The OMOP Cup Grand Prize competition is now open!
Ensuring the safety of prescription drugs has emerged as a major global public health challenge. Several recent events have highlighted the need for new data sources and new algorithms to respond to the challenge. The new data sources have now become available but algorithmic progress has lagged. Here is your chance to develop a groundbreaking approach that can help protect the lives of millions of patients worldwide.
We are providing a large dataset that resembles observational data that can be extracted from insurance claims or electronic medical records. Your method will identify relationships in the data between drugs and medical outcomes (adverse events). The goal is to develop methods that correctly identify true drug-event associations while minimizing false positive findings. Methods will be evaluated by how closely they predict the known relationships that exist in the data.
The OMOP Cup has two competitions:
Each challenge offers a Grand Prize, to be awarded in March 2010. The Challenge 1 winner will receive $10,000. Challenge 2 winner will receive $5,000.
- Challenge 1 explores how well your method works when provided an entire dataset, so the goal is accurate classification of which drugs are associated with which outcomes.
- Challenge 2 evaluates the timeliness of detection of drug-event associations by having your methods run against data sequentially as it accumulates over time.
We recognize detecting adverse drug events in observational healthcare data can be a tough problem to solve, but we’re confident we can do it if we work together and make steady progress. So to keep up the momentum, we’re also offering Progress Prizes to the two methods with the best performance in each challenge by the end of November 2009. In addition, all competitors in the Top 10 will be recognized as OMOP Cup Award Winners and will be invited to submit their methods for publication on the OMOP website.
Total prize money is $20,000! All you have to do to win is come up with a high-performing method, and then share your method and describe what you’ve done with OMOP and the broader research community. We can all learn from your great ideas!
Start Date: September 23, 2009
Progress Prize: November 30, 2009
Grand Prize: March 31, 2010
Participants must register in order to enter the competition.
You can work as an individual or as a team. If you work as a team, each team should select one representative who will submit results and receive the prize. Each participant must register in order to compete. Participants may submit as many solutions as desired, but only one per 24 hour period. Submissions for the Progress Prize will not be accepted after 11:59 p.m. ET on November 30, 2009. The winners will be determined shortly afterwards.
OMOP reserves the right to request participants to resubmit results by executing the method on a different dataset if there are any perceived anomalies with a submission. Participants with promising results may be invited to participate on the OMOP methods development team to implement and test their methods within the OMOP data consortium.
Drug safety is a major public health concern in the United States. Recent high-profile drug safety issues have shed light on the ineffectiveness of the current drug monitoring system. Can intelligent utilization of observational health care data, such as administrative claims and electronic health records, help detect potential safety issues of drugs? Methods for detecting adverse drug events from observational data are immature. This competition intends to address this shortcoming.
Description of Methods Problem:
Methods are concerned with determining the relationship between pharmaceutical drugs (medications) and conditions or health outcomes (potential adverse events). Identification of such associations aims to generate hypotheses from observational data by identifying associations between drugs and conditions for which the relationships were previously unknown. This is an initial step of drug monitoring, where many drug-condition pairs are simultaneously explored to prioritize the drugs and outcomes that warrant further attention. The large number of possible combinations represents a great computational challenge.
For observational analyses, it is important to recognize that the goal is to provide information about associations between drugs and outcomes across a population of interest. The intended objective is not necessarily to conclusively ascertain whether a specific person had a particular outcome due to a particular drug, but instead to infer whether a population of individuals exposed to a product experiences more of the outcome than otherwise expected had they been unexposed. This population-based approach differs from the spontaneous adverse event reporting systems currently used, which considers each data record a specific self-report of a suspected causal association between a drug and an event.
For this competition, we are seeking methods that are computationally feasible, incorporate information of known drug-condition associations, and identify associations from observational data as accurately as possible.
OMOP has created a procedure for producing simulated datasets that model phenomena from real-world observational data (administrative claims and electronic health records) and conform to the OMOP common data model. Access to these documents requires separate registration at omop.fnih.org. In the simulated data, hypothetical persons have observation periods that contain records of fictitious drug exposure and condition occurrence. There are no visits, procedures, or other observations.
The Person table contains a unique identifier for each person along with demographic data. The Observation Period table contains the span of time for which data is captured in the database for a given patient. The Drug Exposure table contains periods of drug usage for each person. The Condition Occurrence table contains records of conditions and when they were observed. Simulated data tables will be made available as tab-delimited text files.
The simulated dataset contains ten million persons, more than 90 million drug exposures from 5000 unique drugs and more than 300 million condition occurrences from 4000 unique conditions over a span of 10 years. For a small subset of the 20 million possible drug-condition combinations, there exists a true causal association between the drug and the condition. We assume that the strength of the causal association remains constant over time. For the remaining combinations, no causal association exists.
The OMOP Methods Competition has two challenges based on the data. Participants are encouraged to compete for either or both challenges. Please go to the Challenges tab for details.
Dear Drugs-Forum readers: We are a small non-profit that runs one of the most read drug information & addiction help websites in the world. We serve over 4 million readers per month, and have costs like all popular websites: servers, hosting, licenses and software. To protect our independence we do not run ads. We take no government funds. We run on donations which average $25. If everyone reading this would donate $5 then this fund raiser would be done in an hour. If Drugs-Forum is useful to you, take one minute to keep it online another year by donating whatever you can today. Donations are currently not sufficient to pay our bills and keep the site up. Your help is most welcome. Thank you.