Unmasking Data’s Healing Powers

UCLA Anderson’s Healthcare Analytics Symposium, hosted by the Morrison Center for Marketing and Data Analytics and UCLA’s Master of Science in Business Analytics, delved deeply into the realm of health care data analytics by bringing together scholars and practitioners in discussions about the power of data and their potential uses in the health care space.

Adjunct Professor of Marketing Andres Terech (Ph.D. ’04), faculty director of the Morrison Center and a UCLA Anderson alumnus, noted several key takeaways from the conference’s industry experts panel, observing, “The health care sector presents a unique opportunity to do more than just sell products or boost profits. It’s a realm where the skills honed in data science can transform lives, shape the destinies of future leaders, and even save lives. Unmasking data’s healing powers lies not only in predicting health outcomes or optimizing operations, but also in driving meaningful behavior change and improving patient care.”

1) Leveraging Analytics to Drive Behavior Change in Health Care

In the realm of health care analytics, it’s not enough to merely predict health outcomes; the true value emerges when these predictions lead to meaningful behavior change. As Kayta Andresen, Cigna Group’s chief digital and analytics officer, put it, “In health care, just being able to predict what will happen isn’t very meaningful. If you can’t change someone’s behavior or have the next action be something someone adopts, that prediction loses its potential impact.” Furthermore, patient satisfaction plays a pivotal role in health care success, as dissatisfied patients are less likely to adhere to prescribed treatments. To navigate this evolving landscape effectively, health care providers must embrace analytic models and digital channels for testing and learning. These platforms provide a unique opportunity to tailor interventions, engage patients and measure their efficacy in real-time.

2) Transitioning from Fragmentation to Navigation to Personalization in Health Care Data

In today’s health care landscape, the data is fragmented. That means the patient experience is fragmented. The difficulty lies not just in the abundance of data but in its cleanliness and standardization. To address this, we must strike a delicate balance between maintaining the raw integrity of data and merging it with other organizations’ information, especially when it comes to preserving the outliers, which may hold the key to identifying rare diseases.

Achieving interoperability in health care data is a complex task. “Primary care providers, where most of the patients are seen, often operate on different systems and technologies that don’t seamlessly merge,” said Mike Thompson, vice president of enterprise data intelligence for Cedars-Sinai. This challenge extends to the integration of health data from more than 325,000 consumer-focused apps, the abundance of which signals the necessity for innovative solutions to personalize the patient experience amid this fragmented data landscape.

However, it’s important to acknowledge that the less glamorous aspects of analytics play a pivotal role in solving this puzzle. As Matthew Foster, director of technology strategy and innovation at Amgen, said, “The less sexy part of data — the data engineering, the data piping, the data interoperability — is key to solving this. It’s not the exciting part, and for most organizations it’s not the first place they want to invest.”

In essence, the transition from fragmentation to personalization in health care data is a journey that requires overcoming challenges related to data cleanliness, interoperability and the integration of diverse data sources. “It’s not just about having access to the data but also about utilizing it effectively to enhance patient care,” said Foster.

3) Navigating Data Usage and Tokenization in Health Care

Creating unique identifiers for de-identified patient data is crucial in understanding diverse populations and data acquisition strategies. There’s a pressing need for innovative technology to address these challenges, including a more nuanced consent engine. The traditional binary approach to data consent needs to give way to more sophisticated models that enable individuals to specify the terms under which they are willing to share their data. Meeting the growing complexity of these data-related endeavors is a formidable challenge that requires careful consideration of ethical and legal frameworks as we move toward a future where data becomes a valuable asset in a rapidly evolving health care landscape. Panelists also discussed the potential of blockchain technology in this context, envisioning how individuals can track their genome data and receive compensation if companies like Amgen use it for groundbreaking discoveries. This entrepreneurial approach to personal data is poised to reshape data-sharing dynamics.

4) Wearables Can Increase Health Equity, but Beware of Data Biases

Integrating wearables into health care has opened new frontiers, particularly in decentralized clinical trials. Shifting from the traditional clinical trial model that often requires patients to be present at a hospital, wearables allow for randomized clinical trials in more patient-friendly settings. This transformative approach empowers researchers to remotely monitor vital signs and relevant health metrics, ultimately unlocking innovative avenues to address health equity concerns. For instance, wearables enable reaching populations that may not have easy access to large hospital centers, such as those in rural communities. However, it’s essential to acknowledge potential biases in wearable data. People who choose to wear and engage with these devices tend to be more health-conscious and engaged in their well-being, potentially skewing the data toward healthier individuals. Recognizing this inherent bias is critical when interpreting wearable data and ensuring that health care interventions remain equitable and inclusive. In essence, wearables offer a promising tool for transforming health care analytics by promoting health equity, but they require careful consideration to mitigate bias and ensure fair representation in research.

5) The Imperative of Human Involvement in Generative AI

Generative AI represents a significant leap in the world of machine learning, ushering in a new era of creativity and infinite potential. While it promises unprecedented capabilities, it’s not without its challenges. Katya Andresen shared that Cigna references generative AI as “infinite interns” because of AI’s computational capability to accomplish many tasks. However, it comes with its quirks. AI systems can be moody and finicky, demanding precise instructions and problem breakdowns. They exhibit short-term memory loss if overloaded with information and are adept at producing plausible and convincing but incorrect answers when faced with uncertainty.

Navigating the world of generative AI, while enticing, requires caution. As the panelists agreed, “AI is better, faster, cheaper, good relief and low risk as long as there’s a human in the loop.” Currently, the use of data in diagnosis, coverage decisions and care decisions is in its early stages, focusing on experimentation to harness the capabilities of these AI “interns.” While the potential is vast, collaboration and checks and balances are crucial, as generative AI is still a relatively early and potentially risky endeavor.

The power of these tools is undeniable, but the expertise of individuals who understand the nuances of real-world situations is equally important. Generative AI should be viewed as a tool for augmenting human intelligence, not replacing it. The physician’s role, for example, remains paramount, with their medical degree and licenses on the line. While generative AI has the potential to revolutionize various fields, including health care, the presence of a human in the loop is the linchpin for success and ethical application.

6) Navigating Analytical Model Accuracy and Adoption

Achieving accuracy and adoption of analytical models in health care demands a nuanced understanding of challenges and the influence of cultural factors. A telling example comes from Cedars Sinai, where an AB testing approach compared expected discharge dates provided by physicians with those generated by an analytical model developed using extensive physician input from the database. The results were revealing: The model outperformed the physicians in accuracy. However, upon reviewing the findings, physicians disclosed their practice of intentionally providing more optimistic discharge dates to expedite patient release, highlighting the complex interplay between clinical judgment and analytical predictions.

Furthermore, an essential aspect of model adoption is the expectation of accuracy. Mike Thompson suggested that physicians might achieve an 80% accuracy rate, which is considered the gold standard because of the inherent complexity of medical decision making. However, even an 80% accuracy rate may not suffice when it comes to analytical models. Physicians often expect these models to approach 100% accuracy, setting a higher standard for their trust and endorsement.

In essence, fostering widespread acceptance of analytical models in health care necessitates navigating these nuances, crafting compelling data-driven narratives, and aligning with the patient-centered care culture. Physicians, driven by their commitment to patient well-being, are trained to check boxes indicating study completion. Thus, it’s imperative not only to demonstrate the accuracy of analytical models, but also to ensure that the researcher has “checked the boxes” during model development. This alignment with physicians’ mission of delivering optimal patient care is pivotal in gaining their trust and fostering model adoption.

7) The Essential Triad of Talent in Health Care Data Science

When searching for talent, Matthew Foster said, “There is an emerging three-legged unicorn.” The first leg entails the capability to comprehend and proficiently execute the intricacies of data science work. The second leg of this triad requires a deep reservoir of business acumen, an understanding of the overarching purpose behind data science initiatives, and the ability to align them with organizational goals. Finally, the third leg emphasizes the crucial skill of effective communication — the capacity to articulate the value story in a compelling narrative.

Individuals or teams with this triad of exceptional skills are pivotal for success in the evolving landscape of health care analytics.