Identifying clinical features in primary care electronic health record studies: methods for codelist development

The effects of the first UK lockdown for the COVID-19 pandemic on primary-care-recorded cancer and type-2 diabetes mellitus records: A population-based quasi-experimental time series study

BACKGROUND

COVID-19 disrupted consulting behaviour, healthcare delivery and cancer diagnostic services. This study quantifies the cancer incidence coded in UK general practice electronic health records and deviations from historical trends after the March 2020 national lockdown. For comparison, we study the coded incidence of type-2 diabetes mellitus, which is diagnosed almost entirely within primary care.

METHODS

Poisson interrupted time series models investigated the coded incidence of diagnoses in adults aged ≥ 18 years in the Clinical Practice Research Datalink before (01/03/2017-29/02/2020) and after (01/03/2020-28/02/2022) the first lockdown. Datasets were stratified by age, sex, and general practice per 28-day aggregation period. Models captured incidence changes associated with lockdown, both immediately and over time based on historical trends.

RESULTS

We studied 189,457 incident cancer and 191,915 incident diabetes records in 1480 general practices over 52,374,197 person-years at risk. During 01/03/2020-28/02/2022, there were fewer incident records of cancer (n = 22,199, 10.49 %, 10.44-10.53 %) and diabetes (n = 15,709, 7.57 %, 7.53-7.61 %) than expected. Within cancers, impacts ranged from no effect (e.g. unknown primary, pancreas, and ovary), to small effects for lung (n = 773, 3.11 %, 3.09-3.13 % fewer records) and female breast (n = 2686, 6.77 %, 6.73-6.81 %), to the greatest effect for bladder (n = 2874, 31.15 %, 31.00-31.31 %). Diabetes and cancer records recovered maximally to 86 % (95 %CI 80.3-92.7 %) and 74 % (95 %CI 70.3-78.6 %) in July 2021 and May 2021, respectively, of their expected values, declining again until the study end.

CONCLUSION

The "missing" cancer and diabetes diagnoses in primary care may comprise delayed or missed diagnoses, reduced incidence associated with excess deaths from COVID-19, and potentially increased non-coded recording of diagnoses. Future validation studies must quantify the concordance between primary care and National Cancer Registration Data and Hospital Episode Statistics over the pandemic era.

Understanding symptom contribution to sex inequality in bladder and renal cancer stage at diagnosis

Background

Understanding sex-specific factors contributing to advanced-stage diagnosis can guide interventions to reduce sex inequality in patients with urological cancers.

Method

We used linked primary care and cancer registry data to examine associations between symptoms and advanced-stage in 1151 bladder cancer and 440 renal cancer patients diagnosed between January 2012 and December 2015 in England. We performed logistic regression, adjusting for sex, age, deprivation and routes to diagnosis, including interaction terms between symptoms and sex and symptoms and age.

Results

Female sex (OR vs. men 1.89 [1.28-2.79]; p = 0.001) and patients presenting with urinary tract infections (OR 2.22 [1.34-3.69]) and abdominal symptoms (OR 2.19 [1.30-3.70]) were associated with increased odds of advanced-stage bladder cancer (vs. haematuria, p = 0.016 for both). Women with haematuria and men with abdominal symptoms (compared with the opposite sex with the same presenting symptom) were more likely to have advanced-stage bladder cancer. Neither sex nor symptom associations were observed for renal cancer.

Conclusion

Non-haematuria symptoms are associated with higher risk of advanced-stage bladder cancer. Greater risk of advanced-stage bladder cancer in women may reflect biological differences in haematuria onset and sex differences during diagnostic process. Identifying higher risk women with haematuria may reduce sex inequalities in bladder cancer outcomes.

Creating a next-generation phenotype library: the health data research UK Phenotype Library

Objective

To enable reproducible research at scale by creating a platform that enables health data users to find, access, curate, and re-use electronic health record phenotyping algorithms.

Materials and Methods

We undertook a structured approach to identifying requirements for a phenotype algorithm platform by engaging with key stakeholders. User experience analysis was used to inform the design, which we implemented as a web application featuring a novel metadata standard for defining phenotyping algorithms, access via Application Programming Interface (API), support for computable data flows, and version control. The application has creation and editing functionality, enabling researchers to submit phenotypes directly.

Results

We created and launched the Phenotype Library in October 2021. The platform currently hosts 1049 phenotype definitions defined against 40 health data sources and >200K terms across 16 medical ontologies. We present several case studies demonstrating its utility for supporting and enabling research: the library hosts curated phenotype collections for the BREATHE respiratory health research hub and the Adolescent Mental Health Data Platform, and it is supporting the development of an informatics tool to generate clinical evidence for clinical guideline development groups.

Discussion

This platform makes an impact by being open to all health data users and accepting all appropriate content, as well as implementing key features that have not been widely available, including managing structured metadata, access via an API, and support for computable phenotypes.

Conclusions

We have created the first openly available, programmatically accessible resource enabling the global health research community to store and manage phenotyping algorithms. Removing barriers to describing, sharing, and computing phenotypes will help unleash the potential benefit of health data for patients and the public.

Review of codelists used to define hypertension in electronic health records and development of a codelist for research

BACKGROUND AND AIMS

Hypertension is a leading risk factor for cardiovascular disease. Electronic health records (EHRs) are routinely collected throughout a person's care, recording all aspects of health status, including current and past conditions, prescriptions and test results. EHRs can be used for epidemiological research. However, there are nuances in the way conditions are recorded using clinical coding; it is important to understand the methods which have been applied to define exposures, covariates and outcomes to enable interpretation of study findings. This study aimed to identify codelists used to define hypertension in studies that use EHRs and generate recommended codelists to support reproducibility and consistency.

ELIGIBILITY CRITERIA

Studies included populations with hypertension defined within an EHR between January 2010 and August 2023 and were systematically identified using MEDLINE and Embase. A summary of the most frequently used sources and codes is described. Due to an absence of Systematized Nomenclature of Medicine Clinical Terms (SNOMED CT) codelists in the literature, a recommended SNOMED CT codelist was developed to aid consistency and standardisation of hypertension research using EHRs.

FINDINGS

375 manuscripts met the study criteria and were eligible for inclusion, and 112 (29.9%) reported codelists. The International Classification of Diseases (ICD) was the most frequently used clinical terminology, 59 manuscripts provided ICD 9 codelists (53%) and 58 included ICD 10 codelists (52%). Informed by commonly used ICD and Read codes, usage recommendations were made. We derived SNOMED CT codelists informed by National Institute for Health and Care Excellence guidelines for hypertension management. It is recommended that these codelists be used to identify hypertension in EHRs using SNOMED CT codes.

CONCLUSIONS

Less than one-third of hypertension studies using EHRs included their codelists. Transparent methodology for codelist creation is essential for replication and will aid interpretation of study findings. We created SNOMED CT codelists to support and standardise hypertension definitions in EHR studies.

5-year mental health outcomes for children and adolescents presenting with psychiatric symptoms to general practitioners in England: a retrospective cohort study

BACKGROUND

Little information is available on the clinical trajectories of children and adolescents who attend general practice (GP) with psychiatric symptoms. We aimed to examine 5-year service use in English primary care for children and adolescents with neurodevelopmental or mental health symptoms or diagnoses.

METHODS

In this retrospective cohort study, we used anonymised primary care health records from the Clinical Practice Research Datalink Aurum database (CPRD-Aurum). We identified children and adolescents (aged 3-18 years) presenting to primary care in England between Jan 1, 2000, and May 9, 2016, with a symptom or diagnosis of a mental health, behavioural, or neurodevelopmental condition. Participants were excluded if they had less than 1 year of follow-up. We followed up participants from their index date until either death, transfer out of the practice, or the end of data collection on May 5, 2021, and for trajectory analysis we limited follow-up to 5 years. We used group-based multi-trajectory models to identify clusters with similar trajectories over 5 years of follow-up for three primary outcomes: mental health-related GP contacts, psychotropic medication prescriptions, and specialist mental health-care contact. We did survival analysis to examine the associations between trajectory-group membership and hospital admission for self-harm or death by suicide, as indicators of severe psychiatric distress.

FINDINGS

We included 369 340 children and adolescents, of whom 180 863 (49·0%) were girls, 188 438 (51·0%) were boys, 39 (<0·1%) were of indeterminate gender, 290 125 (78·6%) were White, 9161 (2·5%) were South Asian, 10 418 (2·8%) were Black, 8115 (2·2%) were of mixed ethnicity, and 8587 (2·3%) were other ethnicities, and the median age at index presentation was 13·6 years (IQR 8·4-16·7). In the best-fitting, seven-group, group-based multi-trajectory model, over a 5-year period, the largest group (low contact; 207 985 [51·2%]) had low rates of additional service contact or psychotropic prescriptions. The other trajectory groups were moderate, non-pharmacological contact (43 836 [13·0%]); declining contact (25 469 [8·7%]); year-4 escalating contact (18 277 [6·9%]); year-5 escalating contact (18 139; 5·2%); prolonged GP contact (32 147 [8·6%]); and prolonged specialist contact (23 487 [6·5%]). Non-White ethnicity and presentation in earlier study years (eg, 2000-2004) were associated with low-contact group membership. The prolonged specialist-contact group had the highest risk of hospital admission for self-harm (hazard ratio vs low-contact group 2·19 [95% CI 2·03-2·36]) and suicide (2·67 [1·72-4·14]).

INTERPRETATION

Most children and adolescents presenting to primary care with psychiatric symptoms or diagnoses have low or declining rates of ongoing contact. If these trajectories reflect symptomatic improvement, these findings provide reassurance for children and adolescents and their caregivers. However, these trajectories might reflect an unmet need for some children and adolescents.

FUNDING

National Institute for Health and Care Research and the Wellcome Trust.

Guideline concordance for timely chest imaging after new presentations of dyspnoea or haemoptysis in primary care: a retrospective cohort study

BACKGROUND

Guidelines recommend urgent chest X-ray for newly presenting dyspnoea or haemoptysis but there is little evidence about their implementation.

METHODS

We analysed linked primary care and hospital imaging data for patients aged 30+ years newly presenting with dyspnoea or haemoptysis in primary care during April 2012 to March 2017. We examined guideline-concordant management, defined as General Practitioner-ordered chest X-ray/CT carried out within 2 weeks of symptomatic presentation, and variation by sociodemographic characteristic and relevant medical history using logistic regression. Additionally, among patients diagnosed with cancer we described time to diagnosis, diagnostic route and stage at diagnosis by guideline-concordant status.

RESULTS

In total, 22 560/162 161 (13.9%) patients with dyspnoea and 4022/8120 (49.5%) patients with haemoptysis received guideline-concordant imaging within the recommended 2-week period. Patients with recent chest imaging pre-presentation were much less likely to receive imaging (adjusted OR 0.16, 95% CI 0.14-0.18 for dyspnoea, and adjusted OR 0.09, 95% CI 0.06-0.11 for haemoptysis). History of chronic obstructive pulmonary disease/asthma was also associated with lower odds of guideline concordance (dyspnoea: OR 0.234, 95% CI 0.225-0.242 and haemoptysis: 0.88, 0.79-0.97). Guideline-concordant imaging was lower among dyspnoea presenters with prior heart failure; current or ex-smokers; and those in more socioeconomically disadvantaged groups.The likelihood of lung cancer diagnosis within 12 months was greater among the guideline-concordant imaging group (dyspnoea: 1.1% vs 0.6%; haemoptysis: 3.5% vs 2.7%).

CONCLUSION

The likelihood of receiving urgent imaging concords with the risk of subsequent cancer diagnosis. Nevertheless, large proportions of dyspnoea and haemoptysis presenters do not receive prompt chest imaging despite being eligible, indicating opportunities for earlier lung cancer diagnosis.

Determining prescriptions in electronic healthcare record data: methods for development of standardized, reproducible drug codelists

Objective

To develop a standardizable, reproducible method for creating drug codelists that incorporates clinical expertise and is adaptable to other studies and databases.

Materials and Methods

We developed methods to generate drug codelists and tested this using the Clinical Practice Research Datalink (CPRD) Aurum database, accounting for missing data in the database. We generated codelists for: (1) cardiovascular disease and (2) inhaled Chronic Obstructive Pulmonary Disease (COPD) therapies, applying them to a sample cohort of 335 931 COPD patients. We compared searching all drug dictionary variables (A) against searching only (B) chemical or (C) ontological variables.

Results

In Search A, we identified 165 150 patients prescribed cardiovascular drugs (49.2% of cohort), and 317 963 prescribed COPD inhalers (94.7% of cohort). Evaluating output per search strategy, Search C missed numerous prescriptions, including vasodilator anti-hypertensives (A and B:19 696 prescriptions; C:1145) and SAMA inhalers (A and B:35 310; C:564).

Discussion

We recommend the full search (A) for comprehensiveness. There are special considerations when generating adaptable and generalizable drug codelists, including fluctuating status, cohort-specific drug indications, underlying hierarchical ontology, and statistical analyses.

Conclusions

Methods must have end-to-end clinical input, and be standardizable, reproducible, and understandable to all researchers across data contexts.

Diagnostic rate estimation from Medicare records: Dependence on claim numbers and latent clinical features

OBJECTIVE

International Classification of Disorders version 10 (ICD-10) codes contribute heavily to healthcare data. Medicare claims and other data-sources are used to constitute study populations and appraise healthcare processes. How variability in claims-per-beneficiary impacts diagnostic determinations is inadequately understood. The objective of this study is so assess distributional properties of Medicare claims, and examine claim rates impact on code utilization and rate determinations.

METHODS

The study population was Medicare beneficiaries aged 75-79.99 with claim(s) in the 5% standard analytical Carrier and Outpatient files, alive and participating in Medicare part B for all 12 months of 2017. Medicare beneficiary files were processed to create records containing all ICD-10 codes specified, key demographics, Part B and vital status, and the total claims for each 2017 beneficiary. Claim number cohorts were characterized.

RESULTS

Beneficiaries meeting inclusion criteria totaled 221,625, these having 7,617,503 claims; 96.4% had between 1 and 120 claims. Median claims were 24 for males (females 25); modal claims were 11 (13). Average distinct codes per beneficiary increased with claims number. The assignment of ICD-10 codes, i.e., 'diagnostic rate estimates' (DRE), increased as claim numbers increased for most codes among those most commonly utilized. For some conditions, mostly benign and age-related, DREs plateaued as claim numbers increased. For other conditions, typically associated with clinical acuity, e.g., chest pain, DREs increased steeply with claims.

CONCLUSIONS

Older adult Medicare beneficiaries aged 75-80 exhibited varying claims activity over the course of a year. Although DRE dependence on claim numbers varies across ICD-10 codes, rate estimates are higher for beneficiaries with claim numbers above the median.

Incidence of 'Low-Risk but Not No-Risk' Features of Cancer Prior to High-Risk Feature Occurrence: An Observational Cohort Study in Primary Care

Diagnosing cancer may be expedited by decreasing referral risk threshold. Clinical Practice Research Datalink participants (≥40 years) had a positive predictive value (PPV) ≥3% feature for breast, lung, colorectal, oesophagogastric, pancreatic, renal, bladder, prostatic, ovarian, endometrial or laryngeal cancer in 2016. The numbers of participants with features representing a 1-1.99% or 2-2.99% PPV for same cancer in the previous year were reported, alongside the time difference between meeting the ≥3% criteria and the lower threshold criteria. A total of 8616 participants had a PPV ≥3% feature, of whom 365 (4.2%) and 1147 (13.3%), respectively, met 2-2.99% and 1-1.99% criteria in the preceding year. The median time difference was 131 days (Interquartile Range (IQR) 27 to 256) for the 2-2.99% band and 179 days (IQR 58 to 289) for the 1-1.99% band. Results were heterogeneous across cancer sites. For some cancers, participants may progress from presenting lower- to higher-risk features before meeting urgent referral criteria; however, this was not usually the case. The details of specific features across multiple cancer sites will allow for a tailored approach to future reductions in referral thresholds, potentially improving the efficiency of urgent cancer referrals for the benefit both of individuals and the National Health Service (NHS).

Are There Ethnic Differences in Recorded Features among Patients Subsequently Diagnosed with Cancer? An English Longitudinal Data-Linked Study

We investigated ethnic differences in the presenting features recorded in primary care before cancer diagnosis.

METHODS

English population-based cancer-registry-linked primary care data were analysed. We identified the coded features of six cancers (breast, lung, prostate, colorectal, oesophagogastric, and myeloma) in the year pre-diagnosis. Logistic regression models investigated ethnic differences in first-incident cancer features, adjusted for age, sex, smoking status, deprivation, and comorbidity.

RESULTS

Of 130,944 patients, 92% were White. In total, 188,487 incident features were recorded in the year pre-diagnosis, with 48% (89,531) as sole features. Compared with White patients, Asian and Black patients with breast, colorectal, and prostate cancer were more likely than White patients to have multiple features; the opposite was seen for the Black and Other ethnic groups with lung or prostate cancer. The proportion with relevant recorded features was broadly similar by ethnicity, with notable cancer-specific exceptions. Asian and Black patients were more likely to have low-risk features (e.g., cough, upper abdominal pain) recorded. Non-White patients were less likely to have alarm features.

CONCLUSION

The degree to which these differences reflect disease, patient or healthcare factors is unclear. Further research examining the predictive value of cancer features in ethnic minority groups and their association with cancer outcomes is needed.

Defining cases of asthma, eczema and allergic rhinitis using electronic health records in the Born in Bradford birth cohort

Key Messages • In electronic health records, the accuracy of diagnostic codes to define outcomes can be uncertain • The accuracy can vary in different settings, doctors and practices, even with validated codes • We recommend definitions combining codes previously described and other codes available in the records.

The association between menstrual cycle characteristics and cardiometabolic outcomes in later life: a retrospective matched cohort study of 704,743 women from the UK

BACKGROUND

Female reproductive factors are gaining prominence as factors that enhance cardiovascular disease (CVD) risk; nonetheless, menstrual cycle characteristics are under-recognized as a factor associated with CVD. Additionally, there is limited data from the UK pertaining to menstrual cycle characteristics and CVD risk.

METHODS

A UK retrospective cohort study (1995-2021) using data from a nationwide database (The Health Improvement Network). Women aged 18-40 years at index date were included. 252,325 women with history of abnormal menstruation were matched with up to two controls. Two exposures were examined: regularity and frequency of menstrual cycles; participants were assigned accordingly to one of two separate cohorts. The primary outcome was composite cardiovascular disease (CVD). Secondary outcomes were ischemic heart disease (IHD), cerebrovascular disease, heart failure (HF), hypertension, and type 2 diabetes mellitus (T2DM). Cox proportional hazards regression models were used to derive adjusted hazard ratios (aHR) of cardiometabolic outcomes in women in the exposed groups compared matched controls.

RESULTS

During 26 years of follow-up, 20,605 cardiometabolic events occurred in 704,743 patients. Compared to women with regular menstrual cycles, the aHRs (95% CI) for cardiometabolic outcomes in women with irregular menstrual cycles were as follows: composite CVD 1.08 (95% CI 1.00-1.19), IHD 1.18 (1.01-1.37), cerebrovascular disease 1.04 (0.92-1.17), HF 1.30 (1.02-1.65), hypertension 1.07 (1.03-1.11), T2DM 1.37 (1.29-1.45). The aHR comparing frequent or infrequent menstrual cycles to menstrual cycles of normal frequency were as follows: composite CVD 1.24 (1.02-1.52), IHD 1.13 (0.81-1.57), cerebrovascular disease 1.43 (1.10-1.87), HF 0.99 (0.57-1.75), hypertension 1.31 (1.21-1.43), T2DM 1.74 (1.52-1.98).

CONCLUSIONS

History of either menstrual cycle irregularity or frequent or infrequent cycles were associated with an increased risk of cardiometabolic outcomes in later life. Menstrual history may be a useful tool in identifying women eligible for periodic assessment of their cardiometabolic health.

Risk of cardiometabolic outcomes among women with a history of pelvic inflammatory disease: a retrospective matched cohort study from the UK

INTRODUCTION

To describe the incidence and prevalence of pelvic inflammatory disease (PID) and to estimate the risk of cardiometabolic outcomes among women with PID compared to women without PID.

METHODS

A UK retrospective matched cohort study using data from The Health Improvement Network. To assess cardiometabolic risk, women (aged ≥ 16 years) with PID were compared to matched controls without PID. Annual prevalence and incidence of PID (1998-2017) were estimated among women aged 16-50 years using annual cross-sectional and cohort analyses, respectively. Adjusted hazard ratios (aHR) and 95% CI for cardiometabolic outcomes were estimated using Cox proportional hazards models. The primary outcome was composite cardiovascular disease (CVD) and its subtypes, including ischaemic heart disease (IHD), heart failure (HF) and cerebrovascular disease. Secondary outcomes were hypertension, and type 2 diabetes mellitus (T2DM).

RESULTS

Among the 715 recorded composite CVD events, the crude incidence rate per 1000 person-years was 1.5 among women with history of PID compared to 1.3 in matched controls. Compared to women without PID (N = 73,769), the aHRs for cardiometabolic outcomes among women with PID (N = 19,804) were: composite CVD 1.10 (95% CI 0.93-1.30); IHD 1.19 (95% CI 0.93-1.53); cerebrovascular disease 1.13 (95% CI 0.90-1.43); HF 0.92 (95% CI 0.62-1.35) hypertension 1.10 (95% CI 1.01-1.20); and T2DM 1.25 (95% CI 1.09-1.43). The prevalence (per 10,000 population) of PID was 396.5 in 1998 and 237 in 2017. The incidence (per 10,000 person-years) of PID was 32.4 in 1998 and 7.9 in 2017.

CONCLUSION

There was no excess risk of composite CVD or its subtypes among women with history of PID compared to matched controls. Findings from our study suggest that history of PID was associated with an increased risk of hypertension and type 2 diabetes mellitus, two major risk factors for CVD. Additional studies are required to support these findings.

Examining methodology to identify patterns of consulting in primary care for different groups of patients before a diagnosis of cancer: An exemplar applied to oesophagogastric cancer

BACKGROUND

Current methods for estimating the timeliness of cancer diagnosis are not robust because dates of key defining milestones, for example first presentation, are uncertain. This is exacerbated when patients have other conditions (multimorbidity), particularly those that share symptoms with cancer. Methods independent of this uncertainty are needed for accurate estimates of the timeliness of cancer diagnosis, and to understand how multimorbidity impacts the diagnostic process.

METHODS

Participants were diagnosed with oesophagogastric cancer between 2010 and 2019. Controls were matched on year of birth, sex, general practice and multimorbidity burden calculated using the Cambridge Multimorbidity Score. Primary care data (Clinical Practice Research Datalink) was used to explore population-level consultation rates for up to two years before diagnosis across different multimorbidity burdens. Five approaches were compared on the timing of the consultation frequency increase, the inflection point for different multimorbidity burdens, different aggregated time-periods and sample sizes.

RESULTS

We included 15,410 participants, of which 13,328 (86.5 %) had a measurable multimorbidity burden. Our new maximum likelihood estimation method found evidence that the inflection point in consultation frequency varied with multimorbidity burden, from 154 days (95 %CI 131.8-176.2) before diagnosis for patients with no multimorbidity, to 126 days (108.5-143.5) for patients with the greatest multimorbidity burden. Inflection points identified using alternative methods were closer to diagnosis for up to three burden groups. Sample size reduction and changing the aggregation period resulted in inflection points closer to diagnosis, with the smallest change for the maximum likelihood method.

DISCUSSION

Existing methods to identify changes in consultation rates can introduce substantial bias which depends on sample size and aggregation period. The direct maximum likelihood method was less prone to this bias than other methods and offers a robust, population-level alternative for estimating the timeliness of cancer diagnosis.

Underlying cancer risk among patients with fatigue and other vague symptoms: a population-based cohort study in primary care

BACKGROUND

Presenting to primary care with fatigue is associated with slightly increased cancer risk, although it is unknown how this varies in the presence of other 'vague' symptoms.

AIM

To quantify cancer risk in patients with fatigue who present with other 'vague' symptoms in the absence of 'alarm' symptoms for cancer.

DESIGN AND SETTING

Cohort study of patients presenting in UK primary care with new-onset fatigue during 2007-2015, using Clinical Practice Research Datalink data linked to national cancer registration data.

METHOD

Patients presenting with fatigue without co-occurring alarm symptoms or anaemia were identified, who were further characterised as having co-occurrence of 19 other 'vague' potential cancer symptoms. Sex- and age-specific 9-month cancer risk for each fatigue-vague symptom cohort were calculated.

RESULTS

Of 285 382 patients presenting with new-onset fatigue, 84% (n = 239 846) did not have co-occurring alarm symptoms or anaemia. Of these, 38% (n = 90 828) presented with ≥1 of 19 vague symptoms for cancer. Cancer risk exceeded 3% in older males with fatigue combined with any of the vague symptoms studied. The age at which risk exceeded 3% was 59 years for fatigue-weight loss, 65 years for fatigue-abdominal pain, 67 years for fatigue-constipation, and 67 years for fatigue-other upper gastrointestinal symptoms. For females, risk exceeded 3% only in older patients with fatigue-weight loss (from 65 years), fatigue-abdominal pain (from 79 years), or fatigue-abdominal bloating (from 80 years).

CONCLUSION

In the absence of alarm symptoms or anaemia, fatigue combined with specific vague presenting symptoms, alongside patient age and sex, can guide clinical decisions about referral for suspected cancer.

Translating and evaluating historic phenotyping algorithms using SNOMED CT

OBJECTIVE

Patient phenotype definitions based on terminologies are required for the computational use of electronic health records. Within UK primary care research databases, such definitions have typically been represented as flat lists of Read terms, but Systematized Nomenclature of Medicine-Clinical Terms (SNOMED CT) (a widely employed international reference terminology) enables the use of relationships between concepts, which could facilitate the phenotyping process. We implemented SNOMED CT-based phenotyping approaches and investigated their performance in the CPRD Aurum primary care database.

MATERIALS AND METHODS

We developed SNOMED CT phenotype definitions for 3 exemplar diseases: diabetes mellitus, asthma, and heart failure, using 3 methods: "primary" (primary concept and its descendants), "extended" (primary concept, descendants, and additional relations), and "value set" (based on text searches of term descriptions). We also derived SNOMED CT codelists in a semiautomated manner for 276 disease phenotypes used in a study of health across the lifecourse. Cohorts selected using each codelist were compared to "gold standard" manually curated Read codelists in a sample of 500 000 patients from CPRD Aurum.

RESULTS

SNOMED CT codelists selected a similar set of patients to Read, with F1 scores exceeding 0.93, and age and sex distributions were similar. The "value set" and "extended" codelists had slightly greater recall but lower precision than "primary" codelists. We were able to represent 257 of the 276 phenotypes by a single concept hierarchy, and for 135 phenotypes, the F1 score was greater than 0.9.

CONCLUSIONS

SNOMED CT provides an efficient way to define disease phenotypes, resulting in similar patient populations to manually curated codelists.

The Effect of Older Age and Frailty on the Time to Diagnosis of Cancer: A Connected Bradford Electronic Health Records Study

Over 60% of cancer diagnoses in the UK are in patients aged 65 and over. Cancer diagnosis and treatment in older adults is complicated by the presence of frailty, which is associated with lower survival rates and poorer quality of life. This population-based cohort study used a longitudinal database to calculate the time between presentation to primary care with a symptom suspicious of cancer and a confirmed cancer diagnosis for 7460 patients in the Bradford District. Individual frailty scores were calculated using the electronic frailty index (eFI) and categorised by severity. The median time from symptomatic presentation to cancer diagnosis for all patients was 48 days (IQR 21-142). 23% of the cohort had some degree of frailty. After adjustment for potential confounders, mild frailty added 7 days (95% CI 3-11), moderate frailty 23 days (95% CI 4-42) and severe frailty 11 days (95% CI -27-48) to the median time to diagnosis compared to not frail patients. Our findings support use of the eFI in primary care to identify and address patient, healthcare and system factors that may contribute to diagnostic delay. We recommend further research to explore patient and clinician factors when investigating cancer in frail patients.

Defining the optimum strategy for identifying adults and children with coeliac disease: systematic review and economic modelling

BACKGROUND

Coeliac disease is an autoimmune disorder triggered by ingesting gluten. It affects approximately 1% of the UK population, but only one in three people is thought to have a diagnosis. Untreated coeliac disease may lead to malnutrition, anaemia, osteoporosis and lymphoma.

OBJECTIVES

The objectives were to define at-risk groups and determine the cost-effectiveness of active case-finding strategies in primary care.

DESIGN

(1) Systematic review of the accuracy of potential diagnostic indicators for coeliac disease. (2) Routine data analysis to develop prediction models for identification of people who may benefit from testing for coeliac disease. (3) Systematic review of the accuracy of diagnostic tests for coeliac disease. (4) Systematic review of the accuracy of genetic tests for coeliac disease (literature search conducted in April 2021). (5) Online survey to identify diagnostic thresholds for testing, starting treatment and referral for biopsy. (6) Economic modelling to identify the cost-effectiveness of different active case-finding strategies, informed by the findings from previous objectives.

DATA SOURCES

For the first systematic review, the following databases were searched from 1997 to April 2021: MEDLINE® (National Library of Medicine, Bethesda, MD, USA), Embase® (Elsevier, Amsterdam, the Netherlands), Cochrane Library, Web of Science™ (Clarivate™, Philadelphia, PA, USA), the World Health Organization International Clinical Trials Registry Platform ( WHO ICTRP ) and the National Institutes of Health Clinical Trials database. For the second systematic review, the following databases were searched from January 1990 to August 2020: MEDLINE, Embase, Cochrane Library, Web of Science, Kleijnen Systematic Reviews ( KSR ) Evidence, WHO ICTRP and the National Institutes of Health Clinical Trials database. For prediction model development, Clinical Practice Research Datalink GOLD, Clinical Practice Research Datalink Aurum and a subcohort of the Avon Longitudinal Study of Parents and Children were used; for estimates for the economic models, Clinical Practice Research Datalink Aurum was used.

REVIEW METHODS

For review 1, cohort and case-control studies reporting on a diagnostic indicator in a population with and a population without coeliac disease were eligible. For review 2, diagnostic cohort studies including patients presenting with coeliac disease symptoms who were tested with serological tests for coeliac disease and underwent a duodenal biopsy as reference standard were eligible. In both reviews, risk of bias was assessed using the quality assessment of diagnostic accuracy studies 2 tool. Bivariate random-effects meta-analyses were fitted, in which binomial likelihoods for the numbers of true positives and true negatives were assumed.

RESULTS

People with dermatitis herpetiformis, a family history of coeliac disease, migraine, anaemia, type 1 diabetes, osteoporosis or chronic liver disease are 1.5-2 times more likely than the general population to have coeliac disease; individual gastrointestinal symptoms were not useful for identifying coeliac disease. For children, women and men, prediction models included 24, 24 and 21 indicators of coeliac disease, respectively. The models showed good discrimination between patients with and patients without coeliac disease, but performed less well when externally validated. Serological tests were found to have good diagnostic accuracy for coeliac disease. Immunoglobulin A tissue transglutaminase had the highest sensitivity and endomysial antibody the highest specificity. There was little improvement when tests were used in combination. Survey respondents (n = 472) wanted to be 66% certain of the diagnosis from a blood test before starting a gluten-free diet if symptomatic, and 90% certain if asymptomatic. Cost-effectiveness analyses found that, among adults, and using serological testing alone, immunoglobulin A tissue transglutaminase was most cost-effective at a 1% pre-test probability (equivalent to population screening). Strategies using immunoglobulin A endomysial antibody plus human leucocyte antigen or human leucocyte antigen plus immunoglobulin A tissue transglutaminase with any pre-test probability had similar cost-effectiveness results, which were also similar to the cost-effectiveness results of immunoglobulin A tissue transglutaminase at a 1% pre-test probability. The most practical alternative for implementation within the NHS is likely to be a combination of human leucocyte antigen and immunoglobulin A tissue transglutaminase testing among those with a pre-test probability above 1.5%. Among children, the most cost-effective strategy was a 10% pre-test probability with human leucocyte antigen plus immunoglobulin A tissue transglutaminase, but there was uncertainty around the most cost-effective pre-test probability. There was substantial uncertainty in economic model results, which means that there would be great value in conducting further research.

LIMITATIONS

The interpretation of meta-analyses was limited by the substantial heterogeneity between the included studies, and most included studies were judged to be at high risk of bias. The main limitations of the prediction models were that we were restricted to diagnostic indicators that were recorded by general practitioners and that, because coeliac disease is underdiagnosed, it is also under-reported in health-care data. The cost-effectiveness model is a simplification of coeliac disease and modelled an average cohort rather than individuals. Evidence was weak on the probability of routine coeliac disease diagnosis, the accuracy of serological and genetic tests and the utility of a gluten-free diet.

CONCLUSIONS

Population screening with immunoglobulin A tissue transglutaminase (1% pre-test probability) and of immunoglobulin A endomysial antibody followed by human leucocyte antigen testing or human leucocyte antigen testing followed by immunoglobulin A tissue transglutaminase with any pre-test probability appear to have similar cost-effectiveness results. As decisions to implement population screening cannot be made based on our economic analysis alone, and given the practical challenges of identifying patients with higher pre-test probabilities, we recommend that human leucocyte antigen combined with immunoglobulin A tissue transglutaminase testing should be considered for adults with at least a 1.5% pre-test probability of coeliac disease, equivalent to having at least one predictor. A more targeted strategy of 10% pre-test probability is recommended for children (e.g. children with anaemia).

FUTURE WORK

Future work should consider whether or not population-based screening for coeliac disease could meet the UK National Screening Committee criteria and whether or not it necessitates a long-term randomised controlled trial of screening strategies. Large prospective cohort studies in which all participants receive accurate tests for coeliac disease are needed.

STUDY REGISTRATION

This study is registered as PROSPERO CRD42019115506 and CRD42020170766.

FUNDING

This project was funded by the National Institute for Health and Care Research ( NIHR ) Health Technology Assessment programme and will be published in full in Health Technology Assessment; Vol. 26, No. 44. See the NIHR Journals Library website for further project information.

Revising the Suspected-Cancer Guidelines: Impacts on Patients' Primary Care Contacts and Costs

OBJECTIVES

This study aimed to explore the impact of revising suspected-cancer referral guidelines on primary care contacts and costs.

METHODS

Participants had incident cancer (colorectal, n = 2000; ovary, n = 763; and pancreas, n = 597) codes in the Clinical Practice Research Datalink or England cancer registry. Difference-in-differences analyses explored guideline impacts on contact days and nonzero costs between the first cancer feature and diagnosis. Participants were controls ("old National Institute for Health and Care Excellence [NICE]") or "new NICE" if their index feature was introduced during guideline revision. Model assumptions were inspected visually and by falsification tests. Sensitivity analyses reclassified participants who subsequently presented with features in the original guidelines as "old NICE." For colorectal cancer, sensitivity analysis (n = 3481) adjusted for multimorbidity burden.

RESULTS

Median contact days and costs were, respectively, 4 (interquartile range [IQR] 2-7) and £117.69 (IQR £53.23-£206.65) for colorectal, 5 (IQR 3-9) and £156.92 (IQR £78.46-£272.29) for ovary, and 7 (IQR 4-13) and £230.64 (IQR £120.78-£408.34) for pancreas. Revising ovary guidelines may have decreased contact days (incidence rate ratio [IRR] 0.74; 95% confidence interval 0.55-1.00; P = .05) with unchanged costs, but parallel trends assumptions were violated. Costs decreased by 13% (equivalent to -£28.05, -£50.43 to -£5.67) after colorectal guidance revision but only in sensitivity analyses adjusting for multimorbidity. Contact days and costs remained unchanged after pancreas guidance revision.

CONCLUSIONS

The main analyses of symptomatic patients suggested that prediagnosis primary care costs remained unchanged after guidance revision for pancreatic cancer. For colorectal cancer, contact days and costs decreased in analyses adjusting for multimorbidity. Revising ovarian cancer guidelines may have decreased primary care contact days but not costs, suggesting increased resource-use intensity; nevertheless, there is evidence of confounding.

Concordance with urgent referral guidelines in patients presenting with any of six 'alarm' features of possible cancer: a retrospective cohort study using linked primary care records

BACKGROUND

Clinical guidelines advise GPs in England which patients warrant an urgent referral for suspected cancer. This study assessed how often GPs follow the guidelines, whether certain patients are less likely to be referred, and how many patients were diagnosed with cancer within 1 year of non-referral.

METHODS

We used linked primary care (Clinical Practice Research Datalink), secondary care (Hospital Episode Statistics) and cancer registration data. Patients presenting with haematuria, breast lump, dysphagia, iron-deficiency anaemia, post-menopausal or rectal bleeding for the first time during 2014-2015 were included (for ages where guidelines recommend urgent referral). Logistic regression was used to investigate whether receiving a referral was associated with feature type and patient characteristics. Cancer incidence (based on recorded diagnoses in cancer registry data within 1 year of presentation) was compared between those receiving and those not receiving referrals.

RESULTS

48 715 patients were included, of which 40% (n=19 670) received an urgent referral within 14 days of presentation, varying by feature from 17% (dysphagia) to 68% (breast lump). Young patients (18-24 vs 55-64 years; adjusted OR 0.20, 95% CI 0.10 to 0.42, p<0.001) and those with comorbidities (4 vs 0 comorbidities; adjusted OR 0.87, 95% CI 0.80 to 0.94, p<0.001) were less likely to receive a referral. Associations between patient characteristics and referrals differed across features: among patients presenting with anaemia, breast lump or haematuria, those with multi-morbidity, and additionally for breast lump, more deprived patients were less likely to receive a referral. Of 29 045 patients not receiving a referral, 3.6% (1047) were diagnosed with cancer within 1 year, ranging from 2.8% for rectal bleeding to 9.5% for anaemia.

CONCLUSIONS

Guideline recommendations for action are not followed for the majority of patients presenting with common possible cancer features. A significant number of these patients developed cancer within 1 year of their consultation, indicating scope for improvement in the diagnostic process.

Assessing Ethnic Inequalities in Diagnostic Interval of Common Cancers: A Population-Based UK Cohort Study

BACKGROUND

This study investigated ethnic differences in diagnostic interval (DI)-the period between initial primary care presentation and diagnosis.

METHODS

We analysed the primary care-linked data of patients who reported features of seven cancers (breast, lung, prostate, colorectal, oesophagogastric, myeloma, and ovarian) one year before diagnosis. Accelerated failure time (AFT) models investigated the association between DI and ethnicity, adjusting for age, sex, deprivation, and morbidity.

RESULTS

Of 126,627 eligible participants, 92.1% were White, 1.99% Black, 1.71% Asian, 1.83% Mixed, and 2.36% were of Other ethnic backgrounds. Considering all cancer sites combined, the median (interquartile range) DI was 55 (20-175) days, longest in lung [127, (42-265) days], and shortest in breast cancer [13 (13, 8-18) days]. DI for the Black and Asian groups was 10% (AFT ratio, 95%CI 1.10, 1.05-1.14) and 16% (1.16, 1.10-1.22), respectively, longer than for the White group. Site-specific analyses revealed evidence of longer DI in Asian and Black patients with prostate, colorectal, and oesophagogastric cancer, plus Black patients with breast cancer and myeloma, and the Mixed group with lung cancer compared with White patients. DI was shorter for the Other group with lung, prostate, myeloma, and oesophagogastric cancer than the White group.

CONCLUSION

We found limited and inconsistent evidence of ethnic differences in DI among patients who reported cancer features in primary care before diagnosis. Our findings suggest that inequalities in diagnostic intervals, where present, are unlikely to be the sole explanation for ethnic variations in cancer outcomes.

Risk of cancer following primary care presentation with fatigue: a population-based cohort study of a quarter of a million patients

BACKGROUND

The management of adults presenting with fatigue presents a diagnostic challenge, particularly regarding possible underlying cancer.

METHODS

Using electronic health records, we examined cancer risk in patients presenting to primary care with new-onset fatigue in England during 2007-2013, compared to general population estimates. We examined variation by age, sex, deprivation, and time following presentation.

FINDINGS

Of 250,606 patients presenting with fatigue, 12-month cancer risk exceeded 3% in men aged 65 and over and women aged 80 and over, and 6% in men aged 80 and over. Nearly half (47%) of cancers were diagnosed within 3 months from first fatigue presentation. Site-specific cancer risk was higher than the general population for most cancers studied, with greatest relative increases for leukaemia, pancreatic and brain cancers.

CONCLUSIONS

In older patients, new-onset fatigue is associated with cancer risk exceeding current thresholds for urgent specialist referral. Future research should consider how risk is modified by the presence or absence of other signs and symptoms. Excess cancer risk wanes rapidly after 3 months, which could inform the duration of a 'safety-netting' period. Fatigue presentation is not strongly predictive of any single cancer, although certain cancers are over-represented; this knowledge can help prioritise diagnostic strategies.

Diagnoses after newly-recorded abdominal pain in primary care: observational cohort study

Background

Non-acute abdominal pain in primary care is diagnostically challenging.

Aim

To quantify the 1-year cumulative incidence of 35 non-malignant diagnoses and nine cancers in adults after newly recorded abdominal pain in primary care.

Design and setting

Observational cohort study of 125 793 Clinical Practice Research Datalink GOLD records.

Method

Participants, aged ≥40 years, had newly recorded abdominal pain between 1 January 2009 and 31 December 2013. Age- and sex-stratified 1-year cumulative incidence by diagnosis is reported.

Results

Most (>70%) participants had no pre-specified diagnoses after newly recorded abdominal pain. Non-malignant diagnoses were most common: upper gastrointestinal problems (gastro-oesophageal reflux disease, hiatus hernia, gastritis, oesophagitis, and gastric/duodenal ulcer) in males and urinary tract infection in females. The incidence of upper gastrointestinal problems plateaued at age ≥60 years (aged 40–59 years: males 4.9%, 95% confidence interval [CI] = 4.6 to 5.1, females 4.0%, 95% CI = 3.8 to 4.2; aged 60–69 years: males 5.8%, 95% CI = 5.4 to 6.2, females 5.4%, 95% CI = 5.1 to 5.8). Urinary tract infection incidence increased with age (aged 40–59 years: females 5.1%, 95% CI = 4.8 to 5.3, males 1.1%, 95% CI = 1.0 to 1.2; aged ≥70 years: females 8.0%, 95% CI = 7.6 to 8.4, males 3.3%, 95% CI = 3.0 to 3.6%). Diverticular disease incidence rose with age, plateauing at 4.2% (95% CI = 3.9 to 4.6) in males aged ≥60 years, increasing to 6.1% (95% CI = 5.8 to 6.4) in females aged ≥70 years. Irritable bowel syndrome incidence was higher in females (aged 40–59 years: 2.9%, 95% CI = 2.7 to 3.1) than males (aged 40–59 years: 2.1%, 95% CI = 1.9 to 2.3), decreasing with age to 1.3% (95% CI = 1.2 to 1.5) in females and 0.6% (95% CI = 0.5 to 0.8) in males aged ≥70 years.

Conclusion

Although abdominal pain commonly remains unexplained, non-malignant diagnosis are more likely than cancer.

Establishing a National Cardiovascular Disease Surveillance System in the United States Using Electronic Health Record Data: Key Strengths and Limitations

Cardiovascular disease surveillance involves quantifying the evolving population-level burden of cardiovascular outcomes and risk factors as a data-driven initial step followed by the implementation of interventional strategies designed to alleviate this burden in the target population. Despite widespread acknowledgement of its potential value, a national surveillance system dedicated specifically to cardiovascular disease does not currently exist in the United States. Routinely collected health care data such as from electronic health records (EHRs) are a possible means of achieving national surveillance. Accordingly, this article elaborates on some key strengths and limitations of using EHR data for establishing a national cardiovascular disease surveillance system. Key strengths discussed include the: (1) ubiquity of EHRs and consequent ability to create a more "national" surveillance system, (2) existence of a common data infrastructure underlying the health care enterprise with respect to data domains and the nomenclature by which these data are expressed, (3) longitudinal length and detail that define EHR data when individuals repeatedly patronize a health care organization, and (4) breadth of outcomes capable of being surveilled with EHRs. Key limitations discussed include the: (1) incomplete ascertainment of health information related to health care-seeking behavior and the disconnect of health care data generated at separate health care organizations, (2) suspect data quality resulting from the default information-gathering processes within the clinical enterprise, (3) questionable ability to surveil patients through EHRs in the absence of documented interactions, and (4) the challenge in interpreting temporal trends in health metrics, which can be obscured by changing clinical and administrative processes.

Intra-abdominal cancer risk with abdominal pain: a prospective cohort primary-care study

Background: Quantifying cancer risk in primary-care patients reporting abdominal pain would inform diagnostic strategies. Aim: To quantify oesophagogastric, colorectal, liver, pancreatic, ovarian, uterine, kidney and bladder cancer risks associated with newly reported abdominal pain with or without other symptoms, signs or abnormal blood tests (i.e. features) indicative of possible cancer. Design and setting: Observational prospective cohort study using Clinical Practice Research Datalink records with English cancer registry linkage. Methods: Participants (N=125,793) aged ≥40 years had newly reported abdominal pain in primary care during 01/01/2009-31/12/2013. The outcomes were 1-year cumulative incidence of cancer, and the composite 1-year cumulative incidence of cancers with shared additional features, stratified by age and sex. Results: With abdominal pain, overall risk was greater in men and increased with age, reaching 3.4% (95%CI 3.0–3.7%; predominantly colorectal cancer 1.9%, 1.6–2.1%) in men ≥70 years, compared with their expected incidence of 0.88% (0.87%–0.89%). Additional features increased cancer risk; for example, colorectal or pancreatic cancer risk with abdominal pain plus diarrhoea at 60–69 and ≥70, respectively, was 3.1% (1.9–4.9%) and 4.9% (3.7–6.4%), predominantly colorectal cancer (2.2%, 2–3.8% and 3.3%, 2.0–4.9%). Conclusions: Abdominal pain increases intra-abdominal cancer risk nearly fourfold in men aged ≥70, exceeding the 3% threshold warranting investigation. This threshold is surpassed for the over-60s only with additional features. These results help direct appropriate referral and testing strategies for patients based on their demographic profile and reporting features. We suggest non-invasive strategies first, such as faecal immunochemical testing, with safety-netting in a shared decision-making framework.

The risk of all-cause and cause-specific mortality in people prescribed mirtazapine: an active comparator cohort study using electronic health records

BACKGROUND

Studies have reported an increased risk of mortality among people prescribed mirtazapine compared to other antidepressants. The study aimed to compare all-cause and cause-specific mortality between adults prescribed mirtazapine or other second-line antidepressants.

METHODS

This cohort study used English primary care electronic medical records, hospital admission records, and mortality data from the Clinical Practice Research Datalink (CPRD), for the period 01 January 2005 to 30 November 2018. It included people aged 18-99 years with depression first prescribed a selective serotonin reuptake inhibitor (SSRI) and then prescribed mirtazapine (5081), a different SSRI (15,032), amitriptyline (3905), or venlafaxine (1580). Follow-up was from starting to stopping the second antidepressant, with a 6-month wash-out window, censoring at the end of CPRD follow-up or 30 November 2018. Age-sex standardised rates of all-cause mortality and death due to circulatory system disease, cancer, or respiratory system disease were calculated. Survival analyses were performed, accounting for baseline characteristics using inverse probability of treatment weighting.

RESULTS

The cohort contained 25,598 people (median age 41 years). The mirtazapine group had the highest standardised mortality rate, with an additional 7.8 (95% confidence interval (CI) 5.9-9.7) deaths/1000 person-years compared to the SSRI group. Within 2 years of follow-up, the risk of all-cause mortality was statistically significantly higher in the mirtazapine group than in the SSRI group (weighted hazard ratio (HR) 1.62, 95% CI 1.28-2.06). No significant difference was found between the mirtazapine group and the amitriptyline (HR 1.18, 95% CI 0.85-1.63) or venlafaxine (HR 1.11, 95% CI 0.60-2.05) groups. After 2 years, the risk was significantly higher in the mirtazapine group compared to the SSRI (HR 1.51, 95% CI 1.04-2.19), amitriptyline (HR 2.59, 95% CI 1.38-4.86), and venlafaxine (HR 2.35, 95% CI 1.02-5.44) groups. The risks of death due to cancer (HR 1.74, 95% CI 1.06-2.85) and respiratory system disease (HR 1.72, 95% CI 1.07-2.77) were significantly higher in the mirtazapine than in the SSRI group.

CONCLUSIONS

Mortality was higher in people prescribed mirtazapine than people prescribed a second SSRI, possibly reflecting residual differences in other risk factors between the groups. Identifying these potential health risks when prescribing mirtazapine may help reduce the risk of mortality.

Preliminary Exploration of Main Elements for Systematic Classification Development: Case Study of Patient Safety Incidents (Preprint)

Background

Currently, there is no holistic theoretical approach available for guiding classification development. On the basis of our recent classification development research in the area of patient safety in health information technology, this focus area would benefit from a more systematic approach. Although some valuable theoretical and methodological approaches have been presented, classification development literature typically is limited to methodological development in a specific domain or is practically oriented.

Objective

The main purposes of this study are to fill the methodological gap in classification development research by exploring possible elements of systematic development based on previous literature and to promote sustainable and well-grounded classification outcomes by identifying a set of recommended elements. Specifically, the aim is to answer the following question: what are the main elements for systematic classification development based on research evidence and our use case?

Methods

This study applied a qualitative research approach. On the basis of previous literature, preliminary elements for classification development were specified, as follows: defining a concept model, documenting the development process, incorporating multidisciplinary expertise, validating results, and maintaining the classification. The elements were compiled as guiding principles for the research process and tested in the case of patient safety incidents (n=501).

Results

The results illustrate classification development based on the chosen elements, with 4 examples of technology-induced errors. Examples from the use case regard usability, system downtime, clinical workflow, and medication section problems. The study results confirm and thus suggest that a more comprehensive and theory-based systematic approach promotes well-grounded classification work by enhancing transparency and possibilities for assessing the development process.

Conclusions

We recommend further testing the preliminary main elements presented in this study. The research presented herein could serve as a basis for future work. Our recently developed classification and the use case presented here serve as examples. Data retrieved from, for example, other type of electronic health records and use contexts could refine and validate the suggested methodological approach.

An evaluation of a national mass media campaign to raise public awareness of possible lung cancer symptoms in England in 2016 and 2017

BACKGROUND

A two-phase 'respiratory symptoms' mass media campaign was conducted in 2016 and 2017 in England raising awareness of cough and worsening shortness of breath as symptoms warranting a general practitioner (GP) visit.

METHOD

A prospectively planned pre-post evaluation was done using routinely collected data on 15 metrics, including GP attendance, GP referral, emergency presentations, cancers diagnosed (five metrics), cancer stage, investigations (two metrics), outpatient attendances, inpatient admissions, major lung resections and 1-year survival. The primary analysis compared 2015 with 2017. Trends in metrics over the whole period were also considered. The effects of the campaign on awareness of lung cancer symptoms were evaluated using bespoke surveys.

RESULTS

There were small favourable statistically significant and clinically important changes over 2 years in 11 of the 15 metrics measured, including a 2.11% (95% confidence interval 1.02-3.20, p < 0.001) improvement in the percentage of lung cancers diagnosed at an early stage. However, these changes were not accompanied by increases in GP attendances. Furthermore, the time trends showed a gradual change in the metrics rather than steep changes occurring during or after the campaigns.

CONCLUSION

There were small positive changes in most metrics relating to lung cancer diagnosis after this campaign. However, the pattern over time challenges whether the improvements are wholly attributable to the campaign. Given the importance of education on cancer in its own right, raising awareness of symptoms should remain important. However further research is needed to maximise the effect on health outcomes.

The impact of changing risk thresholds on the number of people in England eligible for urgent investigation for possible cancer: an observational cross-sectional study

Background

Expediting cancer diagnosis may be achieved by targeted decreases in referral thresholds to increase numbers of patients referred for urgent investigation.

Methods

Clinical Practice Research Datalink data from England for 150,921 adults aged ≥40 were used to identify participants with features of possible cancer equating to risk thresholds ≥1%, ≥2% or ≥3% for breast, lung, colorectal, oesophago-gastric, pancreatic, renal, bladder, prostatic, ovarian, endometrial and laryngeal cancers.

Results

The mean age of participants was 60 (SD 13) years, with 73,643 males (49%). In 2016, 8576 consultation records contained coded features having a positive predictive value (PPV) of ≥3% for any of the 11 cancers. This equates to a rate of 5682/100,000 patients compared with 4601/100,000 Suspected Cancer NHS referrals for these cancers from April 2016–March 2017. Nine thousands two hundred ninety-one patient-consultation records had coded features equating to a ≥2% PPV, 8% more than met PPV ≥ 3%. Similarly, 19,517 had features with a PPV ≥ 1%, 136% higher than for PPV ≥ 3%.

Conclusions

This study estimated the number of primary-care patients presenting at lower thresholds of cancer risk. The resource implications of liberalising this threshold to 2% are modest and manageable. The details across individual cancer sites should assist planning of English cancer services.

Predictive values for different cancers and inflammatory bowel disease of 6 common abdominal symptoms among more than 1.9 million primary care patients in the UK: A cohort study

BACKGROUND

The diagnostic assessment of abdominal symptoms in primary care presents a challenge. Evidence is needed about the positive predictive values (PPVs) of abdominal symptoms for different cancers and inflammatory bowel disease (IBD).

METHODS AND FINDINGS

Using data from The Health Improvement Network (THIN) in the United Kingdom (2000-2017), we estimated the PPVs for diagnosis of (i) cancer (overall and for different cancer sites); (ii) IBD; and (iii) either cancer or IBD in the year post-consultation with each of 6 abdominal symptoms: dysphagia (n = 86,193 patients), abdominal bloating/distension (n = 100,856), change in bowel habit (n = 106,715), rectal bleeding (n = 235,094), dyspepsia (n = 517,326), and abdominal pain (n = 890,490). The median age ranged from 54 (abdominal pain) to 63 years (dysphagia and change in bowel habit); the ratio of women/men ranged from 50%:50% (rectal bleeding) to 73%:27% (abdominal bloating/distension). Across all studied symptoms, the risk of diagnosis of cancer and the risk of diagnosis of IBD were of similar magnitude, particularly in women, and younger men. Estimated PPVs were greatest for change in bowel habit in men (4.64% cancer and 2.82% IBD) and for rectal bleeding in women (2.39% cancer and 2.57% IBD) and lowest for dyspepsia (for cancer: 1.41% men and 1.03% women; for IBD: 0.89% men and 1.00% women). Considering PPVs for specific cancers, change in bowel habit and rectal bleeding had the highest PPVs for colon and rectal cancer; dysphagia for esophageal cancer; and abdominal bloating/distension (in women) for ovarian cancer. The highest PPVs of abdominal pain (either sex) and abdominal bloating/distension (men only) were for non-abdominal cancer sites. For the composite outcome of diagnosis of either cancer or IBD, PPVs of rectal bleeding exceeded the National Institute of Health and Care Excellence (NICE)-recommended specialist referral threshold of 3% in all age-sex strata, as did PPVs of abdominal pain, change in bowel habit, and dyspepsia, in those aged 60 years and over. Study limitations include reliance on accuracy and completeness of coding of symptoms and disease outcomes.

CONCLUSIONS

Based on evidence from more than 1.9 million patients presenting in primary care, the findings provide estimated PPVs that could be used to guide specialist referral decisions, considering the PPVs of common abdominal symptoms for cancer alongside that for IBD and their composite outcome (cancer or IBD), taking into account the variable PPVs of different abdominal symptoms for different cancers sites. Jointly assessing the risk of cancer or IBD can better support decision-making and prompt diagnosis of both conditions, optimising specialist referrals or investigations, particularly in women.

Quantifying the impact of pre-existing conditions on the stage of oesophagogastric cancer at diagnosis: a primary care cohort study using electronic medical records

BACKGROUND

Pre-existing conditions interfere with cancer diagnosis by offering diagnostic alternatives, competing for clinical attention or through patient surveillance.

OBJECTIVE

To investigate associations between oesophagogastric cancer stage and pre-existing conditions.

METHODS

Retrospective cohort study using Clinical Practice Research Datalink (CPRD) data, with English cancer registry linkage. Participants aged ≥40 years had consulted primary care in the year before their incident diagnosis of oesophagogastric cancer in 01/01/2010-31/12/2015. CPRD records pre-diagnosis were searched for codes denoting clinical features of oesophagogastric cancer and for pre-existing conditions, including those providing plausible diagnostic alternatives for those features. Logistic regression analysed associations between stage and multimorbidity (≥2 conditions; reference category: no multimorbidity) and having 'diagnostic alternative(s)', controlling for age, sex, deprivation and cancer site.

RESULTS

Of 2444 participants provided, 695 (28%) were excluded for missing stage, leaving 1749 for analysis (1265/1749, 72.3% had advanced-stage disease). Multimorbidity was associated with stage [odds ratio 0.63, 95% confidence interval (CI) 0.47-0.85, P = 0.002], with moderate evidence of an interaction term with sex (1.76, 1.08-2.86, P = 0.024). There was no association between alternative explanations and stage (odds ratio 1.18, 95% CI 0.87-1.60, P = 0.278).

CONCLUSIONS

In men, multimorbidity is associated with a reduced chance of advanced-stage oesophagogastric cancer, to levels seen collectively for women.

Guideline interval: A new time interval in the diagnostic pathway for symptomatic cancer

BACKGROUND

A standard measure of the cancer diagnostic pathway, diagnostic interval, is the time from "first presentation of cancer" to diagnosis. Cancer presentation may be unclear in patients with multimorbidity or non-specific symptoms, signs or test results ("features"). We propose an alternative, guideline interval, with a more certain start date; namely, when the patient first meets suspected-cancer criteria for investigation or referral.

METHODS

This retrospective cohort study used Clinical Practice Research Datalink (CPRD) and English cancer registry data. Participants, aged ≥55 years, had diagnostic codes for oesophagogastric cancers in 1/1/12-31/12/17. Features of oesophagogastric cancer in the year before diagnosis were identified from CPRD codes for dysphagia, haematemesis, upper-abdominal mass or pain, low haemoglobin, reflux, dyspepsia, nausea, vomiting, weight loss or thrombocytosis. Diagnostic interval was the time from first feature to diagnosis; guidance interval, the time from first meeting criteria in NICE suspected-cancer guidance to diagnosis. Multimorbidity burden was quantified using Adjusted Clinical Groups®. Accelerated failure-time models explored associations between multimorbidity burden and length of both diagnostic and guideline interval.

RESULTS

There were 3,793 eligible participants (69.0 % male), mean age 74.1 years (SD 10.5). 3,097 (81.7 %) presented with ≥1 feature in the year before diagnosis, and 1,990 (52.5 %) met NICE suspected-cancer criteria. The median for both intervals was 11 days in healthy users, and rose with increasing morbidity burden. At very high multimorbidity burden, diagnostic interval was 5.47 (95%CI 3.25-9.20) times longer and guideline interval was 3.91 (2.63-5.80) times longer than for healthy users.

CONCLUSIONS

Guideline interval is proposed as a new measure of the cancer diagnostic pathway. It has a more certain start date than diagnostic interval, and is lengthened less than diagnostic interval in people with a very high multimorbidity burden. Guideline interval has potential for assessing the implementation of suspected-cancer policies.

Changes in presentations with features potentially indicating cancer in primary care during the COVID-19 pandemic: a retrospective cohort study

OBJECTIVES

To investigate how the COVID-19 pandemic affected the number of people aged 50+ years presenting to primary care with features that could potentially indicate cancer, and to explore how reporting differed by patient characteristics and in face-to-face vs remote consultations.

DESIGN, SETTING AND PARTICIPANTS

A retrospective cohort study of general practitioner (GP), nurse and paramedic primary care consultations in 21 practices in South-West England covering 123 947 patients. The models compared potential cancer indicators reported in April-July 2019 with April-July 2020.

MAIN OUTCOME MEASURES

Potential indicators of cancer were identified using code lists for symptoms, signs, test results and diagnoses listed in the National Institute for Health and Care Excellence suspected cancer referral guidance (NG12).

RESULTS

During April-July 2019, 17% of registered patients aged 50+ years reported a potential cancer indicator in a consultation with a GP or nurse. During April-July 2020, this reduced to 11% (incidence rate ratio (IRR) 0.64, 95% CI 0.62 to 0.67, p<0.001). Reductions in potential cancer indicators were stable across age group, sex, ethnicity, index of multiple deprivation quintile and shielding status, but less marked in patients with mental health conditions than without (IRR 0.75, 95% CI 0.72 to 0.79, interaction p<0.001). Proportions of GP consultations with potential indicators of cancer reduced between 2019 and 2020 for face-to-face consultations (IRR 0.84, 95% CI 0.76 to 0.92, p<0.001) and increased for remote consultations (IRR 1.17, 95% CI 1.07 to 1.29, p=0.001), although it remained lower in remote consulting than face-to-face in April-July 2020. This difference was greater for nurse/paramedic consultations (face-to-face: IRR 0.61, 95% CI 0.44 to 0.83, p=0.002; remote: IRR 1.60, 95% CI 1.10 to 2.333, p=0.014).

CONCLUSION

The number of patients consulting with presentations that could potentially indicate cancer reduced during the first wave of the COVID-19 pandemic. Patients should be encouraged to continue contacting primary care for persistent signs and symptoms, and GPs and nurses should be encouraged to probe patients for further information during remote consulting, in the absence of non-verbal cues.

Trends in time to cancer diagnosis around the period of changing national guidance on referral of symptomatic patients: A serial cross-sectional study using UK electronic healthcare records from 2006-17

BACKGROUND

UK primary-care referral guidance describes the signs, symptoms, and test results ("features") of undiagnosed cancer. Guidance revision in 2015 liberalised investigation by introducing more low-risk features. We studied adults with cancer whose features were in the 2005 guidance ("Old-NICE") or were introduced in the revision ("New-NICE"). We compared time to diagnosis between the groups, and its trend over 2006-2017.

METHODS

Clinical Practice Research Datalink records were analysed for adults with incident myeloma, breast, bladder, colorectal, lung, oesophageal, ovarian, pancreatic, prostate, stomach or uterine cancers in 1/1/2006-31/12/2017. Cancer-specific features in the year before diagnosis were used to create New-NICE and Old-NICE groups. Diagnostic interval was time between the index feature and diagnosis. Semiparametric varying-coefficient analyses compared diagnostic intervals between New-NICE and Old-NICE groups over 1/1/2006-31/12/2017.

RESULTS

Over all cancers (N = 83,935), median (interquartile range) Old-NICE diagnostic interval rose over 2006-2017, from 51 (20-132) to 64 (30-148) days, with increases in breast (15 vs 25 days), lung (103 vs 135 days), ovarian (65·5 vs 100 days), prostate (80 vs 93 days) and stomach (72·5 vs 102 days) cancers. Median New-NICE values were consistently longer (99, 40-212 in 2006 vs 103, 42-236 days in 2017) than Old-NICE values over all cancers. After guidance revision, New-NICE diagnostic intervals became shorter than Old-NICE values for colorectal cancer.

CONCLUSIONS

Despite improvements for colorectal cancer, scope remains to reduce diagnostic intervals for most cancers. Liberalised investigation requires protecting and enhancing cancer-diagnostic services to avoid their becoming a rate-limiting step in the diagnostic pathway.

Association of non-malignant diseases with thrombocytosis: a prospective cohort study in general practice

BACKGROUND

Thrombocytosis is an excess of platelets, which is diagnosed as a platelet count >400 × 10⁹/l. An association of thrombocytosis with undiagnosed cancer has recently been established, but the association with non-malignant disease has not been studied in primary care.

AIM

To examine, in English primary care, the 1-year incidence of non-malignant diseases in patients with new thrombocytosis and the incidence of pre-existing non-malignant diseases in patients who develop new thrombocytosis.

DESIGN AND SETTING

Prospective cohort study using English Clinical Practice Research Datalink data from 2000 to 2013.

METHOD

Newly incident and pre-existing rates of non-malignant diseases associated with thrombocytosis were compared between patients with thrombocytosis and age- and sex-matched patients with a normal platelet count. Fifteen candidate non-malignant diseases were identified from literature searches.

RESULTS

In the thrombocytosis cohort of 39 850 patients, 4579 (11.5%) were newly diagnosed with any one of the candidate diseases, compared with 443 out of 9684 patients (4.6%) in the normal platelet count cohort (relative risk [RR] 2.5, 95% confidence intervals [CI] = 2.3 to 2.8); iron-deficiency anaemia was the most common new diagnosis (4.5% of patients with thrombocytosis, RR 4.9, 95% CI = 4.0 to 6.1). A total of 22 612 (57.0%) patients with thrombocytosis had a pre-existing non-malignant diagnosis compared with 4846 patients (50%) in the normal platelet count cohort (odds ratio 1.3, 95% CI = 1.2 to 1.4). There was no statistically significant difference in cancer diagnoses between patients with and without pre-existing disease in the thrombocytosis cohort.

CONCLUSION

Thrombocytosis is associated with several non-malignant diseases. Clinicians can use these findings as part of their holistic diagnostic approach to help guide further investigations and management of patients with thrombocytosis.

Effect of pre-existing conditions on bladder cancer stage at diagnosis: a cohort study using electronic primary care records in the UK

Background

Pre-existing concurrent medical conditions (multimorbidity) complicate cancer diagnosis when they provide plausible diagnostic alternatives for cancer symptoms.

Aim

To investigate associations in bladder cancer between: first, pre-existing condition count and advanced-stage diagnosis; and, second, comorbidities that share symptoms with bladder cancer and advanced-stage diagnosis.

Design and setting

This observational UK cohort study was set in the Clinical Practice Research Datalink with Public Health England National Cancer Registration and Analysis Service linkage.

Method

Included participants were aged ≥40 years with an incident diagnosis of bladder cancer between 1 January 2000 and 31 December 2015, and primary care records of attendance for haematuria, dysuria, or abdominal mass in the year before diagnosis. Stage at diagnosis (stage 1 or 2 versus stage 3 or 4) was the outcome variable. Putative explanatory variables using logistic regression were examined, including patient-level count of pre-existing conditions and ‘alternative-explanations’, indicating whether pre-existing condition(s) were plausible diagnostic alternatives for the index cancer symptom.

Results

In total, 1468 patients (76.4% male) were studied, of which 399 (35.6%) males and 217 (62.5%) females had alternative explanations for their index cancer symptom, the most common being urinary tract infection with haematuria. Females were more likely than males to be diagnosed with advanced-stage cancer (adjusted odds ratio [aOR] 1.62; 95% confidence interval [CI] = 1.20 to 2.18; P = 0.001). Alternative explanations were strongly associated with advanced-stage diagnosis in both sexes (aOR 1.69; 95% CI = 1.20 to 2.39; P = 0.003).

Conclusion

Alternative explanations were associated with advanced-stage diagnosis of bladder cancer. Females were more likely than males to be diagnosed with advanced-stage disease, but the effect was not driven entirely by alternative explanations.

Measuring the impact of national guidelines: What methods can be used to uncover time-varying effects for healthcare evaluations?

We examine the suitability of three methods using patient-level data to evaluate the time-varying impacts of national healthcare guidelines. Such guidelines often codify progressive change and are implemented gradually; for example, National Institute for Health and Care Excellence (NICE) suspected-cancer referral guidelines. These were revised on June 23, 2015, to include more cancer symptoms and test results ("features"), partly reflecting changing practice. We explore the time-varying impact of guideline revision on time to colorectal cancer diagnosis, which is linked to improved outcomes in decision-analytic models. We included 11,842 patients diagnosed in 01/01/2006-31/12/2017 in the Clinical Practice Research Datalink with England cancer registry data linkage. Patients were classified by whether their first pre-diagnostic cancer feature was in the original guidelines (NICE-2005) or was added during the revision (NICE-2015-only). Outcome was diagnostic interval: time from first cancer feature to diagnosis. All analyses adjusted for age and sex. Two difference-in-differences analyses used either a Pre (01/08/2012-31/12/2014, n = 2243) and Post (01/08/2015-31/12/2017, n = 1017) design, or event-study cohorts (2006-2017 vs 2015) to estimate change in diagnostic interval attributable to official implementation of the revised guidelines. A semiparametric varying-coefficient model analysed the difference in diagnostic interval between the NICE groups over time. After model estimation, primary and broader treatment effects of guideline content and implementation were measured. The event-study difference-in-differences and the semiparametric varying-coefficient methods showed that shorter diagnostic intervals were attributable to official implementation of the revised guidelines. This impact was only detectable by pre-to-post difference-in-differences when the pre/post periods were selected according to the estimation results from the varying-coefficient model. Formal tests of the parametric models, which are special cases of the semiparametric model, suggest that they are misspecified. We conclude that the semiparametric method is well suited to explore the time-varying impacts of guidelines codifying progressive change.

Validation of acute exacerbation of chronic obstructive pulmonary disease (COPD) recording in electronic health records: a systematic review protocol

INTRODUCTION

Many patients with chronic obstructive pulmonary disease (COPD) experience a sustained worsening in symptoms termed an acute exacerbation (AECOPD). AECOPDs impact on patients' quality of life and lung function, are costly to health services and are an important topic for research. Electronic health records (EHR) are increasingly being used to study AECOPD, requiring accurate detection of AECOPD in EHRs to ensure generalisable results. The aim of this protocol is to provide an overview of studies that validate AECOPD definitions used in EHRs and administrative claims databases.

METHODS AND ANALYSIS

Medline and Embase will be searched for terms related to COPD exacerbation, EHRs and validation. All studies published between 1 January 1990 and 30 September 2019 written in English that validate AECOPD in EHRs and administrative claims databases will be considered.

INCLUSION CRITERIA

EHR data must be routinely collected; the AECOPD detection algorithm must be compared against a reference standard; and a measure of validity must be calculable. Two independent reviewers will screen articles for inclusion, extract study details and assess risk of bias using QUADAS-2. Disagreements will be resolved by consensus or arbitration by a third reviewer. This protocol has been developed in accordance with the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols checklist.

ETHICS AND DISSEMINATION

This will be a review of previously published literature therefore no ethical approval is required. Results from this review will be published in a peer-reviewed journal. The results can be used in future research to identify occurrences of AECOPD.

PROSPERO REGISTRATION NUMBER

CRD42019130863.

Changes in the presenting symptoms of lung cancer from 2000-2017: a serial cross-sectional study of observational records in UK primary care

Background

Most patients diagnosed with lung cancer present with symptoms. It is not known if the proportions of patients presenting with each symptom has changed over time. Identifying trends in lung cancer’s presenting symptoms is important for medical education and early-diagnosis initiatives.

Aim

To identify the first reported symptom of possible lung cancer (index symptom), and to test whether the percentages of patients with each index symptom changed during 2000–2017.

Design and setting

This was a serial, cross-sectional, observational study using UK Clinical Practice Research Datalink (CPRD) data with cancer registry linkage.

Method

The index symptom was identified for patients with an incident diagnosis of lung cancer in annual cohorts between 1 January 2000 and 31 December 2017. Searches were constrained to symptoms in National Institute for Health and Care Excellence (NICE) suspected-cancer referral guidelines, and to the year before diagnosis. Generalised linear models (with a binomial function) were used to test if the percentages of patients with each index symptom varied during 2000–2017.

Results

The percentage of patients with an index symptom of cough (odds ratio [OR] 1.01; 95% confidence interval [CI] = 1.00 to 1.02 per year; P<0.0001) or dyspnoea (OR 1.05; CI = 1.05 to 1.06 per year; P<0.0001) increased. The percentages of patients with other index symptoms decreased, notably haemoptysis (OR 0.93; CI = 0.92 to 0.95; P<0.0001) and appetite loss (OR 0.94; CI = 0.90 to 0.97; P<0.0001).

Conclusion

During 2000–2017, the proportions of lung cancer patients with an index symptom of cough or dyspnoea increased, while the proportion of those with the index symptom haemoptysis decreased. This trend has implications for medical education and symptom awareness campaigns.

Atopic Eczema in Adulthood and Risk of Depression and Anxiety: A Population-Based Cohort Study

BACKGROUND

Atopic eczema is a common and debilitating condition associated with depression and anxiety, but the nature of this association remains unclear.

OBJECTIVE

To explore the temporal relationship between atopic eczema and new depression/anxiety.

METHODS

This matched cohort study used routinely collected data from the UK Clinical Practice Research Datalink, linked to hospital admissions data. We identified adults with atopic eczema (1998-2016) using a validated algorithm, and up to 5 individuals without atopic eczema matched on date of diagnosis, age, sex, and general practice. We estimated the hazard ratio (HR) for new depression/anxiety using stratified Cox regression to account for age, sex, calendar period, Index of Multiple Deprivation, glucocorticoid treatment, obesity, smoking, and harmful alcohol use.

RESULTS

We identified 526,808 adults with atopic eczema who were matched to 2,569,030 without. Atopic eczema was associated with increased incidence of new depression (HR, 1.14; 99% CI, 1.12-1.16) and anxiety (HR, 1.17; 99% CI, 1.14-1.19). We observed a stronger effect of atopic eczema on depression with increasing atopic eczema severity (HR [99% CI] compared with no atopic eczema: mild, 1.10 [1.08-1.13]; moderate, 1.19 [1.15-1.23]; and severe, 1.26 [1.17-1.37]). A dose-response association, however, was less apparent for new anxiety diagnosis (HR [99% CI] compared with no atopic eczema: mild, 1.14 [1.11-1.18]; moderate, 1.21 [1.17-1.26]; and severe, 1.15; [1.05-1.25]).

CONCLUSIONS

Adults with atopic eczema are more likely to develop new depression and anxiety. For depression, we observed a dose-response relationship with atopic eczema severity.

The usefulness of symptoms alone or combined for general practitioners in considering the diagnosis of a brain tumour: a case-control study using the clinical practice research database (CPRD) (2000-2014)

OBJECTIVES

To evaluate the utility of different symptoms, alone or combined, presented to primary care for an adult brain tumour diagnosis.

DESIGN AND SETTING

Matched case-control study, using the data from Clinical Practice Research Datalink (2000-2014) from primary care consultations in the UK.

METHOD

All presentations within 6 months of the index diagnosis date (cases) or equivalent (controls) were coded into 32 symptom groups. Sensitivity, specificity, positive predictive values (PPVs) and positive likelihood ratios were calculated for symptoms and combinations of symptoms with headache and cognitive features. Diagnostic odds ratios were calculated using conditional logistic regression, adjusted for age group, sex and Charlson comorbidity. Stratified analyses were performed for age group, sex and whether the tumour was of primary or secondary origin.

RESULTS

We included 8,184 cases and 28,110 controls. Seizure had the highest PPV of 1.6% (95% CI 1.4% to 1.7%) followed by weakness 1.5% (1.3 to 1.7) and confusion 1.4% (1.3 to 1.5). Combining headache with other symptoms increased the PPV. For example, headache plus combined cognitive symptoms PPV 7.2% (6.0 to 8.6); plus weakness 4.4% (3.2 to 6.2), compared with headache alone PPV 0.1%. The diagnostic ORs were generally larger for patients <70 years; this was most marked for confusion, seizure and visual symptoms.

CONCLUSION

We found seizure, weakness and confusion had relatively higher predictive values than many other symptoms. Headache on its own was a weak predictor but this was enhanced when combined with other symptoms especially in younger patients. Clinicians need to actively search for other neurological symptoms such as cognitive problems.

Added value and cascade effects of inflammatory marker tests in UK primary care: a cohort study from the Clinical Practice Research Datalink

BACKGROUND

Inflammatory markers (C-reactive protein, erythrocyte sedimentation rate, and plasma viscosity) are commonly used in primary care. Though established for specific diagnostic purposes, there is uncertainty around their utility as a non-specific marker to rule out underlying disease in primary care.

AIM

To identify the value of inflammatory marker testing in primary care as a rule-out test, and measure the cascade effects of testing in terms of further blood tests, GP appointments, and referrals.

DESIGN AND SETTING

Cohort study of 160 000 patients with inflammatory marker testing in 2014, and 40 000 untested age, sex, and practice-matched controls, using UK primary care data from the Clinical Practice Research Datalink.

METHOD

The primary outcome was incidence of relevant disease, including infections, autoimmune conditions, and cancers, among those with raised versus normal inflammatory markers and untested controls. Process outcomes included rates of GP consultations, blood tests, and referrals in the 6 months after testing.

RESULTS

The overall incidence of disease following a raised inflammatory marker was 15%: 6.3% infections, 5.6% autoimmune conditions, 3.7% cancers. Inflammatory markers had an overall sensitivity of <50% for the primary outcome, any relevant disease (defined as any infections, autoimmune conditions, or cancers). For 1000 inflammatory marker tests performed, the authors would anticipate 236 false-positives, resulting in an additional 710 GP appointments, 229 phlebotomy appointments, and 24 referrals in the following 6 months.

CONCLUSION

Inflammatory markers have poor sensitivity and should not be used as a rule-out test. False-positive results are common and lead to increased rates of follow-on GP consultations, tests, and referrals.

Use of multiple inflammatory marker tests in primary care: using Clinical Practice Research Datalink to evaluate accuracy

Background

Research comparing C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), and plasma viscosity (PV) in primary care is lacking. Clinicians often test multiple inflammatory markers, leading to concerns about overuse.

Aim

To compare the diagnostic accuracies of CRP, ESR, and PV, and to evaluate whether measuring two inflammatory markers increases accuracy.

Design and setting

Prospective cohort study in UK primary care using the Clinical Practice Research Datalink.

Method

The authors compared diagnostic test performance of inflammatory markers, singly and paired, for relevant disease, defined as any infections, autoimmune conditions, or cancers. For each of the three tests (CRP, ESR, and PV), sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and area under receiver operator curve (AUC) were calculated.

Results

Participants comprised 136 961 patients with inflammatory marker testing in 2014; 83 761 (61.2%) had a single inflammatory marker at the index date, and 53 200 (38.8%) had multiple inflammatory markers. For ‘any relevant disease’, small differences were seen between the three tests; AUC ranged from 0.659 to 0.682. CRP had the highest overall AUC, largely because of marginally superior performance in infection (AUC CRP 0.617, versus ESR 0.589, P<0.001). Adding a second test gave limited improvement in the AUC for relevant disease (CRP 0.682, versus CRP plus ESR 0.688, P<0.001); this is of debatable clinical significance. The NPV for any single inflammatory marker was 94% compared with 94.1% for multiple negative tests.

Conclusion

Testing multiple inflammatory markers simultaneously does not increase ability to rule out disease and should generally be avoided. CRP has marginally superior diagnostic accuracy for infections, and is equivalent for autoimmune conditions and cancers, so should generally be the first-line test.

Predictive value of inflammatory markers for cancer diagnosis in primary care: a prospective cohort study using electronic health records

Background

Early identification of cancer in primary care is important and challenging. This study examined the diagnostic utility of inflammatory markers (C-reactive protein, erythrocyte sedimentation rate and plasma viscosity) for cancer diagnosis in primary care.

Methods

Cohort study of 160,000 patients with inflammatory marker testing in 2014, plus 40,000 untested matched controls, using Clinical Practice Research Datalink (CPRD), with Cancer Registry linkage. Primary outcome was one-year cancer incidence.

Results

Primary care patients with a raised inflammatory marker have a one-year cancer incidence of 3.53% (95% CI 3.37–3.70), compared to 1.50% (1.43–1.58) in those with normal inflammatory markers, and 0.97% (0.87–1.07) in untested controls. Cancer risk is greater with higher inflammatory marker levels, with older age and in men; risk rises further when a repeat test is abnormal but falls if it normalises. Men over 50 and women over 60 with raised inflammatory markers have a cancer risk which exceeds the 3% NICE threshold for urgent investigation. Sensitivities for cancer were 46.1% for CRP, 43.6% ESR and 49.7% for PV.

Conclusion

Cancer should be considered in patients with raised inflammatory markers. However, inflammatory markers have a poor sensitivity for cancer and are therefore not useful as ‘rule-out’ test.

Code sets for respiratory symptoms in electronic health records research: a systematic review protocol

INTRODUCTION

Asthma and chronic obstructive pulmonary disease (COPD) are common respiratory conditions, which result in significant morbidity worldwide. These conditions are associated with a range of non-specific symptoms, which in themselves are a target for health research. Such research is increasingly being conducted using electronic health records (EHRs), but computable phenotype definitions, in the form of code sets or code lists, are required to extract structured data from these large routine databases in a systematic and reproducible way. The aim of this protocol is to specify a systematic review to identify code sets for respiratory symptoms in EHRs research.

METHODS AND ANALYSIS

MEDLINE and Embase databases will be searched using terms relating to EHRs, respiratory symptoms and use of code sets. The search will cover all English-language studies in these databases between January 1990 and December 2017. Two reviewers will independently screen identified studies for inclusion, and key data will be extracted into a uniform table, facilitating cross-comparison of codes used. Disagreements between the reviewers will be adjudicated by a third reviewer. This protocol has been produced in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Protocol guidelines.

ETHICS AND DISSEMINATION

As a review of previously published studies, no ethical approval is required. The results of this review will be submitted to a peer-reviewed journal for publication and can be used in future research into respiratory symptoms that uses electronic healthcare databases.

PROSPERO REGISTRATION NUMBER

CRD42018100830.

Term sets: A transparent and reproducible representation of clinical code sets

OBJECTIVE

Clinical code sets are vital to research using routinely-collected electronic healthcare data. Existing code set engineering methods pose significant limitations when considering reproducible research. To improve the transparency and reusability of research, these code sets must abide by FAIR principles; this is not currently happening. We propose 'term sets', an equivalent alternative to code sets that are findable, accessible, interoperable and reusable.

MATERIALS AND METHODS

We describe a new code set representation, consisting of natural language inclusion and exclusion terms (term sets), and explain its relationship to code sets. We formally prove that any code set has a corresponding term set. We demonstrate utility by searching for recently published code sets, representing them as term sets, and reporting on the number of inclusion and exclusion terms compared with the size of the code set.

RESULTS

Thirty-one code sets from 20 papers covering diverse disease domains were converted into term sets. The term sets were on average 74% the size of their equivalent original code set. Four term sets were larger due to deficiencies in the original code sets.

DISCUSSION

Term sets can concisely represent any code set. This may reduce barriers for examining and reusing code sets, which may accelerate research using healthcare databases. We have developed open-source software that supports researchers using term sets.

CONCLUSION

Term sets are independent of clinical code terminologies and therefore: enable reproducible research; are resistant to terminology changes; and are less error-prone as they are shorter than the equivalent code set.

The internal validation of weight and weight change coding using weight measurement data within the UK primary care Electronic Health Record

PURPOSE

To use recorded weight values to internally validate weight status and weight change coding in the primary care Electronic Health Record (EHR).

PATIENTS AND METHODS

We included adult patients with weight-related Read codes recorded in the UK's Clinical Practice Research Datalink EHR between 2000 and 2017. Weight status codes were compared to weight values recorded on the same day and positive predictive values (PPVs) were calculated for commonly used codes. Weight change codes were validated using three methods: the percentage (%) difference in kilograms at the time of the code and 1) the previous weight measurement, 2) the weight predicted using linear regression, and 3) the historic mean weight. Weight change codes were validated if estimates were consistent across two out of three methods.

RESULTS

A total of 8,108,481 weight codes were recorded in 1,000,002 patients' EHR. Twice as many were recorded in females (n=5,208,593, 64%). The mean body mass index for "overweight" codes ranged from 31.9 kg/m2 to 46.9 kg/m2 and from 17.4 kg/m2 to 19.2 kg/m2 for "underweight" codes. PPVs for the most commonly used weight status codes ranged from 81.3% (80%-82.5%) to 99.3% (99.2%-99.4%). Across the estimation methods, and using only validated weight change codes, mean weight loss ranged from - 5.2% (SD 5.8%) to -7.9% (SD 7.3%) and mean weight gain from 4.2 % (SD 5.5%) to 7.9 % (SD 8.2%). The previous and predicted weight methods were most consistent.

CONCLUSION

We have developed an internationally applicable methodology to internally validate weight-related EHR coding by using available weight measurement data. We demonstrate the UK Read codes that can be confidently used to classify weight status and weight change in the absence of weight values. We provide the first evidence from primary care that a Read code for unexpected weight loss represents a mean loss of ≥ 5 % in a 6-month period, which was broadly consistent across age groups and gender.