High Performance Computing - Is there a place for it in nursing?

[Denniscr]

Hi all...

I am trying to gauge nursing's interest in High Performance Computing.

I am a nurse who just happens to work on the Windows High Performance Computing team at Microsoft as a program manager. My team recently shipped our product, the first version of an HPC platform called Windows Compute Cluster Server 2003 (www.microsoft.com/hpc). As you may know, many sciences are exploiting the computational power of readily available commodity based clusters running either parallel or serial applications for the purpose of simulation or discovery. During the last two years I have been establishing research labs at universities around the world (http://windowshpc.net/blogs/overview/archive/2006/07/14/237.aspx) primarily to assist my team with open questions regarding our product. During this time, I have come to appreciate the place of simulation (based on computational methodolgies) in our time honored scientific process.

So, where is nursing in all of this? I would say absent. I have spent some time over the last two years looking for nursing researchers that are using HPC. I can only find a small handful of examples. Most of its use is confined to data mining. I have not yet found any evidence of the use of parallel computing for the purpose of simulation.

I have tried to get a few nursing schools interested in HPC but have largely failed for one reason. No one knows what I am talking about (you may feel the same way!). But there is a need for nursing researchers to understand what is possible. They need to know that they do not need to have a computer science or mathematics degree to take advantage of parallel computing on clustered systems.

Nursing researchers must begin to establish interdisciplinary collaborations with other scientific domains for the purpose of exploring complex issues facing nursing. They need to know that “supercomputers” are readily available “off the shelf” and affordable. I have a four node, 10 GFLOP system on my desk – total cost $4,000.00 USD as opposed to the Cray YMP 10 GFLOP system from 1991 at a cost of $40,000,000.00 USD (found at NASA and other gov labs).

The face of computing is changing dramatically. Parallelism is sweeping the industry and the sciences. Is it tough? You bet. Not many people can write programs that take advantage of true parallelism. But many people and companies are inventing the next generation of tools to make it significantly easier. With the advent of multi-core processors (now at 4 cores but soon to be hundreds) previously unheard of computational resources are available to all.

I would like to know you thoughts regarding this. Am I barking up the wrong tree?

Dennis

[rwingo]

Dennis,

Do you have any specific ideas where you think nursing can benefit from parallel computing for simulation? (And by "simulation", I'm assuming you mean some sort of modeling based on some data set)

I'm having a hard time thinking of where HPC might fit in for nursing / nursing informatics other than data mining... but maybe I'm just uneducated in that regard.

I'll reach out to some of my colleagues to see if some of them have any additional insights/opinions on this...

[Denniscr]

HPC can be used in many ways.

1. Speeding up common serial applications that consume large data sets
The compute cluster can be used for distributing common serial applications across many compute nodes. This is common practice among financial institutions. They often run serial apps as the basis of "monte carlo simulations" (also known as parameter sweeps) for the purpose of modeling decisions on their financial positions. Biologists do this with an application called BLAST. The app is scheduled to run thousands of times on many nodes each time with a different subset of genome data. This type of use gets you to your results sooner.

2. Conducting nursing research using simulations based on mathematical models
Simulations are best run on compute clusters using parallel methods. Consider some basic nursing skills...say deep intramuscualar injections. Most nurses accept the Z tracking method as an acceptable method for administering these injections. It seems to make sense. But what if a nursing researcher, using a model based on tissue engineering and computational fluid dynamic principles, produced a simulation that suggested that nurses abandon the Z method in favor of an alternative method. This type of use permits you to vary your experimental variables rapidly and in ways that you may not have considered in the wet lab.

3. Visualization of data
This is one of my favorites. Take any large data set and visualize it. Push it around and conduct "what if" calcuations in real time. This requires a compute cluster and HPC. Consider this...you have a model of an aortic anuerysm. You can "walk" into the model usign a 3-D vizualization CAVE (computer assisted visualization environment). You can then changed the physical attributes of the defect and instantly see the effect of that change. Or consider the nursing practice lab. It is necessary to start an IV on a fellow student. But it is helpful to do this in a CAVE as well. You can see the impact of placing the IV catheter in different anatomical locations and viewing the simulated flow based on changing angles, drip rates, physical position of patient etc... This type of use opens your eyes to a new world and from new perspectives.

These are just three ways to use a compute cluster. There are many other ways. Consider this: Computers are just a resource. Compute clusters increase your available resource and decrease your time to results.

Dennis

[leanthonyrn]

Also, nursing scheduling falls into a very complicated branch of optimization. This branch of optimaization problems are called the NP-complete problems. Please see the link for more information:

-NP complete problems
http://mat.gsia.cmu.edu/mstc/heurnote/node23.html

All,

There certainly is a place for HPC....however, strategically, our country's nursing leadership isn't there yet.

I have been in nursing informatics for quite a while and haven't seen nursing informatics incorporated into AONE to any degree yet. I spoke at AONE in 2005 on standard terminology and the government's initiative in this area and the CNEs/Managers in the room had trouble grasping my message, which was "get on board with nursing informatics." They tend to delegate this area to a Director of Nursing Informatics instead of "jumping in" and learning it themselves. Thus, the absence of strategic planning in nursing informatics in the provider setting.

Of course, there are some out there that "get it" but by enlarge, the critcal mass in nursing leadership isn't there yet.

I would be interested in research in this area.....and can be reached at the following address:

anne.felteau@parkview.com
Anne L. Felteau, PhD, MBA, RN,BC, CNAA, FACHE
Director, Nursing Informatics and Clinical Education
Parkview Health
260-363-3630

[Denniscr]

What will it take to get leadership on board? I would think that whatever strategy is used needs to position this as a "top of mind" issue. Any unified approach to influence leadership must have that in mind as part of the communication and PR plan.

I have faced this issue many times here at MS. How do you take a new technology and introduce it to ambivalent and/or entrenched audiences?

So, is there a strategy among the leadership of Nursing Informatics to broadly influence the broader nursing leadership? If so, has it been effective? If not, why?

Thanks...Dennis

[Denniscr]

leanthonyrn,

P and NP-complete problems are a great use of HPC. Are you currently using a cluster for nurse scheduling?

[rwingo]

There are already some good systems out there for nurse scheduling that don't require supercomputers.

I've been working with Per-Se Technologies' ANSOS One-Staff nurse scheduling application for several years now.

They have a fairly robust schedule creation engine that takes into account quite a few configurable rules and parameters. However, I don't know of many that actually use this functionality - probably because it's rather cumbersome to set up/maintain.

This scheduling engine will generate a schedule for a unit of over 100 employees in less than a minute on an average desktop computer.

Also, IMHO, there's a certain amount of consideration for "human factors" as well as clinical judgement that goes into making safe and appropriate schedules for a nursing unit.

Finally, one of the big buzzwords right now with employee scheduling is employee self scheduling. Many of the scheduling vendors are coming out with employee scheduling portals. Much of the "burden" of creating a nursing schedule is shifting from the managers of nursing units to the employees.