A longitudinal, observational study of the features of transitional healthcare associated with better outcomes for young people with long-term conditions

The study recruited 374 young people from across England and Northern Ireland, on the basis of having one of three conditions – 150 young people with type 1 diabetes mellitus (exemplar of chronic health condition with national standards of care, recruited through five NHS Trusts); 118 young people with autism spectrum disorder (ASD) and additional mental health problems (exemplar of neurodevelopmental disorder, recruited through four NHS Trusts); and 106 young people with cerebral palsy (CP) (exemplar of complex physical disability, recruited through two regional population registers and one NHS Trust). Young people were aged 14 to 18.9 years at recruitment, did not have significant learning disability, and had not transferred to adult healthcare. For each young person, a parent or carer was also invited to participate.

Recruitment was between June 2012 and October 2013. Local researchers visited the young people and parents, usually at home, took informed consent and administered independently completed questionnaires. Visits were arranged annually for 3 years. The local researchers attended joint training each year, and participated in group telephone discussions at around 3-month intervals to maintain consistency of approach. To maximise young person engagement and retention, outcome measures could be completed by post or electronically. At the baseline visit, the nature of the PBFs was discussed with the young person and a log-book was provided. Before each subsequent annual visit, the researcher consulted the young person’s medical records to seek evidence of the PBFs having been provided (for example, inclusion of a member of the adult service at a paediatric appointment). Then, at the visit, the researcher and young person completed a summary sheet recording whether each PBF had been experienced or not in the previous year; the information gathered from medical notes acting as additional prompts for the discussion.

Definition of the nine PBFs is provided in Box 1. ‘Appropriate parent involvement’ represents the perceptions of both the young person and parent being satisfied with level of parent involvement. These were chosen on the basis of being recommended in recent guidance [4], identified as beneficial in a systematic review [11, 14], and following our own analysis of individual studies (Table two in Colver et al. [12]).

We chose measures of satisfaction with services, mental wellbeing, participation and autonomy in appointments. They were chosen to be relevant across conditions and settings and correspond to measures subsequently proposed in international Delphi studies [15, 16]. In particular, we included a measure of mental wellbeing which captures what the young person feels about their life, as recommended by an International, Interdisciplinary Health Care Transition Research Consortium [15, 17]. The measure of participation reflects the importance that the International Classification of Functioning, Disability and Health [18] attaches to social as well as health outcomes.

Satisfaction with services was assessed using Mind the Gap [19]. This scale measures the difference or ‘gap’ between a young person’s ideal service and the service they have received (thus higher scores indicate lower satisfaction). Service satisfaction is expressed as a total score, with subdomain scores for Management of the environment, Provider characteristics and Process issues.

Mental wellbeing was assessed using the Warwick Edinburgh Mental Wellbeing Scale [20]. This 14-item questionnaire collects responses to each item on a 5-point Likert scale (‘none of the time’ 1 to ‘all of the time’ 5). Higher scores (range 14–70) denote higher mental wellbeing.

Participation was measured by the Rotterdam Transition Profile [21]. This captures independence in participation across nine domains. Independence is categorised into three ‘phases’, phase 1 being the least independent (thus higher scores indicate higher participation). A further participation measure, not in the original protocol, was added on the advice of the Programme’s External Advisory Board. This was Autonomy in Appointments, involving three questions about whether the young person makes their own appointments and asks and answers questions themselves (range 3–15) [22]. Higher scores indicate greater autonomy.

Baseline demographic information, including age, sex, index of multiple deprivation (IMD) [23] calculated from postcode in England, and the multiple deprivation measure [24] in Northern Ireland, and a number of socioeconomic indicators, including family composition, was collected for those who participated in the study.

‘Date of transfer’ was defined as the date of the last appointment with a paediatrician or adolescent psychiatrist. The young person’s status at each visit was recorded as still in a children’s service, transferred to dedicated adult service, transferred to General Practitioner in primary care, or lost to follow-up.

In order to maximise useable questionnaire data, the ‘final visit’ was defined as visit 4, or as visit 3 if visit 4 did not take place or questionnaire data were missing.

Our protocol [12] also proposed condition-specific outcomes and data relevant to economic analysis; these data will be or have been reported elsewhere [25, 26].

Age, Satisfaction with services (total and per domain), Mental wellbeing, and Autonomy in appointments were treated as continuous variables. Condition, site, sex, participation and exposure to PBFs were treated as categorical variables. Missing data were handled according to the suggested rules for each outcome measure. In the regression analyses, those with missing data for PBFs (year-by-year and consolidated PBF indicator) were grouped together and included in the modelling.

Representativeness between young people who were retained to final visit and those lost to follow-up, by age, sex, condition, site and socioeconomic status, was assessed using t, Mann–Whitney and χ² tests as appropriate. The Kruskal–Wallis test was used to examine associations between outcome measures and condition. Comparisons of outcome measures between baseline and final visit were tested using Wilcoxon paired sign-rank and χ² tests.

Two approaches to analysis were undertaken to assess the association of PBFs with outcomes across the duration of follow-up. The first approach used the young person’s experience of each PBF during the previous year (i.e. whether the PBF was present (yes) or absent (no) during that year) and tested this against each outcome at the end of that year. Hence, analyses were conducted ‘year-by-year’ and each was a cross-sectional analysis. The time between baseline (visit 1) and visit 2 was called ‘Period’ 1; similarly for periods 2 and 3.

The second approach used the young person’s experience of each PBF throughout the 3 years of the study. This was a longitudinal analysis. It was not appropriate to model the exposure to PBFs by simple frequencies, as young people had varying numbers of contacts with clinicians; in some years, a young person experienced a PBF and the next year might not. Also there was variation in the times at which the PBFs were captured; although intended to be every year, appointments often fell 1 or 2 months either side for practical reasons. We therefore developed a consolidated indicator to estimate the extent to which each PBF was delivered over the duration of the study. The indicator was based on whether the PBF had been experienced or not in each follow-up period. It was defined to be ‘optimal’ or ‘sub-optimal’ for each PBF, based on the following criteria developed by consensus within the members of the research team:

For both approaches, linear or logistic regression modelling was used, depending on the nature of the outcome variable, to test for association between each PBF individually or each consolidated PBF indicator and outcomes. All models were adjusted for age, sex, condition and potential for clustering by site.

A significance threshold of p ≤ 0.01 was used in final models to mitigate multiple testing (but associations at p ≤ 0.05 are presented as supporting evidence). Data were analysed using STATA version 14.

Throughout the 5 years of the Transition Research Programme, a young persons’ advisory group (UP) met each month. All group members had long-term health conditions. The group provided advice on outcome measures, recruitment and data collection, and interpretation of findings. Details of UP’s activities are on the Programme’s website: http://​research.​ncl.​ac.​uk/​transition/​.

To support recruitment and retention during the study, all participants and referring clinicians received regular newsletters (approximately every 9 months) about the progress of the study. Feedback of the results at the end of the study to participants took place in two ways, (1) through displaying results of the research on the website, and (2) a summary of the results was included in the final newsletter sent to every participant who wanted to continue to receive information about the study.

A total of 304 (81.3%) young people remained in the study by visit 2, 259 (69.3%) by visit 3 and 274 (73%) by final visit (235 from visit 4 and 39 from visit 3, see Methods). Of these 274 young people, 58% were male and there were 112 with diabetes, 74 with CP and 88 with ASD. The mean time between baseline visit and final visit was 2.9 years (SD 0.4, range 1.8–3.9).

There were no significant differences between those remaining in the study and those not remaining with respect to sex (p = 0.6), age (p = 0.6), condition (p = 0.6), diabetes sites (p = 0.4) or ASD sites (p = 0.6). However, in Northern Ireland, those with CP lost to follow-up came from areas with, on average, greater socioeconomic deprivation (p = 0.03). Examining socioeconomic factors based on actual circumstances rather than area of residence, there was a significant reduction in the proportion of families with single parents (Additional file 1: Table S1).

Of the 100 participants not remaining in the study, one had died, 28 were lost to follow-up and 71 withdrew. Of those withdrawing, 22 said they were no longer interested, 19 had other commitments, 19 experienced personal issues such as major injuries or severe parental illness, and the remainder gave miscellaneous reasons.

The mean age at final visit was 19.1 years (SD 1.4, range 16.1–22.0). Of the 274 participants at final visit, 49 (18%) remained in child services and 225 (82%) had left child services. Very different proportions by condition transferred to primary care (General Practice) as compared to a dedicated adult service (Table 1).

The average changes in outcomes between baseline and final visits are shown in Additional file 1: Table S2. In summary, satisfaction with services decreased overall but remained stable for those with diabetes; mental wellbeing was steady overall but was always lower for those with ASD and associated mental health problems; participation increased overall but was always higher for those with diabetes; and autonomy in appointments increased overall but was again always higher for those with diabetes.

Table 2 sets out the extent to which participants experienced optimal or suboptimal exposure to PBFs across the period of the study, using the consolidated indicator. Optimal exposure to features of transitional healthcare was 61% for ‘coordinated team’, 53% for ‘age-banded clinic’, 48% for ‘holistic life-skills training’, 42% for ‘promotion of health self-efficacy’, 40% for ‘meeting the adult team before transfer’, 34% for ‘appropriate parent involvement’, and less than 30% for ‘written transition plan’, ‘key worker’ and ‘transition manager for clinical team’. Significantly more young people with diabetes experienced optimal exposure to PBFs compared to those with CP or ASD, particularly ‘meeting the adult team before transfer’ and ‘promotion of health self-efficacy’.

In the year-by-year analysis, there were significant positive associations (p ≤ 0.01) during each period between ‘appropriate parental involvement’ and satisfaction with services (Mind the Gap) overall and in most domains of the instrument (Table 3). These were confirmed by significant associations of the consolidated PBF indicator with Mind the Gap at final visit.

‘Meeting the adult team before transfer’ was significantly associated (p ≤ 0.01) with a number of domains of the Rotterdam Profile in all three periods of the year-by-year analysis and also the consolidated PBF indicator (Table 3).

A number of weaker year-by-year and consolidated PBF indicator associations with other domains of the Rotterdam Profile were seen, but they were not in the predicted direction.

There were significant associations (p < 0.01) of ‘meeting the adult team before transfer’ with ‘autonomy in appointments’ in periods 1 and 3 and with the consolidated PBF indicator (p = 0.02).

Thus, three PBFs of transitional healthcare had significant (p ≤ 0.01) positive associations with better outcomes, namely ‘appropriate parent involvement’, ‘promotion of health self-efficacy’ and ‘meeting the adult team before transfer’. The b-coefficients indicated changes of approximately 0.5 SDs with respect to the satisfaction with services scale (SD 1.5 in our population), wellbeing (SD 7.0 in our population) and autonomy in appointments (SD 3.0 in our population). The odds ratios indicated increased likelihoods of being in a more independent phase of transition. The other six PBFs had few statistically significant positive associations (p ≤ 0.01, Table 3) with better outcomes in the year-by-year analysis, had a number of negative associations, and had no positive associations with the consolidated indicator of exposure to PBFs.

Our study was hypothesis driven, with pre-planned outcome measures that were applicable over condition and setting. These measures examined young people’s satisfaction with services, their wellbeing and participation, rather than focusing on process indicators such as attendance or loss to follow-up. There is a place for both outcomes and process indicators in transition evaluation. However, if service features do not improve the health or well-being of the young person, then it is hard to argue on clinical effectiveness grounds any basis for their adoption regardless of how process indicators may change. Our study has reported on wellbeing and participation which have rarely been used in this area, despite recommendations to do so [15, 17, 30]. Inclusion of a sample of young people with three contrasting conditions raises confidence in the generalisability of the findings to most young people with long-term health conditions. The retention rate of 73% to final visit was a considerable strength in this age group. Our data related to clinical practice for the three very different conditions, across several geographical locations. Unlike many previous studies [31‐35], it was not an evaluation by a local team of their local intervention, which risks observer bias and limits generalisability of findings. Further, we collected data longitudinally over 3 years of transition. One limitation was lower average recruitment from a group with a special need for transitional care, namely those from areas of greater socioeconomic deprivation. However, the difference in overall IMD score on a continuous scale ranging from 0.5 to 87.8, was only 6.1; further, there was no difference in proportion of single parent families as compared to national norms. Regarding attrition, this was more marked by IMD score only for those with CP in Northern Ireland. There was, however, a reduction in the proportion of single parent families, which could be relevant to interpreting our finding that appropriate parent involvement in transition was significantly associated with better outcomes.

There was also lower than intended recruitment from the two disability groupings, so that analyses controlling for condition may have been underpowered. A further potential limitation was the accuracy of exposure to PBFs. The local researchers were trained together each year on this topic and then held the discussion each year with the young person, supplemented by the young person’s notes and inspection of medical notes. The analysis by whether exposure to the consolidated PBF indicator had been ‘optimal’ was a more demanding interrogation of the data than the year-by-year analyses because it required there to be some degree of good practice throughout the 3 years, not just over 1 year. However, the decision rules (described in Methods) about what constitutes optimal exposure to each PBF were determined by the research team and have some subjectivity.

The three PBFs for which we found evidence of positive associations are supported by guidance and corroborating evidence discussed below. The link to current UK guidance (NICE) is also presented for each.

The NICE [4] guideline 43 emphasised ‘Appropriate parent involvement’ throughout its report; it is an overarching principle in section “Background”.1.1, and sections “Background”.2.19–1.2.22 focusing on the involvement of parents. Heath and Farre’s systematic review of studies of parents’ perceptions of their role in Transition [36] concluded that “Parents can be key facilitators of their child’s healthcare Transition, supporting them to become experts in their own condition and care. However, to do so parents require clarification on their role and support from service providers”. Akre’s study [37] found parental satisfaction with their involvement was associated with easier transition from the young person’s point of view. Allen’s study of young people with diabetes emphasised the importance of parents during Transition [38]. A review of qualitative studies [39] and a subsequent study [40] identified the tension that young people experience between seeking autonomy and still needing their parents. Two further recent reports investigated the parent–young person dyad [41, 42] and reached similar conclusions to those of the current study, namely that the parent and young person need to share care but that the change to adult autonomy is dynamic and will continuously change.

In section “Background”.2.17, NICE [4] recommended ‘Promotion of health self-efficacy’. Sattoe and van Staa [43] found that “continuing attention to self-management” was associated with better health-related quality of life. There is conflicting evidence as to whether a structured approach, including motivational techniques, to increase health self-efficacy in diabetes influences glycaemic control [44, 45]. After liver transplant, higher perceived self-management competence was actually associated with poorer clinical outcomes [46]. Mackie [47] showed the benefit of a 1-hour nurse-led intervention to promote knowledge and confidence about one’s condition, in this case congenital heart disease.

In sections “Background”.3.5 and 1.3.6, NICE [4] recommended ‘Meeting the adult team before transfer’. Our definition of meeting the adult team before transfer included clinics where adult and paediatric clinicians consulted jointly. In other studies, such joint clinics have shown improvements in certain indicators in nephrology (transplant rejection) [34], urology (aspects of care) [33] and rheumatology (knowledge of one’s condition) [48]. Crowley’s review [11] found joint clinics were associated with improved outcomes in those with diabetes.

The remaining six PBFs, for which our study found little evidence of benefit, had been included because a number of small published studies suggested they might be associated with improved outcomes. ‘Having a key worker’ was recommended by NICE [4] (sections “Background”.2.5–1.2.10 called for a ‘Named Worker’). Sloper et al. [49] found strong evidence for introducing key workers. The difficulty with ‘key worker’ may be operational rather than the principle; staff changes, due to leaving post, restrictions in job plans, service restructuring, sickness or maternity leave, make it difficult to provide a consistent key worker for all young people with long-term conditions. Having access to “holistic life-skills training” was recommended by NICE (sections “Background”.2.13–1.2.15). It was associated with greater satisfaction with service providers, and with more independent participation in domestic life in the third period of the year-by-year analysis, but not in the analysis by consolidated PBF indicators. Interpretation is difficult because few services provided this type of training. ‘Having a written transition plan’ was associated with greater satisfaction with services during the second study period but was otherwise negatively associated with many outcomes. NICE [4] (section “Background”.3.4) recommended transition planning but did not specifically mention that it should be a written document. Data from our qualitative work [50] showed the potential for careful planning to mitigate some of the disruptive, disorienting consequences of transfer. However, there were conflicting views about a ‘written’ plan. Some professionals said such plans get forgotten or lost, and personal interaction was far more important. On the other hand, lack of a formal plan left many families disoriented and wondering whether services would have sufficient resources to provide for care after transfer. Having a ‘transition manager for clinical team’ had no associations in the year-by-year analysis and had largely negative associations with the consolidated PBF indicator. In pilot studies, some benefits of having such a manager were seen after liver transplant [32] and in rheumatology [51].

Which outcomes should be measured in transition studies? [52‐54] There is a need to go beyond current Delphi studies, and crucially involve young people and their families in developing consensus. Our choice of outcomes was informed by the International Classification of Functioning [18] and by discussion with international transition researchers, and conformed with many of the recommendations of subsequent international surveys [15, 16]. Although carefully chosen and piloted, measures may not be ideal; for example, we found the domain Finances of the Rotterdam Transition Profile not to be sensitive to change by the end of the study, as most young people remained dependent in part on family financial support.