Table 3.
Odds Ratios from Logistic Regression by Data Source of Outcome Definition
| Outcome Definitionb | Exposure: Eczemaa | Exposure: Psoriasisa | |||
|---|---|---|---|---|---|
| OR (95% CI)c | nd | OR (95% CI)c | nd | ||
| UK Biobank | |||||
| Depression | Interview-reported diagnosis | 1.36 (1.27–1.45) | 229,393 | 1.11 (1.01–1.21) | 229,371 |
| Anxiety | Interview-reported diagnosis | 1.30 (1.14–1.47) | 229,393 | 1.29 (1.08–1.52) | 229,371 |
| Depression | PHQ-9 ≥ 10 (survey) | 1.24 (1.09–1.39) | 69,420 | 1.19 (1.00–1.41) | 69,417 |
| Anxiety | GAD-7 ≥ 10 (survey) | 1.20 (1.05–1.37) | 69,737 | 1.19 (0.98–1.44) | 69,734 |
| Linked GP data | |||||
| Depression | ≥ 1 diagnosis pre-interview | 1.56 (1.50–1.62) | 229,378 | 1.38 (1.31–1.46) | 229,356 |
| Anxiety | ≥ 1 diagnosis pre-interview | 1.53 (1.46–1.60) | 229,390 | 1.38 (1.29–1.47) | 229,368 |
| Depression | ≥ 1 diagnosis pre-survey | 1.50 (1.39–1.63) | 64,499 | 1.37 (1.22–1.54) | 64,496 |
| Anxiety | ≥ 1 diagnosis pre-survey | 1.43 (1.30–1.58) | 64,502 | 1.39 (1.21–1.59) | 64,499 |
Notes: aExposures defined using self-reported previous diagnosis at the UK Biobank recruitment interview, or through records in linked GP data prior to the timepoint (at least 1 diagnosis + 2 prescription codes on separate days for eczema; 1 diagnosis for psoriasis). bAt the Initial interview timepoint (in grey), outcomes are defined either as a self-reported previous doctor’s diagnosis, or at least 1 diagnosis code in linked GP data prior to the interview. At the 2016 mental health follow-up survey (70,878 responded), outcomes are defined either as a score of ≥10 in the PHQ-9 score for depression/the GAD-7 score for anxiety, which take into account symptoms in the 2 weeks prior to the survey, or at least 1 diagnosis code in linked GP data prior to the follow-up survey. cOdds ratios (95% confidence intervals) estimated from logistic regression for having a mental illness (adjusted for age, sex, deprivation and ethnicity) comparing people with the respective skin disease to people without the respective skin disease. dNumber of observations that went into the model. Observations with missing values were dropped. “Prefer not to answer” and “Do not know” were treated as missing values. For the follow-up survey timepoint, only used GP data where all of the questions of the mental health follow-up survey were answered.