Bennetts Shannon K, Mensah Fiona K, Westrupp Elizabeth M, Hackworth Naomi J, Reilly Sheena
Department of Paediatrics, The University of MelbourneParkville, VIC, Australia; Murdoch Childrens Research InstituteParkville, VIC, Australia; Judith Lumley Centre, La Trobe UniversityMelbourne, VIC, Australia.
Department of Paediatrics, The University of MelbourneParkville, VIC, Australia; Murdoch Childrens Research InstituteParkville, VIC, Australia; The Royal Children's HospitalParkville, VIC, Australia.
Front Psychol. 2016 Nov 11;7:1710. doi: 10.3389/fpsyg.2016.01710. eCollection 2016.
Parenting behaviors are commonly targeted in early interventions to improve children's language development. Accurate measurement of both parenting behaviors and children's language outcomes is thus crucial for sensitive assessment of intervention outcomes. To date, only a small number of studies have compared parent-reported and directly measured behaviors, and these have been hampered by small sample sizes and inaccurate statistical techniques, such as correlations. The Bland-Altman Method and Reduced Major Axis regression represent more reliable alternatives because they allow us to quantify fixed and proportional bias between measures. In this study, we draw on data from two Australian early childhood cohorts ( = 201 parents and slow-to-talk toddlers aged 24 months; and = 218 parents and children aged 6-36 months experiencing social adversity) to (1) examine agreement and quantify bias between parent-reported and direct measures, and (2) to determine socio-demographic predictors of the differences between parent-reported and direct measures. Measures of child language and parenting behaviors were collected from parents and their children. Our findings support the utility of the Bland-Altman Method and Reduced Major Axis regression in comparing measurement methods. Results indicated stronger agreement between parent-reported and directly measured child language, and poorer agreement between measures of parenting behaviors. Child age was associated with difference scores for child language; however, the direction varied for each cohort. Parents who rated their child's temperament as more difficult tended to report lower language scores on the parent questionnaire, compared to the directly measured scores. Older parents tended to report lower parenting responsiveness on the parent questionnaire, compared to directly measured scores. Finally, speaking a language other than English was associated with less responsive parenting behaviors on the videotaped observation compared to the parent questionnaire. Variation in patterns of agreement across the distribution of scores highlighted the importance of assessing agreement comprehensively, providing strong evidence that simple correlations are grossly insufficient for method comparisons. We discuss implications for researchers and clinicians, including guidance for measurement selection, and the potential to reduce financial and time-related expenses and improve data quality. Further research is required to determine whether findings described here are reflected in more representative populations.
育儿行为通常是早期干预的目标，以促进儿童的语言发展。因此，准确测量育儿行为和儿童语言结果对于敏感评估干预效果至关重要。迄今为止，只有少数研究比较了家长报告的行为和直接测量的行为，而且这些研究受到样本量小和统计技术不准确（如相关性）的阻碍。布兰德 - 奥特曼方法和主轴缩短回归是更可靠的选择，因为它们使我们能够量化测量之间的固定偏差和比例偏差。在本研究中，我们利用来自两个澳大利亚幼儿队列的数据（201名家长和24个月大说话迟缓的幼儿；以及218名经历社会逆境的6 - 36个月大的家长和儿童）来：（1）检查一致性并量化家长报告的测量和直接测量之间的偏差，以及（2）确定家长报告的测量和直接测量之间差异的社会人口学预测因素。从家长及其孩子那里收集了儿童语言和育儿行为的测量数据。我们的研究结果支持布兰德 - 奥特曼方法和主轴缩短回归在比较测量方法方面的实用性。结果表明，家长报告的儿童语言和直接测量的儿童语言之间的一致性更强，而育儿行为测量之间的一致性较差。儿童年龄与儿童语言的差异分数相关；然而，每个队列的方向不同。与直接测量的分数相比，那些将孩子气质评定为更难相处的家长在家长问卷上往往报告较低的语言分数。与直接测量的分数相比，年龄较大的家长在家长问卷上往往报告较低的育儿反应性。最后，与家长问卷相比，说英语以外的语言与录像观察中反应性较低的育儿行为相关。分数分布上一致性模式的差异突出了全面评估一致性的重要性，有力地证明了简单相关性对于方法比较严重不足。我们讨论了对研究人员和临床医生的启示，包括测量选择的指导，以及减少财务和时间相关费用并提高数据质量的潜力。需要进一步研究以确定此处描述的结果是否在更具代表性的人群中得到体现。
