To Develop a Medical Knowledge Management System by Integrating Evidence-Based Medicine and Expert’s Experience

To Develop a Medical Knowledge Management System by Integrating Evidence-Based Medicine and Expert’s Experience

Chun Hao Chang

Yea Wen Wu

Li Liu

a

Graduate Institute of Medical Informatics, Taipei Medical University

b

Taipei Medical University Hospital

E-mail：m110093012@tmu.edu.tw Abstract

The medical environment changes continuously, clinicians confront more challenges than ever. Therefore, Evidence-Based Medicine (EBM) becomes an important basis for medical professionals to make decisions.

However, the medical literature is immense but only small fraction is applicable in answering clinical questions. Unfortunately, the practice of traditional EBM costs enormous money and time. How to help clinicians for their decision making and to select correctly from quantitative historical traditions and expert’s experience by less cost and time consuming timely is the objective of this study. In this paper, we present a mechanism integrating Evidence- Based Medicine、data sources and clinical experience. With text mining tool and expert’s discussion to explore our study, we formulate systematically traditional review articles to define different level scores of data sources and then structure a knowledge map after the discussion of professional groups for conclusions that are evidence-based rather than commentary. With this analytic technique, the hidden and valuable data will be transferred to a methodized overview and can perform as supplementary back up for EBM for clinicians while making decisions to enhance the quality of patient safety.

Key Word : Evidence-Based Medicine 、 Expert ＇ s Clinical Experience、Text Mining、Medical Knowledge System

Introduction

While making medical decision, physicians rely on their clinical experience that they learned from textbooks or the experience of senior experts. But there’s considerable lag for the juncture of personal experience. As nowadays, medical science is fast and progressive developed, today’s recommendation from experts for same medical treatment may become the cause of mortality tomorrow for patient. Sydney Burwell told his students, while he was dean of Harvard Medical School, that half of what they learned from medical school would be wrong in ten years [2]. For medical professionals，to catch up the changes of medical environment is an important issue for them. However, while facing immense medical literatures, they don’t have enough time to read them all.

Further, as the most of medical literatures have various levels of applications, it makes many results of research discovered from medical journals are ignored. How to obtain the most useful clinical information in time by using the latest knowledge management techniques have drawn high attention by many medical professionals.

The concept of EBM （ Evidence-Based Medicine ） comes form Archie Cochrane, a clinical epidemiologist in England. He published a book “Effectiveness and Efficiency: Random Reflections on Health Services” in 1972. This book point out that medical resource should be applied to treatment that proved the effect by rigorous research and could offer rational medical treatment service after. Medical professionals have to accept a complete EMB education to ensure correct search among

important and relevant documents rapidly when facing clinical problems. They use them in clinical work after reading them strictly.

EBM is a kind of medicine to practice the best medical evidence for clinical treatments and medical services for the public or the specific patients [9]. The greatest advantage of EBM is to combine medical education with clinical practice by learning a specific methodologic and statistical technique to satisfy the clinical needs. The learners will find the goal of constant clinical studying and to solve problems by learning self. But it needs to read enormous, spend lots of time and cost much manpower. In this study, we combine EBM with KM

（ Knowledge Management ） as basis for clinical experience and deploy the EBM to KM platform to optimize the benefits of EBM and offer rapid clinical judgments for clinicians.

1. EBM

The EBM is a new paradigm of search、evaluation, using a specific systematic technique to replace the traditional medical paradigm and solve clinical problems which is based on authority. In traditional medical education, physicians treat patients focused on physiological pathology in order to understand disease mechanism and treatment foundation. Nevertheless, with the change of medical science and technology, the best treatment today for patients may have big difference from the ones tomorrow. The most important thing of EMB is to provide a broader point of view from their current inherent viewpoint and response with the provision of top quality medical cares.

2. Knowledge Management

Knowledge is regarded as methodized resources that develop continually by creation, sharing and accumulation. The biggest different between knowledge and the others resources is that knowledge is continuously cumulative. Others resources may become less and less after applications, but knowledge will be accumulated more and more after applications. For all

oral and written traditions in hospital s, files, entities, organizational management system, and even advanced technical network are all regarded the important circulating channels of information and knowledge [1].

Knowledge management use methods and tools defined by experts to support faster dissemination of scientific understanding, discoveries and information, the first step to achieve knowledge management is to adapt, save, learn, organize and innovate. Afterwards, with the help of information utilization, analysis, interpretation and experience sharing, the system can enhance the communication and coordination among members of care teams. Gradually, knowledge management becomes one of the medical industry's important research fields in recent years [3]. It supports medical professionals to use information effectively and flexibly and also bring benefits to both from the individual level to clinical organizations with more effective operation.

Motivation and Background

As medical science is fast and progressively developed nowadays, today’s knowledge will be out of date soon.

On the other side, the limitation from traditional recommendation and individual observation also show the variability of experience. Physicians need to improve their capability and general medical knowledge constantly, not only to guarantee the treatments for patients but also to reduce the possibility of dispute taken place. Therefore, the new paradigm for medical practice is emerging, i.e. EBM.

Although EBM had already practiced and been implemented among lots of countries, many people have questioned its effects. They think EBM is just a sort of learning method; it's difficult to estimate effect in essence, as EBM appears difficulty for implementation.

For example, it’s costly and needs time to practice EBM, like to differ from traditional review articles, systematic formulate questions pertinent to practice, use rigorous methods to appraise and synthesize literature etc. Each process needs time and maybe purchase various database software, computer facility and networks. On the other

side, the EBM system is incomplete, the system mostly used is Medline database whose index system is not yet powerful as expected, the searchers get easily frustrated during data searching who spend plenty of time and energy but may not obtain any useful results.

To improve such situation, in this paper, we take the clinical records as basis to build up a medical knowledge structure and add up methodology to generate the most efficient and optimal retrieval among the different levels of evidence. Further, use methods and tools to support faster dissemination via the internet for knowledge sharing inside organization. With this analytic technique, the hidden and valuable data in hospital will be transferred to a methodized overview and can perform as supplementary back up for EBM for clinicians while making decisions to enhance the quality of patient safety.

Materials and Methods

1. Structure

In this system, we bring the EBM literature, oral and written traditions in hospital and expert’s experience together to build up a medical knowledge system according to different attributes of data sources to define the basic knowledge unit and then generate the basic medical knowledge unit by careful evaluation. Further, KM professional groups will evaluate 、 design and promote to set up a medical knowledge database which we call it as 4E Medical Knowledge Database, i.e.

Evidence、Education、Experience、Electronic。

Figure 1、 4E Medical Knowledge Database 2、Preliminary Data processing

The preliminary data processing is to handle the text mining of data sources including the modules of data pre-processing 、 phrasal database 、 data analysis and

key-word vector.

(a) Data collection

Data collection is proceeded by KM professionals, they search and retrieve the literature which are regarded as useful and valuable. Data sources coming including : evidence-based medicine literature, traditional knowledge and expert’s experience. For the evidence-Based Medicine literature, searching the Evidence-Based Medicine Reviews (EBMR) where collect the publication after 1991 from clinical investigations, relevant research and journals. The EBMR includes two database : Cochrane Database of Systematic Reviews and Best Evidence. The former including 400 important clinical recommendation and latter consists of ACP Journal Club and Evidence-Based Medicine，mainly provide article reviews of clinical journals.

(b) Data filtering

In this study, the chapter, section, printing format and associated relation of data are defined by database software，those are not relevant to the table of data themselves, so we regard them as “noise” and need to filter them before then. After, we manage the review articles and clinic records of medical chart downloaded from Medline and filter the data won’t be used or incorrect, for example, the Chinese expletive and punctuation marks will be filtered in this step. After this process, data can be present according to system setting and enhance following process of handling。

(c) Data Analysis

In this study, we invite the experts to analyze all review articles and delete irrelevant content.

(d) Define Key-word

When import cumulative data to text mining tool，use this text mining tool to manipulate key-word and vocabulary read frequency. Experts appraise those key-word from the highest ratio of vocabulary read frequency and define specific terms to measure how data sources are identified as key-word vector by the

vocabulary frequency and key-word. The key-word vector is decided by total numbers of key-word among entire data sources from database. The medical literature is immense but only a small fraction of it is useful in answering clinical question. To meet this requirement, the system handles the data sources with the recognition of medical background knowledge to facilitate the future analysis.

(e) Conceptual Grouping

Combined with expert’s opinion to categorize and group the article’s compendium, from where to precise the important key-word as the basis of article category and then classify this article。

(f) Building Knowledge Index

Integrate the analysis result of key-word from text mining and the conclusion from expert’s conceptual grouping, discuss again with expert to set up a base of knowledge index also, create the serial numbers and hierarchy of knowledge components and the relationship of different knowledge components.

(g) Structuring Knowledge Map

Use the text mining methodology to for non-systematic review articles, also take the methodized knowledge map to describe the knowledge retrieved from knowledge map and present it clearly by map browsing to increase the interest of user who can explore those systematic information enhance the guiding effectiveness. In this study, there’re three components in the structure, which are topic, association and occurrence. Start to list the topics and sub-topics of specific domain, connect the associated ones by “see” and “also see” mechanism to transfer non-systematic data resources to a structured knowledge map.

2. System Design (a) System Architecture

Mainly apply 3-tier information architecture of World-Wide Web, the Business Rules processed in the Middle Tier，User Interface in Front Tier and database

software in Rear Tier . (b) System Functionality

There’re 5 major functionalities in this system designed for KM professionals : evidence exploration 、 experience hand down、knowledge map、knowledge exchange and knowledge accumulation. To build up a knowledge map through text-mining tool and expert’s knowledge，further, to select quickly the appropriate medical knowledge from the result and suggestion manipulated by text mining tool. This system is web-base available for the purpose of knowledge access and sharing for all medical professionals via Internet anytime and anywhere. Also, this system can help to realize the knowledge management mechanism and bring the optimal benefits inside medical organization。

(c) System Interface

System including evidence exploration 、 experience hand down、knowledge map、knowledge exchange and knowledge accumulation as described as below : I. Evidence Exploration Interface

Import review articles into database ， medical professionals search from phpAdmin interface to obtain knowledge quickly。

Figure 2 Evidence Exploration Interface

II. Hand down knowledge Interface

When all review articles import to database, medical professionals search from phpAdmin interface to learn the historical knowledge from senior experts.

III. Knowledge Map Interface

After filtering of data and use text mining, combined with expert’s knowledge to structure the knowledge map.

Access appropriate medical knowledge by clicking vocabulary which will be associated relevant articles.

Figure 3 Knowledge Map Interface

IV. Knowledge Exchange Interface

Design the knowledge exchange interface to provide a forum for medial professional to discuss and share clinical cases and experience with others. Medical professionals can also use this section of different topics to discuss the problems encountered.

Figure 4 Knowledge Exchange Interface

V. Knowledge Accumulation Interface

To provide a section of different categories according to topics for medial professionals to discuss, access and share clinical cases and experience via Internet at any time and any places. The goal is to realize the knowledge management mechanism and bring the optimal benefits inside medical organization。

Discussion

During the process of this study, we found that there’re various quality level of medical charts and details during text mining, therefore, we add the patient information and recent review of medical journals and books to enhance the accuracy and evidentiary for text mining.

On the other side, the text-mining tool currently used has been limited by single word functionality; it can be replaced by more powerful text-mining tool in the future.

To keep the capacity of system, we set up a specific medical vocabulary base to improve and keep the index and weighted value can be more accurate. By building the medical knowledge management system, there’re several benefits for hospital management as below : 1. Enhance the insufficiency of EBM：The EBM is essential and can improve the retrieval of high quality level studies applicable to clinical practice. It’s also meaningful to explore the knowledge from senior experts. It’s for sure that to integrate together to bring much benefits for hospitals. In general, to practice the EBM does cost much time and money that increases the barrier for execution. In our study, we integrate the EBM and experts’ clinical experience to build up a medical knowledge management system that can reduce dramatically the cost of money and time consuming and money from traditional process.

2. Improve knowledge sharing inside organization：

After implementation of knowledge management application, the medical information can be accumulated、reuse and provide different departments to use the on-line knowledge management mechanism to design proprietary access control and knowledge map.

The core concept for this system is to enhance the knowledge sharing via on-line learning and emphasis the collaborative learning and also, hope to improve learning efficiency. The key research points are ： create collaborative environment，realize knowledge sharing and exchange；design access control mechanism，ensure the security of knowledge sources；build up knowledge map and distribute the well organized knowledge to users。

3. Reduce risk of medical decision：Use community to develop an expert network,，real-time knowledge transmission，effectively guide for new comers, decrease learning curve，improve learning efficiency，reduce the rate of mistakes probability。Also, to help workers to find the solution from previous experience to reduce the

same mistake happened twice, or, reduce the time for working the same job。

4. Enhance administrative efficiency and teaching quality：The medical knowledge database system offers systematic and useful knowledge ， it can improve productivity even create new medical value, enhance the capability of hospital. For problem solving,，this system can also solve the problem of the information or knowledge hand down between resigned employees and new comers. Moreover, can accumulate the long-tem capability of hospital and provide each new clinician a valuable medical knowledge database also to improve the teaching quality.

5. To accumulate and obtain seniority of e-Medical knowledge：From this system, much assets hidden inside hospital can be explored and developed for clinicians by systematic methodology and to promote at higher level of clinical interests or members in matters relating to training effectiveness.

6. Decrease the time consuming of data search : Use explicit methods to categorize traditional review articles and clinical records in that conclusions are store in the archives of knowledge map, clinicians can select and browse effectively the information on various levels of data sources to help them to retrieve the highest level of evidence for a particular clinical question.

Conclusion

There are immense cumulative descriptive medical records scattered in hospital from where we find the precious clinical information and knowledge. Though we are able to apply them by traditional EBM but cost much manpower and time for validity. Also, the traditional EBM system is still far from completion, it’s often frustrated and uncertain in the scientific research process.

In this study, we use the text mining tool, add methodology and clinical filters assessed by professional groups to subject terms and identify the knowledge map can achieve in the most efficient and optimal retrieval for clinicians while finding the highest level of evidence in

answering clinical questions in time. Further, from this system, we can improve the value and benefits from EBM to enhance inter-institutional knowledge sharing, make fully explicit all the risks and benefit of a decision under conditions of uncertainty. On the other side, with the integration of EBM and experience of experts in various levels, this system brings the useful outcome for the development and accumulation of oral and written traditions in hospital from senior experts to reduce the time consuming of junior clinicians for data searching.

This system also helps clinicians to measure the right medical judgments in time and enhance the clinical diagnosis by clearly articulated evidence of EBM.

Reference

[1] 陳雅琴（2001 ），知識經濟的基本觀念介紹與探

討 — 以 APEC 國 家 的 知 識 經 濟 發 展 經 驗 為 例」，APEC 議題研究，P16 - 29。

[2] 陳杰峰,張錦梅,陳新源,李友專,邱文達（2003），萬

芳醫院實證醫學中心建立與醫療品質改善之應 用，引用自 Sydney Burwell, Dean of Harvard Medical School，台北。

[3] 彭雅惠、董正玫，台灣醫療產業知識管理推展的 現況研究，醫務管理期刊，92 年 3 月第 4 卷第 1 期。

[4] Cochrane AL. Archie Cochrane in his won words.

Selections arranged from his 1972 introuduction to

“Effectiveness and Efficiency: Random Reflections on the Health Service”1972. Controlled Clinical Trials. 1989;10(4):428-33.

[5] Cheng AL et al. Cisplatin, etoposide, and weekly high dose 5-fluorouracil and leucovorin infusion (PE-HDFL)- a very effective regimen with good

patient compliance for advanced gastric cancer.

Anticancer Res 1998;18:1276.

[6] Droon H. Guyatt and Christopher Patterson et al:

Diagnosis of Iron-Deficiency Anemia in the Elderly.

The American Journal of Medicine 1990;88:66-70.

[7] Deogun, J.S.; Sever, H.; Raghavan, V. Structural Abstractions of Hypertext Documents for Web-based Retrieva, Proceeding of Ninth International Workshop on Database and Expert Systems Applications, 1998:385 -390.

[8] Luthi, Dolan, and Ballard ： Evidence-Based Healthcare Quality Management in Obstetrics and Gynecology. Clinical Obstetrics and Gynecdogy 1998;41(2):348-360.

[9] David Sackett: Evidence based Medicine：what it is and what it isn't. Br Med J 1996;312:71-7.