Association between nutritional status and subjective health status in chronically ill children attending special schools

Nine primary schools for chronically ill children throughout the Netherlands were invited to voluntarily participate in this study. The screening of the children took place between December 2010 and January 2011. There were no exclusion criteria. Because of the standard nature of the measurements taken regularly by the nurses of these schools, waived informed consent was obtained from our local medical ethical commission (MEC-2015-060). Parents or caregivers were asked for permission through a letter and could refrain from participation without consequences. Two nurses at each school assessed child characteristics, nutritional status, risk of developing malnutrition, and subjective health status in the children.

For all children, age, gender, and diagnosis concerning chronic illness were recorded. The chronic illness was classified as neurologic, respiratory, endocrine, cardiac, metabolic, gastrointestinal, renal, oncological, multiple diagnostic groups, and others. For analysis, these types of chronic illness were classified according to the first question of the nutritional risk-screening tool STRONG_kids (see Fig. 1). Furthermore, the onset of the child’s chronic disease was reported as: the chronic disease was present at birth or the child developed the chronic disease during life. Children were classified with or without a syndromal disorder. Ethnic background was classified as Caucasian and non-Caucasian.

Descriptive analyses were used to describe the study population. WFA, HFA, and WFH deviations from 0 were carried out using one-sample t tests and were represented as standard deviation scores (SDS). Since age was not normally distributed, comparison of continuous age scores between children with malnutrition versus those without malnutrition was completed with Mann–Whitney U test. Pearson’s Chi-square tests were used to test differences in nominal and ordinal distributions between children with malnutrition versus those without malnutrition of: gender, ethnic background, disease present at birth, and prediction of malnutrition (low risk versus moderate/high risk).

To determine the association between malnutrition and subjective health status (continuous VAS scores), first univariate regression analysis was applied. Since VAS scores differed significantly between special schools, these univariate analyses were corrected for site of data collection (transformed as dummy variables). After this, all variables that were significantly associated with VAS scores were forced simultaneously into a multiple regression analysis. Variables that were not significant (p > .050) in the final model were removed (backward elimination procedure), and then, the total explained variance (R ²) was calculated. Site of data collection was forced into the models.

As to the assumptions of multiple regression analysis, we first checked for multicollinearity and calculated the variance inflation factor (VIF). For each model, the average of the VIF’s of the entered functional health variables was around one, which is expedient. The linearity assumption was examined with a plot of the standardized residuals on the Y-axis and the standardized predicted values on the X-axis. No pattern was discovered in this scatterplot. To check normality, the histogram of the regression-standardized residuals was inspected. This histogram was normally distributed. Statistics were conducted using SPSS version 21.0, and a two-sided p value of <0.05 was considered significant.

A total of 642 children were included from nine schools, with a median of 70 children participating from each school (range 39–109), see Supplemental Table A for number of children in each school. Because of the standard nature of the measurements that are taken regularly by the nurses of these schools, all children who were present between December 2010 and January 2011 were included (see Fig. 2). The median age was 9.8 years (IQR 7.7–11.5), 60 % were male, and 71.8 % were Caucasian (see Table 1). Medication use was reported in 419 children (65.3 %) including 106 children using corticosteroids, 25 children using antibiotics, and 15 children using diuretics. Weight and height were available in 99.2 and 99.4 % children, respectively. The mean SDS for WFA (−0.16), WFH (0.41), and HFA (−0.63) all significantly differed from 0 (WFH and HFA p ≤ 0.001; WFA p ≤ 0.01).

The prevalence of children with acute malnutrition was 3.0 %. The prevalence of chronic malnutrition was 13.8 %. Since almost 1 % of the children were both acute and chronically malnourished, the overall prevalence of malnutrition was 16.0 % (see Table 1). No differences in malnutrition were found between boys and girls or between Caucasian and non-Caucasian children.

The four items of the STRONG_kids questionnaire, i.e. ‘high-risk disease’, ‘subjective clinical assessment’, ‘nutritional intake and losses’, and ‘weight loss or poor weight gain’, were scored with ‘yes’ in 30.1, 12.3, 10.1, and 2.6 % of the children, respectively. Based on the STRONG_kids questionnaire, 59.3 % of the children were classified in the low-risk group, 38.5 % in the moderate-risk group, and 2.2 % in the high-risk group for developing malnutrition. The highest prevalence of acute, chronic, and overall malnutrition was found in the high-risk group, though there was no significant difference between the moderate- and high-risk group (see supplemental Table C).

Overall, chronically ill children reported ‘some problems’ or ‘severe problems’ on the health status dimensions mobility (15 %), self-care (17 %), usual activities (19 %), pain/discomfort (22 %), and anxiety/depression (22 %). Their mean VAS score was 73.0 (SD 11.1). Higher age was significantly positively associated with a higher VAS score, r = .21, p < .01.

In Table 2, the counts and percentages of subjective health status problems in children with and without malnutrition, with risk of malnutrition versus those without risk, with their chronic disease present at birth versus those whose disease developed later in time, and for those children with medication usage versus those without medication usage are presented. For children with overall malnutrition, nurses reported significantly more often some/severe problems on the dimensions mobility (24 %) and self-care (24 %) than for children without overall malnutrition (13 and 15 %, respectively) (see Table 2). Nurses also reported more often some/severe problems on the dimensions mobility (19 %), self-care (23 %), and usual activities (23 %) for children whose chronic disease was present at birth, compared with those whose chronic disease developed later in life (respectively, 10, 10, and 19 %).

For children who were at risk of developing malnutrition (STRONG_kids), nurses reported more some/severe problems on usual activities (23 %) and pain/discomfort (29 %), compared with those children who were not at risk (respectively, 16 and 17 %). Nurses also reported more pain/discomfort problems (26 %) in children who used medication compared with children who did not use medication (16 %). Nurses also reported more anxiety/depression problems in 25 % of the children who used medication versus in 16 % of the children who did not use medication.

Children with overall malnutrition obtained a lower VAS score than children without overall malnutrition M = 70.6 (SD = 11.3) versus M = 73.6 (SD = 11.0), t(624) = 2.46, p < .02. As to the risk of malnutrition, those children with a low risk obtained a higher VAS score compared with those with a moderate/high risk M = 74.2 (SD = 11.0) versus M = 71.4 (SD = 11.1), t(626) = 3.11, p < .01. For children whose chronic illness was present at birth, a significantly lower VAS score was reported compared with those who developed their chronic illness during life M = 72.3 (SD = 9.9) versus M = 75.6 (SD = 10.9), t(453) = 3.31, p < .01. Nurses reported a significantly lower mean VAS score in children who used medication than in children who did not M = 71.8 (SD = 10.7) versus M = 75.5 (SD = 11.5), t(626) = 4.07 p < .01.

In univariate regression analysis, corrected for the different special schools, malnutrition in chronically ill children was significantly associated with worse subjective health status; explaining 35 % of its variance (see Table 3). In multivariate regression analysis, all variables that significantly differed between children with malnutrition and those without malnutrition (disease present at birth, risk of developing malnutrition (STRONG_kids), medication usage, and age in years) were forced simultaneously into the model. Corrected for the different special schools, dichotomized risk of developing malnutrition (STRONG_kids) and disease present at birth were no longer significantly associated with VAS scores (p = .33). Medication usage and younger age, together with malnutrition, explained 38 % of the variance of worse subjective health status.

Overall malnutrition was associated with more nurses-reported problems on mobility (capacity for walking about) and self-care (capacity for self-care in terms of washing and dressing). In addition, a moderate or high risk of developing malnutrition (assessed with the screening tool STRONG_kids) was associated with more problems on usual activities (attending school, sport and leisure activities, play, do things with family or friends) and pain/discomfort (to what extent the child experienced pain or discomfort). Disease characteristics such as the presence of the chronic disease at birth and medication usage were also associated with several subjective health dimensions. Remarkably, only chronic medication use was associated with more problems on anxiety/depression dimension (to what extent the child experienced anxious or depressed feelings).

In chronically ill children who are attending special schools, regular measurements of weight and height is standard care. Using these measurements to calculate the SD scores for weight and height and thereby identifying those children with acute (WFH SD scores <−2) and chronic (HFA SD scores <−2) malnutrition might also help to identify children who are at risk of developing problems on subjective health status dimensions. Besides, determination of the risk of malnutrition using a nutritional screening tool like STRONGkids might also be helpful in identifying at-risk children. This is the first study in which the STRONG_kids was applied in a non-clinical setting, and it seemed to be a practical, easy, and reliable tool for assessment of nutritional risk and subjective health status.

Children with a chronic disease in our sample obtained a lower VAS score (M = 73.0) for subjective health compared with children from previous studies with autism spectrum disorder (M = 80.7), with intellectual disorders (M = 79.4), with movement disorders (M = 76.9), and with hearing disorders (M = 88.1) [20]. They also obtained lower VAS scores compared with children with disabilities, ranging from learning disabilities to movement disorders, attending special schools (M = 88.4) [12]. In these two studies, children (who were actually older) reported their own subjective health status, whereas in our study, nurses employed at the special schools assessed the subjective health status of the children. Since in our study children were younger and higher age was associated with a higher VAS score, this may also explain the differences in VAS scores between the studies.

In our study, overall malnutrition, medication usage, and age explained almost 40 % of the variance of the subjective health status VAS. In practice, these variables are easily assessable and may identify those children with a chronic disease, at admission to the special school, who are at risk of a low subjective health status. Overall malnutrition was the sole ‘predictor’ variable in this study that might be influenced with an intervention. Therefore, it might be useful to develop school-feeding intervention programmes. These programmes may help to ameliorate and prevent some of the subjective health status problems. A recent Cochrane review concluded that school meals might have a number of small benefits for children attending regular schools. In younger children, school meals resulted in small improvements in weight, height, school attendance, math performance, and behaviour. Besides, it may have an impact on intelligence scores [21]. For future research, it might also be useful to assess the influence of school-feeding programmes on subjective health status in chronically ill children attending special schools.

This is the first study in a large sample of chronically ill children attending special schools, in which associations between malnutrition, risk of malnutrition, and subjective health were assessed. Moreover, nine out of 11 special schools for chronically ill children in the Netherlands participated in this study, so it is a nationwide study.

Although very few children in these schools did not participate, we had no background information on these non-participating children, so bias may have played a small role. Another limitation is the fact that the one-dimensional EQ-5D questionnaire was used to assess proxy-reported subjective health status instead of a self-reported multidimensional subjective health status measure. For future research, perhaps assessing both self-reported and proxy-reported quality of life, including an emotional reaction to problems, could be more informative.