Introduction of Evidence Based Medicine 曹龍彥 彰化基督教醫院兒科部

壹.序言 醫學知識的更新愈來愈快，學習如何去access，interpret及apply，變成是clinicians一項很重要的挑戰。因此，Evidence base Medicine(證據醫學)就運應而生。 1992年Gordon Guyatt等人開始在McMaster大學發展及推行EBM。1992年John C. Simclair 及 Mickael B. Bracken編寫了一本EBM形式之textbook of Neonatology “effective care of the newborn infant.

EBM之定義 “The conscientious, explicit, and judicious use of current best evidence in making decisions about the care of individual patients.”。 What is EBM? “EBM is the integration of best research evidence with clinical expertise and patient value”

貳、進行證據醫學之五步驟 步驟一：建立一可回答之臨床問題(Asking answerable question) 1. well formed clinical question: a. patient or problem(P) b. intervention(I) c. comparison intervention (if appropriate)(C) d. Outcome(O)

建立一可回答之臨床問題 2.Type of question clinical finding; etiology; differential diagnosis; diagnostic test; prognosis; therapy; prevention; cost –effectiveness ; quality of life. EX. In women caring babies of 24 to 34 week’s question who are threatening to deliver, does corticosteroid (dexa-methasone) compared with no treatment reduce the incidence RDS in their babies?

步驟二:搜尋最佳証據(finding the best evidence) Sources of evidence： textbooks journals， conference proceedings Experts Electronic source: 如CATs; Best Evidence; CoChrane Library; Bibliographic database(Medline, CINAHL) 因Evidence 不斷的改變及有新的證據，故systematic review也須不斷的更新

步驟二:搜尋最佳証據(finding the best evidence) Primary reports之Evidence 常由MedLine去搜尋, search by using MeSH terms，text words， combining them with “ AND” or “OR”，再加上methodological “filters” strategy 才能搜尋到最佳及不會有很長的list of articles。

步驟二:搜尋最佳証據(finding the best evidence) 例：找Reviews 之article (即systematic reviews of the results of RCTs.): Studies of women with threatened preterm delivery that assess the effect of antenatal steroid on the incidence of RDS： ↓ Type：corticosteroid AND respiratory distress syndrome AND (systematic review OR meta-anal*) 即可找出所要之articles.

步驟二:搜尋最佳証據(finding the best evidence) “4S”hierarchical structure system，如下圖 systems synopsis synthesis studies Computerized decision Support system(CDSS) Evidence-based journal abstract Cochrane reviews Original published articles in journal Examples 高 低 Evidence level of articles

步驟三：嚴謹評估所蒐集之證據 進行評估一篇文章時，可以下面之流程圖來分析： Step1：Selection and sampling Step2：randomization Step3：follow up.須注意 5Cs即-- Contamination, cross-over, compliance, co-intervention, count(loss to follow up)會造成bias. Step4：outcome Step5：analysis External validity Exclusions 1 2 3 4 5 Loss to follow up Co-intervention Internal validity Cross over contamination

(1)Appraising therapy articles: 依V(validity),I(importance), P(practice application)之法則去appraisal. 1.Is the study valid?  clearly defined question? Concealed Randomized. “intention –to-treat” analysis? research participators “blinded” groups treated equally Comparable group at the start of the trial? All patients account for at its conclusion

(1)Appraising therapy articles: 2. Are the results important?   Outcome Exposure Event No event Treated a b Control c d Treatment effect measures RR(relative risk) = a / (a+b) ÷c / (c+d) RRR(relative risk reduction) = 1 － RR OR (odds ratios) = ad / bc Risk difference(RD)= a / (a+b) － c / (c+d) NNT = 1/RD Calculation of 95% confidence interval

(1)Appraising therapy articles: 3. Incorporate your patients values and preferences into deciding on a course of action

(2) Appraising diagnosis articles 1. Is the study valid? Clearly defined question? A validated test (Gold standard test)? Test evaluated on an appropriate spectrum of patient? Reference standard applied to all patients?

(2) Appraising diagnosis articles 2. Are the results important ? “What we thought before “ + “ test information” = “what we think after” 或 pretest probability + likelihood ratio = posttest probability

(2) Appraising diagnosis articles Disease + － Test + Result a True positive b False positive － c False negative d True negative a+b c+d a+c b+d Sensitivity = a / (a+c) ※Mnemonic：PID (Positive In Disease)  Specificity = d / (b+d) ※Mnemonic： NIH (Negative In Health) positive predictive value = a / (a+b)     Negative predictive value = d / (c+d)

(2) Appraising diagnosis articles 但 predictive values change as the likelihood of disease changes .so need “ likelihood ratio” LR 之定義： Likelihood of a particular test result in someone with disease Likelihood of the same test result in someone without the disease *Mnemonic：WOWO (With Over With Out) LR (＋) = Sensitivity =true(+) / false(+) 1 － Specificity LR (-) = 1 － Sensitivity =false(+) / false(-) Specificity

(2) Appraising diagnosis articles 計算posttest probability之Steps： Step1：Estimate the pretest probability Step2：Convert the pretest probability to pretest odds pretest odds = pretest probability 1－ pretest probability Step3：Determine post-test odds 即 post-test odds = pretest odds ×LR (negative or positive) Step4：Converting post-test odds to post-test probability 即 posttest probability = posttest odds 1 ＋ posttest odds

(2) Appraising diagnosis articles 例：predict pyloric stenosis by UltraSound if pretest probability = 46%，LR＋ = 77，LR－ =0.03，then， Post-test probability in Test＋ is [ 0.46/(1-0.46)]×77 = 66，then [ 66/ (1 + 66)] = 99% (87%~100%)。 而 posttest probability in Test－ is [ 0.46/(1-0.46)] ×0.03 = 0.256，then [ 0.0256 / (1+0.0256)] = 2.5% (0% ~ 12%)。 所以，UltraSound in predicting pyloric stenosis 很有用。

(2) Appraising diagnosis articles 3. decide to perform the test or not on the basis of your assessment.

(3) Appraising prognosis studies: 1. Is the study valid? Is the sample representative? Is the follow up long enough for the clinical outcome? Was follow up complete? Outcomes measured “blind”?

(3) Appraising prognosis studies: 2. Are the result important? What is the risk of the outcome overtime? Outcome event Yes No experiment a b control c d Relative risk = {a/ (a+b) ÷ c/(c+d) Odd ratio = (a ÷b)/ (c ÷ d) Rare case為 RR=OR

Odd ratio in case – control study: ={Risk present  risk absent [ case group] } {Risk present  risk absent [ control group]} 1000 comparable patient ↙ ↘ Disease absent (500) Disease present (500) if risk present is 5% If risk present is 20% risk absent is 95%， risk absent is 80% Disease present 20/80 (odds of risk being present)= 0.25 Disease absent 5/95 ( odds of risk being present) = 0.05 So，odds ratio =0.25/0.05=5

(3) Appraising prognosis studies: How precise are the estimate？即計算95% confidence intervals around the odds ratio. ※Standard error(SE) =√(p×(1－p)/n) 其中，p = the proportion of the patients who experience the event 所以，95% CI is P% ±1.96 ×SE(%)。

(4)Appraising articles on harm/etiology: 1. Is the study valid? Clearly defined research question? Similar group patients? Same exposure and clinical outcome measurements Follow up complete and long enough? Causative -link

(4)Appraising articles on harm/etiology: 2. Are the results important? In a cohort study: Relative risk =RR={a/(a+b)}/{ c/ (c+d)} In a case-control study: Odds ratio= OR = ad / bc To calculate the NNH for any OR and PEER (patient expected event rate) NNH=[PEER(OR - 1) +1] / PEER (OR-1) × (1-PEER) 在cohort study 時, 其計方算法與 NNT相同。 即reciprocal of the difference in adverse event rate. exposure Adverse outcome Present (case) Absent (control) Yes a b No c d

(4)Appraising articles on harm/etiology: 3.Are the results applicable to your patients?

(5) Appraising systematic reviews: 1. Is the systematic review valid? High quality studies relevant to your question? Comprehensive search and how the reviewers assess the validity of each study? Are the studies consistent clinically and statistically?

(5) Appraising systematic reviews: 基本上, key features of a good review: a.locate all original articles on the topic of interest b.clinical evaluation of the reports c.conclusion from a synthesis of studies which meet pre-set quality criteria (meta-analysis)

(5) Appraising systematic reviews: 2. Are the results important ? 如review有odds ratio ,則 NNT= 1 － [PEER × (1 － OR )] (1 － PEER) ×PEER ×(1 － OR ) 3.Do the results apply to my patients?

步驟四：須評估一篇valid，important的研究論文是否可以應用在您自己病人？ 考慮： Are your patients similar to those of the study ? The study effect on your patient: a. for diagnostic test : 由 pretest probability (prevalence)算出post-test probability. b. for therapy :先求出PEER再算 NNT(for your patient)=1/ (PEER× RRR)

步驟四：須評估一篇valid，important的研究論文是否可以應用在您自己病人？ Is the intervention realistic in your setting? Does the comparison intervention reflect your current practice? What alternative are available? Are the outcomes appropriate to your patient?

Guidelines Development Cost and Effect Analysis Decision Analysis

Qualitative Research A small child runs in from the garden and says, excitedly, “Mummy, the leaves are falling off the trees.” ”Tell me more”, says his mother. “Well, five leaves fell in the first hours, then 10 leaves fell in the second hour..” That child will become a quantitative researcher!

Qualitative Research A second child, when asked “ tell me more”, might reply, “Well, the leaves are big and flat, and mostly yellow or red, and they seem to be falling off some trees but not others. And, Mummy , why did no leaves fall last month?” That child will become a qualitative researcher!

步驟五：評價與稽核 (Evaluation and Audit)： 對所有上述步驟作一self-evaluation，如對於“問題”的建立；是否“遍尋資訊”；是否會嚴謹評估現有證據；是否應用於實際醫療中？並隨時更新您自己的knowledge？有無參與一EBM的team，不斷的求進步。

參、推展Evidence-Based Clinical practice及其限制 EBM是一種clinical skill，此技術應在大學時期就開始訓練，在住院醫師時就會熟練的應用，即如何問問題、搜尋資料、評估資料及綜合證據，而將最佳的證據應用在病人身上；亦可藉由EBM方法，建立practice guideline及consensus recommendation.

參、推展Evidence-Based Clinical practice及其限制 然而EBM至少有三大限制：(1)要達到高水準的搜尋資料及評論技能，令學習者怯步。(2)忙碌的臨床醫師沒時間去精通及應用這些新技能，且資源常不足。(3)證明EBM “work”需要一段時間。

肆、結論 Evidence-Based Medicine 應是一種最佳的自我學習、終身學習的方法。我們期許會在醫學教育，品質改善，臨床服務，甚至醫學研究影重大的影響。然而，由於臨床上病狀之多樣變化及不確定性，醫師各群體之不協同性，加上電腦資訊程式之快速發展，證據醫學仍存在許多尚待解決之問題，有待大家一起努力。

彰基推展EBM之經驗 1996:賴鈺嘉,黃昭聲院長參加Oxford EBM大會,開始推展EBM。 派VS(內科、外科、小兒科、婦產科、家醫科、牙科)到Oxford, Australia研習。 建立EBM Center 電腦：網站link到England EBM所有期刊、相關書籍

彰基推展EBM之經驗 賴鈺嘉主任到各醫院去演講及院內醫師之教導基本的EBM概念 各科EBM教學 護理部之EBN 藥劑部EBP 內科/小兒科/精神科/家醫科/牙科 護理部之EBN 藥劑部EBP

彰基推展EBM之經驗 PBL+EBM style 之journal reading in 精神科/家醫科/小兒科 PGY1,R3,CR→VS之EBM訓練完成作業 (請見PGY1 36小時一般醫學基礎訓練教材) 讀書會,加強critical appraisal之能力 參加EBM之研習會及自己主辦研習會

彰基推展EBM之經驗 推展EBM之困難 1.觀念之改變 2.不易熟練之上網技術及critical appraisal 3.效果不易在短期間看到 4.院長支持最重要

願景:教育訓練標竿 策略目標 核心成果(落後) 績效驅動因素(領先) 財務 降低學習成本 落實EBM臨床,增加醫院口碑附加價值 教學學習時間 醫院評鑑 顧客 滿足顧客的教育系統(提供可近性教育系統) EBM期刊閱讀參加率 滿意度 內部流程 克服時間限制,提高學習可近性 EBM期刊閱讀教學時間控制 EBM期刊閱讀教學品質 學習 提高講師群能力 講師培訓滿意度 教育資訊之可用性 講師適任率(自我評估) 個人目標配合度 教材數目/教材品質 願景:教育訓練標竿

建議 Epidemiologist、Biostatistician 各科種子教師 硬體設備