Completeness and validity of alcohol recording in general practice within the UK: a cross-sectional study

An evaluation of the impact of a national Minimum Unit Price on alcohol policy on alcohol behaviours

BACKGROUND

In 2018, Scotland pioneered national legislation which set a Minimum Unit Price (MUP) of 50 pence (∼US$0.64, €0.59) per unit of UK alcohol sold (8 g/10 ml). To inform policy development, we examine the policy effect using the Alcohol Use Disorders Identification Test (AUDIT-C), employing longitudinal data for over 17 200 individuals.

METHODS

The effect of MUP on AUDIT-C scores is inferred by employing difference-in-difference regression. Pre- and post-intervention alcohol behaviours of individuals from Scotland are compared to a matched 'control' from England. Drinking at hazardous and harmful levels could be identified, as well as the frequency of alcohol consumption, number of drinks and heavy episodic drinking. Estimates adjust for demographic, socioeconomic and health characteristics. Potential inequalities by gender, age and household income are examined.

RESULTS

MUP led to an estimated 5.3% reduction in the number of drinks consumed on drinking occasions, though a statistically significant effect on overall reported AUDIT-C scores or drinking at hazardous levels was not detected, with few differential effects for subgroups.

CONCLUSIONS

Differences in the findings of this research compared to other studies may be explained by differences in population coverage collected in the survey data, compared to more comprehensive, population-wide administrative data, as well as sample attrition.

Screening for Alcohol Use Disorder Among Hospitalised Patients: Learning from a Retrospective Cohort Study in Secondary Care

Background: Excessive alcohol consumption is among the leading causes of hospitalisation in high-income countries and contributes to over 200 medical conditions. We aimed to determine the prevalence and characteristics of alcohol use disorder (AUD), describe the distribution of AUD in ICD-10 discharge diagnosis groups and ascertain any relationship between them in secondary care. Methods: The study group was a retrospective cohort of adult patients admitted to Nottingham University Hospital (NUH) between 4 April 2009 and 31 March 2020. Uni- and multivariable analysis was performed to determine the relationship between AUD and covariable high-risk characteristics and describe the distribution of AUD in ICD-10 discharge diagnosis groups defined by an alcohol-attributable fraction. Results: A total of 44,804 patients (66,440 admissions) were included, with a mean age of 63.1 years (SD ± 19.9); of these, 48.0% (n = 20,863) were male and 71.2% were (n = 30,994) white. AUDIT-C was completed in 97.1% (n = 43,514) of patients, and identified 16.5% (n = 7164) as having AUD, while 2.1% (n = 900) were found to be alcohol-dependent. In patients with AUD, 4.0% (n = 283) had an ICD-10 diagnosis that was alcohol-specific and 17.5% (n = 1255) were diagnosed with alcohol-related disorders; the remainder were not diagnosed with either disorder. Two-thirds (64.7%) of the patients with AUD had associated mental and behavioural disorders. Multivariable logistic regression analysis revealed that patients aged 60-69 had the highest risk of AUD (OR 4.19, 95% CI 3.53-4.99). Being single (OR 1.18, 95% CI 1.11-1.26) and a history of emergency admission (OR 1.21, 95% CI 1.14-1.29) were associated with increased odds of AUD. Conversely, females compared to males (OR 0.34, 95% CI 0.35-0.39), individuals from minority ethnic backgrounds compared to white Caucasians (OR 0.39, 95% CI 0.35-0.45), and those from more deprived areas (IMD quintile 1: OR 0.79, 95% CI 0.74-0.86) had lower odds of AUD. Conclusions: One in six admitted patients had AUD, with a higher risk in males, ages 60-69, and emergency admissions. Mental disorders are highly prevalent among hospitalised patients with AUD. The performance of the AUDIT-C score varied among hospitalised patients based on their ICD-10 diagnosis, which should be considered when implementing universal alcohol screening in these settings.

Incidence of long-term conditions in the Latin American community of London: A validation and retrospective cohort study of 890,922 primary care records, 2005-2022

BACKGROUND

Minoritised populations in the United Kingdom frequently identify in multiple ethnic groupings and therefore little is known of their health needs. There were 136,062 Latin American people recorded in the 2021 UK Census across six different ethnic groups.

AIM

Characterise the incidence of long-term conditions (LTCs) and multiple LTCs (mLTCs) amongst the Latin American community of London. Compare the incidence of LTCs in the Latin American population to other ethnic groups.

DESIGN AND SETTING

Retrospective cohort study using pseudonymised primary care data from 890,922 individuals in an urban, superdiverse area of London from 2005-2022.

METHOD

Latin American individuals were identified using country of birth, language and ethnicity codes, and validated against Census findings. Multivariable competing risks regression models estimated the effect of being Latin American, compared to the White British ethnic group, on incidence of 32 LTCs and risk factors relevant to urban populations.

RESULTS

28,617 Latin American people were identified in this cohort, 3.2% of total. In multivariable analysis, compared to the White British ethnic group, being Latin American was associated with twice the rate of HIV/AIDS (hazard ratio (HR) 2.00; 95% confidence interval (CI) 1.65-2.43), 60% increased rate of diabetes (HR 1.61; 95%CI 1.47-1.77) and almost twice the rate of systemic lupus erythematosus and rheumatoid arthritis (HRs 2.28; 95% CI 1.18-4.38 and 1.69; 95% CI 1.32-2.17 respectively).

CONCLUSION

Using commonly-recorded primary care codes accurately and reliably identifies markedly higher risks of HIV/AIDS, diabetes and joint disease among London's Latin American population. These data can be used to target inclusive and equitable health interventions.

Recording of Alcohol Use Disorder in Electronic Health Records: Developing a Recommended Codelist for Research

Purpose

Electronic health records (EHR) are valuable resources for health research; however, their use is challenging. A validated alcohol use disorder (AUD) codelist for UK primary care is needed to improve population-based research in this patient group. We aimed to develop an AUD codelist for use in the Clinical Practice Research Datalink (CPRD) Aurum database, a UK EHR primary-care database.

Methods

The CPRD code browser was searched using keywords related to alcohol use using a previously developed search strategy. The resulting codes were categorised as AUD if they were: a) diagnostic of AUD, b) indicated alcohol withdrawal, or c) indicated chronic alcohol-related harm (physical or mental). Codes related to alcohol use but not used to define AUD were also classified into relevant categories (alcohol status, acute harm, and alcohol screening). All codes were categorised independently by at least two reviewers (one person reviewed all codes and five reviewers (all practising GPs) each reviewed a subset of codes (100-200 codes each). Disagreements in categorisation were discussed by at least three coders and a consensus was reached. The reliability of categorisation was assessed using kappa statistics.

Results

In total, 556 potential codes related to alcohol use were identified. The Kappa for reliability between coders was moderate for both AUD (0.72) and across all categories (0.62), with substantial variability between coders (AUD: 0.33-0.97; all categories 0.36-0.74). In the final codelist, 138 codes were included as indicating AUD: 38 codes identified which indicated diagnosis of AUD, 14 indicating withdrawal plus 85 codes indicating chronic alcohol-related harm (41 physical health and 44 mental health).

Conclusion

Many codes are used in primary care to record alcohol use and associated harms, and there is substantial variability in how clinicians categorise them. While future work formally validating the codelist against gold standard clinical reviews and qualitative work with General Practitioners is needed for a deeper understanding of coding processes, we have documented here the process used for the development of an AUD codelist within primary care which can be used as a reference for future research.

Healthcare interactions prior to first hospital admission with alcohol-related liver disease

BACKGROUND AND AIMS

To examine the healthcare contacts of patients in the year before an index admission to hospital with alcohol-related liver disease (ArLD) to identify where opportunities for earlier identification of alcohol use disorders (AUD) and ArLD and intervention may occur.

METHODS

A retrospective cohort study using the regional database encompassing NHS organisations across North West London (344 general practitioner [GP] practices, 4 acute hospital trusts and 2 mental health and community health trusts). Patients who had an index admission with ArLD were identified through healthcare coding and compared with a control cohort. Healthcare contacts, blood tests and AUD testing in the year preceding admission were measured.

RESULTS

The ArLD cohort had 1494 participants with an index hospital admission with ArLD. The control cohort included 4462 participants. In the year preceding an index admission with ArLD, 91% of participants had at least one contact with primary care with an average of 2.97 (SD 2.45) contacts; 80% (n = 1199/1494) attended ED, 68% attended an outpatient clinic, and 42% (n = 628/1494) had at least one inpatient admission. Only 9% of the ArLD (137/1494) had formal testing for AUD. Abnormal bilirubin and platelets were more common in the ArLD than the control cohort 25% (138/560) and 28% (231/837), respectively, v 1% (12/1228) and 1% (20/1784).

CONCLUSIONS

Prior to an index admission with ArLD patients have numerous interactions with all healthcare settings, indicating missed opportunities for early identification and treatment.

Population-Based Study of Alcohol-Related Liver Disease in England in 2001-2018: Influence of Socioeconomic Position

INTRODUCTION

England has seen an increase in deaths due to alcohol-related liver disease (ALD) since 2001. We studied the influence of socioeconomic position on the incidence of ALD and the mortality after ALD diagnosis in England in 2001-2018.

METHODS

This was an observational cohort study based on health records contained within the UK Clinical Practice Research Datalink covering primary care, secondary care, cause of death registration, and deprivation of neighborhood areas in 18.8 million residents. We estimated incidence rate and incidence rate ratios of ALD and hazard ratios of mortality.

RESULTS

ALD was diagnosed in 57,784 individuals with a median age of 54 years and of whom 43% had cirrhosis. The ALD incidence rate increased by 65% between 2001 and 2018 in England to reach 56.1 per 100,000 person-years in 2018. The ALD incidence was 3-fold higher in those from the most deprived quintile vs those from the least deprived quintile (incidence rate ratio 3.30, 95% confidence interval 3.21-3.38), with reducing inequality at older than at younger ages. For 55- to 74-year-olds, there was a notable increase in the incidence rate between 2001 and 2018, from 96.1 to 158 per 100,000 person-years in the most deprived quintile and from 32.5 to 70.0 in the least deprived quintile. After ALD diagnosis, the mortality risk was higher for patients from the most deprived quintile vs those from the least deprived quintile (hazard ratio 1.22, 95% confidence interval 1.18-1.27), and this ratio did not change during 2001-2018.

DISCUSSION

The increasing ALD incidence in England is a greater burden on individuals of low economic position compared with that on those of high socioeconomic position. This finding highlights ALD as a contributor to inequality in health.

How has the COVID-19 pandemic affected the delivery of preventive healthcare? An interrupted time series analysis of adults in English primary care from 2018 to 2022

OBJECTIVE

Offering advice and support for smoking, obesity, excess alcohol, and physical inactivity is an evidence-based component of primary care. The objective was to quantify the impact of the pandemic on the rate of advice or referral for these four risk factors.

METHODS

A retrospective cohort study using primary care data from 1847 practices in England and 21,191,389 patients contributing to the Oxford Clinical Informatics Digital Hub. An interrupted time series analysis was undertaken with a single change point (March 2020). Monthly trends were modelled from 1st January 2018 - 30th June 2022 using segmented linear regression.

RESULTS

There was an initial step reduction in advice and referrals for smoking, obesity, excess alcohol, and physical inactivity in March 2020. By June 2022, advice on smoking (slope change -0.02 events per hundred patient years/month (EPH/month); 95% confidence interval (CI) -0.17, 0.21), obesity (0.06 EPH/month; 95% CI 0.01, 0.12), alcohol (0.02 EPH/month; 95% CI -0.01, 0.05) and physical inactivity (0.05 EPH/month; 95% CI 0.01, 0.09) had not returned to pre-pandemic levels. Similarly, smoking cessation referral remained lower (0.01 EPH/month; 95% CI -0.01, 0.09), excess alcohol referral returned to similar levels (0.0005 EPH/month; 95% CI 0.0002, 0.0008), while referral for obesity (0.14 EPH/month; 95% CI 0.10, 0.19) and physical inactivity (0.01 EPH/month; 95% CI 0.01, 0.02) increased relative to pre-pandemic rates.

CONCLUSION

Advice and support for smoking, and advice for weight, excess alcohol and physical inactivity have not returned to pre-pandemic levels. Clinicians and policy makers should prioritise preventive care in COVID-19 recovery plans.

Screening for alcohol use in primary care: assessing French general practitioner practices

Alcohol use is a leading risk factor for premature death and disability. To tackle this issue, more systematic and accurate screening for at-risk consumption is needed in healthcare systems, especially by general practitioners (GPs). We assessed the frequency of at-risk consumption screening by GPs in France. We also identified characteristics associated with more frequent screening and greater use of validated screening tools by these healthcare providers. A cross-sectional survey was conducted among a representative sample of French GPs. Multinomial logistic regressions were used to identify factors associated with more frequent screening and greater use of validated screening tools. Response rate was of 73%. Of the 2412 participants, 42.8% screened all their patients systematically and repeatedly, while 48.0% never used standardized tools to screen potentially at-risk patients. Among other characteristics, being aware of and using the "early identification and brief intervention" screening strategy, and feeling absolutely comfortable talking with patients about reducing or stopping their alcohol use, were both associated with more frequent screening and use of standardized tools. Our results on at-risk alcohol use screening highlight an improvement over data from previous studies. Nevertheless, better training of French GPs in good alcohol screening practices-specifically, increased screening frequency and greater use of standardized tools-may improve identification of at-risk patients.

Multifaceted intervention to increase the delivery of alcohol brief interventions in primary care: a mixed-methods process analysis

BACKGROUND

Brief interventions (BIs) are effective for reducing harmful alcohol consumption, but their use in primary care is less frequent than clinically indicated. The REducing AlCohol- related Harm (REACH) project aimed to increase the delivery of BIs in primary care.

AIM

To assess the effectiveness of the REACH programme in increasing alcohol BIs in general practice and explore the implementation factors that improve or reduce uptake by clinicians.

DESIGN AND SETTING

This article reports on a sequential, explanatory mixed-methods study of the implementation of the REACH project in six general practice clinics serving low-income communities in Melbourne, Australia.

METHOD

Time-series analyses were conducted using routinely collected patient records and semi-structured interviews, guided by the consolidated framework for implementation research.

RESULTS

The six intervention sites significantly increased their rate of recorded alcohol status (56.7% to 60.4%), whereas there was no significant change in the non-intervention practices (344 sites, 55.2% to 56.4%).

CONCLUSION

REACH resources were seen as useful and acceptable by clinicians and staff. National policies that support the involvement of primary care in alcohol harm reduction helped promote ongoing intervention sustainability.

Integration, effectiveness and costs of different models of primary health care provision for people who are homeless: an evaluation study

Background

There is a high prevalence of health problems among single people who are homeless. Specialist primary health care services for this population have been developed in several locations across England; however, there have been very few evaluations of these services.

Objectives

This study evaluated the work of different models of primary health care provision in England to determine their effectiveness in engaging people who are homeless in health care and in providing continuity of care for long-term conditions. It concerned single people (not families or couples with dependent children) staying in hostels, other temporary accommodation or on the streets. The influence on outcomes of contextual factors and mechanisms (service delivery factors), including integration with other services, were examined. Data from medical records were collated on participants' use of health care and social care services over 12 months, and costs were calculated.

Design and setting

The evaluation involved four existing Health Service Models: (1) health centres primarily for people who are homeless (Dedicated Centres), (2) Mobile Teams providing health care in hostels and day centres, (3) Specialist GPs providing some services exclusively for patients who are homeless and (4) Usual Care GPs providing no special services for people who are homeless (as a comparison). Two Case Study Sites were recruited for each of the specialist models, and four for the Usual Care GP model.

Participants

People who had been homeless during the previous 12 months were recruited as 'case study participants'; they were interviewed at baseline and at 4 and 8 months, and information was collected about their circumstances and their health and service use in the preceding 4 months. Overall, 363 participants were recruited; medical records were obtained for 349 participants. Interviews were conducted with 65 Case Study Site staff and sessional workers, and 81 service providers and stakeholders.

Results

The primary outcome was the extent of health screening for body mass index, mental health, alcohol use, tuberculosis, smoking and hepatitis A among participants, and evidence of an intervention if a problem was identified. There were no overall differences in screening between the models apart from Mobile Teams, which scored considerably lower. Dedicated Centres and Specialist GPs were more successful in providing continuity of care for participants with depression and alcohol and drug problems. Service use and costs were significantly higher for Dedicated Centre participants and lower for Usual Care GP participants. Participants and staff welcomed flexible and tailored approaches to care, and related services being available in the same building. Across all models, dental needs were unaddressed and staff reported poor availability of mental health services.

Limitations

There were difficulties recruiting mainstream general practices for the Usual Care GP model. Medical records could not be accessed for 14 participants of this model.

Conclusions

Participant characteristics, contextual factors and mechanisms were influential in determining outcomes. Overall, outcomes for Dedicated Centres and for one of the Specialist GP sites were relatively favourable. They had dedicated staff for patients who were homeless, 'drop-in' services, on-site mental health and substance misuse services, and worked closely with hospitals and homelessness sector services.

Funding

This project was funded by the National Institute for Health and Care Research (NIHR) Health and Social Care Delivery Research programme (HSDR 13/156/03) and will be published in full in Health and Social Care Delivery Research; Vol. 11, No. 16. See the NIHR Journals Library website for further project information.

Development of an algorithm to classify primary care electronic health records of alcohol consumption: experience using data linkage from UK Biobank and primary care electronic health data sources

OBJECTIVES

Develop a novel algorithm to categorise alcohol consumption using primary care electronic health records (EHRs) and asses its reliability by comparing this classification with self-reported alcohol consumption data obtained from the UK Biobank (UKB) cohort.

DESIGN

Cross-sectional study.

SETTING

The UKB, a population-based cohort with participants aged between 40 and 69 years recruited across the UK between 2006 and 2010.

PARTICIPANTS

UKB participants from Scotland with linked primary care data.

PRIMARY AND SECONDARY OUTCOME MEASURES

Create a rule-based multiclass algorithm to classify alcohol consumption reported by Scottish UKB participants and compare it with their classification using data present in primary care EHRs based on Read Codes. We evaluated agreement metrics (simple agreement and kappa statistic).

RESULTS

Among the Scottish UKB participants, 18 838 (69%) had at least one Read Code related to alcohol consumption and were used in the classification. The agreement of alcohol consumption categories between UKB and primary care data, including assessments within 5 years was 59.6%, and kappa was 0.23 (95% CI 0.21 to 0.24). Differences in classification between the two sources were statistically significant (p<0.001); More individuals were classified as 'sensible drinkers' and in lower alcohol consumption levels in primary care records compared with the UKB. Agreement improved slightly when using only numerical values (k=0.29; 95% CI 0.27 to 0.31) and decreased when using qualitative descriptors only (k=0.18;95% CI 0.16 to 0.20).

CONCLUSION

Our algorithm classifies alcohol consumption recorded in Primary Care EHRs into discrete meaningful categories. These results suggest that alcohol consumption may be underestimated in primary care EHRs. Using numerical values (alcohol units) may improve classification when compared with qualitative descriptors.

Alcohol use recording in adults with depression in English primary care: a cross-sectional study

OBJECTIVES

To investigate alcohol use recording in people with newly diagnosed depression in English primary care and individual characteristics associated with the recording of alcohol use.

DESIGN

A population-based cross-sectional study.

SETTING

Primary care data from English practices contributing to the UK Clinical Practice Research Datalink.

PARTICIPANTS

We included adults (18+ years) diagnosed with depression between 1 January 2011 and 1 January 2017 without previous antidepressant use and at least 1 year of registration before diagnosis.

PRIMARY AND SECONDARY OUTCOME MEASURES

We described the proportion of individuals with alcohol use and level of alcohol use recorded at four time points (the date of depression diagnosis, 3 months before or after depression diagnosis, 12 months before or after depression diagnosis and any point pre or postdepression diagnosis). We used logistic regression to investigate individual characteristics associated with alcohol use recording in the 3 months before or after depression diagnosis.

RESULTS

We identified 36 424 adults with depression. 538 (2%) had alcohol use recorded in the 3 months before or after depression diagnosis using formal validated methods such as the Alcohol Use Disorders Identification Test and its abbreviated versions. At each time point, most individuals with alcohol use recorded were low risk drinkers. Alcohol use recording in the 3 months before or after depression diagnosis was associated with male sex (OR=1.38, 95% CI 1.29 to 1.48) and several other individual-level factors.

CONCLUSIONS

Our study shows low levels of alcohol use recording in the 3 months before or after depression diagnosis. Levels of alcohol use recording varied depending on individual characteristics. Incentivised recording of alcohol use will increase completeness, which could improve clinical management and reduce missed opportunities for care in people with depression.

The role of organization of care in GPs' prevention practice

BACKGROUND

General practitioners (GPs) do not systematically include preventive recommendations in their practice, and some characteristics of health care organization are associated with more systematic prevention. But the characteristics of health care organization may act in a nonuniform manner depending on the type of preventive care. Thus, one characteristic can be positively associated with one type of preventive care and negatively associated with another. Our aim was to investigate the association between health care organization in general practice and different areas of preventive care (immunization and addiction prevention), in search of nonuniform associations.

METHODS

We used a representative survey of 1,813 French GPs conducted in 2009. Four preventive care practices were studied: immunization through flu and HPV vaccination, and prevention of addictive behaviors concerning tobacco and alcohol use.Characteristics of GPs' health care organization and the social context of their practice were collected (spatial accessibility to GPs and socioeconomic level of the area of practice). We constructed mixed models to study associations and interactions between the organization variables and preventive care.

RESULTS

Four out of five characteristics of GPs' organization have uneven impacts on different types of preventive care (p-interaction < 10-4). For example, number of daily consultations is associated with better immunization prevention but with poorer prevention counseling in addictive behaviors. In contrast, working with digital medical files is uniformly associated with both types of preventive care (OR = 1.29 [1.15-1.45]; P < 10-4).

CONCLUSION

An approach centered on specific types of preventive care should help deepen our understanding of prevention and possibly help to identify a new typology for preventive care.

Does the Fast Alcohol Screening Test Accurately Distinguish Between Harmful and Severely Dependent Tiers of Alcohol Misuse?

AIMS

Primary aim: to determine the efficacy of FAST (the Fast Alcohol Screening Test) for detecting harmful and dependent levels of alcohol use. Secondary aim: to compare the performance of the FAST to two short forms of the Alcohol Use Disorder Identification Test (AUDIT): the AUDIT-C and AUDIT-3.

METHODS

Data from 3336 individuals in South Wales, compiled from full AUDIT datasets, were examined. AUROC analysis, alongside measures of sensitivity and specificity of the FAST, AUDIT-C and AUDIT-3 were utilized for the identification of harmful and dependent alcohol use.

RESULTS

The FAST demonstrated efficacy in the identification of harmful and dependent levels of alcohol use, with superior performance to both the AUDIT-C and AUDIT-3.

CONCLUSION

The present paper demonstrates the potential of the FAST as a cost- and time-effective method for appropriate screening and signposting in the stepped care model utilized by many health care and treatment services. Further studies are needed to ensure validity, both within the general population and for specific services and populations.

Renin-Angiotensin-Aldosterone System Inhibitors and Risk of Acute Pancreatitis: A Population-Based Cohort Study

INTRODUCTION

There are conflicting reports on the effects of angiotensin-converting enzyme (ACE) inhibitors and angiotensin II receptor blockers (ARBs) on acute pancreatitis incidence.

OBJECTIVE

The aim was to determine whether use of ACE inhibitors and ARBs is associated with the incidence of acute pancreatitis, compared with use of dihydropyridine calcium channel blockers (dCCBs).

METHODS

We assembled two population-based, new-user, active comparator cohorts using the United Kingdom Clinical Practice Research Datalink linked to the Hospital Episode Statistics repository and Office for National Statistics from 1998 to 2018, with follow-up until 2019. The first cohort included 304,083 ACE inhibitor initiators and 194,431 dCCB initiators. The second cohort included 29,160 ARB initiators and 203,610 dCCB initiators. Cox proportional hazards models were fit to estimate hazard ratios (HRs) with 95% confidence intervals (CIs) of acute pancreatitis, comparing ACE inhibitors and ARBs, separately, with dCCBs. Models were weighted using standardized mortality ratio weights generated from calendar time-specific propensity scores.

RESULTS

ACE inhibitors were associated with an increased risk of acute pancreatitis, compared with dCCBs (64.3 vs 45.2 per 100,000 person-years, respectively; HR 1.45, 95% CI 1.15-1.83). The number needed to harm after 2 and 5 years of use was 2438 and 1019, respectively. In contrast, ARBs were not associated with an increased risk of acute pancreatitis, compared with dCCBs (40.1 vs 47.6 per 100,000 person-years, respectively; HR 0.88, 95% CI 0.60-1.31).

CONCLUSIONS

ACE inhibitors were associated with a modest increased risk of acute pancreatitis compared with dCCBs. This association should be balanced with the known clinical benefits of ACE inhibitors in hypertension management. In contrast, no association was observed with ARBs.

Using Electronic Medical Records to Identify Potentially Eligible Study Subjects for Lung Cancer Screening with Biomarkers

Lung cancer screening trials using low-dose computed tomography (LDCT) show reduced late-stage diagnosis and mortality rates. These trials have identified high-risk groups that would benefit from screening. However, these sub-populations can be difficult to access and retain in trials. Implementation of national screening programmes further suggests that there is poor uptake in eligible populations. A new approach to participant selection may be more effective. Electronic medical records (EMRs) are a viable alternative to population-based or health registries, as they contain detailed clinical and demographic information. Trials have identified that e-screening using EMRs has improved trial retention and eligible subject identification. As such, this paper argues for greater use of EMRs in trial recruitment and screening programmes. Moreover, this opinion paper explores the current issues in and approaches to lung cancer screening, whether records can be used to identify eligible subjects for screening and the challenges that researchers face when using EMR data.

Can alcohol consumption in Germany be reduced by alcohol screening, brief intervention and referral to treatment in primary health care? Results of a simulation study

BACKGROUND

Screening, brief intervention and referral to treatment (SBIRT) is a programme to reduce alcohol consumption for drinkers with high alcohol consumption levels. Only 2.9% of patients in primary health care (PHC) are screened for their alcohol use in Germany, despite high levels of alcohol consumption and attributable harm. We developed an open-access simulation model to estimate the impact of higher SBIRT delivery rates in German PHC settings on population-level alcohol consumption.

METHODS AND FINDINGS

A hypothetical population of drinkers and non-drinkers was simulated by sex, age, and educational status for the year 2009 based on survey and sales data. Risky drinking persons receiving BI or RT were sampled from this population based on screening coverage and other parameters. Running the simulation model for a ten-year period, drinking levels and heavy episodic drinking (HED) status were changed based on effect sizes from meta-analyses. In the baseline scenario of 2.9% screening coverage, 2.4% of the adult German population received a subsequent intervention between 2009 and 2018. If every second PHC patient would have been screened for alcohol use, 21% of adult residents in Germany would have received BI or RT by the end of the ten-year simulation period. In this scenario, population-level alcohol consumption would be 11% lower than it was in 2018, without any impact on HED prevalence. Screening coverage rates below 10% were not found to have a measurable effect on drinking levels.

CONCLUSIONS

Large-scale implementation of SBIRT in PHC settings can yield substantial reductions of alcohol consumption in Germany. As high screening coverage rates may only be achievable in the long run, other effective alcohol policies are required to achieve short-term reduction of alcohol use and attributable harm in Germany. There is large potential to apply this open-access simulation model to other settings and for other alcohol interventions.

Can we screen for pancreatic cancer? Identifying a sub-population of patients at high risk of subsequent diagnosis using machine learning techniques applied to primary care data

BACKGROUND

Pancreatic cancer (PC) represents a substantial public health burden. Pancreatic cancer patients have very low survival due to the difficulty of identifying cancers early when the tumour is localised to the site of origin and treatable. Recent progress has been made in identifying biomarkers for PC in the blood and urine, but these cannot be used for population-based screening as this would be prohibitively expensive and potentially harmful.

METHODS

We conducted a case-control study using prospectively-collected electronic health records from primary care individually-linked to cancer registrations. Our cases were comprised of 1,139 patients, aged 15-99 years, diagnosed with pancreatic cancer between January 1, 2005 and June 30, 2009. Each case was age-, sex- and diagnosis time-matched to four non-pancreatic (cancer patient) controls. Disease and prescription codes for the 24 months prior to diagnosis were used to identify 57 individual symptoms. Using a machine learning approach, we trained a logistic regression model on 75% of the data to predict patients who later developed PC and tested the model's performance on the remaining 25%.

RESULTS

We were able to identify 41.3% of patients < = 60 years at 'high risk' of developing pancreatic cancer up to 20 months prior to diagnosis with 72.5% sensitivity, 59% specificity and, 66% AUC. 43.2% of patients >60 years were similarly identified at 17 months, with 65% sensitivity, 57% specificity and, 61% AUC. We estimate that combining our algorithm with currently available biomarker tests could result in 30 older and 400 younger patients per cancer being identified as 'potential patients', and the earlier diagnosis of around 60% of tumours.

CONCLUSION

After further work this approach could be applied in the primary care setting and has the potential to be used alongside a non-invasive biomarker test to increase earlier diagnosis. This would result in a greater number of patients surviving this devastating disease.