openEHR solution for DSS in epidemiology

Dear members of openEHR Clinical List,

I am much honored in joining this mailing list.

My name is Luciana Tricai Cavalini, and at this moment I occupying the position as the Director of the of at in .

My research field is social epidemiology, that is located on a triple boundary among health sciences, social sciences and biostatistics, and the critical concept for this knowledge area is information.

Thus, our research group has been developing some critical approach abour how information systems are incomplete, and we have been trying to reflect on and develop some solutions for our legate systems in a peripheral country such as (please see an iceberg of our research products on http://www.scielo.br/scielo.php?script=sci_arttext&pid=S0034-89102007000100012&lng=pt&nrm=iso&tlng=en).

But we have realized, since we started our research in 1999, that there is a lot to do at the point of care as far as health information systems goes, and being an epidemiologist, that it is critical for health surveillance. Probably outside this healthcare field is more known as “epidemiological surveillance”, but for this moment, I will consider it as a semantic issue.

Patients with suspected cases of epidemic and endemic diseases, with high potential of transmission, both to healthcare professionals and to the community, show up every day on healthcare settings, especially emergency rooms and primary care settings, where healthcare professionals are not completely prepared to deal with all biohazard and disease control issues. I will give you two examples:

  1. A patient with fever and cough for 2-3 weeks is by definition suspected of having tuberculosis, and if this patient is admitted at the hospital and stays there outside the respiratory isolation, other patients (and especially healthcare professionals, dealing with similar cases every day) might be infected. In an average healthcare setting in a country like Brazil, this case will be kept at the end of the line, because victims of traffic accidents and open violence are showing up all the time, but for the entire period the suspected tuberculosis patient is there he is likely to release tuberculosis bacilli to the environment, what, together with another similar cases every day, it increases the risk of hospital tuberculosis infection for everyone sharing that environment. Besides that, there is the fact that family/workplace contacts could be in risk of infection even not being there at the hospital, which sets up a typical slow non-scale network topology for the tuberculosis endemic. This topology has a defined space-time configuration, that can be set into an exponential relationship if one thinks about an acute disease such as meningococcal meningitis;

  2. A patient shows up at the emergency room with a dog bite. An average physician in Brazil will know everything about how to deal surgically with the wound, and hopefully how to deal with tetanus prevention, but rabies prevention for post-aggression events is a little bit more complex algorithm, that includes: (a) the coverage of anti-rabies vaccination for cats and dogs in the region; (b) the clinical and vaccination condition of the animal; (c) the definition of “severity”, that is different for clinicians and epidemiologists, because for us, epidemiologists (OK, I admit, we are picky), a single wound deeper than epidermis is severe, because it reaches neural terminals, and that is the way rabies virus infects, and not by the blood stream. But we are not picky just because this is charming: let us remember that we are dealing with a disease with 100% of case fatality, but 100% of vulnerability, if the post-aggression event is managed correctly. And there is (d), (e) and (f) that you can check on CDC website if you want, but I do not want to bother you anymore with this complexity. But the fact is that our physicians are dealing with that all the time and they must be empowered in their decision-making process.

So, since 2003, our research group is developing the concept of a decision support for epidemiological surveillance to be implemented at every point of care, no matter what is the level of complexity. Because for us, epidemiologists, it is always good when we can identify a suspected case of an epidemic/endemic disease at the primary care setting, but it is better to identify that case at the high-complexity level (e.g., pulmonary surgery) than on the death certificate. But in fact our group started working on the case that is not even identified on the death certificate (what we call ill-defined cause of death, or even under-registration, the worst case), but for us now this is a done deal.

We want to develop a decision support system for epidemiological surveillance based on openEHR specifications. Our technological decision-making is based on the fact that openEHR standards are computable and interoperable, something that is badly needed for disease prevention and control on a national/worldwide level (just think about SARS). So, we need to start building archetypes that must contain the specific context needed for epidemiologic surveillance, but it must make sense to physicians filling the information at the point of care. Our idea is to feedback physicians at the point of care with some information such as: “This patient is a suspected case of tuberculosis. He must be kept in respiratory isolation, etc etc”, or “This patient needs 5 doses of rabies vaccination, etc etc”. I am not discussing with you the layout architecture of this DSS, because this is not the proper forum.

So, my question to this honorable working group is: how can I build archetypes that can fulfill the needs of epidemiologic surveillance field, and at the same time being meaningful for healthcare professionals at every level of care?

I know my question has a high level of sensitivity, but if it is needed, I can go “down” (as we epidemiologists say) to a higher specificity, in order to bring light to this room that, at this moment, is full of heat for our research group.

Thank you (in advance) for your attention,

Luciana Tricai Cavalini, MD, PhD

Department of Epidemiology and Biostatistics

Health

Dear Luciana

You are welcome. I have just replied to Seref on a similar matter and I believe that we can take the EHR to the point where you can write your algorithms to determine the risks/benefits of particular actions - even based on availability of treatments etc. If the EHR and the recording software is doing its own thing then it is very difficult to provide this sort of ‘boost’ to the clinical process. Some vendors will roll their eyes when they read this - but I am not talking about a single clinical system - quite the contrary. A platform will mean that many applications can read and write to the same EHR - it is just the personal health information being available as a service that we are talking about.

How to build archetypes? Well, I would suggest that you take begin to comment on the archetypes in the openEHR repository. How about we establish a group to work on the new Knowledge Manager we are releasing soon with the task of vetting archetypes for suitability for possible and known decision support activities. If there is a strong case for change, then the clinical groups can try and address it. Sometimes it might be necessary to have archetypes that are specifically for supporting some activity - such as a ‘risk of rabies’ archetype that gathers specific information that is not captured in other archetypes and allows the system to ‘boost’ the care in that situation. This could be presented when someone presents with an animal bite.

Does this sound like a good approach?

Cheers, Sam

(attachments)

OceanInformaticsl.JPG

Dear Folks,

Dr. Cavalini’s argument opens up a main concern about interoperability and standards. As Chair of HL7 Brazil, we are carrying out a discussion about the harmonization of HL7 CDA and archetypes, as on some instances there are reciprocal and complemental procedures on the specification of an EHR.

In Brazil, we are developing some initiatives using HL7 v. 3 with CDA, and I am, myself, involved on a development of EHR system using archetypes, with portuguese translation, and I could figure out that archetypes may fulfill some lacks on clinical issues on HL7 and HL7 may solve some inconsistencies with interoperability on administrative issues using archetypes.

Even being novice on archetypes, I am very enthusiastic with its wide open view on managing healthcare information and I look forward deep participating on the openEHR Project.

The suggestion to establish a study group to work with decision support systems will provide the necessary basic grounds to strengthen the rationale on any medical system, including epidemiological surveillance.

Regards,

Dr. Marivan Santiago Abrahão
CHAIR
HL7 Brazil
e-mail : marivan@abrahao.net
website: www.hl7brasil.org.br

Dear Luciana

You are welcome. I have just replied to Seref on a similar matter and I believe that we can take the EHR to the point where you can write your algorithms to determine the risks/benefits of particular actions - even based on availability of treatments etc. If the EHR and the recording software is doing its own thing then it is very difficult to provide this sort of ‘boost’ to the clinical process. Some vendors will roll their eyes when they read this - but I am not talking about a single clinical system - quite the contrary. A platform will mean that many applications can read and write to the same EHR - it is just the personal health information being available as a service that we are talking about.

How to build archetypes? Well, I would suggest that you take begin to comment on the archetypes in the openEHR repository. How about we establish a group to work on the new Knowledge Manager we are releasing soon with the task of vetting archetypes for suitability for possible and known decision support activities. If there is a strong case for change, then the clinical groups can try and address it. Sometimes it might be necessary to have archetypes that are specifically for supporting some activity - such as a ‘risk of rabies’ archetype that gathers specific information that is not captured in other archetypes and allows the system to ‘boost’ the care in that situation. This could be presented when someone presents with an animal bite.

Does this sound like a good approach?

Cheers, Sam

Dear Luciana,

I was a clinician, I am involved in openEHR-based EHR development, and I just did a bit of research on how to use archetypes and openEHR architecture in DSS. Let me try to answer your question and hope to get more feedbacks from other people.

Your question is: “how can I build archetypes that can fulfill the needs of epidemiologic surveillance field, and at the same time being meaningful for healthcare professionals at every level of care?”

Generally speaking, archetypes are used for health care professionals to record health information for a subject of care. Different fields normally need specific archetypes. What archetypes you would need and what information should be included in each archetype should be derived from the data recorded by health professionals. The forms currently used by health care professionals are good start points. However, before creating new archetypes, I would suggest to check whether the existing archetypes published on openEHR web site can be of any use. As a general principle, try to use existing archetypes as much as possible.

You also ask that how to build archetypes being meaningful for healthcare professionals at every level of care. I feel that you are asking that how to build archetypes which can be used for DSS, such as providing appropriate guidelines at the point of care. Don’t know whether I am correct or not? In fact, my personal view is that archetypes should work together with other knowledge components rather than a replacement. Archetypes along with Archetype Query Language (a semantic query language) would be the best approach used to achieve sharing knowledge components (such as Arden Syntax Medical Logic Modules) across different health institutions.

Cheers,

Chunlan

Hi Marivan

Welcome to the list as well.

We have also been doing some work in Australia with HL7 CDA and openEHR. The problem that we have been trying to solve here is how to produce CDA that is consistent without relying on multiple vendors having to re-engineer their systems to produce CDA. You will know that in the HL7 v2 world, certainly in Australia, that this approach has led to every system producing HL7v2 messages that are slightly different. Pathology companies here have to produce at least 35 different versions of the same HL7v2 message to suit all the different vendors that need these messages. This is despite them all being consistent with the Standards Australia version. We have serious concerns that the same issue may occur here with CDA R2 as vendors try to implement this highly abstract and complex specification. This will mean that interoperability will still rely on prenegotiation of messages and costly integration.

We have solved this issue by using an approach that uses openEHR archetypes and templates to define the semantics of some vendor system data. We can then programatically produce a an XML schema that is based on the vendors data as represented by the archetypes and templates. This is called a Template Data Schema (TDS) and from the vendor data,using a single transform, a Template Data Document (TDD) can be produced that conforms to the Schema. Once we have the TDD, it can be transformed to either openEHR data or to CDA R2. The CDA transforms are built using a library of transforms based on archetypes. This means that if the TDD has been built using an archetype that has a transform in the library, we can build that part of the CDA using that transform. Once the library of archetype based transforms is complete, then any TDD can be converted reliably to and from CDA.

This is not just an academic idea - if you are interested, I can demonstrate this process in action. One of the very nice things about this process, is that vendors need only XML expertise to make it work - they don’t need to be experts in either openEHR or CDA. The format of the TDS will become part of the openEHR specification very soon.

regards Hugh

Congratulations on this excellent initiative. Please keep the list informed on progress in this area. It is most important.

Yours sincerely

Wilton Braund

Dr Wilton Braund

Endocrinologist

19 Alexander Ave

ASHFORD 5035

0411 555 723

Hugh, at the moment we are integrating 2.7 million lab orders / results
to SIGA Saúde the São Paulo Health System. We tried to use HL7 v3 -
impossible. Then we decided to use the core content of the HL7 message
and use the XML schema we already have for the private sector defined
by Jussara's team at the Supplementary Health Care Agency. It would be
very nice to have a look at the TDD - to compare with what we have.
Tks.
Beatriz

Hugh Leslie wrote:

Hi, Hugh

hope you’re well. Definitely we´d like to see your implementation. That is of our big interest right now, because we´re developing a new message for health indicators, which health plans have to send to us (the regulatory agency)on a regular basis in order to assess the quality of health care provided from their network. Beatriz knows, it´s called system of information of health plans, and now it will be generated from the administrative messages we already implemented. It would be wonderful if we can introduce right now an openEHR solution in our project.
When can you make the demonstration?
Cheers
Jussara

I’m not on the list, people, I was once but I don’t have the disciplin it requires, so I quit. But I think I must try it again.
cheers
Jussara

I also am very interested in the TDS and TDD ideas. I guess we might face
similar problems when we will roll out the lab reports nationwide over the
next years in Austria.

Beatriz, could you please point me to the "XML schema we already have for
the private sector defined by Jussara's team at the Supplementary Health
Care Agency" for the lab orders / reports?

I would like to compare the content to what our doctors are discussing at
the moment, and to some other things, and slowly get an international
overview.

Thanks,
Stefan Sauermann