-
Notifications
You must be signed in to change notification settings - Fork 24
Outreach Research Subgroup Meeting Notes
a. OMOP Extension tutorial
i. Status on preparation for instructional training (pre-recording/video)
1. Group plans to have their instructional material ready by beginning of September. Following which and prior to the pre-recording there will be a dry run with the entire group of trainers and volunteers.
2. Andrew will organize the pre-recording.
ii. Confirm infrastructure pre-requisites
iii. Status on Oncology story and data points – Asieh
1. Christian will provide traps to insert in the data so we can demonstrate the need for the Oncology Extension.
2. Shilpa will take data points from examples provided by Asieh and build OMOP Oncology extension tables in Excel
3. Team will meet with IQVIA to pre-populate the OMOP database to execute the ATLAS and SQL queries.
4. Team will also work with IQVIA to prepare the synthea database as well as the NAACCR data points so Anastasios can run the Drug Regimen Extraction and Michael can execute the NAACCR ETL respectively.
a. AJCC -> Since AJCC has not agreed to provide us the license for v8, this issue is closed for now. The plan is to extract v8 from NCIt
b. CAP -> Status on CAP license -> Christian to reach out for license. Doesn’t sound like CAP is going to become a standard. If it’s going to be normalized to something else then it’s okay for it to stay the way it is.
c. Center for Cancer Data Harmonization (CCDH) -> Christian is working on putting the examples together on the challenges we addressed in the first meeting. Once we have this together for Melissa, she can schedule the next call with her terminology team to help us understand the challenges.
d. OncoKB licensing discussion – Genomic variant database. Christian is already registered as an academic user of OncoKB user, therefore we can use the OncoKB API for OHDSI research purposes. Sent a message to Denys and Violetta about the same.
a. MMRF -> Need to better understanding what the funding here entails
b. SEER -> Waiting to hear from Donna about funding; Leanne to check if there is funding looking at capturing disease episodes beyond disease first occurrence.
c. Joe grant due in Jan for funding the vocabulary effort
i. Talk to Joe and Scott and develop a funding proposal. Rimma will share the grant that Joe sent. Due in Jan. Would be great if we submit for this proposal.
ii. Collaborative grant between MSK, Nebraska and Odysseus to engage in standardization work
iii. Rimma is meeting Joe this week to talk about the details and she will try to find out from him why he was thinking about this grant for us.
a. Preparing for the symposium (BC Study). Asieh is helping us write a protocol based on OHDSI standards. Plan is to run the study against the OMOP Extension and see if there are any differences / more precision in the results.
b. The complementary part to getting data representation correct is doing the analysis in an optimal way using best practices. In general, we should be trying to implement analytical best practices that the OHDSI community has defined for observation research, but we don’t have one for a simple descriptive survival analysis. There are some interesting work people are doing and weighing in on how we should do for this.
c. Asieh to act as our liaison to other statisticians in the community to engage them and align practices. This coincides and advances the work she's trying to streamline on how we should be approaching this process.
5. Algorithm development project - need to identify cancer recurrence/progression for the Oncology model.
a. Funding status -> Need funding for algorithm work & to obtain curated (chart abstracted) data from NU or Tufts, a project protocol suitable to be submitted to an IRB will need to be drafted. Re-engage with Ruth as she is eager to make active use of OMOP and evidence that can be generated from OMOP.
b. Write up the requirements and have sign-off on the requirements. As a starting point, there is literature on comparing chart extraction to define recurrence and comparing to automated definitions in EHRs from which we can steal ideas.
c. We need to know what data sources we have available to explore what kinds of methods can be used. Are there people that have chart abstracted data available so we can begin to find recurrence events. If not, there are other approaches to just explore EHR data directly and identify patients that had a recurrence based on other signifies like hospice care, death etc.
d. The next steps -> MSK data has already been abstracted by focused researchers so there is validation so we could start using that. Sara will create a script and run it against MSK and compare it to existing data.
- Algorithm development project (Robert M) -> Deriving Progression/Recurrence/Response/and other Episodes from EHR data
What do we need:
There is a need to identify cancer recurrence for the Oncology model. In most current cancer research settings, cancer recurrence/progression/metastasis data is manually curated by humans reading notes, looking at imaging exams reports/actual images and clinic progress notes and inspecting discrete variables in an EHR. The algorithm needs to bring in NLP over text reports Or even machine learning over Dicom.
Need a tool that incorporates an inferential engine that runs on cancer EHR and writes out results such as -> For patient A, from this date to this date -> initial diagnosis, from this date to this date -> first line of treatment, from this date to this date -> remission, from this date to this date -> recurrence etc write this out in a table. The whole pathway of the cancer. Anyone that has the data will run this and get an abstraction of data in the pathway.
An OMOP instance that informs about the patients that had cancer diagnosis on this date is available thru tumor registry loading and we want the algorithm to tell us when they had subsequent episode from that first occurrence.
Why do we need this:
Recurrence is a critical endpoint for cancer patients and a key driver of the burden of disease yet information on recurrence and disease progression is not included in cancer registries. To answer the question, “Can rich clinical and administrative data be mined to identify recurrence events in an automated and scalable manner?” requires defining recurrence, identifying types of predictors that might be informative, sourcing the relevant data streams, and developing a prediction algorithm that has adequate predictive performance. The question of what constitutes adequate predictive performance is crucially linked to the way the predictions are ultimately going to be integrated into an existing registry infrastructure.
How can we do this:
Develop an algorithm to identify patients with cancer recurrence using some ground truth (Data points will emerge from the work. Need to identify which patient the cancer recurred thru chart abstraction or some pre-existing chart abstracted data that has already done the work to time or recurrence for a cancer. NU does not have this. Best to start from scratch and do the chart abstraction from ground up for one type of cancer).
The algorithm would run against an OMOP instance. Output insertions into an OMOP instance the cancer recurrence and time. Time intervals/eras. Use the cohorts identified in #1 to work on problem of assigning a date to the recurrence.
Next Steps:
To obtain curated (chart abstracted) data from NU or Tufts, a project protocol suitable to be submitted to an IRB will need to be drafted.
For funding, re-engage with Ruth as she is eager to make active use of OMOP and evidence that can be generated from OMOP.
From NCI’s perspective, Donna will help advise us on the rules to submit a proposal and find the appropriate team that she can help us connect with for the potential application related to recurrence work at SEER.
Robert will come up with requirements based on the above discussion and get a sign-off on the requirements.
Robert will speak with Makabi and to check if they have funding opportunities for this.
JCO CCI Special Edition Paper publication -> Rimma to setup meeting to discuss this further. Due in July.
- Background/Problem Statement (Challenges related to research across diff assets, there is no standard to get insight out of patients, review of different classification systems, how our Oncology harmonization efforts are going to be used by the OHDSI community to do open science research in accordance with what NIH- FAIR—Findable, Accessible, Interoperable, Reusable guidelines)
- Normalizing ICO-SNOMED
- Normalizing CAP in terms of classification
- Integrating NAACCR, challenges and next steps
- Data model for Oncology (Episodes and connection between cancer diagnosis and modifiers)
-
Long-term -> Any elements of CANMed that are valuable will be incorporated in Hemonc.org in the future. CANMed is the Drug part of OROT. It has the curation of cancer indication for organ. It has curation of 95% certainty of a cancer drug treatment, this is what HemOnc.org does not have. So far, we have manual curated drugs for active treatments.
-
Medium-term -> If Jeremy can provide a list of OncoTree mappings (Ontology system created by 3 institutions), Rimma can provide high level classification which can be mapped to SNOMED.
-
Short-term -> Donna curate the list of ATC/ Hemonc.org agents is the best we can do.
-
Hemonc.org’s disease indications mapping to SNOMED, Jeremy explained that indications are done at a much higher level than diagnoses. Therefore, it does not make sense to map to ICDO -> Do we need anything from Jeremy?
-
Per Christian -> The steps in sequential order are as below:
-
Get standard concepts for Genomic Aberrations (point mutations, insertions, deletions, and copy-number variations.)
-
Get standard concepts for Staging
-
Get standard concepts for Tumor Attributes
-
Map NAACR, CAP, AJCC into these
-
OROT -> Surgery and Radiotherapy old version. Need new version from Valentina.
-
Incorporate high group therapy mappings for Radiotherapy and Surgeries into OMOP vocabulary
-
Definitions of active treatment by CCC19 (Treatment is active) -> 3 months prior to COVID diagnosis on cancer-directed drugs, radiotherapy, surgery, or BMT
-
Definitions of Active/Non-Active Cancer -> Remission/No Evidence/Active Cancer Responding/ Active Cancer Stable/ Active Cancer Worsening/ Active Cancer Unknown
-
Inclusion of cancer diagnosis (Disease is active) -> 3 year back is too short, at least 5 years
- On-hold for now since Nebraska Lexicon is busy with COVID issues
Status on CAP license -> Christian to reach out for license. Has requested feedback from Mik and Michael on their assessment of mapping of histology.
- EU Oncology Workshop a. First goal with the EU Oncology Workshop is to orient the participants towards an introduction of OHDSI. Benefits of a common data model and vocabulary. How can the CDM be leveraged in the Oncology space. b. Further down the line we can prepare to talk about the comparison between registries used in EU/Korea and NAACCR/SEER. How much can be leveraged from the work that’s already been done in NAACCR. c. Prepare some use cases to show in the workshop -> National Cancer registries (EU/Korea/US) need to start to understand how surveillance epidemiology can benefit from the international contrast.
- Focused discussion with Donna a. Understand and define what we’re going to do with SEER, their analytical goals, the capacity to what OHDSI would bring to what SEER is able to do and extending the registry work that we’ve done in NAACCR to SEER databases and the strategy for that. Give Donna some sense of the capacity of linking between EHR and registry as well as tackling issues around recurrence within the OHDSI community.
- We’ve been taking individual institutions who have a mix of EHR and Tumor Registry data and converting their tumor registry to Oncology. SEER is the collection of all the submissions of all the institutions into one national database. By converting their nationalized database to the OMOP CDM, they could take their entire pool of cancer database and compare it to other national registries like Netherlands, Norway etc.
- Documentation a. NAACCR mapping to OMOP -> Rimma b. Derivation of episodes (Treatment Courses and Disease Episodes) -> Rimma c. Chemotherapy Regimen ->?? (Need assignment) d. SQL ETL section, Unit Testing -> Development Subgroup (Need assignment) # No code just references to the code e. Analysis section ->?? (Need assignment) f. Tutorial -> Slide deck that would go through the above sections, describe what the model is, what the vocabulary is and explain them in detail with the above. g. ETA for completion of documentation and slide deck is 3/6. Review with Chan to ensure his needs for the EU Oncology workshop are met.
- Add an effort to highlight/understand/discuss factors that distinguishes the OHDSI Oncology efforts from what other organizations (Trinetics/other efforts) are doing to get Oncology data to support large scale Oncology research. This could be a good review paper and opportunity for a student (informatics) to work on.
- Andrew, Christian and Rimma are invited to publish in the journal of cancer informatics. This is going to be done under the category of using classification systems. Deadline is in May 2020.
- Prep for AMIA a. In the context of standardizing to SNOMED the Nebraska protocols and talk about incorporating Nebraska mapping into OHDSI. Work being done at MSK to standardize pathology report. b. Submit a panel proposal on the topic of standardization of cancer data solutions. Present our solution and invite competitors (mCode, Gini – Jeremy, FDA - Mitra) to present theirs. Deadline to submit the panel proposal is March 11. We need to get other people as well to agree to participant in this panel discussion.
Meeting Date: 11/5/2019
Attendees: Michael Gurley m-gurley@northwestern.edu; Robert Miller rmiller1@tuftsmedicalcenter.org; Anastasios Siapos anastasios.siapos@iqvia.com; Reich, Christian CReich@us.imshealth.com; awilliams15@tuftsmedicalcenter.org; Rimma Belenkaya rimma.belenkaya@gmail.com; Leanne Goldstein goldstein.leanne@gmail.com;Shaadi Mehr mehrs@themmrf.org
- For Issue 178, Issue 153, Issue 154 -> Christian will meet with Andrew this week and come up with a plan and next steps on engagement efforts with the standards organizations. The plan is to reach out to them by the middle of November.
- Issue 155 -> Rimma will reach out to IMO again post recent meeting cancellation and try to setup another session with them. The goal is to see if they will do the vocabulary work for the NAACCR standard.
- Issue 180 -> This is lower priority compared to above issues. Meera is currently evaluating their model to see if it will work for other registries and genomic.
- Issue 192 -> This issue has a pre-requisite Issue 179. Rimma and Dima will meet to discuss Issue 179 in the coming weeks.
- Issue 182 -> Meeting has been scheduled for Monday 11/11 to discuss the 8th edition of the staging system.
Next steps:
- Andrew and Christian to meet this week or next to discuss a plan to engage with NAACCR, SEER, CAP and Nebraska Lexicon.
- Rimma and Dima to meet to discuss steps for the evaluation of Nebraska Lexicon
- Shilpa to create forum post for soliciting use cases for Oncology CDM and Drug Regimen Algorithm
- Shilpa will work with Christian and Andrew to solicit update on progress of data source outreach efforts
Meeting Date: 10/22/2019
Attendees: Michael Gurley m-gurley@northwestern.edu; Robert Miller rmiller1@tuftsmedicalcenter.org; Anastasios Siapos anastasios.siapos@iqvia.com; Reich, Christian CReich@us.imshealth.com; awilliams15@tuftsmedicalcenter.org; Rimma Belenkaya rimma.belenkaya@gmail.com
-
Outreach/Research - Andrew to reach out to SEER to understand the NAACCR-SEER relationship and how we can leverage them to map NAACCR to the standard vocabulary (Andrew to work with Eric Durman)
-
Outreach - Andrew to reach out to NAACCR in the next month and find out if they have any mapping plans and figure out if there is a possibility to piggy back on it.
-
CDM/Vocabulary - Evaluation of Nebraska Lexicon to identify gaps, propose a solution that will work for them (Possibly Michael).
-
Outreach/Research - Reach out to CAP and Campbells (Nebraska Lexicon), share our proposal with them (based on our evaluation in #3) to expedite their mapping efforts and formalize our relationships. (Rimma and Christian will drive this effort)
-
Outreach/Research - Rimma will do an evaluation of the MCOD model, look at their clinical and genomic data mode, identify what's useful for us. At the same time, we have to make friends with them, propose to execute their model with our use cases with retrospective data, developer a crosswalk for interoperability. (Rimma and Christian possibly to work with MCOD?)
-
Outreach/Research - Rimma will arrange a meeting with IMO (include Christian; 2nd week of November) to discuss our proposal with them and recruit them to do vocabulary mapping work for us.
Oncology Working Group Publications/Presentation
Data Model
- Cancer Models Representation
- EPISODE
- EPISODE_EVENT
- MEASUREMENT
- CONCEPT_NUMERIC
- Disease Episode Model
Vocabularies
OMOP Model
- Populating the OMOP Oncology Extension
- NAACCR Tumor Registry
- EHR and Claims