The Psychometric Analysis And Modification, Clinical Application Exploration Of The Infant-Toddler Meaningful Auditory Integration Scale

Objective:The study could be divided into three chapters.In the first chapter,it was intended to investigate and evaluate the psychometric properties of the Infant-Toddler Meaningful Auditory Integration Scale(ITMAIS)comprehensively,by combining Item Response Theory(IRT)and Classical Test Theory(CTT).Based on the results of psychometric analysis,it was intended to preliminarily modify the ITMAIS,and validate the psychometric characteristics of the modified ITMAIS(ITMAIS-m)by using a separately new sample.In the second and third chapter,the purpose was to explore the new clinical applications of ITMAIS.In the second chapter,by combining ITMAIS with other easy-operating audiologic tests,we intended to establish clinical predictive models,in an effort to help clinicians to screen out children with noteworthy hearing loss,as well as preliminarily predict pediatric hearing degrees,when behavioral audiometry and electrophysiological audiometry were unavailable.This may help to carry out the neonatal hearing screening project across the board to some extent.In the third chapter,we intended to evaluate the auditory function of infants with unilateral hearing loss by using ITMAIS,investigating whether ITMAIS could be applied to distinguish the differences of Early Pre-Lingual Auditory Development(EPLAD)between binaural normal hearing and monoaural normal hearing.Materials and Methods:1 First ChapterThe first chapter were divided into two sections.In the first section,by retrospectively collecting the audiometry and ITMAIS data of children visiting West China Hospital,Sichuan University from Nov.2006 to Jun.2017,we used IRT to analyze the psychometric properties of ITMAIS: 1)Three assumptions(unidimensionality,local independence,monotonicity)were tested;2)When ITMAIS met the three assumptions,the item and scale properties of ITMAIS were evaluated by using the Graded Response Model(GRM),including item difficulty and discrimination parameters,item information,test information and differential item functioning;3)Based on the results above,the ITMAIS was preliminarily modified into ITMAIS-m.In the second section,by collecting a separately new sample of children aged between 0 and 24 months with steady and untreated hearing conditions from West China Hospital,Sichuan University between Jul.2018 and Jun.2019,the reliability and validity of ITMAIS-m was verified,including item-total correlation,inter-item correlation,Cronbach α coefficients,split-half reliability,convergent validity(correlations with Littl EARS Auditory Questionnaire(LEAQ),pure tone average,test age)and known-group validity(the discriminative capacity among different hearing degrees).2 Second ChapterIn the second chapter,by retrospectively analyzing the data of children aged between 0 and 24 months,with varied untreated hearing conditions from West China Hospital,Sichuan University between Dec.2012 and Feb.2021,we established clinical predictive models,in an attempt to distinguish three hearing degrees in infants and toddlers: normal-mild,moderate,and severe-profound hearing loss.The study randomly divided the involved participants into training sample and validating sample with a ratio of 2:1.In the step of model construction,the significant variables were screened out in the training sample by using univariate analysis and ordinal multivariate logistic regression.Then the variables were applied to establish three predictive models with different combinations.In the step of model validation,three models were validated in the training and validating samples separately,by evaluating the predictive accuracy,discriminative capacity and calibration plots.Finally,the nomograms of the models were plotted.3 Third ChapterIn the third chapter,by involving children between 0 and 3 years old,with bilateral normal hearing,unilateral hearing loss and bilateral mild hearing loss at West China Hospital,Sichuan University between Jul.2018 and Jun.2019,it was intended to investigate the differences of ITMAIS scoring and the EPLAD patterns of children with different hearing conditions.Results:1.First Chapter1.1 First SectionIn the first section,a total of 1730 children aged between 1.0 and 221.3 months with 3092 ITMAIS assessments were involved.The children were with normal hearing(7.3%),mild(0.4%),moderate(11.5%),severe(13.8%)and profound(66.2%)hearing loss,and different types of hearing loss accounted for 76.3%(sensorineural),13.1%(conductive)and 8.3%(mixed)separately.A total of 62.8% children were evaluated with ITMAIS before auditory intervention,and the remaining 37.2% children were evaluated with ITMAIS after hearing aids or cochlear implantation.In the first section,the main results were listed below: 1)The completing percentage of items in ITMAIS ranged from 97.6% to 99.9%,and the correlations between items and scale varied from 0.83 to 0.91,correlations between items varied from 0.63 and 0.88.2)Three assumptions: the exploratory factor analysis found that the eigenvalue of the first factor was 7.01,and 75% of the variance could be explained by this factor.Items loading on the first factor varied from 0.80 to 0.90.The confirmatory factor analysis found that excepting the RMSEA,the other indexes supported the assumption of unidimensionality(one scale only measures one latent trait).The residual correlation of items in ITMAIS were all below 0.10,which meant that the ITMAIS met the independence assumption(the assessment of one item would not affect the results of other items).Considering the monotonic increasing curves from item step response function and item response function,the ITMAIS met the monotonicity assumption(the score of an item increases when latent trait increases).3)Item-model fit: Item 2,4,7,8,9 misfitted with GRM when evaluating item-model fit.After deleting item 2,the remaining items excepting item 9 fitted well with GRM.4)Item and scale parameters: With the method of GRM,it demonstrated that the difficulty parameters of the items ranged from-1.583 to 1.167,and the discriminative parameters ranged from 2.380 to 5.335.The information of item 2 was 1.758,information of item 3-10 was between 4.487 and 8.938,and the test information was 47.754.The item characteristic curve of item 2 was in mess,and that of item 9 was centralized.5)None of the items had differential item functioning.Considering that the item 2 was inferior in the results of inter-item correlation,unidimensionality,monotonicity,item-model fitting and item characteristic curve when comparing with other items,we tentatively deleted item 2 and re-analyzed the characteristics of the remaining items and scale.The three assumptions were stable after deleting item 2.Furthermore,the deletion of item 2 alleviated the item-model fit,and did not affect the parameters of the items and scale significantly.In consequence,it was concluded that the ITMAIS could be modified to ITMAIS-m by deleting item 2.1.2 Second SectionIn the second section,a total of 450 infants(aged between 0.1 and 24.0 months)with 450 copies of ITMAIS-m were involved.The infants had normal hearing(17.1%),mild(10.9%),moderate(35.6%),severe(12.0%)and profound(24.4%)hearing loss,and the hearing types were mainly sensorineural(98.1%).Children were all evaluated with ITMAIS-m before auditory intervention.In the second section,the main results were listed below: 1)The completing percentage of items in ITMAIS-m ranged from 90.4% to 99.8%,correlations between items and scale varied from 0.70 to 0.85,and correlations between items varied from 0.42 to 0.82.2)The Cronbach α of ITMAIS-m was 0.919,and the split-half reliability was 0.830.3)In the aspect of convergent validity,ITMAIS-m correlated with LEAQ with a very high r of 0.932,and correlated with pure tone average with high r ranging from-0.670 to-0.909 in different age groups.ITMAIS-m correlated with age with r ranging from 0.092 to 0.831.When children’s hearing loss was milder,the correlations between ITMAIS-m and age was higher;when children were with profound hearing loss,there was no significant correlations.4)In the aspect of known-group validity,ITMAIS-m could significantly discriminate among three different hearing degrees,with effect sizes of Cohen d ranging from 0.41 to 5.83.When discriminating between normal-mild and severe-profound hearing loss,the effect sizes were higher,varying from 2.12 to 5.83.When children were older,the hearing degree discriminative power of ITMAIS-m was higher.2.Second ChapterIn the second chapter,a total of 828 infants aged between 0.1 and 24.0 months,with varied hearing status(normal hearing(30.7%),mild(13.6%),moderate(28.3%),severe(9.2%)and profound(18.2%)hearing loss,sensorineural(84.6%),conductive(13.1%)and mixed(2.2%)hearing loss)were involved in.In the step of model construction,from the analysis of univariate and multivariate analysis in the training group(n=552),the distortion product otoacoustic emission(DPOAE),automatic auditory brainstem response(AABR),acoustic reflex(AR),ITMAIS and ITMAIS assessment age were found significantly different among three hearing groups.In order to meet the requirements of different medical organizations,we combined these five variables in different sets and established three predictive models: Audiologic-ITMAIS model(containing the five variables),Audiologic model(containing DPOAE,AABR,AR),and ITMAIS model(containing ITMAIS and ITMAIS assessment age).In the step of model validation,the parameters of the three models were compared both in the training and validating(n=276)sample.Finally,it was found that the Audiologic-ITMAIS model had preferable predictive accuracy(training: 0.778,validating: 0.819)and discriminative power(training Area Under Curve(AUC): moderate or higher hearing loss: 0.945(0.928~0.963),severe or higher: 0.981(0.972~0.990);validating AUC: moderate or higher: 0.960(0.939~0.981),severe or higher: 0.962(0.940~0.985)).The Audiologic model was better in discriminating moderate or higher hearing loss(training AUC: 0.920(0.895~0.944);validating AUC: 0.938(0.908~0.967)),and the ITMAIS model was better in discriminating severeprofound hearing loss(training AUC: 0.936(0.912~0.960);validating AUC: 0.938(0.911~0.965)).3.Third ChapterIn the third chapter,a total of 167 infants aged between 1 and 35 months,with bilateral normal hearing(n=70),unilateral hearing loss(n=65)and bilateral mild hearing loss(n=32)were involved in.In the group of unilateral hearing loss,infants were with mild(9.3%),moderate(13.8%),severe(32.3%)or profound(44.6%)hearing loss,and the type of hearing loss was mainly sensorineural(73.8%).In the group of bilateral mild hearing loss,all infants were with sensorineural hearing loss.In the aspect of item 2 and item 7-10 scoring,the group with unilateral hearing loss were lower than that with bilateral normal hearing(item 2: 0(0,1)vs 1(0,3),p < 0.001;item 7-10: 18.8%(6.3%,43.8%)vs 31.3%(6.3%,70.3%),p = 0.03).After dividing unilateral hearing loss into two subgroups,the scores of total ITMAIS in the subgroup of severe-profound unilateral hearing loss were significantly lower than those in the group of bilateral normal hearing(34.4%(27.8%,60.2%)vs 47.2%(21.5%,75%),p = 0.024).No matter how severe the unilateral hearing loss was,the scores of item 2 and item 7-10 were all lower than those in the group of bilateral normal hearing(mild-moderate unilateral hearing loss vs bilateral normal hearing: item 2: 0(0,0)vs 1(0,3),p = 0.001;item 7~10: 12.5%(6.3%,18.8%)vs 31.3%(6.3%,70.3%),p = 0.013;severe-profound unilateral hearing loss vs bilateral hearing loss: item 2: 0(0,1)vs 1(0,3),p<0.001;item 7~10: 15.6%(0,62.5%)vs 31.3%(6.3%,70.3%),p = 0.002).When age grew,the scores of ITMAIS in each hearing group increased in a logarithmic way,but differed in increasing rate.The increasing rate was the highest in the bilateral normal hearing group,and unilateral hearing loss group took the second place,and the bilateral mild hearing loss group was the lowest.Conclusions:In the first chapter,it is the first time to combine IRT with CTT to explore the psychometric properties of a tool concentrating on the pediatric auditory function evaluation.The results have demonstrated that the ITMAIS met the assumptions of unidimensionality,local independence and monotonicity.The items had great discriminative capacity and item information.However,the coverage of item difficulty parameters was narrow,and item 2 did not do well in some aspects of psychometric analysis.The modified version-ITMAIS-m,demonstrated satisfied reliability and validity.The first chapter constructed a comprehensive and recommendable way to analyze a scale by combining IRT and CTT.In addition,although it is merely a preliminary exploration of ITMAIS modification,and more proofs are needed,it is worthwhile to notice that the ITMAIS-m is highly correlated with pure tone average,which could be a basis for exploring the new clinical application of ITMAIS.In the second chapter,by concentrating on exploring the new clinical applications of ITMAIS,we mainly investigated the ability of ITMAIS in predicting pediatric hearing loss degrees.By combining ITMAIS with other easy-operating audiologic tests to build clinical predictive models,it would be a convenient and robust way to predict hearing loss in infants and toddlers.These models could be used in some preliminary organizations,in an effort to screen out pediatric apparent hearing loss timely and easily,and preliminarily predict hearing degrees when behavioral audiometry or electrophysiological audiometry are unavailable.In the third chapter,we explored the new clinical applications of ITMAIS in another way.Unilateral hearing loss would delay the EPLAD of infants,and ITMAIS has the capacity to find out the difference of EPLAD between binaural normal hearing and monoaural hearing loss.The second and third chapter broadened the clinical applications of ITMAIS,although more investigations in multicenter with large samples are still required.