Use of healthcare services before diagnosis of attention-deficit/hyperactivity disorder: a population-based matched case-control study

Objective To compare use of healthcare services and reasons for attendance by children and young people (CYP) with attention-deficit/hyperactivity disorder (ADHD) versus non-ADHD controls. Design Population-based matched case-control study. Setting English primary care electronic health records with linked hospital records from the Clinical Practice Research Datalink, 1998–2015. Participants 8127 CYP with an ADHD diagnosis aged 4–17 years at the time of diagnosis and 40 136 non-ADHD controls matched by age, sex and general practitioner (GP) practice. Main outcome measures Medical diagnoses, prescriptions, hospital admissions and hospital procedures in the 2 years before diagnosis (or the index date for controls). Results CYP with ADHD attended healthcare services twice as often as controls (rate ratios: GP: 2.0, 95% CI=2.0, 2.1; hospital 1.8, 95% CI=1.8, 1.9). CYP with ADHD attended their GP, received prescriptions and were admitted to hospital for a wide range of reasons. The strongest association for GP attendances, comparing CYP with versus without ADHD, was for ‘mental and behavioural disorders’ (OR=25.2, 95% CI=23.3, 27.2). Common reasons for GP attendance included eye, ear, nose, throat, oral (OR=1.5, 95% CI=1.4, 1.5) and conditions such as asthma (OR=1.3, 95% CI=1.3, 1.4) or eczema (OR=1.2, 95% CI=1.0, 1.3). Conclusions Two years before diagnosis, CYP with ADHD attended healthcare services twice as often as CYP without. CYP with ADHD had increased rates of physical conditions, such as asthma and eczema. These contacts may be an opportunity for earlier recognition and diagnosis of ADHD.


INTRODUCTION
Attention-deficit/hyperactivity disorder (ADHD) is a prevalent neurodevelopmental disorder affecting 3%-5% of children and young people (CYP) 1 characterised by developmentally inappropriate hyperactivity, impulsivity and inattention in multiple settings (eg, school and home). 2 Peak age for diagnosis of ADHD is at 7-9 years. 3Yet, as a neurodevelopmental disorder, ADHD is likely to have been present from an earlier age, with persistent symptoms. 2 4ADHD is under-recognised, with <1% of CYP having a diagnosis in medical records, 1 5 often with considerable delays, which are not well understood. 6Girls are disproportionately affected by underdiagnosis and delays in diagnosis, 7 which results in transitioning from primary/junior school (4-11 years) to senior school (12-17 years) with inadequate support.ADHD is associated with poor mental health and negative outcomes across the life course, [8][9][10][11][12] such as poor relationships and social functioning, low self-esteem, 12 academic underattainment [13][14][15] and injuries. 16utcomes can be improved by specialist services and evidence-based interventions, 1 2 but access to these is often significantly delayed. 6One reason for

WHAT IS ALREADY KNOWN ON THIS TOPIC
⇒ Attention-deficit/hyperactivity disorder (ADHD)  is the most frequent neurodevelopmental disorder among children and young people (CYP), with an estimated prevalence of 3%-5%.⇒ Diagnosis of ADHD in CYP is often delayed or missed and this may impede access to effective treatments.⇒ Previous work has suggested that general practitioners (GPs) have difficulty recognising ADHD.

WHAT THIS STUDY ADDS
⇒ Compared with controls, CYP with ADHD have attended their GP and general hospital services twice as often in the 2 years preceding their diagnosis.⇒ CYP attend for a range of reasons, including mental and physical reasons unrelated to ADHD.⇒ These frequent contacts of CYP with a range of healthcare services may present an opportunity for earlier diagnosis.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY
⇒ Further research is required to investigate the reasons for delayed diagnosis of ADHD.⇒ Parents/caregivers with concerns about ADHD should be encouraged to discuss these with the GP or specialist services.⇒ Care pathways should aim to be holistic so that concerns about physical, mental health and behavioural difficulties can be considered at the same time.⇒ Integrated care systems should be alert to serving the needs of CYP with symptoms of ADHD across health, education and social care services.

Original research
this is that general practitioners (GPs), who are the main gatekeepers to specialist services in the UK, 1 2 have difficulty recognising ADHD, 17 18 leading to delay in referral.Identifying early markers of and wider health presentations associated with undiagnosed ADHD represents an opportunity to improve recognition and outcomes.In this study, we investigate the pattern of healthcare utilisation of CYP with ADHD in the 2 years before diagnosis to explore opportunities for earlier recognition and referral.

Data source
The Clinical Practice Research Datalink (CPRD) is a primary care database containing records of 15.2 million people from 730 GP practices and covers 7% of the UK population. 19 20We conducted a case-control study using medical records from just over half of CPRD practices in England who consented to linked hospital medical records from the Hospital Episodes Statistics (HES) database.HES contains the diagnoses associated with all overnight National Health Service (NHS) admissions and operative procedures. 21Previous work including all ages has shown a high validity of CPRD diagnoses, with a median of 89% of cases confirmed by manual review of GP records. 22The HES-linked CPRD is representative of the UK population in terms of age and sex, covering practices from every Strategic Health Authority (SHA) region of England. 23 24The CPRD contains Read (diagnostic and attendance) codes and drug codes arising from attendances at GPs.It also contains Read-coded correspondence from attendances at secondary care at outpatients, emergency departments, overnight admissions or operative procedures.CPRD contains Index of Multiple Deprivation (IMD) 2015, based on home postcode as a proxy measure of an individual's socioeconomic status. 25HES data are coded using International Classification of Disease, V.10 (ICD10) 26 and procedure codes from the Office of Population Censuses and Surveys, V.4 (OPCS4). 27

Study population
The study period was 1 January 1998 (HES-link started) to 31 December 2015 (data extraction).Individuals were eligible for inclusion if (1) they were 4-17 years of age during the study period, (2) they had been registered for at least 2 years in a primary care practice, which met research data recording standards (known as 'up to standard' and defined using CPRD algorithms examining patterns of data completeness and temporal gaps in recording), (3) they had at least 2 years of medical records prior to the index date.The index date was defined as the date of the first code for ADHD in the matched ADHD patient record.Events in the first 90 days included of medical records were ineligible because historical diagnoses may be incorrectly recorded as new diagnoses in the first few months of registering with a new GP practice. 28ases had at least one drug/diagnostic code for ADHD (see online supplemental tables 1 and 2 for list of codes).
Controls had no record of ADHD in their GP medical records.Up to five controls were randomly selected and matched on age (year of birth), sex, practice and index date.

Reasons for attendance at healthcare services
The medical records from the CPRD GOLD and linked databases were extracted for participants in the 2 years before the index date. 19All codes were manually reviewed and categorised based on existing classification systems (ICD10, 26 British National Formulary chapter headings 29 and OPCS4 chapter headings) 27 and clinical review (PW and VP).

Comparison of reasons for attendance by category between cases and controls
The number of people with one or more attendance in the 2 years prior to index date was counted according to category for: (1)  Read codes (GP attendance and hospital correspondence); (2) drug codes (GP prescriptions); (3) ICD10 (hospital admissions); and (4) OPCS4 codes (hospital procedures).This number was the numerator.For example, a CYP who attended the GP more than once for a respiratory condition would contribute once to the analysis for respiratory Read codes.The denominator was the total number of cases in the study population.The outcome was presented as the proportion of attendances among the case and controls.Conditional logistic regression compared CYP cases to controls.We calculated relative risk estimates as ORs with 95% CIs.Where there were less than 20 cases or controls, the category was omitted from the results for clarity and to protect anonymity.

Reasons for attendance
The three most frequently occurring codes resulting from CYP's healthcare attendances were inspected for each type of code (online supplemental tables 4-7).We compiled a list of physical and mental health conditions for which CYP with ADHD might attend their GP (online supplemental table 3).To identify these conditions, we compiled a list of Read and drug codes (VP and PW).We reviewed the medical records of CYP with versus without ADHD for at least one attendance with the condition.An unadjusted OR was estimated for the condition using conditional logistic regression.
To explore the frequency of healthcare service use between cases and controls, the attendance rates were described for CYP with versus without ADHD separately for Read codes, drug codes, ICD10 codes, OPCS4 codes.The frequency was described for GP attendances versus hospital admissions.
To assess whether patterns of healthcare attendance varied by sex or age, the proportions with each category of attendance were examined separately by sex and age (<11 years vs 12 years+, reflecting later diagnosis of ADHD).Age bands were chosen to reflect UK schooling (infant, junior and secondary) ages.We undertook subgroup analyses assessing the effect on our findings through varying the definition of an ADHD diagnosis as (1) at least two drug codes and at least two diagnosis codes, (2) at least two diagnosis codes and less than two drug codes, (3) at least two drug codes and less than two diagnosis codes and (4) one drug code or one diagnosis code/one drug code and one diagnosis code.Descriptive analyses are reported because of the size of the dataset and the large number of comparisons.
Lists of all Read and drug codes are available from the authors on request.Statistical analysis was performed using Stata V.15.1 (StataCorp, College Station, Texas, USA).

RESULTS
Table 1 shows the sample characteristics (8127 cases and 40 136 controls).
CYP with ADHD were twice as likely as controls to have contact with the health services, prior to diagnosis, regardless of the type of contact (GP consultation, prescription, hospital admission and procedure) (GP: 8 vs 4; hospital 0.2 vs 0.1, times per year) (table 2).

Original research
The odds of cases having attended their GP were greater than controls in all 17 categories (figure 1).The strongest association was for 'mental and behavioural disorders' (OR=25.2,95% CI=23.3,27.2).The odds for cases receiving a prescription were greater in 16 of 17 categories (figure 2).The strengths of the association with a 'circulatory' prescription (OR=2.5, 95% CI=1.7, 3.5) and a nervous system prescription (OR=2.2,95% CI=2.1, 2.4) were similar but the former was an uncommon event.
Figure 3 shows medical conditions recorded in the GP medical records in the 2 years before diagnosis.For 18 out of 19 conditions, the odds were greater for cases.The strongest associations were for 'behaviour codes' (OR=29.7,95% CI=26.7,33.1) and 'learning disability' (OR=10.9,95% CI=8.6, 13.8).
Figure 4 shows admissions to hospital in the 2 years before diagnosis.For 15 out of 17 categories, the odds were greater for cases.The strongest association was for 'mental and behavioural disorders' (OR=10.2,95% CI=8.3, 12.4).
For 12 out of 13 categories, the odds were greater for cases for hospital procedures (online supplemental figure 1).Sexstratified analyses showed little difference between males and females.However, there were exceptions (eg, compared with males, female cases had a stronger association for: GP attendances (musculoskeletal, injuries); prescriptions (nervous   The square shows the OR and the horizontal line shows the 95% CI.ADHD, attention-deficit/hyperactivity disorder; CYP, children and young people; Freq., frequency; GP, general practitioner; LL, lower limit of the 95% CI; UL, upper limit of the 95% CI.

Main findings
In the 2 years preceding their diagnosis, CYP with ADHD used health services significantly more than controls.Their rates of attendance to GPs and hospitals were higher across almost all presentation categories investigated.The largest differences were seen in mental and neurodevelopmental presentations but there were increased rates of physical conditions, such as asthma and eczema.They also received treatment twice as often and were more likely to have records of other physical and mental health conditions.These findings suggest that there were potential opportunities for earlier recognition.

Strengths and weaknesses
To our knowledge, this is the first work to investigate the primary and secondary healthcare attendances (using English NHS healthcare records) prior to diagnosis of ADHD.The CPRD-HES linked database provides the most detailed available picture of healthcare attendances.Previous work has shown that the CPRD-HES linked data are representative of GP and hospital attendances in the UK population and there is no reason to suspect findings from HES-linked practices differ from the overall CPRD. 5 16 22-24 30 However, there were limitations.Although ADHD misdiagnosis is unlikely because it is diagnosed by specialists according to national guidelines, 2 a small number (<1% of CYP) had a record of ADHD in their hospital but not GP records.This may result from delays in communication from hospital.There was a high proportion of attendances for 'behavioural' reasons which implies that GPs may have been aware that some presentations were suggestive of ADHD.GPs   Categories are arranged in order of frequency for CYP with ADHD.ADHD, attention-deficit/hyperactivity disorder; ENT, ear, nose and throat; Freq., frequency; LL, lower limit of the 95% CI; UL, upper limit of the 95% CI.

Original research
use Read for clinical recording and do not primarily collect these data for research.However, national guidelines 2 specify referral to specialist services for assessment of ADHD, perhaps resulting in inherent delays to diagnosis, especially if there are long waiting lists. 6Presentations to healthcare may arise from parents/caregivers having a low threshold for seeking help when their child was unwell for other reasons and when they had other long-term medical conditions, resulting in ascertainment bias.However, there were a greater number of hospital admissions and procedures for CYP with ADHD which might imply more severe presentations, and not accounted for by lower helpseeking thresholds.It is also possible that the true association between health problems and ADHD may be smaller than our estimates.These data were extracted from the CPRD to cover the 1998-2015 period.Recent reviews have implied potential for overdiagnosis of ADHD, 1 31 mainly in the USA and reported under-recognition in the UK CPRD up to 2010. 3 32 However, since cases and controls were matched by age, it is unlikely that temporal trends in recognition (eg, rising recognition of ADHD) were different between cases and controls.Although we matched cases and controls on age, sex and practice, there may have been other confounders (such as parental mental health problems including parental ADHD) that we were unable to account for.Individual postal codes are unavailable from CPRD to protect anonymity.CYP attend healthcare for a wide range of reasons.Due to the nature of Read codes, which include codes for GP administration and symptoms, a high proportion of attendances were classified as 'factors influencing health status/contact with services'.Although the reason for these attendances is unclear, it is unlikely that the effect would have been different between cases and controls.

Comparison with existing literature
Our findings are consistent with previous studies suggesting that CYP with ADHD are at an increased risk of having other mental or physical health disorders, including injuries. 16For example, in a German health insurance database study, 83% with ADHD (vs 20% without) had a comorbid psychiatric diagnosis and 2% (vs 1.3%) had a cardiovascular disorder. 33A Korean health insurance database study reported a range of associations between ADHD and other disorders, for example, nervous system disease (OR=2.59),which was similar to our findings. 34However, the health events were based only on hospital attendances, which may not represent the full range of disorders or healthcare presentations.A global systematic review reported a similar association between ADHD and asthma as our findings. 35The male to female ratio was 4.8:1 in our sample which is similar to previous studies from healthcare data in the UK (5:1) 36 and Germany (3-4:1). 37There is some suggestion that, compared with males, females with ADHD have more injuries, overnight admissions for respiratory conditions, receive more nervous system drugs, and have recorded autism and language impairments.These findings warrant further research and may address the unequal recognition of females with ADHD. 7 Implications for future research CYP with ADHD had a higher risk of mental health or behaviour codes, dyspraxia, dyslexia, autism, tics and insomnia being recorded prior to diagnosis.These codes might suggest an opportunity for earlier diagnosis or may indicate clinicians were already gathering information (eg, related to other neurodevelopmental conditions), which would eventually lead to an ADHD diagnosis.These findings warrant further research.Further research is also required to develop and test interventions to identify ADHD earlier in primary care. 38For example, machine learning in CPRD data may generate a predictive model for automated detection of ADHD among patients with no formal diagnosis, such as those generated for other conditions 39 or in other datasets. 40CYP with ADHD have multiple health needs and long-term vulnerabilities.Research exploring how CYP with ADHD might interface differently with health services might be useful in understanding the overlap of mental and physical health needs among CYP presenting in primary care, especially as CYP with ADHD have high healthcare costs. 32However, our work suggests that CYP with ADHD may also incur greater healthcare costs prior to their diagnosis.

Implications for practice
While acknowledging that GPs may not recognise reasons for attendance as being related to ADHD 18 and that parent/caregiver perceptions of a behaviour problem influence recognition, 18 41 our work suggests there are potential earlier opportunities to identify undiagnosed ADHD.Healthcare practitioners (primary and secondary care), and Integrated care systems in health, education and social care services, should be aware that CYP who attend frequently and for a wide variety of reasons may have additional needs reflecting an unrecognised healthcare problem, such as ADHD.Patient consent for publication Not applicable.
Ethics approval All procedures involving human subjects/patients were approved by CPRD's independent scientific advisory committee (CPRD protocol number 17_269).CPRD data are anonymised, therefore further ethical approval was not required.
Provenance and peer review Not commissioned; externally peer reviewed.

Figure 1
Figure1Attendances to the GP by categories of diagnoses (Read codes), comparing children and young people with (n=8127) versus without (n=40 136) ADHD (n=48 263).Categories are arranged in order of frequency for CYP with ADHD.The square shows the OR and the horizontal line shows the 95% CI.ADHD, attention-deficit/hyperactivity disorder; CYP, children and young people; Freq., frequency; GP, general practitioner; LL, lower limit of the 95% CI; UL, upper limit of the 95% CI.

Figure 2
Figure 2Attendances to the GP by categories of drugs (drug codes), comparing children and young people with (n=8127) versus without (n=40 136) ADHD (n=48 263).Categories are arranged in order of frequency for CYP with ADHD.The square shows the OR and the horizontal line shows the 95% CI.ADHD, attention-deficit/hyperactivity disorder; CYP, children and young people; Freq., frequency; GP, general practitioner; LL, lower limit of the 95% CI; UL, upper limit of the 95% CI.

Figure 3
Figure 3Medical conditions recorded in the GP medical records, comparing children and young people with (n=8127) versus without (n=40 136) ADHD (n=48 263).Categories are arranged in order of frequency for CYP with ADHD.ADHD, attention-deficit/hyperactivity disorder; CYP, children and young people; Freq., frequency; GP, general practitioner; LL, lower limit of the 95% CI; UL, upper limit of the 95% CI.

Figure 4
Figure 4Admissions to hospital by categories of diagnoses (ICD10 codes), comparing children and young people with (n=8127) versus without (n=40 136) ADHD (n=48 263).Categories are arranged in order of frequency for CYP with ADHD.ADHD, attention-deficit/hyperactivity disorder; ENT, ear, nose and throat; Freq., frequency; LL, lower limit of the 95% CI; UL, upper limit of the 95% CI.

Table 1
Characteristics of children and young people with and without ADHD Sex, age and practice were matching variables.Therefore, cases and controls were registered at the same GP practice.ADHD, attention-deficit/hyperactivity disorder ; GP, general practitioner.

Table 2
Rates for contact with health service by type of contact (GP consultation, prescription, hospital admission and procedure) for children and young people with vs without ADHD (n=48 263) Average yearly rate over two years of follow-up.ADHD, attention-deficit/hyperactivity disorder; GP, general practitioner.
AcknowledgementsThis paper represents independent research part funded by the NIHR Maudsley Biomedical Research Centre at South London and Maudsley NHS Foundation Trust and King's College London.The views expressed are those of the author(s) and not necessarily those of the NIHR or the Department of Health and Social Care.Clinical Practice Research Datalink (CPRD) data were provided by the NIHR Biomedical Research Centre, a partnership between Guy's and St Thomas' National Health Service (NHS) Foundation Trust and King's College London.Thanks are due to Dr Peter Schofield for initial statistical input to the analyses.Thanks are due to the Lambeth ADHD parent/caregiver support group for their comments and interpretation of the findings.Contributors VP conceived the idea for this study and supervised ER-P in the conduct of the analyses.VP is the guarantor and accepts full responsibilit for the work and the conduct of the study, had access to the data, and controlled the decision to publish.PW and VP undertook review of the code lists and reasons for attendance.JD, KS and ES-B consulted on the interpretation.ER-P, VP and NB drafted the submitted manuscript.All authors commented on the manuscript as submitted.FundingThis work was supported by the Academy of Medical Sciences Starter Grant.ER-P received salary funding from this grant.VP received salary funding as part of the National Institute for Health and Care Research (NIHR) academic clinical lecturer scheme, hosted by King's College London and as part of the NIHR Clinical Research Network East Midlands scholar scheme, hosted by University of Nottingham.Disclaimer This article/paper/report presents independent research funded by the NIHR.The views expressed are those of the author(s) and not necessarily those of the NHS, the NIHR or the Department of Health.Competing interests VP reports salary support from the National Institute for Health Research (NIHR) academic clinical lecturer scheme, NIHR East Midlands scholarship, a grant from the Academy of Medical Sciences and non-financial support from NIHR Biomedical Research Centre, during the conduct of the study.KS reports being a member of the NICE ADHD Guideline Development Group (NG87; 2016-2018).ES-B reports personal fees from Takeda(Shire), grants from QBTech, personal fees from Neurotech Solutions, personal fees from Medici, outside the submitted work.