Download - 資料的評讀 (II) 診斷與篩檢
資料的評讀 (II)診斷與篩檢
神經內科 王志弘
診斷的過程1.1.Initiation of diagnosis hypothesisInitiation of diagnosis hypothesis
初步診斷 我想這病人可能有 。。。
2.2.Refinement of the diagnostic causesRefinement of the diagnostic causes 修正診斷 他可能不是 X 或 Y ,但到底是何種感染呢? Narrowing the possibilities
3.3.Defining the final diagnosisDefining the final diagnosis 最終診斷 我們應該再做個[ XX 切片]確定,再來治療
初步診斷● 目前有上萬種的診斷疾病
● 如果我們不認識這個疾病,我們就不可能考慮到這個診斷
● 關於疾病發生比例的文獻,讓我們了解各種診斷的可能性( pretest probability )
修正診斷● 根據
– Symptoms, 症狀– Signs, 徵象– Laboratory tests, 實驗室檢查– Imaging, 影像檢查
診斷性試驗的實證
Evidence about “diagnostic tests”
● Is this evidence about the accuracy of diagnostic test
validvalid?
● Is this (valid) evidence show that the test is useful at all?
● How can I applyapply this valid, accurate diagnostic test to a specific patient?
文獻評讀
治療性● Validity (closeness to
the truth)● Impact (size of the
effect)● Applicability
(usefulness in our clinical practice)
診斷性● Is this evidence about
the accuracy of diagnostic test valid?
● Is this (valid) evidence show that the test is useful at all?
● How can I apply this valid, accurate diagnostic test to a specific patient?
ValidValidUsefulUsefulApplyApply
什麼是『正常』
BNP vsLV dysfunction
Is this evidence about the Is this evidence about the accuracy of a diagnostic accuracy of a diagnostic
test valid?test valid?
Validity about Diagnostic Tests
●Diagnostic Test in Question
●Reference (gold) standard
常見的 GOLD STANDARDS
1. 外科或是病理標本 2. 血液培養的菌株3. 風溼熱, Jones Criteria
4. DSM IV ( 精神疾病 )
5. X 光6. 長期追蹤
代表性● 該檢查是否在適當的病患族群中被評估過(尤其是
那些在臨床上會使用此一檢查的對象)● Representative● common presentation of the target disorder● confusing presentations
● include patients, with mild and severe, early and late, treated and untreated cases.
確定性● 無 論 檢 查 結 果 如 何 , 參 考 標 準 ( reference
standard )是否經過確認
● 如何達到確定診斷 ?● 另一個參考標準● 長期追蹤● 確定不會延誤治療
測量● Independent and blind measurement
● Psychiatric disorders
Is this (valid) evidence show that the test is
useful at all?
Sensitivity, Specificity,
Likelihood ratios
整體盛行率Prevalence
= ( a + c) / ( a + b + c +d ) = 809 / 2579 = 31%
Pre-test probability
Positive predictive value
● 陽性預測值 = a / (a + b) = 731 / 1001 = 73%● 檢查陽性( ferritin < 65 )的人當中,真正有病 ~
(缺鐵性貧血)的人的比例
Negative predictive value
● 陰性預測值 = d / (c + d) = 1500 / 1578 = 95%● 檢查陰性( ferritin > 65 )的人當中,真正沒有
病 ~ (缺鐵性貧血)的人的比例
Positive vs Negative Predictive Value
Pre-test probability: 測前機率Post-test probability: 測後機率根據前兩張 slide :
– Positive predictive value: 73%– Negative predictive value: 95%
假設抽血檢查前病人 ( 有病 ) 的測前機率: 50%
如果陽性反應 測後機率: 73%
如果陰性反應 測後機率: 1-95% = 5%
Sensitivity
● 敏感度 = a / (a + c) = 731 / 809 = 90%● 真正有病的人當中,檢查有問題(陽性)的比例
Specificity
● 特異度 = d / (b + d) = 1500 / 1700 = 85%● 真正沒病的人當中,檢查沒問題(陰性)的比例
Likelihood Ratio
● 可能性比率● 陽性結果的可能性比率● LR+ = ( 出現目標疾病的病人中,檢查結果為陽
性的可能性) / ( 沒有出現目標疾病的病人中,檢查結果為陽性的可能性)
● = 敏感度 / (1- 特異度 ) ● = 90% / (1-85%) = 6
Likelihood Ratio
● 陰性結果的可能性比率● LR- = ( 出現目標疾病的病人中,檢查結果為陰
性的可能性) / ( 沒有出現目標疾病的病人中,檢查結果為陰性的可能性)
● = (1- 敏感度 ) / 特異度 ● = (1-90%) / 85% = 0.12
勝算 vs 機率• Odds vs Probability
• 假設有病與沒病的機率分別是 31% , 69%
• 則有病的勝算為: 31%/69% = 0.45
• Study pre-test odds = prevalence / (1-prevalence)
Post-test Odds測後勝算
• LR+ = ( 出現目標疾病的病人中,檢查結果為陽性的可能性) / ( 沒有出現目標疾病的病人中,檢查結果為陽性的可能性)
• LR+(ferritin)= sensitivity/(1-specificity) = 90%/15%=6
• 測後勝算 ( 陽性反應 )
• = 測前勝算 * LR+ =1 * 6 = 6
• 測後勝算 ( 陰性反應 ) = 測前勝算 * LR-
Post-test Odds vs Probability
• 測後機率 = 測後勝算 / ( 測後勝算 +1)
• Study pre-test odds = 0.45
• Study post-test odds = 0.45 *6 = 2.7
• Study post-test probability = 2.7 /(2.7+1) = 73%
Post-test Odds , Probability ( 陰性 )
• 測後機率 = 測後勝算 / ( 測後勝算 +1)
• Study pre-test odds = 0.45
• Study post-test odds = 0.45 *0.12 = 0.054
• Study post-test probability( 有病 ) = 0.054/(0.054+1) = 5%
• 測後沒病機率 = 1-5% = 95% negative predictive value
個人化調整病患測後勝算= 研究測後勝算 * ( 病患測前勝算 / 研究測前勝算 )
檢查是否有用Sensitivity 敏感度 , Specificity 特異度兩者相加減掉 100% (Youden IndexYouden Index)
至少要大於 0, 最好要大於 50%, 理想值是 100%
Rule in / Rule out
SnNout:SnNout:
– high sensitivity, negative result rule out the diagnosis
SpPin: SpPin:
– high specificity, positive result rule in the diagnosis
LR+ = sensitivity / (1-specificity)
LR- = (1-sensitivity) /specificity
D-dimer vs Deep Vein ThrombosisSensitivity: 97.7% Specificity: 46%
Ferritin vs Iron deficiency anemiaSensitivity: 90%, specificity: 85%
How can I How can I applyapply this valid, accurate this valid, accurate diagnostic test to a diagnostic test to a specific patient?specific patient?
Is the diagnostic tests
Available
Affordable
Accurate
PreciseIn our setting
Patients Pre-Test Probability
• From personal experience, prevalence statistics, practice database, primary studies
• The study patients similar to our own ?
• If the disease probability changes after the evidence
Results Affect our management?
1. Move us across a test-treatment threshold
2. Our patient willing to carry it out
3. The consequence of the test help the patient reach his/her goal
Multilevel likelihood ratios
Multiple Tests
Prediction rule
SCREENING
Screening
Early diagnosis of pre-symptomatic disease among well individual in public
Case-finding
Early diagnosis of pre-symptomatic disease among patients who came to us for other unrelated diseases
Harm of Early Diagnosis
• Label: high risk for developing some disease
• False positive screening test
• Early diagnosis may not make people live longer, but it surely makes all of them “sick” longer.
Screening 有效嗎 ?
通常願意來接受篩檢的人,對醫囑的遵從性較高 自然預後較好
早期診斷通常會找到進展較慢的疾病
Diagnostic test
診斷不是用來發現絕對的事實
是用來減少臨床診斷的不確定性