Table 1 Main characteristics of the identified studies reported in these studies

Dexheimer JW, [10]

America,

2007

Retrospective observational study

Two months study period.

Children (aged 2–18 years) seen in the pediatric ED.

4,115 patient visits.

Not explained.

Freetext diagnosis of “asthma exacerbation,” “status asthmaticus,” “wheezing,” or “reactive airway disease.

GP

BN

MMHC

ANN

SVM

Split-sample validation.

Expert BN AUC 0.959 (95% CI: 0.933–0.977), MMHC BN AUC 0.962 (95% CI: 0.935–0.980), ANN AUC 0.936 (95% CI: 0.902–0.961), and GP AUC 0.956 (95% CI: 0.923–0.976).

Moderate

Emeryk A, [31]

Poland,

2023

Six months study period.

90 children (aged 0–17 years) diagnosed with asthma.

Not explained.

Asthma exacerbation level.

RF

AUC values (younger children: 93.2% (95% CI, 92.1%-94.4%) and 93.0% (95% CI, 92.1%-93.9%), older children: 92.4% (95% CI, 90.9%-93.9%) and

92.4% (95% CI, 91.1%-93.7%)).

Farion KJ, [12]

Canada,

2013

Phase 1 (from November 2006 to May 2007), phase 2 (from February 2009 to March 2010).

Children (aged 1–17 years) diagnosed with asthma. Phase 1: 240, Phase 2: 82.

Mild deterioration: brief treatment (less than 4 hours in the ED) and then discharged home; Moderate deterioration: longer, more aggressive treatment in the ED or observation room (4–16 hours total); Severe deterioration: maximum stabilization and hospitalization for ongoing treatment (more than 16 hours in the ED).

Patient’s exacerbation severity.

NB

DT

EDT

SVM

IB1

IB10

Phase 1: NB: AUC 0.74(0.73, 0.76), DT: AUC 0.59(0.57, 0.62), EDT: AUC 0.70(0.68, 0.72), SVM: AUC 0.63(0.61, 0.65), IB1: AUC 0.56(0.54, 0.58) and IB10: AUC 0.68(0.66, 0.70). Phase 2: NB prediction accuracy 70.7%, PRAM accuracy 73.2% and physicians accuracy 78.0%.

Gardeux V, [17]

America,

2017

Three years study period.

23 pediatric asthmatic patients (age not explained).

Not explained.

Asthma exacerbations.

RF

NB

DT

SVM

KNN

Bayesian classifier achieved 74% accuracy (AUC 0.71; two-sided P ¼.039)

Harmon I, [30]

America,

2024

Four years study period from 2018 to 2021.

Children (aged 2–18 years).

991 patient encounters.

Not explained.

Asthma exacerbations.

Transformer

MLP

Multi-layer perceptron-based model had the best performance (F₁ 0.95, specificity 1.00, sensitivity 0.91, negative predictive value 0.98, positive predictive value 1.00.).

Hurst JH, [26]

America,

2022

Six years study period from January 1, 2014 to December 31, 2019.

5982 children (aged 5–18 years) diagnosed with asthma.

Defined as any encounter with an asthma-related ICD9 or − 10 code and a prescription for a systemic steroid.

Asthma-related exacerbation.

LASSO

RF

XGBoost

Three models performed moderately well (AUC 0.730–0.742) over all three time horizons. Decision rule (sensitivity 70%, positive predictive value 13.8% for 180 day, 2.9% for 30 day.

Juhn YJ, [25]

America,

2022

Two years study period from December 13, 2016, to December 12, 2018.

246 children (aged < 18 years) had persistent asthma or met Predetermined Asthma Criteria (PAC).

Defined as an emergency department visit/hospitalization for asthma or an unscheduled visit for asthma requiring oral corticosteroids.

1-year asthma exacerbation risk.

NB

GBM

Asthmatic children with lower SES had greater BER (¼ 1.35 for HOUSES Q1 vs. Q2–Q4) and a higher proportion of missing information related to asthma care (41% vs. 24% for missing asthma severity).

Kim D, [21]

Korea,

2020

One year study period from September 1, 2017 to August 31, 2018.

14 children (aged 6–14 years) diagnosed with asthma.

Not explained.

PERF value.

K-means

MNL

LSTM

On an average level, cluster 2 has a lower mean PEFR than cluster 1 (218.2 vs. 263.2), significant fluctuation of average PEFR values over the study period.

Lee CH, [28]

China,

2011

One years study period in 2015.

33 children (age not explained) diagnosed with asthma.

Not explained.

Asthma attack.

DT

CAR

PBCAR accuracy 86.89% and recall 84.12%, PBDT accuracy 87.52% and recall 85.59.

Luo G, [14]

America,

2015

Two years study period.

180 children (aged 2–18 years) diagnosed with asthma.

Not explained.

Asthma control deterioration.

RF

DS

NB

DNN

SVM

KNN

Best model accuracy 71.8 %, sensitivity 73.8 %, specificity 71.4 %, and AUC 0.757.

Okubo Y, [22]

Japan,

2020

Seven years and nine months study period from July 1, 2010, to March 31, 2018.

54,981 children (aged six months to 15 years) with asthma exacerbation.

Defined according to the International Classification of Diseases, Tenth Revision (ICD-10) codes.

Variation of antibiotic and adjunctive treatment.

HC

Proportions of antibiotic use decreased from 47.2% in 2010 to 26.9% in 2018. Utilization of antitussives, antihistamines, and methylxanthine showed decreasing trends, the use of mucolytics and ambroxol increased.

Overgaard SM, [29]

America,

2022

Not explained.

Children (aged 6–17 years) diagnosed with active asthma (sample size not explained).

Defined as an inpatient/hospitalization visit for asthma diagnosis, or an ED visit for asthma diagnosis, or an outpatient visit of a patient with asthma diagnosis along with usage of oral corticosteroids medications.

Asthma exacerbations.

LR

SVM

RF

NB

MLP

LR model outperformed the other candidates producing a 0.8 AUC-ROC.

Patel SJ, [19]

America,

2018

Four years study period from January 1, 2012, to December 31, 2015.

Children (aged 2–18 years) with asthma exacerbation.

29,392 ED visits.

Defined as concurrent treatment with salbutamol and systemic corticosteroids.

Hospitalization.

DT

LR

RF

GBM

DT AUC 0.72 (95% CI: 0.66–0.77), LR AUC 0.83 (95% CI: 0.82–0.83), RF AUC 0.82 (95% CI: 0.81–0.83), GBM AUC 0.84 (95% CI: 0.83–0.85).

Rezaeiahari M, [27]

America,

2024

Two years study period.

22,631 children (aged 5–18 years) (2,042 patients with asthma exacerbation, 20,589 patients without).

Defined as an ED visit and/or hospitalization.

ED visit, and/or hospitalization.

RF

CRF

The model in the OOB sample AUC 72%, sensitivity 55% and specificity 78%, in the training samples AUC 73%, sensitivity 58% and specificity 77%.

Robroeks CM, [13]

Netherlands,

2013

One year study period.

39 children (aged 6–16 years) diagnosed with asthma.

Defined according to the latest ATS/ERS: 1) an increase in asthma symptoms (dyspnoea, cough and wheezing) and/or use of short acting b2-agonists for o2 days; and/or 2) a need for treatment with oral corticosteroids; and/or 3) a need for hospital admission.

Primary outcome: asthma exacerbation; Secondary outcome: asthma control score, Lung function tests.

SVM

Six VOCs support vector machines (correct classification 96%, sensitivity 100%, specificity 93%). Seven VOCs models (correct classification 91%, sensitivity79%, specificity 100%) compared to patients without exacerbations.

Sanders DL, [9]

America,

2006

Two months study period.

Children (aged 2–18 years) diagnosed with asthma.

3,023 patient visits.

Free-text ED visit diagnosis of “asthma exacerbation”, “status asthmaticus”, “wheezing”, or “reactive airway disease

Probability of asthma exacerbation in patients presenting to the ED.

BN

AUC 0.959 (95% CI = 0.933–0.977). Sensitivity 90%, Specificity 88.3%, PPV 44.7%, NPV 98.8%, PLR 7.69 and NLR 0.11.

Seol HY, [23]

America,

2021

One year study period from December 13, 2016, to December12, 2017.

184 children (aged < 18 years) diagnosed with asthma.

Defined as an emergency department visit/hospitalization for asthma or an unscheduled visit for asthma requiring oral corticosteroids.

Primary outcomes: 1-year asthma exacerbation risk; Secondary outcomes: time required for clinicians to review EHRs for asthma management.

NBC

AE frequency (IG 12% vs. CG 15%, OR: 0.82; 95%CI:0.374–1.96; P:0.626). Mean health care costs (IG -$1,036 [-$2177, $44] vs. CG +$80 [-$841, $1000]; P = 0.12).

Sills MR,[24]

America,

2021

Five years study period from January 1, 2009, to December 31, 2013.

Children (aged 2–21 years) with asthma exacerbation.

9,069 ED visits.

Not explained.

Hospitalization.

RF

LR

Auto ML AUCs 0.914 and 0.942, RF AUCs 0.831 and 0.886, LR AUCs 0.795 and 0.823.

Spyroglou II,[20]

Greece,

2018

Eight years study period from 2008 to 2016.

65 children (aged 1–14.5 years) diagnosed with asthma.

Not explained.

Asthma exacerbations.

NB

TAN

SNBC

Semi-naive network predicted exacerbation with an accuracy 93.84% and sensitivity 90.9%.

Toti G,[16]

America,

2016

Eleven years study period from January 1, 2002, to December 31, 2012.

Pediatric asthmatic patients in pediatric emergency rooms (age and disease status not explained).

20,959 ED visits.

Not explained.

Patients went to the ER with an asthma attack.

ARM

27 rules were reported, with support ranging from 0.54% to 5.82% and FDR < 13%.

Van Vliet D,[15]

Netherlands,

2015

One year study period.

96 children (aged 6–18 years) diagnosed with asthma.

Defined according to the latest ATS/ERS criteria and were classified as moderate or severe.

Asthma exacerbations.

KNN

Model 1 AUC 0.47, Model 2 AUC 0.54 and Model 3 AUC 0.59. The K-nearest neighbor correctly predicted 52% of the exacerbations in the validation dataset.

Van Vliet D,[18]

Netherlands,

2017

One year study period.

96 children (aged 6–18 years) diagnosed with asthma.

Defined according to the latest ATS/ERS criteria and were classified as moderate or severe.

Asthma exacerbations.

RF

First RF correct prediction was 82%, sensitivity 88%, specificity 75% and AUC 0.90. Second RF correct prediction was 65%, sensitivity 63% and specificity 67%.

Xu M,[11]

America,

2011

Five to six years study period.

581 children (aged 5–12 years) with mild-moderate asthma.

Defined as a visit to the emergency room or hospitalization for asthma symptoms during a clinical trial.

Severe asthma exacerbation.

RF

Using 160–320 SNPs AUC 0.66, using 10 SNPs AUC 0.57 and using only clinical features AUC 0.54.