How Healthcare Back-Office Professionals Can Use DvSum CADDI and GenAI Powered Chat to Drive Data-Driven Results.

24 Jul, 2023 •

healthcare use cases

Let’s delve into the transformative potential of combining DvSum CADDI with GenAI Powered Chat to empower healthcare back-office professionals. In today’s rapidly evolving healthcare landscape, data-driven decision-making is paramount. Join us as we explore how these innovative technologies revolutionize the way healthcare back-office professionals harness data insights to drive tangible results and enhance operational efficiency. From streamlining workflows to optimizing patient care, discover how DvSum CADDI and GenAI Powered Chat pave the way for a new era of data-driven excellence in healthcare administration.

The Centers for Medicare & Medicaid Services (CMS) is driving the healthcare industry from a fee-based service to a value-based service.  That means instead of rewarding insurance companies and agencies for delivering more services, the focus is instead on member outcomes.  

Therefore, a key business objective of healthcare payors, commonly insurance companies, is to improve their ratings – STAR ratings and the net promotor score (NPS), which are both a reflection of providing superior member (patient) care as well as a key ingredient in driving successful member outcomes and thus the business successfully.  

To do so, fast and iterative data analysis around urgent questions is required. But there’s a bottleneck of questions and answers to help drive member satisfaction. 

Healthcare back-office professionals, such as executives, managers, product managers, care managers, support, business analysts, and data analysts, are overwhelmed with their own as well as others’ ad hoc questions about how to drive patient care for increased customer satisfaction.  

The challenge: “in-the-moment, iterative analysis” 

To say the challenge is simply “ad hoc analysis” still understates the problem. The ad hoc nature of analysis is also both iterative and “in-the-moment.” Examples of when “in-the-moment” analysis is required include conversations during meetings, phone calls with peers, an urgent follow-up request from an executive, and an ad hoc and urgent business requirement. And it’s compounded by the requirement of both in-depth business intelligence (BI) and data skills. The requirements can’t wait days for an ad hoc analysis by a BI or data team. 

BI reports often give a reasonable starting point for the investigation of an “in-the-moment” hypothesis. BI assets are developed by data analysis experts who understand the data well, and their efforts often solidly address the more obvious 80% of the reporting analysis requirements. But in many scenarios, the healthcare back-office professional needs to go much deeper and pivot to many other related questions – all very quickly and iteratively, and in-the-moment. 

Healthcare back-office professionals need to hypothesize and analyze on-the-fly. The payor cannot afford to offer real-time access to the centralized BI team. And the professional likely does not have the deep BI tools skills nor knowledge of the data and associated markers to get answers within the minutes or seconds required for a given question “in-the-moment.” 

Example healthcare insurance use cases 

The following are examples where healthcare back-office professionals need answers fast and in-the-moment. 

  1. Claim Denial Analysis 

A common scenario is a determination of whether a claim should be denied, and if not, does it warrant a change in policy? A payor’s employee should be able to get a quick answer for “how many claims were denied in past 12 months with this condition?” The attributes behind the claim might not be intuitively known, and thus a strong command of the data attributes is required.  

The analysis can become quite complex and quite confusing very quickly. A given medical condition might have hundreds of codes, with hundreds of procedures, and hundreds of member characteristics.  That can total hundreds more if not thousands of markers for an analyst to hypothesis, sift through, and query. Hundreds of these types of claims may happen every week.  

For example, for a mother seeking a prescribed stroller for her child who has cerebral palsy, should the stroller be covered? To do the research, an analyst might research the claim against policies for claims regarding “durable medical equipment (DME)”. Or should the marker be rather “prosthetic stroller?” In fact, it may be that the policy and code associated with the requirement has neither phrase. Based upon results with cited sources of data, it can then be determined if a particular claim should be denied or not – and whether the claim policy should be reviewed. 

  1. Transfer Rate Analysis 

A common question amongst executives is, “What was our transfer rate in 2022?” In other words, how many member enrollments changed from one plan to another in the past twelve months? 

This often requires involving a data analyst who is unfortunately not available at that time. Instead, it would be helpful for the executive to get a fast answer with the underlying cited sources to advance the topic of his or her question in the moment. 

  1. Member Coverage Analysis  

When a member or provider submits a claim for a pharmaceutical drug, it’s not always clear whether it’s covered under the member’s plan. A drug is sometimes covered under CMS (Medicare, Medicaid, etc.). Sometimes the payor covers it but only on specific plans. And sometimes it’s vice versa: Medicare covers the treatment, but the insurance company does not. To add to the confusion, drugs themselves have different names. There’s a commercial name, a generic drug name, and perhaps other names for it. All of this requires iterative research in the moment to understand what’s covered and under what policy, plan, or government agency.  

  1. Member Segmentation Risk Analysis 

It’s very common to ask questions in a meeting to advance and evolve a hypothesis. For example, a question may arise around Population Health Information analysis. Which patients are most risky that therefore require higher premiums? Or what further upfront information is required to determine potential coverage costs? 

To build a risk profile, the payor requires a longitudinal view of the member, i.e., a flattened data set describing the member’s markers. Analysis of the members and historical conditions, treatments, and success rates lets the payor discover which factors are required to understand filing claims and then build risk segmentation of members. From there, the payor can apply a machine learning (ML) model to existing patients to understand care risk and provide intervention. 

A care manager can also do real-time queries on key questions like “which members have had their annual wellness visit, or their flu shot.” Getting the answers to these questions closes gaps in care which have a direct correlation on improving both STAR ratings and the NPS score.   

  1. Compliance Check for Claim Appeals and Appeasement 

After a claim is denied and the appeal is initiated by either the provider or member, there are strict regulations about how complaints are dispositioned. A compliance report on this process must be submitted for an audit that tracks oral or written confirmation from the member that the claim is closed. The payor analyst needs to find closed cases that do not have an oral or written confirmation of closure; that would be non-compliant closure. There needs to be a fast way to determine which claims were closed incorrectly and then to determine if the closure reason and mechanism was simply buried and needs to be surfaced or if the closure was done incorrectly. A fast, iterative, in-the-moment analysis would help greatly. 

The solution: AI-powered chat that understands the data 

With DvSum CADDI any person, whether technical or not, can simply ask questions naturally and iteratively to their data. CADDI understands how a question might be asked and quickly responds. They don’t have to ask in a constrained way such as “Sort Open Claims by Date = 4Q22- 1Q23”. They just ask, “Which claims are still open the past two quarters?”  

In all situations, CADDI quickly gives both an answer and the results with the sources cited to confirm the reasoning behind the answers. 

Very importantly, CADDI provides ad hoc analysis that is iterative and in-the-moment, in context of:  

 1. No BI expertise required   

  1. For the businessperson asking the question, it means a BI report expert is not required to quickly navigate a report with hundreds of dimensions and filters.  
  1. For the BI and analytics team receiving the question, it means not having to be distracted from core work. 

2. No expertise in the schemas or data landscape required 

Even when it’s not clear which specific data elements are required to answer a question, CADDI can help. It’s not important how to ask, even when it’s particularly hard to understand which data elements amongst hundreds or thousands of potential ones. 

3. Advancing the conversation 

Even if the perfect answer and sources are not quickly available, it’s possible to exclude wrong answers and get important clues as to what might be required to achieve the right answers still quickly.  

Iterative analysis in the moment, saves time thus accelerating solutions. 

Conclusion 

DvSum securely combines OpenAI’s GPT-4 with a powerful underlying data infrastructure to query your healthcare data quickly, easily, and iteratively. This empowers the payor’s healthcare professionals to become more autonomous and productive “in-the-moment” and thus to drive payor satisfaction scores and the overall business. Contact us today to see how DvSum CADDI can help you. 

Share this post:

You may also like