二階段發展篩選策略對於發展遲緩嬰幼兒之效度與預期效用之分析

研究背景與目的：以發展篩選工具早期發現發展遲緩兒童以提供早期介入為療育成功關鍵因素之一。在台灣，簡易兒童發展量表(Simplified Child Developmental Screening Test ，簡稱SiCDeST)已被部分臨床人員使用；然其心理計量學相關資料尚未完整，因此本研究目的之一是探討SiCDeST的再測信度與其效度-多層次概率比(multi-level likelihood ratio)。在臨床上，使用單一發展篩選測驗的問題為，發展遲緩低盛行率常導致低陽性預測率，進而造成過度轉介與家長過度擔心；多階段發展偵測是解決上述問題的方法之一。在台灣，SiCDeST與嬰幼兒綜合發展測驗-篩選量表(Comprehensive Developmental Inventory for Infants and Toddlers-Screening Test，簡稱CDIIT-ST)各有其效度資料，但尚無結合此二工具的二階段發展篩選策略之效度資料；此外，總預期效用值(total expected utility，簡稱TEU)也是作為評估篩選策略之決策指標之ㄧ，因此，本研究另一目的是探討結合SiCDeST與CDIIT-ST的二階段發展篩選策略的效度與TEU。方法:受試者:在再測信度研究中，SiCDeST每個年齡層各收集15對親子。在多層次概率比研究與二階段發展篩選策略效度研究中，從本實驗室過去一個長期追蹤的資料檔中，選取測試月齡在6-29個月，將同時具有SiCDeST、CDIIT-ST與小兒科醫師發展結果判斷的資料選出以作分析，共有406份資料(266個嬰幼兒)。在二階段發展篩選策略TEU研究中，25名專業人員填寫發展篩選評估工具效用值問卷，以得到篩選成本及四種篩選結果(真陽性、真陰性、偽陽性、偽陰性)的效用值。過程:家長在一週內填寫兩次SiCDeST以得到再測信度資料。在SiCDeST多層次概率比研究中，我們將SiCDeST原始分數分為小於60、等於60、70、80、90與100分，探討各分數之概率比值。在二階段發展篩選策略效度研究中，分為二階段發展篩選陽性策略(two-stage positive approach)與二階段發展篩選陰性策略(two-stage negative approach)。發展遲緩兒童的診斷由小兒科醫師根據兩個診斷測驗結果(分別是嬰幼兒綜合發展測驗診斷測驗與貝萊氏兒童發展評估量表第二版)分為發展遲緩兒童與發展正常兒童。在二階段發展篩選策略TEU研究中，用5個視覺類比量尺之效用值問卷得到效用值。資料統計: 次方權重卡柏分析(Quadratic weighted Kappa)用來分析再側信度；以列聯表用算出SiCDeST多層次概率比及二階段發展篩選策略之效度指標值。篩選效度指標包含：敏感度(sensitivity)、特異度(specificity)、預測率(predictive value)、概率比 (likelihood ratio)、約登指數(Youden Index)與診斷勝算比(diagnostic odds ratio)。在二階段發展篩選策略TEU研究中，將四種篩選結果的效用值分別乘上該結果的機率，並加上成本的效用值加總，即得TEU。本研究α值定於0.05(雙尾)。結果：SiCDeST的再測信度顯示重要至極盡完美的一致性(κ=0.63-0.94)。SiCDeST各分數的多層次概率比分別為無限大、7.02、9.36、2.40與0.66。也就是說，二階段陽性發展篩選策略的效度為：敏感度18%、特異度98%、陽性預測率27%、陰性預測率96%、陽性概率比9、陰性概率比0.8、約登指數16%與診斷勝算比11.3；而二階段陰性篩選策略的效度則分別為53%、82%、12%、97%、2.9、0.6、35%與4.8。二階段陽性與陰性篩選策略的TEU分別為0.91與0.73。結論：SiCDeST是一個具有信度且可接受效度的發展偵測工具，但其心理計量學特性仍需進一步探討。根據SiCDeST各分數的多層次概率比，臨床專業人員可以作出適當的臨床決策，如6-29個月兒童之SiCDeST分數小於60的兒童要儘早接受介入或治療; 60或70分的兒童要接受診斷性評估; 80分的兒童需要第二次篩選; 90或100分的兒童則安排下次例行之發展監測。從效度與TEU來看，二階段陽性發展篩選策略比二階段陰性發展篩選策略好。然未來必須進一步於不同場域進行更多研究以探究之。

關鍵字

多階段篩選測驗；兒童；信度；效度；發展遲緩障礙；效用理論

並列摘要

Background and Purposes: It is important to apply reliable and valid developmental screening tests for the early intervention of children with developmental delays (DD) in clinics. The Simplified Child Developmental Screening Test (SiCDeST) is a simple and brief screening test used by some clinicians in Taiwan. However, psychometric properties of the SiCDeST are still limited. Based on evidence-based medicine, the multilevel likelihood ratios of a test are more powerful and useful than one single cut-off point. Therefore, first two purposes of this study are to investigate the test-retest reliability and the multi-level likelihood ratios of the SiCDeST. The problem of administering any single screening test in clinic is the low prevalence rate of DD. The two-stage positive or negative screening approaches are methods to increase the post-test probabilities to wait-test-treatment threshold. The third purpose of this study is then to estimate the validity of the two-stage positive/ negative screening approaches. In addition to validity indices, total expected utility (TEU) is used in the decision making for selecting screening stratgies or tests. However, the TEU of the two-stage screening approaches are not well investigated yet. Therefore, the fourth purpose of this study is to investigate the TEU of the two-stage positive/ negative screening approaches. Methods: Participants: Fifteen dyads were enrolled for the test-retest reliability of the 5 age groups of the SiCDeST, i.e. 6, 9, 12, 18, and 24 months. From the dataset of one previous longitudinal study, there were 266 infants who entered that study at 6-18 months and being followed up at 18-36 months. For data analysis, the children who received tests during 6-29.4 month-of-age and had complete data of the SiCDeST, CDIIT-ST as well as pediatrician’s diagnosis, were included in this study. Therefore, totally 406 data sets were included for analysis. Twenty-five professionals were recruited to fill the questionnaire for the utility estimation of 4 screening outcomes (true positive, true negative, false positive, and false negative) and cost. Procedure: The SiCDeST was filled by parents twice in the time interval of one week for the test-retest reliability study. For multi-level LRs of the SiCDeST, the raw scores of the SiCDeST were set at less than 60, 60, 70, 80, 90, and 100. The scores of the SiCDeST and Comprehensive Developmental Inventory for Infants and Toddlers- Screening Test (CDIIT-ST) were combined for validity analysis in two-stage positive and negative screening approaches. The reference criteria of DD was the diagnosis of one pediatrician who examine or observe the children with the information of two diagnosis tests, the Comprehensive Developmental Inventory for Infants and Toddlers- Diagnosis Test (CDIIT-DT) and Bayley Scales of Infants Development-II (BSID-II). One questionnaire with 5 visual analysis scale (VAS) was developed for estimate the utility values of 4 outcomes and the cost of the screening test. Statistics analysis: Quadratic weighted Kappa was used for the test-retest reliability analysis. Contingency tables were used to calculate the multi-level likelihood ratios of various score of the SiCDeST and validity indices of two-stage screening strategies. Screening validity indices of the two-stage screening approaches include: sensitivity, specificity, positive predictive value, negative predictive value, likelihood ratio positive, likelihood ratio negative, Youden index, and diagnostic odds ratio. Median values of utility for 4 outcomes and cost would be obtained from the questionnaire. Total expected utility would be calculated as the sum of the product of probabilities of four outcomes and its associated median utility, plus the utility of the cost. The α level was set at 0.05 (two tailed). Results: The test-retest reliability of the SiCDeST showed substantial to almost perfect agreement (κ = 0.63-0.94). The multi-level likelihood ratios of the SiCDeST at scores of < 60, 60, 70, 80, 90, and 100 were infinity, 7.02, 9.36, 2.40, and 0.66, respectively. Therefore, children with raw scores of the SiCDeST >= 80 were chosen for the validity and TEU analysis of two-stage screening approaches. The validity indices of the two-stage positive screening approaches were sensitivity 18%, specificity 98%, positive predictive value 27%, negative predictive value 96%, positive likelihood ratio 9, negative likelihood ratio 0.8, Youden index 16%, and diagnostic odds ratios 11.3. The validity indices of the two-stage negative screening approach were 53%, 82%, 12%, 97%, 2.9, 0.6, 35%, and 4.8, respectively. The TEU of the two-stage positive/ negative screening approaches were 0.91 and 0.73, respectively. Conclusions: The SiCDeST is a reliable and acceptable valid surveillance test filled by parents and it can be used in health care settings where the time is limited. However, its psychometric properties need further study. From the results of this study, clinicians may make appropriate decisions base on the multi-level LRs of the SiCDeST. From the results of this study, authors suggest children with raw scores of the SiCDeST < 60 need further intervention, with 60 or 70 need further diagnosis evaluation, of 80 need further second screening test (CDIIT-ST), and of 90 or 100 are arranged for next screening schedule. From the views of validity indices and TEU, two-stage positive screening approach is better than two-stage negative screening approach. Further studies in various setting are needed to clarify the feasibility and validity of the screening approaches before its widespread use.

並列關鍵字

Multiphasic Screening ； Child ； Reliability ； Validity ； Developmental Delay Disorders ； Utility Theory

參考文獻

39. Huang HL, Chuang SF, Jong YJ, Yu L, and Shieh YL. Applicability of BSID-II in diagnosing developmental delay at Kaohsiung area. Kaohsiung J Med Sci 2000; 16: 197-202.

45. Liao HF, Wang TM, Yao G, and Lee WT. Concurrent validity of the Comprehensive Developmental Inventory for Infants and Toddlers with the Bayley Scales of Infant Development-II in preterm infants. J Formos Med Assoc 2005; 104: 731-7.

47. Liao HF, Lee SC, Soong WT, Tseng CC, and Su SC. Factors associated with mental outcomes of non-specific mentally retarded children. JPTAROC 1994; 19:29-37.

80. Wong MK, Chong CK, Wang CM, and Lin HT. Validation of a simplified child developmental screening test in Taiwan. Formosan J Med 1997; 1: 424-39. [in Chinese; English abstract]

81. Wu HY, Liao HF, Yao G Lee WY, Wang TM, and Hsieh JY. Diagnostic accuracy of the motor subtest of Comprehensive Developmental Inventory for Infants and Toddlers (CDIIT) and the Peabody Developmental Motor Scale-Second Edition (PDMS-2) for preschool children. Formos J Med 2005; 9: 312-22. [in Chinese; English abstract]

國際替代計量

二階段發展篩選策略對於發展遲緩嬰幼兒之效度與預期效用之分析

全文下載

主題瀏覽