Cohort description for MADDEC: Mass data in detection and prevention of serious adverse events in cardiovascular disease

The risk for mortality and prevalence of comorbidities of patients treated for cardiovascular diseases are high. Several risk estimation algorithms based on traditionally obtainable clinical information have failed in recognition of patients at risk even when medical interventions would be available. Usually the poor performance of risk prediction algorithms is attributable to heterogeneity in risk factors related hazards between different populations, national health care systems and even hospitals. MADDEC is an ongoing research project focusing on the use of mass data in the development of accurate hospital-level risk prediction algorithms among patients treated for different cardiac conditions. The study population comprises all patients treated (and to be treated) in TAYS Heart Hospital (electronic health records of ~73.000 individuals from a ten-year period) with a special focus on high-risk patients such as patients admitted for myocardial infarction or undergoing major interventions such as cardiothoracic surgery (both ~700 patients annually). The goal is to combine all past, present and future clinical data between years 2007 and 2029.Hospital electronic patient records are combined with a database (KARDIO) designed specifically for research and quality control purposes updated daily by physicians. Additional phenotype information is acquired from bio-signal data from systems monitoring patients in hospital and from portable or mobile devices after discharge. Background and endpoint data of all previous and future hospital admissions, drug reimbursements and disability allowances are collected from national registries. Finally, mortality data will be monitored from national causes of death registry allowing also adjudication of different causes of death for more accurate endpoint definition. All data are integrated to a dedicated noSQL/SQL research database service. The technical aim is to develop and deploy beyond state-of-the-art signal analysis and machine learning methods for hospital-data driven risk analysis. The clinical aim is to develop easily applicable tools for patient-level risk estimation used in facilitation of clinical decision-making. These tools can be used for example in estimating risk of short term-mortality after myocardial infarction or before heavy invasive operations.