Saud Jukaku is a GP (Family Medicine consultant) with an interest in obesity at the King Faisal Specialist Hospital & Research Centre, Al Madinah, Saudi Arabia.
Amrit Lamba is a GP partner at Colindale Medical Centre, North Central London ICB. He has an extended role in primary care cardiometabolic medicine.
We begin with a sad and cautionary case study. In June 2021, Alexander Reid, a 28-year-old man, suffered a fatal cerebral venous sinus thrombosis secondary to Covid-19 Vaccine-Induced Immune Thrombotic Thrombocytopenia.1 His case offers an important learning point.
In April 2021, the Joint Committee on Vaccination and Immunisation had published a statement advising against the use of the AstraZeneca COVID-19 vaccine for adults aged under 30 years old who did not have any underlying health conditions that put them at a higher risk of severe COVID-19 disease.2 However, those who had already received this vaccine as their first dose were advised to receive it as their second dose. Alexander Reid received his second dose of the vaccine on 18 May 2021, and died six weeks later.
The clinician in 2004 had entered his height as 145 cm and his weight as 145 kg, giving a body mass index (BMI) of 69 kg/m² for a boy aged only 11 years old.
Regardless of how safe or risky the vaccine was in general, Alexander Reid, as it turned out, was not eligible to receive the AstraZeneca COVID-19 vaccine at the time when he received it. Our cautionary note relates not to vaccine safety but to the clerical error carried forward by automated systems in primary healthcare.
Alexander Reid received his first dose of the Oxford AstraZeneca vaccine in March 2021, earlier than his age alone would have entitled him to, because of an error made in his GP records seventeen years earlier, in 2004. The clinician in 2004 had entered his height as 145 cm and his weight as 145 kg, giving a body mass index (BMI) of 69 kg/m² for a boy aged only 11 years old. This erroneous BMI was never corrected and resulted in Alexander being classed as a vulnerable person and being incorrectly invited to receive a COVID-19 vaccination earlier than he should have.
The Prevention of Future Deaths Report calls for primary care IT systems with the ability to ‘challenge’ grossly abnormal values from being entered into a patient’s medical records. We strongly agree with such a recommendation, as this is a common problem universally encountered by medical practices.
In 2021, the same year as Alexander Reid’s death, we delivered a programme at a GP surgery in London offering virtual group clinics (VGC) to people living with obesity and a BMI >35 kg/m². During the assessment of patients meeting the programme’s inclusion criteria, we identified several patients with an apparently elevated BMI who in fact had a normal-range weight; historical errors appeared to have been made during the entry of height and weight data into the electronic medical records (EMR). Such instances had not been challenged by the IT system at the time of data entry. Conversely, it appeared likely that some patients were inappropriately excluded from invitation to the programme due to falsely normal BMIs being recorded. We wonder what our colleagues might find if they audited this kind of data for accuracy.
The use of information technology and EMRs in healthcare brings many potential benefits including improving patient safety, efficiency and accuracy of data but it can also create new opportunities for errors.3 This can result in harm to patients as well as affect the validity of clinical research,4 the ability to determine operational performance and financial loss due to underpayment.5
Incorrect recording of weight (or height) in EMRs is potentially an easy mistake to make, requiring only an extra digit or a decimal point in the wrong place to have a significant impact on the final value. The resultant effect for patients extends beyond their classification of disease vulnerability and risk, to include safe medication dose calculations, particularly in paediatric, oncology and critical care settings. A 2017 study on manual transcription errors by healthcare personnel performing point-of-care laboratory tests showed that 3.7% of manual entries were erroneous and 0.5% potentially dangerous, a finding consistent with similar studies in the field.6
The Prevention of Future Deaths Report calls for primary care IT systems with the ability to ‘challenge’ grossly abnormal values from being entered into a patient’s medical records.
Errors in non-clinical data can also have the potential to impact patient care. For example, our VGC programme identified several patients with a BMI >35 kg/m² who did not receive the SMS invitation to the programme because of inaccurate contact details on the GP record. Such errors can either be due to incorrect data entry by GP practice staff or inaccurate data provision by patients during the registration process. Despite accurate patient contact details being essential for communicating with patients in urgent and routine health matters, their erroneous entry in EMRs is regularly encountered across healthcare organisations. One study reported that of 1136 patients who gave their phone numbers during an emergency department visit, only 42.1% could be contacted a week later and almost 28% of the numbers were either wrong or disconnected.7
We suggest three solutions for reducing data entry errors in cases such as the ones mentioned in this article:
- Automating data entry by using direct data transfer from medical devices to EHRs.
- Data validation and verification via machine learning. EHRs should be able to assess and highlight if an entry is likely to be erroneous, whether that is because it is significantly different to a previous reading for the same patient or whether it is generally an unexpected reading in the population as a whole (Such as a BMI of 69).
- Having certain fields that must be filled in and with a certain number of digits e.g. telephone numbers
We welcome a discussion of the potential for automation to amplify human error, discussion of our proposed solutions and other strategies to prevent the problem or harms arising from it.
In a data-driven healthcare system, the accuracy of clinical and non-clinical values, as well as diagnostic coding entered into EMRs, is crucial for delivering appropriate and timely care. Developing “live” IT safety-netting systems to challenge data entry at the time of input is essential to reduce the negative impact of human error on morbidity and mortality. This approach not only mitigates risks but also enhances the potential of data-rich EMRs to improve health outcomes for individuals living with or at risk of disease.
Authors’ DOI: Saud Jukaku declares no competing financial interests. Amrit Lamba reports honoraria from AstraZeneca, grants from Boehringer Ingelheim, honoraria from Boehringer Ingelheim, honoraria from Janssen, honoraria from Eli Lilly, grants from NAPP, honoraria from NAPP, grants from Novo Nordisk, honoraria from Novo Nordisk, and honoraria from Sanofi outside the submitted work.
References
- Courts and Tribunals Judiciary. Alexander Reid: Prevention of future deaths report. April 29, 2024. Available at: https://www.judiciary.uk/prevention-of-future-death-reports/alexander-reid-prevention-of-future-deaths-report/
- NHS England. Frequently Asked Questions: MHRA and JCVI guidance on AstraZeneca COVID-19 vaccine and very rare clotting disorders. April 2021. Available at: https://www.england.nhs.uk/south/wp-content/uploads/sites/6/2021/04/C1249-FAQs-MHRA-and-JCVI-guidance-on-AstraZeneca-COVID-19-vaccine-and-very-rare-clotting-disorders.pdf
- Kaihlanen A, Gluschkoff K, Saranto K et al. The associations of information system’s support and nurses’ documentation competence with the detection of documentation-related errors: Results from a nationwide survey. Health Informatics Journal. 2021. 27(4): 1-12. Available at doi: 10.1177/14604582211054026
- Goldberg S, Niemierko A, Turchin A. Analysis of data errors in clinical research databases. AMIA Annu Symp Proc. 2008 Nov 6;2008:242-6.
- Qian S, Munyisia E, Reid D et al. Trend in data errors after the implementation of an electronic medical record system: A longitudinal study in an Australian regional Drug and Alcohol Service. International Journal of Medical Informatics. 2020. 144: 104292. Available at doi: 10.1016/j.ijmedinf.2020.104292
- Mays J and Mathias P. Measuring the rate of manual transcription error in outpatient point-of-care testing. Journal of the American Medical Informatics Association. 2019. 26(3): 269-272. Available at doi: 10.1093/jamia/ocy170
- Boudreaux E, Ary R, St John B et al. Telephone contact of patients visiting a large, municipal emergency department: can we rely on numbers given during routine registration? Journal of Emergency Medicine. 2000. 18(4): 409-15. Available at doi: 1016/s0736-4679(00)00155-4
