How healthcare organisations are tapping data analytics

Point solutions

Why are healthcare providers choosing point solutions, as opposed to taking the enterprise-wide approach? Is it because some analytics initiatives are driven by specific departments and hence, they take a siloed approach towards analytics? Or is it the nature of the industry where technology providers often develop solutions for specific use cases?

Nakhooda: I think it’s both. It’s a combination of individuals like a radiologist who wants to create a tool for imaging and so they identify a point solution to read images and identify anomalies.

If you look at the supplier landscape in this part of the world, almost all of them are point solutions, and there’s nothing that brings together the three components in our approach to healthcare analytics, that is, to align best practices with analytics and adoption to drive outcome improvements.

For example, let’s say a relatively fit and healthy 50-year-old person goes for a knee replacement surgery and stays in a particular hospital for three days. But the average length of hospital stay for someone with a similar patient profile should be one day. We can then start looking at the hospital’s entire orthopaedic process and see how we can cut two days out of their system.

When they cut the length of stay, they can cut costs and improve outcomes. We mostly see companies focusing on just analytics or best practices through benchmarking, but there are very few I’ve seen that brings all three components in our approach together.

Doing so will enable healthcare providers to provide the best care at the lowest cost with the best outcomes. A lot of people look at analytics to solve a particular pain point but they’re not thinking about the entire healthcare system to look at where the biggest cost and issues are.

In your approach, what do most healthcare organisations struggle with? I know some organisations tend to struggle with data that’s locked up in proprietary healthcare systems and it takes a lot of effort to extract the data and combine it with other datasets to do the analysis. Do you see that as well, and are there any capabilities in your platform to help healthcare organisations overcome these challenges?

Nakhooda: Yes, one of the biggest challenges is data aggregation because all the data sits in a bunch of different systems. In fact, some of the health systems we work with in the US have up to 50 systems because they’ve grown by acquisitions, or they are just very big systems.

To pull the data, standardise it and bring it into a way that is usable, we have over 300 connectors to all the various EMR (electronic medical records) systems out there. Even if it’s a custom system, we know how to map the data very quickly.

For example, we were working with a public hospital on a data-driven outcome improvement project around two clinical areas. One was pneumonia, and the other was on colorectal recovery after surgery. We had to pull data from their EMR system and their data warehouse, and we were able to do that in less than a week and standardise the data.

Another challenge is data protection. There’s always this need to leverage data to improve outcomes but there’s also this fear of a cyber attack. For us, that’s critical because as you can imagine, our business is all around your data and that’s all we do. So, we have put in place very strict security protocols. We comply with high trust, and we have all the correct certifications. We always work very closely with chief information security officers to make sure that they’re comfortable.

It also helps to have strong leadership and data governance structures in place. We have support and consulting capability around setting up a proper data governance structure and auditing controls to make sure only the right people access the right data. Also, some healthcare data is very sensitive and some is not. Sometimes, focusing on less sensitive data can still provide a lot of insight.

Clinical research

Besides improving patient outcomes, do you work with researchers on clinical research projects?

Nakhooda: Yes, we’ve worked with the Singapore Ministry of Health’s office for healthcare transformation to create a predictive analytics tool that identifies who’s more likely to die from Covid-19 using data from our secondary data store in the US. A lot of the hospitals we work with give us secondary data usage rights for data that’s anonymised. As a result, we have one of the largest Covid-19 registries because the US was hit very hard.

Using insights derived from a predictive model, we could then prioritise who gets Covid-19 vaccines first after essential workers, healthcare workers and the elderly. We also identified insights such as males being at a higher at risk of dying from the disease than females.

We are also planning to do a comparative study with another healthcare group on the treatment of Covid-19 in the US and Singapore, such as the different drugs that were used and what the different outcomes were.

Digitising medical records

I suppose to really harness the benefits of your platform, healthcare organisations need to digitise their medical records for a start. How are you working with healthcare institutions in countries where digitisation isn’t as pervasive as compared to more mature markets?

Nakhooda: Great question. We created a partnership with a start-up called Smarter Health which was started by top healthcare executives to help insurance companies and healthcare providers collaborate in a more effective way, especially in developing markets.

Specifically, Smarter Health provides a platform to digitise the data between the hospital and the payer (health insurer) because today after I submit my claim, which is usually manual, the payer decides if the treatment is appropriate and should be reimbursed. While Smarter Health is collecting the data, we’re analysing it to determine if the treatment is appropriate based on the data we have on other patients. We’re basically looking for clinical appropriateness to help payers decide if they should approve or decline a treatment pre-authorisation request, or if more information is needed from the doctor.

But what’s interesting is that we’re happy with just billing data if that’s electronic – and most hospitals have that because they want to get paid. Health systems have this perception that they need to have the perfect data, but we can identify a lot of interesting things from billing data, which contains information on diagnoses and treatments. It has enough information that can tell us quite a bit about the health system and the opportunities to cut costs as well.

How do you make sure the data is accurate?

Nakhooda: We do a lot of data quality work on our platform where we pull data and look for things that just don’t look right, such as a pregnant male or someone who has had his appendix removed twice. The other thing we do is to cross check and look for any “missingness” of data. For example, if a person has not been diagnosed with diabetes, but his lab report says otherwise, we’ll check if there’s an anomaly.

We will also check clinical notes to make sure that the data is correct because if you have the wrong data, you’ll end up having the wrong outcomes. But the nice thing is when you’re looking at large populations, any anomaly sticks out. If it’s a consistent problem, we may recommend the hospital to look at the data entry points. Maybe the clinicians are not entering data correctly, but in many instances we can figure out if something is wrong.