Glossary

Sensitivity

The proportion of actual positive cases correctly identified by a diagnostic test, measuring its ability to avoid false negatives.

Get in touch Learn more

Strategy consultant facilitating AI use case discovery workshop, sticky notes on glass wall, casual corporate meeting.

DIAGNOSTIC ACCURACY METRIC

What is Sensitivity?

Sensitivity is a fundamental statistical measure of a binary classification test's ability to correctly identify true positive cases, quantifying its effectiveness at avoiding false negative errors.

Sensitivity, also known as the true positive rate (TPR) or recall, is the proportion of actual positive cases that a diagnostic test correctly identifies as positive. Mathematically, it is calculated as True Positives / (True Positives + False Negatives). A test with 100% sensitivity perfectly rules out disease when the result is negative, a property known by the acronym SnNout (Sensitive test, Negative result rules out).

In clinical validation study design, sensitivity is always paired with specificity to fully characterize a model's discriminative performance. The metric is threshold-dependent and trades off against the false positive rate across the ROC curve. For life-critical diagnostic AI, high sensitivity is paramount when the consequence of a missed diagnosis is catastrophic, such as in cancer screening or intracranial hemorrhage detection.

DIAGNOSTIC PERFORMANCE METRICS

Core Characteristics of Sensitivity

Sensitivity quantifies a diagnostic test's ability to correctly identify patients who truly have a disease. These core characteristics define its clinical relevance, calculation, and inherent trade-offs.

The Clinical Definition

Sensitivity is the proportion of actual positives correctly identified. It answers the question: 'If a patient has the disease, how likely is the test to return a positive result?' A highly sensitive test is most useful for ruling out disease (SnNout) because a negative result makes the diagnosis unlikely. It is calculated as:

Formula: True Positives / (True Positives + False Negatives)
Ideal Value: 100% (no false negatives missed)

The False Negative Trade-Off

Sensitivity exists in a fundamental inverse relationship with false negatives. A test with 95% sensitivity misses 5% of diseased patients. The clinical cost of a missed diagnosis defines the required sensitivity threshold:

High-stakes screening (e.g., HIV, cancer): Demands sensitivity approaching 99%+ to avoid catastrophic missed cases.
Triage scenarios: Lower sensitivity may be acceptable if a confirmatory test follows.
Threshold tuning: Adjusting the decision boundary to increase sensitivity invariably increases false positives, reducing specificity.

Prevalence Independence

Unlike Positive Predictive Value (PPV) and Negative Predictive Value (NPV), sensitivity is an intrinsic characteristic of the test itself and is theoretically independent of disease prevalence in the population. This makes it a portable metric for comparing tests across different settings. However, spectrum bias can cause apparent sensitivity to vary if the test is evaluated on populations with different disease severity distributions.

Confidence Intervals & Precision

Sensitivity is a point estimate and must be reported with a 95% confidence interval (CI) to convey precision. The CI width is heavily dependent on sample size, specifically the number of true positive cases:

Small sample of diseased subjects: Wide CI, low precision.
Large, well-powered study: Narrow CI, high confidence in the estimate.
Wilson score method is the recommended calculation for binomial proportions like sensitivity, avoiding the errors of the normal approximation interval.

Sensitivity vs. Recall

In machine learning and information retrieval, sensitivity is synonymous with Recall. While 'sensitivity' is the preferred term in clinical diagnostics and epidemiology, 'recall' is used in computer science contexts. Both measure the fraction of relevant instances retrieved:

Clinical context: Sensitivity = TP / (TP + FN)
ML context: Recall = TP / (TP + FN)
Use case: Maximizing recall is critical when missing a positive instance carries a high penalty, such as in fraud detection or disease screening.

The SROC Curve Synthesis

In meta-analyses of diagnostic accuracy, a Summary Receiver Operating Characteristic (SROC) curve is used to synthesize sensitivity and specificity across multiple studies that may have used different thresholds. Unlike a standard ROC curve from a single study, the SROC accounts for threshold effect—the inverse correlation between sensitivity and specificity caused by studies using different cut-points. The area under the SROC curve (SAUC) represents overall diagnostic performance.

SENSITIVITY IN DIAGNOSTIC AI

Frequently Asked Questions

Clear answers to common questions about sensitivity, its calculation, and its critical role in evaluating the clinical performance of diagnostic artificial intelligence systems.

Sensitivity is the proportion of actual positive cases correctly identified by a diagnostic test, measuring its ability to avoid false negatives. It is calculated as True Positives / (True Positives + False Negatives). A highly sensitive test rarely misses a disease when it is present. In medical imaging AI, a model with 99% sensitivity for detecting pulmonary emboli on CT scans will correctly flag 99 out of 100 patients who truly have the condition, with only one false negative slipping through. This metric is crucial for screening applications where missing a diagnosis carries catastrophic consequences.

Enabling Efficiency, Speed & Accuracy

Intelligent Analysis, Decision & Execution

We build AI systems for teams that need search across company data, workflow automation across tools, or AI features inside products and internal software.

Talk to Us

Search across company data

Give teams answers from docs, tickets, runbooks, and product data with sources and permissions.

Useful when people spend too long searching or get different answers from different systems.

Enterprise searchRAGPermissions

Automate internal workflows

Use AI to route work, draft outputs, trigger actions, and keep approvals and logs in place.

Useful when repetitive work moves across multiple tools and teams.

AI agentsWorkflow automationGovernance

Add AI to products and internal tools

Build assistants, guided actions, or decision support into the software your team or customers already use.

Useful when AI needs to be part of the product, not a separate tool.

AI integrationDecision supportModel routing

About the author

Prasad Kumkar

CEO & MD, Inference Systems

Prasad Kumkar is the CEO & MD of Inference Systems and writes about AI systems architecture, LLM infrastructure, model serving, evaluation, and production deployment. Over 5+ years, he has worked across computer vision models, L5 autonomous vehicle systems, and LLM research, with a focus on taking complex AI ideas into real-world engineering systems.

His work and writing cover AI systems, large language models, AI agents, multimodal systems, autonomous systems, inference optimization, RAG, evaluation, and production AI engineering.

Limited slotsGet a Free AI Consultation

How We Work

Custom AI workflows for your Business

One-fit-all AI don't work for modern businesses. At Inferensys, we aim to understand your business & custom requirements; which we use to define most efficient agentic workflows, the data, and the tools for your business.

Talk to Us

Sensitivity

What is Sensitivity?