The mission of the Patient-Centered Outcomes Research Institute is to address questions about health care from the patients’ perspective, such as “What are my treatment options, and what are the benefits and harms of those options?” (Washington & Lipstein, 2011). This proposal is to make it easier for PCORI to advance its mission by improving computer-based tools to: learn from the experience of prior patients and to translate what has been learned into simple terms that patients and clinicians can use to improve their decisions, taking into their patients’ unique circumstances.
We propose to use a method of analysis called Bayes method, in which data on the frequency of a disease in a population is combined with data taken from an individual patient (for example, the result of a diagnostic test) to calculate the chance that the patient has the disease given his or her test result. Clinicians currently use Bayes method when screening patients for disease, but we believe the utility of this methodology extends far beyond its current use.
The recent revolution in information technology has unleashed a torrent of new types of health data, from DNA sequences to functional images of the brain. Further, the electronic health record captures every patient’s sequence of health measurements, diagnoses, and treatments. Our proposal will apply Bayes method so that these new types of health data can be used to support patients and their clinicians in decision making. This novel application of Bayes methods will allow us to make predictions about an individual patient’s health state and health trajectory by using information from both the patient and the larger population.
In addition to developing this new application of the Bayes method, we will facilitate its use by creating and disseminating a software package that will allow other researchers to apply this methodology. The software will be open-source and readily extensible to the R statistical package, and it will include the methodology developed as a part of this research as well as existing methods that facilitate individualized health prediction.
As part of this research, we will test the proposed methods and software on three case studies. Specifically, we will: (1) estimate the frequency with which various pathogens cause children’s pneumonia, and predict which pathogen is likely to be causing a particular child’s pneumonia given her or his clinical data, potentially reducing unnecessary use of antibiotics; (2) compare the estimated effects on survival and quality of life of active surveillance versus immediate treatment for prostate cancer patients with equivocal test results, reducing unnecessary surgeries; and (3) characterize the sources of variation in the types and severity of depression, helping establish the National Network of Depression Centers as a research tool for improving depression care.
In carrying out this research, we believe we will improve patient-centered decision making.
Coley RY, Fisher AJ, Mamawala M, et al., A Bayesian hierarchical model for prediction of latent health states from multiple data sources with application to active surveillance of prostate cancer. Biometrics (August 2016).
Wu Z, Deloria-Knoll M, Zeger SL, Nested partially latent class models for dependent binary data; estimating disease etiology. Biostatistics (2016).
Coley R, Zeger S, Mamawala M, Pienta K, et al., Prostate Cancer: Prediction of the Pathologic Gleason Score to Inform a Personalized Management Program for Prostate Cancer. European Urology (August 2016).