Table 1 Categorization of existing predictive models for asthma development in children
From: A systematic review of predictive models for asthma development in children
Category | Articles (year) | Targeted population | Population size | Prediction target (the dependent variable) | Methods for building the predictive models | Predictors included in the final model | Prediction accuracy |
|---|---|---|---|---|---|---|---|
For the general child population | Children at age three | Asthma development at age 6–13 | Clinical index | Seven predictors collected from a parental questionnaire: early wheeze, early frequent wheeze, parental asthma, eczema, blood eosinophilia, wheezing without colds, and allergic rhinitis | The loose asthma predictive index: sensitivity = 57 %, specificity = 81 %, positive predictive value = 26 %, negative predictive value = 94 % [94] | ||
The stringent asthma predictive index: sensitivity = 28 %, specificity = 96 %, positive predictive value = 48 %, negative predictive value = 92 % [94] | |||||||
Chang et al. [97] (2013) | Children at age 1–3 | 289 | Asthma development at age 6–11 | Clinical index | Early wheeze, early frequent wheeze, parental asthma, eczema, blood eosinophilia, wheezing without colds, allergic sensitization to aeroallergens, allergic sensitization to milk, eggs, or peanuts | Sensitivity = 17 %, specificity = 99 %, positive predictive value = 72 %, negative predictive value = 91 % | |
Amat et al. [100] (2011) | Children under age three with a history of ≥3 wheezing episodes and having been assessed for respiratory wheezing disease using a standardized allergy testing program and a doctor-led ISAAC questionnaire [179, 180] | 227 | Asthma development at age 13 | Sensitivity = 87 %, specificity = 37 %, positive predictive value = 61 %, negative predictive value = 71 %, AUC = 0.62, accuracy = 69 % | |||
Singer et al. [98] (2013) | Children aged 3 months–4 years with recurrent coughing or wheezing | 166 | Asthma development six years later | Clinical index | Early wheeze, early frequent wheeze, parental asthma, eczema, elevated fraction of exhaled nitric oxide (FeNO), wheezing without colds, allergic rhinitis | Sensitivity = 75 %, specificity = 62 %, positive predictive value = 58 %, negative predictive value = 78 % | |
Amin et al. [99] (2014) | Children at age three with ≥1 parent with a positive skin prick test | 589 | Objectively confirmed asthma at age seven | Clinical index | Frequent wheezing, parental asthma, allergic sensitization to ≥1 aeroallergens, a history of eczema, wheezing without a cold, allergic rhinitis, allergic sensitization to milk or egg | Sensitivity = 44 %, specificity = 94 %, positive predictive value = 60 %, negative predictive value = 89 % | |
Klaassen et al. [101] (2015) | Children aged 2–4 years with recurrent wheezing | 198 | Asthma development at age six | Logistic regression | The original asthma predictive index [94], exhaled volatile organic compounds, gene expression | AUC = 0.86 | |
Zhang et al. [88] (2014) | Children aged 2–20 months with the first episode of wheezing | 128 | Multi-trigger wheezing in the next two years | Logistic regression | Wheezing severity score computed using the Preschool Respiratory Assessment Measure scoring scale, number of shed exfoliated airway epithelial cells, family or personal history of atopic disease | Sensitivity = 95 %, specificity = 74 %, positive predictive value = 59 %, negative predictive value = 94 % | |
Children at age four | Persistent wheezing at age 10 (wheezing onset by age four and still wheezing at age 10) | Cumulative risk score | Four predictors collected from a parental questionnaire: family history of asthma, recurrent chest infections at age two, atopic skin prick testing at age four, and absence of nasal symptoms at age one | Sensitivity = 53 %, specificity = 85 %, positive predictive value = 68 %, negative predictive value = 74 % [89] | |||
Sensitivity = 22 %, specificity = 97 %, positive predictive value = 65 %, negative predictive value = 81 % [103] | |||||||
Balemans et al. [90] (2006) | Children at age two | 693 | Asthma development at age 21 | Logistic regression | Four predictors collected from a parental questionnaire: female gender, smoking mother, lower respiratory tract illness before age two, and atopic parents | AUC = 0.66, sensitivity = 53 %, specificity = 70 %, positive predictive value = 20 %, negative predictive value = 91 % | |
Children at age four | Four predictors collected from a parental questionnaire: female gender, smoking mother, lower respiratory tract illness between ages two and four, and atopic parents | AUC = 0.68, sensitivity = 71 %, specificity = 53 %, positive predictive value = 18 %, negative predictive value = 93 % | |||||
Clough et al. [17] (1999) | Children aged 3–36 months with first wheezing in the previous 12 weeks and at least one atopic parent | 97 | Receiving prophylactic antiasthma treatment one year later | Logistic regression | Age, serum soluble interleukin-2 receptor (IL-2R) level | Accuracy = 71 %, sensitivity = 57 %, specificity = 84 %, positive predictive value = 76 %, negative predictive value = 68 % | |
Devulapalli et al. [104] (2008) | Children at age two | 449 | Asthma development at age 10 | Severity score | Three predictors collected from a parental questionnaire: number of episodes of bronchial obstruction, number of months with persistent bronchial obstruction, and number of hospital admissions due to bronchial obstruction | AUC = 0.78, sensitivity = 56 %, specificity = 86 %, positive predictive value = 53 %, negative predictive value = 88 % when the severity score was cut off at 5 | |
Marenholz et al. [105] (2009) | Infants | 871 | Asthma development between ages 7 and 13 | Combination of two attributes | Filaggrin gene mutation, increased immunoglobulin E (IgE) levels to food allergens | Sensitivity = 9 %, specificity = 99 %, positive predictive value = 73 %, negative predictive value = 80 % | |
Infants with eczema | Sensitivity = 17 %, specificity = 100 %, positive predictive value = 100 %, negative predictive value = 72 % | ||||||
Children at age five with an asthma diagnosis | 112 | Continued asthma diagnosis at age 7–14 | Evolutionary algorithm consisting of a neural network and a genetic algorithm [106] | Four predictors collected from a questionnaire: cough, bronchiolitis episodes until age five, wheezing, and asthma diagnosis [106] | Accuracy = 95 % [106] | ||
Principle component analysis for feature extraction, least square support vector machine for classification [107] | 46 predictors collected from a questionnaire [107] | Accuracy = 96 %, sensitivity = 95 %, specificity = 96 % [107] | |||||
Partial least square regression for feature selection, neural network for classification [108] | Nine predictors collected from a questionnaire: wheezing episodes until age five, wheezing episodes between ages three and five, wheezing episodes until age three, weight, waist’s perimeter, seasonal symptoms, FEF25/75, number of family members, and corticosteroids inhaled [108] | Accuracy = 97 %, sensitivity = 96 %, specificity = 100 % [108] | |||||
Correlation analysis for feature selection, neural network for classification [109, 110] | Eight predictors collected from a questionnaire: cough, bronchiolitis episodes until age five, until age three, between ages three and five, at age two, at age three, at age four, and at age five [109] | Accuracy = 100 %, sensitivity = 100 %, specificity = 100 % [109, 110] | |||||
Ten predictors collected from a questionnaire: cough, asthma diagnosis, total number of bronchiolitis episodes until age five, bronchiolitis episodes until age three, between ages three and five, until age four, at age one, at age two, at age three, and at age five [110] | |||||||
Lødrup Carlsen et al. [91] (2010) | Children at birth | 614 | Asthma development by age 10 | Logistic regression | Female gender, family network, alcohol in pregnancy, parental rhinoconjunctivis, parental education, lung function at birth (resistance ≤ median, Ve ≤ median, tPTEF/tE ≤ 0.2) | AUC = 0.74, sensitivity = 75 %, specificity = 64 %, positive predictive value = 35 %, negative predictive value = 91 % | |
Spycher et al. [92] (2012) | Children at birth | 5677 | Asthma development at age 7–8 | Logistic regression | Genetic information | AUC < 0.6 | |
van der Werff et al. [102] (2013) | Children aged 4–14 without asthma | 1042 | Asthma development three years later | Logistic regression | Antibiotic use in the child’s first year of life, family history of atopic diseases, allergic sensitization, and municipality | AUC = 0.69 | |
Smolinska et al. [111] (2014) | Children aged 2–4 with recurrent wheezing symptoms | 252 | Asthma development at age six | Random forest for feature selection, dissimilarity partial least squares discriminant analysis for classification | Measurements of volatile organic compounds excreted in breath | Accuracy = 80 % | |
For the primary care setting | Vial Dupuy et al. [18] (2011) | Children under two presenting recurrent wheezing (≥3 wheezing episodes) to a pediatric pulmonology and allergy center’s outpatient department through primary care physicians’ referral | 200 | Development of persistent asthma at age six | Logistic regression | Family history of asthma, atopic dermatitis, multiple allergen sensitizations | AUC = 0.66, sensitivity = 42 %, specificity = 90 %, positive predictive value = 67 %, negative predictive value = 76 % |
Children aged 0–4 at the first time of having asthma-like symptoms in the primary care setting | 2171 in [27] | Asthma development at age 7–8 | Logistic regression | Eight predictors collected from a parental questionnaire: male gender, post-term delivery, parental education, parental inhaled medication, wheezing frequency, wheeze/dyspnea apart from colds, respiratory infections, and eczema | AUC = 0.74, sensitivity = 36 %, specificity = 91 %, positive predictive value = 32 %, negative predictive value = 92 % | ||
2877 in [9] | Asthma development at age six | Male gender, pre-term birth, parental education, parental inhaled medication, wheezing frequency, wheeze/dyspnea apart from colds, respiratory infections, eczema | AUC = 0.75, sensitivity = 37 %, specificity = 92 %, positive predictive value = 22 %, negative predictive value = 96 % when the asthma risk score corresponding to the model was cut off at 12 | ||||
van der Mark et al. [71] (2014) | Children aged 1–5 previously presented to primary care clinic for recurrent coughing, wheezing, and/or shortness of breath | 438 | Asthma diagnosis at age six | Logistic regression | Age, family history of asthma or allergy, wheezing-induced sleep disturbances, wheezing in the absence of common colds, specific IgE for cat, dog, and house dust mite | AUC = 0.73, positive predictive value = 22 %, negative predictive value = 78 % when the asthma prediction score corresponding to the model was cut off at 3 | |
Eysink et al. [65] (2005) | Children aged 1–4 who visited their primary care physicians for persistent coughing of ≥5 days | 123 | Asthma development at age six | Logistic regression | Age, family history of pollen allergy, wheezing, specific IgE for cat, dog, and house dust mite | AUC = 0.87 | |
Children aged 1–3 who visited their primary care physicians for wheeze or cough | Asthma development 5 years later | Logistic regression | Gender, age, wheeze without colds, wheeze frequency, activity disturbance, shortness of breath, exercise-related wheeze/cough, aeroallergen-related wheeze/cough, eczema, parental history of asthma/bronchitis | AUC = 0.76, sensitivity = 72 %, specificity = 71 %, positive predictive value = 49 %, negative predictive value = 86 % when the asthma prediction score corresponding to the model was cut off at 5 | |||
For bronchiolitis patients | Mikalsen et al. [93] (2013) | Children at age two previously hospitalized for bronchiolitis during infancy | 93 | Asthma diagnosis at age 11 | Logistic regression | Four predictors collected from a parental questionnaire: recurrent wheezing, parental atopy, parental asthma, and atopic dermatitis | Sensitivity = 65 %, specificity = 82 %, positive predictive value ≈ 50 %, negative predictive value ≈ 89 % |