Factors influencing clinician and patient interaction with machine learning-based risk prediction models: a systematic review.

Machine learning (ML)-based risk prediction models hold the potential to support the health-care setting in several ways; however, use of such models is scarce. We aimed to review health-care professional (HCP) and patient perceptions of ML risk prediction models in published literature, to inform future risk prediction model development. Following database and citation searches, we identified 41 articles suitable for inclusion. Article quality varied with qualitative studies performing strongest. Overall, perceptions of ML risk prediction models were positive. HCPs and patients considered that models have the potential to add benefit in the health-care setting. However, reservations remain; for example, concerns regarding data quality for model development and fears of unintended consequences following ML model use. We identified that public views regarding these models might be more negative than HCPs and that concerns (eg, extra demands on workload) were not always borne out in practice. Conclusions are tempered by the low number of patient and public studies, the absence of participant ethnic diversity, and variation in article quality. We identified gaps in knowledge (particularly views from under-represented groups) and optimum methods for model explanation and alerts, which require future research.

Assessing eligibility for lung cancer screening using parsimonious ensemble machine learning models: A development and validation study.

BACKGROUND: Risk-based screening for lung cancer is currently being considered in several countries; however, the optimal approach to determine eligibility remains unclear. Ensemble machine learning could support the development of highly parsimonious prediction models that maintain the performance of more complex models while maximising simplicity and generalisability, supporting the widespread adoption of personalised screening. In this work, we aimed to develop and validate ensemble machine learning models to determine eligibility for risk-based lung cancer screening. METHODS AND FINDINGS: For model development, we used data from 216,714 ever-smokers recruited between 2006 and 2010 to the UK Biobank prospective cohort and 26,616 high-risk ever-smokers recruited between 2002 and 2004 to the control arm of the US National Lung Screening (NLST) randomised controlled trial. The NLST trial randomised high-risk smokers from 33 US centres with at least a 30 pack-year smoking history and fewer than 15 quit-years to annual CT or chest radiography screening for lung cancer. We externally validated our models among 49,593 participants in the chest radiography arm and all 80,659 ever-smoking participants in the US Prostate, Lung, Colorectal and Ovarian (PLCO) Screening Trial. The PLCO trial, recruiting from 1993 to 2001, analysed the impact of chest radiography or no chest radiography for lung cancer screening. We primarily validated in the PLCO chest radiography arm such that we could benchmark against comparator models developed within the PLCO control arm. Models were developed to predict the risk of 2 outcomes within 5 years from baseline: diagnosis of lung cancer and death from lung cancer. We assessed model discrimination (area under the receiver operating curve, AUC), calibration (calibration curves and expected/observed ratio), overall performance (Brier scores), and net benefit with decision curve analysis. Models predicting lung cancer death (UCL-D) and incidence (UCL-I) using 3 variables-age, smoking duration, and pack-years-achieved or exceeded parity in discrimination, overall performance, and net benefit with comparators currently in use, despite requiring only one-quarter of the predictors. In external validation in the PLCO trial, UCL-D had an AUC of 0.803 (95% CI: 0.783, 0.824) and was well calibrated with an expected/observed (E/O) ratio of 1.05 (95% CI: 0.95, 1.19). UCL-I had an AUC of 0.787 (95% CI: 0.771, 0.802), an E/O ratio of 1.0 (95% CI: 0.92, 1.07). The sensitivity of UCL-D was 85.5% and UCL-I was 83.9%, at 5-year risk thresholds of 0.68% and 1.17%, respectively, 7.9% and 6.2% higher than the USPSTF-2021 criteria at the same specificity. The main limitation of this study is that the models have not been validated outside of UK and US cohorts. CONCLUSIONS: We present parsimonious ensemble machine learning models to predict the risk of lung cancer in ever-smokers, demonstrating a novel approach that could simplify the implementation of risk-based lung cancer screening in multiple settings.

Autoimmune diseases and cardiovascular risk: a population-based study on 19 autoimmune diseases and 12 cardiovascular diseases in 22 million individuals in the UK.

BACKGROUND: Some autoimmune diseases are associated with an increased risk of cardiovascular disease. We aimed to determine whether or not this is true, and to what extent, for a broad range of autoimmune conditions. METHODS: In this population-based study, we used linked primary and secondary care records from the Clinical Practice Research Datalink (CPRD), GOLD and Aurum datasets, to assemble a cohort of individuals across the UK who were newly diagnosed with any of 19 autoimmune diseases between Jan 1, 2000, and Dec 31, 2017, younger than 80 years at diagnosis, and free of cardiovascular diseases up to 12 months after diagnosis. We also assembled a matched cohort with up to five individuals matched on age, sex, socioeconomic status, region, and calendar year, who were free of autoimmune disease and free of cardiovascular diseases up to 12 months after study entry. Both cohorts were followed up until June 30, 2019. We investigated the incidence of 12 cardiovascular outcomes and used Cox proportional hazards models to examine differences in patients with and without autoimmune diseases. FINDINGS: Of 22 009 375 individuals identified from the CPRD databases, we identified 446 449 eligible individuals with autoimmune diseases and 2 102 830 matched controls. In the autoimmune cohort, mean age at diagnosis was 46·2 years (SD 19·8), and 271 410 (60·8%) were women and 175 039 (39·2%) were men. 68 413 (15·3%) people with and 231 410 (11·0%) without autoimmune diseases developed incident cardiovascular disease during a median of 6·2 years (IQR 2·7-10·8) of follow-up. The incidence rate of cardiovascular disease was 23·3 events per 1000 patient-years among patients with autoimmune disease and 15·0 events per 1000 patient-years among those without an autoimmune disease (hazard ratio [HR] 1·56 [95% CI 1·52-1·59]). An increased risk of cardiovascular disease with autoimmune disease was seen for every individual cardiovascular disease and increased progressively with the number of autoimmune diseases present (one disease: HR 1·41 [95% CI 1·37-1·45]; two diseases: 2·63 [2·49-2·78]); three or more diseases: 3·79 [3·36-4·27]), and in younger age groups (age <45 years: 2·33 [2·16-2·51]; 55-64 years: 1·76 [1·67-1·85]; ≥75 years: 1·30 [1·24-1·36]). Among autoimmune diseases, systemic sclerosis (3·59 [2·81-4·59]), Addison's disease (2·83 [1·96-4·09]), systemic lupus erythematosus (2·82 [2·38-3·33]), and type 1 diabetes (2·36 [2·21-2·52]) had the highest overall cardiovascular risk. INTERPRETATION: These findings warrant targeted cardiovascular prevention measures, in particular in younger patients with autoimmune diseases, and further research into pathophysiological mechanisms underlying these complications. FUNDING: Horizon 2020 Marie Sklodowska-Curie Actions and European Society of Cardiology.

Benefit, Harm, and Cost-effectiveness Associated With Magnetic Resonance Imaging Before Biopsy in Age-based and Risk-stratified Screening for Prostate Cancer.

Importance: If magnetic resonance imaging (MRI) mitigates overdiagnosis of prostate cancer while improving the detection of clinically significant cases, including MRI in a screening program for prostate cancer could be considered. Objective: To evaluate the benefit-harm profiles and cost-effectiveness associated with MRI before biopsy compared with biopsy-first screening for prostate cancer using age-based and risk-stratified screening strategies. Design, Setting, and Participants: This decision analytical model used a life-table approach and was conducted between December 2019 and July 2020. A hypothetical cohort of 4.48 million men in England aged 55 to 69 years were analyzed and followed-up to 90 years of age. Exposures: No screening, age-based screening, and risk-stratified screening in the hypothetical cohort. Age-based screening consisted of screening every 4 years with prostate-specific antigen between the ages of 55 and 69 years. Risk-stratified screening used age and polygenic risk profiles. Main Outcomes and Measures: The benefit-harm profile (deaths from prostate cancer, quality-adjusted life-years, overdiagnosis, and biopsies) and cost-effectiveness (net monetary benefit, from a health care system perspective) were analyzed. Both age-based and risk-stratified screening were evaluated using a biopsy-first and an MRI-first diagnostic pathway. Results were derived from probabilistic analyses and were discounted at 3.5% per annum. Results: The hypothetical cohort included 4.48 million men in England, ranging in age from 55 to 69 years (median, 62 years). Compared with biopsy-first age-based screening, MRI-first age-based screening was associated with 0.9% (1368; 95% uncertainty interval [UI], 1370-1409) fewer deaths from prostate cancer, 14.9% (12 370; 95% UI, 11 100-13 670) fewer overdiagnoses, and 33.8% (650 500; 95% UI, 463 200-907 000) fewer biopsies. At 10-year absolute risk thresholds of 2% and 10%, MRI-first risk-stratified screening was associated with between 10.4% (7335; 95% UI, 6630-8098) and 72.6% (51 250; 95% UI, 46 070-56 890) fewer overdiagnosed cancers, respectively, and between 21.7% fewer MRIs (412 100; 95% UI, 411 400-412 900) and 53.5% fewer biopsies (1 016 000; 95% UI, 1 010 000-1 022 000), respectively, compared with MRI-first age-based screening. The most cost-effective strategies at willingness-to-pay thresholds of £20 000 (US $26 000) and £30 000 (US $39 000) per quality-adjusted life-year gained were MRI-first risk-stratified screening at 10-year absolute risk thresholds of 8.5% and 7.5%, respectively. Conclusions and Relevance: In this decision analytical model of a hypothetical cohort, an MRI-first diagnostic pathway was associated with an improvement in the benefit-harm profile and cost-effectiveness of screening for prostate cancer compared with biopsy-first screening. These improvements were greater when using risk-stratified screening based on age and polygenic risk profile and may warrant prospective evaluation.

Automated and partly automated contact tracing: a systematic review to inform the control of COVID-19.

Evidence for the use of automated or partly automated contact-tracing tools to contain severe acute respiratory syndrome coronavirus 2 is scarce. We did a systematic review of automated or partly automated contact tracing. We searched PubMed, EMBASE, OVID Global Health, EBSCO Medical COVID Information Portal, Cochrane Library, medRxiv, bioRxiv, arXiv, and Google Advanced for articles relevant to COVID-19, severe acute respiratory syndrome, Middle East respiratory syndrome, influenza, or Ebola virus, published from Jan 1, 2000, to April 14, 2020. We also included studies identified through professional networks up to April 30, 2020. We reviewed all full-text manuscripts. Primary outcomes were the number or proportion of contacts (or subsequent cases) identified. Secondary outcomes were indicators of outbreak control, uptake, resource use, cost-effectiveness, and lessons learnt. This study is registered with PROSPERO (CRD42020179822). Of the 4036 studies identified, 110 full-text studies were reviewed and 15 studies were included in the final analysis and quality assessment. No empirical evidence of the effectiveness of automated contact tracing (regarding contacts identified or transmission reduction) was identified. Four of seven included modelling studies that suggested that controlling COVID-19 requires a high population uptake of automated contact-tracing apps (estimates from 56% to 95%), typically alongside other control measures. Studies of partly automated contact tracing generally reported more complete contact identification and follow-up compared with manual systems. Automated contact tracing could potentially reduce transmission with sufficient population uptake. However, concerns regarding privacy and equity should be considered. Well designed prospective studies are needed given gaps in evidence of effectiveness, and to investigate the integration and relative effects of manual and automated systems. Large-scale manual contact tracing is therefore still key in most contexts.

Publisher Correction: Personalized early detection and prevention of breast cancer: ENVISION consensus statement.

An amendment to this paper has been published and can be accessed via a link at the top of the paper.

Personalized early detection and prevention of breast cancer: ENVISION consensus statement.

The European Collaborative on Personalized Early Detection and Prevention of Breast Cancer (ENVISION) brings together several international research consortia working on different aspects of the personalized early detection and prevention of breast cancer. In a consensus conference held in 2019, the members of this network identified research areas requiring development to enable evidence-based personalized interventions that might improve the benefits and reduce the harms of existing breast cancer screening and prevention programmes. The priority areas identified were: 1) breast cancer subtype-specific risk assessment tools applicable to women of all ancestries; 2) intermediate surrogate markers of response to preventive measures; 3) novel non-surgical preventive measures to reduce the incidence of breast cancer of poor prognosis; and 4) hybrid effectiveness-implementation research combined with modelling studies to evaluate the long-term population outcomes of risk-based early detection strategies. The implementation of such programmes would require health-care systems to be open to learning and adapting, the engagement of a diverse range of stakeholders and tailoring to societal norms and values, while also addressing the ethical and legal issues. In this Consensus Statement, we discuss the current state of breast cancer risk prediction, risk-stratified prevention and early detection strategies, and their implementation. Throughout, we highlight priorities for advancing each of these areas.

Blood Pressure and Risk of Vascular Dementia: Evidence From a Primary Care Registry and a Cohort Study of Transient Ischemic Attack and Stroke.

BACKGROUND AND PURPOSE: Vascular dementia is the second most common form of dementia but reliable evidence on age-specific associations between blood pressure (BP) and risk of vascular dementia is limited and some studies have reported negative associations at older ages. METHODS: In a cohort of 4.28 million individuals, free of known vascular disease and dementia and identified from linked electronic primary care health records in the United Kingdom (Clinical Practice Research Datalink), we related BP to time to physician-diagnosed vascular dementia. We further determined associations between BP and dementia in a prospective population-based cohort of incident transient ischemic attack and stroke (Oxford Vascular Study). RESULTS: For a median follow-up of 7.0 years, 11 114 initial presentations of vascular dementia were observed in the primary care cohort after exclusion of the first 4 years of follow-up. The association between usual systolic BP and risk of vascular dementia decreased with age (hazard ratio per 20 mm Hg higher systolic BP, 1.62; 95% confidence interval, 1.13-2.35 at 30-50 years; 1.26, 1.18-1.35 at 51-70 years; 0.97, 0.92-1.03 at 71-90 years; P trend=0.006). Usual systolic BP remained predictive of vascular dementia after accounting for effect mediation by stroke and transient ischemic attack. In the population-based cohort, prior systolic BP was predictive of 5-year risk of dementia with no evidence of negative association at older ages. CONCLUSIONS: BP is positively associated with risk of vascular dementia, irrespective of preceding transient ischemic attack or stroke. Previous reports of inverse associations in old age could not be confirmed.

Blood pressure lowering for prevention of cardiovascular disease and death: a systematic review and meta-analysis.

BACKGROUND: The benefits of blood pressure lowering treatment for prevention of cardiovascular disease are well established. However, the extent to which these effects differ by baseline blood pressure, presence of comorbidities, or drug class is less clear. We therefore performed a systematic review and meta-analysis to clarify these differences. METHOD: For this systematic review and meta-analysis, we searched MEDLINE for large-scale blood pressure lowering trials, published between Jan 1, 1966, and July 7, 2015, and we searched the medical literature to identify trials up to Nov 9, 2015. All randomised controlled trials of blood pressure lowering treatment were eligible for inclusion if they included a minimum of 1000 patient-years of follow-up in each study arm. No trials were excluded because of presence of baseline comorbidities, and trials of antihypertensive drugs for indications other than hypertension were eligible. We extracted summary-level data about study characteristics and the outcomes of major cardiovascular disease events, coronary heart disease, stroke, heart failure, renal failure, and all-cause mortality. We used inverse variance weighted fixed-effects meta-analyses to pool the estimates. RESULTS: We identified 123 studies with 613,815 participants for the tabular meta-analysis. Meta-regression analyses showed relative risk reductions proportional to the magnitude of the blood pressure reductions achieved. Every 10 mm Hg reduction in systolic blood pressure significantly reduced the risk of major cardiovascular disease events (relative risk [RR] 0·80, 95% CI 0·77-0·83), coronary heart disease (0·83, 0·78-0·88), stroke (0·73, 0·68-0·77), and heart failure (0·72, 0·67-0·78), which, in the populations studied, led to a significant 13% reduction in all-cause mortality (0·87, 0·84-0·91). However, the effect on renal failure was not significant (0·95, 0·84-1·07). Similar proportional risk reductions (per 10 mm Hg lower systolic blood pressure) were noted in trials with higher mean baseline systolic blood pressure and trials with lower mean baseline systolic blood pressure (all ptrend>0·05). There was no clear evidence that proportional risk reductions in major cardiovascular disease differed by baseline disease history, except for diabetes and chronic kidney disease, for which smaller, but significant, risk reductions were detected. ß blockers were inferior to other drugs for the prevention of major cardiovascular disease events, stroke, and renal failure. Calcium channel blockers were superior to other drugs for the prevention of stroke. For the prevention of heart failure, calcium channel blockers were inferior and diuretics were superior to other drug classes. Risk of bias was judged to be low for 113 trials and unclear for 10 trials. Heterogeneity for outcomes was low to moderate; the I(2) statistic for heterogeneity for major cardiovascular disease events was 41%, for coronary heart disease 25%, for stroke 26%, for heart failure 37%, for renal failure 28%, and for all-cause mortality 35%. INTERPRETATION: Blood pressure lowering significantly reduces vascular risk across various baseline blood pressure levels and comorbidities. Our results provide strong support for lowering blood pressure to systolic blood pressures less than 130 mm Hg and providing blood pressure lowering treatment to individuals with a history of cardiovascular disease, coronary heart disease, stroke, diabetes, heart failure, and chronic kidney disease. FUNDING: National Institute for Health Research and Oxford Martin School.

Usual blood pressure, peripheral arterial disease, and vascular risk: cohort study of 4.2 million adults.

OBJECTIVES: To determine the subgroup specific associations between usual blood pressure and risk of peripheral arterial disease, and to examine the relation between peripheral arterial disease and a range of other types of vascular disease in a large contemporary cohort. DESIGN: Cohort study. SETTING: Linked electronic health records from 1990 to 2013 in the United Kingdom. PARTICIPANTS: 4,222,459 people aged 30-90 years, registered at a primary care practice for at least one year and with a blood pressure measurement. MAIN OUTCOME MEASURES: Time to first diagnosis of new onset peripheral arterial disease and time to first diagnosis of 12 different vascular events. RESULTS: A 20 mm Hg higher than usual systolic blood pressure was associated with a 63% higher risk of peripheral arterial disease (hazard ratio 1.63, 95% confidence interval 1.59 to 1.66). The strength of the association declined with increasing age and body mass index (P<0.001 for interaction) but was not modified by sex or smoking status. Peripheral arterial disease was associated with an increased risk of 11 different vascular events, including ischaemic heart disease (1.68, 1.58 to 1.79), heart failure (1.63, 1.52 to 1.75), aortic aneurysm (2.10, 1.79 to 2.45), and chronic kidney disease (1.31, 1.25 to 1.38), but not haemorrhagic stroke. The most common initial vascular event among those with peripheral arterial disease was chronic kidney disease (24.4% of initial events), followed by ischaemic heart disease (18.5% of initial events), heart failure (14.7%), and atrial fibrillation (13.2%). Overall estimates from this cohort were consistent with those derived from traditional studies when we pooled the findings in two meta-analyses. CONCLUSIONS: Raised blood pressure is a strong risk factor for peripheral arterial disease in a range of patient subgroups. Furthermore, clinicians should be aware that those with established peripheral arterial disease are at an increased risk of a range of other vascular events, including chronic kidney disease, ischaemic heart disease, heart failure, atrial fibrillation, and stroke.

Blood pressure lowering in type 2 diabetes: a systematic review and meta-analysis.

IMPORTANCE: Lowering blood pressure (BP) is widely used to reduce vascular risk in individuals with diabetes. OBJECTIVE: To determine the associations between BP-lowering treatment and vascular disease in type 2 diabetes. DATA SOURCES AND STUDY SELECTION: We searched MEDLINE for large-scale randomized controlled trials of BP-lowering treatment including patients with diabetes, published between January 1966 and October 2014. DATA EXTRACTION AND SYNTHESIS: Two reviewers independently extracted study characteristics and vascular outcome data. Estimates were stratified by baseline BP and achieved BP, and pooled using fixed-effects meta-analysis. MAIN OUTCOMES AND MEASURES: All-cause mortality, cardiovascular events, coronary heart disease events, stroke, heart failure, retinopathy, new or worsening albuminuria, and renal failure. RESULTS: Forty trials judged to be of low risk of bias (100,354 participants) were included. Each 10-mm Hg lower systolic BP was associated with a significantly lower risk of mortality (relative risk [RR], 0.87; 95% CI, 0.78-0.96); absolute risk reduction (ARR) in events per 1000 patient-years (3.16; 95% CI, 0.90-5.22), cardiovascular events (RR, 0.89 [95% CI, 0.83-0.95]; ARR, 3.90 [95% CI, 1.57-6.06]), coronary heart disease (RR, 0.88 [95% CI, 0.80-0.98]; ARR, 1.81 [95% CI, 0.35-3.11]), stroke (RR, 0.73 [95% CI, 0.64-0.83]; ARR, 4.06 [95% CI, 2.53-5.40]), albuminuria (RR, 0.83 [95% CI, 0.79-0.87]; ARR, 9.33 [95% CI, 7.13-11.37]), and retinopathy (RR, 0.87 [95% CI, 0.76-0.99]; ARR, 2.23 [95% CI, 0.15-4.04]). When trials were stratified by mean baseline systolic BP at greater than or less than 140 mm Hg, RRs for outcomes other than stroke, retinopathy, and renal failure were lower in studies with greater baseline systolic BP (P interaction <0.1). The associations between BP-lowering treatments and outcomes were not significantly different, irrespective of drug class, except for stroke and heart failure. Estimates were similar when all trials, regardless of risk of bias, were included. CONCLUSIONS AND RELEVANCE: Among patients with type 2 diabetes, BP lowering was associated with improved mortality and other clinical outcomes with lower RRs observed among those with baseline BP of 140 mm Hg and greater. These findings support the use of medications for BP lowering in these patients.

Heart failure care in low- and middle-income countries: a systematic review and meta-analysis.

BACKGROUND: Heart failure places a significant burden on patients and health systems in high-income countries. However, information about its burden in low- and middle-income countries (LMICs) is scant. We thus set out to review both published and unpublished information on the presentation, causes, management, and outcomes of heart failure in LMICs. METHODS AND FINDINGS: Medline, Embase, Global Health Database, and World Health Organization regional databases were searched for studies from LMICs published between 1 January 1995 and 30 March 2014. Additional unpublished data were requested from investigators and international heart failure experts. We identified 42 studies that provided relevant information on acute hospital care (25 LMICs; 232,550 patients) and 11 studies on the management of chronic heart failure in primary care or outpatient settings (14 LMICs; 5,358 patients). The mean age of patients studied ranged from 42 y in Cameroon and Ghana to 75 y in Argentina, and mean age in studies largely correlated with the human development index of the country in which they were conducted (r = 0.71, p<0.001). Overall, ischaemic heart disease was the main reported cause of heart failure in all regions except Africa and the Americas, where hypertension was predominant. Taking both those managed acutely in hospital and those in non-acute outpatient or community settings together, 57% (95% confidence interval [CI]: 49%-64%) of patients were treated with angiotensin-converting enzyme inhibitors, 34% (95% CI: 28%-41%) with beta-blockers, and 32% (95% CI: 25%-39%) with mineralocorticoid receptor antagonists. Mean inpatient stay was 10 d, ranging from 3 d in India to 23 d in China. Acute heart failure accounted for 2.2% (range: 0.3%-7.7%) of total hospital admissions, and mean in-hospital mortality was 8% (95% CI: 6%-10%). There was substantial variation between studies (p<0.001 across all variables), and most data were from urban tertiary referral centres. Only one population-based study assessing incidence and/or prevalence of heart failure was identified. CONCLUSIONS: The presentation, underlying causes, management, and outcomes of heart failure vary substantially across LMICs. On average, the use of evidence-based medications tends to be suboptimal. Better strategies for heart failure surveillance and management in LMICs are needed. Please see later in the article for the Editors' Summary.