The following is a summary of “A validated artificial intelligence-based pipeline for population-wide primary immunodeficiency screening,” published in the JANUARY 2023 issue of Allergy & Immunology by Rider, et al.
It was difficult to identify people who had underlying inborn immune deficiencies and predispositions to infection. Such patients frequently have increased morbidity and less-than-ideal outcomes as a result of the extended diagnostic journey that follows, which highlights the need for deliberate approaches to raising diagnosis rates. For a study, researchers sought to construct and verify a generalizable analytical pipeline for detecting infection susceptibility and the risk of primary immunodeficiency throughout the population.
Weighted rules and a machine learning classifier were used for risk stratification in the prospective, longitudinal cohort research. Data on claims were examined over the course of 30 months iteratively from a varied community (n = 427,110). For new diagnoses, hospital stays, and acute care visits, cohort outcomes were given. The research complied with the TRIPOD (Transparent Reporting of multivariable prediction model for Individual Prognosis Or Diagnosis) guidelines.
Members of the cohort who were initially classified as high risk were proportionately more likely to be given a primary immunodeficiency diagnosis than cohort members who were classified as low-medium risk or who had no claims of interest (9% vs. 1.5% vs. 0.2%; P< .001, chi-square test). An annual individual snapshot of difficulty for triaging referrals was made possible by later machine learning stratification. A single-dense layer neural network architecture was utilized in the top-performing machine learning model for visit-level prediction in this study (area under the receiver-operator characteristic curve = 0.98; F1 score = 0.98).
Identification of people with primary immunodeficiency and precise clinical risk quantification can both be facilitated by a two-step analytical approach.
Reference: jacionline.org/article/S0091-6749(22)01343-4/fulltext