Introduction

There are many familial, social and developmental complexities growing up with perinatally acquired HIV (PHIV), which makes ongoing self-management a chronic condition through adolescence and adulthood challenging [1–8]. This is reflected in the often poorer health outcomes of young people living with PHIV compared to other age groups [9, 10].

Engagement in care (EIC) has been associated with improved patient outcomes and is a focus of global and national targets [11–14]. EIC is especially important in young people with PHIV as they are a particularly vulnerable group. However, few studies have measured EIC in children and young people with PHIV in Europe, and none have looked at characteristics associated with EIC, which could help identify young people at increased odds of disengagement. Furthermore, most studies of EIC in people living with HIV use a simplistic definition of EIC based on the number of clinic visits per year.

In a previous paper [15], we report how we adapted an existing EIC algorithm for adults with HIV in England [16], for use in young people and applied it to young people living with PHIV in the Adolescent and Adults Living with Perinatal HIV (AALPHI) cohort in 2013–2015. AALPHI was a prospective observational study evaluating the impact of HIV infection and ART exposure in young people in England [17, 18]. Here we build on that papers results by exploring a broad range of potential characteristics of EIC. The aim was that these characteristics could be useful to clinicians to identify and provide additional interventions for young people living with PHIV who are more likely to disengage from HIV care.

Materials and methods

Potential characteristic associated with EIC

Published papers of factors known to be associated with EIC were reviewed and expert opinion sought, to select potential characteristics for this analysis. In addition, characteristics that may affect recognised factors associated with EIC were also considered (see S1 File for a list of all variables in the AALPHI and CHIPS cohorts, and S2 File for the rational for all included variables). For example, mental health has been found to be associated with poorer EIC [20] therefore variables associated with mental health status (such as exclusion from school and death of parents) were included. The conceptual framework used for this analysis was developed from the work of Harberer and Mellin [21] on adherence to ART in young people living with PHIV. They recommended that due to the complex challenges of long-term adherence, four categories, or domains, need to be taken into consideration when planning interventions; the child; caregiver(s) and family; the ART regimen; and society and culture [21]. Our study benefited from measurement of a wide range of potential exposure variables, far beyond the more limited core clinical HIV factors included in many analyses. We extended Harberer and Mellin’s domains by grouping these variables into 7 domains (Table 1) and adjusted for a priori variables. A priori variables were time from interview, sex, age, ethnicity, country of birth. Management of missing data are described elsewhere [15].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. A priori and baseline characteristics considered in EIC models, grouped by 7 domains.

https://doi.org/10.1371/journal.pone.0302601.t001

Results

Of 316 young people living with PHIV included in this analysis (Table 2, column a), just under two thirds were female (59%) and median age was 17 years [IQR 15–18]. Most young people were of black ethnicity (86%), and born outside the UK/Ireland (58%). At baseline, the majority of young people were on ART (88%), two thirds (69%) had a suppressed viral load (≤50 c/mL) and the median CD4 count was 597 cells/μL [427, 791]. In total there were 3,585 person-months of follow-up, with 87% (3,126/3,585) of months fulfilling the definition of engaged in care.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Characteristics associated with EIC in domain-specific and combined multivariable models.

https://doi.org/10.1371/journal.pone.0302601.t002

Table 2 shows characteristics associated with EIC from the domain-specific (column b) and combined models (column c). In the combined model, ethnicity, ever self-harmed, adherence self-assessment and viral load were independently associated with EIC. Young people were less likely to be engaged in care if they were of Asian/mixed/other ethnicity (OR 0.44, 95% CI 0.25, 0.78) compared with white or black ethnic groups (white OR 1.05, 95%CI 0.29, 3.81, reference black ethnicity, p = 0.02). They were less likely to be engaged in care if they had ever self-harmed (OR 0.55, 95% CI 0.32, 0.95) compared to those who had not self-harmed (p = 0.03). They were also less likely to be engaged in care if they were on ART and self-assessed their adherence as not so good or bad (OR 0.46, 95% CI 0.25, 0.84) or were not on ART (OR 0.64, 95% CI 0.34, 1.21) compared to those on ART who assessed their adherence as excellent or good (p = 0.04). Finally young people were less likely to be engaged in care if they had a baseline viral load >50c/mL (OR 0.47, 95% CI 0.30, 0.75) compared to ≤50c/mL (p = 0.002).

Discussion

In this paper, we investigated the association between a broad range of characteristics and EIC in young people living with PHIV. To our knowledge this is the first analysis that examines characteristics associated with EIC in young people living with PHIV in the UK, and the first multi-clinic analysis in Europe. Using individual patient data from the AALPHI and CHIPS cohorts, we were able to measure EIC in a more nuanced way than many previous analyses [23–26] by taking into account participants’ clinical characteristics. Variables were grouped into eight domains, and four variables were found to be associated with EIC, each from a different domain. From the a priori domain, young people were less likely to be engaged in care if they were of Asian/mixed/other ethnicity compared to being of black or white ethnicity. From the mental health domain, participants had worse EIC if they reported having ever self-harmed. From the HIV management and experience domain participants were less likely to be engaged in care if they reported having not so good/bad adherence to ART, or if they were not on ART, compared to excellent/good adherence to ART. Finally from the “clinical markers” domain, participants had poorer EIC if they had a viral load >50c/mL compared to ≤50c/mL. Results of sensitivity analyses were reassuringly similar to the main combined model, with ethnicity, self-harm, adherence and viral load remaining associated with EIC across all three sensitivity analyses. In addition, there was evidence from all three sensitivity analyses of a potential effect of older age and worse EIC.

It is important to note that relatively few studies have looked at EIC in young people living with PHIV. Additionally the method of measuring EIC varies widely in the literature, and often most characteristics are from the a priori and HIV clinical markers domains, limiting direct comparisons. This paper used an adapted algorithm for EIC which was originally developed by Howarth et al [16] for adults with HIV in the UK, and results are consistent across the age groups. Howarth et al [16] demonstrated that age was also associated with EIC in univariable analyses, with the proportion of patients engaged in care being lowest in the 16–24 years age group (77% engaged in care, vs. 83% for 25–45 years and 87% for >45 years). However after adjustment for other factors the association was weaker (16–24 years p = 0.09, 25–45 years p = 0.008 vs >45 years). In terms of ethnicity, they also reported that people who identified as being in an ethnic group other than black African or white [16] were less likely to be engaged in care, similar to our findings. Such patients are minority ethnic groups in HIV clinics in the UK, and may have specific barriers to engaging in care. Howarth et al [16] also reported that people who were on ART were more likely to be engaged in care than people who were off ART, similar to our findings. They also found that patients with lower nadir CD4 counts were less likely to be engaged in care, and did not look at the effect of viral load [16]. Our model included both, and CD4 nadir was dropped in favour of viral load.

In addition to Howarth et al, [16] two other studies identified associations between age and EIC. Gray et al [23] in the USA, reported that young people living with PHIV aged 13–17 years had a higher EIC that 18–25 year olds. Similarly, Gebrezgi et al [27] reported that 13–17 year olds had better EIC than 18–20 year olds and 21–24 year olds. This is similar to the trend shown in our analysis with EIC declining as young people aged.

Other studies, where EIC was more simply defined, have also found associations between ethnicity and EIC. In Gray et al’s [23] analysis of young people living with PHIV across the USA (n = 11,747), people of Hispanic/Latino ethnic group were more likely to be engaged in care than people of white ethnicity. In contrast, in a cohort of young people aged 13–24 years with HIV in Florida (n = 2,872), Gebrezgi et al [27] in their cohort of young people with mainly horizontally acquired HIV reported that Hispanics were less likely to be engaged in care than people of non-Hispanic white ethnicity. These contradictory findings are perhaps unsurprising and difficult to compare. Ethnicity in the context of these studies, as well as our own, can be considered a measure of health disparities caused by structural inequalities and discrimination [28]. The complexity of these factors and the impact on engagement in care of young people living with HIV may differ across the different social and political contexts in the UK and US as well as between different groups of people.

In our analysis, self-harm was the only mental health factor, and adherence the only HIV management and experience factor, associated with EIC. Most other studies of EIC included only routinely collected sociodemographic and clinical factors in their analyses, and it may be challenging to enable honest reporting of self-harm and adherence by young people in clinic. A previous analysis from our group suggested that in the AALPHI cohort, self-esteem was strongly associated with self-harm, with each one point increase in the Rosenberg Self-Esteem Scale (higher score = better self-esteem) being associated with a 10% reduction in the odds of self-harm [29]. Similarly in the Avon Longitudinal Study of Parents and Children (ALSPAC) general population (non-HIV specific) study, concurrent depression was associated with a much increased risk of self-harm [30]. Any indication of mental health issues might thus be considered a flag for risk of disengagement with care.

Conclusion

In conclusion, the richness of the AALPHI dataset enabled us to explore the potential effect of a broad range of variables on EIC. Many clinics already use viral load as an indicator that young people may be disengaging with care. This analysis suggests that in addition to viral load, simple assessments of adherence and self-harm (or other mental health measures) may have added benefit in identifying young people most at need of targeted interventions to optimise clinic attendance.

Supporting information

Acknowledgments

AALPHI Steering Committee chairs: D. Gibb, C. Foster. NHS clinics (named alphabetically): LONDON: Chelsea and Westminster NHS Foundation Trust, F. Boag, P. Seery; Great Ormond Street Hospital NHS Foundation Trust, M, Clapton, J. Flynn; Guys and St Thomas’ NHS Foundation Trust, A. Callahgan, A, Mukela; Imperial College Healthcare NHS Trust, S. Ayres, C. Foster; King’s College Hospital NHS Foundation Trust, E. Fuller, E. Hamlyn; Mortimer Market Centre, Central and North West London NHS Foundation Trust, R. Gilson, T. Peake; Newham University Hospital, V. Apea; North Middlesex University Hospital NHS Trust, J. Daniels, C. Wood; Royal Free London NHS Foundation Trust, T. Fernandez, S. Kinloch; St George’s University Hospitals NHS Foundation Trust, S. Storey, K. Prime. REST OF ENGLAND: Alder Hey Children’s NHS Foundation Trust, S. Paulus, A. Riordan; Birmingham Heartlands, Heart of England NHS Foundation Trust J. Daglish, C. Robertson; Bristol Royal Infirmary, University Hospitals Bristol NHS Foundation Trust, J. Bernatoniene, L. Hutchinson, University Hospitals Bristol NHS Foundation Trust, M. Gompel, L. Jennings; Leeds Teaching Hospitals NHS Trust, M. Dowie, S. O’Riordan; University Hospitals of Leicester NHS Trust, W. Ausalut, S. Bandi; North Manchester General Hospital, Pennine Acute Hospitals NHS Trust, P. McMaster, C. Murphy; Royal Liverpool and Broadgreen University Hospitals NHS Trust, M. Chaponda, S Paulus.

Voluntary services (named alphabetically): Blue Sky Trust, C. Dufton, B. Oliver; Body and Soul, J. Marsh; Faith in People, I. Clowes; Positively UK, B. Osoro; Metro Charity, B. Driver.

We thank Alison Howarth and Caroline Sabin for their permission to adapt their adult engagement in care measure for young people.