How can life sciences brands leverage real-world data and AI-driven tools while safeguarding patient privacy?
The growing use of real-world data and AI-driven tools has enabled life sciences brands to streamline HCP engagement tactics and drive patient outcomes. However, this increasingly data-driven landscape presents a new challenge for pharma marketers - leveraging the power of real-world data (RWD) and AI while ensuring patient data privacy.
At the recent MM+M Transform conference, I had the opportunity to join Hemali Lakhani, SVP of Media at CMI Media Group, and Fran Lane, Senior Data Scientist and Privacy Expert at Datavant for a panel exploring what privacy risks pharma brands should take into consideration, and how one life science brand leveraged the power of AI- and RWD-driven models without compromising patient privacy.
Real World Data: What Patient Data is Being Utilized?
To fully optimize the potential of real-world data for improved physician targeting and engagement, life sciences brands need to fully understand patient attributes – not just predicting which patients are treatment-eligible but identifying their treating physician(s) and anticipating upcoming care milestones. We can therefore think about these RWD+AI models as comprising three key stages:
- Patient Prediction Model – the key patient eligibility criteria
- HCP Identification – which provider(s) the patient is likely to see
- Predicted/Identified HCPs – what these patient/provider interactions look like, and when they will occur
For this particular case study, we looked at an ulcerative colitis (UC) brand seeking strategies to identify physicians treating moderate-to-severe patients with uncontrolled disease that were likely to benefit from their therapy. By infusing their messaging across multiple channels – including the EHR – we provided a way to execute “just-in-time" communication with decision-making HCPs and help them deliver better care for patients.
With RWD and AI programs, brands can more easily determine the most impactful channel mix, forecast business impact and understand the projected revenue gain through channel and tactical investment – all while managing data-privacy concerns with care.
Utilizing an AI model driven by patients’ de-identified RWD to drive throughput and conversion, this brand was able to be present for each stage of the decision-making funnel:
- Diagnosis of UC: Triggered messaging by ICD code
- Initial treatment of therapy with 5-ASA, steroids, or other therapies: Triggered messaging based on ICD and NPI codes
- Escalating symptoms and continued treatments: General messages targeted by RWD+AI predictive model
- Recognition of escalation by HCP and selection of follow-on therapy: Specific messages targeted by RWD+AI predictive model
As patients moved through the funnel, this UC brand could prioritize the highest-value patients – those whose condition was escalated – while also engaging physicians to generate brand awareness and familiarity at earlier stages in the care journey. By leveraging this AI-driven model to project each patient’s likely course of treatment, this life sciences brand was able to apply specific patient attributes in order to target and engage their treating physicians – then refine their channel mix to help more patients start on therapy.
Media Planning and the Strategy Behind the Data
We live in an era where we can collect patient-level digital data in a safe, secure, and scalable way through precise and compliant “data puddles” – subsets of data lakes where patient information is stored, cleansed, and activated. Much of this data is deterministic, meaning that with unique identifiers, real-time tailored and omnichannel messaging, pharma brands’ campaigns are more efficient than ever.
To expand brand reach and find the most effective channel mix, the majority of pharma brands – and their agencies-of-record – rely on customized HCP target lists – such as those generated by the RWD+AI model described above. By leveraging highly secure matching technology, media planners are able to maximize reach against their custom list. This technology also allows brands to better understand the HCP affinity, the usage and digital activity of these key stakeholders, driving “next-best-action” marketing strategies, such as coordinating rep visits to in response to specific profile triggers.
As a results, these highly personalized programs allow brands and agencies an unprecedented level of insight and customization around:
- Client goals
- Competitive landscape
- Market size
- Lifecycle stage
- Message complexity
With omnichannel RWD and AI programs that include these considerations, and focus on next best actions, brands can more easily determine the most impactful channel mix, forecast business impact and understand the project revenue gain through channel and tactical investment – assuming they can manage data-privacy concerns with care.
Data Certification: How to Feel Comfortable Using Patient Data
When it comes to using patient data responsibly, education is crucial; patients want to know that their data safety is a top priority, and life science brands want to know they are following internal and external regulations. Life sciences brands can ensure HIPAA compliance and the de-identification of health data through two main methods:
- Safe Harbor – the removal of specific patient information that could be used alone or in combination with other data to identify that patient
- Expert determination – the ability to take use-case into account and minimize patient re-identification risk by having a trusted third-party statistician review the dataset holistically for privacy risk
Expert determination typically offers more options to life science brands, as it allows for case-by-case analysis of a given data set and seeks to quantify what percentage of patients in that data set are at a high risk of identification (or re-identification). If after that analysis, less than 1% are deemed high risk, the overall risk can be considered “very small”, as stipulated by HIPAA.
However, de-identification alone may not always be sufficient – especially when data is being used to train machine-learning (ML) / AI models. In those instances, additional factors that must be considered are:
- How is the data being ingested? Where is it stored?
- What ML / AI models are being used? What data are they outputting?
- Is the ML / AI output appended to the original data? Can individuals be re-identified?
- How will the output be used? Will third parties have access?
For this particular UC brand, it was critical to address these points and more through a personalized assessment and approach. In addition, the brand felt strongly that some of the data being used in the predictive model belonged to the patient – not the clinician. By taking the educational approach described above, being clear about what was being collected and used, and working in partnership with the brand, the project was able to reach a satisfactory solution – without compromising the model integrity or predictive power.
Leveraging Real-World Data and AI-Driven Tools with Care
As AI-powered tools continue to increase in usage and sophistication, they continue to drive the possibilities for dynamic, next-best-action engagement strategies. But without proper safeguards to protect patient privacy, life sciences brands may struggle to secure buy-in across their organization or limit themselves from reaching the full potential of these tools for predictive engagement. As this UC brand found, there is a viable path forward, one that begins with choosing partners that truly understand building omnichannel strategies that align with the care journey, the importance of highly curated, matched NPI lists, and how to properly handle de-identified RWD to maintain patient privacy.
Interested in exploring how your brand can take a similar AI-driven approach to HCP engagement and targeting? Book a meeting with our team today to learn more about our patent-pending technology that drives brand awareness and patient acquisition.