Identifying and handling data bias within primary healthcare data using synthetic data generators.

Advanced synthetic data generators can simulate data samples that closely resemble sensitive personal datasets while significantly reducing the risk of individual identification. The use of these advanced generators holds enormous potential in the medical field, as it allows for the simulation and sharing of sensitive patient data. This enables the development and rigorous validation of novel AI technologies for accurate diagnosis and efficient disease management. Despite the availability of massive ground truth datasets (such as UK-NHS databases that contain millions of patient records), the risk of biases being carried over to data generators still exists. These biases may arise from the under-representation of specific patient cohorts due to cultural sensitivities within certain communities or standardised data collection procedures. Machine learning models can exhibit bias in various forms, including the under-representation of certain groups in the data. This can lead to missing data and inaccurate correlations and distributions, which may also be reflected in synthetic data. Our paper aims to improve synthetic data generators by introducing probabilistic approaches to first detect difficult-to-predict data samples in ground truth data and then boost them when applying the generator. In addition, we explore strategies to generate synthetic data that can reduce bias and, at the same time, improve the performance of predictive models.

The cost of primary care consultations associated with long COVID in non-hospitalised adults: a retrospective cohort study using UK primary care data.

BACKGROUND: The economic impact of managing long COVID in primary care is unknown. We estimated the costs of primary care consultations associated with long COVID and explored the relationship between risk factors and costs. METHODS: Data were obtained on non-hospitalised adults from the Clinical Practice Research Datalink Aurum primary care database. We used propensity score matching with an incremental cost method to estimate additional primary care consultation costs associated with long COVID (12 weeks after COVID-19) at an individual and UK national level. We applied multivariable regression models to estimate the association between risk factors and consultations costs beyond 12 weeks from acute COVID-19. RESULTS: Based on an analysis of 472,173 patients with COVID-19 and 472,173 unexposed individuals, the annual incremental cost of primary care consultations associated with long COVID was £2.44 per patient and £23,382,452 at the national level. Among patients with COVID-19, a long COVID diagnosis and reporting of longer-term symptoms were associated with a 43% and 44% increase in primary care consultation costs respectively, compared to patients without long COVID symptoms. Older age, female sex, obesity, being from a white ethnic group, comorbidities and prior consultation frequency were all associated with increased primary care consultation costs. CONCLUSIONS: The costs of primary care consultations associated with long COVID in non-hospitalised adults are substantial. Costs are significantly higher among those diagnosed with long COVID, those with long COVID symptoms, older adults, females, and those with obesity and comorbidities.

Completeness, agreement, and representativeness of ethnicity recording in the United Kingdom's Clinical Practice Research Datalink (CPRD) and linked Hospital Episode Statistics (HES).

BACKGROUND: This descriptive study assessed the completeness, agreement, and representativeness of ethnicity recording in the United Kingdom (UK) Clinical Practice Research Datalink (CPRD) primary care databases alone and, for those patients registered with a GP in England, when linked to secondary care data from Hospital Episode Statistics (HES). METHODS: Ethnicity records were assessed for all patients in the May 2021 builds of the CPRD GOLD and CPRD Aurum databases for all UK patients. In analyses of the UK, English data was from combined CPRD-HES, whereas data from Northern Ireland, Scotland, and Wales drew from CPRD only. The agreement of ethnicity records per patient was assessed within each dataset (CPRD GOLD, CPRD Aurum, and HES datasets) and between datasets at the highest level ethnicity categorisation ('Asian', 'black', 'mixed', 'white', 'other'). Representativeness was assessed by comparing the ethnic distributions at the highest-level categorisation of CPRD-HES to those from the Census 2011 across the UK's devolved administrations. Additionally, CPRD-HES was compared to the experimental ethnic distributions for England and Wales from the Office for National Statistics in 2019 (ONS2019) and the English ethnic distribution from May 2021 from NHS Digital's General Practice Extraction Service Data for Pandemic Planning and Research with HES data linkage (GDPPR-HES). RESULTS: In CPRD-HES, 81.7% of currently registered patients in the UK had ethnicity recorded in primary care. For patients with multiple ethnicity records, mismatched ethnicity within individual primary and secondary care datasets was < 10%. Of English patients with ethnicity recorded in both CPRD and HES, 93.3% of records matched at the highest-level categorisation; however, the level of agreement was markedly lower in the 'mixed' and 'other' ethnic groups. CPRD-HES was less proportionately 'white' compared to the UK Census 2011 (80.3% vs. 87.2%) and experimental ONS2019 data (80.4% vs. 84.3%). CPRD-HES was aligned with the ethnic distribution from GDPPR-HES ('white' 80.4% vs. 80.7%); however, with a smaller proportion classified as 'other' (1.1% vs. 2.8%). CONCLUSIONS: CPRD-HES has suitable representation of all ethnic categories with some overrepresentation of minority ethnic groups and a smaller proportion classified as 'other' compared to the UK general population from other data sources. CPRD-HES data is useful for studying health risks and outcomes in typically underrepresented groups.

Healthcare workers knowledge of cholera multi-stranded interventions and its determining factors in North-East Nigeria: planning and policy implications.

BACKGROUND: Healthcare workers' (HCWs) knowledge of multi-stranded cholera interventions (including case management, water, sanitation, and hygiene (WASH), surveillance/laboratory methods, coordination, and vaccination) is crucial to the implementation of these interventions in healthcare facilities, especially in conflict-affected settings where cholera burden is particularly high. We aimed to assess Nigerian HCWs' knowledge of cholera interventions and identify the associated factors. METHODS: We conducted a cross-sectional study using a structured interviewer-administered questionnaire with HCWs from 120 healthcare facilities in Adamawa and Bauchi States, North-East Nigeria. A knowledge score was created by assigning a point for each correct response. HCWs' knowledge of cholera interventions, calculated as a score, was recoded for ease of interpretation as follows: 0-50 (low); 51-70 (moderate); ≥ 71 (high). Additionally, we defined the inadequacy of HCWs' knowledge of cholera interventions based on a policy-relevant threshold of equal or lesser than 75 scores for an intervention. Multivariable logistic regression was used to identify the factors associated with the adequacy of knowledge score. RESULTS: Overall, 490 HCWs participated in the study (254 in Adamawa and 236 in Bauchi), with a mean age of 35.5 years. HCWs' knowledge score was high for surveillance/laboratory methods, moderate for case management, WASH, and vaccination, and low for coordination. HCWs' knowledge of coordination improved with higher cadre, working in urban- or peri-urban-based healthcare facilities, and secondary education; cholera case management and vaccination knowledge improved with post-secondary education, working in Bauchi State and urban areas, previous training in cholera case management and response to a cholera outbreak-working in peri-urban areas had a negative effect. HCWs' knowledge of surveillance/laboratory methods improved with a higher cadre, 1-year duration in current position, secondary or post-secondary education, previous training in cholera case management and response to a cholera outbreak. However, HCWs' current position had both positive and negative impacts on their WASH knowledge. CONCLUSIONS: HCWs in both study locations recorded a considerable knowledge of multi-stranded cholera interventions. While HCWs' demographic characteristics appeared irrelevant in determining their knowledge of cholera interventions, geographic location and experiences from the current position, training and involvement in cholera outbreak response played a significant role.

Quality of rheumatoid arthritis recording in United Kingdom Clinical Practice Research Datalink Aurum.

PURPOSE: While several studies have assessed quality and completeness of recording acute medical events in Clinical Practice Research Datalink (CPRD) Aurum, evaluation of additional chronic conditions is warranted. METHODS: We selected patients with a first diagnosis of rheumatoid arthritis (RA) coded in their CPRD Aurum record between 2005 and 2019. We assessed quality of RA diagnosis by evaluating additional information in the patient record that would corroborate the diagnosis. We report recording of diagnoses, prescriptions, labs, and referrals expected to be present based on NICE guidelines for RA management. RESULTS: There were 53 083 patients with a first recorded RA diagnosis during the study period: 43606 (82%) patients had RA drug treatments in their record, 7596 (14%) had supporting codes without drug treatment, and 1881 (4%) patients had only a RA diagnoses recorded in their medical record with no supporting codes or RA treatments. Patients with RA diagnosis only were more likely to be first diagnosed in the earliest time period of study. Labs for diagnosing and monitoring RA were most common among patients with RA treatment. Analgesic and glucocorticoid prescriptions were common in all study patients but were highest among patients with RA treatment. Among patients with RA diagnosis only, the overwhelming majority had only one RA diagnosis recorded (76%). CONCLUSIONS: Our findings suggest that codes expected for monitoring and treatment of RA are routinely recorded in CPRD Aurum. These results support previous assessments, which found data recorded in CPRD Aurum to be of good quality for use in research.

Epidemiology, diagnostics and factors associated with mortality during a cholera epidemic in Nigeria, October 2020-October 2021: a retrospective analysis of national surveillance data.

OBJECTIVES: Nigeria reported an upsurge in cholera cases in October 2020, which then transitioned into a large, disseminated epidemic for most of 2021. This study aimed to describe the epidemiology, diagnostic performance of rapid diagnostic test (RDT) kits and the factors associated with mortality during the epidemic. DESIGN: A retrospective analysis of national surveillance data. SETTING: 33 of 37 states (including the Federal Capital Territory) in Nigeria. PARTICIPANTS: Persons who met cholera case definition (a person of any age with acute watery diarrhoea, with or without vomiting) between October 2020 and October 2021 within the Nigeria Centre for Disease Control surveillance data. OUTCOME MEASURES: Attack rate (AR; per 100 000 persons), case fatality rate (CFR; %) and accuracy of RDT performance compared with culture using area under the receiver operating characteristic curve (AUROC). Additionally, individual factors associated with cholera deaths and hospitalisation were presented as adjusted OR with 95% CIs. RESULTS: Overall, 93 598 cholera cases and 3298 deaths (CFR: 3.5%) were reported across 33 of 37 states in Nigeria within the study period. The proportions of cholera cases were higher in men aged 5-14 years and women aged 25-44 years. The overall AR was 46.5 per 100 000 persons. The North-West region recorded the highest AR with 102 per 100 000. Older age, male gender, residency in the North-Central region and severe dehydration significantly increased the odds of cholera deaths. The cholera RDT had excellent diagnostic accuracy (AUROC=0.91; 95% CI 0.87 to 0.96). CONCLUSIONS: Cholera remains a serious public health threat in Nigeria with a high mortality rate. Thus, we recommend making RDT kits more widely accessible for improved surveillance and prompt case management across the country.

Completeness and representativeness of small area socioeconomic data linked with the UK Clinical Practice Research Datalink (CPRD).

BACKGROUND: The Clinical Practice Research Datalink (CPRD) holds primary care electronic healthcare records for 25% of the UK population. CPRD data can be linked via practice postcode in the UK, and additionally via patient postcode in England, to area-level socioeconomic status (SES) data including the Index of Multiple Deprivation (IMD), the Carstairs Index and the Townsend Deprivation Index; as well as rural-urban classification (RUC). This study aims to describe the completeness and representativeness of CPRD-linked SES and RUC data. METHODS: Patients currently registered at general practices contributing data to the May 2021 snapshots of CPRD GOLD (n=445 587) and CPRD Aurum (n=13 278 825) were used to assess the completeness and representativeness of CPRD-linked SES and RUC data against the UK general population. RESULTS: All currently registered patients had complete SES and RUC data at practice level across the UK. Most English patients in CPRD GOLD (78%), CPRD Aurum (94%) and combined (93%) had SES and RUC data at patient level. Patient-level SES data in CPRD for England were comparable to England's general population (average IMD decile in CPRD 5.52±0.00 vs 5.50±0.02). CPRD UK practices were on average in more deprived areas than the UK general population (6.06±0.07 vs 5.50±0.02). A slightly higher proportion of CPRD patients and practices were from urban areas (85%) as compared with the UK general population (82%). CONCLUSION: Completeness of CPRD-linked SES and RUC data is high. The CPRD populations were broadly representative of the general populations in the UK in terms of SES and RUC.

Symptoms and risk factors for long COVID in non-hospitalized adults.

Severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) infection is associated with a range of persistent symptoms impacting everyday functioning, known as post-COVID-19 condition or long COVID. We undertook a retrospective matched cohort study using a UK-based primary care database, Clinical Practice Research Datalink Aurum, to determine symptoms that are associated with confirmed SARS-CoV-2 infection beyond 12 weeks in non-hospitalized adults and the risk factors associated with developing persistent symptoms. We selected 486,149 adults with confirmed SARS-CoV-2 infection and 1,944,580 propensity score-matched adults with no recorded evidence of SARS-CoV-2 infection. Outcomes included 115 individual symptoms, as well as long COVID, defined as a composite outcome of 33 symptoms by the World Health Organization clinical case definition. Cox proportional hazards models were used to estimate adjusted hazard ratios (aHRs) for the outcomes. A total of 62 symptoms were significantly associated with SARS-CoV-2 infection after 12 weeks. The largest aHRs were for anosmia (aHR 6.49, 95% CI 5.02-8.39), hair loss (3.99, 3.63-4.39), sneezing (2.77, 1.40-5.50), ejaculation difficulty (2.63, 1.61-4.28) and reduced libido (2.36, 1.61-3.47). Among the cohort of patients infected with SARS-CoV-2, risk factors for long COVID included female sex, belonging to an ethnic minority, socioeconomic deprivation, smoking, obesity and a wide range of comorbidities. The risk of developing long COVID was also found to be increased along a gradient of decreasing age. SARS-CoV-2 infection is associated with a plethora of symptoms that are associated with a range of sociodemographic and clinical risk factors.

Presence of Codes for Indication for Use in Clinical Practice Research Datalink Aurum: An Assessment of Benign Prostatic Hyperplasia Treatments.

Background: Assessments of strengths and limitations of new data sources are critical for making decisions about suitability for specific research questions. For some studies, it is necessary to capture a drug's indication for use. Objective: To assess the presence of indications for prescription use in Clinical Practice Research Datalink (CPRD) Aurum (January 1988-June 2021) by describing the proportion of men in CPRD Aurum who had a recorded indication for use of prescriptions for 5-alpha reductase inhibitors (5-ARI), alpha blockers (AB), or tadalafil, which have multiple indications. Methods: From a random sample of 154 practices of CPRD Aurum data, we selected 85,597 male patients with a prescription for a 5-ARI, an AB, or tadalafil. Among these patients, we described presence of codes indicating whether the patient had benign prostatic hyperplasia, hypertension, erectile dysfunction, or alopecia using three indication definitions: narrow (specific diagnoses recorded within one year before and up to 90 days after the prescription), broad (specific diagnoses or supporting clinical codes in the time period described above), and widest (diagnoses or supporting codes recorded at any time before the prescription and up to 90 days after the prescription). Results: Using the narrow indication definition limited to diagnoses only, 39,861 (46.6%) patients' records contained an indication for use. The broad definitions, which additionally included supporting codes, captured indications for 62,912 (73.5%) patients and the widest definition, which additionally included supporting codes and all available data before the first prescription date, captured indications for 71,478 (83.5%) patients. Indications were present more often for prescriptions in 2005 and later (85.9%). Conclusion: The findings of this assessment suggest that CPRD Aurum can be used for studies that require information on treatment indications for BPH and potentially for treatments of other chronic diseases managed in the primary care setting.

Therapies for Long COVID in non-hospitalised individuals: from symptoms, patient-reported outcomes and immunology to targeted therapies (The TLC Study).

INTRODUCTION: Individuals with COVID-19 frequently experience symptoms and impaired quality of life beyond 4-12 weeks, commonly referred to as Long COVID. Whether Long COVID is one or several distinct syndromes is unknown. Establishing the evidence base for appropriate therapies is needed. We aim to evaluate the symptom burden and underlying pathophysiology of Long COVID syndromes in non-hospitalised individuals and evaluate potential therapies. METHODS AND ANALYSIS: A cohort of 4000 non-hospitalised individuals with a past COVID-19 diagnosis and 1000 matched controls will be selected from anonymised primary care records from the Clinical Practice Research Datalink, and invited by their general practitioners to participate on a digital platform (Atom5). Individuals will report symptoms, quality of life, work capability and patient-reported outcome measures. Data will be collected monthly for 1 year.Statistical clustering methods will be used to identify distinct Long COVID-19 symptom clusters. Individuals from the four most prevalent clusters and two control groups will be invited to participate in the BioWear substudy which will further phenotype Long COVID symptom clusters by measurement of immunological parameters and actigraphy.We will review existing evidence on interventions for postviral syndromes and Long COVID to map and prioritise interventions for each newly characterised Long COVID syndrome. Recommendations will be made using the cumulative evidence in an expert consensus workshop. A virtual supportive intervention will be coproduced with patients and health service providers for future evaluation.Individuals with lived experience of Long COVID will be involved throughout this programme through a patient and public involvement group. ETHICS AND DISSEMINATION: Ethical approval was obtained from the Solihull Research Ethics Committee, West Midlands (21/WM/0203). Research findings will be presented at international conferences, in peer-reviewed journals, to Long COVID patient support groups and to policymakers. TRIAL REGISTRATION NUMBER: 1567490.

COVID-19 mortality rate and its associated factors during the first and second waves in Nigeria.

COVID-19 mortality rate has not been formally assessed in Nigeria. Thus, we aimed to address this gap and identify associated mortality risk factors during the first and second waves in Nigeria. This was a retrospective analysis of national surveillance data from all 37 States in Nigeria between February 27, 2020, and April 3, 2021. The outcome variable was mortality amongst persons who tested positive for SARS-CoV-2 by Reverse-Transcriptase Polymerase Chain Reaction. Incidence rates of COVID-19 mortality was calculated by dividing the number of deaths by total person-time (in days) contributed by the entire study population and presented per 100,000 person-days with 95% Confidence Intervals (95% CI). Adjusted negative binomial regression was used to identify factors associated with COVID-19 mortality. Findings are presented as adjusted Incidence Rate Ratios (aIRR) with 95% CI. The first wave included 65,790 COVID-19 patients, of whom 994 (1∙51%) died; the second wave included 91,089 patients, of whom 513 (0∙56%) died. The incidence rate of COVID-19 mortality was higher in the first wave [54∙25 (95% CI: 50∙98-57∙73)] than in the second wave [19∙19 (17∙60-20∙93)]. Factors independently associated with increased risk of COVID-19 mortality in both waves were: age ≥45 years, male gender [first wave aIRR 1∙65 (1∙35-2∙02) and second wave 1∙52 (1∙11-2∙06)], being symptomatic [aIRR 3∙17 (2∙59-3∙89) and 3∙04 (2∙20-4∙21)], and being hospitalised [aIRR 4∙19 (3∙26-5∙39) and 7∙84 (4∙90-12∙54)]. Relative to South-West, residency in the South-South and North-West was associated with an increased risk of COVID-19 mortality in both waves. In conclusion, the rate of COVID-19 mortality in Nigeria was higher in the first wave than in the second wave, suggesting an improvement in public health response and clinical care in the second wave. However, this needs to be interpreted with caution given the inherent limitations of the country's surveillance system during the study.

Systematic approach to outcome assessment from coded electronic healthcare records in the DaRe2THINK NHS-embedded randomized trial.

Aims: Improving the efficiency of clinical trials is key to their continued importance in directing evidence-based patient care. Digital innovations, in particular the use of electronic healthcare records (EHRs), allow for large-scale screening and follow up of participants. However, it is critical these developments are accompanied by robust and transparent methods that can support high-quality and high clinical value research. Methods and results: The DaRe2THINK trial includes a series of novel processes, including nationwide pseudonymized pre screening of the primary-care EHR across England, digital enrolment, remote e-consent, and 'no-visit' follow up by linking all primary- and secondary-care health data with patient-reported outcomes. DaRe2THINK is a pragmatic, healthcare-embedded randomized trial testing whether earlier use of direct oral anticoagulants in patients with prior or current atrial fibrillation can prevent thromboembolic events and cognitive decline (www.birmingham.ac.uk/dare2think). This study outlines the systematic approach and methodology employed to define patient information and outcome events. This includes transparency on all medical code lists and phenotypes used in the trial across a variety of national data sources, including Clinical Practice Research Datalink Aurum (primary care), Hospital Episode Statistics (secondary care), and the Office for National Statistics (mortality). Conclusion: Co-designed by a patient and public involvement team, DaRe2THINK presents an opportunity to transform the approach to randomized trials in the setting of routine healthcare, providing high-quality evidence generation in populations representative of the community at risk.

Quality and Completeness of Myocardial Infarction Recording in Clinical Practice Research Datalink Aurum.

BACKGROUND: Validation studies of the Clinical Practice Research Datalink (CPRD) Aurum database in the UK are critical for making decisions about its suitability and validity for research purposes. OBJECTIVE: To examine data source agreement of myocardial infarction (MI) diagnoses recorded in CPRD Aurum compared with linked Hospital Episode Statistics (HES) data. This comparison provides information on CPRD Aurum data correctness (accuracy, validity) and completeness (presence, missingness). METHODS: Patients with MI diagnoses recorded in either data source were selected from a random sample of 50,000 patients in CPRD Aurum with HES linkage (1997-2017). Correctness was defined as the proportion of MI cases in CPRD Aurum with a concordant MI diagnosis recorded in HES or with strong supporting evidence in either data source. Completeness was defined as the proportion of patients with primary HES-coded MIs with strong supporting evidence that were also present in CPRD Aurum. RESULTS: There were 1260 patients with MI recorded in the CPRD Aurum sample. The overall correctness of the recorded MI diagnoses was 94%: 986 patients (78%) had concordant diagnoses in HES within 90 days; 123 (10%) were concordant with HES, but with an inconclusive date and another 71 (6%) had strong supporting evidence for being a true MI case. There were 1125 patients with MI recorded in HES primary diagnosis fields with strong supporting evidence in either data source. Of these, 880 (78%) were present in CPRD Aurum, with completeness somewhat higher in more recent years. CONCLUSION: MI diagnoses recorded in CPRD Aurum were highly likely to be correct, supporting its use in clinical research studies. Completeness was lower, indicating the need for data linkage for some studies.

Assessing the capacity of symptom scores to predict COVID-19 positivity in Nigeria: a national derivation and validation cohort study.

OBJECTIVES: This study aimed to develop and validate a symptom prediction tool for COVID-19 test positivity in Nigeria. DESIGN: Predictive modelling study. SETTING: All Nigeria States and the Federal Capital Territory. PARTICIPANTS: A cohort of 43 221 individuals within the national COVID-19 surveillance dataset from 27 February to 27 August 2020. Complete dataset was randomly split into two equal halves: derivation and validation datasets. Using the derivation dataset (n=21 477), backward multivariable logistic regression approach was used to identify symptoms positively associated with COVID-19 positivity (by real-time PCR) in children (≤17 years), adults (18-64 years) and elderly (≥65 years) patients separately. OUTCOME MEASURES: Weighted statistical and clinical scores based on beta regression coefficients and clinicians' judgements, respectively. Using the validation dataset (n=21 744), area under the receiver operating characteristic curve (AUROC) values were used to assess the predictive capacity of individual symptoms, unweighted score and the two weighted scores. RESULTS: Overall, 27.6% of children (4415/15 988), 34.6% of adults (9154/26 441) and 40.0% of elderly (317/792) that had been tested were positive for COVID-19. Best individual symptom predictor of COVID-19 positivity was loss of smell in children (AUROC 0.56, 95% CI 0.55 to 0.56), either fever or cough in adults (AUROC 0.57, 95% CI 0.56 to 0.58) and difficulty in breathing in the elderly (AUROC 0.53, 95% CI 0.48 to 0.58) patients. In children, adults and the elderly patients, all scoring approaches showed similar predictive performance. CONCLUSIONS: The predictive capacity of various symptom scores for COVID-19 positivity was poor overall. However, the findings could serve as an advocacy tool for more investments in resources for capacity strengthening of molecular testing for COVID-19 in Nigeria.

Use of Primary Care Data in Research and Pharmacovigilance: Eight Scenarios Where Prescription Data are Absent.

The use of primary care databases has been integral in pharmacoepidemiological studies and pharmacovigilance. Primary care databases derive from electronic health records and offer a comprehensive description of aggregate patient data, from demography to medication history, and good sample sizes. Studies using these databases improve our understanding of prescribing characteristics and associated risk factors to facilitate better patient care, but there are limitations. We describe eight key scenarios where study data outcomes can be affected by absent prescriptions in UK primary care databases: (1) out-of-hours, urgent care and acute care prescriptions; (2) specialist-only prescriptions; (3) alternative community prescribing, such as pharmacy, family planning clinic or sexual health clinic medication prescriptions; (4) newly licensed medication prescriptions; (5) medications that do not require prescriptions; (6) hospital inpatient and outpatient prescriptions; (7) handwritten prescriptions; and (8) private pharmacy and private doctor prescriptions. The significance of each scenario is dependent on the type of medication under investigation, nature of the study and expected outcome measures. We recommend that all researchers using primary care databases be aware of the potential for missing prescribing data and be sensitive to how this can vary substantially between items, drug classes, patient groups and over time. Close liaison with practising primary care clinicians in the UK is often essential to ensure awareness of nuances in clinical practice.

Use of prescribing safety quality improvement reports in UK general practices: a qualitative assessment.

BACKGROUND: Quality improvement (QI) initiatives are increasingly used to improve the quality of care and reduce prescribing errors. The Royal College of General Practitioners (RCGP) and Clinical Practice Research Datalink (CPRD) QI initiative uses routinely collected electronic primary care data to provide bespoke practice-level reports on prescribing safety. The aim of this study was to explore how the QI reports were used, barriers and facilitators to use, long-term culture change and perceived impact on patient care and practices systems as a result of receiving the reports. METHODS: A qualitative study using purposive sampling of practices contributing to the CPRD, semi-structured interviews and inductive thematic analysis. We interviewed general practitioners, pharmacists, practice managers and research nurses. RESULTS: We conducted 18 interviews, and organised themes summarising the use of QI reports in practice: receiving the report, facilitators and barriers to acting upon the reports, acting upon the report, and how the reports contribute to a quality culture. Effective dissemination of reports, and a positive attitude to audit and the perceived relevance of the clinical topic facilitated use. Lack of time and failure to see or act upon the reports meant they were not used. Factors influencing use of the reports included the structure of the report, ease of identifying cases, and perceptions about coding accuracy. GPs and pharmacists used the reports to conduct case reviews and directly contact patients to discuss unsafe prescribing and patient medication preferences. Finally, the reports contributed to the development of a quality culture within practices through promoting audit activity and acting as a reminder of good prescribing behaviours, promoting future patient safety initiatives, contributing to continuing professional development and improving local networks. CONCLUSIONS: This study found the reports facilitated individual case review leading to an enhanced sense of quality culture in practices where they were utilised. Our findings demonstrate that the reports were generally considered useful and have been used to support patient safety and clinical practice in specific cases.

Self-reported diseases and their associated risk factors among camp-dwelling conflict-affected internally displaced populations in Nigeria.

Background Conflict in Nigeria displaced millions of people, and some settled in camp-like locations within the country. Evidence on the association between living conditions and health outcomes among these populations are limited. This study investigated the risk factors associated with illnesses among camp-dwelling internally displaced persons (IDPs) in northern Nigeria. Methods A cross-sectional study was conducted in nine camps in 2016. Self-reported data on socio-demography, resource utilization and disease outcomes were collected. Association between health conditions and various factors, including sanitation and healthcare access, was investigated. Results Data from 2253 IDPs showed 81.1% (CI = 79.5-82.7) experienced one or more health conditions; however, over 20% did not access healthcare services. Most common diseases were malaria, fever, typhoid and diarrhoea. Multivariable logistic regression presented as adjusted odds ratios(aOR) and 95% confidence intervals(CIs) showed factors significantly associated with increased likelihood of illnesses included being female (aOR = 1.53;CI = 1.19-1.96), overcrowding (aOR = 1.07;CI = 1.00-1.36), long-term conditions (aOR = 2.72;CI = 1.88-3.94), outdoor defecation (aOR = 2.37;CI = 1.14-4.94) and presence of disease-causing vectors (aOR = 3.71;CI = 1.60-8.60). Conclusion Most diseases in the camps were communicable. Modifiable risk factors such as overcrowding and poor toilet facilities were associated with increased poor health outcomes. This evidence highlights areas of high priority when planning humanitarian public health interventions.

Generating high-fidelity synthetic patient data for assessing machine learning healthcare software.

There is a growing demand for the uptake of modern artificial intelligence technologies within healthcare systems. Many of these technologies exploit historical patient health data to build powerful predictive models that can be used to improve diagnosis and understanding of disease. However, there are many issues concerning patient privacy that need to be accounted for in order to enable this data to be better harnessed by all sectors. One approach that could offer a method of circumventing privacy issues is the creation of realistic synthetic data sets that capture as many of the complexities of the original data set (distributions, non-linear relationships, and noise) but that does not actually include any real patient data. While previous research has explored models for generating synthetic data sets, here we explore the integration of resampling, probabilistic graphical modelling, latent variable identification, and outlier analysis for producing realistic synthetic data based on UK primary care patient data. In particular, we focus on handling missingness, complex interactions between variables, and the resulting sensitivity analysis statistics from machine learning classifiers, while quantifying the risks of patient re-identification from synthetic datapoints. We show that, through our approach of integrating outlier analysis with graphical modelling and resampling, we can achieve synthetic data sets that are not significantly different from original ground truth data in terms of feature distributions, feature dependencies, and sensitivity analysis statistics when inferring machine learning classifiers. What is more, the risk of generating synthetic data that is identical or very similar to real patients is shown to be low.