Linking patients to the best treatments is critical for the best healthcare outcomes. In oncology, where patient populations are sometimes small and treatments are indicated for multiple tumor types, there is a brief window of time to anticipate when a cancer patient is ready to start initial treatment or switch to the next-line therapy presented. This brief window of time is often difficult to predict—until now.
Advances in big data analytics and artificial intelligence make it possible to predict when patients may need to start initial treatment or move to the next line of treatment, as well as which identifying physicians are treating them. These predictive analytics eliminate the trial-and-error approach to marketing oncology products and empower sales and marketing teams to deliver more precise engagement with physicians. This combination of science and data has proven to deliver impressive results that help influence disease detection, treatment initiation, line of therapy transitions, and disease progression.
Today’s Oncology Marketplace
Oncology research has hit incredible highs in recent years. Advances in immunotherapies and targeted therapies have transformed cancer treatment, promising to provide better quality of life, greater longevity and, in some cases, full remission. These advances are drawing significant attention and investments as the number of approved cancer therapies continues to rise. In 2018, a record 15 new oncology therapeutics were launched, and more than half of them are delivered as an oral formulation, have an orphan indication, or include a predictive biomarker on their label.1
While these innovative therapies are changing the oncology landscape, they also come with rising price tags and a lot of competition. The top 38 cancer drugs account for 80% of total spending, and more than half of cancer drugs in aggregate account for only about 2% of oncology spending.2
With a lack of innovation in reimbursement decisions and an increase in competition, pharma companies can leverage predictive analytics to drive more precise and custom sales and marketing efforts. When used effectively, medical science liaisons, nurse educators and the sales teams can align their efforts to the specific needs of patients and raise awareness of the best treatment options. In turn, lifesaving treatments can be delivered to patients at their point of need while maximizing the sales performance of these innovative drugs.
Precision Marketing for Precision Medicine
Because oncology drugs target smaller, highly specific patient populations, sales teams are much smaller and more spread out. These sales reps may be responsible for an entire state or region and target oncologists who care for a variety of patients with a range of cancers and treatment needs. Sales efforts need to be precisely timed within the narrow window of need when a patient isn’t responding to a first-line treatment but hasn’t yet moved on. To get the best treatment to those patients, sales reps need predictive analytics that tell them which patients may need their treatment and when. They can use the insights generated to time their sales calls, customize messages to individual patient needs, and educate physicians about relevant treatment options.
Cancer patients have critical milestones throughout their disease journey that trigger treatment decisions. AI algorithms can be trained to analyze anonymized patient longitudinal healthcare data to identify these key triggers, such as doctor’s appointments, prescriptions, symptom reports, hospital visits, lab results, and patient profiles. These AI algorithms can be built to precisely predict disease detection, treatment initiation, line of therapy transitions, and disease progression. AI algorithms deliver several benefits, including:
- Detecting undiagnosed patients and identifying patients with the disease who are likely to benefit from a specific therapy.
- Generating lists of high-value healthcare professionals using real-time predictions of eligible patients in the area and the number of patients under their care.
- Building influence maps that show individual providers’ profiles, academic affiliations, industry connections, board certifications, areas of interest, and adoption behavior to inform more strategic marketing campaigns.
This proactive approach to oncology sales and marketing benefits the entire oncology community by ensuring patients have access to the best treatment options, saving physicians’ time, and increasing industry knowledge by providing information when it is most relevant and helping demonstrate the value of medicine for payers who need to justify the high-price tags.
Real-World Results and AI Adoption Trends
Healthcare companies and scientific studies illustrate the impact of using advanced algorithms and modeling approaches to optimize sales strategies in real time. For example, physicians treating patients with slowly growing tumor types often take a “watch and wait” approach, delaying first-line treatment until the disease starts actively progressing. Pharmaceutical companies promoting treatments for these cancers are challenged to align their sales strategies with these decision triggers.
In an IQVIA project focused on addressing this need by identifying patients nearing first-line therapy, the AI algorithm developed is six times more accurate than rules-based models for prediction and nearly twice as accurate as linear regression models for the same condition. Sales teams can use these insights to prioritize physicians who have a high number of potential patients, customize acquisition strategies for new patients, and personalize physician engagement before treatment decisions are made. Predictive alert reports help sales teams adapt their sales strategy and call lists in response to evolving needs.
Across the industry, journals document the growing trend to adopt AI for oncology. For example, in a paper published in the Proceedings of Machine Learning Research, the authors explain how they developed a wide and deep neural network incorporating prior medical knowledge to predict the treatment initiation for Waldenstrom Macroglobulinemia patients. When they compared the accuracy of predictions from their wide and deep neural network with various machine learning models, their approach achieved significantly better results in predicting the start of treatment.3
Similarly, in a paper accepted by the 2019 International Conference on Machine Learning, the authors describe conducting experiments using real-world data to predict the initiation of first-line treatment for Chronic Lymphocytic Leukemia patients and several other diseases. The results show their method can improve prediction over alternative machine learning models. These examples further highlight the use of AI in oncology and its evolving potential to change the way research is conducted.4
Leveraging AI in Oncology and Other Therapeutics Areas
Predictive analytics can have significant impact in many areas of drug development and commercialization. It is especially relevant for oncology, rare disease, and other therapeutic categories. When populations are small and dispersed and have limited treatment options, finding patients at the right time through the right physicians is vital to delivering the best treatment available.
When pharmaceutical companies can access the right data, technology, analytics, and industry expertise, it becomes possible to find patients in need and recommend the best course of treatment at the right time. This will drive tangible benefits for patients, empower physicians, and drive positive sales and healthcare outcomes for stakeholders across the oncology community.
Resources:
1. “Global Oncology Trends 2019.” May 2019. IQVIA Institute.
2. “Global Oncology Trends 2019.” May 2019. IQVIA Institute.
3. “Predicting Treatment Initiation for Waldenstrom Macroglobulinemia Patients via Deep Neural Network with Prior Medical Knowledge, Machine Learning for Healthcare, Proceedings of Machine Learning Research.” 00:1–18, 2019.
4. “Predicting Treatment Initiation from Clinical Time Series Data via Graph-Augmented Time-Sensitive Model,” Proceedings of the 36th International Conference on Machine Learning, Long Beach, June 2019.
