Emotional and behavioral problems of pediatric cancer survivors and their siblings: Concordance of child self‐report and parent proxy‐report

Childhood cancer confronts families with major challenges. The study aimed at developing an empirical and multi‐perspective understanding of emotional and behavioral problems of cancer survivors diagnosed with leukemia and brain tumors and their siblings. Further, the concordance between child self‐report and parent proxy‐report was examined.


| BACKGROUND
Approximately 2200 children are diagnosed with cancer every year in Germany. The most frequent tumor entities are leukemia (30%) and brain tumors (24%). 1 Improved treatment methods increased survival rates to over 81% for 5 years and longer after diagnosis. 2 Thus, more childhood cancer survivors and their families have to deal with the psychological and physical consequences of the disease and its treatment. These can persist even after the end of intensive therapy and might negatively affect the survivor's cognitive, social, and emotional development. 3,4 Research indicates that siblings are likewise at increased risk for mental and social abnormalities. 5 The results of previous research are, however, very heterogeneous. Some studies suggest that siblings of cancer survivors are more likely to have emotional and behavioral problems than their healthy peers, other studies could not confirm corresponding effects. [5][6][7][8] The Childhood Cancer Survivor Study demonstrates that siblings are psychologically unremarkable in general. However, they identified certain subgroups that are at higher risk for long-term mental health issues. 7 Inconsistent results in literature might be explained in part by differences in the sample (age, diagnosis, treatment), time period after diagnosis, and methods. To date, there is no unified method for assessing emotional and behavioral problems in children. Data are not only collected differently (e.g. quantitative, qualitative), but also from different perspectives (self vs. proxy) and using different instruments.
Surveying children is challenging for researchers, as the level of children's abilities varies greatly depending on their age and health status. Hence, proxy-reports by caregivers are frequently utilized methods of data collection. 9 Discrepancies between self-reports and parental proxy-reports, especially concerning children's psychological well-being, are common. 10,11 In an extensive cohort study of children with cancer, only a low level of concordance between child selfreport and parental assessments of depression, anxiety, and psychological stress was reported. 12 Similar results were published regarding siblings of cancer survivors: a study on health-related quality of life indicates that siblings suffer from more serious problems than parents perceive. 13 Eiser et al. demonstrated that observable symptoms were reported more consistently than unobservable symptoms. Further, concordance was higher between parents and chronically ill children than between parents and their healthy children. Effects related to age and gender were not ascertained. 14 Since parents usually are stakeholders in their children's aftercare treatments, it is necessary to validate the concordance of the children's perspective with the parental perspective regarding strengths and difficulties to identify and counteract emotional and behavioral problems at an early stage. Based on the results of previous studies, we expect low concordance between self-and proxyreports, especially with regard to unobservable symptoms. The results of this study can provide a solid base for the development of joint aftercare programs. To avoid bias due to tumor entities, the present study focuses on the most common childhood cancer entities, leukemia, and brain tumors.
To gain new insights into the problems described above, the objectives of this study are: 1. To identify emotional and behavioral problems of pediatric cancer survivors and their siblings from the parental and from the children's perspective 2. To compare the emotional and behavioral problems of survivors, siblings, and a norm sample.
3. To assess concordance and discrepancies between parental proxy-reports and child self-reports.

| METHODS
This study presents a secondary analysis of data from a prospective mixed-methods study, investigating the reintegration of children with brain tumors and leukemia after family-oriented rehabilitation. The analyses of this paper are based on the baseline assessment. 15  complete a self-report. Sufficient knowledge of German and a signed consent form were required. Participants who were considered too stressed by the treatment or had cognitive deficits were excluded.
The study registries identified 662 families who were potentially eligible to participate in the study. Of these, 190 families signed the informed consent form to be contacted. 135 families subsequently participated, resulting in an initial rate of approximately 20%. Reasons for non-participation could not be recorded. After participation, no signed consent forms were available for two families, and one could not correctly complete the questionnaire due to a lack of language skills. Thus, data from 132 families from the study registers were available. 237 families were located at the Bad Oexen rehabilitation clinic who were suitable for potential study participation. 177 families participated in the study, resulting in an initial rate of 75%. Reasons

| Measures
Data were collected using paper-pencil questionnaires. Medical information, including diagnosis and time since the end of treatment were obtained through either parent-reports (study registries) or the physicians (rehabilitation clinic), depending on the recruitment path.
Socio-demographic data were collected via self-developed items. The Strengths and Difficulties Questionnaire (SDQ) was used to assess emotional and behavioral problems.

| Strengths and Difficulties Questionnaire
The SDQ was developed by Goodman 16 to assess emotional and behavioral problems as well as prosocial behavior in children and adolescents. The German version was published by Klasen et al. 17 The parent proxy-report and self-report versions for children between 11 and 17 years of age were used for this study. The answers were collected through a three-point Likert scale. Scores of 0-10 can be calculated for the five subscales (emotional symptoms, conduct problems, hyperactivity, peer relationship problems, prosocial behavior) and a total problem score of 0-40. Cut-off values can be used to classify each scale into the categories "normal", "borderline", and "abnormal". 17 Categories were chosen so that in the representative population sample of Klasen et.al 10% are classified as "borderline" and 10% as "abnormal". 17 The SDQ and instruments with clinical categories such as the CBCL/YSR were equally able to capture behavioral and emotional problems. 18 We decided to use the SDQ as it is shorter and includes a positive behavior scale (prosocial behavior). Both, the self-report and the parent proxy-report versions of the German SDQ have sufficient reliability and validity. 17,19  To identify the emotional and behavioral problems from the parental perspective, the parental questionnaires were evaluated. To avoid duplications one parent proxy-report per family was randomly selected. In 212 families, parents additionally provided information on at least one sibling. If proxy-reports on multiple siblings per family were available, one was randomly selected. This resulted in 309 proxy survivor reports and 212 proxy sibling reports.

| Statistical methods
For the comparison of the parental perspective with the children's perspective, the data of families were used in which the child and at least one parent participated. Again, if both parents participated, one was randomly selected. The level of concordance between child and parent reports was estimated using weighted Cohen's kappa coefficient. The weighted kappa coefficient accounts for different degrees of concordance on ordinal scales. A difference between "normal" and "borderline" is weighted less than a difference between "normal" and "abnormal". The level of concordance can be classified using the ranges and labels set by Regier 22  extent parents rate their children as more or less burdened than children do themselves. To get a better overview of the direction, the following categories of frequency were formed: "parents = child", "parents > child" and "parents < child" based on the cut-off categories of the SDQ. This method cannot account for the weights to the same extent, but provides supplementary content information.

| Children's self-report of strengths and difficulties
Cancer survivors and their siblings did not significantly differ concerning their self-reported emotional and behavioral problems.
Compared to the normative sample survivor and siblings reported significantly higher emotional problems as well as more prosocial behavior (Table 1). In relation to emotional problems, the reports of 7% of the cancer survivors and 13% of the siblings fell into the "borderline" and 10% of each cancer survivors and siblings in the "abnormal" category.

| Parental proxy-report of strengths and difficulties
Overall, parents rated the cancer survivors as significantly more burdened than their healthy siblings. In the parental assessment, survivors received a mean total score of 10.9 (SD = 5.6; 18% each categorized as "borderline", and as "abnormal") while siblings received a mean total score of 9.6 (SD = 5.7; 12% categorized as "borderline" and 15% as "abnormal"). Further, survivors were rated significantly higher than siblings regarding peer relationship problems. According to the parents' ratings, survivors were found to develop significantly more emotional problems, hyperactivity, and peer relationship problems than the norm sample. However, when comparing the siblings to the norm sample, only a higher level of emotional problems was observed in siblings (Table 2).

T A B L E 1
Mean score values of self-reported emotional and behavioral problems in cancer survivors (n = 72), siblings (n = 68), and a norm sample (n = 140).

| Concordance between parent proxy-report and survivor's self-report
The results show significant concordances between parental assessments and survivor's self-report across all scales. The concordance between parents and cancer survivor ranges between "questionable" (prosocial behavior k = 0.215; emotional problems k = 0.270, hyperactivity k = 0.362, conduct problems k = 380) and "good" (peer relationship problems k = 0.476 and total problem score k = 0.422).
However, the direction of non-concordant answers is inconsistent.
The highest concordance of answers was achieved for prosocial behavior and hyperactivity (both 84%) and the lowest for emotional problems (52%). Concerning emotional problems, in 20% of the cases, the parents reported higher scores than the survivors themselves, while in 28% of the cases, survivors rated their emotional problems higher than their parents (Figure 1).

| Concordance between parental report and sibling self-report
The interrater reliability between parents and siblings was significant

F I G U R E 1
Distribution of (non-) concordant assessments of parents and survivors regarding the cut-off categories of the SDQ. Percentage of concordance (cancer survivors = parent: CS = P) and non-concordance (cancer survivors < parent: CS < P and cancer survivors > parent: CS > P) between self-reports (n = 72) and parent proxyreports (n = 72) in total (TPS) and on the subscales emotional problems (EP), conduct problems (CP), hyperactivity (HA), peer relationship problems (PRP) and prosocial behavior (PSB).

1252
-PAUL ET AL. k = 0.304; total problem score k = 0.437). Concerning the direction of the deviations the highest exact concordances of answers were achieved for prosocial behavior (82%), hyperactivity, and conduct problems (each 79%), and the lowest for peer relationship problems (65%) and emotional problems (70%). In 16% of cases, parents rated peer relationship problems higher than siblings themselves. In 21% of the cases, siblings classified their emotional problems higher than their parents (Figure 2). These findings contradict some long-term results reported in the literature, suggesting that minor cancer survivors, especially after brain tumor diseases, are at increased risk of developing behavioral problems in the long term. 29 As our sample was recruited a few months after the end of treatment, it is conceivable that potential behavioral consequences may not be apparent at an early stage.

| DISCUSSION
Hence, this could be an opportune time for interventions to focus on strengthening resources rather than waiting for children to exhibit

| Concordance between parents' perspectives and the children's perspectives
A lack of concordance between child self-report and parent proxyreport has been well documented in the literature. 11 As mostly parents are the stakeholders regarding children's medical and psychosocial support, information on the concordance of children's and parental perspectives is important. The evaluation of this study suggests that parents are quite capable of recognizing their children's F I G U R E 2 Distribution of (non-) concordant assessments of parents and siblings regarding the cut off categories of the SDQ. Percentage of agreement (sibling = parent: S = P) and disagreement (sibling < parent: S < P) and (sibling > parent: S > P) between self-report and parent proxy-report on the scales of total problem score (TPS), emotional problems (EP), conduct problems (CP), hyperactivity (HA), peer relationship problems (PRP) and prosocial behavior (PSB).
burdens. Behavioral problems (e.g. hyperactivity, conduct problems) indicate a higher concordance than internal problems. A possible explanation could be a lack of verbal communication between parents and children. A recent systematic review indicates higher concordances between parent and child self-reports, typically for more observable constructs (i.e., physical functioning vs. social-emotional constructs). 14

| Study limitations
Some limitations should be considered when interpreting the results.

| Clinical implications
As the challenges of parenting a child with cancer might go beyond in essence adequate parental capacities, corresponding psychosocial care services should be offered proactively by health care staff. Both cancer survivors and their siblings should have access to these support services even after the end of intensive treatment. Parent assessment is an important tool to evaluate children's distress and needs. Especially for very young children, children with severe illnesses, and children who are cognitively disabled, the proxy-report is indispensable. However, considering the nonconformity found between parents and children, the parent proxy-report should not be used exclusively. Consideration should be given to a parentindependent screening to assess the support needs of both the cancer survivors and their healthy siblings.

| CONCLUSION
According to self-reports, survivors and siblings have similar emotional burdens, while parents perceived survivors as more supervision; validation; writing -review and editing.