presenting with physical symptoms in older participants. Indeed, although SSD has received little attention in older people [37], physical illnesses in older people are known to be associated with prominent somatization [38,39]. Finally, people infected during the second wave were more likely to present SSD than people who were infected during the first wave. This results appear are at odds with results from previous study [40] and may be explained by the long term impact of the pandemic on mental health and less of psychological support and exceptional measures after the end of the first wave. The awareness of the long-term effects of the disease, including persistent physical symptoms may also have contributed to more preoccupations for these symptoms.

In agreement with the results obtained by previous studies [38,39,40], patients with SSD were found to report significantly higher health care use than other patients with COVID-19. This result confirms the high level of distress of these patients [15], which is likely related to the association of persistent physical symptoms and excessive and disproportionate health-related behaviors, both of them contributing to a risk of overdiagnosis and overtreatment [16,39].

This study presents several limitations. The main limitation is that approximately 41% of eligible patients did not respond to the questionnaire. However, we found significant differences between the groups of respondents and non-respondents for 2 variables only. First, there were more patients with pre-existing physical comorbidities in the non-respondent group. Since these patients are at high-risk of having persistent physical symptoms and/or excessive preoccupations, we may have underestimated the actual proportion of patients with probable SSD in patients with severe COVID-19. Second, there were more people infected during wave 2 in the non-respondent group. This might be related to the less strict lockdown conditions during the second wave, resulting in less availability to answer questionnaires.

The second limitation is that this is a monocentric study conducted in a University Hospital. Although the Lille University Medical Center is a Regional Hospital that admits patients from the whole of northern France, including cities that were differentially impacted by the epidemic, we included an important proportion of patients with a severe form of COVID-19. Third, the SSD diagnosis, which was based on the association of PHQ-15 and SSD-12 scales [19], was not confirmed by a psychiatric assessment. Moreover, we used a French version of SSD-12 that has not been validated yet, and applied the cut-off of the English version. It should also be noted that the PHQ-15 does not include all symptoms that have been described in post-COVID-19. Fourth, variables were collected from medical record at the time of inclusion, so any potential event that occurred after the inclusion period were not considered in the present study. Finally, while this study assessed SSD by considering both physical and psychological symptoms lasting at least 6 months after COVID-19 infection, SSD can appear later. It is therefore too early to consider that the present results reflect the definitive impact of the epidemic, and future studies should investigate the SSD prevalence among patients with COVID-19 > 6 months after the infection.

5. Conclusion

To the best of our knowledge, this study is the first to explore SSD in patients with COVID-19, revealing a prevalence of 10%. This confirms the hypotheses formulated by several authors regarding the high risk of developing SSD in the context of COVID-19 [12,13,16]. Importantly, the identification of SSD should encourage clinicians to move beyond the artificial somatic/psychiatric dualism and contribute to a better alliance based on multi-disciplinary care [13,41]. The diagnoses of SSD or related disorders are often delayed, after patients have received inappropriate and excessive diagnosis, testing, dangerous or ineffective therapies, and multiple, often unnecessary, referrals [42]. A positive diagnosis of SSD is crucial to identify the patients' concerns and enable an appropriate treatment [16]. Psychological interventions, especially cognitive behavior therapy, have been demonstrated to be effective in reducing SSD symptom severity but also secondary outcomes, such as level of functional disability or depressive and anxiety symptoms [43]. In addition, the identification of PTSD as the main risk factor for SSD should improve the detection and care of at-risk patients notably through preventive interventions. This study paves the way for future investigations regarding the long-term consequences of the COVID-19 epidemic on patients' mental health. Considering the great number of people at high-risk of persistent symptoms in the context of COVID-19 pandemic, and the functional consequences of SSD, it appears essential to detect the occurrence of SSD as early as possible.

Conflicts of interest and source of funding

The authors declare no conflicts of interest, this research received no specific grant from any funding agency, commercial or not-for-profit sectors.

Appendix A. Supplementary data

Translation procedure of the french version of The Somatic Symptom Disorder – B Criteria Scale A.

Results of the linear regression model explaining SSD-12 and PHQ-15 scores considering inverse probability weighting

Detailed item-by-item responses at the PHQ-15 scale