De diagnose van COVID-19 wordt bemoeilijkt door zowel geografische als temporele fluctuaties in de prevalentie van de ziekteachtergrond en het brede klinische spectrum van de ziekte. Zowel de achtergrondsprevalentie als de klinische presentatie bepalen de voorafkans (pre-test probability) op ziekte en daarmee de interpretatie van diagnostische modaliteiten en de achterafkans (post-test probability) van een COVID-19-diagnose. Tijdens perioden met een hoge COVID-19-achtergrondprevalentie zou de klinische verdenking hoger moeten zijn en neemt de waarschijnlijkheid van ziekte vóór de test (pre-test probability) toe.

COVID-19 heeft ook diverse klinische presentaties. Het klinische spectrum kan variëren van asymptomatische of presymptomatische infectie tot milde, matige, ernstige en kritieke ziekte. Om specifieke klinische symptomen als leidraad voor de diagnose te gebruiken, heeft het expertiseteam de voorspellende waarde van verschillende veelvoorkomende klinische symptomen bij de diagnose van COVID-19 beoordeeld. Deze omvatten verlies van smaak en geur, diarree, thoracale pijn en trombose. Veelvoorkomende symptomen passend bij COVID-19 zijn beschreven in Tabel 1.

Tabel 1. Symptomen passend bij COVID-19 *

* Gebaseerd op IDSA diagnostic guideline en de RIVM website.

Bij specifieke patiëntenpopulaties komen atypische presentaties van COVID-19 voor waarbij een ​​COVID-19-diagnose dient te worden overwogen. In deze module wordt zowel de voorspellende waarde van specifieke symptomen bij volwassenen als bij specifieke patientenpopulaties, waaronder ouderen en kinderen, beschreven/bekeken. Het is belangrijk om aandacht te hebben voor de meest voorkomende klachten bij ouderen gezien hun frequente opname en het verhoogde risico op ernstig beloop van ziekte (Williamson, 2020).

Description of studies

Atypical symptoms at presentation in adults

Five studies (one systematic review and four clinical studies) reported on the predictive value of loss of smell and taste and diarrhea. One of the studies, a Cochrane review (Struyf 2020), searched for studies assessing the diagnostic accuracy of signs and symptoms for COVID-19 among patients presenting to primary care or hospital outpatient settings. The search was performed on the 27^th of April 2020. The review included both cross-sectional and diagnostic case-control type studies. 16 Studies encompassing 7706 participants were included. In these studies, the prevalence of COVID-19 varied from 5% to 38%. Six studies were included that reported on the diagnostic accuracy of diarrhoea and one study reported on the diagnostic accuracy of loss of smell or loss of taste (see evidence table).

Brendish (2020) reported a clinical study in large UK hospital. The study population included adults (≥18 years old) presenting to the Emergency Department or Acute Medicine Unit with an acute respiratory illness or otherwise clinically suspected of having COVID-19. 1054 adults presenting to hospital with suspected COVID-19 were included, of whom 33.4% tested positive for COVID-19. A subset of 499 patients who received molecular

point-of-care testing (mPOCT) were asked detailed questions about their symptoms prior to the mPOCT result. The remaining patients were tested with RT-PCR. In this group only routinely collected data was used. The prevalence of COVID-19 was 39%.

Haehner (2020) reported the results of a cross-sectional study which included all patients presenting to a coronavirus testing center at a University Hospital in Germany. 500 patients were included and filled in a questionnaire which included questions on symptoms, time course, and an additional self-assessment of the patients’ current smell, taste function, and nasal breathing compared to the level before the onset of symptoms. The percentage of patients who tested positive for COVID-19 was 6.8%.

Payrony (2020) reported the results of an observational study including 391 adult patients with suspected COVID-19 who were tested for SARS-CoV-2 at a university hospital. Symptoms were recorded by attending emergency physicians or resident physicians in a dedicated form detailing the patient’s medical history. The prevalence COVID-19 in this study was 58%.

Çalıca Utku (2020) described 297 patients who adhered to the case definition and who were admitted to a COVID-19 inpatient clinic. Patients were asked if they had fever, cough, sputum production, dyspnea, sore throat, myalgia, lumbago, weakness, rhinorrhea, headache, taste disorder, smell disorder, diarrhea, conjunctivitis and skin rash. In this study, the prevalence of COVID-19 was 48%.

Atypical symptoms at presentation in elderly patients

No studies were included in which the predictive value of delirium, falls, decreased consciousness, hypothermia for COVID-19 in elderly patients was reported.

Atypical symptoms at presentation in children

No studies were included in which the predictive value of symptoms matching Kawasaki disease or Guillain Barre for COVID-19 in children was reported.

Results

Diarrhea

In the Cochrane review (Struyf, 2020), six studies (1733 patients) were included that reported on the sensitivity and specificity of diarrhea as a symptom predictive for COVID-19. In these included studies, the sensitivity ranged from 0.0 to 0.15. The specificity ranged from 0.86 to 0.99. The results of these studies were not pooled in a meta-analysis, due to the small numbers of studies currently available, considerable heterogeneity across studies and the high risk of bias. The positive predictive values were not reported. We calculated that these values ranged from 0.0 to 0.43. However, two of these studies were small (n=19/ n=29) and in four of the studies the target population was COVID-19 pneumonia instead of COVID-19 disease.

Brendish (2020) reported a sensitivity of 0.373 [95%CI 0.3 to 0.451] for the presence of diarrhea at presentation. The specificity was 0.824 [95%CI 0.771 to 0.867] and the positive predictive value 0.576 [95%CI 0.477 to 0.668]. Diarrhea was present in 99 patients at presentation (20%).

In the study by Çalıca Utku (2020) the prevalence of diarrhea was 16% (49 of the 297 included patients). The sensitivity of this symptom for COVID-19 was 0.217 [95%CI 0.157 to 0.291], the specificity 0.883 [95%CI 0.823 to 0.925] and the positive predictive value 0.633 [95%CI 0.493 to 0.753].

Payrony (2020) reported on the diagnostic value of gastrointestinal complaints. Gastrointestinal symptoms, including such as vomiting, diarrhea, or abdominal pain, were present in 94 of the 391 included patients (24%). The percentages were comparable in COVID-19 positive and COVID-19 negative patients, resulting in a sensitivity of 0.236 [95%CI 0.185 to 0.295], specificity of 0.753 [95%CI 0.682 to 0.812] and positive predictive value of 0.564 [95%CI 0.463 to 0.66].

Loss of smell or loss of taste

The Cochrane review included one study (870 participants) which reported on the sensitivity and specificity of loss of smell or loss of taste. In this study the sensitivity was 0.23 (95%CI 0.16 – 0.30) and the specificity 0.99 (95%CI 0.98 – 0.99). The positive predictive value is 0.795 [95%CI 0.655 to 0.888].

In the study by Brendish (2020), anosmia was present in 66 patients in a subgroup of 499 patients who were tested with mPOCT only (13%). The sensitivity of the presence of anosmia for COVID-19 was 0.331 [95%CI 0.259 to 0.412], the specificity 0.912 [95%CI 0.867 to 0.943] and the positive predictive value 0.712 [95%CI 0.594 to 0.807].

In the study by Haehner (2020), 69 patients (14%) presented with olfactory loss. For the sole symptom “smell loss”, the sensitivity was 0.647 [95%CI 0.479 to 0.785], the specificity 0.899 [95%CI 0.868 to 0.923] and the positive predictive value 0.319 [95%CI 0.221 to 0.436]. Excluding patients with a blocked nose increased the positive predictive value to 63% (specificity 97% sensitivity of 65%).

Thirty-four (8.7%) patients included in the study by Payrony (2020) reported anosmia. The sensitivity of anosmia for COVID-19 was 0.14 (95%CI 0.10–0.19), the specificity 0.98 (95%CI 0.95–1.00) and the positive predictive value 0.91 (95%CI 0.76–0.98).

Çalıca Utku (2020) reported that a smell disorder was present in 50 patients (17%). The sensitivity of a smell disorder for the presence of COVID-19 was 0.231 [95%CI 0.169 to 0.306]. The specificity was 0.89 [95%CI 0.83 to 0.93] and the positive predictive value 0.66 [95%CI 0.522 to 0.776]. A taste disorder was present in 72 patients (24%). The sensitivity of this symptom was 0.357 [95%CI 0.283 to 0.438]. The positive predictive value 0.708 [95%CI 0.595 to 0.801]

Level of evidence of the literature

The level of evidence regarding the predictive value of symptoms was downgraded by 3 levels because of inconsistency (PPV range from no predictive value to moderate) and study limitations (risk of misclassification of symptoms and disease; unclear whether symptoms were measured independent from PCR test results).

A systematic review of the literature was performed to answer the following questions:

What is the predictive value of atypical symptoms for COVID-19 in adults, elderly, and children who presented at the hospital? More specifically,

• What is the predictive value of loss of smell, loss of taste, diarrhea, thoracic pain, or thrombosis for COVID-19 in adults who presented at the hospital?
• What is the predictive value of delirium, falls, decreased consciousness, hypothermia for COVID-19 in elderly who presented at the hospital?
• What is the predictive value of symptoms matching Kawasaki disease or Guillain Barre for COVID-19 in children who presented at the hospital?

P1: adults who present themselves at the hospital with loss of smell, loss of taste, diarrhea, thoracic pain, or thrombosis

I: -

C: -

O: (positive) predictive value for COVID-19

P2: elderly patients who present themselves at the hospital with delirium, falls, decreased consciousness, hypothermia

I: -

C: -

O: (positive) predictive value for COVID-19

P3: Children who present themselves at the hospital with symptoms matching Kawasaki disease or Guillain Barre

I: -

C: -

O: (positive) predictive value for COVID-19

Relevant outcome measures

The guideline development group considered the positive predictive value as a critical outcome measure for decision making.

Search and select (Methods)

The databases Medline (via OVID) and Embase (via Embase.com) were searched with relevant search terms until 14 October 2020. Three separate searches were performed: for adults, elderly patients, and children. The detailed search strategy is depicted under the tab Methods. The systematic literature searches resulted in 3342 hits (adults), 126 hits (elderly) and 110 hits (children). Studies were selected based on the following criteria clinical studies and/ or systematic reviews with a meta-analysis, reporting presenting data about the predictive value (diagnostic accuracy) of selected atypical symptoms for COVID-19 or data that can be used to calculate the predictive value.

Concerning adults, sixteen studies were initially selected based on title and abstract screening. After reading the full text, eleven studies were excluded (see the table with reasons for exclusion under the tab Methods), and five studies were included.

For the question concerning elderly patients, nine studies were initially selected based on title and abstract screening. After reading the full text, all studies were excluded (see the table with reasons for exclusion under the tab Methods), and no studies were included.

Concerning children, nine studies were initially selected based on title and abstract screening. After reading the full text, all studies were excluded (see the table with reasons for exclusion under the tab Methods), and no studies were included.

Results

Five studies were included in the analysis of the literature. Important study characteristics and results are summarized in the evidence tables. The assessment of the risk of bias is summarized in the risk of bias tables.