What factors predict youth mental health service use?

Mental health disorders are common in both children (18% in 7-16 year olds) and young people (22% in 17-24 year olds) (NHS Digital, 2022), a priority which is reflected within NHS England’s long term plan (NHS, 2019). This has only gotten worse since the COVID-19 pandemic, with a recent systematic review finding that suicidal thoughts and emergency department visits in children and young people have increased (Madigan et al., 2023; read Molly’s Mental Elf blog to learn more).

Ensuring that children and young people have timely access to mental healthcare is vital; but the need to tailor this care to reflect the individual needs of this population has also been long recognised (Department of Health, 2015). One way of tailoring services is by first identifying groups of services users with similar characteristics, who may subsequently have similar needs. This has been done to a greater extent with adult populations, but less is known about children and young people. To help work towards this, Julian Edbrooke-Childs and colleagues (2022) conducted a secondary analysis of real-world NHS data to explore factors that may predict mental health service use in children and young people.

Ensuring that children and young people have timely access to tailored mental healthcare is a priority – but how do we identify needs to tailor this care to? Analysing large-scale administrative datasets may hold the answer.

Methods

The authors conducted a secondary analysis of the Mental Health Services Data Set (2016-17 and 2017-18), comprising routinely collected administrative data from all publicly funded community services in England. The authors defined an episode of care as comprising of at least two attended care contacts that were less than 180 days apart, where:

Care was conducted either face-to-face, over the telephone, or not stated
Age at the start of the episode was between 0 and 25 years
The episode of care was closed
Complete data on presenting difficulties was available.

Thirty-two predictors of service use were explored in analyses, focusing on presenting difficulties (e.g., substance use difficulties, internalisation difficulties, schizoaffective difficulties) demographics (e.g., age, ethnicity, socio-economic status [SES]), and source of referral (e.g., self-referral, social care).

Results

The dataset analysed included 27,362 episodes of care across 39 services. The mean age of the population was 13 years old (SD = 4.71, range = 0-25 years), with an equal proportion of males and females. The authors found that the number of care contacts varied widely between services, accounting for almost a quarter (24%) of all variation between individuals. Episodes of care were for 27,033 young people, meaning that <3% of young people had two care contacts between 2016 and 2018.

Presenting difficulties

To simplify analyses, and to make most of the data, the authors categorised presenting difficulties into 12 categories:

Internalising (e.g., social anxiety, low mood; n = 20,034, 73.22%)
Relational (e.g., peer or family relationship difficulties; n = 17,764, 64.92%)
Externalising (e.g., conduct disorder; n = 12,641, 46.20%)
Neurodevelopmental (e.g., attention deficit hyperactivity disorder; n = 11,592, 42.37%)
Self-harm (n = 7,465, 27.28%)
Post-traumatic stress disorder (PTSD; n = 4,877, 17.82%)
Risk management (n = 4,601, 16.82%)
Emerging personality disorder (n = 4,164, 15.22%)
Eating disorder (n = 3,560, 13.01%)
Schizoaffective (e.g., bipolar disorder, psychotic symptoms; n = 3,443, 12.58%)
Self-care (n = 2,952, 10.79%)
Substance use (n = 2,628, 9.60%)

Of the 12 identified predictors, 7 presenting difficulties were statistically significant predictors of service use (number of care contacts):

Substance use (6.29 additional contacts, 95% CI 5.06 to 7.53, p < .001)
Eating disorder (4.30 additional contacts, 95% CI 3.29 to 5.30, p < .001)
Self-care (2.76 additional contacts, 95% CI 1.67 to 3.85, p < .001)
Emerging personality disorder (2.01 additional contacts, 95% CI 1.02 to 3.01, p < .001)
Schizoaffective (1.94 additional contacts, 95% CI 0.88 to 3.00, p < .001)
Internalising (-2.00 fewer contacts, 95% CI -2.83 to -1.18, p < .001)
Relational (-3.25 fewer contacts, 95% CI -4.10 to -2.40, p < .001)

Children and young people with substance use difficulties required the most additional number of contacts (on average, 6 additional contacts compared to those without substance use difficulties) with mental health services.

Children and young people with relational difficulties required the least contact (on average, 3 fewer contacts compared with those without relational difficulties).

Demographics

Of the 13 identified predictors, 4 demographic factors were statistically significant predictors of service use:

Not reported ethnicity vs White (-2.78 fewer contacts, 95% CI -3.74 to -1.83, p < .001)
Mixed-race vs White (2.63 additional contacts, 95% CI 0.76 to 4.49, p = .006)
Female vs Male (1.64 additional contacts, 95% CI 0.94 to 2.33, p < .001)
Age (0.71 additional contacts, 95% CI 0.62 to 0.80, p < .001)

Source of referral

Of the 7 identified predictors, 3 factors related to referral source were statistically significant predictors of service use:

Social care/youth justice vs primary care (4.17 additional contacts, 95% CI 2.49 to 5.85, p < .001)
Mental health vs primary care (3.98 additional contacts, 95% CI 2.59 to 5.37, p < .001)
Child health vs primary care (2.63 additional contacts, 95% CI 0.89 to 4.37, p = .001)

Out of 32 predictors of service use, difficulties with substance use was the strongest predictor of contact with mental health services, with young people requiring an average of 6 more cases of contact.

Conclusions

The authors of the study concluded that: “young people with substance use, eating disorders, self-care difficulties, or schizoaffective difficulties had higher numbers of care contacts than young people without these presenting difficulties” indicating that those with more complex difficulties are receiving more treatment, although there may still be issues with disengagement and treatment drop-out.

Regarding the higher number of contacts for children and young people referred through social care/youth justice, the authors suggested that: “Referral through this route is less likely to be voluntary, and it may be that engagement with services was higher as this was compulsory.”

Referral through mental health services was also higher than primary care, potentially reflecting longer waiting times (Edbrooke-Childs & Deighton, 2020). Although these findings are interesting, most of these trends require further research to unpick why this might be the case.

Study findings suggest that young people with more complex difficulties are receiving more treatment – however, large differences in number of contacts between services accounted for approximately a quarter of variation between individuals.

Strengths and limitations

Research on children and young people and their contact with mental health services is much less developed than for adults in mental health services, meaning this study is a welcome addition to the literature. Additional strengths include the use of a large-scale national dataset of 39 services, resulting in increased statistical power and confidence in the results, alongside greater representation, reliability, and generalisability of the study findings.

However, there are some limitations:

Missing data was a significant limitation of the findings, with a considerable number of episodes of care excluded from the analysis due to a lack of information about presenting difficulties (373,957 of 424,940). These findings therefore reflect a small proportion of the Mental Health Services Dataset, and make it unclear as to whether the included episodes of care are reflective of NHS service provision as a whole, or whether there was something systematically different about these episodes of care. However, this was not the fault of researchers, but an inherent limitation of the data source and conducting secondary analyses. Further work should be conducted to improve data quality so that more informative analyses are possible in future.
Following the above, data may have been collected from different services in different ways, with the authors having no control over consistency. This may have impacted the reliability and comparability of the data.
Additionally, the same young person could have been attending multiple services, and therefore represented in the dataset multiple times, potentially causing concerns for the validity of the results.
Finally, the authors highlight that only one researcher coded the data, potentially introducing bias and inaccuracies to the dataset. In the future, double-screening and noting the inter-rater reliability between coders would be useful in determining accuracy.

A considerable proportion of episodes of care were excluded from the study due to a lack of information about presenting difficulties, highlighting a need for improved reporting within services.

Implications for practice

Edbrooke-Childs et al. (2022) provide a useful study that may contribute to the future planning of mental health service provision for children and young people. For example, ethnicity appears to impact number of contacts, with both mixed-race and Black children and young people requiring more contact than White children and young people (although the association between Black and White children and young people was non-significant). Commissioners might want to consider the geographical demographics of services within their remit, and whether additional resources are needed to help meet demand.

The authors also identified that children and young people struggling with substance use, eating disorders, self-care, schizoaffective disorders, and emerging personality disorders were more likely to need additional contact with mental health services than other conditions. These associations are not unexpected given the literature, but nonetheless provide useful data for services to reflect on how best to support these groups of children and young people.

The study raises interesting questions regarding why those presenting with internalising or relational difficulties experienced less contact with mental health services than other conditions. This could suggest that children and young people presenting with these difficulties may require less intensive intervention; alternatively, it could just reflect that there was more attrition (drop-out) in these groups. Future research should explore what factors may be driving these potential associations, perhaps in a different dataset or a prospective study.

Edbrooke-Childs et al (2002) identified important differences between services in number of care contacts, which accounted for approximately a quarter of all variation. It is unclear what the implications of these findings are. Does it reflect that some services are better at engaging children and young people than others? If so, then there is the potential to investigate why this is the case and whether their good practice can be replicated in other services. Alternatively, it may reflect that some services are providing less effective treatment (or due to other factors not identified in the study) so require more contacts for either similar or less benefits. Again, this emphasises the need for additional studies with more comprehensive data.

The study raises interesting questions regarding why children and young people presenting with internalising or relational difficulties experienced less contact with mental health services than other presenting difficulties.

Statement of interests

None.

Links

Primary paper

Edbrooke-Childs, J., Rashid, A., Ritchie, B., & Deighton, J. (2022). Predictors of amounts of child and adolescent mental health service use. European Child & Adolescent Psychiatry, 1-8.