Joint Effects Of Risk Factors Research Articles

Applications of machine learning methods in traffic crash severity modelling: current status and future directions

ABSTRACT As a key area of traffic safety research, crash severity modelling has attracted tremendous attention. Recently, there has been growing interest in applying machine learning (ML) methods in this area. However, the lessons and experience learned so far have not been systematically documented and summarised. This is the first article that surveys studies on ML applications in crash severity modelling and has the following major contributions: (1) it provides a comprehensive and critical review of current research efforts; (2) it summarises the successful experience and main challenges (e.g. data and methodology); and (3) it identifies promising research opportunities towards accurate and reliable crash severity modelling and results interpretation. The review results suggest that imbalanced data remains a major issue. Under- and over-samplings are often used to balance crash severity data despite their limitations. Some studies use local sensitivity analysis (LSA) to interpret ML modelling results but ignore the strict assumptions of LSA and omit the joint effects of risk factors. Moreover, very few studies consider the accuracy and reliability of ML model evaluation metrics. Other issues include spatiotemporal correlations, causality, model transferability and heterogeneity. This paper concludes by providing suggestions on model selection and modification to address the identified issues and recommendations for future research. For example, employing advanced ML methods such as graph convolutional networks (GCN) to model spatiotemporal correlations; exploring innovative ways of applying ML methods; and leveraging new developments in ML (e.g. interpretable ML) to derive causal relationships and interpret modelling results.

Abstract 2280: Burden of lung cancer in Australia avoidable by modifications to lifestyle-related risk factors

Abstract Aim: The cancer burden avoidable through risk factor modification can be quantified using the Population Attributable Fraction (PAF) which combines estimates of the prevalence of the risk factor exposure in the population and the strength of the exposure-cancer association. PAF for risk factors causally related to cancer is most accurately estimated from cohort studies, using analytical approaches that account for death as a competing risk. We provide the first such estimates of the lifestyle-related avoidable lung cancer burden. Methods: Seven contemporary Australian cohort studies (N = 367,772), with comprehensive data on exposure to lifestyle-related risk factors, were linked to the Australian Cancer Database and National Death Index to identify lung cancers and deaths from any cause. The risk factors were harmonised across the studies and the data pooled. The strength of the exposure-cancer and exposure-death associations were estimated using a proportional hazards model, adjusting for age, sex and the other lifestyle exposures. Age- and sex-specific exposure prevalence was estimated from the Australian National Health Survey 2011-2012. These estimates were then combined to calculate the PAFs and their 95% confidence intervals (CI) using a newly developed method accounting for competing risk of death. Results: During the first 10-years follow-up, 2,030 incident lung cancers and 20,348 deaths were ascertained as first events. Smoking, physical inactivity and insufficient fruit intake were statistically significantly associated with both lung cancer incidence and death. Most of the lung cancer burden (PAF = 73%, CI = 69-76%) was attributable to current or past smoking, with 31% (CI = 27-35%) of the burden attributable to current smokers. Physical activity below Australian recommendations (< 150 minutes of moderate or 75 minutes of vigorous exercise per week) explained 14% (CI = 6-22%) and fruit intake below Australian recommendations (< 2 daily serves) explained 6% (CI = 0-12%) of the lung cancer burden. Each risk factor contributed more to the lung cancer burden in men than in women. Overall, 78% (CI = 74-81%) of the lung cancer burden was attributable to current or past smoking, physical inactivity and insufficient fruit intake. The combined contribution of modifiable risk factors (i.e. excluding past smoking) was 44% (CI = 37-50%). Based on projected Australian cancer incidence rates, this would translate to 59,000 avoidable lung cancers in the next 10 years. Conclusions: According to these joint PAF estimates, the current lung cancer burden in Australia could be reduced by up to half in 10 years by quitting smoking and increasing physical activity and fruit intake to the recommended levels. As the combined effect of risk factors is rarely additive, knowing both individual and joint effects of risk factors on cancer burden is essential in planning cancer interventions and public health policies and predicting their potential impact. Citation Format: Maarit A. Laaksonen, Maria Arriaga, Peter Hull, Karen Canfell, Robert MacInnis, Emily Banks, Graham Giles, Paul Mitchell, Robert Cumming, Barbara-Ann Adelstein, Julie Byles, Dianna J. Magliano, Jonathan Shaw, Anne Taylor, Kay Price, Vasant Hirani, Claire M. Vajdic. Burden of lung cancer in Australia avoidable by modifications to lifestyle-related risk factors [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2017; 2017 Apr 1-5; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2017;77(13 Suppl):Abstract nr 2280. doi:10.1158/1538-7445.AM2017-2280

Individual and joint effects of risk factors for onset widespread pain and obesity - a population-based prospective cohort study.

Widespread musculoskeletal pain (WSP) and obesity frequently co-occur and may have shared risk factors. We aimed to investigate whether four dichotomized risk factors individually or jointly increase the risk for the onset of WSP and onset of obesity. Persons aged 34-76years in 2004 living in Ullensaker municipality, Norway, responded to questionnaires in 2004 and 2010 (n=1553). Using causal interaction analyses, we examined whether baseline obesity and WSP, poor sleep quality, mental distress and poor physical fitness jointly increased the risk of new onset WSP (≥3 pain sites leading to disability the last year) and new onset obesity (self-reported BMI ≥30kg/m(2) ) in persons without WSP (n=1270) or without obesity (n=1300) at baseline respectively. The mean (SD) age was 51 (12.1) years and 56% were female. The incidence of WSP and obesity were 9.1% and 5.4%. Mental distress and poor sleep quality individually and jointly with poor physical fitness increased WSP onset risk (relative excess risk due to interaction [RERI]=1.90, 95% CI, 0.39-3.42 and RERI=1.43, 95% CI, 0.10-2.76). Poor physical fitness individually increased the risk for new onset obesity, and baseline WSP and poor sleep quality jointly (RERI=1.87, 95% CI, 0.49-3.24). The presence of more risk factors was dose-dependently associated with onset WSP and to a lesser extent with onset obesity. The onset of WSP and the onset of obesity were results of joint effects of exposures. Poor physical fitness was a key covariate in increasing the risk for both conditions. WHAT DOES THIS STUDY ADD?: In a general population, the new onset of widespread pain and new onset of obesity were results of joint effects of risk factors and particularly poor physical fitness. The study may aid in the identification of patients at risk of future disability.

Estimates of Global and Regional Premature Cardiovascular Mortality in 2025

United Nations member states have agreed to reduce premature cardiovascular disease (CVD) mortality 25% by 2025. Global CVD risk factor targets have been recommended. We produced estimates to show how selected risk factor reduction would affect CVD mortality for different regions and countries. We used mortality, risk factor, and relative risk data from the Global Burden of Disease, Risk Factors, and Injuries (GBD) 2013 study to project CVD mortality for 188 countries up to the year 2025. We disaggregated observed CVD mortality in 1990 and 2013 into deaths attributable and unattributable to hypertension, tobacco smoking, diabetes mellitus, and obesity using an age- and sex-specific population-attributable fraction. Risk factors were projected to 2025 assuming that current trends continue. Counterfactual scenarios were then constructed reflecting CVD premature mortality if United Nations risk factor targets are achieved in the year 2025, adjusting for joint effects of risk factors. We estimate 7.8 million premature CVD deaths in 2025 if current risk factor trends continue. Premature CVD deaths would be reduced to 5.7 million if these risk factors targets are achieved as a result of a 26% reduction for men and a 23% reduction for women in the global risk of premature CVD death. Globally, decreasing the prevalence of hypertension accounted for the largest risk reduction, followed by a reduction in tobacco smoking for men and obesity for women, but these results varied by region. The impact of meeting all risk factor targets on CVD mortality varied widely by region and sex. The United Nations target of a 25% reduction in premature CVD mortality by the year 2025 appears achievable for some countries, but more aggressive risk factor targets may be required if all regions are to reach this goal. Without these reductions in CVD risk factors, many countries will see no change or even an increase in premature CVD mortality.

Impact of Hepatitis C Seropositivity on the Risk of Coronary Heart Disease Events

Chronic infections have been shown to enhance atherogenicity. However, the association between chronic hepatitis C (HCV) and coronary heart disease (CHD) remains controversial. We examined the risk for CHD events in patients with HCV with an emphasis on the risk of CHD events with active infection. We conducted a retrospective cohort study using the Enterprise Data Warehouse at the University of Arkansas for Medical Sciences. HCV positive and negative patients were identified based on serology and incident CHD events were studied. Patient characteristics at entry were compared either by analysis of variance/F-test (continuous variables) or by a Chi-squared test (categorical variables). The joint effect of risk factors for incident CHD was evaluated using logistic regression. A total of 8,251 HCV antibody positive, 1,434 HCV RNA positive and 14,799 HCV negative patients were identified. HCV antibody and RNA positive patients had a higher incidence of hypertension, diabetes mellitus, obesity and chronic lung disease, but lower serum cholesterol levels compared to HCV negative patients (p< 0.001). HCV seropositive patients had a higher incidence of CHD events when compared to controls (4.9% vs. 3.2%, p<0.001). In the HCV cohort, patients with detectable HCV RNA had a significantly higher incidence of CHD events when compared to patients who were only HCV antibody positive with no detectable RNA (5.9% vs. 4.7%, p=0.04). In multivariate logistic regression analysis, both HCV antibody positivity (OR 1.32, 95% CI 1.09-1.60, p<0.001) and HCV RNA positivity (OR 1.59, 95% CI 1.13-2.26, p<0.001) were independent risk factors for incident CHD events. In conclusion, there is increased incidence of CHD events in HCV seropositive patients and the incidence is much higher in patients with detectable HCV RNA when compared to patients with remote infection who are only antibody positive. Lipid profile does not appear to be a good cardiovascular risk stratification tool in HVC patients.

Joint Effect of Modifiable Risk Factors on the Risk of Aneurysmal Subarachnoid Hemorrhage

The joint effect of risk factors on the risk of aneurysmal SAH (aSAH) has been studied sparsely. We examined the potential synergism between cigarette smoking, hypertension, and regular alcohol consumption and the risk of aSAH in a prospective, population-based cohort of participants from the Nord-Trøndelag Health Study and the Tromsø Study in Norway. Interaction was assessed on additive and multiplicative scales. We identified 122 cases of aSAH over 977 895 person-years of follow-up. Interaction was observed between current smoking and hypertension on the additive scale, (relative excess risk because of interaction, 6.40; 95% CI, 0.88-11.92, adjusted for sex and age). We found no significant interaction between hypertension and regular alcohol consumption or current cigarette smoking and regular alcohol consumption on the additive scale. No significant interaction was detected on the multiplicative scale. The joint effect of current smoking and hypertension on the risk of aSAH was stronger than was the sum of the independent effects of each factor. Persons at risk of aSAH should be advised of a markedly stronger risk for aSAH with the combination of current smoking and hypertension. In addition, the finding suggests that combining smoking cessation and blood pressure lowering may have an extra risk reduction effect on preventing aSAH.

Abstract 4484: A novel objective, image-based phenotypic assay reveals higher α-radiation-induced γ-H2AX level in peripheral blood lymphocytes as a risk predictor for lung cancer

Abstract Intermediate phenotypic biomarker measures the summary results of genetic variations and may enhance predictive power in cancer risk assessment. In this study, we developed a novel objective, image-based phenotypic assay to measure α-radiation induced DNA double strand breaks (DSBs) in peripheral blood lymphocytes (PBLs) and used it as a potential biomarker for lung cancer risk prediction. γ-H2AX formation in the chromatin flanking DNA DSBs is one of the most sensitive DNA damage markers. In an ongoing lung cancer case-control study, the baseline and α-radiation-induced γ-H2AX immunofluorescence intensities in the PBLs from 403 untreated lung cancer patients and 307 controls subjects were assessed, and the ratio of the induced γ-H2AX level to the baseline γ-H2AX level was used as a variable to evaluate the sensitivity to mutagen challenge. The mean γ-H2AX ratio was significantly higher in cases than controls (case vs. control: 1.46±0.14 vs. 1.41±0.12, p &amp;lt; 0.001). A significantly decreasing trend in γ-H2AX ratio with age was observed for both cases (p for trend =0.003) and controls (p for trend =0.011). Dichotomized at the median level of the γ-H2AX ratio in controls, individuals with high ratio exhibited a significantly increased risk of lung cancer (OR=2.33, 95%CI: 1.67-3.23). There was also a dose-response relationship with lung cancer risk in tertile and quartile analysis (p for trend &amp;lt; 0.001 for both). Analysis of the joint effects of risk factors revealed an apparently elevated risk of lung cancer for subjects with a higher γ-H2AX ratio who were ever smokers, compared with subjects with a low ratio who were never smokers (OR=12.68, 95%CI: 6.84-23.51, p &amp;lt; 0.001). This is the first study applying relative quantification of the γ-H2AX level in PBL to assess the risk for developing lung cancer. Susceptibility to DNA DSBs might help to identify individuals at high risk for this cancer. α-radiation-induced DNA damage as determined with the new γ-H2AX assay could be a novel phenotypic risk factor of lung cancer. Citation Format: {Authors}. {Abstract title} [abstract]. In: Proceedings of the 103rd Annual Meeting of the American Association for Cancer Research; 2012 Mar 31-Apr 4; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2012;72(8 Suppl):Abstract nr 4484. doi:1538-7445.AM2012-4484

Frequency of and predictors for patient-reported medical and medication errors in Switzerland

To analyse the frequency of and identify risk factors for patient-reported medical errors in Switzerland. The joint effect of risk factors on error-reporting probability was modelled for hypothetical patients. A representative population sample of Swiss citizens (n = 1306) was surveyed as part of the Commonwealth Fund's 2010 lnternational Survey of the General Public's Views of their Health Care System's Performance in Eleven Countries. Data on personal background, utilisation of health care, coordination of care problems and reported errors were assessed. Logistic regression analysis was conducted to identify risk factors for patients' reports of medical mistakes and medication errors. 11.4% of participants reported at least one error in their care in the previous two years (8% medical errors, 5.3% medication errors). Poor coordination of care experiences was frequent. 7.8% experienced that test results or medical records were not available, 17.2% received conflicting information from care providers and 11.5% reported that tests were ordered although they had been done before. Age (OR = 0.98, p = 0.014), poor health (OR = 2.95, p = 0.007), utilisation of emergency care (OR = 2.45, p = 0.003), inpatient-stay (OR = 2.31, p = 0.010) and poor care coordination (OR = 5.43, p <0.001) are important predictors for reporting error. For high utilisers of care that unify multiple risk factors the probability that errors are reported rises up to p = 0.8. Patient safety remains a major challenge for the Swiss health care system. Despite the health related and economic burden associated with it, the widespread experience of medical error in some subpopulations also has the potential to erode trust in the health care system as a whole.

The societal benefits of reducing six behavioural risk factors: an economic modelling study from Australia

BackgroundA large proportion of disease burden is attributed to behavioural risk factors. However, funding for public health programs in Australia remains limited. Government and non-government organisations are interested in the productivity effects on society from reducing chronic diseases. We aimed to estimate the potential health status and economic benefits to society following a feasible reduction in the prevalence of six behavioural risk factors: tobacco smoking; inadequate fruit and vegetable consumption; high risk alcohol consumption; high body mass index; physical inactivity; and intimate partner violence.MethodsSimulation models were developed for the 2008 Australian population. A realistic reduction in current risk factor prevalence using best available evidence with expert consensus was determined. Avoidable disease, deaths, Disability Adjusted Life Years (DALYs) and health sector costs were estimated. Productivity gains included workforce (friction cost method), household production and leisure time. Multivariable uncertainty analyses and correction for the joint effects of risk factors on health status were undertaken. Consistent methods and data sources were used.ResultsOver the lifetime of the 2008 Australian adult population, total opportunity cost savings of AUD2,334 million (95% Uncertainty Interval AUD1,395 to AUD3,347; 64% in the health sector) were found if feasible reductions in the risk factors were achieved. There would be 95,000 fewer DALYs (a reduction of about 3.6% in total DALYs for Australia); 161,000 less new cases of disease; 6,000 fewer deaths; a reduction of 5 million days in workforce absenteeism; and 529,000 increased days of leisure time.ConclusionsReductions in common behavioural risk factors may provide substantial benefits to society. For example, the total potential annual cost savings in the health sector represent approximately 2% of total annual health expenditure in Australia. Our findings contribute important new knowledge about productivity effects, including the potential for increased household and leisure activities, associated with chronic disease prevention. The selection of targets for risk factor prevalence reduction is an important policy decision and a useful approach for future analyses. Similar approaches could be applied in other countries if the data are available.

Examining interactions between risk factors for psychosis

For complex multifactorial diseases it seems likely that co-exposure to two risk factors will show a greater than additive relationship on disease risk. To test whether greater than additive relationships occur between risk factors for psychosis. A cohort study of 50 053 Swedish conscripts. Data on IQ, cannabis use, psychiatric diagnoses, disturbed behaviour and social relations assessed at age 18 were linked to admissions with any non-affective psychoses over a 27-year follow-up period. Statistical interactions between risk factors were examined under both additive and multiplicative models. There was some evidence of interaction for eight of the ten combinations of risk factors under additive models, but for only one combination under multiplicative models. Multiplicative models describe the joint effect of risk factors more adequately than additive ones do. However, the implications of finding interactions as observed here, or for most interactions reported to date, remain very limited.

Joint effects of hypertension, smoking, dyslipidemia and obesity and angiotensin-converting enzyme DD genotype on albuminuria in Taiwanese patients with type 2 diabetes mellitus

Objective We investigated the individual and joint effects of hypertension, smoking, dyslipidemia, and obesity and angiotensin-converting enzyme (ACE) DD genotype on albuminuria in Taiwanese type 2 diabetic patients. Designs and methods ACE genotypes were determined in 519 (287 men and 232 women) patients aged 58.5 (SD: 9.0) years. Among them, 240 had albuminuria (urinary albumin-to-creatinine ratio ≥ 30 μg/mg). Logistic regression was used to evaluate the individual and joint effects of risk factors and DD classified by two-by-four table. Results The adjusted odds ratios were significant for hypertension, smoking and obesity but not for DD and dyslipidemia in models evaluating individual effects. However, while analyzing the joint effects of DD and hypertension, smoking, dyslipidemia and obesity, the respective adjusted odds ratios were 3.253 (1.261–8.391), 3.016 (1.086–8.376), 2.385 (1.010–5.630) and 2.508 (1.117–5.631). Conclusion Hypertension, smoking, dyslipidemia and obesity jointly play an important role with DD genotype in mediating albuminuria.

Interplay between mutagen sensitivity and epidemiological factors in modulatinglung cancer risk

Few studies have assessed mutagen sensitivity and lung cancer (LC) risk associations in the context of multiple epidemiological risk factors. We evaluated mutagen sensitivity as a susceptibility marker and explored the interplay of the genetic marker and multiple epidemiologic risk factors in modulating LC risk. This largest case-control study included 977 newly diagnosed LC patients and 977 controls, matched by age, gender, ethnicity and smoking status. Cases exhibited significantly higher mutagen sensitivity than controls in bleomycin (0.76 vs. 0.62 breaks/cell, p < 0.001) and benzo[a]pyrene diol epoxide (BPDE) assays (0.70 vs. 0.61 breaks/cell, p < 0.001). Mutagen sensitivity also exhibited dose-response relationship with LC risk in quartile analysis (p for trend <0.001). In smokers, history of emphysema, absence of hay fever history, LC in first-degree relatives, belomycin sensitivity, and high pack-year were identified to be the top 5 significant risk factors in a stepwise logistic regression model (odds ratios (ORs) of 2.69, 1.91, 1.84, 1.73 and 1.67, respectively). Analyses of joint effects of risk factors showed that compared to the reference group, subjects with no exposure to any of the aforementioned risk factors, the ORs for exposure to 1, 2, 3, 4, 5 or more of the risk factors were 1.56, 2.39, 3.75, 6.74 and 17.39, respectively (p for trend <0.001). This study strongly supports mutagen sensitivity as a predisposition factor for LC and demonstrates the importance of assessing multiple risk factors to comprehensively assess LC risk. This new integrative approach should facilitate identification of high-risk subgroups and has important implications in LC prevention.

Joint effects of risk factors for stroke and transient ischemic attack in a German population

Single, modifiable risk factors for stroke have extensively been studied. In contrast, differences of their combined effects among stroke and transitoy ischemic attack (TIA) have been rarely investigated. The aim of the present study was to assess single and joint effects of risk factors on the incidence of stroke and TIA and to compare their magnitudes in a large population-based German cohort. Incident cases of stroke and TIA were identified among 25,538 participants (aged 35-65 at baseline) of the European Prospective Investigation into Cancer and Nutrition (EPIC)-Potsdam Study. Relative risks for stroke and TIA related to modifiable risk factors were estimated using Cox proportional hazard models. During 4.3 years of follow-up 100 stroke cases and 112 TIA cases occurred. Incidences of stroke and TIA were 91.7 and 102.7 per 100,000 person-years, respectively. Relative risks for ischemic stroke (RR 5.12, 95% CI 1.49-17.6, p for trend<0.0001) and for TIA (RR 3.08, 95% CI 1.00-9.44, p for trend<0.024) were highest among participants having 4 or 5 modifiable risk factors. 58.5% of ischemic strokes and 26.2% of TIA cases were attributable to the 5 risk factors hypertension, diabetes mellitus, high alcohol consumption, hyperlipidemia, and smoking. Our data indicate that classical risk factors may explain almost 60% of ischemic stroke but only one in four TIA cases. Analysing potential differences of known risk factors between ischemic stroke and TIAs and the identification of other determinants of ischemic attacks are important steps to better explain the burden of stroke.

A high-risk group for prostatism: a population-based epidemiological study in Korea.

To evaluate the effect of sociodemographic, dietary and physical factors on prostatism in Korean men aged 50 and over. A community-based cross-sectional epidemiological study was performed in Yonchon County, Korea. The Korean version of the International Prostate Symptom Score (IPSS) was used to assess the severity of prostatism. Data on occupation, marital status, education, smoking habits, alcohol intake, daily consumption of nutrients, body mass index, abdominal circumference, waist-to-hip ratio, blood pressure, serum glucose, cholesterol, triglyceride and high-density lipoprotein (HDL) levels were analysed. The age-adjusted relative risk of these factors was calculated for moderate to severe prostatism (IPSS > or = 8). A multivariate analysis of all significant factors was performed to examine the joint effect of risk factors. Of 514 subjects, 119 (23.2%) had moderate to severe prostatism, the risk for which was related to age and alcohol consumption; waist-to-hip ratio (which represents the degree of abdominal obesity) and the serum level of HDL showed a biphasic association with prostatism in the multivariate analysis. In addition to previously reported risk factors, these data suggest that there might be an association between the development of prostatism and abnormal lipid metabolism.

The joint effect of risk factors on endometrial cancer

The joint effect of risk factors on endometrial cancer was examined by applying general statistical models to the data of a hospital-based case-control study conducted in Copenhagen, Denmark. The analysis included 149 cases of histologically confirmed adenocarcinoma of the endometrium and 154 age-matched controls with cervical cancer. Information on risk factors derived from the medical records. Estrogen use and body mass were found to be the main predictors of endometrial cancer risk. In the model proposed, women who ever used non-contraceptive estrogens had a 10-fold increased risk irrespective of their weight and height. Among non-users of estrogen, the risk of endometrial cancer rose with increasing body mass, the largest showing a five-fold increased risk. These data provide further evidence of the significant role that excess estrogens play, whether exogenous from replacement therapy or endogenous from enhanced androgen conversion, in the etiology of endometrial cancer.