Children undergoing medical procedures, anaesthesia, and surgery experience significant psychological and physiological reactions. The unfamiliar environment, the equipment and routines, fear of separation, needles, and the medical procedure are well documented as sources of these negative reactions (27–29). These reactions lead to short- and long-term maladaptive behaviours such as irritation, aggression, incoherency, uncooperativeness, eating problems, and sleep disturbances (4) and fear of healthcare professionals or medical treatment (8, 9). In addition, they are associated with poor anaesthesia induction compliance (IC) (30), emergence delirium (5), increased need for sedation or rescue analgesia (31), and prolonged pain and recovery (5). To address these psychological and physiological reactions, research has been undertaken on the use of pharmacological and non-pharmacological interventions to reduce pre-operative anxiety.

The OVID databases that were selected were MEDLINE, EMBASE, PsycINFO, and HMIC. A mix of keywords and Medical Subject Headings (MESHs) was used to search for themes. The search was carried out in February 2022 using the complete syntax with truncation for each database as outlined in Appendix B. Limitations were added for the period (2000–2022), English language, and age.

The preliminary search returned 931 records; 363 duplicate records were identified, and 176 records were removed. A total of 730 records remained, and these progressed to the stage of title and abstract screening (57). Two reviewers screened titles and abstracts for the 730 records for eligibility against the PICOS, resulting in 655 articles for exclusion, 41 articles for stage full-text screening, and 34 conflicts. After consultation with a third reviewer, 17 (58–74) articles remained for full paper review. The Cohen's Kappa score (75) for the screened title and abstract was 0.682, with a 95% proportionate agreement, and for the full paper review, a score of 0.907 was obtained, with a 96% proportionate agreement, demonstrating substantial agreement among the reviewers. Figure 1 outlines the searching and screening process diagrammatically using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) (76) flow chart. Appendix C in the Supplementary material shows the full-text screening selection process questions.

Data relevant for extraction were considered against the aims and objectives of the review (77). Supplementary Table S2 in Appendix D sets out the data extraction fields. For any randomised control trials, version 2 of the Cochrane Risk of Bias tool for randomised trials (RoB2) (78) was chosen, given that it is the standard recommended for Cochrane Reviews. For any non-randomised control trials or included quasi-experimental study designs, the Critical Appraisal Skills Programme (CASP) (79) was chosen, given its wide use in systematic assessment of the relevance and results of research.

The synthesis and analysis were first assessed, on the basis of the degree of homogeneity (80, 81), in terms of four aspects: patient characteristics, the intervention and comparators of the studies, the reported outcomes and timeframes over which they were measured, and the similarity of the results. If homogeneity is determined to be insignificant and heterogeneity significant, then a narrative synthesis would be undertaken on the study and participant characteristics, findings from the quality assessment, and the measurements used and reported outcomes. To meet the primary objective of this review, an evaluation of the development (design) of the DHIs was also undertaken. The results were then used to determine any correlation to the evaluation of the DHI development and findings from the studies using a measure of effect. DHI descriptions were evaluated using the information provided within the relevant studies, and where this was insufficient, related articles were sought out. For some studies, no related information was available, and the DHIs were, therefore, assessed using only the information provided in the included paper.

The digital health interventions in the studies are aimed at changing behaviour to reduce pre-operative anxiety through education, information, and coping strategies. The TDF was chosen to evaluate the design and development of the digital health interventions within the context of behavioural science, as it is a validated tool for assessing implementation issues, supporting intervention design, and analysing interventions (24, 26, 82). The DHI evaluation was undertaken using a scoring system against a 16-domain framework. The 16 domains constituted the 14 domains from the TDF (24) and two additional domains. The definitions of the 14 domains from the TDF were adapted from Cane et al. (24) and Smalley et al. (82) with two additional domains added. The additional domains identified as relevant in assessing the development of the DHIs, and added to create a modified TDF, were 1.

input from one or more healthcare professionals, children, and parents, and

For each domain in the modified TDF, the domain descriptions were used to develop a criteria checklist to guide the evaluation of the DHIs. The criteria checklist considered what information, activities, techniques, or actions the DHIs should incorporate to meet the domain descriptions. This was tested against a sample of the DHIs to refine the criteria checklist. Each DHI was then assessed against each domain criteria checklist and a score applied depending on whether the DHI fully met, partially met, or did not meet the requirements in the criteria checklist. Table 1 sets out the criteria checklist used to evaluate the DHIs against the modified 15-domain TDF. The scoring system applied to the 13 domains from the TDF was “1” if the DHI fully met the criteria, “0.5” if the DHI partially met the criteria, and “0” if either the DHI did not meet the criteria or insufficient information was provided. The scoring for the co-production domain (described in Table 1 as “input into the development of DHI”) was “1” if the paper evidenced development involved healthcare professionals, children, and parents, “0.5” if the paper evidence development only involved one or two of these groups, and “0” if the paper did not evidence involvement from these groups. The scoring applied to the domain for use of behavioural and/or design frameworks in DHI development was “1” if the paper explicitly evidenced their use and “0” if the paper did not evidence their use. The scores were summed to provide an overall score out of 15 for each of the DHIs in the included studies, with those scoring higher assumed to have optimal design and development through meeting more of the modified TDF domains. The scores were also summed to provide totals on how many of the DHIs scored fully (given a score of 1) or partially (given a score of 0.5) against each domain. These scores were then used to determine any correlation between the DHI designs and health outcomes.

To determine any correlation between the evaluation of the development of the DHIs and the reported outcomes, quantitative data was converted into a summary statistic. Specifically, this examined what outcomes were measured and how, whether there was a noticeable measure of effect, and how it correlated to the scoring from the DHI evaluation. To ensure that the data analysis met the requirement of systematic review transparency, established reporting guidelines were followed (83).

The effect size measure and direction was calculated where feasible using a standardised mean difference, Cohen's d, Glass's delta, and Hedges’ g (84), or other appropriate statistical calculations such as a Chi-square p-value calculation (85). Where data are presented in studies using median and interquartile range (IQR) and where there is no evidence of significantly skewed data, median and IQR was converted to an estimated mean and standard deviation (SD), using an online calculator (86) developed from research by Wan et al. (87), Lou et al. (88), and Shi et al. (89, 90). Where estimate mean and SD can be derived, the results were used to calculate the effect size. Similarly, where the mean is provided but not SD, SD was calculated using the RevMan Calculator (91), with the subsequent effect size also calculated. Table 2 outlines the scoring criteria to determine the direction of the effect.

The studies were mostly conducted in developed countries, with three in the United Kingdom (58, 65, 72), two in Canada (59, 62), four in South Korea (60, 61, 63, 70), and one each in the United States (71), Thailand (64), Portugal (67), Turkey (66), the Netherlands (68), Italy (69), and Japan (73). The study by Dehghan et al. (74) was conducted in Iran. Study durations varied, with six studies being conducted over 8 months or less, eight being between 10 and 18 months, two at 20 and 23 months, respectively, and one not stating the duration. All DHIs were utilised pre-operatively. The length of the DHIs ranged from 344 s (66) to a maximum of 45 min (59), with four studies (58, 64, 65, 72) not stating the length and the rest being between 4 and 15 min.

The DHIs trialled in the studies are divided into four main types—VR (59–61, 63, 68, 70, 74), audio-visual presentations (64, 66, 69, 73), web-based programs or presentations (62, 65, 71, 72), and educational interactive multi-media applications (58, 67). All DHIs incorporated a tour or information, in varying levels of detail, about the hospital environment and equipment, but only 11 studies (58–61, 63, 66–68, 70, 71, 73) explicitly stated that the information included details of the staff involved. Of the seven studies using a VR-based DHI (59–61, 63, 68, 70, 74), Stunden et al. (59), Eijlers et al. (68), and Ryu et al. (70) incorporated interactive elements, with the rest being informational video tours. The DHIs by Bray et al. (58), Wright et al. (62), Wantanakorn et al. (64), Fernandes et al. (67), and Fortier et al. (71) also incorporated interactive elements such as games and chatbots.

Except for five studies (63, 67, 68, 73, 74), all other studies used usual care in the control group, and this comprised standard verbal information and/or information leaflets. Of those studies using usual care, four were three-arm parallel randomised control trials and involved a comparator, and these were a Child Life Program (CLP) (59), handwashing game (65), voice recording (66), and cartoon strip (72). Park et al. (63) used the same video tour for the control group but without the mirror display for parents to watch simultaneously as their child as used for the intervention. Fernandes et al. (67) used a video game as a comparator and no intervention as the control. Eijlers et al. (68) and Wakimizu et al. (73) used audio-visual tour/information as the control, with the latter being the same as for the DHI intervention group but only viewed once a week in advance of the procedure. Dehghan et al. (74) used parental presence as the control.

The setting for the studies was linked to where the intervention DHIs were used. The majority were used once in the hospital either on the day before the procedure (64, 69) or on the same day as the procedure (59–61, 63, 67, 68, 70, 72), with four of the same-day DHIs being one hour pre-operatively. Hatipoglu et al. (66) presented the DHI once, 1 week in advance of the procedure during hospital admission. The DHIs for the rest of the studies were used either at home (58, 62, 71) or both at home and in the hospital (65, 73), but for all five of these studies, the DHIs could be accessed by children and parents more than once. For the studies where the DHIs could be used at home, one (71) was made available a week before and up to 7 days after the procedure, three (62, 65, 73) were made available a week before the procedure, and one (58) 3 days before the procedure. It is unclear in the Dehghan et al. (74) study when the DHI was used relevant to the procedure, but it is assumed that the setting was in hospital post the randomisation of participants.

The total sample size across the 17 studies was 1,726 children, with sample sizes ranging between 40 and 200. The ages of the children ranged between 2 and 14 years, with three studies (65, 71, 73) including only younger children between the ages of 2 and 7 years. The reporting of sex across the studies was not consistent, with seven studies (59–62, 65, 68, 70) reporting the sex breakdown of only those included in the analysis and the rest reporting the sex breakdown of the children randomised. In total, of the sex breakdown reported, there were 980 males and 718 females. The only studies to report on child ethnicity were Wright et al. (62) and Fortier et al. (71). Eight studies (58, 59, 62, 64, 66, 67, 69, 73) included baseline information on the number of previous surgeries and/or hospitalisations by the children.

Inclusion criteria across all 17 studies were children within the studies specified age range, undergoing the relevant included procedures and without any cognitive impairments. Children were explicitly excluded from 11 studies (59–64, 67, 68, 70, 71, 73) with visual and/or developmental and/or auditory delays. Language was an exclusion in eight studies, with Stunden et al. (59), Wright et al. (62), Fortier et al. (71), and Campbell et al. (72) limited to English, Fernandes et al. (67) limited to Portuguese, Eijlers et al. (68) limited to Dutch, Liguori et al. (69) limited to Italian, and Wakimizu et al. (73) limited to Japanese. A history of seizures or epilepsy was an exclusion criterion in six (59–61, 63, 68, 70) of the seven VR DHIs, with Dehghan et al. (74) stating the only exclusion as “stress or special problems in using eyeglass or headphone in [virtual reality exposure therapy]” (p. 3).

Parents were included in 10 studies (58, 59, 62, 63, 65, 67, 68, 70, 71, 73). Six studies (58, 65–68, 71) reported baseline information on the educational socioeconomic status of the child's parents. Fortier et al. (71) also included information on parental income. Parental age was reported in seven studies (62, 65–67, 69, 71, 73) and parental ethnicity was reported only by Wright et al. (62).

The procedures that the children were undergoing across the studies were mostly for surgery, including elective and ambulatory surgery (60–63, 66–71, 74). The types of surgery differed across the studies, but the most noted were otolaryngology; ophthalmic; orthopaedic; dental; ear, nose, and throat (ENT); urology; herniorrhaphy; and tonsillectomy. The other procedures included tooth extractions (65, 72), magnetic resonance imaging (MRI) (59), and bone marrow aspirations (64). Bray et al. (58) included children undergoing both invasive (surgery, cannulation, and blood tests) and non-invasive procedures (x-ray or ultrasound). Wakimizu et al. (73) included only children undergoing herniorrhaphy.

There were 15 unique DHIs across the 17 included studies, with the same DHI used in three (60, 61, 63). The DHIs were scored against the 15 domains in the modified TDF, where 1, 0.5, or 0, respectively, meant that it either fully, partially, or did not demonstrate the domain. Supplementary Table S7 in Appendix F provides the results of the DHI assessment against the 15 domains in the modified TDF, while Table 3 offers a commentary for each.

None of the 15 domains was fully evidenced across all the DHIs, with 35% evidencing eight or more domains and 65% evidencing seven or fewer domains. The DHIs by Wright et al. (62) and Fortier et al. (71) fully evidenced the most domains, with 13 met in each. The first nine domains outlined in Table 3 were met in each of these three studies, with differences occurring in the remaining six domains, namely, intentions, goals, social influence, optimism, co-production, and use of a behaviour framework. Stunden et al. (59) scored the next highest fully evidencing 11 domains, meeting the first nine and those for intentions and goals. Ryu et al. (70) scored the next highest, fully evidencing 10 domains, with all but that for emotion in the first nine being met, as well as goals and optimism. Dehghan et al. (74) did not evidence any domains fully, and the DHIs used by Huntington et al. (65), Campbell et al. (72), and Wakimizu et al. (73) fully evidenced only one domain and two domains each, respectively. This was attributed to insufficient information, as opposed to simply not meeting the domain. The remaining DHIs varied, with between three and nine domains fully evidenced. On average, the DHIs fully met 5.4 domains with a standard deviation of 4.17 and partially met 2.5 domains with a standard deviation of 1.19.

All studies assessed the outcomes of the intervention, with these being self-reported by children and parents, observed by clinicians or researchers, or extracted from medical records. The primary and secondary outcomes included assessments across five categories: 1.

emotions and feelings,

The risk of bias across the 16 randomised control trials was generally concerning, with 68.8% having an overall result of some concern (60–63, 65–70, 73) and 31.1% an overall result of high risk (59, 64, 71, 72, 74). Risks were linked to the process for randomisation or the inability to confirm whether a pre-specified analysis plan was finalised before the results were unblinded for analysis. Figure 2 provides an overall summary of bias as a percentage for the six domains.

Figure 3 provides a breakdown of the risk of bias for each study against the six domains, namely randomisation process (D1), deviations from intended interventions (D2), missing outcomes data (D3), measurement of the outcomes (D4), selection of the reported results (D5), and overall bias.

All studies used a random group allocation sequence, with this being computerised in eight studies (60–63, 66, 67, 70–72). The method of randomisation varied in the rest of the studies, including drawing lots, using concealed envelopes, allocating on bed numbers, or using an allocation ratio. Randomisation process bias (D1) for seven studies (60–65, 67) was low, with this being attributed to confirmed allocation sequence concealment and no noted baseline differences among the groups to suggest problems. Conversely, seven studies (66, 68–70, 72–74) were determined as having some concern due to insufficient information on the allocation sequence concealment but no notable baseline differences among the groups. Dehghan et al. (74) provided insufficient information to determine whether baseline differences among the groups suggested a problem with the randomisation process. Due to the potential for allocation knowledge to influence participant bias, the studies by Stunden et al. (59) and Fortier et al. (71) were determined to have a high risk of randomisation process bias, as both confirmed that blinding to allocation was not possible.

Bias due to deviations from intended interventions (D2) was low across 50% of studies. Of the five studies considered to have some concern of bias in this domain, three (59, 62, 63) were attributed to insufficient information on deviations from intended intervention groups. Ryu et al. (61) had one deviation from the DHI group due to dizziness using the VR, although the child was not reassigned and was excluded from the analysis. Fernandes et al. (67) reassigned 15 children after randomisation because of ethical concerns over children sharing the same ward and being in different groups. The potential bias from this change in the group was deemed to be of some concern but not high risk, as participants were unaware of their group allocation until receiving the intervention. Wantanakorn et al. (64), Campbell et al. (72), and Dehghan et al. (74) were considered at a high risk of bias in this domain because of insufficient information to determine whether participants, carers, and people delivering the interventions were aware of group assignment, whether any deviations from the intended groups occurred, and whether an appropriate analysis was used to estimate the effect of assignment to intervention.

Bias due to missing outcome data (D3) was low across 88% of the studies and considered high for two studies. Ryu et al. (60) excluded three participants from analysis because of a data collection failure, and given the small sample size, it was considered that this could have impacted the outcomes, thus having a potentially high risk of bias. Dehghan et al. (74) provided insufficient information on whether data were available for all or nearly all participants, thus also having a higher risk of bias.

Bias for measurement of outcome (D4) was deemed low in 50% of studies (60, 61, 63, 65, 66, 68, 70, 71) as the same appropriate outcome measures among the groups were used and the outcome assessors were blinded. In contrast, 43.8% of studies either provided insufficient information to conclude whether the outcome assessors were blinded (67, 69, 73) or provided evidence to suggest that they were not blinded (59, 62, 64), resulting in some concern of bias. Campbell et al. (72) likewise had some concern of bias in this domain, but this was due to an inability to align the sample size in the result data, meaning insufficient information was provided to decide whether measurement or ascertainment of the outcome differed among the groups. Although the same appropriate measures were used for the outcomes among the groups in Dehghan et al. (74), insufficient information was provided to determine whether the outcome assessors were blinded. As a knowledge of group interventions could lead to bias, and it was not possible to determine whether it was likely that the outcomes could have been influenced by this knowledge, it was considered that this study was at a high risk of bias.

Most studies (68.8%) had some concern for bias in D5 “selection of the reported results”. This was due to an inability to confirm whether the outcome data were analysed following a finalised pre-specified analysis plan before unblinded outcome data were made available for analysis. This according to Cochrane RoB2 guidelines (94) means that there is an unclear risk for reporting bias. For 10 studies (62, 64–67, 69, 71–74), a trial protocol was not obtained, and although the studies generally set out the analysis plan, it was not viable to confirm whether it was finalised before unblinded analysis. Five studies had a low risk of bias in this domain, with four (60, 61, 63, 70) due to a finalised pre-specified analysis plan being reported in the trial protocol and one (68) due to the analysis plan being followed and the outcome assessors being blinded.

Medical trials entail a comprehensive understanding of clinical ethics, with those involving children complicated by stricter standards than those involving adults (95). In addition, paediatric medical trials entail a careful balancing of benefit against risk and a consideration of the evolving stages of a child's development and an informed parental, often family-centred, decision making (96). These stricter ethical standards and requirements, together with fewer eligible participants, result in paediatric medical trials being more challenging and less frequent (95, 97). The outcome is that paediatric medical trials are often not supported by class I evidence, having a higher probability of bias and lower external validity. These issues correlate with the studies included in this systematic review and the overall higher risk of bias.

All studies included in this review assessed child anxiety either as an emotion or as a feeling or behavioural response. Compared with children in the control group(s), 14 studies (82%) showed that children using the DHIs were associated with lower anxiety levels and the DHI had a positive impact, with this corresponding to the result of the effect size calculations where they could be calculated. This differed for three studies (17%), which showed anxiety levels were similar and the DHIs had no or little impact and effect. Given that higher pre-operative anxiety is a predictor of negative behavioural changes, the results for measures such as emergence delirium, induction behaviour, and induction compliance were mixed, although they were considered only in a small number of the included studies. For the three studies (60, 68, 71) that measured ED, only one found its occurrence lower in children prepared using the DHI. For the studies looking at induction behaviour (61, 65) and induction compliance (62, 63, 70), one study found improved induction behaviour and two found improved induction compliance in children using the DHI. Some of these health improvements are linked to higher scoring within the modified TDF and the first finding of this study.

The first finding is that paediatric preparation DHIs scoring higher against the modified TDF are more likely to be associated with reduced anxiety and reduced negative behavioural changes, as they will provide detailed information on the planned procedure and encompass information on coping with emotions, feelings, and anxiety (1, 13). Bray et al. (58), Stunden et al. (59), Wright et al. (62), Ryu et al. (70), and Fortier et al. (71) were the highest scoring studies against the modified TDF, having fully met 10 or more domains with 8 of these in common. The eight domains that were commonly met were knowledge, skills, behavioural regulation, environmental context and resources, belief about capabilities, beliefs about consequences, and reinforcement. This is attributed to the DHIs including (1) detailed information on the hospital environment, staff, equipment, and relevant procedure; (2) interactive elements such as games, quizzes, rewards, actions, or activities; and (3) breathing or coping exercises or modelling videos. The children using the DHIs in four of these studies were associated with lower anxiety levels (58, 62, 70, 71), lower occurrence of emergence delirium (71), and higher induction compliance (62, 70). This finding indicates that DHIs that incorporate these domains and are used as preparation interventions could be associated with reduced anxiety levels and other negative behavioural changes. An anomaly to this finding is the study by Stunden et al. (59). Stunden et al. (59) did not find any impact on child anxiety levels nor any difference in the key measure for head movement in their randomised control trial. This inconsistency is not likely to impact the first finding of this review for two reasons: (1) the trial was conducted with a simulated MRI and paid volunteers; and (2) all children reported no anxiety at baseline. This contrasts with the other four studies where the DHIs were used in preparation for real paediatric procedures, and varying levels of anxiety were reported at baseline. Nevertheless, the design of this DHI is considered relevant to the evaluation against the modified TDF. This finding cannot be extrapolated to all studies that reported positive health outcomes, given that the DHI scoring varied against the modified TDF. Despite this, the lack of meeting this finding can be attributed to either one or more of the remaining three findings, or insufficient information available in the study paper to make a judgement, thus resulting in a zero score.

The second finding is that preparation DHIs scoring higher against the modified TDF are more likely to have used co-production and a behavioural framework in their design and development. Aufegger et al. (21) stated that children and their parents prefer “easily digestible, non-medical explanations as to what to expect during the treatment process [together with information] on how to prepare”, whereas healthcare professionals suggest that information on policies, the hospital environment, staff roles and responsibilities, and patient flow timings are useful. In addition, Bray et al. (52) found that children valued coping strategy information as it enabled emotional self-regulation and provided more information about the procedure. Of the five DHIs scoring the highest against the modified TDF, three (58, 62, 71) fully met the co-production and behaviour framework domains. No information was provided in the papers by Stunden et al. (59) and Ryu et al. (70) to determine whether a behavioural framework was used, but both partially met the domain for co-production, having involved healthcare professionals in the DHI development. Fernandes et al. (67) used a behavioural framework and co-produced the DHI with healthcare professionals and children, with this DHI being the sixth highest scoring one. In the context of the findings from Aufegger et al. (21) and Bray et al. (52), the DHI in these studies all incorporated detailed information about the hospital environment, staff, equipment, and procedure, and the five highest scoring DHIs included interactive elements, coping strategies, or self-regulation feedback. The association between a higher modified TDF score and health outcomes is linked to the hypothesis in the primary objective of this review. Preparation DHIs that are co-designed and grounded in behavioural science can result in reduced pre-operative anxiety and improved health outcomes. However, further research is required to validate this finding.

The three higher scoring studies (58, 62, 71) that explicitly stated had used behavioural frameworks in designing and developing the DHIs were associated with lower levels of child anxiety, lower occurrence of emergence delirium, and higher induction compliance. In the context of theory-driven intervention design, execution, and reporting, behavioural frameworks such as the TDF offer an approach to understand and/or explain what influences the success of intervention implementation. Through understanding and explaining what influences will contribute to successful implementation, interventions aimed at changing behaviours can be designed and developed accordingly. Similarly, this study suggests that behavioural frameworks, such as the TDF, can be used to assess an intervention design and development in the context of implementation evaluation, thus supporting refinement of the intervention design.

The third finding is that the type of preparation DHI plays an important role in achieving a higher score against the modified TDF, with this being intrinsically linked to interactivity and rewards or achievements. In a previous qualitative study (17), children wanted preparation information that is easy to access, comprehensible, engaging, and child-friendly, as they believed that all of this will aid in the alleviation of their worries. This builds on the previous two findings, reiterating the value of interactive DHIs that incorporate games, quizzes, rewards, actions, or activities. Here again, the top six and the seventh highest scoring DHIs against the modified TDF were all interactive, being an educational multi-media App (58, 67), a VR-MRI App (59), a video App with games (64), a web-based program (62, 71), and a VR video game (70). An anomaly to this finding is the educational multi-media App by Huntington et al. (65) that scored very low against the modified TDF. However, this is due to the lack of information in the study paper to fully assess the DHI. The remaining DHIs were mostly non-interactive video tours, VR information, or web-based click-through presentations. Consequently, a correlation was further identified between the domain for optimism and the domains for skills, reinforcement, intentions, and goals. This was observed in three of the highest scoring DHIs by Stunden et al. (59), Ryu et al. (70), and Fortier et al. (71). All these scored fully or partially in the optimism, intention, and goal domains, and all fully scored in the reinforcement and skill domains. Stunden et al. (59) used interactive cues (skills) and real-time feedback on movement within the MRI stimulation (reinforcement and intention) to encourage stillness (goal), and when this was achieved, the child advanced to the next level (optimism). Ryu et al. (70) and Fortier et al. (71) used interactive games (skills) and breathing and coping exercises (reinforcement and intention) to advance through the steps or modules (goals), receiving health points and a completion certificate, respectively (optimism). Both these DHIs used interactive game elements to reinforce behaviour, such as chasing the germ monster after instructions in the recovery room and placing the anaesthesia mask on animals. These findings suggest that integrating interactive elements (skills and reinforcement) with feedback or rewards (intentions) could be used to drive certain actions or behavioural changes (goals) by creating the desire (optimism) to achieve the feedback or reward. Furthermore, for two of the studies, it is associated with improved outcomes. An irregularity to this correlation was the DHI by Campbell et al. (72). It failed to meet the reinforcement, intention, and optimism domains, but it partially met the skill and goal domains through the provision of a list of activities to prevent tooth decay at the end of the web-based presentation.

The fourth finding is that the timing and location of the preparation DHI lends itself to a higher score against the modified TDF. Three of the highest scoring DHIs, by Bray et al. (58), Wright et al. (62), and Fortier et al. (71), were provided for use at home by children and parents, as many times as they liked, between a week and 3 days before the procedure. These three DHIs were also associated with lower anxiety levels in the children using the DHI, and for two, lower occurrences of emergence delirium (71) and higher induction compliance, respectively (62). This suggests that the use of the DHI in the comfort of the child's own home, within a few or more days before the planned procedure, may contribute to reduced pre-operative anxiety and improved health outcomes.

This study's strength is that it evaluates the design and development of DHIs used in preparing children for hospital procedures, correlating this against effectiveness in improving outcomes. Previous systematic reviews (6, 19, 20, 22, 23, 98–100) have predominately focused on the type of health interventions used and their effectiveness in improving health outcomes and/or reducing pre-operative anxiety, stress, and pain. In addition, some of these reviews included non-digital health interventions (98–100) and those used for distraction (6, 100). This study has specifically evaluated DHIs used for preparation.

There are limitations to this study. The first and second limitations relate to the search strategy and data extraction. While the search strategy was considered comprehensive, it was limited to papers in English published within the last 22 years, with the period being to ensure the relevance of the studies. When snowballing references of included papers and previous systematic reviews, a few papers published before the year 2000 may have been relevant for inclusion.

The third was the inability to conduct a meta-analysis because of the presence of heterogeneity across the included studies. Consequently, effect sizes were calculated, but not all studies reported the mean and standard deviation. It was, therefore, necessary to convert the median and interquartile ranges into a mean and standard deviation to then calculate the effect size. However, due to insufficient information to determine proximity to a normal distribution, the results may potentially be skewed. Some data reported in the studies were not amendable to calculating the effect size, and for these studies, the results were only narratively synthesised.

Finally, the level of information contained within some of the study papers to describe the DHIs was minimal, with supporting resources not found. This was a factor in the inability to draw meaningful conclusions against many of the modified TDF domains.

The quality of the studies was predominately moderate, with five studies having an overall high risk of bias. However, when considering the individual risk of bias in each of the five domains, it generally ranged from low to some concern, with most of the concerns linked either to an uncertainty of, or to a confirmed lack of, blinding of participants or those assessing the data, or to a lack of information in the papers to make a judgement. This was within the domains for “randomisation process” and “selection of reported results”, with the latter predominately linked to uncertainty on whether the analysis plan was finalised before results were assessed and the trial protocol not being readily available to verify, rather than the results being biased.

It is considered that this study is the first to use an adapted version of the TDF to assess the design and development of DHIs used to prepare children for hospital procedures. The four key findings from this study suggest that the TDF can be used to analyse the effects of preparation DHIs, and by using theory-driven behavioural science, their design can be redressed accordingly to improve health outcomes. While these findings contribute to this field of study, further research is required to validate the findings. Furthermore, research is required to understand the developmental costs of these preparation DHIs and whether they are cost-effective against the traditional form of pre-operative preparation.