Machine learning models of healthcare expenditures predicting mortality: A cohort study of spousal bereaved Danish individuals.

BACKGROUND: The ability to accurately predict survival in older adults is crucial as it guides clinical decision making. The added value of using health care usage for predicting mortality remains unexplored. The aim of this study was to investigate if temporal patterns of healthcare expenditures, can improve the predictive performance for mortality, in spousal bereaved older adults, next to other widely used sociodemographic variables. METHODS: This is a population-based cohort study of 48,944 Danish citizens 65 years of age and older suffering bereavement within 2013-2016. Individuals were followed from date of spousal loss until death from all causes or 31st of December 2016, whichever came first. Healthcare expenditures were available on weekly basis for each person during the follow-up and used as predictors for mortality risk in Extreme Gradient Boosting models. The extent to which medical spending trajectories improved mortality predictions compared to models with sociodemographics, was assessed with respect to discrimination (AUC), overall prediction error (Brier score), calibration, and clinical benefit (decision curve analysis). RESULTS: The AUC of age and sex for mortality the year after spousal loss was 70.8% [95% CI 68.8, 72.8]. The addition of sociodemographic variables led to an increase of AUC ranging from 0.9% to 3.1% but did not significantly reduce the overall prediction error. The AUC of the model combining the variables above plus medical spending usage was 80.8% [79.3, 82.4] also exhibiting smaller Brier score and better calibration. Overall, patterns of healthcare expenditures improved mortality predictions the most, also exhibiting the highest clinical benefit among the rest of the models. CONCLUSION: Temporal patterns of medical spending have the potential to significantly improve our assessment on who is at high risk of dying after suffering spousal loss. The proposed methodology can assist in a more efficient risk profiling and prognosis of bereaved individuals.

Predicting mortality risk after a fall in older adults using health care spending patterns: a population-based cohort study.

OBJECTIVE: To develop a prognostic model of 1-year mortality for individuals aged 65+ presenting at the emergency department (ED) with a fall based on health care spending patterns to guide clinical decision-making. DESIGN: Population-based cohort study (n = 35,997) included with a fall in 2013 and followed 1 year. METHODS: Health care spending indicators (dynamical indicators of resilience, DIORs) 2 years before admission were evaluated as potential predictors, along with age, sex and other clinical and sociodemographic covariates. Multivariable logistic regression models were developed and internally validated (10-fold cross-validation). Performance was assessed via discrimination (area under the receiver operating characteristic curve, AUC), Brier scores, calibration and decision curve analysis. RESULTS: The AUC of age and sex for mortality was 72.5% [95% confidence interval 71.8 to 73.2]. The best model included age, sex, number of medications and health care spending DIORs. It exhibited high discrimination (AUC: 81.1 [80.5 to 81.6]), good calibration and potential clinical benefit for various threshold probabilities. Overall, health care spending patterns improved predictive accuracy the most while also exhibiting superior performance and clinical benefit. CONCLUSIONS: Patterns of health care spending have the potential to significantly improve assessments on who is at high risk of dying following admission to the ED with a fall. The proposed methodology can assist in predicting the prognosis of fallers, emphasising the added predictive value of longitudinal health-related information next to clinical and sociodemographic predictors.

Danish Pathology Life Course (PATHOLIFE) cohort: a register based cohort extending upon a national tissue biobank.

PURPOSE: The Danish Pathology Life Course (PATHOLIFE) cohort was established to facilitate epidemiological research relating histological and cytological features extracted from patient tissue specimens to the rich life course histories, including both prior and future register data, of the entire Danish population. Research results may increase quality of diagnosis, prognosis and stratification of patient subtypes, possibly identifying novel routes of treatment. PARTICIPANTS: All Danish residents from 1 January 1986 to 31 December 2019, totalling 8 593 421 individuals. FINDINGS TO DATE: We provide an overview of the subpopulation of Danish residents who have had a tissue specimen investigated within the Danish healthcare system, including both the primary sector and hospitals. We demonstrate heterogeneity in sociodemographic and prognostic factors between the general Danish population and the above mentioned subpopulation, and also between the general Danish population and subpopulations of patients with tissue specimens from selected anatomical sites. Results demonstrate the potential of the PATHOLIFE cohort for integrating many different factors into identification and selection of the most valuable tissue blocks for studies of specific diseases and their progression. Broadly, we find that living with a partner, having higher education and income associates with having a biopsy overall. However, this association varies across different tissue and patient types, which also display differences in time-to-death and causes of death. FUTURE PLANS: The PATHOLIFE cohort may be used to study specified patient groups and link health related events from several national health registries, and to sample patient groups, for which stored tissue specimens are available for further research investigations. The PATHOLIFE cohort thereby provides a unique opportunity to prospectively follow people that were characterised and sampled in the past.

Cohort profile: the MUNICH Preterm and Term Clinical study (MUNICH-PreTCl), a neonatal birth cohort with focus on prenatal and postnatal determinants of infant and childhood morbidity.

PURPOSE: The MUNICH Preterm and Term Clinical (MUNICH-PreTCl) birth cohort was established to uncover pathological processes contributing to infant/childhood morbidity and mortality. We collected comprehensive medical information of healthy and sick newborns and their families, together with infant blood samples for proteomic analysis. MUNICH-PreTCl aims to identify mechanism-based biomarkers in infant health and disease to deliver more precise diagnostic and predictive information for disease prevention. We particularly focused on risk factors for pregnancy complications, family history of genetically influenced health conditions such as diabetes and paediatric long-term health-all to be further monitored and correlated with proteomics data in the future. PARTICIPANTS: Newborns and their parents were recruited from the Perinatal Center at the LMU University Hospital, Munich, between February 2017 and June 2019. Infants without congenital anomalies, delivered at 23-41 weeks of gestation, were eligible. FINDINGS: Findings to date concern the clinical data and extensive personal patient information. A total of 662 infants were recruited, 44% were female (36% in preterm, 46% in term). 90% of approached families agreed to participate. Neonates were grouped according to gestational age: extremely preterm (<28 weeks, N=28), very preterm (28 to <32 weeks, N=36), late preterm (32 to <37 weeks, N=97) and term infants (>37+0 weeks, N=501). We collected over 450 data points per child-parent set, (family history, demographics, pregnancy, birth and daily follow-ups throughout hospitalisation) and 841 blood samples longitudinally. The completion rates for medical examinations and blood samples were 100% and 95% for the questionnaire. FUTURE PLANS: The correlation of large clinical datasets with proteomic phenotypes, together with the use of medical registries, will enable future investigations aiming to decipher mechanisms of disorders in a systems biology perspective. TRIAL REGISTRATION NUMBER: DRKS (00024189); Pre-results.

Biomarker evaluation under imperfect nested case-control design.

The nested case-control (NCC) design has been widely adopted as a cost-effective sampling design for biomarker research. Under the NCC design, markers are only measured for the NCC subcohort consisting of all cases and a fraction of the controls selected randomly from the matched risk sets of the cases. Robust methods for evaluating prediction performance of risk models have been derived under the inverse probability weighting framework. The probabilities of samples being included in the NCC cohort can be calculated based on the study design ``a previous study'' or estimated non-parametrically ``a previous study''. Neither strategy works well due to model mis-specification and the curse of dimensionality in practical settings where the sampling does not entirely follow the study design or depends on many factors. In this paper, we propose an alternative strategy to estimate the sampling probabilities based on a varying coefficient model, which attains a balance between robustness and the curse of dimensionality. The complex correlation structure induced by repeated finite risk set sampling makes the standard resampling procedure for variance estimation fail. We propose a perturbation resampling procedure that provides valid interval estimation for the proposed estimators. Simulation studies show that the proposed method performs well in finite samples. We apply the proposed method to the Nurses' Health Study II to develop and evaluate prediction models using clinical biomarkers for cardiovascular risk.

Referral Patterns for Patients with Nonalcoholic Fatty Liver Disease.

The incidence of nonalcoholic fatty liver disease (NAFLD) is rapidly increasing. This study evaluates the referral pattern of patients with NAFLD. A cohort study evaluating all patients with NAFLD referred to a single Gastroenterology Department from January 2017 to June 2020. Electronic patient referral letters were reviewed, and patients with NAFLD were diagnosed using standardized tests as part of a prospective cohort study. Predictors of nonalcoholic steatohepatitis (NASH) with significant (≥F2) fibrosis were evaluated in logistic regression analyses. In total, 323 (18.6%) of 1735 patients referred to the Gastro Unit during the study period were diagnosed with NAFLD. Patients were referred from general practitioners (62.5%) or other hospital departments (37.5%). Most referral letters included information suggesting a possible diagnosis of NAFLD (patient history, blood tests, or diagnostic imaging) or used the nonspecific general diagnosis suspected disease (Z.038). Out of 110 patients referred for a liver biopsy, 71 (22%) had NASH with significant fibrosis (F2 n = 39, F3 n = 19, F4 n = 13). Thirty-nine of these patients were referred from the primary sector. A logistic regression analysis (adjusted for age and gender) including all 323 patients showed that type 2 diabetes was the only significant independent predictor of NASH with fibrosis.

Pericardial, But Not Hepatic, Fat by CT Is Associated With CV Outcomes and Structure: The Multi-Ethnic Study of Atherosclerosis.

OBJECTIVES: The study sought to determine the associations between local (pericardial) fat and incident cardiovascular disease (CVD) events and cardiac remodeling independent of markers of overall adiposity. BACKGROUND: The impact of pericardial fat-a local fat depot encasing the heart-on myocardial function and long-term CV prognosis independent of systemic consequences of adiposity or hepatic fat is an area of active debate. METHODS: We studied 4,234 participants enrolled in the MESA (Multi-Ethnic Study of Atherosclerosis) study with concomitant cardiac magnetic resonance imaging and computed tomography (CT) measurements for pericardial fat volume and hepatic attenuation (a measure of liver fat). Poisson and Cox regression were used to estimate the annualized risk of incident hard atherosclerotic CVD (ASCVD), all-cause death, heart failure, all-cause CVD, hard coronary heart disease, and stroke as a function of pericardial and hepatic fat. Generalized additive models were used to assess the association between cardiac magnetic resonance indices of left ventricular (LV) structure and function and pericardial fat. Models were adjusted for relevant clinical, demographic, and cardiometabolic covariates. RESULTS: MESA study participants with higher pericardial and hepatic fat were more likely to be older, were more frequently men, and had a higher prevalence of cardiometabolic risk factors (including dysglycemia, dyslipidemia, hypertension), as well as adiposity-associated inflammation. Over a median 12.2-year follow-up (interquartile range: 11.6 to 12.8 years), pericardial fat was associated with a higher rate of incident hard ASCVD (standardized hazard ratio: 1.22; 95% confidence interval: 1.10 to 1.35; p = 0.0001). Hepatic fat by CT was not significantly associated with hard ASCVD (standardized hazard ratio: 0.96; 95% confidence interval: 0.86 to 1.08; p = 0.52). Higher pericardial fat was associated with greater indexed LV mass (37.8 g/m2.7 vs. 33.9 g/m2.7, highest quartile vs. lowest quartile; p < 0.01), LV mass-to-volume ratio (1.2 vs. 1.1, highest quartile vs. lowest quartile; p < 0.01). In adjusted models, a higher pericardial fat volume was associated with greater LV mass (p < 0.0001) and concentricity (p < 0.0001). CONCLUSIONS: Pericardial fat is associated with poorer CVD prognosis and LV remodeling, independent of insulin resistance, inflammation, and CT measures of hepatic fat.

Kernel machine testing for risk prediction with stratified case cohort studies.

Large assembled cohorts with banked biospecimens offer valuable opportunities to identify novel markers for risk prediction. When the outcome of interest is rare, an effective strategy to conserve limited biological resources while maintaining reasonable statistical power is the case cohort (CCH) sampling design, in which expensive markers are measured on a subset of cases and controls. However, the CCH design introduces significant analytical complexity due to outcome-dependent, finite-population sampling. Current methods for analyzing CCH studies focus primarily on the estimation of simple survival models with linear effects; testing and estimation procedures that can efficiently capture complex non-linear marker effects for CCH data remain elusive. In this article, we propose inverse probability weighted (IPW) variance component type tests for identifying important marker sets through a Cox proportional hazards kernel machine (CoxKM) regression framework previously considered for full cohort studies (Cai et al., 2011). The optimal choice of kernel, while vitally important to attain high power, is typically unknown for a given dataset. Thus, we also develop robust testing procedures that adaptively combine information from multiple kernels. The proposed IPW test statistics have complex null distributions that cannot easily be approximated explicitly. Furthermore, due to the correlation induced by CCH sampling, standard resampling methods such as the bootstrap fail to approximate the distribution correctly. We, therefore, propose a novel perturbation resampling scheme that can effectively recover the induced correlation structure. Results from extensive simulation studies suggest that the proposed IPW CoxKM testing procedures work well in finite samples. The proposed methods are further illustrated by application to a Danish CCH study of Apolipoprotein C-III markers on the risk of coronary heart disease.

Alcohol intake and risk of acute coronary syndrome and mortality in men and women with and without hypertension.

Although a light to moderate alcohol intake is associated with a lower risk of acute coronary syndrome (ACS), alcohol is also associated with risk of hypertension, which in turn is a strong risk factor of ACS. We examined whether middle-aged men and women with hypertension also benefit from a light to moderate alcohol intake in relation to risk of ACS and overall mortality. We used data from 57,053 men and women, aged 50-64, who participated in the Danish Diet, Cancer and Health study. Information on alcohol intake (amount and frequency) was reported by the participants. Hypertension status was assessed at baseline by combining blood pressure measurements and self-reports. During follow-up, 860 and 271 ACS events occurred among men and women. Irrespective of alcohol intake, participants with hypertension had a higher risk than participants with normal blood pressure. Alcohol intake was associated with a lower risk of ACS among participants both with and without hypertension and there was no evidence of interaction between alcohol intake and hypertension. Those who drank moderately had a lower mortality than abstainers and those who drank heavily; and for all levels of alcohol intake, participants with hypertension had a higher risk than participants with normal blood pressure. Results were similar for men and women. These findings indicate that a light to moderate alcohol intake has similar effects on the risk of ACS in men and women with and without hypertension.

EPIC-Heart: the cardiovascular component of a prospective study of nutritional, lifestyle and biological factors in 520,000 middle-aged participants from 10 European countries.

EPIC-Heart is the cardiovascular component of the European Prospective Investigation into Cancer and Nutrition (EPIC), a multi-centre prospective cohort study investigating the relationship between nutrition and major chronic disease outcomes. Its objective is to advance understanding about the separate and combined influences of lifestyle (especially dietary), environmental, metabolic and genetic factors in the development of cardiovascular diseases by making best possible use of the unusually informative database and biological samples in EPIC. Between 1992 and 2000, 519,978 participants (366,521 women and 153,457 men, mostly aged 35-70 years) in 23 centres in 10 European countries commenced follow-up for cause- specific mortality, cancer incidence and major cardiovascular morbidity. Dietary information was collected with quantitative questionnaires or semi-quantitative food frequency questionnaires, including a 24-h dietary recall sub-study to help calibrate the dietary measurements. Information was collected on physical activity, tobacco smoking, alcohol consumption, occupational history, socio-economic status, and history of previous illnesses. Anthropometric measurements and blood pressure recordings were made in the majority of participants. Blood samples were taken from 385,747 individuals, from which plasma, serum, red cells, and buffy coat fractions were separated and aliquoted for long-term storage. By 2004, an estimated 10,000 incident fatal and non-fatal coronary and stroke events had been recorded. The first cycle of EPIC-Heart analyses will assess associations of coronary mortality with several prominent dietary hypotheses and with established cardiovascular risk factors. Subsequent analyses will extend this approach to non-fatal cardiovascular outcomes and to further dietary, biochemical and genetic factors.