1. Introduction
The global burden of disease reports a great vulnerability in children under five years of age [
1]. The prioritisation of interventions targeting this age group has increased considerably in the last two decades [
2,
3,
4], and appropriate outcome measures need to be identified to monitor the impact of these interventions. Major regulatory bodies, including the United States (USA) Food and Drug Administration (FDA) and The United Kingdom (UK) National Institute for Health and Care Excellence (NICE) [
5], recommend that, in addition to other health or clinical outcomes in paediatric submissions, measurement of Health-Related Quality of Life (HRQoL) should be included. There has been a concomitant interest in measuring and valuing HRQoL in children under five years in recent years [
6,
7,
8].
There are many generic measures available to measure HRQoL in children younger than five years [
6,
9,
10]. The Pediatric Quality of Life Inventory (PedsQL) is one of the more commonly used generic health instruments which has versions available across childhood [
6,
9,
10]. It was developed from the World Health Organisation core health dimensions to be used across the paediatric population [
11,
12,
13]. There are versions available for infants (1–12 months), young children (2–4 years and 5–7 years), children (8–12 years) and adolescents (13–18 years). The PedsQL has been used to measure outcomes of children across a range of conditions and settings including developmental delays [
14,
15], toddlers with very low birth weight [
16], post burn injury [
17], different levels of physical activity [
18] and those exposed to political violence [
19]. The PedsQL scoring algorithm is a summation of the items included on the measure and reference can be made to general population or country data [
13]. One of the limitations of the PedsQL and other generic measures is that currently they do not have any societal preference-based scores for economic evaluations.
The development and testing of preference-based measures are of particular value in economic evaluations and decision making. Preference-based measures elicit a societal-preference-based score, which allows for the calculation of quality-adjusted life-years (QALYs). QALYs are quantified on a scale ranging from 0 (death) to 1 (full health) with equal intervals allowing for losses and gains to be aggregated [
20]. A review of NICE appraisals identified that there have been several appraisals submitted for review in this age group and it is anticipated that this will continue or increase in the future [
8]. The applications reviewed most often included adult preference-based measures and rarely consider the view of the child or their family [
8]. Further evidence is required on performance of HRQoL measurements, especially preference-based measures, in young children. This will ensure that regulatory bodies, scientists and clinicians can make an informed recommendation for inclusion of a HRQoL measure. To our knowledge, there are currently four preference-based measures available for children under five years: the Health-Related Quality of Life Utility Measure for Pre-School Children (HuPs) [
21,
22,
23,
24], the Infant Health-Related Quality of Life Instrument (IQI) [
25,
26,
27], the EQ-5D-Y Proxy [
28,
29] and The Toddler and Infant (TANDI) Health-Related Quality of Life measure [
30,
31,
32].
The HuPs was based on the Health Utilities Index (HUI) and is recommended for children aged 2.5–5 years. It was developed in Canada and Australia and includes 12 items including emotion, hearing, speech, ambulation, dexterity, learning and remembering, thinking and problem solving, pain, behaviour, general health, self-care and vision [
21]. The earlier version of the instrument has been used to measure the outcome of children admitted to intensive care as infants [
22,
23,
24], children with neuroblastoma [
33,
34,
35] and in a Dutch community sample [
36]. We were unable to identify a preference-based score for the measure. The IQI was developed in multi-national sample (China–Hong-Kong, UK, USA, New Zealand, Singapore) and is recommended for infants aged 1–12 months. The IQI has an associated preference-based scores. However, it is unclear if it is appropriate to extend the age range of the IQI to include toddlers or pre-school children, as it focuses on sleeping, feeding, breathing, stooling, mood, skin and interaction [
25,
26,
27].
The EQ-5D-Y Proxy version was developed by adapting the adult version of the EQ-5D-3L to include youth-friendly wording and examples for the five dimensions included in the adult version (mobility, self-care, usual activities, pain or discomfort and emotions). The measure is recommended for self-complete in children from age 8 years and proxy completion from 4–7 years [
29]. However, there is limited evidence of its performance as a proxy measure in young children [
37]. The international protocol for valuation of the EQ-5D-Y has been published [
38], and national value sets developed [
39,
40], but little is known about how the preference-based score will perform in children younger than the recommended age of 8 years.
The TANDI is a new measure developed for proxy completion for children aged 1–36 months [
30,
31] that is amenable to preference-based valuation. It was developed in South Africa on the basis of a review of the literature and stakeholder input, including international expert opinion. It is similar in presentation to the EQ-5D-Y Proxy, but dimensions were developed from the bottom up. The dimensions include movement, play, pain, relationships, communication and eating. The dimensions are norm-referenced by defining each as being ‘at an age-appropriate level’. The dimension of Pain includes a reference to observable pain behaviour of inconsolable crying, restless movement and grimacing. The norm referencing of the dimensions means that they may be applied across a broader age range, although initially only tested on young children. To explore the effect of extending the target age range, this study aims to test the feasibility and validity of the TANDI in children aged 3–4 years.
2. Materials and Methods
2.1. Participants
Children 3–4 years of age attending pre-schools and a tertiary paediatric hospital in Cape Town, South Africa were recruited. The tertiary paediatric hospital manages children in a 300-bed in-patient facility and in specialist paediatric out-patient clinics. The pre-schools included in this study accept typically developing children, some of whom may have minor health conditions. The pre-schools were from the same referral region as the hospital. The pre-schools were randomly selected from schools recommended by the department of education. The children’s HRQoL questionnaires were completed by their parent or caregivers.
Participants were recruited according to their birthday and included from the day they turned three years of age until the day before their fifth birthday. Caregivers who were able to complete English questionnaires were included, as some of the measures are not available in local South African languages. Children at the paediatric hospital were eligible for inclusion if their children had a known acute or chronic health condition. All children attending the selected pre-schools were eligible for inclusion. Children who were in the intensive care unit were excluded as they were considered medically unstable or critically ill and participation in the study would be distressing (
Figure 1).
The sample size was calculated to identify a difference in proportions of problems reported on the TANDI dimension scores, with a small effect size (0.4), between those with and without a health condition. A sample of 60 children was required in each group ensuring 90% power and a significance of 0.05.
2.2. Measures
2.2.1. Toddler and Infant (TANDI) Health-Related Quality of Life Measure
The TANDI was developed for children aged 1–36 months for proxy completion [
30,
31]. It consists of six dimensions including Movement, Play, Pain, Relationships, Communication and Eating. The dimensions are scored across three levels of severity (no problems, some problems, or a lot of problems), and general health is scored on a Visual Analogue Scale (VAS) from 0 (worst) to 100 (best). Problems on the TANDI are described, similarly to the EQ-5D instruments [
29], by a six-digit code. For example, the TANDI health state 111223 describes someone with no problems with Movement, no problems with Play, no problems with Pain, some problems with Relationships, some Communication and a lot of problems with Eating. The best health state described by the instrument is coded as 111111, describing ‘no problems’ in each of the dimensions. Thus, the TANDI has 729 (3
6) unique health states. The TANDI does not have a preference-based score; therefore, a level sum score (LSS), similar to that used on the EQ-5D, was used to describe the responses on the descriptive system where the level labels are treated as numeric data with the best possible score (1 + 1 + 1 + 1 + 1 + 1) = 6 and the most severe score is (3 + 3 + 3 + 3 + 3 + 3) = 18 [
41]. The TANDI was designed to be amenable to developing preference weights in the future.
2.2.2. EQ-5D-Y
The EQ-5D-Y Proxy version 1 is a youth-friendly instrument requiring the respondent to rate the child’s health from their own (proxy) perspective [
29]. The youth measure has five dimensions: Mobility, Looking after Myself, Usual Activities, Pain or Discomfort and Worried, Sad or Unhappy. Each of the dimensions is rated on a severity scale of ‘no’, ‘some’ or ‘a lot’ of problems. Proxy respondents further rate the child’s global health on a VAS between 0 and 100 (worst to best health) [
28,
42]. At the time of data analysis, there were two published EQ-5D-Y preference-based scores available for Slovenia [
39] and Japan [
40]. The Slovenian preference-based scores were used in this study.
2.3. Pediatric Quality of Life Inventory (PedsQL)
The PedsQL generic core scales includes proxy versions for toddlers aged 2–4 years [
43]. The PedsQL includes four dimensions: physical-functioning (8 items), emotional-functioning (5 items), social-functioning (5 items) and school-functioning (5-items). Each of the 23 items are scored on a Likert scale from 0 (never a problem) to 4 (almost always a problem). All item scores are reversed and converted to a scale between 0 and 100, i.e., 0 = 100, 1 = 75, 2 = 50, 3 = 25, 4 = 0. Scores for the dimensions are computed by the summation of item scores over the number of items. The total PedsQL score, which gives an overall HRQoL score, is similarly calculated by the sum the dimension scores divided by four (number of dimensions). A higher PedsQL score suggests a better HRQoL.
2.4. Procedure
Ethical approval was obtained from the University of Cape Town, Human Research Ethics Committee of the Faculty of Health Sciences (HREC: 825/2017) before the study commenced. Approval was gained from all of the relevant authorities. Children were either recruited from the children’s hospital on the day of their scheduled out-patient appointment or during admission to the in-patient facility. Healthy children were recruited from pre-schools through information flyers that were sent home to the parents/caregivers.
The researcher provided the parents/caregivers with a detailed explanation of the study and the caregivers who consented were included in the study. Caregivers completed the measures with paper and pencil on the same day. The treating clinician was further asked to categorise the child’s health state on that day as mild, moderate, or severe.
A research pack was sent home with children attending the pre-schools. This included an explanation of the study, informed consent and the HRQoL measures for completion (TANDI, PedsQL and EQ-5D-Y Proxy). The caregivers were requested to return the completed research packs to school within one week if they wished to participate.
3. Data Analyses and Management
Children with acute or chronic health conditions attending the paediatric hospital were collapsed into a single group, labelled health condition, and compared to healthy children recruited from the pre-schools. As the group of children with acute and chronic illness was heterogeneous, expected differences could not be hypothesised. However, it was anticipated that healthy children would report a better HRQoL.
The frequency of TANDI responses to each dimension were compared across the two groups with Chi-square (
x2) or Fisher exact statistics. The ceiling and floor effects were categorised as those who reported no problems (111111) or a lot of problems (333333) across all six TANDI dimensions. The proportion of reporting was compared between those with and without a health condition. The LSS score was calculated to summarise the TANDI dimension scores. The Slovenian value set, which ranges from −0.691 to 1.000, was used to calculate the index score of the EQ-5D-Y. The known-group validity was assessed for the TANDI median LSS, and the VAS score using the Mann–Whitney U-test. The discriminate validity of the TANDI was determined across those with a health condition categorised as mild, moderate, or severe by the Kruskal–Wallis H-Test. Spearman’s correlation coefficient was used to establish the concurrent validity of the PedsQL and EQ-5D-Y dimension responses. Pearson’s r was used to explore the concurrent validity between TANDI LSS and the VAS score, and EQ-5D-Y Proxy preference-based score and summary scores on the PedsQL. Correlation coefficients were interpreted according to Cohen: 0.1–0.29 low association, 0.3–0.49 moderate association and ≥0.5 high association [
44].
Caregivers were asked which of the questionnaires were best able to describe the health state of their child. The proportion of their responses was compared across those with a health condition and those without with Chi-square (x2).
5. Discussion
This was the first study to extend the age range of the TANDI to children older than 3 years. The inclusion of norm referencing of the dimensions was anticipated to make the TANDI suitable for children older than the original target population of children, younger than 3 years [
31]. This was found to be the case, as the instrument demonstrated feasibility and reliability in this group of children.
As no other studies reported the ceiling effect of the TANDI with reporting of problems across all dimensions (111111), comparison was made to the ceiling effect of the EQ-5D-Y in older children. The ceiling effect of healthy children and those with a health condition in this study was comparable to the EQ-5D-Y in children over eight in both the general Swedish population [
45] with cystic fibrosis and functional disability [
46,
47], respectively. At a dimension level, the TANDI showed a similarly high proportion of ‘no problems’ in healthy preschoolers and toddlers and infants [
31] compared to those with a health condition. As anticipated, the reporting of unique health profiles was higher in those with a health condition to those without, this affirms the TANDI’s ability to capture the difference in health condition and severity thereof [
46]. Furthermore, there is no concentration around select health profiles, which is advantageous for a wide distribution of values in the future, and the potential for measuring a change in health state is increased [
41].
The TANDI dimension of Relationships had a higher reporting of problems in the healthy group in this study whereas in younger children those with a health condition reported significantly more problems than the general population [
31]. This could be attributed to reference to interaction with family members which may be more appropriate in toddlers and infants than pre-school children. Furthermore, the dimension of Communication was significantly different between health groups in toddlers and infants but not in this study, owing to the large number of healthy children reporting ‘some problems’. However, those with a health condition reported a higher number of ‘a lot of problems’. At a composite level, the TANDI showed good known-group validity and was able to discriminate between the levels of severity of the health condition. Although the severity rating by the clinician was subjective there seemed to have been a broad agreement between the caregiver and clinician. The TANDI was able to discriminate between mild and severe, which suggests that a more granular instrument might not be able to discriminate between levels that are closer together (e.g., 5 levels). It is suggested that future research explore the discriminate validity in those with disease groups that are known to differ in severity.
The TANDI showed improved concurrent validity to the PedsQL than reported for the EQ-5D-Y Proxy [
37]. This could be attributed to the fact that the TANDI and PedsQL items are more similar and developmentally appropriate for children aged 3–4 years than the dimensions included in the EQ-5D-Y Proxy. In young children, the development of skills is integrated across areas of functioning resulting in more TANDI dimensions and PedsQL items showing an association than hypothesised. It was, however, evident that the PedsQL did not measure problems with Pain or Eating. Thus, for children with a health condition, where pain and/or eating is important it would be recommended to use the TANDI or EQ-5D-Y-Proxy.
It has been previously reported that the EQ-5D-Y Proxy dimensions did not perform as well in children aged 3–4 years as they did in older children [
28,
37]. This is not surprising as the measure was adapted from the existing adult version. The TANDI dimensions, in contrast, were identified based on input from stakeholders and literature on child development [
30]. The caregiver respondents in the current study reported that the TANDI was better able to describe their child’s health if they had a health condition. Thus, the TANDI would be recommended for use in children 3–4 years with a health condition. For studies that include infants and pre-school children or for longitudinal studies for infants and toddlers, the TANDI would be a suitable and valid instrument. In a study where pre-school children with a health condition are included with older school-going children or with a longitudinal component into school-going age, the EQ-5D-Y Proxy would be recommended. However, the poor dimension performance of looking after myself, due to the developmental age of the child, as previously reported should be anticipated [
37]. The PedsQL would be recommended for studies with typically developing children who report minor health impairments as its sub-scales do not have such a high ceiling effect and the caregiver respondents in this study considered it to describe the health of their child well.
Study Limitations
The healthy group attended a pre-school in the same referral region that the paediatric hospital serves to ensure that children from similar socio-economic circumstances were recruited; the sample was, however, not matched. The results appear to reflect the age group; however, they cannot be generalised to the Western Cape or South Africa, as no data on ethnicity, language, or socioeconomic status were collected for comparison to the general population of the Western Cape. The LSS used for the TANDI is a crude summary of dimensions and indicates the performance of the instrument in the absence of utility values. The limitation of the LSS includes that different health profiles may have the same score although the severity may differ [
48]; it assumes that the difference between levels of report is equal and that dimensions are equally weighted [
49]. The TANDI was designed to be amenable to developing preference weights and further research is needed to determine tariffs.