HemOnc Today's PharmAnalysis
HemOnc Today's PharmAnalysis
September 27, 2019
3 min read

Algorithm could help leverage data for pediatric oncology research

You've successfully added to your alerts. You will receive an email when new content is published.

Click Here to Manage Email Alerts

We were unable to process your request. Please try again later. If you continue to have this issue please contact customerservice@slackinc.com.

Charles A Phillips, MD
Charles A. Phillips

Investigators have developed an algorithm that appeared to accurately identify children with leukemia and lymphoma who would be appropriate for future clinical studies.

“Accurately identifying patient cohorts is key to designing better research,” Charles A. Phillips, MD, pediatric oncologist at Children’s Hospital of Philadelphia, said in a press release. “Because not every patient in large data sets would be appropriate for a clinical study, having a tool to separate signals from the ‘noise’ will help researchers leverage data to design pragmatic, real-world studies in patients with a range of different cancers. For instance, we could better evaluate nausea medicines or detect factors that influence the rates of infections in patients with central line placements.”

HemOnc Today spoke with Phillips about what prompted this research, the results of the study and the potential clinical implications of the findings.

Question: What prompted this research?

Answer: As researchers and the field in general have started to explore pragmatic clinical trials and using real-world electronic health record data, more research networks are starting to incorporate data extracted directly from the EHR. Although this creates a lot of opportunities in terms of data access, it can bring with it a unique set of challenges. Another point is the EHRs can pick-up a lot of ‘noise,’ especially around diagnoses. For pediatric leukemia and lymphoma, there are a lot of false positives in the EHR and multiple reasons why we expect this would happen. Considering these points, we need to take steps to make the data usable. We set out to develop a relatively simple method to more accurately identify children with a hematologic cancer and a large EHR-derived database.

Q: How did you conduct the research?

A: We used three PEDSnet sites for this study — Children’s Hospital of Philadelphia, Children’s Hospital Colorado and Seattle Children’s Hospital. In the early stages, we tinkered with different versions of identification algorithms. The most accurate algorithms were a combination of two diagnoses within a 90-day window with chemotherapy administration that is consistent with a hematologic malignancy treatment plan and visits to a hematologist-oncologist. We applied this general algorithm at CHOP and used data from Children’s Hospital Colorado and Seattle Children’s Hospital to confirm this.

Q: What did you find?

A: Using this algorithm allowed us to accurately identify patients with a hematologic cancer. We also did a sensitivity analysis in which we compared the number of chemotherapy administrations (one vs. two) and found that requiring two chemotherapy administrations was more likely to give us a list of patients who were not coming to these institutions for things such as early-phase clinical trials. Instead, they were coming for a bone marrow transplant or an early-phase study, and were more likely to receive less traditional chemotherapy. What we took away from this was to have the most homogenous population possible, we would recommend two chemotherapy regimens as part of the filtering algorithm to identify patients because it gives the most accurate results.


Q: Were you surprised by what you found?

A: Initially, we were surprised by how many false positives were in the chart. This stressed to us the importance of the work.

Q: What are the clinical implications of the findings?

A: A lot of my research is focused on supportive care. For a lot of those things, such as nausea control, the exact subtype of cancer does not matter as much. We could identify patients with an algorithm and then look at those who received an antiemetic of interest and supportive care research. This algorithm is really good for trying to accurately capture patients with pediatric cancer who have a hematologic malignancy. If someone wanted a specific focus, this algorithm can still be useful in that it could reduce the amount of work that it takes to identify cases.

Q: Are there plans for additional research?

A: Next steps would include performing similar work for solid organ and neurologic malignancies. We would also like to be able to delineate patients with specific types of hematologic malignancies to provide more specificity for future research. Finally, we would like to utilize this data as a foundation for additional supportive care research that examines outcomes of interest detectable in the EHR, such as infections.

Q: Is there anything else that you would like to mention?

A: This work would not have been possible without the support of PEDSnet, Children’s Hospital Colorado and Seattle Children’s Hospital. There had not been a pediatric cancer-specific work in that network before, so it was exciting and fun to be able to do this. I am grateful to have wonderful collaborators to be a part of this. – by Jennifer Southall


Phillips CA, et al. Pediatr Blood Cancer. 2019;doi:10.1002/pbc.27876.

For more information:

Charles A. Phillips, MD, can be reached at Children’s Hospital of Philadelphia, Roberts Center for Pediatric Research, 2716 South St., 11th Floor, Office 11392, Philadelphia, PA 19146; email: phillipsc2@email.chop.edu.

Disclosure: Phillips reports no relevant financial disclosures.