key: cord-0824690-o70sahmk
authors: Fernández-de-las-Peñas, César; Palacios-Ceña, Domingo; Gómez-Mayordomo, Víctor; Florencio, Lidiane L; Cuadrado, María L.; Plaza-Manzano, Gustavo; Navarro-Santana, Marcos
title: Prevalence of Post-COVID-19 Symptoms in Hospitalized and Non-Hospitalized COVID-19 Survivors: A Systematic Review and Meta-Analysis
date: 2021-06-16
journal: Eur J Intern Med
DOI: 10.1016/j.ejim.2021.06.009
sha: 0d4c48e0ea89480cc9214d44d90a2a0b6d0de61c
doc_id: 824690
cord_uid: o70sahmk

BACKGROUND: : Single studies support the presence of several post-COVID-19 symptoms; however, no meta-analysis differentiating hospitalized and non-hospitalized patients has been published to date. This meta-analysis analyzes the prevalence of post-COVID-19 symptoms in hospitalized and non-hospitalized patients recovered from COVID-19 . METHODS: : MEDLINE, CINAHL, PubMed, EMBASE, and Web of Science databases, as well as medRxiv and bioRxiv preprint servers were searched up to March 15, 2021. Peer-reviewed studies or preprints reporting data on post-COVID-19 symptoms collected by personal, telephonic or electronic interview were included. Methodological quality of the studies was assessed using the Newcastle-Ottawa Scale. We used a random-effects models for meta-analytical pooled prevalence of each post-COVID-19 symptom, and I² statistics for heterogeneity. Data synthesis was categorized at 30days, 60days, and ≥90 days after . RESULTS: : From 15,577 studies identified, 29 peer-reviewed studies and 4 preprints met inclusion criteria. The sample included 15,244 hospitalized and 9,011 non-hospitalized patients. The methodological quality of most studies was fair. The results showed that 63.2%, 71.9% and 45.9% of the sample exhibited ≥one post-COVID-19 symptom at 30, 60, or ≥90days after onset/hospitalization. Fatigue and dyspnea were the most prevalent symptoms with a pooled prevalence ranging from 35% to 60% depending on the follow-up. Other post-COVID-19 symptoms included cough (20-25%), anosmia (10-20%), ageusia (15-20%) or joint pain (15-20%). Time trend analysis revealed a decreased prevalence 30days after with an increase after 60days . CONCLUSION: : This meta-analysis shows that post-COVID-19 symptoms are present in more than 60% of patients infected by SARS-CoV‑2. Fatigue and dyspnea were the most prevalent post-COVID-19 symptoms, particularly 60 and ≥90 days after.

4 heterogeneity was expected. An I 2 value ≥75% was considered to indicate serious heterogeneity. We were not able to assess funnel plot asymmetry due to an insufficient number of studies investigating the same post-COVID-19 symptom at a particular follow-up. We calculated sample size-weighted mean scores for each study reporting data alongside 95% confidence intervals (95%CI) in addition to any potential meta-analytical summary effect on the pooled prevalence data for each post-COVID-19 symptom. Data synthesis was categorized by time after onset/hospitalization into three follow-up periods (symptoms at 30 days, 60 days, and ≥90 days). To determine the time-course of post-COVID-19 symptoms over time (from onset to ≥90 days after), Freeman-Tukey double arcsine transformation was conducted using the escalc function in the metafor package. The rma.mv (meta-analytic multilevel random effect model with moderators via linear mixed-effect models) was used to carry out a multilevel metanalysis with three levels to identify time and time*subgroup effect. For meta-analyses of studies reporting outcomes at multiple time points, it may be reasonable to assume that the true effects are correlated over time according to an autoregressive structure; therefore, a heteroscedastic autoregressive (HAR) model was adopted. Grouping by gender was not possible due to lack of data (see discussion section)

For quantitative data (age, days at hospital), overall means and standard deviations (SD) were calculated using the pool.groups function from the dmetar package. Median and interquartile range (IQR) were converted to mean and SD as described by Luo et al. [10] . When necessary, data were estimated from graphs with the GetData Graph Digitizer v.2.26.0.20 software.

There was no funding source for this study.

Patients were not involved in the study since this was a meta-analysis of the literature.

The selection process is shown in Figure 1 . The electronic search identified 15,577 potential titles. After removing duplicates and papers not directly related to post-COVID-19 symptoms, 64 studies remained. Twenty-six (n=26) were excluded after title/abstract examination. One preprint was excluded because it analyzed risk factors and clusters but not detailed specific post-COVID-19 symptoms [11] ; one study was excluded because it was a case series [12] ; another one because mortality rate, not post-COVID-19 symptoms, was analyzed [13] ; and the last one because it included children, not adults, with .

A total of 29 published studies and five medRxiv preprints [44] [45] [46] [47] [48] were initially included in the review/metaanalysis ( Figure 1 ). One preprint [44] was excluded because the same study has been posteriorly published in a peer-reviewed journal [30] . Therefore, a total of 29 peer-reviewed studies and four medRxiv preprints [45] [46] [47] [48] were included in the systematic review and meta-analysis.

The characteristics of the COVID-19 populations of the included studies are shown in Table 1 . The total sample comprised 24,255 COVID-19 survivors (52.26% female; mean ± SD age: 47.8±16.6 years); 15,244 were hospitalized (42.7% female; age: 48.6± 17.4) whereas 9,011 (70.2% female; age: 44.3±14.8) were non-hospitalized patients. The mean length of hospital stay due to SARS-CoV-2 infection was 12.5 days (SD 6.8) . From those hospitalized, 402 patients (8%) required ICU admission (mean stay: 15±14.6 days).

Almost 50% of the total sample exhibited at least one pre-existing comorbidity (one: 26.3%, 95%CI 25.3-28.0%; two: 17.6%, 95%CI 15.1-20.5%; ≥ three: 25.6%, 95%CI 11.4 -47.8%) with hypertension (22.9%, 95%CI 16.2-31.5%) and obesity (22.2%, 95%CI 13.9 -33.5%) being the most prevalent. Pre-existing comorbidities were, in general, more prevalent in hospitalized patients than in non-hospitalized patients. Table 2 summarizes the pooled prevalence of demographic and clinical data of COVID-19 survivors separated by hospitalization. Hospitalization data were collected from medical records in all studies.

Thirty studies (88%) were cross-sectional, just one was of good quality (3/3 stars) , 28 were considered of fair quality (2/3 stars), and two of poor quality (1/3 stars). One was a longitudinal cohort study with high methodological quality (8/9 stars) , and two were case-control studies of poor quality (5/9 stars, with 0 stars in the comparability domain). No disagreement between authors was observed. Table 3 presents the Newcastle-Ottawa Scale scores for each study and a summary of every item.

Supplementary Table summarizes which study assessed each COVID-19 onset symptom and each post-COVID-19 symptom. Sixteen studies (48.5%) collected the post-COVID-19 data by telephonic interviews, whereas ten studies (30%) collected data face-to-face interviews.

Pooled data of symptoms at onset and post-COVID-19 symptoms experienced by the total sample, including both hospitalized and non-hospitalized COVID-19 patients, are shown in Table 4 . In the total sample, the most common symptoms experienced at SARS-CoV-2 infection were fatigue (63.4%), cough (60.2%), fever (55.3%), ageusia (46.0%), anosmia (45.7%) and dyspnea (44.1%). Among hospitalized patients, the most common onset symptoms at hospital admission included cough (65.2%), fever (59.45%), fatigue (48.0%), dyspnea (50.9%), anosmia (34.3%) and ageusia (34.0%). In non-hospitalized patients, the most common onset symptoms were fatigue (71.89%), myalgia (59%), cough (56%), fever (52.5%), anosmia (51.9%), and ageusia (51.8%). Most pooled data showed high level of heterogeneity (I 2 ≥75% Overall, thirty days after onset/hospital admission (mean: 30.3±6.3 days), the most frequent post-COVID-19 symptoms were cough (18.6%), anosmia (16.5%), ageusia (15.7%), dyspnea (13.2%), fatigue (11.7%) and confusion (8%), without significant differences between the hospitalized and non-hospitalized patients (Table 4) .

Overall, sixty days after onset or hospitalization (mean: 60.4±6.6 days), the most frequent post-COVID-19 symptoms were fatigue (56.2%), dyspnea (27.2%), chest pain (23.6%), headache (19.8%), joint pain (19%), and cough (18.9%). Non-hospitalized individuals showed higher prevalence of sore throat (67%), headache (48%) and anosmia (37%) than hospitalized patients (4%, 6 11%, and 11.5%, respectively), but the differences did not reach statistical significance due to the heterogeneity in the comparison (Table 4) .

More than ninety days after onset/hospitalization (mean: 118.4±40.0 days), the most frequent post-COVID-19 symptoms included fatigue (35.3%), dyspnea (26.3%), anosmia (11%), myalgia (10.9%), joint pain (10.3%), and ageusia (10%). At this followup period, non-hospitalized patients reported significantly higher prevalence of anosmia (15.5% vs. 8.1%, P=0.012), chest pain (14.9% vs. 7.7%; P=0.02), sputum (10.7 vs. 3.4, P=0.002), and vertigo (12.7% vs. 4.2%, P=0.02) than hospitalized patients (Table 4) .

Of the twenty-one studies [15, 16, 18, [22] [23] [24] [25] 27, 29, [32] [33] [34] [35] [36] [37] [40] [41] [42] [43] 46, 47] investigating the presence of post-COVID-19 symptoms in hospitalized patients, four analyzed symptoms 30 days after hospital discharge [15, 33, 41, 43] , nine showed a followup period of 60 days [15, 18, 24, 27, 29, 36, 37, 42, 47] , whereas ten reported symptoms ≥90 days after discharge [16, 22, 23, 25, 26, [33] [34] [35] 40, 46] . Overall, hospitalized COVID-19 patients were assessed a mean of 83.6±48.4 after hospital discharge. Among twelve studies [17, [19] [20] [21] 26, 28, 30, 31, 38, 39, 45, 48] with non-hospitalized patients, four studies evaluated post-COVID-19 symptoms 30 days after onset [19, 31, 38, 45] two had a follow-up of 60 days [30, 45] , whereas seven analyzed symptoms after ≥90 days [17, 20, 21, 26, 28, 45, 48] . The sample of non-hospitalized patients was assessed a mean of 73.9±46.4 days after onset of symptoms.

Within hospitalized patients, the most common post-COVID-19 symptoms included: cough (26.6%), skin rashes (14%), ageusia (11.4%), anosmia (11.1%), confusion (9.3%) and dyspnea (9.2%) 30 days after hospitalization; fatigue (53.9%), dyspnea (24.4%), joint pain (22.8%), chest pain (21.0%), cough (13.8%), and anosmia (11.5%) 60 days after hospitalization; and fatigue (38.5%), dyspnea (33.3%), cough (10.4%), myalgia (9.7%), joint pain (9.4%) and palpitations (9.1%) ≥90 days after hospitalization (Fig. 2 ).

Within non-hospitalized patients, the most common post-COVID-19 symptoms were anosmia (19.9%), ageusia (18.3%), dyspnea (15.7%), cough (13.9%), fatigue (11.8%), and headache (10.9%) 30 days after the onset of symptoms; sore throat (67.0%), fatigue (63.2%), headache (48.2%), cough (40.7%), dyspnea (39.9%), and anosmia (37.7%) 60 days after symptom onset; and fatigue (29.8%), dyspnea (19.1%), anosmia (15.5%), chest pain (14.9%), and ageusia (13.2%) ≥90 days after (Fig. 2) . Figure 2 graphs the time-course of the eight most prevalent symptoms from onset/ hospitalization to 30, 60 and ≥90 days after in hospitalized and non-hospitalized patients. The random effect model showed significant effect for time (all, P<0.001) for fatigue, dyspnea, headache, myalgias, cough, anosmia and ageusia symptoms, but not for chest pain: symptoms dropped at 30 days relative to baseline and raised up again at 60 and ≥90 days after. Significant group*time effects were also found showing that this tendency was more pronounced in hospitalized than non-hospitalized patients.

This systematic review/meta-analysis revealed that more than 60% of COVID-19 survivors exhibit at least one post-COVID-19 symptom for more than 30days after onset or hospitalization. The prevalence of each symptom in isolation was 10-15% at 30 days and 40-60% at 60 days or longer after onset/hospitalization (Fig. 2) . Fatigue and dyspnea were the most prevalent post-COVID-19 symptoms in hospitalized and non-hospitalized patients, particularly at 60 and ≥90 days of follow-up, whereas the prevalence of other symptoms, e.g., headache, anosmia, ageusia, chest pain, or palpitations, was lower and highly variable.

The preprint meta-analysis by Lopez-Leon et al observed that fatigue, headache, attention disorder, hair loss or dyspnea were the most frequent post-COVID-19 symptoms [6] . They reported overall prevalence of post-COVID-19 symptoms without distinction between hospitalized/non-hospitalized patients or considering the follow-up period [6] ; therefore, the comparison between prevalence rates is not feasible. Another systematic review have reported that main post-COVID-19 sequelae were postinfectious fatigue, persistent reduced lung function and carditis; however, this review did not pooled data on post-COVID symptoms since it focused on functional impairments [49] . Another meta-analysis reported that the most common respiratory post-COVID-19 symptoms reported by hospitalized COVID-19 survivors included fatigue, dyspnoea, chest pain, and cough showing prevalence rates of 52%, 37%, 16% and 14%, respectively between 3 weeks and 3 months after hospital discharge [50] . These prevalence data are similar to our pooled data observed at 60days follow-up; however, Cares-Marambio et al [50] pooled studies without distinction on follow-up periods. Our systematic review/meta-analysis examined the prevalence of post-COVID-19 symptoms considering if patients were hospitalized or not and also separated by follow-up periods. We were able to identify 29 peer-reviewed studies as well as four medRxiv preprints providing prevalence data on post-COVID-19 symptoms from both hospitalized and non-hospitalized COVID-19 survivors at different follow-up periods; the highest number of studies pooled to date; however, most studies were of fair methodological quality and also showed high heterogeneity in their results. Nevertheless, it should be remarked that more and more studies assessing post-COVID-19 symptoms will be published and future updated meta-analyses will be needed.

The most common symptoms experienced by patients at onset/hospitalization in the overall sample were fatigue, cough, fever, ageusia, anosmia and dyspnea in agreement with a previous meta-analysis showing similar symptoms at SARS-CoV-2 infection [51] . Nevertheless, some differences in prevalence rates can be found. Compared to the current meta-analysis, Alimohamadi et al found similar prevalence of cough (58.5%), but higher prevalence of fever (81.2%) and lower rate of fatigue (38.5%) [51] . There is clear evidence supporting that clinical manifestations of COVID-19 are highly heterogeneous.

A relevant finding was that post-COVID-19 symptoms experienced 30days after onset/hospitalization decreased dramatically in prevalence as compared to the acute phase but increased 60days after (Fig. 2) . The reasons of these findings are still unknown and need to be confirmed in well-designed longitudinal studies; however, it should be noted that most prevalence data were based on a small number of studies and comparisons had large heterogeneity. In fact, studies conducted in Europe reported higher prevalence rates of fatigue (50-70%) or dyspnea (30-40%) as post-COVID-19 symptoms [15] [16] [17] [18] 20, 27, 37, [40] [41] [42] whereas Chinese studies reported, in general, lower prevalence rates of these symptoms (12-20%) [22, 23, 32, 43] . Factors such as younger age and lower pre-existing medical comorbidities in Chinese studies could explain these discrepancies; however, the magnitude of these different prevalence rates would suggest other relevant factors e.g., racial disparities [52] or blood type [53] . Future studies investigating the epidemiology of post-COVID-19 symptoms attending to these factors are needed.

The occurrence of respiratory symptoms following SARS-CoV-2 infection is similar to that present in severe acute respiratory syndrome (SARS) survivors, who also exhibit symptoms 6-12 months after the infection [54], but contrasts with that observed after community-acquired bacterial pneumonia where almost all patients are asymptomatic 10 days after the infection [55] . In addition, a main difference between SARS-CoV-2 and other respiratory infectious diseases is the presence of a plethora of post-infectious symptoms, e.g., joint pain, ageusia, anosmia, chest pain, nausea, headaches or palpitation, affecting systems other than the respiratory system. This meta-analysis confirms the presence of several post-COVID-19 symptoms supporting a multisystemic involvement; it also shows that time-course of symptoms fluctuates depending on the follow-up period and whether the COVID-19 patient was hospitalized or not. These considerations are highly important to properly define the timeframe of post-COVID-19 symptoms [7] .

To determine the underlying mechanisms behind these symptoms is beyond the scope of the current review, but two main hypotheses are currently discussed, although not alone. First, a prolonged pro-inflammatory response (hyper-inflammatory cytokine storm) related to SARS-CoV-2 infection can provoke an atypical response of the immune system and mast cells, promoting a cascade of events affecting the respiratory, immune, and central nervous systems [56] . Second, social and emotional factors around COVID-19 pandemic, e.g., posttraumatic stress, hospitalization, treatments received, catastrophic social alarm, lockdown, laboral and familiar situations, and psychological disorders, such as anxiety or depression, may contribute to these post-COVID-19 symptoms.

Although the underlying mechanisms explaining this plethora of symptoms are unknown, their complexity and heterogeneity supports that post-COVID-19 sequalae will need from a multidisciplinary approach [57]. The results of this review and meta-analysis summarizing prevalence rates of post-COVID-19 symptoms should be considered according to its strengths and weaknesses. The main strength was the rigorous methodology applied for literature search, study selection, screening for eligibility, assessment of methodological quality, and pooling analysis of prevalence data from more than 30 studies. Nevertheless, some weaknesses should be also recognized. First, a meta-regression could not be conducted because of the presence of serious/large heterogeneity between the studies. In fact, most of comparisons showed large heterogeneity. Second, the small number of studies in some comparisons limit the generality of the current results. Similarly, the number of patients requiring ICU admission was small, so no conclusions regarding this population can be achieved. Third, just two studies reported prevalence data separately by gender [22, 25] ; however, they reported different follow-ups and different post-COVID-19 symptoms; therefore, gender differences were not possible to be analyzed. Fourth, most studies included Caucasian subjects, with just four including Chinese people and none including African people; therefore, racial influence on the presence of post-COVID-19 symptoms remains unknown. Finally, post-COVID-19 symptoms were mostly self-reported by the patients themselves and collected by telephonic interview, electronical websites, postal or face-to-face interviews (table 1). Development of specific patient-reported outcome measures (PROM) for COVID-19 will be helpful to obtain homogeneous data. Interestingly, Tran et al have recently developed the long COVID Symptom and Impact Tools, which could help for more standardized collection of post-COVID-19 symptoms [58] .

This systematic review and meta-analysis investigating prevalence rates of post-COVID-19 symptoms provides updated data on the presence of persistent post-COVID-19 symptoms in COVID-19 survivors; however, it opens several questions for future studies. First, due to the relapsing and remitting nature of post-COVID-19 symptoms, it is important to identify those time frames where these symptoms should be considered as residual (post-acute COVID) or as real (long-term) post-COVID-19 symptom. In fact, time frames are important for proper description of post-COVID-19 symptomatology [7] . For instance, symptoms appearing soon (i.e., the first 30 days after symptoms onset) after recovery from acute infection have been considered as post-acute sequelae of COVID-19 (PASC), whereas symptoms appearing later, i.e., 3 months or longer, after infection could be considered as the real post-COVID-19 syndrome [7] . Second, identification of risk factors associated with post-COVID-19 symptoms is crucial. Some studies included in this review identified, by using multivariate analyses, potential risk factors, such as older age [15, 17, 38] , female gender [22, 23, 25, 41, 46] , longer hospital stance [15] , pre-existing comorbidities [17] , or number of symptoms at the acute stage [15, 17] associated with a higher number of post-COVID-19 symptoms. However, contradictory findings were also observed. For instance, whereas some studies reported that females were more prone to exhibit post-COVID-19 symptoms when compared with males [22, 23, 25, 41, 46] , others did not find such association with female gender [21, 24, 26, 30, 45, 47] . The heterogeneity in the methodology between the studies could explain these discrepancies in the results and does not permit to determine firm conclusions. Studies investigating risk factors associated with post-COVID-19 symptoms are urgently needed to promote focus on this issue in healthcare systems and, thereby, facilitate counselling and management strategies for these patients. A relevant topic for considering in future studies would be a potential participation of the patients into the designs since COVID-19 patients are highly active and their point of view may be crucial for designing studies according to their needs [59] . Studies investigating underlying mechanisms explaining post-COVID-19 symptoms are needed for better management of this group of individuals, the long-haulers [4] .

This review/meta-analysis has revealed that more than 60% of individuals infected by SARS-CoV-2 exhibited at least one post-COVID-19 symptom after onset or hospital admission. Fatigue and dyspnea were the most prevalent post-COVID-19 symptoms experienced by both hospitalized and non-hospitalized patients, particularly 60 and ≥90 days after onset/ hospitalization. The prevalence rate of other post-COVID-19 symptoms including headache, anosmia, ageusia, chest pain, joint pain or palpitations was lower and more variable. Early identification of post-COVID-19 symptoms will ensure immediate action and counselling of these "long haulers", who may otherwise struggle with unrecognized and unmanaged symptoms.

All authors contributed to the study concept and design. CFdlP, DMP, GPM and MNS conducted literature review. MNS, MLC and LLF did the statistical analysis. All authors contributed to interpretation of data. CFdlP, VGM, MLC and LLF contributed to drafting the paper. All authors revised the text for intellectual content and have read and approved the final version of the manuscript.

No funds were received for this study

This study will not share any individual data or document from any participant.

The lead author affirms that the manuscript is an honest, accurate, and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as originally planned have been explained

No conflict of interest is declared by any of the authors <bib id="bib3" type="Periodical"><number> [3] </number>Marshall M. The lasting misery of coronavirus long-haulers. Nature 2020:339-41.</bib> <bib id="bib4" type="Periodical"><number>[4]</number>Rubin R. As their numbers grow, COVID-19 "long haulers" stump experts. JAMA 2020;324:1381-3.</bib> <bib id="bib5" type="Periodical"><number> [5] </number>Nabavi N. Long covid: How to define it and how to manage it. BMJ 2020; 370: m3489.</bib> <bib id="bib6" type="Periodical"><number> Carvalho et al. 2020[15] 

A novel coronavirus from patients with pneumonia in China

Eyes irritation Carfi et al

Sore Throat Carfi et al

 Venturelli et al. 2021[33]