A survey on clinical natural language processing in the United Kingdom from 2007 to 2022.

Much of the knowledge and information needed for enabling high-quality clinical research is stored in free-text format. Natural language processing (NLP) has been used to extract information from these sources at scale for several decades. This paper aims to present a comprehensive review of clinical NLP for the past 15 years in the UK to identify the community, depict its evolution, analyse methodologies and applications, and identify the main barriers. We collect a dataset of clinical NLP projects (n = 94; £ = 41.97 m) funded by UK funders or the European Union's funding programmes. Additionally, we extract details on 9 funders, 137 organisations, 139 persons and 431 research papers. Networks are created from timestamped data interlinking all entities, and network analysis is subsequently applied to generate insights. 431 publications are identified as part of a literature review, of which 107 are eligible for final analysis. Results show, not surprisingly, clinical NLP in the UK has increased substantially in the last 15 years: the total budget in the period of 2019-2022 was 80 times that of 2007-2010. However, the effort is required to deepen areas such as disease (sub-)phenotyping and broaden application domains. There is also a need to improve links between academia and industry and enable deployments in real-world settings for the realisation of clinical NLP's great potential in care delivery. The major barriers include research and development access to hospital data, lack of capable computational resources in the right places, the scarcity of labelled data and barriers to sharing of pretrained models.

Recording problems and diagnoses in clinical care: developing guidance for healthcare professionals and system designers.

BACKGROUND: Accurate recording of problems and diagnoses in health records is key to safe and effective patient care, yet it is often done poorly. Electronic health record systems vary in their functionality and ease of use, and are not optimally designed for easy recording and sharing of clinical information. There is a lack of professional consensus and guidance on how problems and diagnoses should be recorded. METHODS: The Professional Record Standards Body commissioned work led by the Royal College of Physicians Health Informatics Unit to carry out a literature review, draft guidance, carry out an online consultation and round table discussion, and produce a report including recommendations for systems. A patient workshop was held to explore patient preferences for mechanisms for sharing diagnosis information between primary and secondary care. RESULTS: Consensus was reached among medical specialties on key elements of diagnosis recording, and draft guidance was produced ready for piloting in a variety of care settings. Patients were keen for better ways for diagnosis information to be shared. DISCUSSION: Improving the recording of diagnoses and problems will require a major effort of which the new guidance is only a part. The guidance needs to be embedded in training, and clinical systems need to have improved, standardised functionality. Front-line clinicians, specialist societies, clinical informaticians and patients need to be engaged in developing information models for diagnoses to support care and research, accessible via user-friendly interfaces.

Ethnicity and the first diagnosis of a wide range of cardiovascular diseases: Associations in a linked electronic health record cohort of 1 million patients.

BACKGROUND: While the association of ethnic group with individual cardiovascular diseases has been studied, little is known about ethnic differences in the initial lifetime presentation of clinical cardiovascular disease in contemporary populations. METHODS AND RESULTS: We studied 1,068,318 people, aged ≥30 years and free from diagnosed CVD at baseline (90.9% White, 3.6% South Asian and 2.9% Black), using English linked electronic health records covering primary care, hospital admissions, acute coronary syndrome registry and mortality registry (CALIBER platform). During 5.7 years median follow-up between 1997-2010, 95,224 people experienced an incident cardiovascular diagnosis. 69.9% (67.2%-72.4%) of initial presentation in South Asian <60 yrs were coronary heart disease presentations compared to 47.8% (47.3%-48.3%) in White and 40.1% (36.3%-43.9%) in Black patients. Compared to White patients, Black patients had significantly lower age-sex adjusted hazard ratios (HRs) for initial lifetime presentation of all the coronary disease diagnoses (stable angina HR 0.80 (95% CI 0.68-0.93); unstable angina- 0.75 (0.59-0.97); myocardial infarction 0.49 (0.40-0.62)) while South Asian patients had significantly higher HRs (stable angina- 1.67 (1.52-1.84); unstable angina 1.82 (1.56-2.13); myocardial infarction- 1.67 (1.49-1.87). We found no ethnic differences in initial presentation with heart failure (Black 0.97 (0.79-1.20); S Asian 1.04(0.87-1.26)). Compared to White patients, Black patients were more likely to present with ischaemic stroke (1.24 (0.97-1.58)) and intracerebral haemorrhage (1.44 (0.97-2.12)). Presentation with peripheral arterial disease was less likely for Black (0.63 (0.50-0.80)) and South Asian patients (0.70 (0.57-0.86)) compared with White patients. DISCUSSION: While we found the anticipated substantial predominance of coronary heart disease presentations in South Asian and predominance of stroke presentations in Black patients, we found no ethnic differences in presentation with heart failure. We consider the public health and research implications of our findings. TRIAL REGISTRATION: NCT02176174, www.clinicaltrials.gov.

Neutrophil Counts and Initial Presentation of 12 Cardiovascular Diseases: A CALIBER Cohort Study.

BACKGROUND: Neutrophil counts are a ubiquitous measure of inflammation, but previous studies on their association with cardiovascular disease (CVD) were limited by small numbers of patients or a narrow range of endpoints. OBJECTIVES: This study investigated associations of clinically recorded neutrophil counts with initial presentation for a range of CVDs. METHODS: We used linked primary care, hospitalization, disease registry, and mortality data in England. We included people 30 years or older with complete blood counts performed in usual clinical care and no history of CVD. We used Cox models to estimate cause-specific hazard ratios (HRs) for 12 CVDs, adjusted for cardiovascular risk factors and acute conditions affecting neutrophil counts (such as infections and cancer). RESULTS: Among 775,231 individuals in the cohort, 154,179 had complete blood counts performed under acute conditions and 621,052 when they were stable. Over a median 3.8 years of follow-up, 55,004 individuals developed CVD. Adjusted HRs comparing neutrophil counts 6 to 7 versus 2 to 3 × 109/l (both within the 'normal' range) showed strong associations with heart failure (HR: 2.04; 95% confidence interval [CI]: 1.82 to 2.29), peripheral arterial disease (HR: 1.95; 95% CI: 1.72 to 2.21), unheralded coronary death (HR: 1.78; 95% CI: 1.51 to 2.10), abdominal aortic aneurysm (HR: 1.72; 95% CI: 1.34 to 2.21), and nonfatal myocardial infarction (HR: 1.58; 95% CI: 1.42 to 1.76). These associations were linear, with greater risk even among individuals with neutrophil counts of 3 to 4 versus 2 to 3 × 109/l. There was a weak association with ischemic stroke (HR: 1.36; 95% CI: 1.17 to 1.57), but no association with stable angina or intracerebral hemorrhage. CONCLUSIONS: Neutrophil counts were strongly associated with the incidence of some CVDs, but not others, even within the normal range, consistent with underlying disease mechanisms differing across CVDs. (White Blood Cell Counts and Onset of Cardiovascular Diseases: a CALIBER Study [CALIBER]; NCT02014610).

White cell count in the normal range and short-term and long-term mortality: international comparisons of electronic health record cohorts in England and New Zealand.

OBJECTIVES: Electronic health records offer the opportunity to discover new clinical implications for established blood tests, but international comparisons have been lacking. We tested the association of total white cell count (WBC) with all-cause mortality in England and New Zealand. SETTING: Primary care practices in England (ClinicAl research using LInked Bespoke studies and Electronic health Records (CALIBER)) and New Zealand (PREDICT). DESIGN: Analysis of linked electronic health record data sets: CALIBER (primary care, hospitalisation, mortality and acute coronary syndrome registry) and PREDICT (cardiovascular risk assessments in primary care, hospitalisations, mortality, dispensed medication and laboratory results). PARTICIPANTS: People aged 30-75 years with no prior cardiovascular disease (CALIBER: N=686 475, 92.0% white; PREDICT: N=194 513, 53.5% European, 14.7% Pacific, 13.4% Maori), followed until death, transfer out of practice (in CALIBER) or study end. PRIMARY OUTCOME MEASURE: HRs for mortality were estimated using Cox models adjusted for age, sex, smoking, diabetes, systolic blood pressure, ethnicity and total:high-density lipoprotein (HDL) cholesterol ratio. RESULTS: We found 'J'-shaped associations between WBC and mortality; the second quintile was associated with lowest risk in both cohorts. High WBC within the reference range (8.65-10.05×109/L) was associated with significantly increased mortality compared to the middle quintile (6.25-7.25×109/L); adjusted HR 1.51 (95% CI 1.43 to 1.59) in CALIBER and 1.33 (95% CI 1.06 to 1.65) in PREDICT. WBC outside the reference range was associated with even greater mortality. The association was stronger over the first 6 months of follow-up, but similar across ethnic groups. CONCLUSIONS: Clinically recorded WBC within the range considered 'normal' is associated with mortality in ethnically different populations from two countries, particularly within the first 6 months. Large-scale international comparisons of electronic health record cohorts might yield new insights from widely performed clinical tests. TRIAL REGISTRATION NUMBER: NCT02014610.

Net clinical benefit of warfarin in individuals with atrial fibrillation across stroke risk and across primary and secondary care.

OBJECTIVE: To investigate net clinical benefit (NCB) of warfarin in individuals with atrial fibrillation (AF) across stroke risk and across primary and secondary care. METHODS: We conducted a linked electronic health record cohort study of 70 206 individuals with initial record of diagnosis of AF in primary (n=29 568) or secondary care (n=40 638) in England (1998-2010). We defined stroke risk according to the CHA2DS2-VASc score, and followed individuals over a median 2.2 years for 7005 ischaemic strokes (IS) and for 906 haemorrhagic strokes (HS). We calculated incidence rates (IRs) and 95% CIs per 100 person-years (PYs) (IR (95% CI)/100 PY) of IS and HS, with and without use of warfarin, and the NCB (ie, number of IS avoided) per 100 PYs of warfarin use (NCB (95% CI)/100 PY). RESULTS: Compared with individuals with initial record of diagnosis in secondary care, those in primary care had lower scores of IS risk (CHA2DS2-VASc≤2: 30.8% vs 20.6%), and lower overall incidence of IS (IR (95% CI)/100 PY: 2.3 (2.2 to 2.4) vs 4.3 (4.2 to 4.4), p value=0.00); however among individuals with CHA2DS2-VASc=0, 1 or 2 there were no differences in IS rate between those with initial record of diagnosis in primary care or secondary care (IR (95% CI)/100 PY: 0.2 (0.1 to 0.3) vs 0.3 (0.2 to 0.5), p value=0.16), (IR (95% CI)/100 PY: 0.6 (0.4 to 0.7) vs 0.7 (0.6 to 0.9), p value=0.08) and (IR (95% CI)/100 PY: 1.1 (1.00 to 1.3) vs 1.4 (1.2 to 1.6), p value=0.05), respectively. For CHA2DS2-VASc=0, 1 and 2, IRs of IS with versus without warfarin were (IR (95% CI)/100 PY: 0.4 (0.2 to 0.8) vs 0.2 (0.1 to 0.3), p value=0.16), (IR (95% CI)/100 PY: 0.4 (0.3 to 0.7) vs 0.7 (0.6 to 0.8), p value=0.03) and (IR (95% CI)/100 PY: 0.8 (0.7 to 1.0) vs 1.4 (1.3 to 1.6), p value=0.00), respectively. We found a significant positive NCB of warfarin from CHA2DS2-VASc≥2 in men (NCB (95% CI)/100 PY: 0.5 (0.1 to 0.9)) and from CHA2DS2-VASc≥3 in women (NCB (95% CI)/100 PY: 1.5 (1.1 to 1.9)). CONCLUSIONS: CHA2DS2-VASc accurately stratifies IS risk in individuals with AF across both primary and secondary care. However, the incidence rate of ischaemic stroke at CHA2DS2-VASc=1 are lower than previously reported, which may change the decision to start anticoagulation with warfarin in these individuals.

Low eosinophil and low lymphocyte counts and the incidence of 12 cardiovascular diseases: a CALIBER cohort study.

BACKGROUND: Eosinophil and lymphocyte counts are commonly performed in clinical practice. Previous studies provide conflicting evidence of association with cardiovascular diseases. METHODS: We used linked primary care, hospitalisation, disease registry and mortality data in England (the CALIBER (CArdiovascular disease research using LInked Bespoke studies and Electronic health Records) programme). We included people aged 30 or older without cardiovascular disease at baseline, and used Cox models to estimate cause-specific HRs for the association of eosinophil or lymphocyte counts with the first occurrence of cardiovascular disease. RESULTS: The cohort comprised 775 231 individuals, of whom 55 004 presented with cardiovascular disease over median follow-up 3.8 years. Over the first 6 months, there was a strong association of low eosinophil counts (<0.05 compared with 0.15-0.25×10(9)/L) with heart failure (adjusted HR 2.05; 95% CI 1.72 to 2.43), unheralded coronary death (HR 1.94, 95% CI 1.40 to 2.69), ventricular arrhythmia/sudden cardiac death and subarachnoid haemorrhage, but not angina, non-fatal myocardial infarction, transient ischaemic attack, ischaemic stroke, haemorrhagic stroke, subarachnoid haemorrhage or abdominal aortic aneurysm. Low eosinophil count was inversely associated with peripheral arterial disease (HR 0.63, 95% CI 0.44 to 0.89). There were similar associations with low lymphocyte counts (<1.45 vs 1.85-2.15×10(9)/L); adjusted HR over the first 6 months for heart failure was 2.25 (95% CI 1.90 to 2.67). Associations beyond the first 6 months were weaker. CONCLUSIONS: Low eosinophil counts and low lymphocyte counts in the general population are associated with increased short-term incidence of heart failure and coronary death. TRIAL REGISTRATION NUMBER: NCT02014610; results.

Long-term healthcare use and costs in patients with stable coronary artery disease: a population-based cohort using linked health records (CALIBER).

AIMS: To examine long-term healthcare utilization and costs of patients with stable coronary artery disease (SCAD). METHODS AND RESULTS: Linked cohort study of 94 966 patients with SCAD in England, 1 January 2001 to 31 March 2010, identified from primary care, secondary care, disease, and death registries. Resource use and costs, and cost predictors by time and 5-year cardiovascular disease (CVD) risk profile were estimated using generalized linear models. Coronary heart disease hospitalizations were 20.5% in the first year and 66% in the year following a non-fatal (myocardial infarction, ischaemic or haemorrhagic stroke) event. Mean healthcare costs were £3133 per patient in the first year and £10 377 in the year following a non-fatal event. First-year predictors of cost included sex (mean cost £549 lower in females), SCAD diagnosis (non-ST-elevation myocardial infarction cost £656 more than stable angina), and co-morbidities (heart failure cost £657 more per patient). Compared with lower risk patients (5-year CVD risk 3.5%), those of higher risk (5-year CVD risk 44.2%) had higher 5-year costs (£23 393 vs. £9335) and lower lifetime costs (£43 020 vs. £116 888). CONCLUSION: Patients with SCAD incur substantial healthcare utilization and costs, which varies and may be predicted by 5-year CVD risk profile. Higher risk patients have higher initial but lower lifetime costs than lower risk patients as a result of shorter life expectancy. Improved cardiovascular survivorship among an ageing CVD population is likely to require stratified care in anticipation of the burgeoning demand.

How Does Cardiovascular Disease First Present in Women and Men? Incidence of 12 Cardiovascular Diseases in a Contemporary Cohort of 1,937,360 People.

BACKGROUND: Given the recent declines in heart attack and stroke incidence, it is unclear how women and men differ in first lifetime presentations of cardiovascular diseases (CVDs). We compared the incidence of 12 cardiac, cerebrovascular, and peripheral vascular diseases in women and men at different ages. METHODS AND RESULTS: We studied 1 937 360 people, aged ≥ 30 years and free from diagnosed CVD at baseline (51% women), using linked electronic health records covering primary care, hospital admissions, acute coronary syndrome registry, and mortality (Cardiovascular Research Using LInked Bespoke Studies and Electronic Records [CALIBER] research platform). During 6 years median follow-up between 1997 and 2010, 114 859 people experienced an incident cardiovascular diagnosis, the majority (66%) of which were neither myocardial infarction nor ischemic stroke. Associations of male sex with initial diagnoses of CVD, however, varied from strong (age-adjusted hazard ratios, 3.6-5.0) for abdominal aortic aneurysm, myocardial infarction, and unheralded coronary death (particularly >60 years), through modest (hazard ratio, 1.5-2.0) for stable angina, ischemic stroke, peripheral arterial disease, heart failure, and cardiac arrest, to weak (hazard ratio <1.5) for transient ischemic attack, intracerebral hemorrhage, and unstable angina, and inverse (0.69) for subarachnoid hemorrhage (all P<0.001). CONCLUSIONS: The majority of initial presentations of CVD are neither myocardial infarction nor ischemic stroke, yet most primary prevention studies focus on these presentations. Sex has differing associations with different CVDs, with implications for risk prediction and management strategies. CLINICAL TRIAL REGISTRATION: URL: http://www.clinicaltrials.gov. Unique identifier: NCT01164371.

Type 2 diabetes and incidence of cardiovascular diseases: a cohort study in 1·9 million people.

BACKGROUND: The contemporary associations of type 2 diabetes with a wide range of incident cardiovascular diseases have not been compared. We aimed to study associations between type 2 diabetes and 12 initial manifestations of cardiovascular disease. METHODS: We used linked primary care, hospital admission, disease registry, and death certificate records from the CALIBER programme, which links data for people in England recorded in four electronic health data sources. We included people who were (or turned) 30 years or older between Jan 1, 1998, to March 25, 2010, who were free from cardiovascular disease at baseline. The primary endpoint was the first record of one of 12 cardiovascular presentations in any of the data sources. We compared cumulative incidence curves for the initial presentation of cardiovascular disease and used Cox models to estimate cause-specific hazard ratios (HRs). This study is registered at ClinicalTrials.gov (NCT01804439). FINDINGS: Our cohort consisted of 1 921 260 individuals, of whom 1 887 062 (98·2%) did not have diabetes and 34 198 (1·8%) had type 2 diabetes. We observed 113 638 first presentations of cardiovascular disease during a median follow-up of 5·5 years (IQR 2·1-10·1). Of people with type 2 diabetes, 6137 (17·9%) had a first cardiovascular presentation, the most common of which were peripheral arterial disease (reported in 992 [16·2%] of 6137 patients) and heart failure (866 [14·1%] of 6137 patients). Type 2 diabetes was positively associated with peripheral arterial disease (adjusted HR 2·98 [95% CI 2·76-3·22]), ischaemic stroke (1·72 [1·52-1·95]), stable angina (1·62 [1·49-1·77]), heart failure (1·56 [1·45-1·69]), and non-fatal myocardial infarction (1·54 [1·42-1·67]), but was inversely associated with abdominal aortic aneurysm (0·46 [0·35-0·59]) and subarachnoid haemorrhage (0·48 [0·26-0.89]), and not associated with arrhythmia or sudden cardiac death (0·95 [0·76-1·19]). INTERPRETATION: Heart failure and peripheral arterial disease are the most common initial manifestations of cardiovascular disease in type 2 diabetes. The differences between relative risks of different cardiovascular diseases in patients with type 2 diabetes have implications for clinical risk assessment and trial design. FUNDING: Wellcome Trust, National Institute for Health Research, and Medical Research Council.

Heterogeneous associations between smoking and a wide range of initial presentations of cardiovascular disease in 1937360 people in England: lifetime risks and implications for risk prediction.

BACKGROUND: It is not known how smoking affects the initial presentation of a wide range of chronic and acute cardiovascular diseases (CVDs), nor the extent to which associations are heterogeneous. We estimated the lifetime cumulative incidence of 12 CVD presentations, and examined associations with smoking and smoking cessation. METHODS: Cohort study of 1.93 million people aged ≥30years, with no history of CVD, in 1997-2010. Individuals were drawn from linked electronic health records in England, covering primary care, hospitalizations, myocardial infarction (MI) registry and cause-specific mortality (the CALIBER programme). RESULTS: During 11.6 million person-years of follow-up, 114859 people had an initial non-fatal or fatal CVD presentation. By age 90 years, current vs never smokers' lifetime risks varied from 0.4% vs 0.2% for subarachnoid haemorrhage (SAH), to 8.9% vs 2.6% for peripheral arterial disease (PAD). Current smoking showed no association with cardiac arrest or sudden cardiac death [hazard ratio (HR)=1.04, 95% confidence interval (CI) 0.91-1.19).The strength of association differed markedly according to disease type: stable angina (HR=1.08, 95% CI 1.01-1.15),transient ischaemic attack (HR=1.41, 95% CI 1.28-1.55), unstable angina (HR=1.54, 95% CI 1.38-1.72), intracerebral haemorrhage (HR=1.61, 95% CI 1.37-1.89), heart failure (HR=1.62, 95% CI 1.47-1.79), ischaemic stroke (HR=1.90, 95% CI 1.72-2.10), MI (HR=2.32, 95% CI 2.20-2.45), SAH (HR= 2.70, 95% CI 2.27-3.21), PAD (HR=5.16, 95% CI 4.80-5.54) and abdominal aortic aneurysm (AAA) (HR=5.18, 95% CI 4.61-5.82). Population-attributable fractions were lower for women than men for unheralded coronary death, ischaemic stroke, PAD and AAA. Ten years after quitting smoking, the risks of PAD, AAA (in men) and unheralded coronary death remained increased (HR=1.36, 1.47 and 2.74, respectively). CONCLUSIONS: The heterogeneous associations of smoking with different CVD presentations suggests different underlying mechanisms and have important implications for research, clinical screening and risk prediction.

Completeness and diagnostic validity of recording acute myocardial infarction events in primary care, hospital care, disease registry, and national mortality records: cohort study.

OBJECTIVE: To determine the completeness and diagnostic validity of myocardial infarction recording across four national health record sources in primary care, hospital care, a disease registry, and mortality register. DESIGN: Cohort study. PARTICIPANTS: 21 482 patients with acute myocardial infarction in England between January 2003 and March 2009, identified in four prospectively collected, linked electronic health record sources: Clinical Practice Research Datalink (primary care data), Hospital Episode Statistics (hospital admissions), the disease registry MINAP (Myocardial Ischaemia National Audit Project), and the Office for National Statistics mortality register (cause specific mortality data). SETTING: One country (England) with one health system (the National Health Service). MAIN OUTCOME MEASURES: Recording of acute myocardial infarction, incidence, all cause mortality within one year of acute myocardial infarction, and diagnostic validity of acute myocardial infarction compared with electrocardiographic and troponin findings in the disease registry (gold standard). RESULTS: Risk factors and non-cardiovascular coexisting conditions were similar across patients identified in primary care, hospital admission, and registry sources. Immediate all cause mortality was highest among patients with acute myocardial infarction recorded in primary care, which (unlike hospital admission and disease registry sources) included patients who did not reach hospital, but at one year mortality rates in cohorts from each source were similar. 5561 (31.0%) patients with non-fatal acute myocardial infarction were recorded in all three sources and 11 482 (63.9%) in at least two sources. The crude incidence of acute myocardial infarction was underestimated by 25-50% using one source compared with using all three sources. Compared with acute myocardial infarction defined in the disease registry, the positive predictive value of acute myocardial infarction recorded in primary care was 92.2% (95% confidence interval 91.6% to 92.8%) and in hospital admissions was 91.5% (90.8% to 92.1%). CONCLUSION: Each data source missed a substantial proportion (25-50%) of myocardial infarction events. Failure to use linked electronic health records from primary care, hospital care, disease registry, and death certificates may lead to biased estimates of the incidence and outcome of myocardial infarction. TRIAL REGISTRATION: NCT01569139 clinicaltrials.gov.