Table 4

Risk of bias assessment (using PROBAST) based on four domains across 107 studies that created prediction models for coronavirus disease 2019

AuthorsRisk of bias
ParticipantsPredictorsOutcomeAnalysis
Hospital admission in general population
Original review
 DeCaprio et al8HighLowHighHigh
Update 2
 Jiang et al90HighUnclearHighHigh
Diagnosis
Original review
 Feng et al10LowUnclearHighHigh
 Lopez-Rincon et al35UnclearLowLowHigh
 Meng et al12HighLowHighHigh
 Song et al31HighUnclearLowHigh
Update 1
 Martin et al41HighHighHighHigh
 Sun et al40LowLowUnclearHigh
 Wang et al43LowUnclearUnclearHigh
 Wu et al45HighUnclearLowHigh
Update 2
 Batista et al69UnclearUnclearLowHigh
 Brinati et al74UnclearUnclearLowHigh
 Chen et al78HighHighLowHigh
 Diaz-Quijano et al81HighHighLowHigh
 Kurstjens et al95UnclearLowHighHigh
 Mei et al101HighUnclearUnclearHigh
 Menni et al102HighUnclearUnclearHigh
 Soares et al109UnclearUnclearLowHigh
 Tordjman et al110LowUnclearUnclearHigh
 Zhao et al117HighHighUnclearHigh
Diagnosis of severity
Original review
 Yu et al25UnclearUnclearUnclearHigh
Update 1
 Zhou et al46UnclearLowHighHigh
Update 2
 Benchoufi et al71HighLowLowHigh
 Chassagnon et al77LowLowLowHigh
 Li et al97UnclearUnclearUnclearHigh
 Lyu et al99LowUnclearUnclearHigh
 Wang et al114UnclearHighLowHigh
 Zhu et al118LowLowHighHigh
Diagnostic imaging
Original review
 Barstugan et al32UnclearUnclearUnclearHigh
 Chen et al27HighUnclearLowHigh*
 Gozes et al26UnclearUnclearHighHigh
 Jin et al11HighUnclearUnclearHigh†
 Jin et al33HighUnclearHighHigh*
 Li et al34LowUnclearLowHigh
 Shan et al29UnclearUnclearHighHigh†
 Shi et al36HighUnclearLowHigh
 Wang et al30HighUnclearLowHigh
 Xu et al28HighUnclearHighHigh
 Song et al24UnclearUnclearLowHigh
 Zheng et al38UnclearUnclearHighHigh
Update 1
 Abbas et al47HighUnclearUnclearHigh
 Apostolopoulos et al48HighUnclearHighHigh
 Bukhari et al49UnclearUnclearUnclearHigh
 Chaganti et al50HighUnclearLowUnclear
 Chowdhury et al39HighUnclearUnclearHigh
 Fu et al51HighUnclearUnclearHigh
 Gozes et al52HighUnclearUnclearHigh
 Imran et al53HighUnclearUnclearHigh*
 Li et al54LowLowUnclearHigh
 Li et al55HighUnclearHighHigh*
 Hassanien et al56UnclearUnclearUnclearHigh*
 Tang et al57UnclearUnclearHighHigh
 Wang et al42LowUnclearUnclearHigh
 Zhang et al58HighUnclearHighHigh
 Zhou et al59HighUnclearHighHigh*
Update 2
 Angelov et al64HighUnclearHighHigh
 Arpan et al65UnclearUnclearUnclearHigh
 Bai et al66HighUnclearHighHigh
 Bassi et al68HighUnclearHighHigh
 Borghesi et al72HighUnclearUnclearHigh
 Born et al73HighUnclearUnclearHigh
 Castiglioni et al76UnclearUnclearLowHigh
 Guiot et al82HighUnclearLowHigh
 Hu et al86HighUnclearHighHigh
 Islam et al87HighUnclearHighHigh
 Kana et al91HighUnclearHighHigh*
 Karim et al92HighUnclearHighHigh
 Khan et al93HighUnclearHighHigh*
 Kumar et al94HighUnclearUnclearHigh*
 Moutounet-Cartan103UnclearUnclearUnclearHigh
 Ozturk et al104HighUnclearUnclearHigh
 Rahimzadeh et al105HighUnclearUnclearHigh
 Rehman et al106HighUnclearUnclearHigh
 Singh et al107HighUnclearUnclearHigh
 Ucar et al107HighUnclearUnclearHigh
 Wu et al115HighUnclearUnclearHigh
Prognosis
Original review
 Bai et al9LowUnclearUnclearHigh
 Caramelo et al19HighHighHighHigh
 Lu et al20LowLowLowHigh
 Qi et al21UnclearLowLowHigh
 Shi et al37HighHighHighHigh
 Xie et al7LowLowLowHigh
 Yan et al22LowHighLowHigh
 Yuan et al23LowHighLowHigh
Update 1
 Huang et al60UnclearUnclearUnclearHigh
 Pourhomayoun et al61LowLowUnclearHigh
 Sarkar et al44HighHighHighHigh
 Wang et al42LowLowLowHigh
 Zeng et al62LowLowLowHigh
Update 2
 Al-Najjar et al63UnclearUnclearUnclearHigh
 Barda et al67LowLowHighHigh
 Bello-Chavolla et al70UnclearUnclearLowHigh
 Carr et al75LowLowLowHigh
 Chassagnon et al77LowLowLowHigh
 Colombi et al79HighUnclearUnclearHigh
 Das et al80LowLowLowHigh
 Gong et al119LowLowhighHigh
 Guo et al83LowHighUnclearHigh
 Hu et al84HighLowLowHigh
 Hu et al85LowUnclearLowHigh
 Ji et al88LowLowLowHigh
 Jiang et al89UnclearUnclearUnclearHigh
 Levy et al96LowLowLowHigh
 Liu et al98LowLowLowHigh
 McRae et al100HighHighHighHigh
 Singh et al108lowUnclearHighHigh
 Vaid et al112UnclearHighHighHigh
 Vazquez Guillamet et al113HighLowUnclearHigh
 Zhang et al116LowUnclearUnclear/low‡High

PROBAST=prediction model risk of bias assessment tool.

  • * Risk of bias high owing to calibration not being evaluated. If this criterion is not taken into account, analysis risk of bias would have been unclear.

  • Risk of bias high owing to calibration not being evaluated. If this criterion is not taken into account, analysis risk of bias would have been low.

  • Zhang et al evaluated two outcomes: death (low risk of bias) and a composite poor outcome (unclear risk of bias).