The number of approved oncology therapies continues to rise, with 63 oncology drugs launching within the past five years. With so many approved drugs in the oncology marketplace and several new drug launches on the horizon, pharma companies are competing to identify and engage niche patient segments that could benefit from their drugs. However, the task of identifying the right patients at the right time can be extremely challenging due to narrow drug labels and small patient populations for different types of cancer.
A way to overcome this challenge is to develop an advanced analytics capability to predict key events along the oncology patient journey. Such a capability can enable pharma sales and marketing teams to dynamically target oncologists who are likely to treat patients eligible for their drug. The pharma patient support programs can use this capability to predict potential patient drop-offs from therapy and proactively support such patients. Pharma medical teams can also use this capability to identify the right clinical trial sites that are likely to receive patients who match the clinical trial eligibility criteria.
Setting up an oncology patient journey prediction capability is a four-step process:
1. Map out the oncology patient journey. The first key step is mapping out the oncology patient journey and prioritizing the key events that are critical for brand success. For example, for drugs indicated for first or later lines of therapy, the predictive modelling should focus on identifying patients who are likely to move to approved lines of therapy. Pharma companies can initiate the patient journey mapping exercise using NCCN guidelines, a recognized standard for the use of drugs and biologics, biomarker testing, imaging and radiation therapy for patients with cancer. They can refine the nuances of the patient journey by engaging their internal experts and by conducting research with oncology practices.
2. Acquire and integrate real-world patient data. The second key step is the acquisition and integration of real-world patient data sources such as claims, EMR, lab testing data and specialty pharmacy data. One or more of these sources can be used to support a predictive modelling capability. Pharma companies should develop a systematic data strategy to acquire (or create), curate and connect data sources that are most relevant to predicting the events of interest, such as patient progression from first- to second-line therapy. Several factors should be evaluated including data richness, granularity, capture rates, refresh rates, acquisition costs and connectivity with other data sources. Lack of a single, comprehensive, ready-to-use patient-level data source is a reality as of today, so some trade-offs across these dimensions are necessary. Some companies have successfully developed and operationalized predictive capabilities using only one or two data sources, and they are evaluating and integrating additional data sources on their road map.
3. Develop and test predictive algorithms. The third key step is developing and testing algorithms to predict key events of interest. Pharma companies can develop and test various algorithms using retrospective modelling on real-world data. They should pay special attention to algorithms used to determine the changes in line of therapy for different types of cancer. Several factors such as the evolution of treatment pathways, variability of the treatment process across different oncologists, and the gaps in underlying real-world data should be considered. Incorporating the inputs of sales, marketing and medical teams can help refine the algorithms and help with the required buy-in to adopt such predictive models. Companies are increasingly using machine learning algorithms combined with a robust feedback mechanism to develop and deploy predictive models that can evolve and get better over time.
4. Set up a scalable analytics platform to develop and disseminate predictions. The final step is setting up a scalable analytics platform that enables the development of predictive models and the dissemination of predictive insights and triggers to relevant stakeholders. Pharma companies should look for the following capabilities in an analytics platform: a robust patient data management module to ingest and integrate real-world patient data sources, an intelligent business rules engine that enables the creation of longitudinal patient cohorts with a variety of eligibility rules, an analytics work bench that enables the development and tuning of predictive models using machine learning and advanced analytics, and a business insights consumption module that enables the dissemination of predictive insights to sales, marketing, patient support and medical teams.
An oncology patient journey prediction capability can be developed and matured over time by smartly combining these key components. Pharma companies that have developed this capability are receiving glowing feedback from their field sales team regarding the impeccable timing of predictive insights. Their field sales teams are now getting timely access to busy oncologists who need to treat a relevant cancer patient. The ability to predict key events along the oncology patient journey can deliver transformative results for pharma companies launching new oncology drugs, and it can make the right drugs accessible to the right cancer patients at the right time.
