A Clinical Prediction Model to Assist Screening Patients for Scabies in Primary Care

To develop and test a clinical prediction model based on demography and symptoms to assist in screening for scabies diagnosis in primary care.

Data from a scabies prevalence survey conducted in Auckland, New Zealand (NZ) were analysed using logistic regression to predict diagnosis, defined by either clinical criteria (International Alliance for the Control of Scabies, IACS) or quantitative polymerase chain reaction (qPCR) for scabies derived from a skin swab. Best subsets regression was used to select predictors in the final models, using Akaike's information criterion. Model performance was assessed using internal validation with bootstrap resampling.

From a survey of 181 children aged between 8 months and 14 years, 105 were available for analysis with complete questionnaires. Age and symptoms of itch (in the child and their close contacts) were retained as predictors of the model where clinical scabies diagnosis was the outcome. In addition to these predictors, whether household members were affected by insect bites, skin sores or blisters was also retained as a predictor of qPCR-based diagnosis. Accuracy was good for both models, with discrimination and calibration indices favourable (Harrell's concordance index 0.96 [IACS] and 0.85 [qPCR]; calibration slope 0.82 [IACS] and 0.77 [qPCR]).

Simple logistic models based on common symptoms of scabies, especially itch in the child and 'simultaneous itch in the family', accurately predict scabies status. Such tools have potential use in supporting screening and improving diagnosis of scabies in community and primary care settings.