A competing risks model is a survival analysis framework that accounts for events that preclude the occurrence of the primary event of interest. Unlike standard survival models that treat non-primary events as independent censoring, competing risks analysis recognizes that a subject who dies from a non-target cause is no longer at risk for the event under study, requiring distinct statistical handling.
Glossary
Competing Risks Model
A survival analysis framework that accounts for events that preclude the occurrence of the primary event of interest, using metrics like the Cumulative Incidence Function.
SURVIVAL ANALYSIS FRAMEWORK
What is a Competing Risks Model?
A statistical framework for analyzing time-to-event data where subjects are at risk of multiple mutually exclusive events, and the occurrence of one event precludes the occurrence of others.
The framework relies on the Cumulative Incidence Function (CIF) rather than the Kaplan-Meier estimator to avoid overestimation bias. Key submodels include the cause-specific hazard model, which estimates the instantaneous rate of a specific event type, and the Fine-Gray subdistribution hazard model, which directly models covariate effects on the CIF without requiring proportional hazards on cause-specific hazards.
CORE MECHANISMS
Key Features of Competing Risks Models
Competing risks models address scenarios where a subject can experience one of several mutually exclusive events, and the occurrence of one event precludes the others. This framework moves beyond standard survival analysis by decomposing risk into distinct, estimable components.
01
Cause-Specific Hazard
The instantaneous rate of occurrence of a specific event type among subjects who have not yet experienced any event. It is defined as:
h_k(t) = lim(Δt→0) P(t ≤ T < t+Δt, D=k | T ≥ t) / Δt- Represents the direct biological force of mortality or failure from cause
k. - A high cause-specific hazard for one event type does not necessarily translate to a high observed incidence if the competing event occurs first.
02
Cumulative Incidence Function (CIF)
The probability of experiencing a specific event by time t while accounting for the fact that competing events remove subjects from the risk set. Key properties:
CIF_k(t) = P(T ≤ t, D=k)- Unlike 1-KM, the CIF correctly estimates the absolute risk of an event in the presence of competing risks.
- The sum of all CIFs at time
tequals the total probability of having failed from any cause by that time.
03
Subdistribution Hazard (Fine-Gray)
Models the hazard directly linked to the Cumulative Incidence Function rather than the cause-specific hazard. Distinct characteristics:
- The risk set includes subjects who have experienced a competing event, assigning them a time-dependent weight.
- Produces a Subdistribution Hazard Ratio (SHR) that quantifies the relative change in the CIF.
- Preferred when the clinical question is about absolute risk prediction rather than etiological mechanisms.
04
Latent Failure Time Framework
A theoretical construct assuming each subject has a potential failure time for every event type, but only the earliest is observed. Critical limitations:
- Relies on the unverifiable assumption of independence between latent times.
- Non-identifiability problem: observed data cannot distinguish between independent and dependent latent times.
- Modern practice favors the observable quantities approach using cause-specific hazards and CIFs over latent time models.
05
Proportional Subdistribution Hazards Assumption
The Fine-Gray model assumes that the effect of a covariate on the subdistribution hazard is constant over time. Diagnostic checks include:
- Schoenfeld-type residuals plotted against time to detect non-proportionality.
- If violated, consider time-varying coefficients or stratified models.
- A covariate may satisfy proportional cause-specific hazards but violate proportional subdistribution hazards, and vice versa.
06
Direct Binomial Regression
An alternative to the Fine-Gray model that directly models the CIF at a specific time point using a binomial family with a link function (e.g., log, logit). Advantages:
- Avoids the proportional subdistribution hazards assumption entirely.
- Estimates absolute risk differences or risk ratios at clinically meaningful time horizons.
- Particularly useful when interest focuses on a fixed landmark time rather than the entire hazard trajectory.
METHODOLOGICAL COMPARISON
Competing Risks vs. Standard Survival Analysis
Key distinctions between the competing risks framework and standard survival analysis when multiple mutually exclusive event types are present.
| Feature | Standard Survival Analysis | Competing Risks Model | Fine-Gray Subdistribution Model |
|---|---|---|---|
Primary estimand | Survival function S(t) and hazard rate h(t) | Cause-specific hazard and Cumulative Incidence Function (CIF) | Subdistribution hazard and CIF |
Handling of competing events | Censored at time of competing event | Modeled as separate event type | Modeled as part of the risk set |
Risk set definition | Subjects event-free at time t | Subjects event-free at time t | Subjects event-free OR who experienced competing event prior to t |
Interpretation of risk | Direct probability of remaining event-free | Direct probability of experiencing event k in presence of competing events | Direct covariate effects on CIF |
Independence assumption | Assumes non-informative censoring | No independence required between event types | No independence required between event types |
1 - Kaplan-Meier interpretation | Valid when no competing events | Overestimates incidence when competing events present | Not applicable |
Covariate effect interpretation | Cause-specific hazard ratio | Cause-specific hazard ratio | Subdistribution hazard ratio |
Total probability constraint | Not constrained across event types | Sum of CIFs across all event types ≤ 1 | Sum of CIFs across all event types ≤ 1 |
COMPETING RISKS CLARIFIED
Frequently Asked Questions
Clear, technically precise answers to the most common questions about competing risks analysis in survival modeling, designed for clinical trial statisticians and oncology informaticians.
A competing risks model is a survival analysis framework that explicitly accounts for events that preclude or alter the probability of observing the primary event of interest. Unlike standard survival analysis using the Kaplan-Meier estimator or Cox proportional hazards model, which treats all non-primary events as independent censoring, competing risks methodology recognizes that a patient who dies from a non-cancer cause can no longer experience cancer-specific death. The key distinction lies in the estimand: standard survival estimates the hypothetical risk in a world where competing events are removed, while competing risks estimates the observed risk in the real world where multiple event types compete. This is quantified through the Cumulative Incidence Function (CIF), which estimates the probability of experiencing a specific event type by time t while accounting for the presence of other event types. The cause-specific hazard models the instantaneous rate of a particular event among those still at risk, whereas the subdistribution hazard from the Fine-Gray model models the instantaneous rate among those who have not yet experienced the event of interest, including those who experienced a competing event in the risk set.
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
Read more
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
Read more
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
Read more
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.
LinkedIn
Limited slotsGet a Free AI Consultation
Partnered with leading AI, data, and software stack.
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.
01
Review the use case
We understand the task, the users, and where AI can actually help.
Read more02
Pick the right approach
We define what needs search, automation, or product integration.
Read more03
Build the first useful version
We implement the part that proves the value first.
Read more04
Improve from there
We add the checks and visibility needed to keep it useful.
Read moreThe first call is a practical review of your use case and the right next step.
Talk to Us