*Making Medical Decisions Nishitarumizu 2001.3.9 ***1. Make a clinical estimate (before testing) of the likelihood of the disease as odds. Clue : Odds = Probability /(1- Probability) How do we come up with pre-test probability? Careful history and physical examination, and a few simple labs Knowledge Epidemiology, History (age,gender, prevalence, natural course…) Traditional? textbooks ,studies (Table 1), report Experience Many cases? (last case?) Judgement Clinical reasoning (Heuristics…) ***2. Estimate or calculate the action threshold for the treatment you would advise. &br()Use your patient's subjective preferences in estimating the impact of the various outcomes. Clue : Harm = Probability of harm times its impact score Clue : Improvement = Probability of improvement of disease multiplied by its impact score. Clue : Action threshold = Harm / Improvement (as probability) ***3. Select the test that best suits the scenario if more than one test is available. Seek the test that has the greatest power to change your mind; if your pre-test estimate is above the action threshold , choose the test with the lowest negative likelihood ratio (rule-out). If it is below the action threshold, select the test with the highest positive likelihood ratio(rule-in). These are just the "numbers". If a given test entails disutility such as great cost, risk, pain, or other negative implications, clinical judgment (the final arbiter of all things clinical ) may rightfully override the mathematically "correct" decision in favor of a more humane one. How do we keep up to date with the best evidence concerning tests? Texts such as Ref.2 Electronic texts Best Evidence… Internet PubMed "Clinical Queries" ***4. Calculate what the post-test estimates of likelihood would be if the test results were positive and if the test results were negative. Clue : Post-test Odds = Pre-test Odds × Likelihood Ratio Positive Likelihood Ratio = LR+ = Sensitivity/(1-Specificity) Negative Likelihood Ratio = LR- = (1- Sensitivity) / Specificity Sensitivity: PID , SnNout Specificity: NIH ,SpPin Multilevel likelihood ratios (Table 2) Sequential tests ***5. If your test would take you from below the action threshold (pre-test odds) to above the action threshold ( post-test odds) or from above the action threshold to below the action threshold, go ahead and test. If it would not do so, there is no point in testing because the test results would not change your mind. ***6. Based on the actual test result, if your post-test estimate of disease likelihood is above the action threshold, treat. If below, observe. (例題) A氏が36時間持続する咽頭痛のため来院した.身体所見では38.5℃の発熱と両側の扁桃の 滲出物がみられたが、リンパ節腫脹やその他所見はみられなかった.その他の病歴にも異常 はみられなかった. あなたは以下について考える. 1. 対症療法だけやろうか?何もしないでおくか? 2. 簡易溶連菌検査をやってこれで判断しようか? 3. 検査しないで抗生剤を出しちゃおうか? Step-1: Pretest estimate 臨床所見からは溶連菌性の扁桃炎である可能性は20%.(Table 3 or Ref.2) 咽頭痛のみで熱、扁桃滲出物、前脛部リンパ節腫張のどれもなければ3%未満だが、 前脛部の圧痛を伴うリンパ節腫張があれば、40%以上の確率で溶連菌性の扁桃炎である. Pre-test probability 20% Pre-test Odds 0.25 Step-2: Estimating the action threshold Harm 経口ペニシリンの下痢や皮疹がでる確率は4%.アナフィラキシーの確率は非常に 低い(1/200,000)ので無視できる.その他の副作用1%としても、 Harm frequency 5% Harm impact =0.1 (0-1) Improvement 経口ペニシリンで改善する確率はほぼ100%.これにより溶連菌後腎炎や リウマチ熱の頻度は減るが、これらの頻度は一般には成人では非常に少ないとされている. (小児でも0.4%)そのため治療によるimpactは個人により変動する.例えば元気な学生では 多少の下痢や皮疹はあまり問題にならないのでそのimpactは0. 1ぐらいかもしれないが、 慢性疾患のある高齢者では食事がとれずその効果のimpactは0.9かもしれない. Improvement frequency 100% Improvement impact =0.1 Action threshold= Harm/Improvement = (0.1× 0.05)/ (0.1×1) =0.05 5% Step-3: Assessing Testing 今自分のPre-test probability が20%で、Action thresholdの5%を超えているので 単純に治療したくなるが、ここで検査を行うことで今の自分の気持ちが変わるかどうか を考えてみる.(検査が必要かどうか) 咽頭培養の溶連菌性扁桃炎における感度Sensitivityは90%で、特異度Specificityは85%で ある.すると、(Figure 1) Positive Likelihood Ratio = Sensitivity /(1-Specificity) = 6 Negative Likelihood Ratio = (1- Sensitivity) / Specificity = 0.12 なので Post-test estimate of likelihood (if positive) = 67% Odds 1.5 Post-test estimate of likelihood (if negative) = 3% Odds 0.03 Step-4: Interpreting the test results. もし咽頭培養が陰性なら自分のPre-test probabilityは20%から3%まで低下するため、 Action thresholdの5%を下回ることとなり、この場合治療による危険度のほうが高くなる. (検査の結果によって自分の気持ちが変わり得る) Step-5: Confirming the pre-test estimate. 検査によって自分の考えが変わるということは、自分のpre-test estimateの考え方が正しく なかった可能性はないだろうか? ここで検査を行うかどうか考え、pre-test estimateをもう一度考えることが、今後の自分の 臨床能力をあげることにつながるのでは? (Figure 1) A likelihood ratio nomogram &ref(Making2.jpg) References 1. Making Medical Descisions An approach to clinical decision making for practicing physicians. Richard Gross ACP 2. Diagnostic strategies for common medical problems 2nd. ACP 3. Evidence-based Medicine D.L. Sackett 2nd 4. Learning Clinical Reasoning (Table 3) Strep pharyngitis probability score: Score Interpretation 0 0.64% probability 1 1.71% probability 2 4.53% probability 3 11.42% probability 4 25.95% probability 10 1.34% probability 11 3.55% probability 12 9.1% probability 13 21.39% probability 14 42.52% probability 20 2.96% probability 21 7.65% probability 22 18.38% probability 23 37.97% probability 24 62.47% probability Tonsillar exudate? Yes (1 points) No (0 points) Swollen or tender anterior cervical lymph nodes? Yes (1 points) No (0 points) Fever? Yes (1 points) No (0 points) Absence of cough (except in smokers)? Yes (1 points) No (0 points) Prevalence of group A strep pharyngitis in patient's community 5% (0 points) 10% (10 points) 20% (20 points)