Article Text

Prescribing errors on admission to hospital and their potential impact: a mixed-methods study
  1. Avril Janette Basey1,2,
  2. Janet Krska3,
  3. Thomas Duncan Kennedy4,
  4. Adam John Mackridge2
  1. 1Pharmacy Department, Royal Liverpool University Hospital, Liverpool, UK
  2. 2School of Pharmacy and Biomolecular Sciences, Liverpool John Moores University, Liverpool, UK
  3. 3Medway School of Pharmacy, The Universities of Greenwich and Kent at Medway, Chatham, Kent, UK
  4. 4Acute Medical Unit, Royal Liverpool University Hospital, Liverpool, UK
  1. Correspondence to A J Basey, Pharmacy Department, Royal Liverpool University Hospital, Prescot Street, Liverpool L7 8XP, UK; A.Basey{at}


Background Medication errors are an important cause of morbidity and mortality and adversely affect clinical outcomes. Prescribing errors constitute one type of medication error and occur particularly on admission to hospital; little is known about how they arise.

Aim This study investigated how doctors obtain the information necessary to prescribe on admission to hospital, and the number and potential impact of any errors.

Setting English teaching hospital—acute medical unit.

Methods Ethics approval was granted. Data were collected over four 1-week periods; November 2009, January 2010, April 2010 and April 2011. The patient admission process was directly observed, field notes were recorded using a standard form. Doctors participated in a structured interview; case notes of all patients admitted during study periods were reviewed.

Results There were differences between perceived practice stated in interviews and actual practice observed. All 19 doctors interviewed indicated that they would sometimes or always use more than one source of information for a medication history; a single source was used in 31/68 observed cases. 7/12 doctors both observed and interviewed indicated that they would confirm medication with patients; observations showed they did so for only 2/12 patients. In 66/68 cases, the patient/carer was able to discuss medication, 14 were asked no medication-related questions. Of 688 medication charts reviewed, 318 (46.2%) had errors. A total of 851 errors were identified; 737/851 (86.6%) involved omission of a medicine; 94/737 (12.8%) of these were potentially significant.

Conclusions Although doctors know the importance of obtaining an accurate medication history and checking prescriptions with patients, they often fail to put this into practice, resulting in prescribing errors.

  • Medication reconciliation
  • Medication safety
  • Health services research
  • Pharmacists
  • Hospital medicine

Statistics from


Medication errors are recognised as an important cause of morbidity and mortality in hospitals. In recent years, awareness has been raised in England through government reports published in 2000,1 20012 and 2004.3 International guidance regarding the problems arising when patients transfer between care settings has been published by WHO.4 ,5 Medication errors, especially omission of prescribed medicines,6 have an adverse impact on clinical care and may also have a financial impact.7 ,8 Prescribing errors constitute one type of medication error and have been defined as occurring: ‘when, as a result of a prescribing decision or prescription writing process, there is an unintentional significant (1) reduction in the probability of treatment being timely and effective, or (2) increase in the risk of harm when compared with generally accepted practice’.9

WHO has defined unintentional medication discrepancies as those ‘in which the prescriber unintentionally changed, added or omitted a medication the patient was taking prior to admission’.5

A medication history should comprise a list of all medication currently being taken by the patient including any medication recently started by the general practitioner (GP), over the counter (OTC) medicines and herbal remedies.8 ,10 Medicines reconciliation on admission to hospital is the process of collecting information to prepare the patient's current medication history, verifying this list against the current hospital medication chart, identifying any discrepancies and taking appropriate action. Guidance is available for good medicines reconciliation practice which should result in an accurate prescription,5 ,10 however, there is little published data indicating how prescribers ensure that the prescription written on admission to hospital is correct.

This study investigated how doctors obtain the information necessary to write a prescription on admission to hospital. Outcomes were assessed in terms of the inpatient prescribing error prevalence (unintentional medication discrepancies) together with the potential patient impact.


Setting and overall methodology

The study was carried out in the acute medical unit (AMU) of a large English teaching hospital. On arrival at the AMU, patients are seen by a doctor who is responsible for taking a history (including medication), assessing the patient, making a provisional diagnosis, documenting a management plan, ordering initial investigations and writing the admission prescription. This process is known as clerking. Ward pharmacists routinely carry out medicines reconciliation as soon as possible following admission; they document and follow-up any discrepancies identified. At the time of the study, all prescriptions on admission were hand written by doctors on paper charts.

A triangulated mixed-methods approach was used involving: direct observation of the admission process, interviews with healthcare staff and case note review. There were four one-week study periods: November 2009 (1), January 2010 (2), April 2010 (3) and April 2011 (4). Study periods were selected to enable the participation of as many medical staff as possible within the constraints of the AMU rotas and to identify any changes over time. The range of study periods also helped overcome any bias resulting from seasonal variations in healthcare workload and experience of staff. Case note review and direct observation of a sample of admissions was carried out for all four study periods. Interviews were undertaken with all admitting staff observed during period 4 and further purposively selected staff to ensure that all grades working on AMU were included.


Staff gave informed consent for observations; patients or their carers could exclude the researcher at any time. They were purposively selected to maximise both the range of grades involved and variation in time and day of admission. Doctors clerk patients in order of arrival; hence patients whose admissions were observed were included if the clerking doctor agreed to participate in the study. Hospital staff were aware that the study concerned the hospital admission process but not specifically prescribing. During observations, data about medication were recorded on a standard form with additional field notes. The pharmacist researcher interacted socially with staff, but there was no professional interaction, unless inappropriate clinical management with potential for serious adverse consequences was detected. Any questions directed to the pharmacist researcher about medication were answered to avoid the focus of the study becoming known; all such incidents were recorded.


Interviews with staff took place as soon as practical following observation period 4, using a structured approach to ascertain their training experiences, perceptions of prescribing error rates and usual practices when taking a medication history and prescribing on admission (schedule available online).

Case note review

Case notes for all patients admitted during each study period were reviewed retrospectively, following discharge or death, to establish whether medicines were prescribed on admission, and whether any discrepancies were identified by the ward pharmacist during medicines reconciliation. All prescribing errors identified were documented in the case notes for review by medical staff; the appropriate doctor was alerted immediately by the ward pharmacist if urgent action was necessary, in accordance with routine pharmacy practice. If no medicines reconciliation had been carried out by a pharmacist during the hospital stay the notes were excluded from the analysis.

For the purposes of the study, prescribing errors were considered to be: unintentional omission of medication; unintentional prescribing of medication; incorrect medicine devices; and dose changes for which no justification could be identified. Minor discrepancies such as missing details for sustained release (SR) or enteric coated (EC) preparations, were not included as it was not possible to reliably identify these retrospectively.

The medicines reconciliation documentation in the case notes and pharmacist endorsements on prescriptions were reviewed and prescribing errors noted. All errors of omission were coded as red (significant or catastrophic, long-term patient impact); amber (significant, short-term patient impact), or green (negligible patient impact) using the UK Medicines Information (UKMI) tool for assessing possible harm from omitted or delayed medicines.11 This tool has clearly defined categories, requiring minimal interpretation, hence coding was carried out by only one researcher.

Data analysis

Data from case note reviews and observations from all four data collection periods were pooled, as there were no significant variations in working practices or procedures between the study periods. Data from the observations and interviews were categorised where necessary. Descriptive analysis was carried out using SPSS V.17; statistical tests were carried out using Minitab V.16. Where case notes and/or prescription charts were missing these cases were excluded from the relevant analyses.


The dataset for period 4 is slightly larger as more pharmacist hours were allocated to AMU resulting in more medication histories being checked (χ2 test p=0.035) and more patients experiencing an error being identified (χ2 test p=0.025), but other than this, similar numbers of admission observations, interviews and case note reviews were included in all four periods (table 1). The number of patients experiencing an error of omission and the proportion of red, amber and green errors were also broadly similar in all four periods (χ2 test p=0.201); the data were therefore pooled for analysis.

Table 1

Details of datasets and number of prescribing errors in each study period


A total of 68 admissions were observed, involving 35 doctors (four consultant/specialist registrar, four specialist trainee ST4/5 (6–7 years postregistration), nine specialist trainee ST1/2 (3–4 years postregistration) and 18 foundation F1/2 (0–2 years postregistration).

In 66/68 (97%) cases, the pharmacist researcher assessed that the patient or carer was able to discuss medication issues; two patients were too unwell to do so. However, 14/66 (21%) of the patients or their carers able to provide information were not asked about medicines; all but one of these were taking regular medication. Of the two patients unable to provide verbal information, the care home Medication Administration Record (MAR) chart was used for one and the medication chart from the referring hospital for the other.

No medication history was documented in the case notes for one patient who was discharged within a few hours. The most common sources used to obtain the medication history for the remaining 67 patients were: printed letters from the GP or Walk-In Centre (33/67; 49%); verbal information provided by patient (26/67; 39%); patients’ own medicines (19/67; 28%) and hand written letters from the GP (14/67; 21%). Only six patients had their GP repeat prescription order form with them; these were used in five cases. A single source was used to determine the medication history in almost half the cases (31/67; 46%); for 16/31 (52%) of these, additional sources were overtly available but not used. Two sources were used in 27/67 (40%) cases, three in eight (12%) cases and four in just one case. There was no significant difference in the number of sources used by F1/F2 doctors in comparison with other doctors (Mann–Whitney U test; p=0.904). The printed information provided in GP summaries was misinterpreted on nine occasions during eight patient admissions; examples are shown in box 1. The pharmacist researcher was asked for assistance with prescribing on 13 occasions (box 2), and prescribing errors witnessed are shown in box 3.

Box 1

Prescribing errors observed as a result of misinterpretation of printed general practitioner (GP) summaries

  • Summary states Fragmin (dalteparin) 25000 units/mL 0.6 mL daily.

  • Doctor initially prescribed 2500 units daily, and when challenged, changed this to 25 000 units daily; dose should be 15 000 units daily (research pharmacist intervened).

  • Buprenorphine patch 5 mcg/h weekly 4.

  • Doctor interpreted as four patches every week rather than one patch every week; 4 weeks supply.

  • Spironolactone two daily at the bottom of page 1; strength 25 mg at the top of page 2 of GP summary.

  • Doctor assumed strength was 100 mg and prescribed 200 mg daily; should be 50 mg daily.

  • Fluticasone 250 mcg/salmeterol 25 mcg inhaler (Seretide 250 evohaler) 2 puffs twice daily.

  • Prescribed as fluticasone 250 mcg inhaler 2 puffs twice daily.

Box 2

Research pharmacist assistance sought

  • Dose therapeutic dalteparin based on body weight

  • Dalterpain dose reduction for a patient with renal impairment

  • Identifying inhalers from patient descriptions of colour and shape

  • Confirming appropriateness of medication: Oramorph (morphine sulfate oral solution) for breathlessness in patient with severe COPD

  • Identification of white and yellow tablets in a blister pack

  • How to prescribe tiotropium inhaler 18 micrograms daily

  • Identification of new diabetes tablet beginning with ‘S’—sitagliptin

  • How to prescribe calcium carbonate 1.5 g/cholecalciferol 400 units—Adcal D3

  • To access general practitioner (GP) summary using EMIS web*—passwords not issued to rotational medical staff

  • Dose of paracetamol in liver disease?

  • Appropriate non-steroidal anti-inflammatory to be started

  • Dose of fondaparinux for probable pulmonary embolism as patient is allergic to dalteparin

  • Appropriate antibiotic for patient who has a chest infection, penicillin allergy and had recent course of erythromycin from GP

*Web-based computer system used by many GPs.

Box 3

Additional prescribing errors witnessed during writing of admission prescription

  • Tiotropium inhaler missed off prescription

  • MST (morphine sulfate sustained release) prescribed as 40 mg Mane and 30 mg Nocte using a blister pack; should be 30 mg twice daily but 40 mg twice daily on Mondays and Thursdays when the patient has dressing changes

  • Ramipril prescribed 5 mg daily; should be twice daily and furosemide 20 mg three daily prescribed as 20 mg mane

  • Tacrolimus prescribed; should be prescribed by brand—Prograf

  • Calcichew prescribed; should be cinacalcet (research pharmacist intervened)

  • Calcichew prescribed should be Calcichew D3 Forte

  • Regular medication fluoxetine and vitamin B compound strong omitted for no apparent reason

A prescription chart was written by the admitting doctor for 56/68 (82%) patients; no chart was written for 10/68 (15%) patients, and another doctor had already written the chart elsewhere in the hospital in 2/68 (3%) cases. The prescription written on admission was confirmed with the patient in only 12/56 (21%) cases; in 37/56 (66%) cases, the prescriber made no attempt to confirm that the medicines prescribed matched those which the patient was actually taking. In two (2/56; 4%) cases, only urgently required medication was prescribed, and in five (5/56; 9%) cases confirmation was not possible due to illness. Seven of the 12 (58%) doctors both observed and interviewed indicated that they would confirm medication with the patient before prescribing; however, the observations showed they did so for only 2/12 (17%) patients whom they admitted.

Medicines reconciliation was completed by a pharmacist for 42 of the 56 (75%) medication charts written during the observations; 25/42 (59%) were accurate, eight (8/42; 19%) contained one prescribing error, seven (7/42; 17%) had two errors and two (2/42; 5%) had three errors.


Nineteen doctors were interviewed, comprising 73% of the 26 working on AMU during study period 4: two consultants, nine specialist trainees and eight foundation years. Twelve of the 19 were responsible for admitting 20 of the patients observed in the study. Sixteen (16/19; 84%) reported receiving undergraduate training in medication history-taking; however, 9/16 (56%) were unable to recall the details. The majority of doctors were unaware of the proportion of patients at risk, with 13/19 (68%) estimating that no more than 30% of medication charts written on admission would have a prescribing error, and 16/19 (84%) estimating that fewer than 10% of such errors may be potentially serious.

When asked to list the sources they usually use to obtain a medication history, the most common responses were the patient (17/19; 89%), patients’ own medicines (16/19; 84%), GP repeat medication order form (15/19; 79%), previous discharge prescription (14/19; 74%), telephone GP for information (13/19; 68%), GP/Walk-In Centre letter (10/19; 53%).

Fourteen of the 19 interviewees indicated they would sometimes use more than one source to check a medication history; a further four said that they would always do so. Common reasons given for using more than one source were: information given by patients is not reliable (6/19; 32%) and patients may not take their medication as prescribed (3/19; 16%). Five interviewees stated that clinical anomalies also prompted them to check medication thoroughly. Examples were: a patient with epilepsy who has brought their own medicines but none are antiepileptic medicines; a patient taking letrozole but no history of breast cancer. Warfarin and insulin were cited as causing particular problems in identifying the current dosage regimen, and the difficulties in obtaining accurate information outside of normal working hours when GP surgeries are closed were highlighted.

The majority (11/19; 58%) said that they would ‘sometimes’ confirm all regular medication taken with patients before writing the admission prescription, and some (6/19; 32%) said that they would discuss newly initiated medicines but not the patients’ ‘regular’ medicines. Reasons given for not discussing with patients were: incapacity due to illness (8/19; 42%) and too time consuming (1/19; 5%).

All 19 doctors thought that prescriptions should be checked for accuracy and appropriateness; 17/19 (89%) indicated that this should take place within 24 h of prescribing, and 18/19 (95%) felt that pharmacists were the most appropriate professionals to perform the check. One doctor felt that checking on the next working day would be adequate, and another that 24–48 h after prescribing was appropriate. However, it was recognised that anyone involved with medication should also take the opportunity to check, for example, doctors on ward rounds and nurses administering medicines. Five doctors spontaneously indicated that they had a responsibility to self-check prescriptions that they had written.

Suggestions for reducing prescribing errors included: better access to GP prescription data (6/19; 32%) especially out of hours; integrated Information technology (IT) systems (2/19; 11%) and improved training for medical students and F1 doctors (10/19; 53%). Two senior interviewees suggested that increasing availability of pharmacists to provide accurate medication histories prior to patients being clerked by a doctor may be helpful.

Case note review

A total of 1015 patient case notes were identified of which 930 (91.6%) were followed-up. In 54 cases, the relevant admission documentation was not available, and for a further 66 the original medication chart was missing, leaving 810 cases suitable for analysis. Medicines reconciliation was completed by a pharmacist for 688/810 (84.9%) of patients; 4155 medicines should have been prescribed (average 6.0 per patient) and 851 errors were identified, therefore, 20.5% of items which should have been prescribed had an error. The errors involved 318/688 (46.2%) patients; each of whom experienced an average of 2.7 errors; the most common error was omission of a medicine (737/851; 86.6%). The overall error rate was 1.2 errors per patient for whom medicines reconciliation was completed.

Details of the types of prescribing errors identified and the potential impact of omissions are shown in table 2; 94/737 (12.8%) of omissions were classified as having the potential to have a significant long-term or short-term effect. Most of those classified as red (significant or catastrophic, long-term patient impact) involved antiepileptic medicines (5/7). The majority of errors (502/851; 59%) were rectified within 24 h, and over two-thirds (587/851; 69.0%) within 48 h of being identified and highlighted by pharmacists.

Table 2

Details of type of prescribing errors identified during case note review and potential impact of errors of omission


This study provides a novel insight into how prescriptions are written, and the possible causes of prescribing errors on admission to hospital. It adds to the published literature regarding the proportion of medical patients who experience a prescribing error, and the potential impact of these errors.

The sources actually used for obtaining medication histories during the observations matched those most frequently cited by staff during the interviews as being sources they commonly used. Although almost all doctors interviewed indicated that they would sometimes or always use more than one source to confirm medication histories, the observations showed that a single source was used in almost half the cases. This is at variance with national guidance for England and Wales8 and other published guidance10 ,12 ,13 which suggests that at least two sources should be used. WHO definition for ‘Best Possible Medication History’ states that the patient should be interviewed where possible, and advocates the use of more than one source.5

From the observations it was apparent that several patients or carers were able to provide information about medicines but were asked no relevant questions, despite this being an integral part of the standard hospital clerking model.14 The numbers in the study were too small to suggest any particular reason for this omission, however, during the interviews six doctors did allude to the unreliability of information provided by patients.

Although seven of the 12 doctors both interviewed and observed indicated that they would confirm current medication with the patient, the observations showed that these doctors only did so in two of the 12 patients they admitted between them, suggesting that although the theory is understood, application in practice was not simple. Overall, confirmation of the prescription with the patient occurred infrequently despite overt acknowledgement by three doctors that patients may not take their medicines as prescribed; a UK study has shown that up to 11% of errors identified on admission may result from a patient decision to alter their treatment regimen.15 One doctor commented that time pressures were an issue when talking to patients about medication. A recent study showed that medication history-taking for medical patients takes 10–20 min15; national guidance for England and Wales suggests that 15 min is needed for the ‘average’ non-elective patient.16 Both the EQUIP study17 and the PROTECT programme18 reported time pressures and high workload as being contributory factors to prescribing errors.

The error rates found in the present study are similar to those reported elsewhere, although there are difficulties in making comparisons between studies as ‘prescribing error’ is not always defined19 and results may be expressed in different ways.20 ,21 The present study found prescribing errors in 20.5% of the medicines which should have been prescribed which is comparable with results published in a recent English study which reported a rate of 16.3% for medical admissions.22 A systematic review found that overall prescribing errors affect 50% of patients20 which is similar to the 46.2% of patients affected in the present study. The most common error in the present study was omission of a medicine usually taken by the patient (86.6%), which is in line with the findings of studies from Belgium,23 Sweden24 and Wales.25

All doctors, on qualification, should be able to establish an accurate medication history26; this has been highlighted as a core skill necessary for safe prescribing by the British Pharmacological Society.27 ,28 Limited information is available in the literature regarding the most effective way to train medical students to prescribe29 ,30; only two papers providing specific medicines reconciliation guidance for medical students or junior doctors have been identified.12 ,13 These papers confirm the need for at least two sources to be used and highlight some of the common pitfalls.

As the majority of the prescribing errors in the present study were omissions, a tool specifically developed to estimate the potential impact of omitted medicines was used to categorise them11; 12.8% of omissions were assessed as red or amber and, therefore, had the potential for some clinical impact on the patient. However, the majority of omissions were likely to have a minimal impact which is in line with the findings of a recent meta-analysis.31 Few studies have attempted to assess the impact of prescribing errors and those that have used different tools. A study from Wales using an adapted version of a tool developed by the National Patient Safety Agency32 classified 20% of errors as ‘major’ or ‘moderate’,25 other studies using consensus panels to estimate impact have reported 32.9% of errors could potentially cause moderate discomfort or clinical deterioration,33 and 26% were potentially serious.34 The majority of doctors interviewed were unaware of the proportion of patients at risk of prescribing errors, with most estimating error rates of below 30%, in contrast with the 50% reported in the literature20 and 46.2% found in the present study.

A meta-analysis published in 2012 concluded that there are limited data regarding the most effective interventions to improve medicines reconciliation,35 however, the present study did suggest some actions which may prove particularly successful. Raising awareness of both the level of risk and the potential seriousness of many errors may help to reduce error rates, but doctors may also require practical guidance regarding the need to check more than one source, and especially the need to confirm the medication history with the patient before prescribing, whenever possible. Training in areas where knowledge was found to be lacking, for example, colours and types of inhaler, preferably by supervising medical students while taking medication histories and providing feedback to staff about actual errors may also be beneficial. However, training alone may not result in a significant reduction in errors as doctors appeared to know the theory but failed to apply it in practice, which suggests that other factors are contributing to the problem. Staff comments about difficulties arising outside of normal working hours when access to GP information was limited are also important considerations. Expanding the use of electronic systems, such as EMIS web,36 or a system similar to the emergency care summary used in Scotland,37 and facilitating access to GP records for junior staff who provide the ‘out of hours’ services in hospitals, may go some way to addressing this issue.

Earlier involvement of the pharmacy team in the admission process was suggested by two senior doctors. Studies from Scotland,38 Australia39 and the USA40 have shown that fewer doses are missed if a pharmacist completes a medication history in the ED, before relatives leave, taking with them the vital information available from patients’ own medication. Studies in the UK,41 ,42 USA43 and Belgium23 have demonstrated that pharmacist-documented medication histories are more accurate than those gathered by doctors. This finding is supported by the requests for assistance from the pharmacist researcher during 13 (19%) of the 68 admissions observed and the need to intervene on two occasions to prevent a serious prescribing error. Perhaps it is time to rethink the patient journey on admission to hospital, and involve pharmacy staff in obtaining a medication history on AMU before the patient is seen by the doctor, to prevent prescribing errors arising rather than identifying and correcting them later.

Strengths and limitations

The main strength of this study is in the triangulation of data derived from interviews with a proportion of the staff observed, helping to explain some of the findings from the observations, while the case note review provided real outcome data. This is in contrast with many published studies which focus on the number of prescribing errors rather than investigating the cause of such errors.

Limitations are that the study was carried out in one hospital and involved relatively small numbers of observations and interviews, therefore, the results may not be representative of other hospitals. Only doctors observed in period 4 were interviewed, but additional interviews maximised the proportion of AMU staff included. The interviews were carried out sequentially, and although all staff agreed to keep the subject matter confidential, it is impossible to be certain that confidentiality was maintained, however there is no evidence that this impacted on the data integrity.

Minor discrepancies, such as missing SR or EC preparations, were excluded, which may have resulted in a reduced number of prescribing errors being recorded. Independent pharmacist medicines reconciliation was only available for 67% of the total number of patients admitted during the study periods, due, in part, to limitations in the capacity of the pharmacy service and unavailability of the necessary documentation. However, as patients’ mean age was very similar for both groups, and the three most common presenting complaints were identical, there is no reason to suspect that either the number of regular medicines or the number of prescribing errors would differ between those patients whose prescriptions were, and those whose were not, reviewed by a pharmacist.

The researcher is a member of the AMU staff which may have impacted on behaviour during observations.


The study interviews showed that medical staff have the necessary knowledge to establish an accurate medication history and are aware of the potential pitfalls, but observations showed that theoretical knowledge is frequently not put into practice. Therefore, a reduction in prescribing errors could be achieved if a mechanism can be found to implement existing guidance effectively. Improved awareness training highlighting the extent of the problem may be beneficial, but improving access to patient medication histories and alternative strategies for involving pharmacists should also be considered.


The authors thank Dr PH Rowe, Reader, School of Pharmacy and Biomedical Sciences Liverpool John Moores University for assistance with statistical analysis.


Supplementary materials

  • Supplementary Data

    This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Files in this Data Supplement:


  • Contributors AJB, JK, TDK and AJM were responsible for the planning and design of the study, AJB was responsible for data collection and analysis. AJB wrote the first draft; JK, TDK and AJM provided critical revision, all authors read and approved the final manuscript.

  • Funding The study was supported by an unrestricted educational grant from Pfizer UK paid to the Royal Liverpool University Hospital.

  • Competing interests The authors declare that: AJB had an unrestricted educational grant from Pfizer UK paid to the Royal Liverpool University Hospital for the submitted work, PhD fees were paid by the Royal Liverpool and Broadgreen University Hospitals NHS Trust.

  • Ethics approval The study was approved by the National Research Ethics Service (Liverpool Central REC) and Liverpool John Moores University Ethics Committee. Research Governance approval was granted by Royal Liverpool University Hospital.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement Participants gave informed consent for data sharing in publications. Consent was not obtained for the sharing of the dataset, but the presented data are anonymised and the risk of identification is low; dataset available from the corresponding author at

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.