Clinical Decision Support in Open MRS ?

[Hoc NGUYEN THANH]

Hi everyone,

We're a small team in a software company in Vietnam who has been working in Open MRS for the last 6 months.

Now we want to develop a Clinical Decision Support in Open MRS for our users: outpatient clinics which serves HIV patients.

After 2 days looking at the Open MRS modules list, I see 2 modules "Calculation" and "Notifiable Condition Detector" which we can reuse part of its code/idea.

Our use-cases:

1. The user must be able to mange (add/remove/update/applicable) medical practice guidelines to use in the CDS. These guidelines may come from various sources.
2. Based on the historical data of the patient, CDS module should be able to give useful recommendation/indication
3. Based on the clinical results, CDS module has to give warning or prediction on the status of the patient in the upcoming time (e.g. disease will get worse)
4. Based on the clinical results, CDS module has to give recommendation such as changing of regimen (e.g. no improvement on a health status of the patient).
5. CDS module has to give warning if there is drug interaction in the regimen given by the doctor
6. CDS module should be able to detect possible errors in data manipulation of the users, e.g. wrong patient, wrong dose, wrong drug, etc.

Challenges for this CDS module are:
- How to design to let the user define/update medical practice guidelines via user interface as powerful/flexible as possible
- The rules processing engine must be efficient (it should not take too much time to process rules)

Any suggestion/comment will be appreciated. Thanks.

--
Regards,
Hoc, NGUYEN THANH
Email: than...@gmail.com
Mobile phone: (0084) (0)908.129.309

[Lluis Martinez]

Hi,

I did some prolog and lisp in university, and I'd like to learn about
java based rule engines. I can give you a hand in the module, but my
availability is reduced (around 10 hours/week).

Best regards

> --
> OpenMRS Developers: http://go.openmrs.org/dev
> Post: d...@openmrs.org
> Unsubscribe: dev+uns...@openmrs.org
> *** OpenMRS has moved from Subversion to Git. Read more:
> http://go.openmrs.org/git ***
>
>

[judy wawira]

Hello

Have you had a look at the clinical/patient summary module?

At AMPATH we have been using CDSS for many years now.Basically we use arden syntax for the rules, and borrow a lot of functionality from the CHICA work on openmrs. We are able to run paper summaries,and mobile based summaries for our CDSS program. Our summaries are postdated i.e.. they trigger when a patient is overdue for some test/ intervention.

We have not really implemented prioritization of reminders, but we let the program decide on what are the key reminders that they want implemented every quarterly and their weight in importance.At present we have reminders for adult HIV care, pediatric HIV care and Antenatal care. Over the next weeks we will be rolling out primary care reminders and oncology care reminders. We are working on connectivity between our sites, and thus are refining the flowsheet module in openmrs to support real time care/reminder system. Due to remote implementation of the standalone machines with a list of reminders, we are able to download new reminders from production and update remote sites with these reminders using the current system.

Specific focus on your questions;

> 1. The user must be able to mange (add/remove/update/applicable) medical
> practice guidelines to use in the CDS. These guidelines may come from
> various sources.

Any CDSS implementation must incorporate data quality review. We have data assistants and train the health care providers to make corrections on the paper summary which is then corrected in the main system. We model rules around clinical care guidelines that AMPATH and the Kenya Ministry of Health program decides as the standard of care

> 2. Based on the historical data of the patient, CDS module should be able to
> give useful recommendation/indication

We focus on attainable variables and use derived concepts to recommend and action that a clinician can consider when attending to the patient

> 3. Based on the clinical results, CDS module has to give warning or
> prediction on the status of the patient in the upcoming time (e.g. disease
> will get worse)

This is predictive modeling. I think a s afirst pass of clinical sumamry it may be too ambitious(its on our wish list). Depending on the implementation, our experience is taht you have to improve the data quality to even be able to manipulate it for such level of modeling

> 4. Based on the clinical results, CDS module has to give recommendation such
> as changing of regimen (e.g. no improvement on a health status of the
> patient).

We rely on radiology/ lab results / pharmacy or drug regimens and key scales liek WHO staging to inform our actions

 
> 5. CDS module has to give warning if there is drug interaction in the
> regimen given by the doctor

 see above

> 6. CDS module should be able to detect possible errors in data manipulation
> of the users, e.g. wrong patient, wrong dose, wrong drug, etc.

 

On its own, we correct errors at indivudal levels e.g., adultas with pediatroic reminders, etc. 

On  large scale implementation, several tools like the Data integrity module done in openmrs has relally helped us clean the data.

Good luck!

Judy

--
Judy

[Hoc NGUYEN THANH]

Hi Lluis. 

Good to hear that. I will contact you once we get something clearer.

Hoc

2012/10/11 Lluis Martinez <llu...@gmail.com>

[Hoc NGUYEN THANH]

Hi Judy,

Thanks for sharing. It's interesting!

Thanks,

Hoc

2012/10/11 judy wawira <judyw...@gmail.com>

[Timothy Cook]

While not specific to OpenMRS. You may find the approach used in EGADSS HELPFUL.
http://egadss.sourceforge.net/

Cheers,
Tim

[Tammy Dugan]

Hoc NGUYEN THANH,

This is exactly what the chica module does. I received your forwarded email from Vibha Anand. I am planning to help you out. I have just been busy this week and haven't had a chance to respond. See responses below:

Our use-cases:

1. The user must be able to mange (add/remove/update/applicable) medical practice guidelines to use in the CDS. These guidelines may come from various sources.

The dss module and arden parser (which we both wrote) will translate clinical guidelines defined with arden syntax in mlm files. The mlm files can be dropped in a directory, compiled on the fly into java rules implementing the openmrs Rule interface, and then compiled into class files that are available for execution by the Openmrs Logic Service. You can also create java rules directly and drop them in a directory for on the fly compilation as well.

2. Based on the historical data of the patient, CDS module should be able to give useful recommendation/indication

We do this with chica using the mlms.

3. Based on the clinical results, CDS module has to give warning or prediction on the status of the patient in the upcoming time (e.g. disease will get worse)

We do this with chica using the mlms.

4. Based on the clinical results, CDS module has to give recommendation such as changing of regimen (e.g. no improvement on a health status of the patient).

We do this with chica using the mlms.

5. CDS module has to give warning if there is drug interaction in the regimen given by the doctor

We have not done this with chica yet but could in theory.

6. CDS module should be able to detect possible errors in data manipulation of the users, e.g. wrong patient, wrong dose, wrong drug, etc.

Chica uses paper based questionairres so this in not applicable.

Challenges for this CDS module are:
- How to design to let the user define/update medical practice guidelines via user interface as powerful/flexible as possible

We do this by directly editing the mlms as text. The long term goal is to build a rule builder.

- The rules processing engine must be efficient (it should not take too much time to process rules)

Chica's rule engine is extremely efficient. We can execute 50 rules or more in 1 or 2 seconds.

Chica uses the dss module and  Logic module and openmrs 1.7.x. The calculation module, to my knowldege, is not thread safe for rule execution unless the appropriate changes have been made.


Thanks,

Tammy Dugan

-- 
Tammy Dugan
Technical Lead
Child Health Informatics Research and Development Lab (CHIRDL)
Children's Health Services Research
IU School of Medicine
317-278-6926

[Tammy Dugan]

We also support rule prioritization. Here are a list of publications about CHICA, including two randomized controlled trials proving its effectiveness:

http://www.ncbi.nlm.nih.gov/pubmed/22744960 (RCT)
http://www.ncbi.nlm.nih.gov/pubmed/21672910 (RCT)

http://www.ncbi.nlm.nih.gov/pubmed/22508925
http://www.ncbi.nlm.nih.gov/pubmed/21346962
http://www.ncbi.nlm.nih.gov/pubmed/18998823
http://www.ncbi.nlm.nih.gov/pubmed/16779000
http://www.ncbi.nlm.nih.gov/pubmed/15360800


Tammy

[Ben Wolfe]

The Calculation Module (a simplification of the Logic Module) allows this already.  The part that needs implemented is where to integrate and use those Logic Rules (or Calculations)

https://wiki.openmrs.org/display/docs/Calculation+Module

https://wiki.openmrs.org/display/docs/Logic+Service

The Chica system that Tammy refers to uses MLMs to create the Java Logic Rules. 

What is the status of turning different language (arden, groovy) into Calculations in the Calc Module?

Ben

[Darius Jazayeri]

We haven't implemented "user-defined calculations, stored in the database" in the calculation module, though this should be straightforward to do, especially copying what exists in the logic module now.

-Darius

[Tammy Dugan]

I would definitely say that the dss module and chica (using the Logic Service directly) are more robust and better tested than the calculation module, at this point, since we have been using them in production for clinical care for 3 and half years now.

Tammy