Boys Town National Research Hospital, Boys Town, Neb.
National Jewish Health, Denver, Colo.
J Allergy Clin Immunol Pract. 2024 Aug;12(8):2092-2101.e4. doi: 10.1016/j.jaip.2024.04.050. Epub 2024 May 3.
Complex models combining impairment-based control assessments with clinical characteristics and biomarkers have been developed to predict asthma exacerbations. The composite Asthma Impairment and Risk Questionnaire (AIRQ) with adjustments for demographics (age, sex, race, and body mass index) predicts 12-month exacerbation occurrence similarly to these more complex models.
To examine whether AIRQ exacerbation prediction is enhanced when models are adjusted for a wider range of clinical characteristics and biomarkers.
Patients aged 12 years and older completed monthly online surveys regarding exacerbation-related oral corticosteroid use, emergency department or urgent care visits, and hospitalizations. Univariate logistic regressions to predict exacerbations were performed with sociodemographics, comorbidities, exacerbation history, lung function, blood eosinophils, IgE, and FeNO. Significant (P ≤ .05) variables were included in multivariable logistic regressions with and without AIRQ control categories to predict 12-month exacerbations (log odds ratio [95% Wald confidence interval]). Model performances were compared.
Over 12 months, 1,070 patients (70% female; mean [SD] age, 43.9 [19.4] years; 22% non-White; body mass index [SD], 30.6 [8.7]) completed one or more survey (mean [SD], 10.5 [2.8] surveys). In the multivariable analysis, AIRQ control category adjusted for significant clinical characteristics and biomarkers was predictive of one or more exacerbations: odds ratio (95% CI) not well-controlled versus well-controlled: 1.93 (1.41-2.62), very poorly controlled versus well-controlled: 3.81 (2.65-5.47). Receiver operating characteristic area under the curve (AUC) for this more complex model of exacerbation prediction (AUC = 0.72) did not differ from AIRQ (AUC = 0.70). Models with AIRQ performed better than those without AIRQ (AUC = 0.67; P < .05).
Costly and time-consuming complex modeling with clinical characteristics and biomarkers does not enhance the strong exacerbation prediction ability of AIRQ.
已开发出结合基于损伤的控制评估与临床特征和生物标志物的复杂模型，以预测哮喘加重。经过调整以适应人口统计学因素（年龄、性别、种族和体重指数）的综合哮喘损伤和风险问卷（AIRQ）与这些更复杂的模型相似，可预测 12 个月的加重发作。
检验当模型调整为更广泛的临床特征和生物标志物时，是否能提高 AIRQ 加重预测能力。
年龄在 12 岁及以上的患者完成了有关加重相关口服皮质类固醇使用、急诊或紧急护理就诊和住院情况的每月在线调查。采用单变量逻辑回归分析来预测人口统计学、合并症、加重史、肺功能、血液嗜酸性粒细胞、IgE 和 FeNO 与加重相关的变量。具有统计学意义（P ≤.05）的变量包括在包含和不包含 AIRQ 控制类别的多变量逻辑回归中，以预测 12 个月的加重（对数优势比[95% Wald 置信区间]）。比较了模型性能。
在 12 个月期间，有 1070 名患者（70%为女性；平均[标准差]年龄为 43.9 [19.4]岁；22%为非白人；体重指数[标准差]为 30.6 [8.7]）完成了一次或多次调查（平均[标准差]为 10.5 [2.8]次）。在多变量分析中，经过调整以适应显著的临床特征和生物标志物的 AIRQ 控制类别可预测一次或多次加重：控制不佳与控制良好的比值比（95%CI）为 1.93（1.41-2.62），控制极差与控制良好的比值比为 3.81（2.65-5.47）。这个更复杂的加重预测模型（AUC=0.72）的接受者操作特征曲线（ROC）下面积（AUC）与 AIRQ 没有差异（AUC=0.70）。包含 AIRQ 的模型比不包含 AIRQ 的模型性能更好（AUC=0.67；P <.05）。
用临床特征和生物标志物进行昂贵且耗时的复杂建模并不能增强 AIRQ 对加重的预测能力。
