Seasonally adjusted laboratory reference intervals to improve the performance of machine learning models for classification of cardiovascular diseases.

BACKGROUND: Variation in laboratory healthcare data due to seasonal changes is a widely accepted phenomenon. Seasonal variation is generally not systematically accounted for in healthcare settings. This study applies a newly developed adjustment method for seasonal variation to analyze the effect seasonality has on machine learning model classification of diagnoses. METHODS: Machine learning methods were trained and tested on ~ 22 million unique records from ~ 575,000 unique patients admitted to Danish hospitals. Four machine learning models (adaBoost, decision tree, neural net, and random forest) classifying 35 diseases of the circulatory system (ICD-10 diagnosis codes, chapter IX) were run before and after seasonal adjustment of 23 laboratory reference intervals (RIs). The effect of the adjustment was benchmarked via its contribution to machine learning models trained using hyperparameter optimization and assessed quantitatively using performance metrics (AUROC and AUPRC). RESULTS: Seasonally adjusted RIs significantly improved cardiovascular disease classification in 24 of the 35 tested cases when using neural net models. Features with the highest average feature importance (via SHAP explainability) across all disease models were sex, C- reactive protein, and estimated glomerular filtration. Classification of diseases of the vessels, such as thrombotic diseases and other atherosclerotic diseases consistently improved after seasonal adjustment. CONCLUSIONS: As data volumes increase and data-driven methods are becoming more advanced, it is essential to improve data quality at the pre-processing level. This study presents a method that makes it feasible to introduce seasonally adjusted RIs into the clinical research space in any disease domain. Seasonally adjusted RIs generally improve diagnoses classification and thus, ought to be considered and adjusted for in clinical decision support methods.

A deep learning algorithm to predict risk of pancreatic cancer from disease trajectories.

Pancreatic cancer is an aggressive disease that typically presents late with poor outcomes, indicating a pronounced need for early detection. In this study, we applied artificial intelligence methods to clinical data from 6 million patients (24,000 pancreatic cancer cases) in Denmark (Danish National Patient Registry (DNPR)) and from 3 million patients (3,900 cases) in the United States (US Veterans Affairs (US-VA)). We trained machine learning models on the sequence of disease codes in clinical histories and tested prediction of cancer occurrence within incremental time windows (CancerRiskNet). For cancer occurrence within 36 months, the performance of the best DNPR model has area under the receiver operating characteristic (AUROC) curve = 0.88 and decreases to AUROC (3m) = 0.83 when disease events within 3 months before cancer diagnosis are excluded from training, with an estimated relative risk of 59 for 1,000 highest-risk patients older than age 50 years. Cross-application of the Danish model to US-VA data had lower performance (AUROC = 0.71), and retraining was needed to improve performance (AUROC = 0.78, AUROC (3m) = 0.76). These results improve the ability to design realistic surveillance programs for patients at elevated risk, potentially benefiting lifespan and quality of life by early detection of this aggressive cancer.

Temporal patterns of multi-morbidity in 570157 ischemic heart disease patients: a nationwide cohort study.

BACKGROUND: Patients diagnosed with ischemic heart disease (IHD) are becoming increasingly multi-morbid, and studies designed to analyze the full spectrum are few. METHODS: Disease trajectories, defined as time-ordered series of diagnoses, were used to study the temporality of multi-morbidity. The main data source was The Danish National Patient Register (NPR) comprising 7,179,538 individuals in the period 1994-2018. Patients with a diagnosis code for IHD were included. Relative risks were used to quantify the strength of the association between diagnostic co-occurrences comprised of two diagnoses that were overrepresented in the same patients. Multiple linear regression models were then fitted to test for temporal associations among the diagnostic co-occurrences, termed length two disease trajectories. Length two disease trajectories were then used as basis for constructing disease trajectories of three diagnoses. RESULTS: In a cohort of 570,157 IHD disease patients, we identified 1447 length two disease trajectories and 4729 significant length three disease trajectories. These included 459 distinct diagnoses. Disease trajectories were dominated by chronic diseases and not by common, acute diseases such as pneumonia. The temporal association of atrial fibrillation (AF) and IHD differed in different IHD subpopulations. We found an association between osteoarthritis (OA) and heart failure (HF) among patients diagnosed with OA, IHD, and then HF only. CONCLUSIONS: The sequence of diagnoses is important in characterization of multi-morbidity in IHD patients as the disease trajectories. The study provides evidence that the timing of AF in IHD marks distinct IHD subpopulations; and secondly that the association between osteoarthritis and heart failure is dependent on IHD.

Cohort profile: Copenhagen Hospital Biobank - Cardiovascular Disease Cohort (CHB-CVDC): Construction of a large-scale genetic cohort to facilitate a better understanding of heart diseases.

PURPOSE: The aim of Copenhagen Hospital Biobank-Cardiovascular Disease Cohort (CHB-CVDC) is to establish a cohort that can accelerate our understanding of CVD initiation and progression by jointly studying genetics, diagnoses, treatments and risk factors. PARTICIPANTS: The CHB-CVDC is a large genomic cohort of patients with CVD. CHB-CVDC currently includes 96 308 patients. The cohort is part of CHB initiated in 2009 in the Capital Region of Denmark. CHB is continuously growing with ~40 000 samples/year. Patients in CHB were included in CHB-CVDC if they were above 18 years of age and assigned at least one cardiovascular diagnosis. Additionally, up-to 110 000 blood donors can be analysed jointly with CHB-CVDC. Linkage with the Danish National Health Registries, Electronic Patient Records, and Clinical Quality Databases allow up-to 41 years of medical history. All individuals are genotyped using the Infinium Global Screening Array from Illumina and imputed using a reference panel consisting of whole-genome sequence data from 8429 Danes along with 7146 samples from North-Western Europe. Currently, 39 539 of the patients are deceased. FINDINGS TO DATE: Here, we demonstrate the utility of the cohort by showing concordant effects between known variants and selected CVDs, that is, >93% concordance for coronary artery disease, atrial fibrillation, heart failure and cholesterol measurements and 85% concordance for hypertension. Furthermore, we evaluated multiple study designs and the validity of using Danish blood donors as part of CHB-CVDC. Lastly, CHB-CVDC has already made major contributions to studies of sick sinus syndrome and the role of phytosterols in development of atherosclerosis. FUTURE PLANS: In addition to genetics, electronic patient records, national socioeconomic and health registries extensively characterise each patient in CHB-CVDC and provides a promising framework for improved understanding of risk and protective variants. We aim to include other measurable biomarkers for example, proteins in CHB-CVDC making it a platform for multiomics cardiovascular studies.

Disease trajectory browser for exploring temporal, population-wide disease progression patterns in 7.2 million Danish patients.

We present the Danish Disease Trajectory Browser (DTB), a tool for exploring almost 25 years of data from the Danish National Patient Register. In the dataset comprising 7.2 million patients and 122 million admissions, users can identify diagnosis pairs with statistically significant directionality and combine them to linear disease trajectories. Users can search for one or more disease codes (ICD-10 classification) and explore disease progression patterns via an array of functionalities. For example, a set of linear trajectories can be merged into a disease trajectory network displaying the entire multimorbidity spectrum of a disease in a single connected graph. Using data from the Danish Register for Causes of Death mortality is also included. The tool is disease-agnostic across both rare and common diseases and is showcased by exploring multimorbidity in Down syndrome (ICD-10 code Q90) and hypertension (ICD-10 code I10). Finally, we show how search results can be customized and exported from the browser in a format of choice (i.e. JSON, PNG, JPEG and CSV).

TAILS N-terminomics and proteomics reveal complex regulation of proteolytic cleavage by O-glycosylation.

Proteolytic processing is an irreversible post-translational modification functioning as a ubiquitous regulator of cellular activity. Protease activity is tightly regulated via control of gene expression, enzyme and substrate compartmentalization, zymogen activation, enzyme inactivation, and substrate availability. Emerging evidence suggests that proteolysis can also be regulated by substrate glycosylation and that glycosylation of individual sites on a substrate can decrease or, in rare cases, increase its sensitivity to proteolysis. Here, we investigated the relationship between site-specific, mucin-type (or GalNAc-type) O-glycosylation and proteolytic cleavage of extracellular proteins. Using in silico analysis, we found that O-glycosylation and cleavage sites are significantly associated with each other. We then used a positional proteomic strategy, terminal amine isotopic labeling of substrates (TAILS), to map the in vivo cleavage sites in HepG2 SimpleCells with and without one of the key initiating GalNAc transferases, GalNAc-T2, and after treatment with exogenous matrix metalloproteinase 9 (MMP9) or neutrophil elastase. Surprisingly, we found that loss of GalNAc-T2 not only increased cleavage, but also decreased cleavage across a broad range of other substrates, including key regulators of the protease network. We also found altered processing of several central regulators of lipid homeostasis, including apolipoprotein B and the phospholipid transfer protein, providing new clues to the previously reported link between GALNT2 and lipid homeostasis. In summary, we show that loss of GalNAc-T2 O-glycosylation leads to a general decrease in cleavage and that GalNAc-T2 O-glycosylation affects key regulators of the cellular proteolytic network, including multiple members of the serpin family.