Predictive Model Identifies Severity-Related Factors in Atopic Dermatitis

Increasing age, higher eosinophil level, higher economic status, and urban residence are associated with severe atopic dermatitis (AD), according to study findings published in Dermatitis.

Researchers conducted a cross-sectional study of 31 factors and their association with AD severity to develop a clinical predictive model for AD. A questionnaire was completed by 900 patients with AD aged 1 week to 101 years to collect information on potential risk factors.

Participants’ AD severity was measured with use of the scoring atopic dermatitis (SCORAD) index, the eczema area and severity index, the investigator’s global assessment scales, and the dermatology life quality index (DLQI). The primary outcome was a 3-tiered composite ordinal outcome based on the SCORAD score classification.

The researchers created a predictive model based on findings from the multivariate ordinal logistic regression analysis.

Among the cohort, 375 participants (41.67%) had mild AD, 384 (42.67%) had moderate AD, and 141 (15.67%) had severe AD.

Outdoor activity (P =.042), total serum immunoglobulin E (P =.004), eosinophil count (P <.001), onset age (P <.001), age (P <.001), bath frequency (P <.001), bath duration (P =.002), residence (P <.001), feeding (P <.001), skin prick test (P <.001), household income (P <.001), sleep duration (P =.015), hair coloring (P =.012), disinfectants exposure (P <.001), emollients use (P <.001), antibiotics exposure (P =.003), alcohol drinking (P =.023), consumption of sugar-sweetened drinks (P =.003), consumption of pouched or boxed food (P =.034), and dietary bias (P <.001) varied among the 3 groups.

The participants’ SCORAD scores were positively correlated with age (r, 0.175; P <.001), onset age (r, 0.162; P <.001), urban residence (r, 0.161; P <.001), eosinophil count (r, 0.176; P <.001), alcohol consumption (r, 0.090; P =.007), consumption of pouched or boxed food (r, 0.086; P =.010), household monthly income (r, 0.339; P <.001), formula feeding (r, 0.249; P <.001), bath frequency (r, 0.169; P <.001), bath duration (r, 0.122; P <.001).

The SCORAD scores were correlated negatively with sleep duration (r, -0.092; P =.006), positive skin prick test (r, -0.072; P =.031), dietary bias (r, -0.112; P =.001), disinfectants exposure (r, -0.117; P <.001), hair coloring (r, -0.094; P =.005), consumption of sugar-sweetened drinks (r, -0.094; P =.005), antibiotic exposure (r, -0.100; P =.003), and emollient application (r, -0.141; P <.001).

In the multivariable model, age, residence, eosinophil level, feeding, disinfectant exposure, emollients use, bath duration, and household income status were significant risks and included in the prediction model.

For predicting the probabilities of moderate or severe disease with use of the bootstrap method, the concordance index (c-index) was 0.73 (95% CI, 0.70-0.76), and the bias-adjusted c-index was 0.72, which suggested that the nomogram has good predictive value.

Limitations of the study include use of a convenience sample from a single hospital as the study population, and external validation is required. Also, because of the cross-sectional design, a causal relationship between each variable and AD severity could not be established.

“Well-calibrated prediction models were established that accurately identified and discriminated between mild and moderate or severe [atopic dermatitis] patient outcomes,” conclude the researchers.

This article originally appeared on Dermatology Advisor