Table 1.
Association between eczema and ADD/ADHD in children, adolescents and adults
| Study | ADD/ADHD
|
||||
|---|---|---|---|---|---|
| Total subjects | % Prev (95% CI) | Crude OR (95% CI) | Adjusted OR (95% CI) | Adjusted OR 2 (95% CI) | |
| Pooled NHIS (children)a | |||||
| No eczema | 163 522 | 6·4 (6·2–6·5) | 1 (ref) | 1 (ref) | 1 (ref) |
| Eczema | 17 277 | 9·1 (8·7–9·5) | 1·46 (1·36–1·57) | 1·35 (1·24–1·46) | 1·11 (1·01–1·21) |
| Pooled NSCH (children)b | |||||
| No eczema | 152 934 | 7·8 (7·7–8·0) | 1 (ref) | 1 (ref) | 1 (ref) |
| Eczema | 18 588 | 9·6 (9·2–10·0) | 1·45 (1·28–1·63) | 1·48 (1·29–1·70) | 1·25 (1·08–1·45) |
| Pooled NHIS and NSCH (children)c | |||||
| No eczema | 316 456 | 7·1 (7·0–7·2) | 1 (ref) | 1 (ref) | 1 (ref) |
| Eczema | 35 865 | 9·4 (9·1–9·7) | 1·46 (1·36–1·57) | 1·36 (1·22–1·49) | 1·14 (1·03–1·26) |
| NHIS 2012 (adults aged 18+ years) | |||||
| No eczema | 32 072 | 4·0 (3·7–4·3) | 1 (ref) | 1 (ref) | 1 (ref) |
| Eczema | 2483 | 7·5 (5·9–9·1) | 1·94 (1·52–2·48) | 1·73 (1·35–2·22) | 1·61 (1·25–2·06) |
Binary logistic regression models were constructed with ADHD as the dependent variable and eczema as the independent variable. Multivariate logistic regression models were then constructed that included eczema, sex, age, race, household income, highest level of household education, U.S. vs. foreign birthplace, insurance coverage, lifetime history of asthma, and 1-year history of hay fever and food allergy as the independent variables. A second multivariate model was created that additionally controlled for healthcare utilization in the past year. Adjusted prevalence ORs and 95% CIs were estimated. ADD/ADHD, attention deficit (hyperactivity) disorder; CI, confidence interval; NHIS, National Health Interview Survey; NSCH, National Survey of Children’s Health; OR, odds ratio; prev, prevalence; ref, reference; bold indicates significance.
Pooled analyses were performed by merging the datasets and dividing NHIS sample weights by the number of studies (n = 17).
Pooled analyses were performed by merging the datasets and dividing NSCH sample weights by the number of studies (n = 2).
Due to differences in sampling methodology between NHIS and NSCH, sample weights could not be combined. Meta-analyses of multivariate regression analysis effects were performed using a robust variance estimation method.