Imagine if your pharma company’s field force received timely alerts such as “6 patients predicted to be screened at Cleveland Clinic this week, 3 of which are expected to be positive on a certain biomarker.” This notice would allow them to adapt a planned call or initiate a new one in real time, bringing timely and insightful information directly to oncologists discussing treatment regimens and patient therapies.

The status quo

While most companies already invest in broadly proactive triggers intended to generate better provider target lists, some work on predictive triggers and alerts based on patient diagnosis, treatment initiation, line of therapy progression, and more. But what typically happens next with these alerts? Are they delivered to the field force in timely fashion and in the platform they already use? Can they quickly act on these alerts?

Implementation and delivery of traditional triggers and alerts requires some level of buy-in from the field force—the people who ultimately benefit from the information—to learn how to best apply these insights and turn them into actions. Unfortunately, more often than not, the field force experience isn’t at the center of design and implementation, and triggers/alerts are left ignored or received too late to be relevant as they’re in siloed systems and disjointed from the field force’s daily activities. 

Putting the end user at the center

If reimagined and implemented well, the predictive model outputs could inform a variety of market shaping activities, including efficient and timely targeting of HCPs and patients, effective use of personal and non-personal promotion, physician education and messaging, and supporting patient engagement.  Amid market saturation, the prevalence of smaller patient populations, and lack of differentiation, the ability to proactively find the right patients becomes that much important.

[The] most innovative oncology drugs are often approved as later line therapies to treat specific subpopulations with a cancer type. For these therapies, 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... In one project, the AI algorithm is six times more accurate than rules-based models for prediction, and nearly twice as accurate as linear regression models for the same condition.¹

Putting knowledge directly into the field force’s hands

Key applications of predictive analytics in Oncology targeting can identify undiagnosed patients by analyzing medical utilization patterns, ultimately leading to the identification of new patients eligible for treatment in specific disease areas. Identification also may be accomplished by projecting treated and diagnosed patients at the HCP level by drug or by tumor types across all major treatment channels, including pharmacy, clinic, and hospital. Other applications proactively predict forthcoming lines of therapy transition and identify target populations for treatment. For example, it is possible to predict and generate a target list of HCPs whose patients are failing on their first line therapy and are about to transition to a second line treatment.

Using a single or simple combination of clinical events, while useful, may not account for the complexity of clinical progression for all therapy initiations or switches. […] patient specific factors can be supplemented with HCP-driven variables. For example, historical HCP prescribing tendencies can help a model understand if a specific patient is treated by someone that is likely to prescribe a new therapy, or to transition them to a specific medication.² 

Even if not investing in predictive alerts, pharma companies should enable their field force with event-driven programs enabling a more targeted commercial approach. This could be significant when, for instance, an oncologist bills for a first-time visit from a specific cancer patient of interest (newly diagnosed); a patient starts drug or other therapy of interest; a new patient starts and/or is followed by one additional line of chemotherapy; a patient therapy is switched and/or specific lab or biomarker test is ordered.

Ultimately, these outputs help companies design targeted customer engagement activities around patient acquisition and/or retention that better aligns with patient journeys. 

Additionally, a key aspect of advanced analytics and machine learning application is supporting brand performance, especially for those underperforming ones. 

Brand teams have an understandably difficult time determining which performance levers will yield the largest impact to improve the brand’s overall performance. Which investments will have the greatest impact – and at what cost? […] Having identified the number of opportunity patients (those who are not on the brand, but could be), the brand team must next understand why these patients are not on the brand. Which commercial tactics – messaging, promotional execution (both in person and non-personal), and access – have not had the desired effect?³

From hypotheticals to implementation

Applying a full suite of possibilities may appear daunting, especially as some other aspects of truly using the digitally transformative platforms have not been established, but focus needs to be on the first step and first application which should drive speed to action: Allowing the customer-facing roles to seize the opportunities and increase the value of every interaction.

The best platforms provide commercial teams with access to integrated internal and external data to understand physician and patient behavior, preferences, and needs. […] Forward-thinking pharma companies are already using machine learning and advanced analytics to transform their commercial strategies, and they are seeing enhanced sales results.⁴

Ultimately, this means providing the insights in the field force’s contextual workflow maximizing the adoption and usefulness. This also means having ability to prioritize the most important alerts to be sent to the end users and supporting various personas (Reps, MSLs, KAMs, etc.). Finally, the solution should track the adoption rate of suggestions and impact on the business, as well as automatically improve and optimize recommendations over time.

Fundamentally, the ability to drive insights lies with the analytics and delivery technology that companies use. The oncology data-to-insights ecosystem is simple to define, but often complex to deliver due to number of data challenges, finding the right domain expertise to derive meaningful insight and evidence from the data at hand, and ultimately to deliver insights to the field in a timely manner. Working with a trusted partner that has strong domain expertise, employs innovative technology, delivers quickly and at scale, understands data, and has advanced analytics capabilities, can help make these use cases a reality.

¹ Predictive Analytics in Oncology: How Artificial Intelligence Drives Greater Precision for Pharma Brands

² Using artificial intelligence to predict disease progression: Targeting therapy transition with precise HCP and patient engagement

³ Optimizing Brand Performance: An Evidence-Based Approach Powered by Machine Learning

⁴ Leveraging artificial intelligence and machine learning to drive commercial success