You have probably benefited from artificial intelligence (AI) by finding your next binge-worthy show using a Netflix recommendation or by telling Alexa to shuffle songs by Kelly Clarkson. But, using AI in healthcare fundraising has until now been the stuff of fantasy, or at the very least used by extremely early adopters.
Its use in fundraising has expanded as AI and machine learning have become more common. The good news is that it has huge potential benefits.
AI is the ability for machines to “learn, reason, and act for themselves.” Most AI you encounter is facilitated by machine learning. Machine learning refers to using algorithms to find patterns in large amounts of data and applying those patterns to predict future behaviors.
Most fundraising professionals recognize that individuals give because of how they feel about an organization, not just based on the fact that they are wealthy and have extra money in their pocket.
Experts have said that data is nice but that the experience an individual has with your organization is much more indicative of how they will give. Despite this conclusion, grateful patient screening processes have primarily used capacity to identify prospects because wealth is relatively straightforward to quantify.
Non-AI predictive models use only a handful of measurements and are static, meaning they are a snapshot of a particular moment in time and quickly become outdated. With machine learning, it is possible to analyze more data points on a more regular basis to more accurately predict which patients are likely to give.
The biggest hurdle to getting the project off the ground is educating everyone involved — fundraising staff, clinical staff, IT, and compliance — not just that the program was happening but why it was happening. Once everyone understands the purpose and the potential benefits, they were more likely to dedicate the time needed to get the program set up and to offer feedback on how to deploy it successfully.
Because the model relies on large amounts of patient data, working with IT and compliance to understand your data sources is important. Work with IT to build out the electronic medical record files that feed the model and rely on IT to refine the data queries over time.
Working with sensitive patient data requires being creative with data collection. When compliance staff don’t feel comfortable with sharing a particular data point used in the model, work to identify alternative data points that might serve the same purpose.
Because the predictive model is constantly evolving, getting ongoing feedback is key to its continued success. Recording physicians’ reaction to the patients the model suggests critical, even if the feedback is negative.
