Oral medicine acceptance in infants and toddlers: measurement properties of the caregiver-administered Children’s acceptance tool (CareCAT)

The CareCAT tool enables longitudinal measurements of behavioral responses during a treatment for up to 4 weeks. It is a single page diary with a 5-point nominal scale, which provides one descriptor of positive acceptance behavior; ‘swallows well’ and four descriptors of negative acceptance behavior, i.e. ‘refusal’, ‘spitting’, ‘vomiting’, and ‘medication not taken’. ‘Swallows well’ characterizes that the child received and retained the oral medicine. A child’s ‘refusal’ refers to behavior hampering the medicine reception partly or as for ‘medication not taken’ completely when a child is totally unwilling. The descriptors ‘spitting up’ and ‘vomiting’ are options for behaviors indicating that the medicine has been received but not completely been retained. Each of these descriptors is shown along with a pictogram representing it (Fig. 1). After each medicine administration the user reports observed behavior by ticking boxes that correspond to the point in time when the medicine was administered as well as to the relevant descriptor(s). As they are not mutually exclusive, descriptors (one or several) are chosen that best represent the child’s behavior. A completed diary may be interpreted by tallying the scores and generate proportions of how often one or several descriptors were reported out of all events of medicine administration. The scores can provide information on a child’s acceptance behavior and also indirectly inform about the number of potentially incomplete ingestions.

The CareCAT is intended to be utilized in clinical practice when a detailed picture of behavioral challenges during a child’s treatment period is warranted. In a clinical trial, it might allow to estimate the general acceptance of a medication in a population of children of different ages. The CareCAT has been designed to be administered by caregivers in their home environment but can also be used by health professionals. Informants require basic literacy and numeracy skills for the tool to be correctly used.

This cross-sectional study examines inter-rater and intra-rater reliability of the CareCAT tool when used by caregivers of children under the age of 5 years receiving oral antibiotics. It also explores the users’ understanding of the tool descriptors, an aspect of content validity.

Data were collected between April and June 2016. As the first step, we obtained informed consent, after which we gathered information about the child’s medication and age, as well as the caregiver’s age, language, education and socio-economic status. The participant was then introduced to the tool through a standardized protocol aimed at minimizing bias. Detailed explanations were given about the five descriptors and the time-structure of the tool, after which the reporting process had to be practiced using six given real-life examples (Additional file 1). This procedure was repeated until all these examples had been accurately scored. Subsequently, the caregiver administered one dose of the oral antibiotic to the child, which was video-recorded using a smart phone. The caregiver and the external observer(s) independently scored their observations on the CareCAT tool. To minimize bias, caregivers were not made aware that their scores would be compared with those of the observers. The scoring was done individually without any communication between the assessors. After having scored on the tool, we asked caregivers to describe in their own words what they had seen their child doing, and how they would explain each descriptor. The three observers were also requested to score a sequence of video clips of children receiving oral medicine that had been recorded during the study. The videos were shown in two rounds, first in a systematic consecutive order, and second in a randomized order. To minimize recall bias at least five days had to pass between the day of administration and the first video-view, and at least 3 days between the first and second video-view.

General participant information, CareCAT scores of the different users, as well as the caregivers’ explanation of the tool descriptors were captured and quality checked using EpiData Entry software 3.1. Interviews were tape-recorded, consecutively transcribed in Xhosa or English, and the Xhosa transcriptions subsequently translated into English. After reading through the complete transcripts, we captured a summary of the individual explanations of the tool descriptors.

The measurement properties of the CareCAT tool were examined by: 1) exploring similarities and differences in scoring patterns among the different users, 2) assessing the agreement of scores between different users (inter-rater) and the reproducibility of scores by the observers (intra-rater), and 3) evaluating the caregivers’ understanding of each tool descriptor by reviewing their individual explanations. Scoring patterns of caregivers, observers and tool developer were reported descriptively. The user’s scores were analyzed in 2 groups, first as raw scores called ‘detailed scores’, and second after having divided them into 2 categories called ‘dichotomized scores’. The dichotomization was done firstly with regards to the child’s ‘acceptance behavior’ (positive/negative), which was considered positive if a child’s behavior was scored solely ‘swallows well’; and negative for all other scoring categories. We secondly categorized the scores focusing on whether the oral medicine had been received and retained by the child entirely or not, here called ‘medicine ingestion’ (complete/incomplete). We assumed that scoring ‘refusal’ and ‘swallows well’ in combination represented a child showing dislike, but still swallowing the medicine while the combination of ‘swallows well’ with ‘spitting up’ and/or ‘vomiting’ indicated some loss of medicine. Therefore, a medicine ingestion was considered complete if scored ‘swallows well’ alone or in combination with ‘refusal’, and incomplete for all other scoring categories. To assess the tool’s reliability, we cross-tabulated scores and calculated percentage agreement and unweighted Cohen’s kappa (κ) coefficients of agreement (with 95% confidence intervals, 95% CI) for nominal scales, using statistical software package of SPSS 23 and Microsoft Excel. While percent agreement represents the proportion of scores classified into the same categories by either two users or replicate observations of the same user, Kappa statistic measures the frequency of exact agreement while discounting ‘the proportion of agreement expected by chance alone’ [21, 22]. We further categorized Kappa values according to Landis and Koch’s criteria: as ‘poor’ if less than 0.2, ‘fair’ if between 0.21 and 0.4, ‘moderate’ if between 0.41 and 0.61, and ‘strong’ if above 0.61 [23]. Cross-tabulations were used to compare the agreement of scores between the users of the tool. The caregivers’ scores refer to multiple caregivers of which each solely scored their own child. The observer’s score refers to scores of three different observers, of whom only the score of the observer who introduced and instructed the caregiver was used. The summarized explanations of each tool descriptor given by caregivers were first coded; later codes were merged to main categories of themes that were displayed in a table, with examples of participants’ quotes.

The study was approved by the University of Western Cape’s Faculty Research Committee and the City of Cape Town, South Africa. Participants gave written consent for their participation. A separate consent form was used for the video-recording of the participant’s child receiving the medicine. Participants were informed that the material would be reviewed and the children’s behavior scored within the study team to determine whether CareCAT is used always in the same way.

The scoring of caregivers, observers and the tool developer resulted in a total of 12 scoring categories (Table 2), with the scoring pattern of the caregivers being slightly different from that of the observers and the tool developer. While the caregivers predominantly chose to report their observation in the form of a single score (88%), this was less frequent among the observers and the tool developer (58% and 62%, respectively). For example, caregivers reported ‘refusal’ alone in 11% of the administrations; but ‘refusal’ and ‘swallows well’ for 5%. In contrast, the observers reported ‘refusal’ alone in 1% and ‘refusal’ in combination with ‘swallows well’ in 16%. ‘Vomiting’ and ‘medication not taken’ were rare and usually reported in combination with ‘spitting’ and/or ‘refusal’ (Table 2).

After dichotomizing the scores into positive or negative acceptance behavior, the CareCAT tool’s intra- and inter-rater agreement proved strong, irrespective of who was the user (Table 3). Importantly, 15% of the caregivers’ scores categorized as positive acceptance behavior were categorized as negative by the observers. The opposite – acceptance behavior categorized as positive by the observer’s but negative by the caregiver’s score – was rare (1%).

When the scores were dichotomized on the basis of complete or incomplete medicine ingestion, the inter-rater agreement between the caregivers and observers was moderate (Table 3). There were a few occasions in which the ingestion of medicine was categorized as incomplete by the caregivers’ scores, but as complete by observers, and vice-versa (8 and 10%). Inter-rater agreement between observers and tool developer was strong, whereas the intra-rater agreement of the observers varied slightly between moderate and strong.

Examples of the caregivers’ explanations of the tool descriptors are presented in Table 4. The descriptor ‘swallows well’ was explained as the actual act of swallowing, as exemplified by one caregiver stating that her child would drink and swallow the medicine well similar to how the child usually eats. On the other hand, many caregivers referred to ‘swallows well’ as the absence of negative child behavior. A caregiver reported, for example, that the child would ‘not give me a hard time when she drinks the medicine’. The explanation of ‘refusal’ entailed physical action such as aiming to prevent - but also to reverse - the intake of medicine by spitting and induced vomiting. One caregiver reported that her child ‘…would beat the spoon with the medicine, close her mouth, move her body and cry’. While most caregivers generally defined ‘spitting’ or ‘vomiting’ correctly, few also mentioned the loss of medicine through ‘spilling’ or ‘overflow’. ‘Medication not taken’ was described as actions resulting in no intake of medicine.

Expecting acceptance data to be biased, e.g. by informants reporting in a socially desirable manner, we chose tool descriptors that would avert the focus of caregivers from feeling assessed in their ability to administer a medicine to paying attention to the actual behavior of the child. The tendency to put themselves, the child or the medicine into a favorable light might have affected caregivers’ and observers’ scores differently [34, 35]. It might have lead caregivers to report negative behavior with a higher threshold. A different scoring pattern of the observers with their tendency to report negative behavior in parallel with ‘swallows well’ might show their focus on the medicine intake, no matter if the child displayed negative behavior or not. Rephrasing the descriptor as ‘swallows well’ with ‘neither refusal, spitting nor vomiting observed’ is one possibility in controlling this element of reporting bias. Furthermore, the tool could provide two scales – one to report completeness of ingestion and another for child acceptance behavior. However, adjusting the tool might be at the expense of the tool’s simplicity, which consequently might require higher literacy and numeracy levels of the user population and more instructions to the user. We consider it a strength that the tool with its current design is not restricted to being used by health professionals only; indeed, the results show that caregivers with different educational level enrolled here could self-administer the tool.

Our approach to dichotomize the CareCAT scores separately for acceptance behavior or completeness of medicine ingestion, intentionally deviates from others [20]. We believe that focusing only on the completeness of medicine ingestion, irrespective of child’s behavior, might also show a tendency to report the acceptance of a medication in a desirable manner.

This study has several strengths: first, by focusing only on the age group of infants and toddlers, detection of observations typical for this group increased [36]. Second, by testing validity and reliability, we were addressing the lack of non-validated tools in this field. Furthermore, the tool was tested in the key population, for which it was developed in a real-life setting. By enrolling caregivers who were casual attendees at a healthcare facility, we have demonstrated the tool’s use in assessing asymptomatic as well as sick children when receiving antibiotics prescribed in practice. With a sample size of more than 100 rated observations, it also fulfilled by far the minimum requirement of participants in reliability studies [19].

While this study focused on a relevant knowledge gap in the field of the use of medicines in pediatrics, it has some limitations. It took place at a healthcare facility and not in the natural home environment, which we believe could have been perceived as rather intrusive. Future research should involve the tool’s implementation at a caregiver’s home with completion on multiple occasions. Another limitation is that the diversity achieved in response was not as homogeneously distributed as desired. This is related to the frequency of behaviors, such as vomiting, which occurred rarely, as reported by others [20, 37], and particularly depends on the palatability of the medicine.