Sensitivitas dan Spesifisitas

01The 2x2 Contingency Table: The Foundation of Everything

Before any formula can be understood, you must internalize the 2x2 table. Every single calculation in this topic derives from this table. On your exam, whenever you see a question about diagnostic accuracy, your first action should be to draw this table.

The gold standard (also called the reference standard) is the definitive method of diagnosing the disease. Every other test is compared against it. For example, biopsy is the gold standard for cancer; culture is the gold standard for infection. The gold standard defines the "truth" in the table: who truly has the disease and who does not.

True Positive (TP): The test correctly identifies a person WITH disease. The test says "positive," and the person indeed has the disease.

False Positive (FP): The test incorrectly labels a healthy person as having the disease. The test says "positive," but the person does NOT have the disease. This is a Type I error (alpha error).

False Negative (FN): The test incorrectly labels a diseased person as healthy. The test says "negative," but the person DOES have the disease. This is a Type II error (beta error).

True Negative (TN): The test correctly identifies a person WITHOUT disease. The test says "negative," and the person indeed is disease-free.

The exam will typically give you raw numbers in a vignette and expect you to fill in the table, then compute one or more values from the formulas below. The hardest part is not the math; it is correctly placing the numbers into the right cells. Pay close attention to which number represents the gold standard result and which represents the new test result.

02Core Definitions, Formulas, and Clinical Meaning

Sensitivity (True Positive Rate)

$$Sensitivity=\frac{TP}{TP+FN}=\frac{A}{A+C}Sensitivity\; =\; \backslash frac\{TP\}\{TP\; +\; FN\}\; =\; \backslash frac\{A\}\{A\; +\; C\}$$

Sensitivity answers: Of all people who truly HAVE the disease, what proportion does the test correctly identify as positive?

A test with high sensitivity catches nearly everyone with the disease. It has very few false negatives. This makes it ideal for screening, because you do not want to miss anyone. The clinical mnemonic is SnNOut: when a test with high Sensitivity yields a Negative result, it helps rule Out the disease. The logic: if the test is extremely good at detecting disease, and yet it comes back negative, you can be confident the person likely does not have the disease.

Sensitivity is calculated by reading the table vertically, looking only at the "Disease Present" column. It is a property of the test itself and does not change when you apply the test to populations with different disease prevalence.

Specificity (True Negative Rate)

$$Specificity=\frac{TN}{TN+FP}=\frac{D}{B+D}Specificity\; =\; \backslash frac\{TN\}\{TN\; +\; FP\}\; =\; \backslash frac\{D\}\{B\; +\; D\}$$

Specificity answers: Of all people who truly do NOT have the disease, what proportion does the test correctly identify as negative?

A test with high specificity correctly excludes nearly all healthy individuals. It has very few false positives. This makes it ideal for confirmation, because when it says positive, you can trust it. The clinical mnemonic is SpPIn: when a test with high Specificity yields a Positive result, it helps rule In the disease. The logic: if the test almost never falsely calls a healthy person positive, and yet it comes back positive, you can be confident the person likely has the disease.

Specificity is also read vertically from the table, this time from the "Disease Absent" column. Like sensitivity, it is intrinsic to the test and does not change with prevalence.

The Inverse Relationship

Sensitivity and specificity have an inverse relationship. As you lower the cutoff value for a diagnostic test to increase sensitivity (catch more cases), you will inevitably increase false positives, thus decreasing specificity. As you raise the cutoff to increase specificity (fewer false alarms), you will miss more true cases, thus decreasing sensitivity. This tradeoff is the basis of the ROC curve discussed later.

03Predictive Values: The Clinician's Perspective

While sensitivity and specificity describe the test's accuracy, predictive values describe what the test result means for an individual patient sitting in front of you. This is where exam questions become tricky.

Positive Predictive Value (PPV)

$$PPV=\frac{TP}{TP+FP}=\frac{A}{A+B}PPV\; =\; \backslash frac\{TP\}\{TP\; +\; FP\}\; =\; \backslash frac\{A\}\{A\; +\; B\}$$

PPV answers: If my patient tests positive, what is the probability that they actually have the disease?

PPV is read horizontally from the "Test Positive" row.

Negative Predictive Value (NPV)

$$NPV=\frac{TN}{TN+FN}=\frac{D}{C+D}NPV\; =\; \backslash frac\{TN\}\{TN\; +\; FN\}\; =\; \backslash frac\{D\}\{C\; +\; D\}$$

NPV answers: If my patient tests negative, what is the probability that they truly do not have the disease?

NPV is read horizontally from the "Test Negative" row.

The Critical Role of Prevalence

This is one of the most frequently tested concepts on the exam:

PPV and NPV change with disease prevalence. Sensitivity and specificity do not.

The relationship works as follows:

When prevalence increases (more people in the population have the disease):

• PPV increases (a positive result is more likely to be a true positive, because there are proportionally more diseased individuals)

• NPV decreases (a negative result is less reliable, because there are more diseased people who might be missed)

When prevalence decreases (fewer people have the disease):

• PPV decreases (more of the positive results are false positives from the large pool of healthy individuals)

• NPV increases (a negative result is very reassuring, because there was little disease to begin with)

This is precisely why screening a low-prevalence population with even a moderately good test produces many false positives. The exam loves to test this concept by placing a screening test in a general population versus a high-risk population and asking how the results differ.

04Likelihood Ratios: Combining Sensitivity and Specificity into One Number

Likelihood ratios (LRs) tell you how much a test result changes the probability of disease. They are independent of prevalence (like sensitivity and specificity), but unlike sensitivity and specificity, they combine both metrics into a single clinically useful number.

Positive Likelihood Ratio ($LR+LR+$)

$$LR+=\frac{Sensitivity}{1-Specificity}=\frac{TPrate}{FPrate}LR+\; =\; \backslash frac\{Sensitivity\}\{1\; -\; Specificity\}\; =\; \backslash frac\{TP\backslash \; rate\}\{FP\backslash \; rate\}$$

The $LR+LR+$ tells you: How much more likely is a positive test result in someone WITH the disease compared to someone WITHOUT the disease?

Negative Likelihood Ratio ($LR-LR-$)

$$LR-=\frac{1-Sensitivity}{Specificity}=\frac{FNrate}{TNrate}LR-\; =\; \backslash frac\{1\; -\; Sensitivity\}\{Specificity\}\; =\; \backslash frac\{FN\backslash \; rate\}\{TN\backslash \; rate\}$$

The $LR-LR-$ tells you: How much less likely is a negative test result in someone WITH the disease compared to someone WITHOUT?

A perfect test would have an $LR+LR+$ approaching infinity and an $LR-LR-$ approaching zero.

Using LRs with Pre-Test and Post-Test Probability

In clinical practice and on the exam, LRs connect to Bayesian reasoning:

$$Post\text{-}testodds=Pre\text{-}testodds\times LRPost\backslash text\{-\}test\backslash \; odds\; =\; Pre\backslash text\{-\}test\backslash \; odds\; \backslash times\; LR$$

where $Odds=\frac{Probability}{1-Probability}Odds\; =\; \backslash frac\{Probability\}\{1\; -\; Probability\}$

The pre-test probability is essentially the prevalence or the clinician's clinical estimate of disease probability before ordering the test. The LR modifies this probability. A high $LR+LR+$ shifts the probability upward after a positive result; a low $LR-LR-$ shifts it downward after a negative result.

05The ROC Curve (Receiver Operating Characteristic Curve)

The ROC curve is a graphical representation of the tradeoff between sensitivity and specificity at all possible cutoff values for a diagnostic test.

Key concepts for the exam:

Area Under the Curve (AUC): The AUC quantifies the overall ability of the test to discriminate between diseased and non-diseased individuals. An AUC of 1.0 represents a perfect test. An AUC of 0.5 represents a test with no discriminating ability (equivalent to random chance, visualized as the diagonal line from the bottom-left to the top-right corner). The closer the curve hugs the upper-left corner, the better the test.

Choosing the optimal cutoff point: The point on the ROC curve closest to the upper-left corner is often considered the best balance between sensitivity and specificity. However, clinically the optimal cutoff depends on the consequences of missing a disease (favoring sensitivity) versus the cost of false alarms (favoring specificity).

If the exam asks you to compare two diagnostic tests, the test with the higher AUC is the better overall test.

06Screening vs. Confirmatory Tests: How Exam Vignettes Apply These Concepts

Exam vignettes will frequently present a two-step testing strategy: first a sensitive screening test, then a more precise confirmatory test for those who screen positive. Understand that this is a deliberate sequential strategy, and the exam may ask you WHY we use a sensitive test first (to minimize false negatives and not miss anyone) and WHY we follow up with a specific test (to eliminate false positives and avoid misdiagnosis).

07Indonesian Regulatory Context

The Indonesian Medical Competency Standards (SKDI) have served as the minimum standard for medical education and physician professional competence since they were first ratified by the Indonesian Medical Council (KKI) in 2006. The current edition, SKDI 2019, was compiled based on evaluation and revision of the 2012 standards. The SKDI is endorsed by the KKI and developed by the Indonesian Association of Medical Education Institutions (AIPKI) together with the Indonesian College of Physicians.

The KKI issued an updated edition in August 2024 through Decision No. 193/KKR/KEP/VIII/2024, incorporating new competencies for artificial intelligence, machine learning, big data analytics, and interdisciplinary and transdisciplinary approaches to medical education.

Within SKDI 2024, the Scientific Literacy competency area requires graduates to evaluate data, arguments, and evidence scientifically, draw scientific conclusions in health problem management, and use clinical data and rational investigations to establish a diagnosis. This competency domain is where diagnostic test analysis lives: understanding sensitivity, specificity, predictive values, likelihood ratios, and evidence-based selection of diagnostic tests is a required competency for all Indonesian medical graduates.

The national licensing examination for Indonesian medical graduates is the UKMPPD (Uji Kompetensi Mahasiswa Program Profesi Dokter), which consists of a computer-based test (CBT) and an OSCE (Objective Structured Clinical Examination). Questions on diagnostic test accuracy, the 2x2 table, and the effects of prevalence on PPV/NPV are standard content on the CBT component and fall under the biostatistics and epidemiology blueprint.

08Worked Practice Problem

Vignette: A new rapid test for Dengue NS1 antigen is being evaluated at a Puskesmas. The test is applied to 500 febrile patients. Blood culture and paired serology (gold standard) ultimately confirm Dengue in 200 patients. Results of the rapid test: 180 patients with confirmed Dengue test positive. 30 patients without Dengue also test positive.

Step 1: Build the 2x2 table.

(FN = 200 - 180 = 20; TN = 300 - 30 = 270)

Step 2: Calculate.

$Sensitivity=\frac{180}{180+20}=\frac{180}{200}=0.90=90\mathrm{\%}Sensitivity\; =\; \backslash frac\{180\}\{180\; +\; 20\}\; =\; \backslash frac\{180\}\{200\}\; =\; 0.90\; =\; 90\backslash \%$

$Specificity=\frac{270}{270+30}=\frac{270}{300}=0.90=90\mathrm{\%}Specificity\; =\; \backslash frac\{270\}\{270\; +\; 30\}\; =\; \backslash frac\{270\}\{300\}\; =\; 0.90\; =\; 90\backslash \%$

$PPV=\frac{180}{180+30}=\frac{180}{210}=0.857=85.7\mathrm{\%}PPV\; =\; \backslash frac\{180\}\{180\; +\; 30\}\; =\; \backslash frac\{180\}\{210\}\; =\; 0.857\; =\; 85.7\backslash \%$

$NPV=\frac{270}{270+20}=\frac{270}{290}=0.931=93.1\mathrm{\%}NPV\; =\; \backslash frac\{270\}\{270\; +\; 20\}\; =\; \backslash frac\{270\}\{290\}\; =\; 0.931\; =\; 93.1\backslash \%$

$LR+=\frac{0.90}{1-0.90}=\frac{0.90}{0.10}=9.0LR+\; =\; \backslash frac\{0.90\}\{1\; -\; 0.90\}\; =\; \backslash frac\{0.90\}\{0.10\}\; =\; 9.0$

$LR-=\frac{1-0.90}{0.90}=\frac{0.10}{0.90}=0.111LR-\; =\; \backslash frac\{1\; -\; 0.90\}\{0.90\}\; =\; \backslash frac\{0.10\}\{0.90\}\; =\; 0.111$

Interpretation: The test has good sensitivity and specificity. The $LR+LR+$ of 9.0 provides moderate evidence to rule in the disease. The $LR-LR-$ of 0.111 provides moderate evidence to rule out the disease.

Now the prevalence twist: If this same test is deployed in a general outpatient setting during non-endemic season, where the prevalence of Dengue among febrile patients drops from 40% (200/500) to 5%, what happens? Sensitivity and specificity remain at 90%. However, PPV drops dramatically (because the denominator of "all test-positives" is now dominated by false positives from the large pool of non-Dengue patients), and NPV rises (because there is very little disease to miss).

09Differential Concepts and Distractors

10Traps and High-Yield Pearls

The single most common way students get questions on this topic wrong is by confusing sensitivity/specificity with PPV/NPV. A typical trap vignette will present a highly sensitive and highly precise test, then place it in a very low-prevalence population. The student assumes that because the test has excellent numbers, the positive result is reliable. The correct answer is that the PPV is low in a low-prevalence setting, and most positive results are false positives. The test is still good; it is just being applied to the wrong population. This is the core competency: understanding that the clinical utility of any test depends on the population in which it is used.

Another common trap involves the SnNOut/SpPIn mnemonics applied backwards. Students memorize these shortcuts but then apply them incorrectly. Remember: SnNOut means a sensitive test with a negative result rules out the disease. SpPIn means a specific test with a positive result rules in the disease. A sensitive test with a positive result tells you relatively little (it might just be a false positive). A specific test with a negative result tells you relatively little (it might just be a false negative).

A third trap targets the inverse relationship between sensitivity and specificity. When a vignette describes lowering a diagnostic cutoff (e.g., calling any blood glucose above 100 mg/dL "diabetic" instead of 126 mg/dL), you will capture more true cases (sensitivity increases) but also sweep in more healthy people (specificity decreases). The exam expects you to recognize this tradeoff without being explicitly told, and it may ask what happens to each metric when a threshold changes.

Finally, always remember that sensitivity and specificity describe the performance of the test. PPV and NPV describe the meaning of the result for the patient. The exam is testing whether you can shift between these two perspectives and understand that a test can be excellent in its own right (high sensitivity and specificity) but clinically misleading when applied in the wrong context (low PPV in a low-prevalence population). This conceptual flexibility is the core skill being assessed.