Challenges and recommendations for high quality research using electronic health records.

Harnessing Real World Data is vital to improve health care in the 21st Century. Data from Electronic Health Records (EHRs) are a rich source of patient centred data, including information on the patient's clinical condition, laboratory results, diagnoses and treatments. They thus reflect the true state of health systems. However, access and utilisation of EHR data for research presents specific challenges. We assert that using data from EHRs effectively is dependent on synergy between researchers, clinicians and health informaticians, and only this will allow state of the art methods to be used to answer urgent and vital questions for patient care. We propose that there needs to be a paradigm shift in the way this research is conducted - appreciating that the research process is iterative rather than linear. We also make specific recommendations for organisations, based on our experience of developing and using EHR data in trusted research environments.

Impact of penicillin allergy records on antibiotic costs and length of hospital stay: a single-centre observational retrospective cohort.

INTRODUCTION: Patients with a penicillin allergy record are usually prescribed non-penicillin antibiotics and have worse health outcomes. This study explored the impact of penicillin allergy records on antibiotic treatment costs and patient length of stay. METHODS: Patients prescribed a systemic antibacterial agent between April 2016 and March 2018 in a 750-bed English hospital were included in this study. The following data were extracted for each patient: age, sex, comorbidities, infection treated, antibiotic usage (defined daily dose), hospital length of stay and penicillin allergy status. Multi-variable log-linear modelling was used to determine associations between patients labelled as penicillin allergic and total antibiotic costs and length of stay. Using the above models, the potential reductions in total cost and hospital bed-days of 'delabelling' patients with a penicillin allergy record were estimated. RESULTS: Penicillin allergy records were present in 14.3% of hospital admissions and were associated with an increase in non-penicillin antibiotic prescribing, a 28.4% increase in antibiotic costs and 5.5% longer length of hospital stay compared with patients without a penicillin allergy record. Patients with a penicillin allergy record accounted for an excess antibiotic spend of £10,637 (2.61% of annual antibiotic drug spend) and 3522 excess bed-days (3.87% of annual bed-days). Delabelling 50% of patients with a self-reported penicillin allergy record would save an estimated £5501 in antibiotic costs and £503,932 through reduced excess bed-days. CONCLUSION: Delabelling patients with a self-reported allergy record has potential to reduce antibiotic costs, but its biggest cost impact is via a reduction in excess bed-days.

A retrospective impact analysis of the WannaCry cyberattack on the NHS.

A systematic analysis of Hospital Episodes Statistics (HES) data was done to determine the effects of the 2017 WannaCry attack on the National Health Service (NHS) by identifying the missed appointments, deaths, and fiscal costs attributable to the ransomware attack. The main outcomes measured were: outpatient appointments cancelled, elective and emergency admissions to hospitals, accident and emergency (A&E) attendances, and deaths in A&E. Compared with the baseline, there was no significant difference in the total activity across all trusts during the week of the WannaCry attack. Trusts had 1% more emergency admissions and 1% fewer A&E attendances per day during the WannaCry week compared with baseline. Hospitals directly infected with the ransomware, however, had significantly fewer emergency and elective admissions: a decrease of about 6% in total admissions per infected hospital per day was observed, with 4% fewer emergency admissions and 9% fewer elective admissions. No difference in mortality was noted. The total economic value of the lower activity at the infected trusts during this time was £5.9 m including £4 m in lost inpatient admissions, £0.6 m from lost A&E activity, and £1.3 m from cancelled outpatient appointments. Among hospitals infected with WannaCry ransomware, there was a significant decrease in the number of attendances and admissions, which corresponded to £5.9 m in lost hospital activity. There was no increase in mortality reported, though this is a crude measure of patient harm. Further work is needed to appreciate the impact of a cyberattack or IT failure on care delivery and patient safety.

One-year outcomes for congenital diaphragmatic hernia.

Background: Congenital diaphragmatic hernia (CDH) is a congenital anomaly with high mortality and long-term morbidity. The aim of this study was to benchmark trends in 1-year and hospital volume outcomes for this condition. Methods: This study included all infants born with CDH in England between 2003 and 2016. This was a retrospective analysis of the Hospital Episode Statistics database. The main outcomes were: 1-year mortality, neonatal length of hospital stay (nLOS), total bed-days at 1 year and readmission rate. The association between hospital volume and outcomes was assessed for specialist paediatric surgery centres. Results: A total of 2336 infants were included (incidence 2·5 per 10 000 live births). No significant time trends were found in incidence and main outcomes. Some 1491 infants (63·8 per cent) underwent surgical repair. The 1-year mortality rate was 31·2 per cent. Median nLOS and total bed-days were 17 and 19 days respectively. The readmission rate in specialist paediatric centres was 6·3 per cent. Higher mortality was associated with birthweight lower than 1 kg (OR 5·90, 95 per cent c.i. 1·03 to 33·75), gestational age of 36 weeks or less (OR 1·75, 1·12 to 2·75) and black ethnicity (OR 2·13, 1·03 to 4·48). Only 4·0 per cent had extracorporeal membrane oxygenation, which was associated with higher mortality (OR 5·34, 3·01 to 9·46), longer nLOS (OR 3·70, 2·14 to 6·14) and longer total bed-days (OR 3·87, 2·19 to 6·83). Specialist paediatric centres showed variation in 30-day mortality (4·6 per cent with 84 per cent coefficient of variation), nLOS (median 25 (i.q.r. 15-42) days) and total bed-days (median 28 (i.q.r. 16-51) days), but no significant volume-outcome relationship. Conclusion: Key outcomes for CDH were similar to those of other developed countries. High variation among specialist paediatric centres was found and should be investigated further to explore the value of regionalization of care.

Factors that impact on the burden of Escherichia coli bacteraemia: multivariable regression analysis of 2011-2015 data from West London.

BACKGROUND: The incidence of Escherichia coli bacteraemia in England is increasing amid concern regarding the roles of antimicrobial resistance and nosocomial acquisition on burden of disease. AIM: To determine the relative contributions of hospital-onset E. coli bloodstream infection and specific E. coli antimicrobial resistance patterns to the burden and severity of E. coli bacteraemia in West London. METHODS: Patient and antimicrobial susceptibility data were collected for all cases of E. coli bacteraemia between 2011 and 2015. Multivariable logistic regression was used to determine the association between the category of infection (hospital or community-onset) and length of stay, intensive care unit admission, and 30-day all-cause mortality. FINDINGS: E. coli bacteraemia incidence increased by 76% during the study period, predominantly due to community-onset cases. Resistance to quinolones, third-generation cephalosporins, and aminoglycosides also increased over the study period, occurring in both community- and hospital-onset cases. Hospital-onset and non-susceptibility to either quinolones or third-generation cephalosporins were significant risk factors for prolonged length of stay, as was older age. Rates of mortality were 7% and 12% at 7 and 30 days, respectively. Older age, a higher comorbidity score, and bacteraemia caused by strains resistant to three antibiotic classes were all significant risk factors for mortality at 30 days. CONCLUSION: Multidrug resistance, increased age, and comorbidities were the main drivers of adverse outcome. The rise in E. coli bacteraemia was predominantly driven by community-onset infections, and initiatives to prevent community-onset cases should be a major focus to reduce the quantitative burden of E. coli infection.