Developing a data product in Healthcare: How to Start Navigating Conversations to Uncover the Business Question


November 2023

You’ve read the blog post "Tell me What You Want, What You Really, Really Want: How to Identify the Real Business Question" – and are now curious about (or struggling with) how to apply it to the healthcare setting.

Background 

In healthcare, we can bring value in many different ways. Suppose we’re working for a hospital. Then our collaborators, which may include executives, physicians, residents, researchers/PIs, technicians, and/or nurses, are typically looking to improve the quality of care. This is reflected by any number of things, including:

Vague Request

Suppose you're met with the PI (or another collaborator), and their vague request sounds like:

"Can you help us improve the quality of care?"

Here’s one way to navigate the discovery phase template with the collaborator.

1. Customer Value 

I hope we can agree with the collaborator that "The hospital’s mission is to help our patients feel better."

2. Objective 

Because the request is so vague, we’ll need to discuss how the collaborator envisions that this request will help the hospital grow/succeed, e.g. which customer are we helping?

The goal is to try to get as specific as possible, on what outcome we’ll try to reach together. 

For example, one objective that may help our clinical teams be less overwhelmed is trying to reduce their load during patient rounds. (At a high level, patient rounds are a discussion of each patient’s situation with the whole clinical care team, one patient at a time.)

3. Strategy 

To achieve the objective, we will do the following: 

a. Decrease/increase 1+ of 14 levers

By looking at this chart and thinking about the steps/events of what happens during rounds – and transition probabilities between them – we can see that there’s a lot we still don’t know about our request. That is, should we be focusing on line 10: reducing the churn of high-value customers, by focusing on reducing their load? 

And is the "customer" here, that we’ll be trying to reduce the load of, everyone doing rounds, or just the physicians (who get paid the most), or just the nurses whom we can’t retain for more than 6 months? Or should we focus on someone else entirely?

Suppose, for this example, that we agree that need to help retain our nurses, by trying to reduce their load during rounds.

b. Summarized by metric

How should we define “load”? Is it the length of time it takes to do rounds? Or should it reflect another aspect of rounding?

Suppose, for our example, we agree that we’d like to cut down the time for nurses to prepare for rounds from 30 minutes to 25.

c. Process of pulling the lever to add customer value

First, we need to understand, minute-by-typical-minute (!) how those 30 minutes are currently spent! Can we observe the next few rounds – and try to understand how the preparations happen?

As you observe/learn, try to understand the nurses’ current workflow – and what happens every minute. That is, what are all the systems, files, screens, and log-ins that a nurse needs to access/check/confirm to create a discussion plan for the rounds? Is the plan created on paper, that needs to be entered into multiple systems manually? Does the nurse get interrupted during this time? If so, for what, when, and how often? Are rounds typically at the same time; if not, how much of a heads-up is given for prep? Is there anything else we should be considering? 

Based on our understanding of the workflow, let’s discuss with our collaborator what is the smallest – very specific – thing we can try to improve about this process.

For example, how quick/simple/painless would it be to:

The answer to this question will be the scope of the experiment we’ll be trying out, on a small subset of prep for the rounds – for one nurse to start (e.g., our nurse champion), to evaluate if the suggestion is promising in changing outcomes, e.g. reducing the load for nurses; or it will help us conclude that something else should be considered.

d. With an (approximate or accurate) recommendation

Are we OK to reduce the prep time by any amount, or do we need to reduce it by at least 5 minutes? 

e. Partners implementing the process of pulling the lever

Before we do any development, who are all the people we need to make aware and get buy-in from, for this to happen?

That is, in our example, do we need the Head Nurse to approve the process, along with a nurse champion? Do we need Engineering and EHR Analyst support to help us pre-populate information? Do all the physicians and residents need to be made aware of the (smallest) proposed change? Whom else do we need to loop in, such that no one is surprised?

Depending on what the proposed change is, this may be a long list (!).

4. Outcome 

Implementing the strategy above should result in: 

a. Company growth of size (size)

It costs over $60K to replace an experienced nurse. If our small change, which we picked to be relatively easy to implement, can help the hospital retain even 1 nurse, the ROI for this project will be XL.

b. Over the next (timeframe)

We will compare 6-month retention rates of nurses, starting the comparison after the experiment goes live. 

c. With a deadline of (timeframe)

The original deadline of 1 week was unrealistic because of the number of partners that needed to be on board for the experiment to start. Instead, it will be 1 month after the experiment goes live to come up with a recommendation for the next steps, e.g. trying something else to reduce load or expanding the experiment to more nurses, as they prep for more rounds. 

5. Additional Context Is there anything else we should know?

This would be a great time for our collaborator to share that we’ve tried answering this request before, what was tried in the past, what worked and didn’t – and what were some lessons learned. Also, if there’s any additional context we should know, RE policies, politics, retirement plans, etc. – this would be a great time to share that. 🙂

This way, we don’t spend time in development that was tried before and didn’t work, and try something else instead.

Hopefully, we now have a better idea of what our collaborator is looking for. With this detailed context, we as a Data Team, can now make a more informed prioritization decision based on effort, outcomes, and deadlines.

Good luck! 

You May Also Like