Data were extracted from the CPRD (Clinical Practice Research Datalink) . CPRD contains clinically rich data collected in a non-interventional manner from the daily record-keeping of primary-care physicians in the UK. These data include demographics, medical history, test results, outpatient letters, and prescriptions. There are, in total, 143 million acceptable person-years of computerised data in CPRD, and the dataset is broadly representative of the UK population. Following record-linkage to NHS hospital episode statistics (HES), CPRD additionally contains details of inpatient admissions for a proportion of practices in England. The data extract used in this study includes records up to June 2012. Ethical approval for this study was granted by the CPRD Independent Scientific Advisory Committee on 1st March 2012, protocol number 12_025R2.
Patients were selected from CPRD if they had received two or more diagnoses for ADHD in their clinical history, or they had received at least one diagnosis of ADHD and at least one prescription for a medicine licensed for the management of ADHD. For cases where there was no prescription for an ADHD medication, the requirement of two or more diagnoses was used to avoid selecting for patients with only a provisional diagnosis recorded by the GP prior to assessment by a specialist. Under NICE guidelines, diagnosis should be made by a mental health specialist; therefore the second diagnosis is used to confirm that the patient has ADHD. A medicine used for the management of ADHD was defined as a product containing one of the following drugs: dexamfetamine, methylphenidate or atomoxetine. Pemoline (indicated for hyperkinetic syndrome but not generally available in the UK after 1997)  and modafinil (not licensed for the management of ADHD nor for use in children)  were not used for case selection. The study index date was the date of ADHD presentation, taken as the earlier of their first recorded diagnosis date for ADHD or their first prescription for a medicine used in the management of ADHD.
Cases were excluded from the analysis if they had a history of narcolepsy. In order to identify incident cases only, cases with less than six months’ “wash-in” for relevant parameters were also excluded (Additional file 1: Figure S1). No exclusion criteria based on age were implemented; however, the results were split by age group (0–5, 6–17 and ≥18 years) because licensed and recommended treatments vary by age. For example, atomoxetine and methylphenidate are not licensed in children younger than 6 years. In addition, the NICE guidelines do not recommend pharmacological treatment in preschool children. After school leaving age (≤18), NICE recommends that patients should be reassessed before transfer to adult services to ensure that continuing treatment into adulthood is still warranted and to facilitate transition. In addition, only atomoxetine is licensed for the treatment of ADHD in adults.
The healthcare costs and resource use of the ADHD group were compared to a randomly drawn control group of patients matched on year of birth, gender and GP practice. Control patients had no history of ADHD and had received no prescription for a medication indicated for ADHD.
Cases and controls were included in an annual cost calculation if they had a complete year’s observation for the year in question in both CPRD and CPRD-linked HES. Therefore, patients were excluded from the analysis of the costs for year 1 if they had less than 12 months’ observations from the index date to the last date of any prescription or the censor date, whichever was earlier. For year 2, patients were excluded if they did not have a complete year of data from 366 days to 730 days following their index date. The same rule was applied for the calculation of costs for years three through five.
Diagnostic incidence of ADHD
The incidence of diagnosed ADHD was calculated by dividing the number of new cases of ADHD each year by the number of person-years at risk in the CPRD data set for the same year (including those registered but with no GP attendance).
The number of person-years of people without ADHD was calculated by adding the number of days each patient had been present in the CPRD database for each specific year. Patients were included in the denominator until the earliest of their death date, transferred-out date, or ADHD presentation date. Patients who did not meet the selection criteria for the study were included in the denominator data.
On the date of ADHD presentation, cases were included in the numerator portion of the incidence calculation for that specific year.
The incidence of treated ADHD was calculated using the same method. For calculations of incidence by gender and age group only those patients of the appropriate age or gender were included in the numerator and denominator parts of the incidence calculation.
Diagnostic prevalence of ADHD
The point prevalence of diagnosed ADHD was calculated each year by dividing the number of patients with ADHD on 1st July (mid-year point) of that year by the total number of patients registered in CPRD on that date.
A patient was included as a prevalent case if they met the selection criteria for the study, their ADHD presentation date was prior to 1st July of the specific year, and the later of their last ADHD diagnosis or last prescription for an ADHD medication was after 1st of July of that year. However, in order to allow for an adequate washout period (more than 12 months), prevalence was only calculated from 1998 to 2009. A washout period was considered necessary as the chance of receiving a diagnosis for ADHD following the mid-year point reduces as the time between the mid-year point and the last collection date for the database becomes shorter.
This was the total number of patients registered in CPRD on 1st July of the specific year.
For calculations of prevalence by gender and age group only those patients of the appropriate age and gender were included in the numerator and denominator parts of the prevalence calculation.
Estimation of the cost of healthcare in CPRD
Resource use and costs were applied to the following areas of patient care: prescriptions, primary-care contacts, investigations, hospital admissions, and outpatient appointments. The aim was to calculate the overall cost of treating an individual with ADHD not just the cost of treating the ADHD itself. Annual costs for the first five years following the index date were estimated.
Each prescription item listed in CPRD was attributed a net ingredient cost (NIC) from the corresponding year of the Prescription Cost Analysis (PCA) [22, 23]. The NIC refers to the cost of the drug before discounts and does not include any dispensing costs or fees . All NICs were adjusted for inflation to 2011 prices . Either an exact match was made or the British National Formulary (BNF) taxonomy was utilised to attribute an average NIC per item for the BNF sub-paragraph, section or chapter.
Outpatient attendance costs
Outpatient events were identified from CPRD’s consultation table if they had a consultation type indicative or suggestive of an outpatient appointment. The outpatient department and whether the consultation was a first or follow-up visit were used to allocate each appointment to an outpatient tariff .
Cost of investigations
Investigations were identified, including both pathology and diagnostic services. Several reference sources were used to attribute a cost to these tests [26–29]. Laboratory tests carried out on the same day were grouped into test panels where appropriate in order to take account of any reduction in cost of carrying out more than one test at the same time .
Each consultation was classified by consultation type (e.g. surgery appointment, clinic, home visit, telephone consultation) and staff type (e.g. GP, practice nurse, mental health nurse, district nurse) and then assigned an average cost as listed in the Unit Cost of Health and Social Care 2010 from the Personal Social Services Research Unit (PSSRU) . Where average cost per hour was the only cost published in the Unit Costs of Health and Social Care, the UK GP workload survey  was utilised in order to determine the average length of the consultation. From this figure, the average cost per consultation could be calculated.
CPRD-linked HES records allowed us to cost inpatient admissions. From the care pathway outlined in the NICE guidelines for ADHD, we would not expect patients to be routinely admitted to hospital as a direct result of their ADHD . However, children and adolescents with ADHD may be more prone to other problems requiring admission such as accidents or self-harm .
Data from inpatient admissions recorded in HES were processed into Healthcare Resource Groups (HRGs) using HRG-4 grouper . The HRGs were then matched to NHS Reference Costs 2009–2010 . It was not possible to differentiate between elective or emergency day-case admissions from the data available, and so costs were averaged by ratio of each admission type. Data on procedures were not available and so all costs were inflated by 17.5%: the average difference between procedural and non-procedural admissions.