Wir freuen uns Ihnen folgende internationaleÂ Keynote LexturesÂ anbieten zu können:
âMaking Healthcare Processes Work: Clinical-guidelines based Decision-Support Systemsâ
Mor Peleg, Haifa (Israel)
Moderation: Prof. Dr. Richard Lenz (Erlangen)
Healthcare organizations are facing the challenge of delivering high-quality affordable services to their patients. To help support this challenge, medical Informatics research offers formalisms for developing clinical guideline-based decision-support systems (DSSs). On the other hand, business process management (BPM) offers IT support for healthcare processes using workflow technology. By integrating aspects from these two approaches, there is hope for achieving better healthcare process support.
I will start by reviewing trends in Medical Informatics research on clinical-guideline based DSSs that utilize BPM. Several groups who have been developing clinical guideline formalisms have seen the value that BPM methods and tools could bring when they are integrated with guideline formalisms. I will present the work that I have done on the GLIF guideline modeling language, highlighting the way in which it uses standards to specify the medical concepts that clinical actions and decision refer to, the patient information model, and the decision criteria that refer to concepts and data, and at the same time, supports flexible processes. I will then review the work that I have done on mapping the GLIF and EON clinical guideline-modeling formalisms into a workflow model. Using methods and tools developed by the BPM community, I verified that clinical guidelines encoded in GLIF and EON encoded guideline satisfy soundness and liveness properties. People from the medical informatics community have often argued that the BPM formalisms are not appropriate for supporting healthcare processes because they do not offer enough flexibility. Together with the group of van der Aalst, we compared various guideline formalisms in terms of their support of flexible workflow patterns and contrasted the guideline modeling representations with declarative representations in terms of flexibility. Others have extended this work by using formal methods based on Petri Net semantics to prove satisfaction of workflow patterns by the PROforma guideline formalism.
I will then continue to review trends in BPM research on supporting healthcare processes. These works address healthcare processes, including action management more than decision-support. They include healthcare process design for flexibility, verification and testing of healthcare process models (these two topics overlap with issues reviewed in the first half of the talk), utilizing clinical semantics for IT support, and process mining & learning. I will focus on the topic of process learning and present work that my group has done on mining healthcare process instance data to learn what the best path should be in order to achieve desired outcomes for patients with different contextual characteristics.
Many of the works reviewed in my talk were presented in PROhealth workshops.
Ulrich Mueller (Institut für Medizinische Soziologie und Sozialmedizin, Philipps-Universität Marburg)
The German Council for Social and Economic Data (RatSWD) has installed a working group for the preparation of a National Death Index / National Mortality Database. A National Death Index covers all deaths in a country registering vital data, basic biographical information and a detailed documentation of the cause(s) of death for each case. Many other countries do have such National Death Indices, some already for decades. A National Death Index is an indispensible infrastructure for social, public health and medical research. Registration of vital events plus cause-of-death documentation is the oldest and most extensive public health surveillance system. International compatibility of such databases opens many options for morbidity and mortality risk identification, but requires more common features than proper use of the International Classification of Diseases (ICD).
General data quality and procedural standards recommendations should at least cover these areas: (1) Speed (In the US National Death Index data records are available only 18-24 months later), (2) Completeness (in smaller European countries, a substantial proportion of the population spends some years abroad, thus, there should be an exchange of mortality data between countries of birth); (3) Record linkage capacity; (4) Multi-causality features in the cause-of-death documentation;(5) Convertibility between coding systems; (5) Efficient, Information preserving privacy protection features; (6) Data quality documentation; (7) High bioethics standards.
Specific recommendations will have to consider the federal system of government in Germany with three layers of government (federal, state, municipal) both in legislation and administration, with universal, but very decentralized registration of every birth and death and all residents and moves, and a completely separated processing of death certificates.
The recommendations of the working group will be presented and discussed by a small panel of experts, comprising Karl-Heinz Jöckel (Essen), Sabine Luttmann (Bremen), Ulrich Mueller (Marburg), Mechthild Vennemann (Münster).
nota bene Dr. med. Sabine Luttmann Bremer Institut für Präventionsforschung und Sozialmedizin Universität Bremen
“EMR Systems and Secondary Use for the Retrieval of Similar Case Reports”
Katsuhiko Takabayashi, Chiba University (Japan)
Moderation: Prof. Dr. Alfred Winter (Leipzig)
Electronic medical record (EMR) systems have been implemented in more than 39 percent of major Japanese hospitals, and a large number of discharge summaries are now stored electronically. By utilizing these discharge summaries, we performed morphological analysis and text mining with TF*IDF method in three hospitals and compared the vectors of the same DPC (Diagnosis and Procedure Complex; Japanese version of DRG) code between the hospitals. Even though there are different styles of discharge summaries in these hospitals, we obtained proper extracts from the summaries of the same DPC codes, or in other words, the computer could make a proper diagnosis of the DPC codes out of more than 80 percent of the cases in every hospital. Furthermore, the integrated summary data of these hospitals also showed favorable results, suggesting that a huge combined database of many hospitals can be utilized for text mining.ãThis fact may also allow us to retrieve the similar complicated or unsolved cases or reclassify the disease entity by text mining. Especially, to retrieve similar case reports is one of the important activities for clinicians when they encounter an unknown case to make a diagnosis or to treat patients. The Japanese Society of Internal Medicine (JSIM) holds more than 30 clinical meetings yearly in seven districts throughout Japan in which a lot of rare case reports have been presented. We have built a database of more than 15,000 case reports from these meetings as well as extractions of case reports from MEDLINE and conducted morphological analysis and text mining. Now we provide to Japanese physicians to search for similar cases through the internet by using TF*IDF method from the JSIM homepage. They can search for similar cases not only by keywords, but from the patients abstracts. We will apply this system for the database in other Japanese medical associations for the medical services. In addition to that we are planning to integrate the discharge summaries of 42 National University Hospitals which have more than one million discharge summaries per year and start the same kind of retrieval service which we hope will contribute to medical progress.