Diagnostic Test Calculator: Zero to Hero Tutorial

This tutorial takes you from the foundations of diagnostic testing through sensitivity, specificity, predictive values, likelihood ratios, ROC concepts, calculator use, clinical interpretation, reporting, and common mistakes. It is written for clinical, public health, laboratory, and research contexts where an index test is compared with a reference standard.

Prerequisites and Background Concepts
What Is Diagnostic Test Evaluation?
Diagnostic Study Design
The Mathematics Behind Diagnostic Accuracy
Assumptions and Bias
Using the Diagnostic Test Calculator
Predictive Values and Prevalence
Likelihood Ratios and Clinical Decisions
ROC Curves and Threshold Selection
Worked Examples
Common Mistakes and How to Avoid Them
Troubleshooting
Quick Reference Cheat Sheet

1. Prerequisites and Background Concepts

1.1 Index Test and Reference Standard

The index test is the test being evaluated. The reference standard is the best available method for classifying true disease status. Diagnostic accuracy depends on how well the index test agrees with the reference standard.

1.2 Disease Status and Test Status

Diagnostic test evaluation uses a 2x2 table:

	Disease +	Disease -	Total
Test +	TP	FP	TP+FP
Test -	FN	TN	FN+TN
Total	TP+FN	FP+TN	n

Where:

TP = true positive.
FP = false positive.
FN = false negative.
TN = true negative.

1.3 Screening vs. Diagnosis

Screening is applied to people without known disease to identify who needs follow-up. Diagnosis is applied when disease is suspected. The same test can have different clinical value in these settings because prevalence and consequences differ.

2. What Is Diagnostic Test Evaluation?

Diagnostic test evaluation asks:

How well does this test classify disease compared with a reference standard?

The answer requires more than one metric. Sensitivity and specificity describe test classification performance. Predictive values describe what test results mean in a particular population. Likelihood ratios describe how much a result shifts disease probability.

3. Diagnostic Study Design

3.1 Ideal Design

An ideal diagnostic accuracy study:

Enrolls a representative spectrum of patients.
Applies the index test and reference standard to all participants.
Blinds test interpreters to the other result.
Uses a clinically credible reference standard.
Reports uncertainty with confidence intervals.

3.2 Common Sources of Design Error

Bias	What happens
Spectrum bias	Test evaluated in patients unlike the target population
Verification bias	Only some participants receive the reference standard
Incorporation bias	Index test is part of the reference standard
Review bias	Test readers know the reference result
Selection bias	Diseased and non-diseased groups are sampled unrealistically

4. The Mathematics Behind Diagnostic Accuracy

4.1 Sensitivity

Sensitivity is the proportion of diseased people correctly testing positive:

$Sensitivity = \frac{TP}{TP + FN}$

High sensitivity is useful when missing disease is costly.

4.2 Specificity

Specificity is the proportion of non-diseased people correctly testing negative:

$Specificity = \frac{TN}{TN + FP}$

High specificity is useful when false positives are costly.

4.3 Predictive Values

Positive predictive value:

$PPV = \frac{TP}{TP + FP}$

Negative predictive value:

$NPV = \frac{TN}{TN + FN}$

4.4 Likelihood Ratios

Positive likelihood ratio:

$LR+ = \frac{Sensitivity}{1 - Specificity}$

Negative likelihood ratio:

$LR- = \frac{1 - Sensitivity}{Specificity}$

4.5 Diagnostic Odds Ratio

The diagnostic odds ratio summarizes test discrimination:

$DOR = \frac{TP \times TN}{FP \times FN} = \frac{LR+}{LR-}$

Higher values indicate better discrimination, but DOR is less directly clinical than likelihood ratios.

5. Assumptions and Bias

Diagnostic accuracy estimates assume:

Disease status is classified by a valid reference standard.
The same reference standard is applied consistently.
Test and disease statuses are independent observations.
The study population matches the intended clinical population.
The index test threshold is pre-specified or clearly justified.

When these assumptions fail, accuracy can look better than it will be in practice.

6. Using the Diagnostic Test Calculator

Step-by-Step Guide

Step 1: Define the clinical question.

Example: "How well does a rapid antigen test detect PCR-confirmed infection?"

Step 2: Identify the reference standard.

Use the best available disease classification method.

Step 3: Build the 2x2 table.

	Disease +	Disease -
Test +	TP	FP
Test -	FN	TN

Step 4: Enter raw counts.

Use counts, not percentages.

Step 5: Review all outputs.

Read sensitivity, specificity, PPV, NPV, LR+, LR-, and confidence intervals.

Step 6: Translate the result clinically.

State what a positive and negative result mean in the population studied.

7. Predictive Values and Prevalence

PPV and NPV depend strongly on prevalence:

$PPV = \frac{Sensitivity \times Prevalence}{Sensitivity \times Prevalence + (1-Specificity)(1-Prevalence)}$

$NPV = \frac{Specificity(1-Prevalence)}{(1-Sensitivity)Prevalence + Specificity(1-Prevalence)}$

As prevalence falls, PPV usually falls and NPV usually rises. This is why screening programs can generate many false positives even with a good test.

8. Likelihood Ratios and Clinical Decisions

Likelihood ratios help move from pre-test probability to post-test probability.

Rules of thumb:

Value	Interpretation
LR+ > 10	Strong evidence for disease
LR+ 5 to 10	Moderate evidence for disease
LR+ 2 to 5	Weak evidence for disease
LR+ = 1	No diagnostic value
LR- < 0.1	Strong evidence against disease
LR- 0.1 to 0.2	Moderate evidence against disease

Likelihood ratios are often more transportable than predictive values because they are less directly tied to prevalence.

9. ROC Curves and Threshold Selection

When a test has a continuous score, the threshold determines sensitivity and specificity. Lowering the threshold usually increases sensitivity and lowers specificity.

Youden's Index:

$J = Sensitivity + Specificity - 1$

The threshold with the highest $J$ balances sensitivity and specificity, but clinical costs may justify a different threshold.

AUC interpretation:

AUC	Interpretation
0.5	No discrimination
0.7 to 0.8	Acceptable
0.8 to 0.9	Excellent
> 0.9	Outstanding

10. Worked Examples

Example 1: Rapid Test vs. PCR

	Disease +	Disease -
Test +	90	30
Test -	10	870

$Sensitivity = 90/(90+10) = 0.90$

$Specificity = 870/(870+30) = 0.967$

$PPV = 90/(90+30) = 0.75$

$NPV = 870/(870+10) = 0.989$

$LR+ = 0.90/(1-0.967) = 27.3$

$LR- = (1-0.90)/0.967 = 0.10$

Interpretation: A positive result provides strong evidence for disease, while a negative result provides strong evidence against disease in this population.

Example 2: Low-Prevalence Screening

Assume sensitivity 90%, specificity 95%, and prevalence 1%.

Among 10,000 screened people:

Expected diseased: 100.
True positives: 90.
False negatives: 10.
Non-diseased: 9,900.
False positives: 495.
True negatives: 9,405.

$PPV = 90/(90+495) = 0.154$

Even with a good test, most positive results are false positives when prevalence is low.

11. Common Mistakes and How to Avoid Them

Mistake 1: Treating Sensitivity as PPV

Sensitivity answers "Among diseased people, how many test positive?" PPV answers "Among positive tests, how many truly have disease?"

Mistake 2: Ignoring Prevalence

Predictive values change across populations.

Mistake 3: Using Accuracy Alone

Accuracy can be misleading when disease is rare.

Mistake 4: Evaluating Only Obvious Cases and Healthy Controls

This can exaggerate test performance.

Mistake 5: Choosing a Threshold After Seeing Results Without Saying So

Threshold optimization should be transparent and validated.

12. Troubleshooting

Problem	Likely cause	What to do
PPV is unexpectedly low	Low prevalence	Review expected false positives
Accuracy is high but sensitivity is poor	Disease is rare	Focus on sensitivity and specificity
LR values are undefined	Specificity or sensitivity is 0 or 1	Check sparse cells and exact intervals
Results do not generalize	Spectrum bias	Evaluate in target population
Sensitivity and specificity trade off	Threshold issue	Review ROC curve and clinical costs

13. Quick Reference Cheat Sheet

Metric	Formula	Meaning
Sensitivity	$TP/(TP+FN)$	Detects disease
Specificity	$TN/(TN+FP)$	Excludes non-disease
PPV	$TP/(TP+FP)$	Positive result credibility
NPV	$TN/(TN+FN)$	Negative result credibility
LR+	$Sensitivity/(1-Specificity)$	Evidence from positive test
LR-	$(1-Sensitivity)/Specificity$	Evidence from negative test
Youden's J	$Sensitivity+Specificity-1$	Threshold balance

Reporting Template

Compared with [reference standard], [index test] had sensitivity [value], specificity [value], PPV [value], NPV [value], LR+ [value], and LR- [value] in a population with [prevalence] disease prevalence.

Final Checklist

Define the index test.
Define the reference standard.
Enter TP, FP, FN, and TN correctly.
Report confidence intervals.
Interpret PPV and NPV in context of prevalence.
Discuss verification, spectrum, and threshold bias.

Diagnostic Test Calculator Tutorial