Research Methods

The Cardiovascular Disease and Outcomes Research Program uses several research methods with the data from the Heart Disease in Communities cohort.


The Heart Disease in Communities cohort represents the entire experience of a community, and thus is uniquely positioned for epidemiologic studies.

The cohort is representative of the population of the Upper Midwest. Patients were enrolled in the cohort if they provided authorization for the use of their medical record for research without any restriction on age, race and ethnicity, socioeconomic status, and insurance status.

In contrast, data from standard clinical trials of patients with heart disease aren't suitable for risk prediction relevant to real-life practice because of their inherent selection biases amplified by the persistent underenrollment of older adult patients in trials.

Other data from convenience samples, voluntary registries, claims data or multi-institutional collections of retrospective cases are all subject to notable biases, incomplete capture of exposures and outcomes, and limited follow-up.

Hence, for research on the outcomes of heart disease to be inferentially appropriate and clinically relevant, we must study unselected (all comers) community-dwelling people living with heart disease and composed of incident (not subject to incidence prevalence bias) cases. In addition, the Heart Disease in Communities cohort focuses on patients with heart disease, and thus addresses an important gap in the current portfolio of cardiovascular cohorts, which enrolled mostly people free of disease.

Randomized trials

Our randomized trials include:

Blood pressure monitoring study. The PCORnet Blood Pressure Home Monitoring (BP HOME) Study is a patient-level randomized controlled trial that is comparing the effectiveness of home blood pressure monitoring with versus without a linked smartphone application for helping patients with uncontrolled hypertension achieve a reduction in systolic blood pressure. The trial is being conducted within the National Patient-Centered Clinical Research Network (PCORnet), which supports a research network that enables distributed querying of electronic health record (EHR) data in a common data model. We're using data from the electronic health record, an online patient portal and the home BP monitor (in the smartphone-linked arm) to collect outcome data for six months after enrollment. The primary outcomes are reduction in systolic blood pressure by clinic measurements and a patient satisfaction outcome measured by the Net Promotor Score, derived from self-reported likelihood of recommending the device to a friend.

Aspirin dosing study. The ADAPTABLE study is a pragmatic clinical trial. Pragmatic trials are designed to reflect real-world care by recruiting broad populations of patients, embedding the trial into usual health care settings, and leveraging data from electronic health records to produce results readily applicable to patient care. ADAPTABLE, which stands for Aspirin Dosing: A Patient-Centric Trial Assessing Benefits and Long-Term Effectiveness, is embracing a new paradigm of patient engagement in clinical research, whereby patient partners work alongside researchers in all aspects of the trial, including designing the protocol, consent form, study portal and study materials.

Data science

The emerging relevance of data science in public health research has coincided with the massive explosion of information readily available for scientific inquiry. Advanced computational methods provide scalable solutions for extracting, integrating and analyzing complex data from a variety of sources.

Our research teams work with investigators to merge big data analytics with domain expertise in addressing the pressing research questions in cardiovascular disease today.

Next-generation phenotyping and electronic epidemiology. Electronic epidemiology (e-epidemiology) refers to the adoption of digital technology as part of the ongoing transformation of epidemiology.

Electronic health record-based approaches enable more nimble and near real-time epidemiologic investigations in larger populations, and EHR data optimally lend themselves to secondary analyses. An essential prerequisite to this work is establishing the validity, reliability and scalability of electronic tools.

Under the auspices of the eMERGE Network and within our epidemiology cohorts, electronic health record algorithms for heart failure and dementia have been developed for population research and can be readily applied to secondary analysis of existing data sets. These tools are critical to revealing informative data patterns in the EHR and driving novel research directions.

Machine learning. Our access to high-performance computing environments affords us the ability to use the latest machine learning algorithms in our big data applications. These methods provide our investigators with scalable data analysis solutions to uncover novel risk factors embedded in the electronic health record, to process complex imaging and text data sources, and to develop powerful risk-prediction algorithms for cardiovascular diseases and related outcomes.

Omics. We maintain a biobank of patients who have granted consent from cohorts with myocardial infarction and heart failure within the main Heart Disease in Communities cohort. Sample types include buffy coat, extracted DNA, serum and plasma. Prospective recruitment is ongoing for eligible patients with suspected heart disease for genetic studies with subsequent clinical validation of disease status.

Patient-provided data survey

We surveyed patients with heart failure in an 11-county region of Southeast Minnesota guided by the Chronic Care Model. Through a PCORI Clinical Data Research Network, we also surveyed patients with heart failure from three other health care centers around the United States.

The aim of this study is to understand the distribution of patient-centric factors, including social support, self-management and health literacy in these patients. We also have linked survey responses to data from the electronic health record, including demographics, comorbidities and short-term outcomes.