Skip to main content

Table 3 Description of the different diagnostic models for COVID-19 in the systematic review of prediction models to diagnose COVID-19 from the start of the epidemic to March 2021

From: A systematic review of prediction models to diagnose COVID-19 in adults admitted to healthcare centers

Reference

Population

Gold standard

Number of candidate predictors

Type of predictors

Sample size

Modeling method

Final model

Aldobyany [12]

All patients screened for COVID-19 at a tertiary care facility

RT-PCR test

11

Exposure risk

Clinical signs Symptoms Clinical history

1435

Score

Logistic regression

Exposure risk: A history of travel abroad during the 14 days

prior to symptoms onset.

OR

Visiting or being a resident of a high-risk area

for COVID-19 in the kingdom during the 14

days prior to symptom onset.

OR

A close physical contact with a confirmed case

of COVID-19 or MERS-COV in the past

14 days.

OR

An exposure to camel or camel products in the

past 14 days.

OR

Working in a healthcare facility

= 3 points

Fever or recent history of fever = 2 points

Cough = 2 points

Shortness of breath = 2 points

Nausea, vomiting, and/or diarrhea = 1 point

Chronic renal failure, CAD/heart failure,

Immunocompromised patient = 1 point

Maximum = 11 points

Discriminative score: 4 points or 5 points

Multivariate analysis (not used as a model)

Sex (male): OR = 1.47 (1.03–2.09), p = 0.034

Healthcare workers and their family members: OR = 1.99 (1.14–3.50), p = 0.016

Fever: OR: 2.98 (1.97–4.5), p < 0.001

Moderate disease severity: OR = 5 (1.23–20), p < 0.001

Bar [13]

Adults admitted to emergency room

RT-PCR test

21

Clinical data

Imaging

Laboratory test

100

Logistic regression

Logistic regression:

Intercept: β = − 1.95

qSOFA score > 1: β = 0.05; OR = 1.05 (1.01–1.10)

Upper sites B-lines ≥3: β = 0.42; OR = 1.52 (1.31–1.79)

Lower sites thickened pleura: β = 0.55; OR = 1.73 (1.49–1.98)

Lower sites consolidation: β = 0.87; OR = 2.39 (2.07–2.69)

Posterolateral sites thickened pleura: β = 0.68; OR = 1.97 (1.72–2.22)

Callejon-Leblic [17]

Individuals aged 18 or older that were suspected of having COVID-19

RT-PCR test

11

Demographics

Clinical data Symptoms

777

Logistic regression

Machine learning

x = − 1.76 + 0.88 × ((1 if VAS for loss of smell ≥21) or (0 if VAS for loss of smell < 21)) + 1.83 × ((1 if VAS for loss of taste ≥44) or (0 if VAS for loss of taste < 44)) + 0.79 × ((1 if VAS for dyspnea ≥28) or (0 if VAS for dyspnea < 28)) + 0.61 × ((1 if fever) or (0 if no fever)) + 0.70 × ((1 if diarrhea) or (0 if no diarrhea)) − 1.13 × ((1 if female) or (0 if male))

Loss of smell: OR (95%CI) = 2.42 (1.3–4.5), p = 0.0053

Loss of taste: OR (95%CI) = 6.21 (3.21–12.04), p < 0.001

Dyspnea: OR (95%CI) = 2.21 (1.34–3.64), p = 0.002

Fever: OR (95%CI) = 1.84 (1.18–2.87), p = 0.007

Diarrhea: OR (95%CI) = 2.02 (1.29–3.16), p = 0.002

Sex (Male): OR (95%CI) = 3.11 (1.97–4.90), p < 0.001

Fink [18]

Patients’ age > 18 years admitted to hospital for minimum 24 h

RT-PCR test

9

Demographics

Symptoms

Laboratory test

581

Logistic regression

Final logistic regression:

Age: β = 0.018; SE = 0.008, z = 2.18, P > z = 0.029, 95%CI = 0.001–0.033

Fever: β = 1.547; SE = 0.2931, z = 5.28, P > z = 0.000, 95%CI = 0.973–2.122

Maximal FiO2: β = 0.022; SE = 0.007, z = 2.94, P > z = 0.003, 95%CI = 0.007–0.0372

CRP: β = 0.0045; SE = 0.0016, z = 2.174 P > z = 0.006, 95%CI = 0.001–0.0078

Normal chest X-ray: β = − 1.113497; SE = 0.2879424, z = − 3.87, P > z = 0.000, 95%CI = − 1.677854- -0.5491403

Neutrophiles: β = − 0.1611107; SE = 0.0321778, z = − 5.01, P > z = 0.000, 95%CI = − 0.2241779- -0.0980434

Intercept: β = − 0.0367507; SE = 0.6817362, z = − 0.05, P > z = 0.957, 95%CI = − 1.372929-1.299428

Gupta-Wright [19]

Patients were included in this study if they were admitted via the acute medical team between 2 March and 3 May 2020 inclusive.

RT-PCR test

15

Demographics

Comorbidities

Symptoms

Laboratory test

2940

Logistic regression

Score

Intercept = −4.0 (− 4.4 - -3.6)

Age (50–70): β = 0.53 (0–0.41); Point = 1

Sex (Male): β = 0.23 (0.3–0.73); Point = 1

Ethnicity - Asian: 0.6 (0.4–0.82); Point =1

Ethnicity - Black: 0.62 (0.3–0.93); Point = 1

Ethnicity - Mixed/other: 0.81 (0.4–1.2); Point = 1

Ethnicity - Unknown: 0.57 (0.3–0.85); Point = 1

Cough, fever or shortness of breath: 1.13 (0.9–1.35); Point = 1

National Early Warning Score 2 (NEWS) > 5: 0.87 (0.7–1.05); Point = 2

CRP > 50: 1.13 (1–1.32); Point = 2

Lymphocytes < 1: 0.54 (0.4–0.73); Point = 1

Chest X-ray - Lung infiltrates: 1.32 (1.1–1.59); Point = 2

Chest X-ray - Other abnormality: 0.66 (0.3–0.98); Point = 2

Huang [20]

All suspected patients at admission

RT-PCR test

71

Demographics

Comorbidities

Vital signs

Symptoms

CT imaging

Laboratory test

475

Logistic regression

Score

Logistic regression:

Epidemiological exposure histories (OR:13.32, 95%CI, 6.39–27.75), symptoms

of weakness/fatigue at admission (OR:4.51, 95%CI, 1.70–11.96),

heart rate less than 100 beat/min at admission (OR:3.80; 95%CI,

2.00–7.22), imaging characteristics of bilateral pneumonia (OR:3.60,

95%CI, 1.83–7.10), neutrophil count less than equal to 6.3 × 109 /L at

admission (OR: 6.77, 95%CI, 2.52–18.19), eosinophil count less than equal to 0.02 × 109/L at admission (OR:3.14, 95%CI, 1.58–6.22), glucose more than equal to 6 mmol/L at admission (OR:2.43, 95%CI, 1.04–5.66), D-dimer more than equal to 0.5 mg/L at admission (OR:3.49; 95%CI, 1.22–9.96), and CRP less than 5 mg/L at admission (OR:3.83, 95%CI, 1.86–7.92)

Score:

Epidemiological exposure histories: 13 points

Neutrophil count, × 109/L, ≤6.3: 7 points

Weakness/fatigue: 5 points

Bilateral pneumonia: 4 points

Heart rate (beat/min), < 100: 4 points

CRP, mg/L, < 5: 4 points

Eosinophil count, × 109/L, ≤0.02: 3 points

D-dimer, mg/L, ≥0.5: 3 points

Glucose, mmol/L, ≥6: 2 points

Kurstjens [21]

Overcrowding of emergency departments

RT-PCR test

8

Demographics

Laboratory test

CT imaging

Model population: 375

Validation population: 592

Score

Age: ≤75: 0 points, 76–79: 1 point, 80+: 2 points

Sex: Female: 0 point, Male: 1 point

CRP, mg/L: 0–9: 0 point, 10–14: 1 point, 15–38: 2 points, 39–69: 3 points, 70–193: 2 points, 194–303: 1 point, 304+: 0 point

Ferritin, μg/L: ≤15: − 1 point, 16–179: 0 point, 180–301: 1 point, 302–538: 2 points, ≥539: 3 points

LDH, U/L: ≤257: 0 point, 258–265: 1 point, 266–397: 2 points, ≥398: 3 points

ALC, 10^9/L: ≤1.2: 1 point, ≥1.3: 0 point

ANC, 10^9/L: ≤5.1: 0 point, 5.2–7.9: − 1 point, 8.0–9.0: − 2 points, 9.1–10.3: − 3 points

Chest X-ray: no infiltrate: 0 point, unilateral infiltrate: 1 point, bilateral infiltrate: 4 points

McDonald [22]

All patients greater than 18 years presenting to a single academic ED who were tested for COVID-19 during this index ED evaluation

RT-PCR test

44

Clinical data

Demographics

Symptoms

Contact tracing

Laboratory test

1026

Logistic regression

Random forest

XGBoost

Final logistic regression:

Intercept: β = − 36

WBC: β = − 0.59, OR = 0.6 (0.4–0.8)

Temperature: β = 0.44, OR = 1.6 (1.2–2.0)

Known exposure: β = 1.51, OR = 4.5 (2.4–8.7)

Positive chest X-ray: β = 1.69, OR = 5.4 (3–10)

Nakakubo [23]

Patients suspected of having COVID-19

RT-PCR test

31

Demographics

Comorbidities

Patient history

Symptoms

Respiratory failure

Laboratory test

CT imaging

Differential diagnosis

131

Score

Blood test score

WBC < 8000 (count/μL) = 1 point

Eosinophil < 50 (count/μL) = 1 point

Procalcitonin < 0.5 (ng/mL) and CRP ≥ 0.5 (mg/dL) = 1 point

CT imaging score

Ground glass opacity; with or without consolidation = 1 point

Multipolar or bilateral lesions = 1 point

Subpleural or lower lung dominant distribution = 1 point

No atypical signs = 1 point

Alternative diagnosis score (choose one)

More likely other diagnosis = 0 point

Hard to determine = 2 points

More likely COVID-19 = 4 points

Maximum 11 points

0–4 points: Low risk

5–7 points: Moderate risk

8–11: High risk

Plante [24]

Patients aged ≥20 years from 66 hospitals

RT-PCR test

29

Laboratory test

Training dataset: 12183

Sensitivity analysis dataset: 7842

External validation dataset: 172754

XGBoost

XGBoost model

Eosinophils (%) 23.62

Calcium total (mg/dL) 16.85

AST (IU/L) 9.45

WBC (K/uL) 6.73

Basophils (%) 6.36

RDW (%) 5.13

RBC (m/uL) 5.11

Albumin (g/dL) 4.65

Bilirubin total (mg/dL) 4.30

MCV (fL) 3.98

MCH (pg) 3.49

Sodium (mEq/L) 3.29

Bicarbonate (mEq/L) 2.97

BUN (mg/dL) 2.40

Chloride (mEq/L) 1.68

Total score/100

Sung [16]

Patients who were admitted to the hospital from the emergency room to regular floors and tested for COVID-19

RT-PCR test

38

Demographics

Medical history

Contact tracing

Laboratory test

Symptoms

Chest X ray imaging

Development cohort = 203

Validation cohort = 135

Logistic regression

Score

Nursing facility residence: 9.63 (3.02–30.67), p < 0.001, Points: 10/2/3

Obesity: 2.93 (1.32–6.51), p = 0.008, Points: 3/1/1

Hypoxia: 3.52 (1.58–7.83), p = 0.002, Points: 4/1/1

Sick contact: 10.47 (2.67–41.04), p = 0.001, Points: 10/2/3

Constitutional symptom (fever, chills or myalgia): 2.31 (1.05–5.10), p = 0.038, Points: 2/1/1

Respiratory symptom (cough or shortness of breath): 4.36 (1.73–10.98), p = 0.002, Points: 4/1/1

Gastrointestinal symptom (nausea, vomiting or diarrhea): 4.11 (1.59–10.64), p = 0.004, Points: 4/1/1

Leukocytosis: 0.33 (0.12–0.90), p = 0.030, Points: − 3/− 1/− 1

Tordjman [14]

Subjects with clinical suspicion of SARS-CoV-2 infection

Both RT-PCR test and CT-Scan

31

Demographics

Comorbidities

Symptoms

Laboratory test

Derivation cohort: 100

Validation cohort: 300

XGBoost

Logistic regression:

Basophils < 4G/L: β = 1.66, OR = 5.27, p-value = 0.03

Eosinophils < 0.06 G/L: β = 1.98, OR = 7.23, p-value = 0.01

Lymphocytes < 1.3 G/L: β = 2.54, OR = 12.72, p-value = 0.001

Neutrophils <5G/L: β = 2.10, OR = 8.17, p-value = 0.006

Score:

Eosinophils < 0.06 G/L: 1 point

Lymphocytes < 1.3 G/L: 2 points

Neutrophils <5G/L: 1 point

Basophils < 0.04 G/L: 1 point

Score, maximum 5 points: 0–1 - > low probability, 2–3: intermediate probability, ≥ 4 points

Vieceli [15]

Patients aged 18 or older admitted to hospital due to suspected COVID-19

RT-PCR test

43

Clinical data

Laboratory test

X ray imaging

100

Logistic regression

Logistic regression:

Leukocyte count < 7.7 × 103 mm3: OR(CI) = 17.63 (3.68–84.7), p-value < 0.001

LDH > 273 U/L: OR(CI) = 5.42 (1.18–24.7), p-value = 0.03

Any chest radiographic abnormality: OR(CI) =27.8 (2.5–309.1), p-value = 0.007

Score:

Two points for LDH > 273 U/L, three for leukocyte count ≤7.7 × 103 per mm3 and four for any chest radiography abnormality. A result ≥5 points was considered positive.

  1. Abbreviations: ALC Absolute Lymphocyte Count, ANC Absolute Neutrophil Count, AST ASpartate aminotransferase, BUN Blood Urea Nitrogen, CAD Coronary Artery Disease, CI Confidence Interval, CRP C-Reactive Protein, CT-scan/imaging Computed Tomography-scan/imaging, ED Emergency Department, LDH Lactate DeHydrogenase, MCH Mean Corpuscular Hemoglobin, MCV Mean Corpuscular Volume, OR Odds Ratio, qSOFA quick Sepsis-related Organ Failure Assessment, RBC Red Blood Cells, RDW Red cells Distribution Width, RT-PCR reverse-transcription polymerase chain reaction, SE Standard Error, VAS Visual Analogic Scale, WBC White Blood Cells