Table 2 Studies dealing with SC models and allergic diseases

Prosperi et al. [6]

To predict asthma severity

383 children with asthma (age 6–18 years)

Use of unsupervised statistical learning techniques, such as exploratory factor analysis (EFA), hierarchical clustering (HC) to identify asthma phenotypes

Lung function, inflammatory and allergy markers, family history, environmental exposures, body mass index, age of asthma onset and medications

BN

Significant recognition of asthma severity

Farion et al. [12]

To predict asthma exacerbation

322 Phase 1: 240 children (age 1–17 years)

Phase 2: 82 children (age 1–17 years)

Phase 1: selection of the most accurate machine learning model with WEKA tool

Phase 2: comparison of performance of BN with PRAM score and Physicians

42 attributes corresponding to the patient’s history, current asthma exacerbation, primary assessment and a selected secondary assessment

BN

Phase 1: Accuracy = 68%

Phase 2: Accuracy = 70.7%

Finkelstein et al. [13]

To predict asthma exacerbation

26 adult asthmatic patients

Use of a modern tele-monitoring system at home

Daily self-reports

SVM

Accuracy = 80%

Sensitivity = 84%

Specificity = 80%

Sanders et al. [10]

To predict asthma exacerbation

4023 patients (age 2–18 years)

Use of a SC model to identify patients eligible for asthma care guidelines

Past diagnoses, allergies, family history, medications, social history and vital signs (temperature, respiratory rate, and oxygen saturation)

BN

Accuracy = 96%

Sensitivity = 90%

Specificity = 88.3%

Dexheimer et al. [11]

Same of Sanders et al. [10]

4023 patients (age 2–18 years)

Comparison of machine learning models to best identify patients eligible for an asthma care guideline

Same of Sanders et al. [10]

ANN,BN, Gaussian processes (GP)

BN accuracy = 96%

GP accuracy = 95.6%

ANN accuracy = 94%

Pifferi et al. [20]

To classify asthma control levels

77 patients (age 7.5–17 years)

Assessment of spirometry and fractional exhaled nitric oxide (FeNO) measurements to classify asthma control according to GINA guidelines

1st model: values of spirometry;

2nd model: values of FeNO;

3rd model: values of spirometry and FeNO

ANN

3rd model achieved best performances of classification

Accuracy = 86.4%

Pifferi et al. [17]

To classify asthmatic vs control

123

90 asthmatic children (age 9–16 years) and 33 controls (age 12–13 years),

Pattern recognition analysis of the exhaled breath temperature curve

The rate of temperature increase and the mean plateau value

ANN

Accuracy = 5%

Sensitivity = 77.2%

Specificity = 99%

Jaing et al. [21]

To classify how children manage their asthma

305 children (age 5–14 years)

Each participant was given 10 asthma-based problems and asked to manage them

Each management decision and its order

ANN

Significant classification of five major classes representing different approaches to solving an acute asthma case

Kharroubi et al. [24]

To classify health state of patient with asthma

307 subjects

Estimation of a preference-based index for asthma (five-dimensional asthma quality of life utility index)

99 features about health statuses

BN

BN model is more appropriate than conventionally used parametric random-effects model

Hirsch et al. [4]

To classify asthmatic vs control

6825 adults (age ≥ 16)

Respiratory questionnaires were analyzed by experts and compared with results provided by neural network

12 answers provided by the respiratory questionnaire (wheezing, chest tightness, shortness of breath, night cough), family history of asthma and associated

conditions of hay fever or eczema

ANN

Accuracy = 74%

Chatzimichail et al. [19]

To classify asthmatic vs control

112 children (age 7–14 years)

Three step analysis:1-feature selection with Principal Component Analysis, 2-pattern classification, 3-performance evaluation

46 prognostic factors including data on asthma, allergic diseases,

and lifestyle factors

SVM

Accuracy = 95.54%

Sensitivity = 95.45%

Specificity = 95.59%

Goulart et al. [32]

To classify allergic conjunctivitis vs control

102

48 with allergic conjunctivitis and 54 controls (age 3–14 years)

Allergic conjunctivitis questionnaires were analyzed by experts and compared with results provided by neural network

7 items selected from a questionnaire of 15 answers

ANN

Accuracy = 100%

Takahashi et al. [33]

To classify atopic dermatitis vs control

4610 answers, 2714 infants (12 months old) and 1896 children (2 years old)

To analyze the predictive accuracy of the predictive model for effect of atopic dermatitis in infancy, from the data of the epidemiological survey

Family history (father, mother, siblings, grand-father, grand-mother), food restriction, food allergy, age, food restriction of mother, egg introduced time, cow’s milk introduced time

ANN

Accuracy = 96.4%

Sensitivity = 88.6%

Specificity = 99.5%

Christopher et al. [31]

To classify allergic rhinitis vs control

872 patients of all age groups

Allergic rhinitis reports of intradermal skin tests were analyzed by experts and compared with results provided by neural network

Patient’s history and 40 clinically relevant allergens

ANN

Accuracy = 88.31%

Sensitivity = 88.3%

Specificity = 88.2%

De Matas et al. [23]

To predict the clinical effect of salbutamol

23

12 healthy volunteers and 11 mild asthmatics

In vivo and in vitro data of human subjects were analyzed using SC modeling

Demographic data and urinary levels of salbutamol and metabolite

ANN

Accuracy = 83.5%

De Matas et al. [22]

To predict the clinical effect of salbutamol

18 mild-moderate asthmatic patients

SC modeling to predict the bronchodilator response at 10 (T10) and 20 (T20) min after receiving each of the 4 doses for each of the 3 different particle sizes

Aerodynamic particle size (APS), body surface area (BSA), age, pre-treatment forced expiratory volume in one-second (FEV1), forced vital capacity, cumulative emitted drug dose and bronchodilator reversibility

ANN

Accuracy = 88%

Gandhi et al. [25]

To predict hypersensitivity reaction

2458 reports concerning thrombotic events selected from AERS (adverse event reporting system) database

Retrospective analysis focused on thrombotic events associated with C1 esterase inhibitor products

Adverse events; demographic and administrative information; drug/biologic information; report sources; patient outcomes; drug therapy start and end dates; indications for use/diagnosis

BN

Potential signals of C1 esterase inhibitor product—associated thrombotic events among patients with hereditary angioedema were identified

Naranjo et al. [28]

To predict the posterior probability of a drug (BARDI tool)

51 patients

BARDI tool, calculates the posterior probability of a drug being the cause based on epidemiologic and case data

Reactions after receiving aromatic anticonvulsants

BN

Accuracy = 93%

Sensitivity = 94%

Specificity = 50%

Lanctot et al. [29]

To predict the posterior probability of a drug (BARDI tool)

27 cases of skin reactions

BARDI, combined with the LTA, a biochemical test that determines the percent of cell death because of toxic metabolites of a drug

Skin reactions associated with sulfonamide therapy

BN

Accuracy = 96%

Sensitivity = 79%

Specificity = 38%

Lanctot et al. [30]

To predict the posterior probability of a drug (BARDI tool)

106 challenging cases

BARDI, compared with the Adverse Drug Reaction Probability Scale (APS)

Drug- and nondrug-induced adverse events

BN

BN model discriminate better than ADR drug from nondrug-induced cases.

Kadoyama et al. [26]

To predict hypersensitivity reaction caused by anticancer agents

1,644,220 reported cases from 2004 to 2009 (AERS database)

SC model to detect important pattern related to anticancer agents-associated adverse events

Adverse events; demographic and administrative information; drug/biologic information; report sources; patient outcomes; drug therapy start and end dates; indications for use/diagnosis

BN

Potential signals were detected for paclitaxel-associated mild, severe, and lethal hypersensitivity reactions, and docetaxel-associated lethal reactions

Sakaeda et al. [27]

To predict hypersensitivity reactions caused by platinum agents

1,644,220 reported cases from 2004 to 2009 (AERS database)

The BN analysis aims to search for previously unknown patterns and automatically detect important signals, i.e., platinum agent-associat d adverse events, from such a large database

Adverse events; demographic and administrative information; drug/biologic information; report sources; patient outcomes; drug therapy start and end dates

BN

Significant association between the platinum agent-and mild, severe, and lethal hypersensitivity reactions

Lurie et al. [16]

To classify asthma severity

113

patients (age 42.9 ± 16.3 years)

Implementation of a fuzzy model able to combine patients’ and doctors’ asthma perceptions

Doctor assessment, variables self-assessed by patients (dyspnea, perceived treatment efficacy, asthma-related quality of life questionnaire (AQLQ)), patients’ sociodemographic characteristics, and asthma characteristics

FL

Accuracy = 73%

Zolnoori et al. [8]

To classify asthma control level

42 asthmatic patients

Implementation of a fuzzy model able to estimate the level of asthma control and help physicians to manage their patients more effectively

Respiratory symptom severity, bronchial obstruction, asthma instability, current treatment and quality of life

FL

Accuracy = 100%

Zolnoori et al. [14]

To classify asthma exacerbation

25 patients

Implementation of a fuzzy model able to estimate the level of asthma exacerbation and help physicians to manage their patients more effectively

Status of breathless, status of wheeze, status of alertness, status of respiratory rate, status of talk, status of pulse/min heart rate, value of PEF after initial bronchodilator, value of paCO₂, value of SaO₂%

FL

Accuracy = 100%

Zolnoori et al. [15]

To classify asthma severity

28 patients

Implementation of a fuzzy model able to estimate the four categories of asthma severity and help physicians to manage their patients more effectively

Bronchial obstruction, response to drug, skin prick test, severity of respiratory symptoms, instability of asthma, IgE value, quality of life

FL

Accuracy = 100%

Zolnoori et al. [18]

To classify asthmatic vs control

278

139 asthmatic patients and 139 non-asthmatic patients (age range 6–18)

Implementation of a fuzzy model to help physicians to manage their patients more effectively

Medical history, environmental factors, allergic rhinitis, genetic factors, consequences of asthma on lung tissues, response to laboratory tests and response to short-term medicine

FL

Sensitivity = 88%

Specificity = 100%