Clustering of developmental delays in Bavarian preschool children – a repeated cross-sectional survey over a period of 12 years

Background While most children display a normal development, some children experience developmental delays compared to age specific development milestones assessed during school entry examination. Data exist on prevalence of delays in single areas, but there is lack of knowledge regarding the clustering patterns of developmental delays and their determinants. Methods During the observation period 1997-2008, 12 399 preschool children (5-7 years of age) in one district of Bavaria, Germany, were assessed in twelve schooling-relevant development areas. The co-occurrence of developmental delays was studied by means of Pearson’s correlation. Subsequently, a two-step cluster algorithm was applied to identify patterns of developmental delays, and multinomial logistic regression was conducted to identify variables associated with the specific patterns. Results Fourteen percent of preschool children displayed developmental delays in one and 19% in two or more of the studied areas. Among those with at least two developmental delays, most common was the combination of delays in "fine motor skills" + "grapho-motor coordination" (in 9.1% of all children), followed by "memory/concentration" + "endurance" (5.8%) and "abstraction" + "visual perception" (2.1%). In the cluster analysis, five distinct patterns of delays were identified, which displayed different associations with male gender and younger age. Conclusions While developmental delays can affect single areas, clustering of multiple developmental delays is common. Such clustering should be taken into account when developing diagnostic tests, in pediatric practice and considering interventions to reduce delays.


Background
In the international comparison, developmental delays are defined and assessed differently across countries [1][2][3]. This is not only the case for single developmental delays, but even more for the co-occurrence of delays. In Germany, the term "performance deficits" was defined with focus on relevant skills for entering 1 st class of primary school [3]. In this definition only the occurrence of single development delays was recognized [3]. In contrast, in the U.S.A. and in Canada primarily specific combinations of developmental disabilities were in the focus of interest. The term "development disability" was used for developmental delays which manifested before the 18 th birthday and affected daily functioning in three or more of the following areas: capacity for independent living, economic self-sufficiency, learning, mobility, language receptive and expressive, selfcare, and self-direction [4]. Thus, it is not surprising that in the Anglo-Saxon countries the incidence of combined developmental delays received more attention in the relevant literature than in German-speaking countries.
The acquisition of various skills in the context of individual development is a very differentiated process and varies from child to child [5,6]. Although the vast majority of children in the Western industrialized countries experience an intact somatic, psychological and social development, a variety of developmental trajectories can be observed and developmental delays can be identified [1,7]. Most of previous studies considered delays independently of each other, with focus on motor and language development on the one side [2], and cognitive or mental delays in specific patient populations on the other side [8,9]. In the diagnostic practice for a non-negligible number of preschool children not only a single, isolated developmental delay, but a clustering of delays can be observed. Publications in the area of combined developmental delays appeared first in the early 1990s. Bishop [10] and Nicholoson and Fawcett [11] noticed a combination of delays in development of coordination and language disorders in children. According to other authors, language disorders frequently were associated with attention disorders [9,12,13] or with abnormalities in motor skills, attention and psychosocial development [3,14,15]. While the observations of co-occurrence of developmental delays were made, no exact frequencies were reported. This fact represents a considerable deficit, because it is known that especially combined developmental delays usually have moderate or strong expression, while isolated delays have rather a mild expression [16]. Further, combined developmental delays have a tendency to persistence [17]. The knowledge of these facts might be important for diagnostics in the field of childhood development.
The present study aimed to assess the co-occurrence of developmental delays using data from a school entry examination, which is mandatory for preschool children in Germany and therefore provides an unselected populationbased, non-clinical sample. Prevalences of single developmental delays were subject to previous analyses in the same District of Bavaria [18,19].

Study design
The present investigation is based on a repeated survey using the framework of the mandatory school entry examination in Germany and therefore including each year the complete age cohort at about 5-6 years [20]. In the presence of a severe disease of the eyes or hearing, the child was not considered for standardized school entry examination and not included in the sample. We included in the analysis all children participating in school entry examinations in the years 1997-2008 who had primary residence in the Lower Bavarian District Dingolfing-Landau (N = 12 399).

Content of the examination
In the study district, a manual of the Working Group "School and Youth Health Care in the Public Health Service" was used in a slightly modified form [21]. The corresponding test battery was designed to assess four dimensions of development with corresponding subareas -in total 12 developmental areas ( Table 1). All tests used in the diagnostics are standardized, and every child had to absolve the complete examination. If a test could not be performed as requested, this was considered as a developmental delay in the corresponding subarea.
Diagnosis and documentation of findings were performed by the investigation team of the School Health Service in the district of Dingolfing-Landau. During the entire twelve years study period, this medical team was composed of the same personnel and used the same approach. The analysis is based on anonymized data obtained in these routine examinations and was approved by the ethics committee of the University of Bremen.

Statistical analyses
For data analysis, software package SPSS 19.0 was used [22]. First, we performed descriptive analyses of the sample. Second, in order to assess if some delays are more often combined with others, we used Pearson's correlation coefficient. Given the large sample size, even marginal correlations were significant. Therefore, instead of using significance criterion, we used Cohen's classification of effect sizes for interpretation and focused only on correlations of 0.5 or higher which are considered strong [23]. Next, among children with at least one delay, we studied the clustering of delays beyond just a combination of two delays by means of a two-step cluster algorithm [24]. Finally, we used multinomial logistic regression analysis to identify variables independently associated with specific patterns of delays, considering children with "no delays" as the reference group.

Sociodemographic characteristics
The District of Dingolfing-Landau has over 91,000 inhabitants. The area has a rural infrastructure besides one industrial factory of automobiles. About 93% of the population has the German nationality. Average age of the examined children was 5.95 years (standard deviation 0.39); 51.7% of all children were male, and 89.5% had the German nationality.

Description of prevalence of delays
Of all 12,399 preschool children examined during the study period, approximately two-thirds did not demonstrate any delays, 14.2% had one, 6.8% had two, and the remaining 11.6% -three or more developmental delays. Highest co-occurrences of delays were found for fine body coordination and grapho-motor coordination (9.1%) and for memory/concentration and endurance (5.8%) ( Table 2).

Correlations of developmental delays in different areas
Developmental delays of fine motor coordination and grapho-motor coordination showed the strongest correlation (Pearson's correlation coefficient r = 0.78, Table 2). Also, a high correlation was found between developmental delays in the subareas of memory/concentration and endurance (r = 0.66), and in the subareas of capacity for abstract thinking and visual perception (r = 0.54, Table 2). The remaining correlation coefficients were below 0.5. Despite the differences in strength, all correlations were highly significant (p < 0.0001).

Patterns of concurrent delays
In the cluster analysis restricted to children with a least one developmental delay, five distinct patterns of developmental delays were identified (Table 3). We described the patterns based on most frequent areas of impairment in the corresponding pattern using the following algorithm: first, all those which were recorded in at least 50% of cases; second, if there was only one area above 50%, a second area with high ratings was included; and third, if multiple delays to be included in the definition differed by less than 5%, they were all included in the description of the given pattern. The first pattern were isolated disorders in pronunciation of speech, the second patterncombined delays in subareas of pronunciation, grammar, rhythm of speech and psychosocial development, the third pattern -deficits of subareas of memory/concentration, endurance, abstraction and visual perception. The fourth pattern was dominated by delays of fine body coordination and grapho-motor coordination. The fifth pattern was a combination of cognitive and motor developmental delays (fine motor coordination, grapho-motor coordination,  memory/concentration, endurance). In the patterns three to five, delays in some further areas had also high prevalence.

Variables associated with patterns of delays
Compared to children without any delays, male and younger children had a higher risk for any combination of delays (Table 4). The effects of both factors were less pronounced for disorders of language development (Pattern 1 or 2), while they were substantially stronger for combinations with delays in fine body coordination (Pattern 4) and motor development (Pattern 5). Migration background was associated with a lower risk for isolated delays of pronunciation (Pattern 1) and a higher risk for all other patterns (Pattern 2 to 5) (Table 4).

Discussion
The analysis of data from school-entry examinations in a Lower Bavarian district revealed co-occurrence of delays in closely related development areas in bivariate analysis and a clustering of delays into five distinct patterns associated with sex, age and migration status.
In studies of selected, clinical populations, authors often noted co-occurrence of developmental delays. For *Presented are the percentages of children presenting delays among those identified as members of the specific cluster; the dominating delays for each cluster are marked in bold and used for the description of the cluster. example, Kadesjö and Gillberg [25] noted that 6.1% of children in their study population had combined delays in motor body coordination and ability of attention. In several further studies, children with specific developmental disorders were assessed for further impairments [26][27][28][29][30]. Generally, these studies provided evidence for clustering of developmental delays, but their results are not directly comparable with the current study in the normal, unselected population. In addition, all cited studies have in common that they assessed only a narrow selection of developmental delays. Clustering of multiple developmental delays was not formally investigated yet. Eldred and Darrah [5] used cluster analysis to study developmental delays, but only considering gross motor coordination. In our study, a cluster analysis was carried out with respect to multiple development areas. The patterns we identified are interesting from the point of view of diagnostics on the one side and prevention on the other side. With respect to prevention of negative consequences of the delays, single delays can be addressed by single interventions, while combined delays would require a combination of interventions addressing several aspects at the same time or in sequence. For example, isolated delays in pronunciation can be directly addressed by speech therapy while combined with problems in the use of grammar might require learning the language. The question is interesting -which cannot be answered by the current crosssectional study-if the delays are independent of each other or if possibly delays in some areas negatively affect developmental chances in other areas: for example, contribute delays in cognitive development to motor development (pattern 5)? In addition, in the patterns 3 and 5 also other developmental delays beyond those used to name the clusters were rather frequent. In such case, particularly these patterns can be seen as complex delays.
Male sex and younger age were consistently identified as being associated with a higher risk of single development delays in previous analyses of the same data [18,19], now they shown to be also associated with combined delays. As for younger age, this is not surprising, since the instrument is assessing the development with respect to abilities required for schooling. Some younger children might not have achieved this developmental stage yet. In such case, there might not be a true developmental delay at individual level, but the assessment is conducted too early. Migration background was less commonly associated with isolated pronunciation problemslikely not because pronunciation problems were less common in the migrants, but rather because in migrants they were more often associated with other delays.
The strengths of the analysis are the large, unselected sample from the normal population, collected from consecutive years and examined by the same medical team.
A limitation of our analysis is that only dichotomous outcomes: presence or absence of delays was studied and no information about the severity of delays was collected. A more detailed knowledge of the severity of the delays would allow a better understanding of the need of intervention. Also, we did not study the improvement of delays over time, and it is not fully clear which of those represent just a variation of individual development and which some form of a permanent pathology. In addition, the clinical implications of combined delays are not clear then their long term consequences were not studied yet. We also cannot determine, if interventions would help reducing the burden of delays, but we assume that even if the delays can spontaneously resolve over time, interventions could improve the adjustment of the children.

Conclusions
Most preschool children are going through an intact development without significant deficits in the acquisition of skills relevant for schooling. However, some children display development delays, and those with delays often have not just a single delay but rather there is some cooccurrence of delays in form of specific patterns. This co-occurrence of delays in multiple areas should be considered in designing intervention strategies as addressing several areas in a parallel fashion might be particularly effective. In the future, more attention should be paid to combined developmental delays, especially regarding combinations of delays of motor function and of cognition. Furthermore, factors associated with specific patterns should be studied more in detail, to identify unfavorable constellations. Also, there is a need to study long term outcomes of children with combined developmental delays in a longitudinal manner. We initiated such study in the region where the reported data was collected.