脑瘫患儿精细运动功能测试量表的单维性和信度研究

目的 分析脑瘫患儿精细运动功能测试（FMFM）量表的单维性、内在信度和外在信度。方法 共纳入696例脑瘫患儿为研究对象，男481例（69.1%），女215例（30.9%）；年龄2～183个月，平均（30.0±25.9）个月，0～3岁496例（71.3%），>3岁200例（28.7％）；其中痉挛型四肢瘫239例（34.3%）、痉挛型双瘫212例（30.4%）、痉挛型偏瘫185例（包括2例单瘫）（26.6%）、徐动型30例（4.3%），肌张力障碍型21例（3.0%），共济失调型9例（1.3%）。由86项测试项目组成FMFM的取样量表，所有研究样本均接受过至少1次FMFM取样量表测试。采用Rasch分析中的局部评分模型（PCM）对696例样本和86项取样量表进行分析，以项目不适合标准(FITI)中的均方来确定量表的单维性，同时确定FMFM量表的最终入选项目，然后分析由这些项目组成的FMFM量表的测试独立性和样本独立性，通过选取研究样本中最初的23例进行重测信度研究（间隔1～7 d），同样选取最初的49例进行评分者间信度研究。结果 经过Rasch分析的3轮筛选，从86项取样量表项目中筛除25项，剩余的61项中仅有3项属于不适合项目，占总项目的4.9%（3/61），表明其中的绝大多数项目具有良好的单维性，由此形成正式的FMFM量表。通过分析不同项目状态下的样本能力分值之间的相关性，显示FMFM量表具有很好的测试独立性；通过在分析了不同样本状态下的项目难度值的相关性后，确定FMFM量表具有很好的样本独立性。FMFM量表还具有良好的重测信度（ICC=0.989 3，95%CI：0.975 3～0.995 4）和测试者间信度（ICC=0.996 1,95%CI：0.993 2～0.997 8）。结论 脑瘫患儿FMFM量表具有良好的单维性、内在信度和外在信度，为将来使用FMFM量表评价奠定了良好的基础，FMFM量表的效度和反应度还有待于进一步研究。

Study on reliability and unidimension of the Fine Motor Function Measure Scale for children with cerebral palsy

Objective To analyze unidimension，internal reliability and external reliability of the Fine Motor Function Measure Scale （FMFM）for Children with Cerebral Palsy.Methods 696 children with cerebral palsy participated in the study. 481 males（69.1%）and 215 females，the average age was 30.0 months（SD:25.9months），ranged from 2 to 183 months；Types of CP in the children were spastic quadriplegia（n=239，34.3%），spastic diplegia（n=212，30.4%），spastic hemiplegia（n=185，26.6%），athetoid（n=30，4.3%），dystonic（n=21，3.0%），ataxic（n=9，1.3%）.The 86 test items formed FMFM sampling scale. All samples were measured by FMFM sampling scale at least once.86 test items of FMFM sampling scale and 696 samples were analyzed by Rasch analysis using the partial credit model（PCM）. Unidimension of scale was determined by item misfit criterion(FITI), and the FITI was set as mean square values (MnSq) from 0.6 to 1.4, At the same time ultimate selected items of the FMFM were determined.Then test-free person measurement and sample-free item calibration of these items components of FMFM were analyzed. The first 23 samples were included for test retest reliability study with a interval of 1 to 7 days. 49 samples were selected for interscorer reliability analysis.ResultsAfter three screenings, 25 items were deleted from 86 items of FMFM sampling scale. Among the remained 61 items, the misfit items were 3, less than 5% of all, suggesting that the majority items of this measure were good in unidimension, The mean standard error of item difficulty measuring was 0.09 (range 0.06-0.15), and the reliability of items was 1.00, showing that these items had good reliability. The item separation value was 38.09, showing that these items had good stratification of difficulty.The mean standard error of the sample ability measure was 0.30 (range 0.21-1.05). The reliability of samples was 0.99, showing that the FMFM had good reliability. The item separation value was 9.13, showing that the FMFM had good stratification of difficulty. finally formal FMFM Scale of 61 items was composed. At the same time the difficulty order and criterion of the ultimate selected item were determined. Correlations between sample ability scores of different item conditions were analyzed and results indicated that FMFM had good test-free person measurement, The Pearson correlation between sample ability scores under different item conditions was very high(r=0.976-0.995). Correlations between items difficulty values of different sample conditions were also analyzed and results confirmed that FMFM had good sample-free item calibration. The Pearson correlation between item difficulty under three sample conditions was very high(r=0.993-0.998). FMFM had good test retest reliability （ICC=0.989 3，95%CI：0.975 3-0.995 4）as well as good interscorer reliability（ICC=0.996 1,95%CI：0.993 2-0.997 8）.Conclusions FMFM had excellent unidimension，internal reliability and external reliability，which would lay good foundation for future usage of FMFM. Further studies about validity and responsiveness of FMFM are needed.