Better Health Through Better Data

Dartmouth Institute Researcher James O’Malley Receives Funding for a 3-Year Project Aimed at Enhancing Statistical Methods in Order to Improve the Quality of Health Information for Patients with Vascular Disease

As a Professor of Biostatistics at The Dartmouth Institute and the Department of Biomedical Data Science, James O’Malley spends much of his time thinking about how better statistical methods can be used to help improve patient care. Recently, for example, O’Malley and his research team consisting of Todd MacKenzie, Philip Goodney, Doug Staiger, Glyn Elwyn and Doug Hill were awarded over $1 million by the Patient-Centered Outcomes Research Institute (PCORI) to conduct research that could improve the quality of health information provided to patients needing to choose between different treatment options. With this particular project, they will be working to improve the quality of information available to patients with carotid artery disease facing decisions between various vascular surgery procedures and medical management.

“A physician could be providing the patient with the best information she has so that’s what the patient bases his health care decision on,” O’Malley says. “But that information might not be as accurate as it can be. That’s the first aim of this project to help patients get the most accurate information possible.”

To illustrate his point, O’Malley explains that a doctor may relate information about 5-year survivability rates to his patients with surgery (let’s say 70%) vs. non- surgical interventions, such as medication (let’s say 85%). However, because of limitations in the way the data is collected, the ‘real’ numbers may be closer to 80% and 75%, incenting a different decision.

“It seems like a small variance but that type of information can make a difference for patients making such an important decision about their health,” he adds.

Patient billing records and registries such as the Society for Vascular Surgery’s Vascular Quality Initiative (VQI) are currently among the most important sources of data on vascular patients’ treatment and health outcomes. The advantage of information gathered from these sources is they involve large numbers of patients and reflect “real-world complexities,” according to O’Malley.

However, the problem with gathering data from such sources is that there is no safeguard against what’s known as unmeasured confounding variables. Although the terminology may sound complex, what it refers to is simply factors that could affect both the treatment a patient receives and their subsequent health outcomes but that are not included in the registry or patient billing information. If ignored, these factors could impact patient outcome results.

The only way to overcome unmeasured confounding in the VQI and other patient registries is to find a variable, known as an instrument variable (IV), which predicts the treatment a patient receives but otherwise does not affect the outcome. The best IVs tend to have a random quality to them and are often not things you would automatically think of. For example, the ambulance company that picks up a patient in an acute care situation can have an impact on which treatment he or she receives but otherwise be innocuous.

“There’s a bit of an art to it,” O’Malley notes.

In order to provide the most precise information for patients, O’Malley and his team are developing and evaluating IV methodology for the Cox model, a model designed for analyzing the time until an outcome such as stroke or heart attack occurs. The most challenging statistical component of the research is accounting for some patients having unknown event times while simultaneously overcoming the unmeasured confounding problem.

“There is little research at the intersection of these two major areas in statistical methodology so there is a potential for widespread impact, which is really exciting,” O’Malley says.

The second but equally important goal of the project is to determine the best format for presenting risk information to patients. To achieve their goal, the team will develop and solicit feedback on a number of different ways of presenting risk information, including the use of pictures and tables depicting different statistical quantities, to patients suffering from carotid artery disease.

“Already we have been very fortunate to have worked with two patients with asymptomatic carotid stenosis (a known risk factor for stroke), and they provided us with invaluable feedback for the project,” O’Malley says, citing an example to illustrate the fact that how information is presented is often just as important as the information itself.

“Survivability has an accumulative effect,” he explains. “So when a patient is looking at their chances of surviving five years into the future, it’s important to understand that if you survive the first year, it’s like putting money in the bank and getting interest earned. In effect, just because you may have the same probability of surviving for 5 years under each treatment option does not mean that you are expected to live the same amount of time over the next 5 years under each option.”

In order to ensure that patients are at the center of their research, the two patients whom the team has already worked with will be joining a nine-member Patient Engagement and Governance Committee.

“Patients and their doctors need better evidence to guide choices among treatment options,” O’Malley adds. “By providing the tools to collect more accurate information and then determining how to present that information in a more effective way, we will enable patients to make better treatment decisions.”

*All proposed projects, including requested budgets and project periods, are approved subject to a programmatic and budget review by PCORI staff and the negotiation of a formal award contract.

Link to original article publication: