TES DIAGNOSTIKTri Nur Kristina

Studi Tes Diagnostik

Desain penelitian untuk mengetahui seberapa baik alat diagnostik dapat membedakan antara pasien yang menderita/ tidak menderita penyakit

Tes Diagnostik

- Memprediksi adaya penyakit

Tes Diagnostik yang ideal memberi jawaban yang benar: Hasil (+) pada org yang menderita penyakit Hasil (-) pada orang yang tidak menderita penyakit

Murah, cepat, aman, sederhana, tidak menyakitkan, reliabilitasnya baik (reliable)

Predictor variable: Hasil tes

Outcome variable: Ada / tidak adanya penyakit

Struktur tes diagnostik

Hasil tes predictor variable:

Dichotomous: Test (+) or (-)

Categorical: Glycosuria (++++); (+++); (++); (+); (-)

Continuous: ..mg of glucose/ dl

Adanya Penyakit sebagai outcome variable:

Outcome variable pada studi tes diagnostik adalah ada/ tidak adanya penyakit yang ditentukan oleh alat diagnostik yang menjadi standar baku emas ( a gold standard)

Standar baku emas selalu (+) pada pasien dengan penyakit dan (-) pada yang tidak menderita penyakit

Sensitivitas: Proporsi dari subjek penelitian dengan penyakit yang memiliki hasil (+) mengindikasikan seberapa baik alat diagnostik yang dites dalam mengidentifikasi orang yang menderita Penyakit

Specifisitas:Proporsi dari subjek penelitian tanpa penyakit yang memiliki hasil (-) mengindikasikan seberapa baik alat diagnostik yang dites dalam mengidentifikasi orang yang tidak menderita penyakit

Sensitivity: a a + cSpecificity: d b + d

Tes Diagnostik(Hasil Test)Gold StandardPenyakit (+)Penyakit (-)

(+)TPaFPb(-)FNcTNd

Choice of a cutoff pointIf the result of diagnstic test yield in continuous data, a decision must be made as to what will constitute a (+) test, a value called cutoff point

This decision requires trading an increase in sensitivity for a decrease in specificity, or vice versa

The investigator must weigh the relative importance of the sensitivity and specificity of the diagnostic test, and set the cutoff point accordingly

One way to do this is to consider the implications of the two possible errors

If false (+) must be avoided (e.g. to determine dangerous surgery), the cutoff point might be set to maximize the test specificity

If false (-) must be avoided (e.g. screening for neonatal phenil ketonuria), the cutoff point should be set to ensure a high test sensitivity

ROC curvesAnother way to set the cutoff point is to use receiver operator characteristic (ROC) curves.

The investigator selects several cutoff points, and determines the sensitivity and specificity at each point

Graph sensitivity as function of (1 specificity) or false (+) rate

20406080100020406080100(50)(25)(400)(200)(100)1- SpecificitySensitivitySerum ALT (U/L) among patients with and without hepatitis

Usually the best cutoff point is where the ROC curve turns the corner in this case, when ALT is about 50 U/L

The curves for different diagnostic test can be compared, in which the better a test, the closer its curve is to the upper left corner

Positive Predictive ValuePPV (Positive Predictive Value): The probability that a person with a (+) test, actually has a disease

a PPV = a + c

Negative Predictive ValueNPV (Negative Predictive Value): The probability that a person with a (-) test, actually does not has a disease

NPV = d c + d

Jumlah sampel untuk Tes Diagnostik N = ( Z2 1- /2 ) P Q d2 P = sensitifitas (diharapkan 80 %) Q = 1 - P = 0,2 d = 0,2 ( hasil penelitian tak jauh dari 20 % dari proporsi yang sebenarnya) Dari tabel, Z2 1-a/2 adalah 1,96 (tingkat kepercayaan 95% )

N = (1,96)2 (0,8) (0,2) = 0,615 = 15,37 ( 0,2 ) 2 0,04

Jumlah sampel dari penderita = 15

Bila Prevalensi penyakit = 35 % ----> Jumlah sampel yang tidak menderita

100 - 35 X 15 = 28 35

Jumlah sampel untuk tes diagnostik = 15 + 28 = 43

The risk of bias increases if the outcome is known to the person who measuring the predictor

Bias even possible if predictor is measured first (new promising tool result +), can influence the person who measuring the gold standard Especially if the test is difficult to classify

Borderline result (common in a person with borderline disease)

Strategy: BLINDINGMeasurement bias

7 Steps in planning a diagnostic test study

Determine whether there is a need for a new diagnostic test

Set the sampling criteria: Describe the way in which patient will be selected

Define the test and the gold standard: Have a reasonable gold standard (feasible, ethical, affordable for both DX test & gold standard)

Gold standard and Diagnostic test can be applied in a standardized and blinded

Estimate the sample size required to achieve 95% CI for the test sensitivity and specificity

6. Ensure that enough subjects are available

7. Report the result in term of sensitivity, specificity, and predictive value

*