This study presents a comprehensive assessment tool CDSC developed for Chinese infant, which is a multidisciplinary observation scale to assess different developmental domains in infants. an assessment tool should provide rigorous result to precisely capture or depict a significant alternation of developmental domains in infants as they grow rapidly. It is changeling to precisely portray the developmental trajectory of infants within one year. Since the reliability of the CDSC was examined by the Classical Testing Theory in the previous studies, this study attempted to provide the complementary evidence to promote the clinical application of the CDSC based on the Rasch Model. The overall evidence suggests that the CDSC items present acceptable fitness to the Rasch Model, and the items also construct a unidimensional testing structure that demonstrate powerful differential function to evaluate the developmental level of Chinese Infant. Besides, several items showed the unfitness to the Rasch Model, the infants’ response patterns were found to deviate from the prediction model. However, our study found that the removal of these unfitting items would jeopardize the integrity of the instrument structure. Certain items are identified as items with DIF across age groups and gender. Therefore, our study finding supported that the CDSC can be applied in clinical practice with excellent psychometric properties to measure the developmental level of infant under 12 months old with developmental delay.
The CDSC subscales are established based on the item banks collected by different members based on their expertise before, and the subscales targeted different developmental domains can be used independently from one another. Therefore, the clinician can select those subscales essential for the needs of infants.
The primary result has distinguished serval misfitting items, and reanalysis is done to evaluate the impact of these misfitting items on the instrument psychometric properties. These misfitting items with abnormal MNSQ and ZSTD value indicate the unexpected respond pattern when participants encountered these items, and more analysis is needed to evaluate the impact of these misfitting items. Our result shows that the misfitting items did not jeopardized the instrument structure, and these items have made contribution to maintain the integrity of the test framework. The responds of the participants to these items seem to be deviate from the Rasch Model prediction, they still function as mile stones in corresponding domain. We suggest that the misfitting items representing a dimension partially belong to the concept of corresponding developmental domains, and they also contain other irrelevant information representing another dimension. As previous studies recommend, the misfitting items that do not jeopardize the psychometric quality of the instrument should be reformulated to reflect the intent content more clearly and precisely, for example action speed, attention, engagement, etc.(Wæhrens et al., 2021; Wilson et al., 2011).
It is well known that the cause of DIF is the manifestation of multidimensionality in a test(Erhart, Ravens-Sieberer, Dickinson, & Colver, 2009). The PCA of residual shows that the items in the testing domains fulfills the judge criteria for the unidimensionality assumption. These findings support that the subscales are measuring single latent traits respectively. However, the identification of items with DIF indicated that certain items may be influenced by secondary dimensions. The test of the unidimensionality did not deny the existence of secondary dimensions, and the result showed that these secondary dimensions only explained 1.4%-2.5% of the measurement variance.
Our finding identified 2 items with gender specific DIF, and this result denotes that participants with the same developmental level in corresponding domains responds differently to the items of the CDSC dependent on gender. Generally, these two items seem to advantage the female than the male with the same ability. The number of items that presented DIF corresponding to less than 5% of the total items evaluated in corresponding subscale. For items with age groups DIF (older or younger than 6.5 months), most of these items advantage the older infants. This is not surprising that aging is associated with more access to more sensory input or living experience that allows individuals to learn more skills. Therefore, DIF for age may be explained by the living experience difference that exists among growing infants. Our hypothesis concurred with related studies(Gorton et al., 2011; Morales-Murillo, García-Grau, McWilliam, & Grau Sevilla, 2021). Hence, further efforts are needed to explore this relation, and to determine the influence of living experience and personal factors on child developmental level in different domains.
The results of the Rasch analysis of the SCDC presents the hierarchy order of item difficulty. The item difficulty was corresponding with the developmental trajectory of Infant. Floor and ceiling effects were defined as the proportion of response that scored either higher than 95%(ceiling) or lower than 5%(floor) possible score for that domain(Park, Hong, & Park, 2021). This study report 0.5% (Gross motor), 2% (Fine motor), 0.5% (Adaptive behavior), 1.5% (Speech and language), 1% (Social behavior) ceiling effect, indicating less than 5 persons (200-person total) obtain high score in each domain. However, our sample was collected from inpatient infants with developmental delay or certain risk factors, hence those participants with maximum score may still display important developmental lag in emerge of the milestone that were not detected by the CDSC. Therefore, the clinician my possible observe poor movement quality or abnormal movement strategies in infants with high CDSC scores. Future application of the CDSC should focus more on its utilization in treatment planning based on the observation rather than simply the measurement outcome.
This study also reported 6% (Gross motor), 1.5% (Fine motor), 11% (Adaptive ability), 3.5% (Speech and language), 5.5% (Social behavior) in floor effect, indicating 11 persons in average obtain low score in each domain. As the Rasch model assume, Infant with lower developmental level tend to achieve less growth milestones. This can be explained that infant with developmental delay caused by various factors (Extremely preterm, extremely low birthweight) present much lower developmental level than those born at term in gross motor, fine motor, cognition, and language domains(Kwong et al., 2022).
The person and item separation index were within the recommended range (> 2), hence we suggest that the CDSC items were enough to differentiate person into at least 12 distinct development level. Further, the person and item reliability were within satisfactory range as well(> 0.8), and the result supports that the CDSC is reliable enough if the same testing items were applied on different sample with the same participants characteristic or our study sample received another item set that belongs to the CDSC test structure.
In brief, the CDSC has demonstrated good psychometric properties in our Rasch Model analysis. Based on the previous psychometric properties’ studies of the CDSC, we believe that the CDSC is a reliable tool to describe the developmental trajectory of Infant under 12 months old, the testing items have demonstrated promising unidimensionality and reasonable fitness to Rasch Model. Further, the items content still required further calibration and redefinition to reflect the testing content more clearly. The future study also needs to examine the differential item functioning of the CDSC items on sample with different characteristics (preterm/term, low weigh born/extremely low weigh born, etc.).
Implications For Clinical Practice
Our results provide preliminary evidence for the psychometric properties of the CDSC and indicate that the clinician should be confident with the assessment results. However, our study also identified certain short comings that may jeopardize the accuracy of the assessment outcomes. Certain items may provide misinformation to the real growth trajectory of the infants. Further, the clinicians should also pay special attention to the potential bias caused by gender or age groups that infants with the same developmental level may respond to certain items differently dependent on their age groups.
Study Limitation
Three limitations should be noted in this study. Firstly, the sample recruited in our study was mainly from those with developmental delay, but the developmental level varies from person to person due to different causal factors(extremely low birth weight, extremely preterm, e.g.). Our result is limited to this population. Secondly, the ceiling and floor effect of the CDSC indicate that the items still need more calibration to improves the precision and efficiency of the existing version. Therefore, future studies may adopt our research results to revise the CDSC based on data from different samples, because the developmental trajectory may vary in children with developmental delay caused by various factors. Thirdly, our result cannot be related to the previous studies statistically due to the different testing theories and software. In brief, our studies achieved the similar psychometric quality of the CDSC as previous studies reported by using different mathematically methods, and we also proposed a way to revise the current version to achieve better construct validity based on the Rasch Model.