How Clinical Practice Research Datalink data are used to support pharmacovigilance

Developing electronic health records as a source of real-world data for veterinary pharmacoepidemiology

Spontaneous reporting of adverse events (AEs) by veterinary professionals and the public is the cornerstone of post-marketing safety surveillance for veterinary medicinal products (VMPs). However, studies suggest that most veterinary AEs remain unreported. Veterinary medicine regulators, including the United Kingdom Veterinary Medicines Directorate and the European Medicines Agency, have included the exploration of big data utilization to support pharmacovigilance efforts in their regulatory strategies. In this study, we describe the application of veterinary electronic healthcare records (EHRs) from the SAVSNET veterinary first opinion informatics system to conduct pharmacoepidemiological analyses. Five VMP-AE pairs were selected for investigation in a proof-of-concept study, where drug exposure was identified from semi-structured treatment data and AEs from the unstructured free-text clinical narrative. Dictionaries were developed to identify AEs based on standard terminology. The precision of these dictionaries improved when they were expanded using word vectorization and expert opinion. A key strength of first-opinion EHR datasets is their ability to enable cohort studies and facilitate calculations of absolute incidence and relative risk. Thus, we demonstrate that unstructured free-text clinical narratives can be used to identify outcomes for veterinary pharmacoepidemiological studies and, consequently, support and expand pharmacovigilance efforts based on spontaneous AE reports.

The Clinical Practice Research Datalink (CPRD) Mother-Baby Links: A Data Resource Profile

PURPOSE

Maternal exposures before, during and after pregnancy can affect the infant. It is therefore important that researchers study mothers and their children. The CPRD GOLD Mother-Baby Link (MBL) algorithm was applied to the CPRD Aurum database, to extend the useful tool. Here, we present the algorithm and data resource profiles of the CPRD MBLs.

METHODS

Records of female patients registered with a CPRD practice between the 1st January 1987 and the 1st June 2023 were searched for evidence of delivery. Infants born and registered between 1st January 1987 and 1st June 2023 were matched to mothers on practice and household indicators. The resulting MBLs were characterised.

RESULTS

The CPRD MBL algorithm was applied to the CPRD databases resulting in nearly four-million mother-baby pairs: 2.4-million in CPRD Aurum. Mothers in the CPRD GOLD and CPRD Aurum MBL's were similar in terms of age; mean age 29.6 years (SD = 5.7) vs. 30.2 years (SD = 5.7), and length of GP registration; mean = 14.4 years (SD = 10.9) vs. mean = 13.7 (SD = 10.9). The median number of matches was slightly higher in the CPRD GOLD MBL; 2 (IQR = 1, 2) vs. 1 (IQR = 1, 2). The number of matches in both databases peaked in 2008-2011, followed by a steady decline to 2023.

CONCLUSION

The CPRD MBL's offer a valuable tool for researchers to study the mother-infant relationship. Extending the CPRD MBL algorithm to CPRD Aurum has increased the capacity for researchers to investigate rarer exposures and outcomes.

Excess health care use is significantly and persistently reduced following diagnosis of late-onset epilepsy

OBJECTIVE

The incidence of late-onset epilepsy (LOE) is rising, and these patients may use an excess of health care resources. This study aimed to measure pre-/post-diagnostic health care use (HCU) for patients with LOE compared to controls.

METHODS

This was an observational open cohort study covering years 1998-2019 using UK population-based linked primary care (Clinical Practice Research Datalink [CPRD]) and hospital (HES) electronic health records. The participants included patients with incident LOE enrolled in CPRD and 1:10 age-, sex-, and general practice-matched controls. The exposure was incident LOE (diagnosed at age ≥65) using a 5-year washout. The main outcome was all HCU (primary care [PC], accident and emergency [A&E], admitted patient and outpatient care) using inverse proportional weighting to PC use and HCU by setting. An interrupted time-series analysis was used to examine pre-/post-diagnostic HCU between patients with LOE and controls over 4 years either side of diagnosis/matching date. An adjusted mixed-effects negative binomial regression was used for post-diagnosis HCU interactions.

RESULTS

Of 2 569 874 people ≥65 years of age, 1048 (4%) developed incident LOE. Mean weighted total HCU increased by 32 visits per patient-year (95% confidence interval [95% CI]: 13-50, p = .003) until LOE diagnosis, and then dropped by a mean of 60 visits per patient-year (95% CI: -81 to -40). There was an acute rise and fall over the 1-2 years immediately pre-/post-diagnosis. Incident HCU remained higher for LOE compared to controls post-diagnosis (adjusted incidence rate ratio: 1.72; 95% CI: 1.65-1.70; p < .001), including A&E, outpatient, and admitted care.

SIGNIFICANCE

Health care use demonstrates an acute on chronic rise over the 4 years before diagnosis of LOE. To what extent the partial reversal of the acute pre-diagnosis rise, and the mediators of the accelerated increase compared to controls are attributed to epilepsy, comorbid and bidirectional disease states, or a combination of both warrants further exploration.

Healthcare utilisation and associated costs for methadone versus buprenorphine recipients: Examination of interlinked primary and secondary care electronic health records in England

INTRODUCTION

More evidence for patterns of healthcare utilisation and associated costs among people receiving opioid agonist therapy (OAT) is needed. We investigated primary and secondary healthcare usage and costs among methadone and buprenorphine recipients in England.

METHODS

We conducted a cohort study using the Clinical Practice Research Datalink GOLD and Aurum databases of patients who were prescribed OAT between 1 January 2007 and 31 July 2019. The cohort was linked to Hospital Episode Statistics admitted patient care, outpatient and emergency department data, neighbourhood- and practice-level Index of Multiple Deprivation quintiles and mortality records. Negative binomial regression models were applied to estimate weighted rate ratios (wRR) of healthcare utilisation. Total and mean costs were calculated using Unit Costs of Health and Social Care and the National Healthcare Service Payment by Results National Tariffs.

RESULTS

Among 12,639 patients observed over 39,016 person-years, we found higher rate of hospital admissions (wRR 1.18; 1.08-1.28) among methadone compared with buprenorphine recipients. The commonest hospital discharge diagnoses among methadone patients were infectious diseases (19.2%), mental and behavioural disorders (17.0%) and drug-related poisoning (16.5%); the three commonest among buprenorphine patients were mental and behavioural diseases (21.5%), endocrine (13.8%) and genitourinary system diseases (13.1%). Methadone patients had similar mean costs compared with buprenorphine patients (cost difference: £539.01; 432.11-1006.69).

DISCUSSION AND CONCLUSIONS

Differences in healthcare utilisation frequency for methadone versus buprenorphine recipients were observed. The differences in associated costs were mainly driven by hospital admissions. These findings offer valuable insights for optimising care strategies and resource allocation for OAT recipients.

Assessing the feasibility of anticholinergic burden scales and measures in administrative data: A systematic review

AIM

This systematic review aimed to identify and evaluate the quality and adaptability of existing anticholinergic burden scales and measures by using administrative dispensing data.

METHOD

A comprehensive literature search was conducted using the Medline, Embase, CINAHL, and Google Scholar databases from 2001 to 2022. Studies that introduced, updated, or modified anticholinergic burden scales and measures were included in this review. Quality assessment considered various aspects, including scoring systems, tool development criteria, and specific requirements tailored for administrative data.

RESULTS

Twenty-eight anticholinergic burden scales and measures were identified in 14 countries. The Modified Anticholinergic Risk Scale excelled in the scoring system, while the German Anticholinergic Burden Scale stood out in the scale development process. However, significant variability was observed in methodologies, medication listings, and adaptability to administrative data. Quality assessment considers aspects such as potency, dose, exposure duration, longitudinal measurement, clinical interpretation, and compatibility with administrative data variables. The evaluation also considered tool development criteria including evidence for medication selection, panel expertise, relevance, updating methods, international applicability, validation, and clinical guidance.

CONCLUSION

This review emphasizes the importance of adaptable and robust tools that can work well with administrative data to ensure patient safety and better health outcomes, given the ongoing evolution of anticholinergic medications. The findings of this systematic review provide valuable insights for clinicians and researchers in selecting the most appropriate anticholinergic burden scale or measure according to their specific needs and data sources. This systematic review was registered with PROSPERO (registration ID CRD42023423959).

Clinical Outcomes in Patients with CKD and Rapid or Non-rapid eGFR Decline: A Report from the DISCOVER CKD Retrospective Cohort

INTRODUCTION

This analysis examined the baseline characteristics and clinical outcomes of patients with chronic kidney disease (CKD) and rapid or non-rapid estimated glomerular filtration rate (eGFR) decline, using retrospective data from DISCOVER CKD (ClinicalTrials.gov, NCT04034992).

METHODS

Data (2008-2020) were extracted from UK Clinical Practice Research Datalink, US TriNetX, US Limited Claims and Electronic Health Record Dataset, and Japan Medical Data Vision. Patients with CKD (two consecutive eGFR measures < 75 mL/min/1.73 m2 recorded 90-730 days apart) were included. Rapid eGFR decline was defined as an annual decline of > 4 mL/min/1.73 m2 at 2 years post-index; non-rapid eGFR decline was defined as an annual decline of ≤ 4 mL/min/1.73 m2. Clinical outcomes assessed included all-cause mortality, kidney outcomes (composite risk of kidney failure [progression to CKD stage 5] or > 50% eGFR decline, and kidney failure alone), cardiovascular events-including major adverse cardiovascular events (MACE; non-fatal myocardial infarction/stroke and cardiovascular death)-and all-cause hospitalization.

RESULTS

Across databases, rapid eGFR decline occurred in 13.7% of 804,237 eligible patients. Mean annual eGFR decline ranged between - 6.21 and - 6.86 mL/min/1.73 m2 in patients with rapid eGFR decline versus between - 0.11 and - 0.77 mL/min/1.73 m2 in patients with non-rapid eGFR decline. Rapid eGFR decline was associated with increased comorbidity burden and medication prescriptions. Across databases, the composite risk of kidney failure or > 50% decline in eGFR was significantly greater in patients with rapid versus non-rapid eGFR decline (P < 0.01); all-cause mortality, kidney failure alone, MACE, and all-cause hospitalization each significantly increased in two databases (P < 0.01-0.05).

CONCLUSION

Understanding patient factors associated with rapid eGFR decline in patients with CKD may help identify individuals who would benefit from proactive management to minimize the risk of adverse outcomes.

TRIAL REGISTRATION

ClinicalTrials.gov identifier, NCT04034992.

Improving Diagnosis and Care for Patients With Sarcoma: Do Real-World General Practitioners Data and Prospective Data Collections Have a Place Next to Clinical Trials?

There has been growing interest in the use of real-world data (RWD) to address clinically and policy-relevant (research) questions that cannot be answered with data from randomized controlled trials (RCTs) alone. This is, for example, the case in rare malignancies such as sarcomas as limited patient numbers pose challenges in conducting RCTs within feasible timeliness, a manageable number of collaborators, and statistical power. This narrative review explores the potential of RWD to generate real-world evidence (RWE) in sarcoma research, elucidating its application across different phases of the patient journey, from prediagnosis to the follow-up/survivorship phase. For instance, examining electronic health records (EHRs) from general practitioners (GPs) enables the exploration of consultation frequency and presenting symptoms in primary care before a sarcoma diagnosis. In addition, alternative study designs that integrate RWD with well-designed observational RCTs may offer relevant information on the effectiveness of clinical treatments. As, especially in cases of ultrarare sarcomas, it can be an extreme challenge to perform well-powered randomized prospective studies. Therefore, it is crucial to support the adaptation of novel study designs. Regarding the follow-up/survivorship phase, examining EHR from primary and secondary care can provide valuable insights into identifying the short- and long-term effects of treatment over an extended follow-up period. The utilization of RWD also comes with several challenges, including issues related to data quality and privacy, as described in this study. Notwithstanding these challenges, this study underscores the potential of RWD to bridge, at least partially, gaps between evidence and practice and holds promise in contributing to the improvement of sarcoma care.

Multiple adverse outcomes associated with antipsychotic use in people with dementia: population based matched cohort study

OBJECTIVE

To investigate risks of multiple adverse outcomes associated with use of antipsychotics in people with dementia.

DESIGN

Population based matched cohort study.

SETTING

Linked primary care, hospital and mortality data from Clinical Practice Research Datalink (CPRD), England.

POPULATION

Adults (≥50 years) with a diagnosis of dementia between 1 January 1998 and 31 May 2018 (n=173 910, 63.0% women). Each new antipsychotic user (n=35 339, 62.5% women) was matched with up to 15 non-users using incidence density sampling.

MAIN OUTCOME MEASURES

The main outcomes were stroke, venous thromboembolism, myocardial infarction, heart failure, ventricular arrhythmia, fracture, pneumonia, and acute kidney injury, stratified by periods of antipsychotic use, with absolute risks calculated using cumulative incidence in antipsychotic users versus matched comparators. An unrelated (negative control) outcome of appendicitis and cholecystitis combined was also investigated to detect potential unmeasured confounding.

RESULTS

Compared with non-use, any antipsychotic use was associated with increased risks of all outcomes, except ventricular arrhythmia. Current use (90 days after a prescription) was associated with elevated risks of pneumonia (hazard ratio 2.19, 95% confidence interval (CI) 2.10 to 2.28), acute kidney injury (1.72, 1.61 to 1.84), venous thromboembolism (1.62, 1.46 to 1.80), stroke (1.61, 1.52 to 1.71), fracture (1.43, 1.35 to 1.52), myocardial infarction (1.28, 1.15 to 1.42), and heart failure (1.27, 1.18 to 1.37). No increased risks were observed for the negative control outcome (appendicitis and cholecystitis). In the 90 days after drug initiation, the cumulative incidence of pneumonia among antipsychotic users was 4.48% (4.26% to 4.71%) versus 1.49% (1.45% to 1.53%) in the matched cohort of non-users (difference 2.99%, 95% CI 2.77% to 3.22%).

CONCLUSIONS

Antipsychotic use compared with non-use in adults with dementia was associated with increased risks of stroke, venous thromboembolism, myocardial infarction, heart failure, fracture, pneumonia, and acute kidney injury, but not ventricular arrhythmia. The range of adverse outcomes was wider than previously highlighted in regulatory alerts, with the highest risks soon after initiation of treatment.

Non-fatal overdose risk associated with prescribing opioid agonists concurrently with other medication: Cohort study conducted using linked primary care, secondary care and mortality records

BACKGROUND AND AIMS

An apparently protective effect of opioid agonist treatment (OAT) on all-cause and cause-specific mortality risk has been widely reported. Non-fatal overdose (NFO) often precedes subsequent drug-poisoning deaths. We hypothesized that benzodiazepines, gabapentinoids, antipsychotics, antidepressants, Z-drugs or opioids increase the NFO risk when co-prescribed with OAT.

DESIGN

We conducted a cohort study using the Clinical Practice Research Datalink GOLD and Aurum databases. The cohort was linked to Hospital Episode Statistics admitted patient care data (HES-APC), neighbourhood- and practice-level Index of Multiple Deprivation quintiles and mortality records from the Office for National Statistics.

SETTING

Primary care in England.

PARTICIPANTS

We studied patients with opioid use disorder, aged 18-64 years, who were prescribed OAT (15155 methadone and 5743 buprenorphine recipients) between Jan 1, 1998, and Dec 31, 2017.

MEASUREMENTS

The main outcome examined was NFO risk during co-prescription of OAT with benzodiazepines, antipsychotics, gabapentinoids, antidepressants, Z-drugs or opioids. Overdose was defined according to International Classification of Diseases codes from the HES-APC data set. Negative binomial regression models were used to estimate weighted rate ratios (wRR) for NFO during co-prescription of OAT and benzodiazepines, antipsychotics, gabapentinoids, antidepressants, Z-drugs or opioids with periods of exclusive OAT usage.

FINDINGS

Among 20 898 patients observed over 83 856 person-years, we found an elevated overdose risk that resulted in hospital admission during co-prescription of OAT with benzodiazepines [wRR: 1.45; 95% confidence interval (CI) = 1.26-1.67], gabapentinoids (wRR = 2.22; 95% CI = 1.77-2.79), Z-drugs (wRR = 1.60; 95% CI = 1.31-1.96), antipsychotics (wRR = 1.85; 95% CI = 1.53-2.25) and opioids (wRR = 1.28; 95% CI = 1.02-1.60). The risk ratio for antidepressant co-prescriptions was below unity (wRR = 0.90; 95% CI = 0.79-1.02) but this result was not statistically significant.

CONCLUSION

Elevated risk of non-fatal overdose among opioid agonist treatment recipients is associated with concurrent use of medication prescribed for other reasons.

Pharmacoepidemiology: An Overview

The aims of this review are to provide a comprehensive overview of the definition and scope of pharmacoepidemiology, to summarize the study designs and methodologies used in the field, to discuss the future trends in the field and new methodologies to address bias and confounding, and finally to give some recommendations to clinicians interested in pharmacoepidemiologic research. Because drug efficacy and safety from randomized clinical trials do not reflect the real-world situation, pharmacoepidemiological studies on drug safety monitoring and drug effectiveness in large numbers of people are needed by healthcare professionals and regulatory institutions. We aim to highlight the importance of pharmacoepidemiologic research in informing evidence-based medicine and public health policy. The development of new designs and methodologies for the generation of valid evidence, as well as new initiatives to provide guidance and recommendations on how to incorporate real-world evidence into the drug development process, are reported on. In addition, we have touched on the implication of artificial intelligence in the management of real-world data. This overview aims to summarize all important aspects to consider when conducting or interpreting a pharmacoepidemiologic study.

Association of comorbid-socioeconomic clusters with mortality in late onset epilepsy derived through unsupervised machine learning

BACKGROUND AND OBJECTIVES

Late-onset epilepsy is a heterogenous entity associated with specific aetiologies and an elevated risk of premature mortality. Specific multimorbid-socioeconomic profiles and their unique prognostic trajectories have not been described. We sought to determine if specific clusters of late onset epilepsy exist, and whether they have unique hazards of premature mortality.

METHODS

We performed a retrospective observational cohort study linking primary and hospital-based UK electronic health records with vital statistics data (covering years 1998-2019) to identify all cases of incident late onset epilepsy (from people aged ≥65) and 1:10 age, sex, and GP practice-matched controls. We applied hierarchical agglomerative clustering using common aetiologies identified at baseline to define multimorbid-socioeconomic profiles, compare hazards of early mortality, and tabulating causes of death stratified by cluster.

RESULTS

From 1,032,129 people aged ≥65, we identified 1048 cases of late onset epilepsy who were matched to 10,259 controls. Median age at epilepsy diagnosis was 68 (interquartile range: 66-72) and 474 (45%) were female. The hazard of premature mortality related to late-onset epilepsy was higher than matched controls (hazard ratio [HR] 1.73; 95% confidence interval [95%CI] 1.51-1.99). Ten unique phenotypic clusters were identified, defined by 'healthy' males and females, ischaemic stroke, intracerebral haemorrhage (ICH), ICH and alcohol misuse, dementia and anxiety, anxiety, depression in males and females, and brain tumours. Cluster-specific hazards were often similar to that derived for late-onset epilepsy as a whole. Clusters that differed significantly from the base late-onset epilepsy hazard were 'dementia and anxiety' (HR 5.36; 95%CI 3.31-8.68), 'brain tumour' (HR 4.97; 95%CI 2.89-8.56), 'ICH and alcohol misuse' (HR 2.91; 95%CI 1.76-4.81), and 'ischaemic stroke' (HR 2.83; 95%CI 1.83-4.04). These cluster-specific risks were also elevated compared to those derived for tumours, dementia, ischaemic stroke, and ICH in the whole population. Seizure-related cause of death was uncommon and restricted to the ICH, ICH and alcohol misuse, and healthy female clusters.

SIGNIFICANCE

Late-onset epilepsy is an amalgam of unique phenotypic clusters that can be quantitatively defined. Late-onset epilepsy and cluster-specific comorbid profiles have complex effects on premature mortality above and beyond the base rates attributed to epilepsy and cluster-defining comorbidities alone.

Independent Associations of Incident Epilepsy and Enzyme-Inducing and Non-Enzyme-Inducing Antiseizure Medications With the Development of Osteoporosis

Importance

Both epilepsy and enzyme-inducing antiseizure medications (eiASMs) having varying reports of an association with increased risks for osteoporosis.

Objective

To quantify and model the independent hazards for osteoporosis associated with incident epilepsy and eiASMS and non-eiASMs.

Design, Setting, and Participants

This open cohort study covered the years 1998 to 2019, with a median (IQR) follow-up of 5 (1.7-11.1) years. Data were collected for 6275 patients enrolled in the Clinical Practice Research Datalink and from hospital electronic health records. No patients who met inclusion criteria (Clinical Practice Research Datalink-acceptable data, aged 18 years or older, follow-up after the Hospital Episode Statistics patient care linkage date of 1998, and free of osteoporosis at baseline) were excluded or declined.

Exposure

Incident adult-onset epilepsy using a 5-year washout and receipt of 4 consecutive ASMs.

Main Outcomes and Measures

The outcome was incident osteoporosis as determined through Cox proportional hazards or accelerated failure time models where appropriate. Incident epilepsy was treated as a time-varying covariate. Analyses controlled for age, sex, socioeconomic status, cancer, 1 or more years of corticosteroid use, body mass index, bariatric surgery, eating disorders, hyperthyroidism, inflammatory bowel disease, rheumatoid arthritis, smoking status, falls, fragility fractures, and osteoporosis screening tests. Subsequent analyses (1) excluded body mass index, which was missing in 30% of patients; (2) applied propensity score matching for receipt of an eiASM; (3) restricted analyses to only those with incident onset epilepsy; and (4) restricted analyses to patients who developed epilepsy at age 65 years or older. Analyses were performed between July 1 and October 31, 2022, and in February 2023 for revisions.

Results

Of 8 095 441 adults identified, 6275 had incident adult-onset epilepsy (3220 female [51%] and 3055 male [49%]; incidence rate, 62 per 100 000 person-years) with a median (IQR) age of 56 (38-73) years. When controlling for osteoporosis risk factors, incident epilepsy was independently associated with a 41% faster time to incident osteoporosis (time ratio [TR], 0.59; 95% CI, 0.52-0.67; P < .001). Both eiASMs (TR, 0.91; 95% CI, 0.87-0.95; P < .001) and non-eiASMs (TR, 0.77; 95% CI, 0.76-0.78; P < .001) were also associated with significant increased risks independent of epilepsy, accounting for 9% and 23% faster times to development of osteoporosis, respectively. The independent associations among epilepsy, eiASMs, and non-eiASMs remained consistent in propensity score-matched analyses, cohorts restricted to adult-onset epilepsy, and cohorts restricted to late-onset epilepsy.

Conclusions and Relevance

These findings suggest that epilepsy is independently associated with a clinically meaningful increase in the risk for osteoporosis, as are both eiASMs and non-eiASMs. Routine screening and prophylaxis should be considered in all people with epilepsy.

Pharmacovigilance: reporting requirements throughout a product's lifecycle

Comprehensive methods for evaluating safety are needed to objectively assess the full risk profile of a medication. The confidence of the prescribing provider in the safety and effectiveness of pharmaceuticals is extremely important. Pharmacovigilance is a key component of drug safety regulatory processes and is paramount for ensuring the safety profile of medications used to treat patients. All participants in the healthcare system, including healthcare providers and consumers, should understand and meaningfully engage in the pharmacovigilance process; healthcare providers should integrate pharmacovigilance into everyday practice, inviting feedback from patients. This narrative review aims to give an overview of the main topics underlying pharmacovigilance and drug safety in pharmaceutical research phase after the authorization of a drug in the United States. The US Food and Drug Administration guidance and post-approval regulatory actions are considered from an industry perspective.

Plain language summary

Regulatory processes that ensure the safety of drugs is monitored Government agencies regulate the safe use of medicinal products. By determining and enforcing pharmacovigilance, the monitoring of drugs for potential risks, they safeguard the welfare of consumers of medicines. Comprehensive, documented methods for evaluating the safety of a drug during its development and its subsequent use allow identification of any risks associated with the drug's use throughout its lifetime. The comprehensive identification of safety issues associated with a drug is improved when all parties involved in the development and use of drugs participate in the pharmacovigilance process. For example, clinicians should regularly ask their patients if they are experiencing any issues with their treatment, and patients should be encouraged to report problems they encounter with a particular medication to their healthcare provider. This narrative review provides an overview of the main topics underlying pharmacovigilance and drug safety after approval of a drug in the United States. Guidelines and actions from the US Food and Drug Administration are considered from an industry perspective.

Prevalence of anabolic steroid users seeking support from physicians: a systematic review and meta-analysis

OBJECTIVES

To estimate the overall prevalence of androgenic-anabolic steroids (AAS) users seeking support from physicians. Secondary objectives are to compare this prevalence in different locations and among subpopulations of AAS users, and to discuss some of the factors that could have influenced the engagement of AAS users with physicians.

DESIGN

Systematic review and meta-analysis.

DATA SOURCES

MEDLINE, PsycINFO, Web of Science and SciELO were searched in January 2022.

ELIGIBILITY CRITERIA

Quantitative and qualitative studies reporting the number of AAS users who sought support from physicians, with no restrictions of language or time of publication.

DATA EXTRACTION AND SYNTHESIS

Two independent reviewers extracted data and assessed the quality of studies, including publication bias. A random-effects meta-analysis was performed to estimate the overall prevalence of AAS users seeking support from physicians, followed by pooled prevalence rates by studies' location and the subpopulation of AAS users.

RESULTS

We identified 36 studies published between 1988 and 2021, involving 10 101 AAS users. The estimated overall prevalence of AAS users seeking support from physicians is 37.12% (95% CI 29.71% to 44.52%). Higher prevalence rates were observed in studies from Australia (67.27%; 95% CI 42.29% to 87.25%) and among clients of the needle and syringe exchange programme (54.13%; 95% CI 36.41% to 71.84%). The lowest prevalence was observed among adolescent AAS users (17.27%; 95% CI 4.80% to 29.74%).

CONCLUSION

Our findings suggest that about one-third of AAS users seek support from physicians, with remarkable differences between locations and subpopulations of AAS users. Further studies should investigate the factors influencing the engagement of AAS users with physicians.

PROSPERO REGISTRATION NUMBER

CRD42020177919.

Using primary care databases for addiction research: An introduction and overview of strengths and weaknesses

Primary care databases extract and combine routine data from the electronic patient records of various participating practices on a regular basis. These databases can be used for innovative and relevant addiction research, but such use requires a thorough understanding of how data were originally collected and how they need to be processed and statistically analysed to produce sound scientific evidence. The aims of this paper are therefore to (1) make a case for why primary care databases should be considered more frequently for addiction research; (2) provide an overview of how primary care databases are constructed; (3) highlight important methodological and statistical strengths and weaknesses of using primary care databases for research; and (4) give practical advice about how a researcher can get access to databases. Three major primary care databases from the UK serve as examples: Clinical Practice Research Datalink (CPRD), The Health Improvement Network (THIN), and QResearch.

Protocol for a Delphi consensus study to select indicators of high-quality general practice to achieve Quality Equity and Systems Transformation in Primary Health Care (QUEST-PHC) in Australia

Background

High-quality general practice has been demonstrated to provide cost-effective, equitable health care and improve health outcomes. Yet there is currently not a set of agreed comprehensive indicators in Australia. We have developed 79 evidence-based indicators and their corresponding 129 measures of high-quality general practice. This study aims to achieve consensus on relevant and feasible indicators and measures for the Australian context.

Methods

This Delphi consensus study, approved by WSU Human Research Ethics Committee, consists of three rounds of online survey with general practice experts including general practitioners, practice nurses and primary health network staff. The identified indicators and measures are grouped under an attribute framework aligned with the Quadruple Aim, and further grouped under structures, processes and outcomes according to the Donabedian framework. Participants will rate each indicator and measure for relevance and feasibility, and provide comments and recommendations of additional indicators or measures. In the last round, participants will also be asked their views on the implementation of a quality indicator tool. Each indicator and measure will require ≥70% agreement in both relevance and feasibility to achieve consensus. Aggregated ratings will be statistically analysed for response rates, level of agreement, medians, interquartile ranges and group rankings. Qualitative responses will be analysed thematically using a mixed inductive and deductive approach.

Discussion

This protocol will add to the current knowledge of the translation of performance guidelines into quality practice across complex clinical settings and in a variety of different contexts in Australian general practice. The Delphi technique is appropriate to develop consensus between the diverse experts because of its ability to offer anonymity to other participants and minimise bias. Findings will contribute to the design of an assessment tool of high-quality general practice that would enable future primary health care reforms in Australia.

The Association between Nursing Skill Mix and Patient Outcomes in a Mental Health Setting: Protocol for an Observational Feasibility Study

International research on nursing skill mix has focused primarily on medical and surgical patient populations. The association between nursing skill mix and clinical outcomes for psychiatric inpatients has not been explored. The aim of this study is to establish the feasibility of extracting and linking nurse and inpatient data from hospital administrative data sources. This is an observational study. Data will be extracted from hospital administrative sources and linked together. Patient information will include duration and number of psychiatric hospital admissions. We will extract information on the educational preparation of nurses working in the participating hospitals to enable us to calculate estimates of the nursing skill mix. The study will be conducted in two psychiatric inpatient services in Australia. Our study will test the feasibility of extracting and linking nursing skill mix and patient data in a mental health setting and will inform the methodological development of an appropriately powered observational study. Australian and New Zealand clinical trial registry: ACTRN12619001337167p.

Characterization of the deterioration of diabetes control in patients with a subsequent diagnosis of pancreatic cancer: A descriptive study

BACKGROUND

Deterioration of diabetes control can be the first harbinger of pancreatic cancer. However, little is known about how to distinguish patients with pancreatic cancer-related diabetes deterioration from those with type 2 diabetes progression. We aimed to characterize the glycated hemoglobin (HbA1c) and body weight profile of pancreatic cancer patients with deteriorating diabetes before the cancer diagnosis.

METHODS

Using data from the UK-based Clinical Practice Research Datalink (CPRD) GOLD, we established a study population including pancreatic cancer patients with diabetes deterioration in the >0.5-3 years before the cancer diagnosis and non-cancer patients with deterioration of type 2 diabetes (comparison group). Patients were considered to have diabetes deterioration if their glucose-lowering treatment was intensified. We characterized the longitudinal trajectories of HbA1c and body weight in pancreatic cancer patients compared with non-cancer patients before and after treatment intensification.

RESULTS

The mean absolute increase in HbA1c from the pre-deterioration period, i.e. the time >1-2 years before treatment intensification, to the time of treatment intensification, was 1.5% ± 1.6% in pancreatic cancer patients vs. 0.9% ± 1.4% in non-cancer patients. After treatment intensification, mean HbA1c remained elevated in pancreatic cancer patients, while it returned to the pre-deterioration level in non-cancer patients. Body weight decreased by 1.9% ± 6.4% in cancer patients and increased by 0.3% ± 5.2% in non-cancer patients between the pre-deterioration period and treatment intensification, on average.

CONCLUSIONS

Pancreatic cancer-related diabetes deterioration may frequently be characterized by pronounced increases in HbA1c, persistent elevation of HbA1c after treatment intensification, and concomitant weight loss.

Association of Enzyme-Inducing Antiseizure Drug Use With Long-term Cardiovascular Disease

Importance

Enzyme-inducing antiseizure medications (eiASMs) have been hypothesized to be associated with long-term risks of cardiovascular disease.

Objective

To quantify and model the putative hazard of cardiovascular disease secondary to eiASM use.

Design, Setting, and Participants

This cohort study covered January 1990 to March 2019 (median [IQR] follow-up, 9 [4-15], years). The study linked primary care and hospital electronic health records at National Health Service hospitals in England. People aged 18 years or older diagnosed as having epilepsy after January 1, 1990, were included. All eligible patients were included with a waiver of consent. No patients were approached who withdrew consent. Analysis began January 2021 and ended August 2021.

Exposures

Receipt of 4 consecutive eiASMs (carbamazepine, eslicarbazepine, oxcarbazepine, phenobarbital, phenytoin, primidone, rufinamide, or topiramate) following an adult-onset (age ≥18 years) epilepsy diagnosis or repeated exposure in a weighted cumulative exposure model.

Main Outcomes and Measures

Three cohorts were isolated, 1 of which comprised all adults meeting a case definition for epilepsy diagnosed after 1990, 1 comprised incident cases diagnosed after 1998 (hospital linkage date), and 1 was limited to adults diagnosed with epilepsy at 65 years or older. Outcome was incident cardiovascular disease (ischemic heart disease or ischemic or hemorrhagic stroke). Hazard of incident cardiovascular disease was evaluated using adjusted propensity-matched survival analyses and weighted cumulative exposure models.

Results

Of 10 916 166 adults, 50 888 (0.6%) were identified as having period-prevalent cases (median [IQR] age, 32 [19-50] years; 16 584 [53%] female), of whom 31 479 (62%) were diagnosed on or after 1990 and were free of cardiovascular disease at baseline. In a propensity-matched Cox proportional hazards model adjusted for age, sex, baseline socioeconomic status, and cardiovascular risk factors, the hazard ratio for incident cardiovascular disease was 1.21 (95% CI, 1.06-1.39) for those receiving eiASMs. The absolute difference in cumulative hazard diverges by more than 1% and greater after 10 years. For those with persistent exposure beyond 4 prescriptions, the median hazard ratio increased from amedian (IQR) of 1.54 (1.28-1.79) when taking a relative defined daily dose of an eiASM of 1 to 2.38 (1.52-3.56) with a relative defined daily dose of 2 throughout a maximum of 25 years' follow-up compared with those not receiving an eiASM. The hazard was elevated but attenuated when restricting analyses to incident cases or those diagnosed when older than 65 years.

Conclusions and Relevance

The hazard of incident cardiovascular disease is higher in those receiving eiASMs. The association is dose dependent and the absolute difference in hazard seems to reach clinical significance by approximately 10 years from first exposure.

Treatment patterns in pediatric patients with multiple sclerosis in Germany-a nationwide claim-based analysis

BACKGROUND

The manifestation of multiple sclerosis (MS) in childhood and adolescence occurs in 3%-5% of all MS cases. However, the immunomodulatory and symptomatic treatment options in this population group are still limited.

OBJECTIVE

We aimed to elucidate the prescription frequency of medications used in pediatric patients with multiple sclerosis (PwMS) compared with the general population, considering the entire spectrum of medications prescribed.

METHODS

Based on nationwide outpatient drug prescription data and statutory health insurance (SHI) physicians' claims data from 2018, we conducted a population-based cross-sectional study in Germany. Children and adolescents aged ⩽17 years (n = 11,381,939) diagnosed with MS (n = 613), and a matched (age, sex, and health insurance sector) control group (n = 6130) were included. The prescription prevalence was measured as the proportion of MS patients with ⩾1 prescription.

RESULTS

Of the 613 pediatric PwMS with a median age of 16 years, 403 (65.7%) were female. For 15 out of the 18 different active agents analyzed, PwMS had a significantly higher prescription prevalence than the control group (Fisher's exact test: p ⩽ 0.037). The most frequently prescribed drugs in PwMS were ibuprofen (28.4%; anti-inflammatory drug), cholecalciferol (23.0%; vitamin D3), and interferon beta-1a (21.5%; disease-modifying drug, DMD). The proportions of DMD prescriptions and antibiotic prescriptions were higher among PwMS aged 15-17 years than among those ⩽14 years (DMD: 43.4% vs 34.2%, p = 0.05; antibiotic: 34.1% vs 24.8%, p = 0.031). In contrast, younger PwMS were more likely to receive a prescription for anti-inflammatory/anti-rheumatic drugs (36.6% vs 26.5%, p = 0.02).

CONCLUSION

Our study analyzing real-world medication data showed that interferon beta, anti-inflammatory drugs, and vitamins play an essential role in the treatment of pediatric PwMS. Future research should evaluate longitudinal treatment patterns of pediatric PwMS, paying particular attention to the time of diagnosis, time of first DMD initiation, and therapy switches.

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.

The Risk of Muscular Events Among New Users of Hydrophilic and Lipophilic Statins: an Observational Cohort Study

BACKGROUND

Statins are effective lipid-lowering drugs for the prevention of cardiovascular disease, but muscular adverse events can limit their use. Hydrophilic statins (pravastatin, rosuvastatin) may cause less muscular events than lipophilic statins (e.g. simvastatin, atorvastatin) due to lower passive diffusion into muscle cells.

OBJECTIVE

To compare the risk of muscular events between statins at comparable lipid-lowering doses and to evaluate if hydrophilic statins are associated with a lower muscular risk than lipophilic statins.

DESIGN/SETTING

Propensity score-matched cohort study using data from the United Kingdom-based Clinical Practice Research Datalink (CPRD) GOLD.

PATIENTS

New statin users. Cohort 1: pravastatin 20-40 mg (hydrophilic) vs simvastatin 10-20 mg (lipophilic), cohort 2: rosuvastatin 5-40 mg (hydrophilic) vs atorvastatin 10-80 mg (lipophilic), and cohort 3: simvastatin 40-80 mg vs atorvastatin 10-20 mg.

MAIN MEASURES

The outcome was a first record of a muscular event (myopathy, myalgia, myositis, rhabdomyolysis) during a maximum follow-up of 1 year.

KEY RESULTS

The propensity score-matched cohorts consisted of 1) 9,703, 2) 7,032, and 3) 37,743 pairs of statin users. Comparing the risk of muscular events between low-intensity pravastatin vs low-intensity simvastatin yielded a HR of 0.86 (95% CI 0.64-1.16). In the comparison of moderate- to high-intensity rosuvastatin vs equivalent doses of atorvastatin, we observed a HR of 1.17 (95% CI 0.88-1.56). Moderate- to high-intensity simvastatin was associated with a HR of 1.33 (95% CI 1.16-1.53), when compared with atorvastatin at equivalent doses.

LIMITATIONS

We could not conduct other pairwise comparisons of statins due to small sample size. In the absence of a uniform definition on the comparability of statin doses, the applied dose ratios may not fully match with all literature sources.

CONCLUSIONS

Our results do not suggest a systematically lower risk of muscular events for hydrophilic statins when compared to lipophilic statins at comparable lipid-lowering doses.

Epidemiology of Distal Renal Tubular Acidosis: A Study Using Linked UK Primary Care and Hospital Data

INTRODUCTION

Distal renal tubular acidosis (dRTA), or RTA type 1, a rare inherited or acquired disease, is a disorder of the distal tubule caused by impaired urinary acid secretion. Due to associated conditions and nonspecific symptoms, dRTA may go undetected. This analysis aims to estimate the prevalence of dRTA in the UK Clinical Practice Research Datalink (CPRD) databases and extrapolate it to European Union Five (EU5) populations.

METHODS

A retrospective analysis was conducted using the CPRD GOLD database and linked Hospital Episode Statistics (HES) data to identify diagnosed and potentially undiagnosed or miscoded patients (suspected patients). Patients' records with at least one diagnosis code for dRTA, RTA, specific autoimmune diseases, or renal disorders recorded between January 1987 and November 2017 were obtained and analyzed. An algorithm was developed to detect potentially undiagnosed/miscoded dRTA, based on associated conditions and prescriptions.

RESULTS

A total of 216 patients with diagnosis of RTA or dRTA were identified (with 98 linked to hospital data), and 447 patients were identified as having suspected dRTA. dRTA prevalence for 2017 was estimated between 0.46 (recorded cases, of which 22.1% were considered primary) and 1.60 when including suspected cases (7.6% primary) per 10,000 people. Prescription and clinical records of diagnosed patients revealed a wide range of comorbidities and a need for pharmacological treatment to manage associated symptoms.

CONCLUSION

The study provides new estimates of dRTA prevalence in Europe and suggests that patients may often be unreported or miscoded, potentially confounding appropriate disease management.

Vaccine Safety Datalink infrastructure enhancements for evaluating the safety of maternal vaccination

BACKGROUND

Identifying pregnancy episodes and accurately estimating their beginning and end dates are imperative for observational maternal vaccine safety studies using electronic health record (EHR) data.

METHODS

We modified the Vaccine Safety Datalink (VSD) Pregnancy Episode Algorithm (PEA) to include both the International Classification of Disease, ninth revision (ICD-9 system) and ICD-10 diagnosis codes, incorporated additional gestational age data, and validated this enhanced algorithm with manual medical record review. We also developed the new Dynamic Pregnancy Algorithm (DPA) to identify pregnancy episodes in real time.

RESULTS

Around 75% of the pregnancy episodes identified by the enhanced VSD PEA were live births, 12% were spontaneous abortions (SABs), 10% were induced abortions (IABs), and 0.4% were stillbirths (SBs). Gestational age was identified for 99% of live births, 89% of SBs, 69% of SABs, and 42% of IABs. Agreement between the PEA-assigned and abstractor-identified pregnancy outcome and outcome date was 100% for live births, but was lower for pregnancy losses. When gestational age was available in the medical record, the agreement was higher for live births (97%), but lower for pregnancy losses (75%). The DPA demonstrated strong concordance with the PEA and identified pregnancy episodes ⩾6 months prior to the outcome date for 89% of live births.

CONCLUSION

The enhanced VSD PEA is a useful tool for identifying pregnancy episodes in EHR databases. The DPA improves the timeliness of pregnancy identification and can be used for near real-time maternal vaccine safety studies.

PLAIN LANGUAGE SUMMARY

Improving identification of pregnancies in the Vaccine Safety Datalink electronic medical record databases to allow for better and faster monitoring of vaccination safety during pregnancy Introduction: It is important to monitor of the safety of vaccines after they have been approved and licensed by the Food and Drug Administration, especially among women vaccinated during pregnancy. The Vaccine Safety Datalink (VSD) monitors vaccine safety through observational studies within large databases of electronic medical records. Since 2012, VSD researchers have used an algorithm called the Pregnancy Episode Algorithm (PEA) to identify the medical records of women who have been pregnant. Researchers then use these medical records to study whether receiving a particular vaccine is linked to any negative outcomes for the woman or her child.Methods: The goal of this study was to update and enhance the PEA to include the full set of medical record diagnostic codes [both from the older International Classification of Disease, ninth revision (ICD-9 system) and the newer ICD-10 system] and to incorporate additional sources of data about gestational age. To ensure the validity of the PEA following these enhancements, we manually reviewed medical records and compared the results with the algorithm. We also developed a new algorithm, the Dynamic Pregnancy Algorithm (DPA), to identify women earlier in pregnancy, allowing us to conduct more timely vaccine safety assessments.Results: The new version of the PEA identified 2,485,410 pregnancies in the VSD database. The enhanced algorithm more precisely estimated the beginning of pregnancies, especially those that did not result in live births, due to the new sources of gestational age data.Conclusion: Our new algorithm, the DPA, was successful at identifying pregnancies earlier in gestation than the PEA. The enhanced PEA and the new DPA will allow us to better evaluate the safety of current and future vaccinations administered during or around the time of pregnancy.

Framework for the public health assessment of electronic cigarettes

BACKGROUND

Electronic cigarettes (e-cigarettes) are relatively new products with substantial public health impacts. Evidence on their effects is diverse and emerging rapidly, presenting challenges to high-quality policymaking and decision-making. This paper addresses these challenges by developing and presenting a framework for the public health assessment of e-cigarettes, using the Australian context as an example.

METHODS

Framework development involved stakeholder engagement, development of guiding principles, and consideration of existing relevant frameworks and the evidence requirements of current policy options, identified in published and grey literature.

RESULTS

Guiding principles include the need for the framework to: be evidence based; include consideration of the likely balance of benefits and risks of e-cigarettes, uncertainty and safety; support equity; support the ongoing application of evidence to high-quality policy and practice; and consider potential competing interests. The framework draws upon: health technology assessment; health impact assessment; environmental health risk assessment; healthcare recommendations evidence evaluation; consumer goods regulation; medicine and chemical scheduling; tobacco product evaluation; previous reviews and the precautionary principle. Final framework components are: (1) characterisation of products under consideration; (2) definition of populations of interest; (3) characterisation of tobacco smoking, control and impacts on health and well-being; (4) review of evidence on patterns of e-cigarette use; (5) review of evidence on e-cigarette use and health outcomes; (6) assessment of likely risks, benefits and safety; (7) identification and assessment of policy options to optimise health outcomes.

CONCLUSIONS

Structured and ongoing public health assessment of e-cigarette use is likely to support health through enhancing evidence-based decision-making.

Temporal relationship between osteoarthritis and comorbidities: a combined case control and cohort study in the UK primary care setting

Objective

To determine the burden of comorbidities in OA and their temporal relationships in the UK.

Methods

The Clinical Practice Research Datalink (CPRD) GOLD was used to identify people with incident OA and age, gender and practice matched non-OA controls from UK primary care. Controls were assigned the same index date as matched cases (date of OA diagnosis). Associations between OA and 49 individual comorbidities and multimorbidities (two or more comorbidities excluding OA) both before and after OA diagnosis were estimated, adjusting for covariates, using odds ratios (aORs) and hazard ratios (aHRs), respectively.

Results

During 1997–2017, we identified 221 807 incident OA cases and 221 807 matched controls. Of 49 comorbidities examined, 38 were associated with OA both prior to and following the diagnosis of OA and 2 (dementia and systemic lupus erythematosus) were associated with OA only following the diagnosis of OA. People with OA had a higher risk of developing heart failure [aHR 1.63 (95% CI 1.56, 1.71)], dementia [aHR 1.62 (95% CI 1.56, 1.68)], liver diseases [aHR 1.51 (95% CI 1.37, 1.67)], irritable bowel syndrome [aHR 1.51 (95% CI 1.45, 1.58)], gastrointestinal bleeding [aHR 1.49 (95% CI 1.39, 1.59)], 10 musculoskeletal conditions and 25 other conditions following OA diagnosis. The aOR for multimorbidity prior to the index date was 1.71 (95% CI 1.69, 1.74), whereas the aHR for multimorbidity after the index date was 1.29 (95% CI 1.28, 1.30).

Conclusions

People with OA are more likely to have other chronic conditions both before and after the OA diagnosis. Further study on shared aetiology and causality of these associations is needed.

Can the Use of Bayesian Analysis Methods Correct for Incompleteness in Electronic Health Records Diagnosis Data? Development of a Novel Method Using Simulated and Real-Life Clinical Data

Background: Patient health information is collected routinely in electronic health records (EHRs) and used for research purposes, however, many health conditions are known to be under-diagnosed or under-recorded in EHRs. In research, missing diagnoses result in under-ascertainment of true cases, which attenuates estimated associations between variables and results in a bias toward the null. Bayesian approaches allow the specification of prior information to the model, such as the likely rates of missingness in the data. This paper describes a Bayesian analysis approach which aimed to reduce attenuation of associations in EHR studies focussed on conditions characterized by under-diagnosis.

Methods: Study 1: We created synthetic data, produced to mimic structured EHR data where diagnoses were under-recorded. We fitted logistic regression (LR) models with and without Bayesian priors representing rates of misclassification in the data. We examined the LR parameters estimated by models with and without priors. Study 2: We used EHR data from UK primary care in a case-control design with dementia as the outcome. We fitted LR models examining risk factors for dementia, with and without generic prior information on misclassification rates. We examined LR parameters estimated by models with and without the priors, and estimated classification accuracy using Area Under the Receiver Operating Characteristic.

Results: Study 1: In synthetic data, estimates of LR parameters were much closer to the true parameter values when Bayesian priors were added to the model; with no priors, parameters were substantially attenuated by under-diagnosis. Study 2: The Bayesian approach ran well on real life clinic data from UK primary care, with the addition of prior information increasing LR parameter values in all cases. In multivariate regression models, Bayesian methods showed no improvement in classification accuracy over traditional LR.

Conclusions: The Bayesian approach showed promise but had implementation challenges in real clinical data: prior information on rates of misclassification was difficult to find. Our simple model made a number of assumptions, such as diagnoses being missing at random. Further development is needed to integrate the method into studies using real-life EHR data. Our findings nevertheless highlight the importance of developing methods to address missing diagnoses in EHR data.

Identifying undetected dementia in UK primary care patients: a retrospective case-control study comparing machine-learning and standard epidemiological approaches

BACKGROUND

Identifying dementia early in time, using real world data, is a public health challenge. As only two-thirds of people with dementia now ultimately receive a formal diagnosis in United Kingdom health systems and many receive it late in the disease process, there is ample room for improvement. The policy of the UK government and National Health Service (NHS) is to increase rates of timely dementia diagnosis. We used data from general practice (GP) patient records to create a machine-learning model to identify patients who have or who are developing dementia, but are currently undetected as having the condition by the GP.

METHODS

We used electronic patient records from Clinical Practice Research Datalink (CPRD). Using a case-control design, we selected patients aged >65y with a diagnosis of dementia (cases) and matched them 1:1 by sex and age to patients with no evidence of dementia (controls). We developed a list of 70 clinical entities related to the onset of dementia and recorded in the 5 years before diagnosis. After creating binary features, we trialled machine learning classifiers to discriminate between cases and controls (logistic regression, naïve Bayes, support vector machines, random forest and neural networks). We examined the most important features contributing to discrimination.

RESULTS

The final analysis included data on 93,120 patients, with a median age of 82.6 years; 64.8% were female. The naïve Bayes model performed least well. The logistic regression, support vector machine, neural network and random forest performed very similarly with an AUROC of 0.74. The top features retained in the logistic regression model were disorientation and wandering, behaviour change, schizophrenia, self-neglect, and difficulty managing.

CONCLUSIONS

Our model could aid GPs or health service planners with the early detection of dementia. Future work could improve the model by exploring the longitudinal nature of patient data and modelling decline in function over time.