My Journey in the field of Medical Informatics
Yu-Chuan (Jack) Li, M.D., Ph.D., FACMI, FIAHSI, FACHI
2018.08.08
2018.08.08
I started to learn programming language on my own since high school. As a amateur, I published my first software package with my classmate as 4th medical students in Taipei Medical University, Taiwan. In 1988, that was the first software package you can buy that run on MS-DOS to compute personal calories intake/consumption and provide nutritional/exercise suggestions. I was surprised to learn that how little we have tapped from the power of computer to help in the health and medical fields.
I started to write a “differential diagnosis expert system” in 1989 with a group of medical students and we had a rudimentary “knowledge engineering” group that work with the students with the best scores (“experts”) in my class. The whole program can run on a low-density floppy disk with a storage of 360KB and we manage to finish the program with 3 internal medicine diagnoses in the knowledge base when we all went into internship program and never had any time to go further.
I first heard about the University of Utah Medical Informatics program in 1990 and I decided that I had to spend my next several years pursuing my enthusiasm in this field. I enter the program in the fall of 1991 and I was immediately overwhelmed by the progress Dr. Homer Warner has made in his project “Iliad”. I was in awe with the 2,200 diagnostic knowledge frames (both probabilistic and deterministic) he and his group of knowledge engineers and experts had created in the past 7 years. I spend my whole graduate years studying Iliad and eventually developed algorithms to transform its knowledge base into an 11,000 node Bayesian network (probably the largest in the world at that time). As my advisor, Dr. Homer Warner had taught me a lot more than academic research, but also how live a life with integrity and endless curiosity.
I returned to Taipei Medical University in 1995 after I finished my Ph.D. in Utah. With strong support from the school, we started the first Center for Biomedical Informatics (CBI) in Taiwan and then the first dedicated graduate institute for Medical Informatics in Asia in the fall of 1998. Graduate Institute of Medical Informatics (GIMI) has since then educated more than 800 master and doctoral graduate students. Our alumni spreaded across all the major hospitals and government agencies, as well as biomedical industries. GIMI currently has about 100 graduate students (15% international) and 14 full-time faculty members.
I was elected as the president of the Taiwan Association for Medical Informatics (TAMI) at 1999. We had an IMIA GAM meeting in Taipei together with our national MIST (Medical Informatics Symposium in Taiwan) Conference in 2002. In 2003, I was elected as the president of Asia Pacific Association for Medical Informatics (APAMI) and I was dedicated to bringing Medical Informatics education to all the Asia Pacific countries in my 3-year term, focusing on those under-served areas in south-east Asia.
Taiwan’s healthcare providers started a wide-spread computerization effort since 1995, the same year the National Health Insurance Agency (NHIA) was established to implement a universal-coverage type health insurance to the whole population of 23 million people. The patients are free to visit any hospital/clinics and are all covered by the NHIA. No gatekeeper mechanism were in place and a patient is free to visit as many specialists at any time he or she wishes. The NHIA demanded that all insurance claims be electronic, or a surcharge for data entry would be applied. All the 600 hospitals and 12,000 clinics (private and public) were covered by this policy and therefore had to revamp or install computerized hospital/clinical systems. It started mostly from cost-capture and then extended to clinical systems like EHR and CPOE.
In about 5 years, almost all the hospitals and 80% of the clinics are using CPOE as a rule. We started to propose to the government a generic architecture for medical information exchange among all the healthcare providers in 1999 (read more). In 2002, we developed an HL7 CDA R2-based XML template called TMT (Taiwan electronic Medical record Template) that would later evolved into a national standard format for EHR interoperability (read more). In 2004, a Health Smart Card (HSC) infrastructure were implemented nationwide and everyone were each issued a patient’s HSC. All the 45,000 physicians were each issued an additional RSA-version of the HSC called the “physician’s card”. Some key information like allergy, vaccination, past diagnoses, past medications, major illness, major exams, organ donation flags...etc. were stored in the patient’s HSC. When both the patient’s and the physician’s HSC are inserted into the same authenticated NHIA-branded HSC reader, a physician would be able to read these key information from the patient’s HSC and write back some of the relevant information. Some level of continuity of care was achieved at this stage.
The next milestone in terms of Medical Informatics development in Taiwan happened in 2011 when the Taiwan Health Cloud (THC) project debuted. The THC had three components, namely, a Medical cloud, a Care cloud and a Wellness cloud. All under the auspice of the national cloud computing initiative. The Medical cloud later evolved into the PharmaCloud in 2014 - that allows all the physicians to review a 3-month medication history (from all the healthcare providers visited) with the HSC infrastructure as authentication. The medication reconciliation problem were finally resolved by the PharmaCloud approach. After 3 years of PharmaCloud’s introduction, it was further evolved into HealthCloud (2017) that actually provides, in addition to a complete medication history, diagnoses, labs, exams, reports and even imaging studies in real time during the patient encounter. Taiwan has finally achieved a high-level EHR interoperability that is sufficient for most patient encounters to support a safe and sounds continuity of care.
Another achievement is the born of a national personal health record (PHR) for everyone called My Health Bank (MHB). It started as a web-based program that you can log in and check your past visits data and cost in 2015. In the beginning, the service had a low adoption rate due to the clanky way of authentication (you need a card reader at home and insert your HSC) and over-simplified interface. It is still significant since it was the only channel for patients to retrieve their own health data to a end-user device. Fortunately, MHB has also evolved into a smart phone app (IOS and Android) that you can download and authenticate yourself through the phone account and some serial number imprinted on the HSC. This App version of MHB is of much better usability and better user experience. We expect that the adoption rate would rise and there are also an increasing number of privately developed apps looking to value-add the data channeled out of MHB for personal health maintenance. A personal health data value-add marketplace is forming rapidly. Although we are still in the early stage, we do foresee a future that Apps and wearables, along with other IoT (Internet of Things) devices will work together to improve our personal health. MHB did provide a solid foundation for these possibilities to grow.
My journey in the field of Medical Informatics has been very exciting and very rewarding so far. With the awakening of AI (Artificial Intelligence), AR (Augmented Reality) and IoT, Informaticians like ourselves will be in the front-row seats for the real golden age that is coming sooner than later.
I started to write a “differential diagnosis expert system” in 1989 with a group of medical students and we had a rudimentary “knowledge engineering” group that work with the students with the best scores (“experts”) in my class. The whole program can run on a low-density floppy disk with a storage of 360KB and we manage to finish the program with 3 internal medicine diagnoses in the knowledge base when we all went into internship program and never had any time to go further.
I first heard about the University of Utah Medical Informatics program in 1990 and I decided that I had to spend my next several years pursuing my enthusiasm in this field. I enter the program in the fall of 1991 and I was immediately overwhelmed by the progress Dr. Homer Warner has made in his project “Iliad”. I was in awe with the 2,200 diagnostic knowledge frames (both probabilistic and deterministic) he and his group of knowledge engineers and experts had created in the past 7 years. I spend my whole graduate years studying Iliad and eventually developed algorithms to transform its knowledge base into an 11,000 node Bayesian network (probably the largest in the world at that time). As my advisor, Dr. Homer Warner had taught me a lot more than academic research, but also how live a life with integrity and endless curiosity.
I returned to Taipei Medical University in 1995 after I finished my Ph.D. in Utah. With strong support from the school, we started the first Center for Biomedical Informatics (CBI) in Taiwan and then the first dedicated graduate institute for Medical Informatics in Asia in the fall of 1998. Graduate Institute of Medical Informatics (GIMI) has since then educated more than 800 master and doctoral graduate students. Our alumni spreaded across all the major hospitals and government agencies, as well as biomedical industries. GIMI currently has about 100 graduate students (15% international) and 14 full-time faculty members.
I was elected as the president of the Taiwan Association for Medical Informatics (TAMI) at 1999. We had an IMIA GAM meeting in Taipei together with our national MIST (Medical Informatics Symposium in Taiwan) Conference in 2002. In 2003, I was elected as the president of Asia Pacific Association for Medical Informatics (APAMI) and I was dedicated to bringing Medical Informatics education to all the Asia Pacific countries in my 3-year term, focusing on those under-served areas in south-east Asia.
Taiwan’s healthcare providers started a wide-spread computerization effort since 1995, the same year the National Health Insurance Agency (NHIA) was established to implement a universal-coverage type health insurance to the whole population of 23 million people. The patients are free to visit any hospital/clinics and are all covered by the NHIA. No gatekeeper mechanism were in place and a patient is free to visit as many specialists at any time he or she wishes. The NHIA demanded that all insurance claims be electronic, or a surcharge for data entry would be applied. All the 600 hospitals and 12,000 clinics (private and public) were covered by this policy and therefore had to revamp or install computerized hospital/clinical systems. It started mostly from cost-capture and then extended to clinical systems like EHR and CPOE.
In about 5 years, almost all the hospitals and 80% of the clinics are using CPOE as a rule. We started to propose to the government a generic architecture for medical information exchange among all the healthcare providers in 1999 (read more). In 2002, we developed an HL7 CDA R2-based XML template called TMT (Taiwan electronic Medical record Template) that would later evolved into a national standard format for EHR interoperability (read more). In 2004, a Health Smart Card (HSC) infrastructure were implemented nationwide and everyone were each issued a patient’s HSC. All the 45,000 physicians were each issued an additional RSA-version of the HSC called the “physician’s card”. Some key information like allergy, vaccination, past diagnoses, past medications, major illness, major exams, organ donation flags...etc. were stored in the patient’s HSC. When both the patient’s and the physician’s HSC are inserted into the same authenticated NHIA-branded HSC reader, a physician would be able to read these key information from the patient’s HSC and write back some of the relevant information. Some level of continuity of care was achieved at this stage.
The next milestone in terms of Medical Informatics development in Taiwan happened in 2011 when the Taiwan Health Cloud (THC) project debuted. The THC had three components, namely, a Medical cloud, a Care cloud and a Wellness cloud. All under the auspice of the national cloud computing initiative. The Medical cloud later evolved into the PharmaCloud in 2014 - that allows all the physicians to review a 3-month medication history (from all the healthcare providers visited) with the HSC infrastructure as authentication. The medication reconciliation problem were finally resolved by the PharmaCloud approach. After 3 years of PharmaCloud’s introduction, it was further evolved into HealthCloud (2017) that actually provides, in addition to a complete medication history, diagnoses, labs, exams, reports and even imaging studies in real time during the patient encounter. Taiwan has finally achieved a high-level EHR interoperability that is sufficient for most patient encounters to support a safe and sounds continuity of care.
Another achievement is the born of a national personal health record (PHR) for everyone called My Health Bank (MHB). It started as a web-based program that you can log in and check your past visits data and cost in 2015. In the beginning, the service had a low adoption rate due to the clanky way of authentication (you need a card reader at home and insert your HSC) and over-simplified interface. It is still significant since it was the only channel for patients to retrieve their own health data to a end-user device. Fortunately, MHB has also evolved into a smart phone app (IOS and Android) that you can download and authenticate yourself through the phone account and some serial number imprinted on the HSC. This App version of MHB is of much better usability and better user experience. We expect that the adoption rate would rise and there are also an increasing number of privately developed apps looking to value-add the data channeled out of MHB for personal health maintenance. A personal health data value-add marketplace is forming rapidly. Although we are still in the early stage, we do foresee a future that Apps and wearables, along with other IoT (Internet of Things) devices will work together to improve our personal health. MHB did provide a solid foundation for these possibilities to grow.
My journey in the field of Medical Informatics has been very exciting and very rewarding so far. With the awakening of AI (Artificial Intelligence), AR (Augmented Reality) and IoT, Informaticians like ourselves will be in the front-row seats for the real golden age that is coming sooner than later.
李友專,英文名Yu-Chuan (Jack) Li,MD.、Ph.D.,台北醫學大學副校長、醫學資訊研究所教授,台灣醫學資訊協會榮譽理事長,台北市立萬芳醫院原副院長,皮膚科主治醫師。2010年10月當選美國醫學資訊領域最高學術榮譽委員會美國醫學資訊學院(ACMI)院士,成為唯一一位來自亞洲的院士,被稱為台灣醫學資訊之父。
一台兼容機養成了一生興趣
上世紀80年代初,隨著個人電腦逐漸走入家庭,台灣的一些公司也開始推出仿造APPLE II的個人電腦。1981年夏天,初中畢業的李友專正在放暑假,一天父親幫他拿回一張電腦銷售傳單,價格新台幣9,999元。那時他根本不知道什麼是電腦,只在科幻電影裡看到過那些有很多燈,會一閃一閃的東西。於是父親為他買回了第一台電腦Golden II,還特意為他配了一台彩色顯示器。拿到電腦,他迫不及待地按說明書寫出了他平生第一個程式,就是用BASIC畫了一匹馬。「其實那只是螢幕上的幾個亮方塊,勉強可以說像馬,但從此我對電腦產生了濃厚的興趣」李友專說。
從高一到高三,李友專花了很多時間在電腦上,電腦程式編寫水準有了一定的提高。「當時金庸的武俠小說在學生中非常風靡,戲劇性也很高,於是我們就想用BASIC,在APPLE II上把《天龍八部》做成電腦遊戲。而且我們把劇本、角色、人物編輯器和動畫編輯器等等都寫好了。後來到高三了,要準備大學聯考才最終放棄。而那時是1984年,真正的《天龍八部》電腦遊戲在十幾年之後才面世。」
身在北醫,心向電腦
大學聯考填寫志願時,父親希望他當醫生,於是他的前幾個志願是醫學院,後幾個志願都是電腦專業。但李友專第一志願就被台北醫學大學(在台灣也稱北醫)錄取了。「當時我真的有點遺憾,但一個男人畢竟要養家糊口,在台灣當醫生是個很體面的職業,收入也很好,而電腦恐怕只能當興趣了。」
1984年李友專進入台北醫學大學醫學系,一進校他就瞭解到北醫有很多社團,但當時學校不僅沒有電腦社團,就連電腦教室都沒有,於是他就發起成立電腦社。 「當時學校說成立社團要做學生調查,於是我們就做了3,000份調查問卷,沒想到報名的同學都排起隊了。然後我們拿著一份聯署的名單和1,000多份調查表遞交給學校,學校以為是要搞學生運動,馬上就同意了。」
電腦社要做的第一件事就是舉辦暑期電腦培訓班,當時社團只有3台電腦,其中2台是李友專的,另一台是其他同學的。電腦不夠,他們就向全校同學募集;沒有場地,他們就去找補習班借教室。就這樣他們募集了20台電腦,在免費借用的教室開辦了第一期電腦培訓班。「當時人家開玩笑說,創電腦社團要自己帶電腦,那要創游泳社團是不是要自己挖游泳池呀。」電腦社成立後不久,就成為北醫的第二大社團,最多的時候達到100多人,而在那之後一年多北醫才建起了學校的電腦教室。
將IT與醫學相結合的最初目的竟然是少背書
台北醫學大學醫學系是7年制,前4年是醫學基礎課程,第5年開始學習臨床醫學課程,而後兩年則是在醫院實習。「到了五年級,社團已經玩了5年,也覺得有些累了,這時開始想真正做些事情了。當時正好開始學習被譽為內科學聖經的《哈里森內科學》。那本書非常厚,字還非常小,看一遍都要很長時間,就更不要說把每一頁裡的很多內容背下來。不過,當時班裡真有一些同學很能K書(背書),能夠把所有的東西都背下來,我們背後都叫他們‘K棍’。但我和另外幾個同學屬於想東想西的學生,沒有辦法背下來。於是我們就想能不能寫一個程式,把某個疾病的症狀、流行病、診斷和治療等內容寫進去,那輸入一個症狀電腦就可以幫助診斷了,這樣就可以不背那麼多書了。
當時,我們5個人用Turbo Pascal從零開始寫,我自己負責儲存的部分,其他幾個同學負責邏輯的部分,我們做了一張紙本的表格,上面是疾病名稱,下面有幾個格子,每個格子一種症狀,而且標明每種症狀對於疾病診斷重要性的權重分值。然後,我們先請班上最會K書同學幫助填格子,我們自己寫程式,就這樣到五年級結束的時候,我們竟然完成了三個疾病的鑒別診斷,當然這三個疾病是完全不相關的。後來我到了美國才發現,他們已經做了2,200個疾病,而且不僅有權重,還有敏感度和特異性。其實我們講的權重和特異性事實上就是一種條件機率,因此我的博士論文也和條件機率有關。」
醫學是太太,電腦是情人
大學5年級的時候,有人給李友專一張宣傳單,介紹美國的醫學資訊學系。當時的台灣,對於這種在醫學院裡面,把醫學和電腦結合起來的醫學資訊專業瞭解甚少,很多人都沒聽說過。於是,他一邊準備托福和GRE,一邊透過電子郵件聯繫學校(當時還沒有internet,只有一個國際間大學相互連通的學術網BITNET)。
1991年李友專從台北醫學大學畢業,當時有同學問他畢業後要去做哪一科,他說要去美國念醫學資訊。在台灣,醫學系的畢業生都想早點完成專科醫師訓練,就可以行醫或自己開業了。而出國念醫學資訊的還從來沒有過,所以同學對李友專說:「你瘋了」。但當時李友專的父母卻並沒有阻攔他,他們非常瞭解自己兒子的喜好,非常尊重兒子的選擇。就這樣,1991年7月李友專來到美國猶他大學醫學院醫學資訊系攻讀博士學位。
到了美國,李友專驚訝地發現,雖然當初他們做了3個疾病的電腦輔助鑒別診斷軟體已經費了很大力。他的導師,猶他醫學院醫學資訊系創始人Homer R. Warner博士,早在30年前就已經開始做相同的工作了。心臟科醫生出身的Warner博士,每周3次找各科的專科醫師提供相關知識,並且採用更專業和具有邏輯的方法,用了6年的時間完成了2,200多個病種的電腦輔助鑒別診斷。
李友專説:「我當時非常驚訝,也非常興奮,就像看到了美麗新世界。這件事給我留下了深刻的印象,並將影響我的一生。從那時起,我覺得終於可以把IT和醫學結合在一起,成為我畢生的追求,成為可以把大部分時間和精力花在上面的事業。」曾經有人和李友專開玩笑說:「醫學是你的太太,電腦是你的情人。」現在他終於可以把「太太和情人」合二為一了。
具有資訊中心主任頭銜的住院醫師
1990年,李友專在長庚紀念醫院實習時,遇到了一位同樣對電腦非常痴迷的泌尿外科主任。之後,他與這位主任一起做了個腎臟癌症鑒別診斷的軟體,相關論文還在醫學雜誌上發表了。「我覺得很有意思,原來這種論文也可以在正規的醫學雜誌上發表。從那時起我開始真正把電腦和醫學臨床結合起來,而那位主任也成為了我長久的良師益友。」
1995年1月,原本準備去史丹佛做博士後研究員的李友專,被當時台北醫學大學附設醫院副院長,現在台北醫學大學校長邱文達召回台灣,籌建醫院的資訊中心,並被任命為資訊中心主任。1995年7月,醫院皮膚科招收住院醫師,李友專在美國的很多老師都是放射、病理、心臟等科醫師,最後都放棄了臨床專業,專門從事醫療資訊學。「但是我覺得自己學了那麼多年臨床醫學,如果放棄未免可惜。而且如果放棄臨床,將來和醫生溝通就可能產生問題。如果不能好好地瞭解醫生和患者的痛苦,就沒法做出他們真正需要的東西。為此我申請了皮膚科住院醫師,並被錄取。」但是,既要掌管全院的資訊系統,又要完成專科醫生訓練,談何容易。他只能白天做住院醫師,晚上做資訊中心的工作,夜裡值班就睡在電腦下面。即便是這樣,台北醫學大學附設醫院網站還是在1996年4月上線了,成為台灣醫學院校和醫院建立的第一個網站。1997年醫院基於Novel IPX局域和瀏覽器的臨床資訊查詢系統建成,可以實現檢驗、檢查、報告以及超聲波、內視鏡的JPG影像查詢(當時還沒有基於TCP/IP的網路)。
健保推進台灣醫院資訊化建設
在台灣,醫院的資訊化建設大都從收費系統開始。李友專説:如果一個醫院連收費都沒有管好,該收的費都收不上來,那醫院肯定沒法經營。1996年台灣開始實行健康保險,當時健保局規定非常嚴格。如果醫院提交紙本的給付申報資料,每筆需要收取5元登錄費用,而申報電子資料則不收費。5元雖然不多,但如果一天看1,000個病人就是5,000元,看得越多交得越多,而且是永遠收下去。這樣就逼著醫院實現電腦化,建立健保系統,所以台灣的醫院資訊化建設真正是被健保逼上梁山的。
當時台灣健保實行論量計酬,所以系統必須要收集醫院裡的所有資訊,包括一個針頭,一塊紗布,這也是為什麼後來醫院走向精細化全成本核算的原因之一。同時,這也導致醫院的系統變得非常複雜,從而為資訊的再利用打下了基礎。所以,當時台灣的醫院如果能夠實現所有收費都可以被電腦捕捉,並能夠轉成健保所需檔案,然後傳給健保局,最終健保局能夠下載,那麼這家醫院就算實現了完整的電腦化。
台灣醫院資訊系統的建設特點
台灣醫院內的資訊系統大致可分四大類,包括一般行政系統、醫務行政系統、臨床系統和基礎建設系統。 一般行政系統包括人事、會計等系統,這些系統和一般公司使用的系統都是一樣的,所以完全可以外包。醫務行政系統包括醫療保險、ADT系統等等,它們雖是醫院特有的,但大部分醫院都使用相同的系統,這樣的系統或多或少可以找到外包廠商,所以可以自己開發,也可以外包。臨床系統是醫院獨有的,而且每個醫院都不完全一樣,這樣的系統很難找到外包,因此各個醫院大都是自己開發。而基礎建設系統包括電子郵件、網站等系統,這樣的系統完全可以在市場上買到,醫院只要自己維護就可以了。
利用IT技術支持臨床和決策是醫院的關鍵
醫院的核心是臨床醫療和患者服務,如何利用IT技術幫助臨床和服務才是最重要的目標,而這其中的關鍵是臨床決策系統(Clinical Decision Supporting, CDS )。 李友專認為:現在的臨床決策系統做得還很不夠,還有太多CDS可以做的東西卻還沒有做。現在報紙上每年都有醫院開錯藥物劑量的報導,把5cc寫成了5g,把QD寫(一天一次)成QID(一天四次),導致患者藥物過量產生危險。其實在開藥的過程中只要加進去一個簡單的單日最大劑量檢查,進行提示和報警,這些問題就可以解決。目前,台灣一些大醫院已經把CDS做進了系統,但很多小醫院卻還沒有,而正是這些小醫院才更容易出現問題。
李友專說:實現資訊化不是簡單地將資訊以數字化的形式保存到電腦裡,而更重要的是應用。我們現在幾乎已經把所有的資訊存進了電腦,但沒有好好地利用。我們現在是資訊豐富,而知識貧乏。下一步的一個重點就是如何運用更多的知識,來幫助醫療臨床和流程。人一定是會犯錯誤的,尤其是在中國大陸和台灣這樣患者量大,看病密度非常高的醫院,即使有再嚴格的訓練、條例和制度,醫護人員也一定會有所疏忽,會犯錯誤,所以一定要徹底運用CDS。其實,CDS的應用並不一定多先進,即便是很初階、簡單、單一的應用也沒關係,只要這些應用可以減少錯誤,增加安全性就有意義。「我的願望就是所有的醫院實現電腦化之後,要盡全力提高安全性,減少錯誤。如果這裡邊有100分的話,我們已經做的只有5分或10分,剩下的空間還大得不得了,還有很多事情等著我們去做。」
一台兼容機養成了一生興趣
上世紀80年代初,隨著個人電腦逐漸走入家庭,台灣的一些公司也開始推出仿造APPLE II的個人電腦。1981年夏天,初中畢業的李友專正在放暑假,一天父親幫他拿回一張電腦銷售傳單,價格新台幣9,999元。那時他根本不知道什麼是電腦,只在科幻電影裡看到過那些有很多燈,會一閃一閃的東西。於是父親為他買回了第一台電腦Golden II,還特意為他配了一台彩色顯示器。拿到電腦,他迫不及待地按說明書寫出了他平生第一個程式,就是用BASIC畫了一匹馬。「其實那只是螢幕上的幾個亮方塊,勉強可以說像馬,但從此我對電腦產生了濃厚的興趣」李友專說。
從高一到高三,李友專花了很多時間在電腦上,電腦程式編寫水準有了一定的提高。「當時金庸的武俠小說在學生中非常風靡,戲劇性也很高,於是我們就想用BASIC,在APPLE II上把《天龍八部》做成電腦遊戲。而且我們把劇本、角色、人物編輯器和動畫編輯器等等都寫好了。後來到高三了,要準備大學聯考才最終放棄。而那時是1984年,真正的《天龍八部》電腦遊戲在十幾年之後才面世。」
身在北醫,心向電腦
大學聯考填寫志願時,父親希望他當醫生,於是他的前幾個志願是醫學院,後幾個志願都是電腦專業。但李友專第一志願就被台北醫學大學(在台灣也稱北醫)錄取了。「當時我真的有點遺憾,但一個男人畢竟要養家糊口,在台灣當醫生是個很體面的職業,收入也很好,而電腦恐怕只能當興趣了。」
1984年李友專進入台北醫學大學醫學系,一進校他就瞭解到北醫有很多社團,但當時學校不僅沒有電腦社團,就連電腦教室都沒有,於是他就發起成立電腦社。 「當時學校說成立社團要做學生調查,於是我們就做了3,000份調查問卷,沒想到報名的同學都排起隊了。然後我們拿著一份聯署的名單和1,000多份調查表遞交給學校,學校以為是要搞學生運動,馬上就同意了。」
電腦社要做的第一件事就是舉辦暑期電腦培訓班,當時社團只有3台電腦,其中2台是李友專的,另一台是其他同學的。電腦不夠,他們就向全校同學募集;沒有場地,他們就去找補習班借教室。就這樣他們募集了20台電腦,在免費借用的教室開辦了第一期電腦培訓班。「當時人家開玩笑說,創電腦社團要自己帶電腦,那要創游泳社團是不是要自己挖游泳池呀。」電腦社成立後不久,就成為北醫的第二大社團,最多的時候達到100多人,而在那之後一年多北醫才建起了學校的電腦教室。
將IT與醫學相結合的最初目的竟然是少背書
台北醫學大學醫學系是7年制,前4年是醫學基礎課程,第5年開始學習臨床醫學課程,而後兩年則是在醫院實習。「到了五年級,社團已經玩了5年,也覺得有些累了,這時開始想真正做些事情了。當時正好開始學習被譽為內科學聖經的《哈里森內科學》。那本書非常厚,字還非常小,看一遍都要很長時間,就更不要說把每一頁裡的很多內容背下來。不過,當時班裡真有一些同學很能K書(背書),能夠把所有的東西都背下來,我們背後都叫他們‘K棍’。但我和另外幾個同學屬於想東想西的學生,沒有辦法背下來。於是我們就想能不能寫一個程式,把某個疾病的症狀、流行病、診斷和治療等內容寫進去,那輸入一個症狀電腦就可以幫助診斷了,這樣就可以不背那麼多書了。
當時,我們5個人用Turbo Pascal從零開始寫,我自己負責儲存的部分,其他幾個同學負責邏輯的部分,我們做了一張紙本的表格,上面是疾病名稱,下面有幾個格子,每個格子一種症狀,而且標明每種症狀對於疾病診斷重要性的權重分值。然後,我們先請班上最會K書同學幫助填格子,我們自己寫程式,就這樣到五年級結束的時候,我們竟然完成了三個疾病的鑒別診斷,當然這三個疾病是完全不相關的。後來我到了美國才發現,他們已經做了2,200個疾病,而且不僅有權重,還有敏感度和特異性。其實我們講的權重和特異性事實上就是一種條件機率,因此我的博士論文也和條件機率有關。」
醫學是太太,電腦是情人
大學5年級的時候,有人給李友專一張宣傳單,介紹美國的醫學資訊學系。當時的台灣,對於這種在醫學院裡面,把醫學和電腦結合起來的醫學資訊專業瞭解甚少,很多人都沒聽說過。於是,他一邊準備托福和GRE,一邊透過電子郵件聯繫學校(當時還沒有internet,只有一個國際間大學相互連通的學術網BITNET)。
1991年李友專從台北醫學大學畢業,當時有同學問他畢業後要去做哪一科,他說要去美國念醫學資訊。在台灣,醫學系的畢業生都想早點完成專科醫師訓練,就可以行醫或自己開業了。而出國念醫學資訊的還從來沒有過,所以同學對李友專說:「你瘋了」。但當時李友專的父母卻並沒有阻攔他,他們非常瞭解自己兒子的喜好,非常尊重兒子的選擇。就這樣,1991年7月李友專來到美國猶他大學醫學院醫學資訊系攻讀博士學位。
到了美國,李友專驚訝地發現,雖然當初他們做了3個疾病的電腦輔助鑒別診斷軟體已經費了很大力。他的導師,猶他醫學院醫學資訊系創始人Homer R. Warner博士,早在30年前就已經開始做相同的工作了。心臟科醫生出身的Warner博士,每周3次找各科的專科醫師提供相關知識,並且採用更專業和具有邏輯的方法,用了6年的時間完成了2,200多個病種的電腦輔助鑒別診斷。
李友專説:「我當時非常驚訝,也非常興奮,就像看到了美麗新世界。這件事給我留下了深刻的印象,並將影響我的一生。從那時起,我覺得終於可以把IT和醫學結合在一起,成為我畢生的追求,成為可以把大部分時間和精力花在上面的事業。」曾經有人和李友專開玩笑說:「醫學是你的太太,電腦是你的情人。」現在他終於可以把「太太和情人」合二為一了。
具有資訊中心主任頭銜的住院醫師
1990年,李友專在長庚紀念醫院實習時,遇到了一位同樣對電腦非常痴迷的泌尿外科主任。之後,他與這位主任一起做了個腎臟癌症鑒別診斷的軟體,相關論文還在醫學雜誌上發表了。「我覺得很有意思,原來這種論文也可以在正規的醫學雜誌上發表。從那時起我開始真正把電腦和醫學臨床結合起來,而那位主任也成為了我長久的良師益友。」
1995年1月,原本準備去史丹佛做博士後研究員的李友專,被當時台北醫學大學附設醫院副院長,現在台北醫學大學校長邱文達召回台灣,籌建醫院的資訊中心,並被任命為資訊中心主任。1995年7月,醫院皮膚科招收住院醫師,李友專在美國的很多老師都是放射、病理、心臟等科醫師,最後都放棄了臨床專業,專門從事醫療資訊學。「但是我覺得自己學了那麼多年臨床醫學,如果放棄未免可惜。而且如果放棄臨床,將來和醫生溝通就可能產生問題。如果不能好好地瞭解醫生和患者的痛苦,就沒法做出他們真正需要的東西。為此我申請了皮膚科住院醫師,並被錄取。」但是,既要掌管全院的資訊系統,又要完成專科醫生訓練,談何容易。他只能白天做住院醫師,晚上做資訊中心的工作,夜裡值班就睡在電腦下面。即便是這樣,台北醫學大學附設醫院網站還是在1996年4月上線了,成為台灣醫學院校和醫院建立的第一個網站。1997年醫院基於Novel IPX局域和瀏覽器的臨床資訊查詢系統建成,可以實現檢驗、檢查、報告以及超聲波、內視鏡的JPG影像查詢(當時還沒有基於TCP/IP的網路)。
健保推進台灣醫院資訊化建設
在台灣,醫院的資訊化建設大都從收費系統開始。李友專説:如果一個醫院連收費都沒有管好,該收的費都收不上來,那醫院肯定沒法經營。1996年台灣開始實行健康保險,當時健保局規定非常嚴格。如果醫院提交紙本的給付申報資料,每筆需要收取5元登錄費用,而申報電子資料則不收費。5元雖然不多,但如果一天看1,000個病人就是5,000元,看得越多交得越多,而且是永遠收下去。這樣就逼著醫院實現電腦化,建立健保系統,所以台灣的醫院資訊化建設真正是被健保逼上梁山的。
當時台灣健保實行論量計酬,所以系統必須要收集醫院裡的所有資訊,包括一個針頭,一塊紗布,這也是為什麼後來醫院走向精細化全成本核算的原因之一。同時,這也導致醫院的系統變得非常複雜,從而為資訊的再利用打下了基礎。所以,當時台灣的醫院如果能夠實現所有收費都可以被電腦捕捉,並能夠轉成健保所需檔案,然後傳給健保局,最終健保局能夠下載,那麼這家醫院就算實現了完整的電腦化。
台灣醫院資訊系統的建設特點
台灣醫院內的資訊系統大致可分四大類,包括一般行政系統、醫務行政系統、臨床系統和基礎建設系統。 一般行政系統包括人事、會計等系統,這些系統和一般公司使用的系統都是一樣的,所以完全可以外包。醫務行政系統包括醫療保險、ADT系統等等,它們雖是醫院特有的,但大部分醫院都使用相同的系統,這樣的系統或多或少可以找到外包廠商,所以可以自己開發,也可以外包。臨床系統是醫院獨有的,而且每個醫院都不完全一樣,這樣的系統很難找到外包,因此各個醫院大都是自己開發。而基礎建設系統包括電子郵件、網站等系統,這樣的系統完全可以在市場上買到,醫院只要自己維護就可以了。
利用IT技術支持臨床和決策是醫院的關鍵
醫院的核心是臨床醫療和患者服務,如何利用IT技術幫助臨床和服務才是最重要的目標,而這其中的關鍵是臨床決策系統(Clinical Decision Supporting, CDS )。 李友專認為:現在的臨床決策系統做得還很不夠,還有太多CDS可以做的東西卻還沒有做。現在報紙上每年都有醫院開錯藥物劑量的報導,把5cc寫成了5g,把QD寫(一天一次)成QID(一天四次),導致患者藥物過量產生危險。其實在開藥的過程中只要加進去一個簡單的單日最大劑量檢查,進行提示和報警,這些問題就可以解決。目前,台灣一些大醫院已經把CDS做進了系統,但很多小醫院卻還沒有,而正是這些小醫院才更容易出現問題。
李友專說:實現資訊化不是簡單地將資訊以數字化的形式保存到電腦裡,而更重要的是應用。我們現在幾乎已經把所有的資訊存進了電腦,但沒有好好地利用。我們現在是資訊豐富,而知識貧乏。下一步的一個重點就是如何運用更多的知識,來幫助醫療臨床和流程。人一定是會犯錯誤的,尤其是在中國大陸和台灣這樣患者量大,看病密度非常高的醫院,即使有再嚴格的訓練、條例和制度,醫護人員也一定會有所疏忽,會犯錯誤,所以一定要徹底運用CDS。其實,CDS的應用並不一定多先進,即便是很初階、簡單、單一的應用也沒關係,只要這些應用可以減少錯誤,增加安全性就有意義。「我的願望就是所有的醫院實現電腦化之後,要盡全力提高安全性,減少錯誤。如果這裡邊有100分的話,我們已經做的只有5分或10分,剩下的空間還大得不得了,還有很多事情等著我們去做。」
開啟台灣醫療資訊專業
今日北醫|2001.12