Comment by Sharon Dooley on April 27, 2012 at 23:47

After learning about CDSS and probability in decisions - I never thought I would use anything in mathematics in my career as a midwive other that drug calculations! Although I still cant quite get my head around the algorithms and calculations, I can understand the wider implications and use such as you are looking at - its the background stuff behind what the clinician sees when using CDSS. Really complicated but really useful.

Comment by Mark Barrios on April 28, 2012 at 0:48

It wouldn't be as complicated as you thought if you are the user, you will just let the machine do the thinking for you but flexibility and behavioural change are really necessary and very challenging at times.

Comment by Senthil on April 28, 2012 at 1:49

I have studied Machine Learning algorithms like Genetic Algorithm, Bayesian Networks etc with respect to computer science. Though the concept is evolved from probability statistics it can be applied on varied science field. One Such application is like application of Bayesian methods in Medical Images for segmentation of cancer cells :-

1. Bayesian Method in Medical Image

2. Bayesian fused classification of medical images

I am happy that people are exploring methods from statistics into applied Health science which could utilize existing computer programs. But the issue is with the Data. Generic public data for decision making may sometimes fail to support individual health assessment and more individual data has implications on personalized care ignoring a common method for all. Proposition of input data and the output - decision is still supervised under manual guidance of experienced doctors. i envision computerized statistical methods producing standardized output decisions which has caliber of experienced doctors, meaning machine learns from doctor decisions and with each supervised decision of doctor - the program itself becomes experienced.

Comment by Mark Barrios on April 28, 2012 at 11:56

Thanks for the input Senthil!

As GA is an example of latest innovative technique in CDSS, it uses the Bayesian network which applies probability statistics and a specialised method of gene expression data analysis.

Like other forms of technological innovations, this concept is not an absolute perfection. This method has issues on parameters such as the size of the solution population, the rate of mutation and crossover, and the selection methods and criteria which can significantly affect its performance. For example, if the solution population size is too small, the genetic algorithm may have exhausted all the available solutions before the process can identify an optimal solution. If the rate of genetic mutation is too high, the process may be changing too fast for the selection to ever bring about convergence, resulting in the failure of generating an optimal solution. Some clinicians are also hesitant using it because of less transparency in coming up with the solution.

These are some of the flaws and limitations that I would like to recommend to consider when conducting further research and product development on this type of innovation. Nevertheless, existing reviews and evaluations have proven promising benefits.

Comment by Bhavisha Daya on April 29, 2012 at 8:20

Hi Mark,
CDSS is certainly a very interesting topic particular due to the vastness of the different areas that it is applicable. Also, the use of decision support in this context would certainly allow for the best options to be viewed in an unbiased manner.

Coming from a health science background, my understanding of computer science is limited, but I would be interested in know the greater implications of the use of such a tool, particularly relating to the patient experiences and the possible impact on their autonomy.

Reading up on the www.oncotypedx.com site of the patient stories they talk of how the Oncotype DX test supported their views, it would be interesting to know the results of the opposite.

I wonder if the factors such as the patient's values, autonomy and decision making would either limit or promote the use and development of such technologies in the future.

Comment by Mark Barrios on April 29, 2012 at 23:49

Thanks for the comment Bhavisha.

Most CDSS support the clinician in decision-making and in treatment planning which subsequently must be discussed with the patient. Patient autonomy, values, and decision making are another points for discussion to assess if they can limit or promote the use of CDSS as systems developers are always after at creating methods that will enhance patient care. Although we cannot deny the fact that patients are still the centre of care. I came across with a study that might interest you about evaluating Adjuvant!Online as less attractive to patients:

http://onlinelibrary.wiley.com.ezproxy.auckland.ac.nz/doi/10.1002/p...

This is only one of those few studies on patient satisfaction with CDSS. As it is a relatively new technology in health care, it calls for more studies and trials focusing on clinician's performance and patient satisfaction.

On the other hand, if a CDSS is clinically proven, well-established or widely used and reliable, such as the AJCC TNM staging system (http://www.cancerstaging.org/mission/whatis.html), patients are more likely to use them.

Comment by Giselle Pascual on April 30, 2012 at 0:33

Hi Mark, 

Just curious about whether or not you looked at the cost effectiveness of a GA approach to CDSS in your essay. I read a systematic review which shows the effectiveness of CDSS at the clinical level, but it states that studies on the impacts on efficiency, workload and cost are sparse so it couldn't shed any light on this (these were not limited to GA CDSS might I add). So I was just wondering if you had come across anything that would show this. From a theoretical perspective, you'd think that costs could be reduced and waste decreased because CDSS improves patient safety and medical error, which we know to be very expensive to the health care system.

Comment by Mark Barrios on April 30, 2012 at 22:06

Hi Giselle

I did not discuss much about its cost-effectiveness as most evaluations and descriptions in literature always mention its capability to reduce cost, improve service efficiency or even clinician's performance and when integrated into clinical workflow can increase accuracy and better clinical outcomes. These studies might help you:

on clinician's performance: http://jama.ama-assn.org/content/293/10/1223.full.pdf+html

on clinical practice: http://www.ncbi.nlm.nih.gov.ezproxy.auckland.ac.nz/pmc/articles/PMC...

reducing cost: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1513678/

overall benefits: http://www.annals.org/content/early/2012/04/20/0003-4819-157-1-2012...

Comment by Mark Barrios on April 30, 2012 at 22:31

As I came across numerous and varied literature on CDSS I notice some of them are free and can easily be accessed by end-users, such as AJCC staging system, although most of them aren't free particularly those focused on specific needs, eg., genetic assessment.

If designing your own CDSS, it will initially be a bit costly but if this is considered as an investment and in the long run will improve service provision, patient outcome and will reduce cost then it is a practical solution.

I reckon CDSS has the power to revolutionise clinical practice especially in oncology service, especially if the system is supported by clinical trials and evidence-based studies.

Add a Comment