How do clinicians use post-COVID syndrome diagnosis? Analysis of clinical features in a Swedish COVID-19 cohort with 18 months’ follow-up: a national observational cohort and matched cohort study

Introduction

Infection with SARS-CoV-2 resulting in COVID-19 exhibits a wide range of symptoms including respiratory and pulmonary symptoms during the acute phase of disease.1–3 The key characteristics include remarkably variable individual health responses from those who are asymptomatic or have mild symptoms to those with life-threatening severe disease or even death.1 2 An increasing number of patients suffer from disabling long-term consequences, post-COVID syndrome (PCS, also known as Long COVID, post-acute sequelae of COVID-19 or post-acute COVID-19 syndrome). Furthermore, the symptomatology of PCS varies between patients creating a challenge both for disease diagnosis, management and public health overview of the extent of the problem.

The global consensus on how PCS should be diagnosed and treated is still developing based on the efforts from the WHO, patient organisations and the Centre of Disease Control in providing guidelines and International Classification of Diseases (ICD) codes for diagnosis. PCS is thought to affect multiple organs,4–6 and potentially initial severity of COVID-19 could be a risk factor for PCS.6–8 Our understanding of the magnitude of how much vaccines protect against PCS is evolving.9 10 Breakthrough infections following vaccination and reinfection with SARS-CoV-2 also associate with long-term organ sequelae.9 11

The symptoms attributed to PCS are still being characterised, and there could be several different clinical phenotypes; however, dyspnoea and fatigue seem to be central symptomatic components of PCS.5 6 12–15 Post-intensive care syndrome (PICS) refers to a set of physical, cognitive and psychological impairments that survivors of intensive care often experience, with issues like muscle weakness, memory loss and depression.16 This could be similar to issues found in PCS.

To understand how clinicians attribute symptoms and PCS diagnosis, we examined risk factors and associated clinical features in patients who received a PCS ICD code in a total population of COVID-19 cohort in Sweden. To avoid classifying symptomatology due to PICS, COVID-19 patients with PCS were compared with disease severity-matched COVID-19 patients with no PCS diagnosis.

Methods

Results

A total of 16 151 (1.5%) received a PCS diagnosis (online supplemental file 1). To understand the symptomatology how medical doctors use the novel ICD-10 codes for PCS, we examined the population who had received a diagnosis (first or recurrent) at least 3 months after the COVID-19 date. There were 6389 (0.61%) individuals with a healthcare visit where a PCS diagnostic code was registered from 1 044 820 COVID-19 patients 3 months post-COVID-19 (figure 1 and table 1). A majority of patients who were diagnosed with PCS had been hospitalised, or had needed intensive care, or non-invasive and mechanical ventilation (61%), compared with 4.4% in the nationwide COVID-19 cohort (table 1). However, 33% of individuals with PCS diagnosis had not had contact with outpatient or inpatient clinics during the acute phase of initial COVID-19. Internal medicine (24.7%) and infection medicine (14.2%) outpatient and inpatient clinics registered the majority of PCS diagnostic codes in Sweden (online supplemental file 1). Geographically, 39.9% of all PCS visits were administered at Stockholm county council (online supplemental file 1), a county that accounts for 23% of the Swedish population (Statistics Sweden, 2021). The distribution of comorbidities according to the Charlson comorbidity grouping for the COVID-19 cohort, PCS individuals and matched control individuals is shown in online supplemental file 1.

Request reuse permission

Factors associating with receiving a PCS diagnosis by physicians in the COVID-19 cohort

We estimated risk factors for PCS diagnosis first in the full cohort of COVID-19-positive individuals prior to matching for disease severity or other demographic factors. The strongest risk factor for PCS diagnosis was needing mechanical ventilation during the acute phase of COVID-19 (adjusted OR (model 5 (adjusted for age, sex, education and comorbidities) 114.7, 95% (CI) 105.1 to 125.3) (figure 2 and online supplemental file 1). In model 5 including all potential risk factors, there was an approximately 25% and 61% higher risk of being diagnosed with PCS for the age groups 60–79 and 80+ years, respectively, compared with the reference group (20–39 years). Men had a lower risk of receiving a PCS diagnosis compared with women (0.78 (0.74 to 0.82)). Having only secondary and primary education compared with tertiary education associated with a lower risk of PCS diagnosis, whereas income did not significantly associate with PCS diagnosis (figure 2). Higher wCCI associated with PCS diagnosis, and specifically having had peripheral vascular disease, cerebrovascular disease, chronic obstructive pulmonary disorder, rheumatic disease, diabetes with complications, renal disease, ascites and malignancy associated significantly with an increased risk of receiving a PCS diagnosis (online supplemental file 1).

Request reuse permission

Figure 2

Multivariable logistic regression of association between potential risk factors (COVID-19 severity, age, sex, comorbidities, education, income) during the acute phase and receiving a PCS diagnosis during visits to outpatient and inpatient clinics at least 3 months after COVID-19 onset. All COVID-19-positive individuals are included in this analysis. The multivariable logistic regression includes all variables in the figure (ie, model 5 from online supplemental file 1). AOR, adjusted OR, PCS, post-COVID syndrome; wCCI weighted Charlson Comorbidity Index.

Discussion

Currently, the pathophysiology and understanding of mechanisms of PCS are evolving including persistent infection, autoimmunity or dysfunction of blood clotting.19 20 24 The risk factors and clinical diagnostic features in COVID-19 patients with long-term sequelae that result in a PCS diagnosis have remained elusive, and most of the literature on PCS is based on follow-up studies of hospitalised patients25 26 and questionnaire data.7 10 12 27 28 A recent study was published using healthcare data covering 25% of the Israeli population29; however, no studies have been performed on a total country’s population including all individuals who have a laboratory-verified positive test for SARS-CoV-2 in the country regardless of initial COVID-19 severity. This approach allows us to attribute diagnosis of PCS and related symptomatology, and describe the characteristics of physicians who use the PCS diagnosis with regard to clinical specialty and geography.

In the total population of Sweden, we examined all laboratory-verified test-positive SARS-CoV-2 individuals and discovered that the initial COVID-19 severity was the strongest associating factor for physicians assigning a PCS diagnosis in an outpatient/inpatient setting. We found that Swedish physicians during the first and second waves of the pandemic primarily associated dyspnoea and fatigue among other symptoms with PCS. In addition, we identified several diagnoses that were present prior to COVID-19, and thereby not likely associated with clinical features of PCS.

It is noteworthy that the PCS code is novel and likely used differently across practices and geographical location. This is also supported by our results and in our study, we see the highest usage in the Stockholm metropolitan area. Furthermore, there likely is not yet a harmonised national standard across different locations on how to use ICD-10 codes for PCS. Our study provides insight into how these novel codes are used in Sweden and the symptomatology that prompts healthcare providers and medical doctors to assign the code. Yet, the novel ICD-10 codes likely do not capture diagnoses for all PCS patients.

The main strength of our study is the use of the entire population of laboratory-verified SARS-CoV-2 test-positive individuals in Sweden (approximately 1 million) during 1 February 2020–25 May 2021 with historical data from visits to outpatient/inpatient clinics from 1997. This approach allows population-based estimates and offers a clinical perspective, which is complementary to self-reported symptoms in questionnaire-based studies, with the added benefit of identifying the specific diagnoses that were new following COVID-19 onset. The Nordic countries are unique in having PINs for every individual in the population, with Sweden having a numerically larger population, providing the infrastructure that is needed for our study.30

A weakness is the lack of primary care data, which are not registered centrally in Sweden; and additionally, the access to outpatient/inpatient clinics and follow-up was not standardised. The PCS incidence is most likely conceivably larger than what is represented in the data, due to initial physician lack of awareness of PCS and associated novel ICD-10 codes for diagnosis. Therefore, we examined how clinicians attribute the code to patients who received the PCS diagnosis. This is made obvious in our study where the proportion of individuals diagnosed with PCS by a physician in an outpatient/inpatient setting during the early pandemic phases was 0.61%. This is considerably lower than reported in other studies ranging from 14.1%,7 12.5%,28 6.2%15 to 3.3% of the UK population according to the UK Office for National Statistics.31

The initial COVID-19 severity was a strong associating factor for physicians assigning a PCS diagnosis to patients in our study. This is corroborated in earlier reports.6–8 Another associating factor for physicians assigning a PCS diagnosis in our study was female sex, as found in other studies.15 25 26 31 32

Dyspnoea, malaise/fatigue/post-viral fatigue syndrome and abnormal pulmonary imaging/findings were some of the most prominent clinical features that physicians associated with PCS. These clinical features were also enhanced in the COVID-19 patients with PCS, even when compared with COVID-19 patients without a PCS diagnosis but with the same initial severity of disease. The majority of the PCS patients had not received these diagnoses prior to COVID-19 onset. Since we compare intensive care unit (ICU) patients with PCS with ICU patients without PCS, our findings cannot be attributed to PICS only.16 The PCS cohort has a characteristic profile compared with the remaining COVID-19 cohort. Although most of the literature on PCS is based on follow-up studies of hospitalised patients,25 26 our findings are in line with the diverse symptomatology previously reported by questionnaire data,7 12 27 28 and recently also with electronic health record data from Veteran’s affairs data.6 9 Furthermore, systematic reviews and meta-analyses identify fatigue and dyspnoea as common long-term symptoms.12–14

In our study, pulmonary embolism was present in 5% of the PCS cohort, which was a new diagnosis to the majority; and the OR was 15-fold higher when compared with disease severity-matched COVID-19-positive control individuals. In a previous study, we found that the risk of pulmonary embolism was highest for COVID-19 patients in need of intensive care, and still significantly increased in the time period of 3–6 months after COVID-19 indicating a potential association between pulmonary embolism and dyspnoea for this particular patient group.19 This indicates that the PCS cohort has a distinct symptomatic/disease profile as seen by the clinicians from the remaining COVID-19 cohort.

While the mechanisms of PCS are emerging, there are several potential likely disease mechanisms.24 For example, COVID-19 has been demonstrated to damage vasculature.33 Consequently, defects in blood supply and tissue oxygenation may contribute to more severe clinical picture particularly in patients where pulmonary damage through severe disease has occurred, and in patients who have dysfunction of peripheral vasculature. Furthermore, independent of disease severity, the dysfunction of the autonomic nervous system may be caused by autoimmunity, epitope spreading or direct persistent infection.34 35 Interestingly, in our study, 2% of individuals had SARS-CoV-2 detected in the same visit where they received a PCS diagnosis by a physician, raising the possibility of reinfection or persistent infection. Similarly, the clinical picture in PCS may relate to novel infections or opportunistic secondary infections such as those caused by Echerichia coli or other bacterial infections as we observe are enhanced in the PCS cohort in our study, or reactivation of latent pathogens including Epstein-Barr or varicella virus.24 34 35

In summary, we found that Swedish physicians associated the following with PCS: respiratory dysfunction, neurological and cognitive disorders, and cardiac and circulatory dysfunction. Initial COVID-19 severity was the strongest associating factor for receiving a PCS diagnosis by a physician. More studies are required to further characterise this emerging syndrome, and particularly in a primary care setting.