Big data and specialty drugs: A high-traffic intersection
April 21, 2020
This much we know: healthcare data is on a roll. Now is our chance to ensure the growing data stays accurate, accessible, and actionable.
Which sector expanded its data capacity the most between 2016 and 2018? The answer, according to a Dell Technologies survey, is healthcare.1 While the financial sector, telecommunications and IT boast a larger absolute quantity of data, healthcare data showed the highest rate of growth.
When you consider the volume of information processed in healthcare, this surge comes as no surprise. With new technology that makes it possible to translate physician notes, patient self-reports, and radiographic images into hard numbers, the byte count increases with each passing day.
In the high-stakes world of specialty drugs, this data boom represents an opportunity to capture real-world drug performance, use our funding dollars wisely, and ultimately match the right patients to the right drugs—but only if done right.
Getting creative
The quest for robust, actionable data has forced researchers and policymakers to grapple with difficult questions about data collection, privacy, and sharing. Adding further complexity to the picture, each province—and each institution, for that matter—has different regulations around data capture.
These hurdles have not prevented the sector from moving forward with new data initiatives, from pilot projects to publications, with each new success fostering the competence and confidence to reach still higher. No longer in its infancy, the data collection ecosystem has entered a more mature phase characterized by greater creativity and collaboration.
Alberta is leading the charge in data connectivity with the potential to generate real-world evidence (RWE). The province’s Oncology Outcomes Initiative,2 for example, serves as a model of data capture along the oncology product lifecycle. Manitoba has announced the upcoming launch of MindSet, a data platform that will link the province’s disparate pockets of clinical data.5 In a best-case scenario, academic, industry, and policy researchers could tap into MindSet’s rich data vein for their studies. When will Ontario and Quebec, which account for over half of Canada’s population, catch up? Encouraging efforts include the ConnectingOntario ClinicalViewer portal3 and the Quebec Health Record,4 which combine patient data from such sources as physician lab results, imaging reports and hospital visits, though these initiatives fall short of RWE generation.
South of the border, the health data economy has been busting out of its own sector. Earlier this year, in an unprecedented deal, hospitals granted access to patient information to Microsoft, IBM and Amazon.6 The agreement, which respected patient privacy laws, brought to light hospitals’ central role as data repositories and brokers. Similar deals may be coming to a hospital near you.
Looking still further ahead, a “data superhighway” that connects data from patient populations across our country no longer seems so difficult to imagine.
Real-world “magic”
A lot of the information being collected falls under the umbrella of real-world data (RWD). As defined by the FDA, RWD consists of “data relating to patient health status and/or the delivery of health care routinely collected from a variety of sources,” such as electronic health records, product and disease registries, and data generated by patients.7
So where does that leave data from randomized clinical trials (RCTs), the hallowed standard of medical evidence? As an all-star clinical trials group recently argued in a New England Journal of Medicine opinion piece, RWD can’t replace the “magic” of randomization, and only RTCs can be trusted to capture smaller treatment effects.8 Rather than supplanting RTCs, RWD can complement them by establishing the long-term efficacy and safety of a drug, including its effects in specific populations.8
RWD can even do some of the grunt work for clinical trials. For example, electronic medical records can help identify large numbers of patients eligible for a trial, avoiding the inefficiency of enrolling small numbers of patients at multiple sites around the world.8
Juicing the data
Canada’s specialty drug space draws on several sources of RWD, with patient support programs (PSPs) a notable addition over the past couple of years. Exemplifying this trend was a recent analysis using data from Taiho Canada’s PSP for tifluridine/tipiracil, a treatment for metastatic colon cancer.19 The researchers’ conclusion: the rapid rate of enrolment in the PSP reflects a great clinical need, and prior chemotherapy is a factor in treatment discontinuation.9
As noted earlier, efforts to link the disparate sources of RWD have been ramping up. For instance, some PSPs have been working to integrate laboratory services into their data sets, in view of improving workflow processes such as scheduling blood tests and communicating results to physicians.
At the same time, more and more data from patient registries and questionnaires has been filling gaps left by clinical research. The Canadian Kidney Cancer Information System helped researchers assess quality indicators for patients undergoing renal carcinoma surgery, while the Canadian Melanoma Research Network Patient Registry shed light on the real-world effects of new therapies for the disease.10 In another instance, a survey completed by over 1,700 patients revealed a high level of satisfaction with the infliximab infusion experience, significantly exceeding patients’ baseline expectations.11
The data is also becoming more structured. Until recently, data analytics has depended on the manual extraction of clinical variables from patient charts and research databases—a laborious and costly exercise.9 Natural language processing (NLP) technology is changing all that. The technology can convert narrative chart reports into research-grade data sets. A Toronto-based data service provider is pairing this technology with AI methodologies to extract variables from clinical texts and turning them into row-column datasets.12 Applying this process to the University Health Network in Toronto, the company successfully aggregated more than 1,400 records for lung cancer patients.12 In a similar effort based in Alberta, NLP technology extracted machine-readable data—with 98.4% accuracy—from chart reports of multiple myeloma patients, thus helping to build a cost-effective infrastructure for advancing cancer care.13
All told, the specialty medicine sector is not only producing more data, but beginning to use it creatively to generate insights that drive access and clinical decisions. To serve our patients with complex medical needs, we need to do more of the same. A lot more.
References
1. Dell Technologies Global Data Protection Index.
2. Karim S. From real-world data to real-world patients: the O2 initiative (PowerPoint presentation). https://www.cspscanada.org/wp-content/uploads/Karim-O2-Program-Description-FINAL-CSPS-RWE-Workshop-12032019-updated-11302019.pdf
3. ConnectingOntario ClinicalViewer. eHealth Ontario. https://www.ehealthontario.on.ca/en/for-healthcare-professionals/connectingontario
4. Québec Health Record, Québec. http://bit.ly/2xxORxO
5. MindSet data platform. http://mindsetmb.ca/
6. Evans M. Wall Street Journal, January 20, 2020. https://www.wsj.com/articles/hospitals-give-tech-giants-access-to-detailed-medical-records-11579516200
7. Real-world evidence. FDA. https://www.fda.gov/science-research/science-and-research-special-topics/real-world-evidence
8. Phend C. MedPage Today, Feb. 12, 2020. https://www.medpagetoday.com/publichealthpolicy/clinicaltrials/8485
9. Samawi HH. Curr Oncol 2019;26:319-329.
10. Cheema PK, Kuruvilla S. Current Oncology 2019; 26. https://current-oncology.com/index.php/oncology/article/view/5151
11. Jones J et al. Manag Care 2017; 26:41. https://www.ncbi.nlm.nih.gov/pubmed/28273042
12. Leibtag A. Real-world data curation to transform medical investigation. Pentavere Research Group.
19. H.H. Samawi et al. Current Oncology 2019; 26. https://current-oncology.com/index.php/oncology/article/view/5107/4121