Designing the smart hospital for tomorrow, today

AUG 10, 2019 @ 7AM
Earlier, every record kept in the hospital was on paper including their risk instructions and now 45 years later – we are at the point of developing artificial intelligence within our systems. Was it easy? No. Were there things we learnt? Yes. Were there mistakes that we made? Yes.

Artificial intelligence
What does one basic thing good intelligence need? Good data. If you do not have good data and standardized nomenclature for that data – you will never get the information that you need moving forward.
 
“In the United States, the third largest cause of death is medical errors that have taken place. It actually beats out pulmonary disease. So, one point to illuminate a lot of errors, is through the use of electronic systems that are going to be far more accurate which is going to give us a lot of good information. However, I’m sure you have heard of the principal – ‘garbage in garbage out’. So, you have to put in good information,” said James Meier, business development manager, patient care analytics at Philips while presenting at the Future of Connected Care: Placing Patient Experience at the Centre of Care symposium recently.

“Health systems are challenged to more with less. The healthcare landscape is changing where there is a shift to value based / patient-centric care, costs are increasing, while reimbursements are uncertain and there is a need to increase the return on investment (ROI) while maintaining capacity.

Providers must drive operational efficiencies, but often struggle – enterprise lack appropriate performance management tools and systems to help drive this. Data silos, different systems, lack of coordination among departments and clinical and operational factors that are considered separately manifests in problems throughout the care cycle.

“Common pain points include patients accumulating in the emergency department awaiting admission, inefficient discharge between units, inefficient discharge flows to lower cost setting which again mismatches in bed and staff capacity with patient flow.

“Over the course of my history – I have seen where every hospital wanted its own clinical configuration. Also, it developed different standards for what they put in there. One person would call something – heart rate – another person would call it HR. This makes it very hard to get those reports and that type of information out of systems because the nomenclature was different.

“One example in the United States of one of our health care system that decided to create its own terminology and use a special company to develop reports and that company went out of business. Each hospital group that had purchased that system has spent between USD2 – USD10 million for that system and they got zero reports out of them.

“You don’t want to be in that position,” emphasized James adding that: you want to be able to use standards that can be utilized no matter what hospital there is. We know that every hospital does have its own unique needs, every intensive care unit (ICU) is a little bit different, but there are certain standards that we can help to put to affect so that data and reports become common.

Map of data flow
A hospital is an interconnected, interdependent system of care.
 
“Standardize unit level care by optimizing and standardizing care major flow settings, identify clinical readiness to move metrics and dashboards, reduce unnecessary length of stay. Co-ordinate care flow across the enterprise by building enterprise flow dashboards connecting care settings and implement decision support rules to progress patients through care settings. 
 
“You have to have a map of your data flow. Obviously, everybody feeds it to everybody – whether it comes home, the emergency department, the operating room, post operation or emergency care into general ward and ICU and critical care.

So, it is crucial to map out your flow and look at where your patients enter the system, exit and whether they return to the system. This however, should begin by looking at how to keep them from entering the system in the first place through the levels of programs that are out there.

The question is; with lots of information at our finger tips, how do we start putting them all together? Networks are just as hard these days because everyone has a different system in each network too and if you try to pull that information together – it becomes more and more of a hard task.

So how do we start from admission to discharge – making this much easier to do, much more fluid – when a patient comes in where the information flows with the patient?

Have any of you ever had a case where patients come in to ED then travelled immediately to the operating room, to the post anesthesia care unit and then go into the ICU? Do you have a complete record on that patient when they arrive? Do you always know what medications or the volume of fluids they were given in the operating room?

“Nine times out of 10 – they are not going to know that information until you get a lot of paperwork from people because paperwork has not caught up yet. The file hasn’t been done. That information should be seamless and in a true electronic environment – you should have that information flowing with you and built in so that you have a better picture of your patient when they enter your care area.

Opportunities for AI across the health continuum
To develop tools, we have to talk about:
  • Clinical decision support – AI combines large amounts of medical data to generate more holistic view of patients to support clinicians’ decision-making process.
  • Operational optimization – Supports clinicians to optimize workflows such as planning, procedure and selecting the right exam for the right patient.
  • Population health management – Predictive insights in patient population helps healthcare providers to take preventative actions, reduce risks and save costs.
  • Empowering patients – As AI gets embedded into solutions for home care and healthy living, this will enable people to take control over their own health.
“You also need that information to help you make policies for your hospital as to how to treat your patients.

“If you are an academic medical center in particular, researchers are going to need that information. If you think about it – 12 bed ICU with a typical vital sign monitor. The average vital sign monitor puts out 35 vital signs every 15 seconds over the course of a 24-hour time period – that’s 2.5 million data points of unverified information. That’s just your vital signs monitor.

“If you are a clinician in an ICU – anytime you walk into your patient’s room – you will probably have 2000 data points to consider. And you would have to start comparing that with the 2000 data points the day before, the day before that and the day before that. How does that help you improve your patient? To look at each sets of values and figure out what was going on would probably take you an average of 45-minutes per patient. If you have 3 – 4 patients that you are responsible for or if you are a physician who has to see 10 patients in the ICU – you don’t have that kind of time to spend doing that.

“So how are we going to improve the care moving forward?

Big data is wide, deep and dense
We are talking about more detailed data than ever before. So how do you start recognizing patterns and putting it together? How do you do this with less resources than you’ve had before?
 
We also know that hospitals start merging into groups as societies evolve because you can’t afford to have a CT scanner every 6 blocks down the road as it is too costly. So, as hospitals start to merge, they start to share the resources and information.
 
“As they merge and have all grown up on different information, its hard to put all that data together and start getting all that information out of it. Without good data, you will never get to artificial intelligence and there is where everyone wants to be these days.
 
But how do we get there?
​
Healthcare challenges to introduction of AI
First of all, we have to figure out ways to integrate that data and build tools that are reliable. You have to keep in mind as you go out and speak to others – what is it that you need for your hospital.
 
How do you take this information and start putting it together? You would need to be able to put it together for validation purposes so that you could start analyzing that information together. You are going to have to test that information and then you need to start sharing. As your systems evolve, and you start merging your hospital, you need to start applying those standards with every hospital within the country. These means there are lots to be done as we move into the realm of AI. 

EXPLORE FURTHER