How AI in Healthcare Can Address the Explosion of EMR Data

Artificial intelligence (AI) is disrupting virtually every industry, and healthcare is no exception. But deployed correctly, AI in healthcare can move from mere disruptor to a true value-add. In fact, it could be the solution to healthcare’s biggest challenges. 

McKinsey’s recent report notes that changes in payer and revenue models, inflation, and labor shortages are accelerating creative innovations to solve these challenges. AI is one potential solution. 

In this article, we’re going to look at one specific healthcare challenge: the explosion of data as a result of EMR interoperability. We’ll walk through why this change is happening, why it’s critical to address it as soon as possible, and how AI solves the problem. 

The hidden risk & costs of EMR interoperability 

Electronic medical records (EMRs) are the digital versions of patients’ medical charts. These records include data like diagnoses, tests, medications, treatment plans, and more. Because EMRs house patient data in digital systems, it’s easier to keep the data up to date, track trends over time, and, perhaps most importantly, share patient data among other providers when needed. 

Typically, EMR data sharing takes place through an Application Programming Interface (API) exchange. One system queries another and the relevant data is shared in real time, making it simple and easy for providers to query patient data. This increases the interoperability not only of technology platforms, but various healthcare practices. Given that the average patient has 19 providers in their network of care, this interoperability is critical for achieving optimal patient outcomes. 

Despite these benefits, however, EMRs can increase risk and costs among providers. Here are a few ways this can happen.

Data warehousing costs

Querying and transferring data incurs costs, no matter which cloud platform you use (Snowflake or otherwise). The ease of use with which patient data can end up with your cloud costs getting out of control faster than you’d expect. 

According to Sage’s The New Healthcare C-Suite Agenda: 2024-2025 Report, 60% of healthcare executives say that EMR optimization is their top technological initiative, while 46% cite a system-wide objective of cost reduction. Achieving both objectives requires tackling the data warehousing problem by optimizing Snowflake performance and costs. 

Poor data quality

Whether we’re talking conflicting patient IDs, data quality issues on one side or another, different definitions of data types, coding errors, etc., data sharing carries the risk of error or corruption. Correcting decisions made because of bad data can be costly and a significant time investment. 

What’s more, poor data governance can equally put your data quality at risk. After all, if you put good data into a bad process, you’ll end up with bad data. 

Privacy & security risks

Although healthcare technology vendors are typically very rigorous about maintaining HIPAA compliance and keeping patient data secure, there’s always a risk of data being shared with a non-authenticated vendor. 

What’s more, as healthcare companies acquire more data, the likelihood of becoming a target of a cyberattack goes up. Currently, the cost of a data breach sits at over $4.88 million, according to IBM. So organizations must either risk a massive cost or invest in more cybersecurity measures. 

Data overwhelm and fatigue

Just because healthcare providers have access to more patient data doesn’t mean those data are helpful. In fact, when providers are overwhelmed with data, they’re more likely to ignore it. 

As such, there’s a serious need for solutions that contextualize, summarize, and help providers surface the most relevant insights from their data. This is necessary to turn an otherwise complex data set into information that improves patient outcomes. 

How AI provides a solution to growing warehousing costs in healthcare

While AI can help address all the problems listed above, in this article we’re going to focus exclusively on one: the explosion in data warehousing costs that can come from EMR interoperability. 

Because Keebo deals exclusively with Snowflake users, we’ll focus on health systems that use Snowflake as their primary data cloud. 

There are certainly some governance practices that can help limit the data available for querying. But at the end of the day, there’s only so much you can do before you start limiting the volume and speed at which insights become available to fellow providers, payers, and vendors. 

An alternative approach is to reduce cost by optimizing Snowflake performance. If you don’t have to burn so many credits to query your growing databases, you can mitigate the increase in costs.

Why AI is important for optimizing Snowflake

And now we turn to the question of AI in healthcare and its relevance to addressing the warehousing problem. Across most Snowflake users (not just in healthcare), the common approach to Snowflake is manual optimization. 

The impulse behind this makes sense: after all, you want to ensure you’re not over-optimizing and taking resources from critical warehouses. It’s easier—so it seems—to have an experienced engineer do this.

But it’s important to ask the question: is constantly adjusting Snowflake the best use of your engineers’ time? 

Because the real-time nature of cloud computing means that queries are going to fluctuate every minute of every hour of every day: 

• User count fluctuates (ideally, grows)
• Queries become more (or less) frequent and complex
• Your rapidly scaling platform requires additional cloud resources
• You outgrow your current warehouse space and need to provision more resources

If you try to manually adjust your warehouses yourself, it can easily become a full-time job. This means your engineers aren’t building new pipelines, keeping the data clean and governed, maintaining security, etc. 

How to optimize Snowflake costs through AI

At Keebo, we help our users deploy advanced AI algorithms that constantly monitor dozens of parameters, adjusting them dynamically to avoid over- or under-provisioning. 

Most importantly, these algorithms work constantly, saving you pennies here, and dollars there. Compound that over 24 hours, days, weeks, and months, and it can have an outsized (positive) effect on your budget. 

And if you have specific service-level agreements or other performance requirements you need, you can always put guardrails on the AI. So let’s say you want a maximum query latency of 360 seconds, you can ensure you’ll always have the resources to handle those requests. 

Final thoughts on the value of AI in healthcare

Although it’s been around for a while, the last few years have seen a major increase in interest and use of AI in healthcare. But AI doesn’t have to be scary. In fact, when deployed against critical challenges health systems face in 2024, it can be a major value-add.

For data teams specifically, AI-powered Snowflake optimization from Keebo can reduce your costs by 25% or more. To see how to make this happen for your health system, talk with one of our in-house experts today.