OSLER inHealth Advisory Board Minutes
PCORI OSLER inHealth Advisory Board Meeting
Thursday, February 18, 2016
Vincent Carey, Yates Coley, Mary Cooke, Francesca Dominici, Todd Fojo, Patrick Heagerty, William Lewis, Martin Morgan, Desi Stringer, Zhenke Wu, Scott Zeger, Risha Zuckerman
Welcome and Introductions
Dr. Scott Zeger welcomed everyone and acknowledged the PCORI philosophy of having a diverse range of stakeholders on project boards.
Grant Progress and Discussion
Dr. Zeger introduced the co-investigators and administrative staff on the project. He gave background information on the project, the three case studies, and the software package in development. He shared that the role of the Board is to participate in discussions and advise on major choices.
Diagnosing and Evaluating Therapies for Depression
Dr. Zeger provided an overview of the case study and shared how the PHQ-9 data will be used to guide clinicians on a patient’s trajectory and best treatment options. The research group started developing the methods using data on schizophrenia because the PHQ-9 data from the National Network of Depression Centers was in the early stages of collection. Dr. Todd Fojo provided background on depression and how a patient’s state is currently scored, using the PHQ-9, to gauge severity. Over the duration of a patient’s treatment, multiple measurements are gathered, which allows for the plotting of a trajectory. This can then be compared to the trajectories of other patients, allowing for subsets of patients with similar trajectories to be identified.
Active Surveillance of Prostate Cancer
Dr. Yates Coley gave background information on what active surveillance is and why it may be chosen over curative treatments, i.e. surgery, radiation. The research team has data on PSA (measured every 6-12 months) and prostate biopsy (annually). A biopsy is scored on a Gleason scale with a score of 6/7 needed to be eligible for the active surveillance program. At Johns Hopkins, we have over 1,300 patients enrolled since 1995. The decision-making tool developed aims to predict the true state of the prostate, not biopsy results. This model can be used throughout a patient’s participation in the program to make educated decisions. The research team has talked with patients already to learn what they want to know, and how they want information to be presented, and what information they can take home with them. The team is currently working on an assessment plan for outcomes and attitudes for when the tool is put into the clinical workflow. The tool will be a part of the OSLER inHealth package.
Dr. William Lewis (patient stakeholder) commented that this information makes the decision to seek treatment or not more complicated than before because of the use of probabilities. A patient will need explanations of the implications of a Gleason 6 or 7 score and the selection of active surveillance versus curative treatments.
Diagnosing Childhood Pneumonia
Dr. Zhenke Wu shared how this case study aims to diagnose childhood pneumonia, which can be caused by about 30 pathogens. He shared the importance of clinicians knowing what is causing the pneumonia in order to treat it properly. Because some diagnostic tests can be invasive and inappropriate for young children, PERCH only does peripheral diagnostic measurements, i.e. x-ray, nasopharyngeal flora, etc. The model shows all the pathogens and if/where they appear in the specific tests. The model allows clinicians to see exactly what exists in the child to make proper treatment choices. Dr. Wu then walked the group through the results obtained through the model. The model can be applied using the following software: baker: Bayesian Analysis Kit for Etiology Research. The next step for the model is to make a version of it that allows for multiple infections to exist simultaneously.
Proposal for OSLER inHealth Package
Dr. Vincent Carey talked about the basic modeling strategy and what was proposed initially in the grant. He shared the open questions the research team is trying to answer: How to bring the model to clinicians and how to gather information directly from patients? How to bridge from the data stream to the statistician’s workbench? Dr. Carey reviewed the timeline of deliverables by semester for the duration of the project. He brought up concerns about the data collected in electronic medical records (EMR) and how they need to engage more with the EMR Informatics community to solve these complex problems.
Feedback was provided throughout the meeting, but more formal feedback was given from each member of the Board in a round table fashion, as facilitated by Dr. Francesca Dominici. A written summary was sent to Ms. Risha Zuckerman following the meeting.
Can some EMR features and data be extracted to enable more informative analyses of existing research databases (e.g. claims)?
What are the questions that can be uniquely addressed with the EMR data?
Potential outcome framework in the context of supervised learning for PC surveillance
What needs to be done before these tools can be widely used?
Internal standards need to be set for the validity of the proposed algorithms and the analyses of EMRs
What needs to be demonstrated for analyses to be clinically valid?
Ongoing validation necessary for the calibration of a prediction model
There are different degrees of validity for the case studies
Local calibration is necessary
Is there an integrated, concerted effort to develop or identify commonalities across tasks and share the infrastructure?
One thing that is lacking: understanding of evidence-based medicine
Doctors do not have the statistical fact base for informed choice
How will this help the patient become more involved in decisions?
How is this going to improve the ability of the patient to be involved in his health care?
How is this particular software going to be helpful?
Dr. Zeger then gave very brief replies to several questions that were raised:
- Each problem can be fit with one piece of software that will be ready in about six months.
- The research team expects that there are characteristics of models that can be validated.
- The research team is starting a project to get research quality data into the EMR.