Diagnostic Value Of Adenotonsil Size Combined With Pediatric Sleep Questionnaire In Children With Obstructive Sleep Apnea

Objective: To investigate the diagnostic value of the adenotonsil size combined with the pediatric sleep questionnaire(PSQ)for obstructive sleep apnea syndrome(OSAS)in children,and to determine the best model by comparing and analyzing the efficacy of different prediction models.Methods: One hundred and thirteen children aged 2-16 years who were admitted to the Department of Otolaryngology,Head and Neck Surgery,The First Affiliated Hospital of Anhui Medical University from November 2021 to November 2022 with clinical symptoms of snoring,open-mouth breathing and breath-holding/awakening during nighttime sleep were included.The diagnosis of OSAS was made by the clinician based on the results of nocturnal sleep monitoring,and the children were divided into OSAS and non-OSAS groups.Clinical examinations such as adenoids and tonsils,basic information and results of the PSQ were collected from all patients.The differences in basic information,clinical examination results and questionnaire information between the two groups were first compared,and then the independent risk factors for OSAS were investigated by single-factor and multi-factor logistic regression analysis.The diagnostic value of each model was evaluated using the receiver operating characteristic(ROC)curve and the area under curve(AUC).The Hosmer-Lemeshow test was used to evaluate the calibration of each model and the calibration curves were plotted,the integrated discrimination improvement(IDI)was calculated to compare the performance differences between the models.Finally,the best prediction model was identified through the analysis and evaluation of AUC and IDI,and the best prediction model was visualized by plotting a nomogram.Results: In terms of demographic characteristics,the body mass index(BMI)of the OSAS group was greater than that of the non-OSAS group,with a statistically significant difference between the two groups(P < 0.05).However,there was no statistically significant difference in the distribution of age,sex,height,weight,household registration between the two groups.In terms of clinical data,the OSAS group had higher levels of adenoidal hypertrophy,tonsillar hypertrophy,PSQ and nocturnal sleep monitoring parameters(apnea-hypopnea index(AHI),obstructive apnea index(OAI)and oxygen desaturation index(ODI))than the non-OSAS group,while the lowest pulse oxygen saturation(LSp O2)was lower than that of the non-OSAS group.There was no statistically significant difference between the two groups in terms of clinical symptoms(snoring,open mouth breathing,breath-holding/wakefulness),duration of snoring and comorbidities(simple adenoid hypertrophy,tonsillar with adenoid hypertrophy,chronic tonsillitis,secretory otitis media,allergic rhinitis).2.In exploring the independent risk factors for OSAS,four statistically significant(P < 0.05)indicators(BMI,adenoidal-nasopharyngeal ratio(ANR),tonsillar hypertrophy(TH),PSQ)were screened using single-factor and multi-factor logistic regression analysis,and showed that the four indicators could be used as independent risk factors for causing OSAS.Then,using the occurrence of OSAS as the dependent variable and the statistically significant indicators in the above results as predictors of the model,multiple logistic regression models were developed by individual indicator or combined indicators,and a series of model validation and evaluation were conducted.The ROC curves were used to analyze the diagnostic value of the different models and it was found that the diagnostic value of ANR was the highest among the single indicator,with an AUC=0.747(P <0.001),a sensitivity of 73.4% and a specificity of 64.7%,the diagnostic value of BMI+ANR+TH +PSQ was the highest among the multiple indicators,with an AUC=0.873(P <0.05),a sensitivity of 89.9% and a specificity of70.6%,while the model of adenotonsil size combined with the pediatric sleep questionnaire(ANR+TH+PSQ)had an AUC=0.845(P <0.05),a sensitivity of 91.1%and a specificity of 61.8%.K-fold cross-validation(K=10)was used to internally validate each model,and it was found that most of the models remained highly accurate after validation.The Hosmer-Lemeshow test was used to evaluate the calibration of each model and the results showed no statistically significant differences between the predicted and true values of all models(P >0.05),and the calibration curve results showed that each model had good predictive effect.The IDI value was calculated to compare the difference in performance between the different models,and the results showed that the model of the ANR+TH had better predictive performance than those the model of ANR,TH alone,with a statistically significant difference(P <0.01),the model of the ANR+TH+PSQ had better predictive ability than those built with two of the indicators,with a statistically significant difference(P <0.01).However,by comparing the models with any three of BMI,ANR,TH and PSQ,it was found that the IDI values between the models were small and the differences were not statistically significant(P >0.05),and the predictive performance of the four models was considered to be comparable.It was worth noting that the predictive ability of the model built with BMI+PSQ +ANR+TH indicators was better than that of the above four models,and the difference was statistically significant(P <0.05).The model of BMI+PSQ+ANR+TH had the best predictive ability,and combined with the results of ROC curve analysis,the model of BMI+PSQ+ANR +TH has the highest AUC.Therefore,the model of BMI+PSQ+ANR+TH can be the best predictive model.Conclusion: 1.BMI,ANR,TH,PSQ and nocturnal sleep monitoring parameters(AHI,OAI and ODI)were higher,and LSp O2 was lower in the OSAS group compared to the non-OSAS group.2.BMI,PSQ,ANR and TH were found to be independent risk factors for the development of OSAS in children with snoring.3.AUC analysis showed that among the models built with single indicator,ANR had the highest diagnostic value.Among the models built with multiple indicators,the combination of BMI+PSQ+ANR+TH had the highest diagnostic value.The model of ANR+TH+PSQ also had good diagnostic value.4.The comparative analysis of IDI values showed that the model built with three indicators was better than the model built with two indicators,the model built with two indicators was better than the model built with single indicator.5.The best prediction model was the model built with BMI+PSQ+ ANR+TH,which was validated by K-fold crossover and performed well in terms of differentiation,calibration and prediction performance,and the model was visualized by drawing nomograms,which made it simple and easy to use in clinical practice.It could be used to provide convenience to medical practitioners.