TABLE 1.
Distribution of the main methodologies used in data‐driven eczema and atopic dermatitis (AD) publications
| Discipline | Methodology | Number of publications | |||
|---|---|---|---|---|---|
| Total (of 620) | Only eczema (of 255) | Only AD (of 265) | Both (of 100) | ||
| Multivariate statistics (MS) | Cluster analysis | 206 | 84 | 90 | 32 |
| Factor analysis | 89 | 26 | 46 | 17 | |
| Logistic regression | 56 | 24 | 17 | 15 | |
| Latent class/Transition models | 55 | 31 | 12 | 12 | |
| Principal component analysis | 46 | 17 | 20 | 9 | |
| Discriminant analysis | 28 | 8 | 16 | 4 | |
| Markov model | 28 | 8 | 12 | 8 | |
| Structural equation modelling | 11 | 5 | 4 | 2 | |
| Mixture model | 7 | 3 | 4 | 0 | |
| Correspondence analysis | 3 | 2 | 1 | 0 | |
| Latent variable model | 3 | 2 | 1 | 0 | |
| Canonical correlation | 1 | 0 | 1 | 0 | |
| Machine learning and artificial intelligence (ML&AI) | Artificial neural networks (including CNNs) | 67 | 44 | 21 | 2 |
| Machine learning | 48 | 25 | 17 | 6 | |
| Support vector machine | 36 | 24 | 11 | 1 | |
| Artificial intelligence | 17 | 8 | 7 | 2 | |
| Decision trees | 17 | 7 | 6 | 4 | |
| Deep learning | 13 | 9 | 3 | 1 | |
| Natural language processing | 12 | 7 | 5 | 0 | |
| Random forests | 9 | 6 | 2 | 1 | |
| Supervised learning | 2 | 2 | 0 | 0 | |
| Unsupervised learning | 1 | 1 | 0 | 0 | |
| Bayesian statistics (BS) | Bayesian framework | 14 | 8 | 5 | 1 |
| Bayesian network | 5 | 4 | 1 | 0 | |
| Random‐effects Bayesian network meta‐analysis | 4 | 0 | 1 | 3 | |
| Bayesian machine learning model | 3 | 2 | 0 | 1 | |
| Bayesian spatial and temporal models | 3 | 1 | 2 | 0 | |
| Naïve Bayesian classifier | 2 | 2 | 0 | 0 | |
| Bayesian meta‐regression | 2 | 0 | 1 | 1 | |
| Bayesian model averaging | 2 | 2 | 0 | 0 | |
| Bayesian latent class analysis | 1 | 1 | 0 | 0 | |
| Random‐effects Bayesian hierarchical model | 1 | 0 | 1 | 0 | |
Note: Search strings representing the methodologies were searched for in the title, abstract, and keywords of the multivariate statistics (MS) and machine learning and artificial intelligence (ML&AI) publications. Methods for Bayesian statistics (BS) were determined manually. Number of publications are given for the total collection and additionally separated according to the term used, only eczema, only AD, or both. For each method, the highest number of publications between only eczema and only AD is bolded.