STATISTICAL METHODS
Clinical, Survival & Epidemiological
Statistical Consulting for Clinical Outcomes
Our support with clinical and epidemiological statistical methods matches each analysis to the estimand your study asks about, whether time to an event, a risk or rate, diagnostic accuracy or a causal contrast, from survival models to propensity scores.
Clinical, Survival & Epidemiological is the family of clinical and epidemiological statistical methods that estimate how long until an outcome occurs, how often it occurs, how well a test identifies disease and what an exposure or intervention changes, usually from trial or observational follow-up data. Within Statistical Methods, these methods are separated by the effect measure the question needs, by how follow-up and censoring are handled and by how confounding is addressed. OnlineSPSS is a specialist statistical consultancy for academic, clinical and professional research, and the analysis is carried out in-house by PhD-level statisticians.
METHODS IN THIS CATEGORY
Clinical, Survival & Epidemiological Methods
Twelve methods: time-to-event models for censored follow-up, diagnostic accuracy measures for tests and markers, causal inference methods for observational data, and epidemiological methods for disease frequency.
01
Survival Analysis
Models the time until an event when follow-up is incomplete, handling censoring and describing risk over time with survival and hazard functions.
Time-to-event
Censoring
02
Kaplan-Meier Analysis
Estimates the survival function without a model, shows it as a curve with numbers at risk, and compares groups with the log-rank test.
Survival curves
Log-rank
03
Cox Regression
Relates covariates to the hazard of an event and reports hazard ratios with confidence intervals, after checking proportional hazards.
Hazard ratios
Covariates
04
Competing Risks Analysis
Estimates cumulative incidence of each event when another event can prevent it, and models covariates with cause-specific or Fine-Gray hazards.
Competing events
Fine-Gray
05
ROC Curve Analysis
Plots sensitivity against 1 minus specificity across thresholds and summarizes discrimination with the area under the curve and its interval.
Diagnostic cutoffs
AUC
06
Sensitivity & Specificity (Diagnostic Accuracy)
Compares a binary test with a reference standard, estimating sensitivity and specificity, with predictive values that depend on prevalence.
Test accuracy
STARD
07
Relative Risk & Odds Ratios
Contrasts the risk or odds of a binary outcome between exposure groups, with the effect measure chosen to suit the design and outcome frequency.
Binary outcomes
Effect measures
08
Propensity Score Matching (PSM)
Balances measured confounders between exposed and unexposed groups by matching on the propensity score, then checks overlap and balance.
Confounding
Covariate balance
09
Instrumental Variable Analysis
Addresses unmeasured confounding with an instrument that influences the outcome only through the exposure, under assumptions that are stated.
Instruments
Endogeneity
10
Regression Discontinuity (RD)
Estimates a local effect at the cutoff where a score assigns treatment, by comparing units just above and just below the threshold.
Cutoff designs
Local effect
11
Interrupted Time Series
Models a series measured before and after an intervention to estimate the change in level and trend, allowing for autocorrelation.
Intervention
Time trend
12
Incidence & Prevalence Analysis
Estimates new cases per unit of person-time and existing cases in a population, with confidence intervals and standardization where needed.
Person-time
Prevalence
METHOD COMPARISON
Choosing Between Methods in This Family
Start from the effect measure the question needs, then check how follow-up, confounding and outcome type constrain it; when treatment was not randomized, the design usually matters more than the model.
| Research situation | Typical method |
|---|---|
| Time to an event with censored follow-up | Survival Analysis |
| Survival curves and a group comparison | Kaplan-Meier Analysis |
| Time to event adjusted for several covariates | Cox Regression |
| Another event can prevent the event of interest | Competing Risks Analysis |
| Continuous marker separating cases from non-cases | ROC Curve Analysis |
| Binary test compared with a reference standard | Sensitivity & Specificity (Diagnostic Accuracy) |
| Binary outcome compared between exposure groups | Relative Risk & Odds Ratios |
| Observational exposure with measured confounders | Propensity Score Matching (PSM) |
| Unmeasured confounding and a valid instrument | Instrumental Variable Analysis |
| Treatment assigned by a score cutoff | Regression Discontinuity (RD) |
| One series observed before and after a change | Interrupted Time Series |
| New and existing cases in a defined population | Incidence & Prevalence Analysis |
COMMON ANALYTICAL ISSUES
Common Issues We See in Clinical, Survival & Epidemiological
Most invalid inference in clinical data comes from decisions taken before a model is fit: the time origin, the event definition, the effect measure and the handling of confounding.
Time Origin and Event Definition Follow-up that starts at different points, or an event coded differently across groups, shifts every estimate. Censoring must be told apart from the event, and its reasons recorded, before any curve or model is fit.
Effect Measure Not Matched to the Design Hazard, risk, rate and odds ratios answer different questions. An odds ratio for a common outcome lies further from one than the risk ratio, so the measure must fit the outcome, the design and the follow-up.
Confounding and Missing Follow-Up Adjustment cannot repair unmeasured confounding, and participants lost to follow-up are often not lost at random. Causal wording needs a design or method that supports it, not only an adjusted regression.
Discrimination Mistaken for Accuracy The area under the curve summarizes discrimination, not calibration, and predictive values change with prevalence. Reporting guidelines such as STARD and TRIPOD ask for more than one accuracy figure.
Unsure whether your question needs a hazard, a risk, an accuracy measure or a causal contrast? We can review the design and data and confirm an approach.
ANALYTICAL SUPPORT
From Method Selection to Reportable Results
As part of our Quantitative Data Analysis service, clinical and epidemiological statistical methods are carried from the choice of estimand to results reported with their uncertainty.
Method Specification We confirm the time origin, the event or exposure definition and the estimand, and document why the chosen method answers that question for your design.
Assumptions & Diagnostics Proportional hazards, overlap and balance, instrument strength and calibration are judged through diagnostics and sensitivity analyses, not one significance test.
Effect Estimates & Uncertainty Hazard ratios, cumulative incidence, risk or odds ratios, AUC or treatment effects are reported with confidence intervals, and numbers at risk where relevant.
Publication-Ready Outputs Tables and figures are prepared in line with the CONSORT, STROBE, STARD or TRIPOD items that apply, with scripts or syntax provided where applicable.
SOFTWARE
Statistical Software for Clinical Research
Each analysis is run in the software that fits the method and the project's reproducibility needs, with syntax or scripts provided where applicable.
RESEARCH DISCIPLINES
Clinical Methods Across Research Fields
Clinical research statistical methods are used wherever outcomes are followed or exposures compared, and each field has its own designs, data sources and reporting expectations.
RESEARCHER VOICE
What Researchers Say
Feedback from two researchers, on matching a method plan to the study design, and on analysis built around the clinical question, patient-level data and reporting standards.
“Our clinical dataset had censoring, uneven follow-up, and several possible effect measures, so choosing the right analysis was not straightforward. The team separated the survival, diagnostic, and epidemiological questions clearly and gave us a method plan that matched the study design rather than the software menu.”
“The main value was having statisticians who understood clinical study design, patient-level data, repeated measures, missing follow-up, and reporting standards. We did not have to explain why those details mattered, and the analysis was structured around the clinical question rather than around the software.”
OnlineSPSS provides the agreed methodological, analytical and statistical reporting support and claims no intellectual contribution. The clinical question, interpretation, conclusions and authorship remain with the researcher or research team.
FREQUENTLY ASKED QUESTIONS
Questions About Clinical, Survival & Epidemiological
Short answers to what researchers ask before sharing a clinical or epidemiological project, from choosing a method to confidentiality and the assessment process.
Which method fits my clinical or epidemiological question?
Start from the estimand: the effect measure the question needs, such as a hazard, risk, rate, odds, diagnostic accuracy or a causal contrast, decides the method. Censored time to an event points to survival methods and a test against a reference standard to diagnostic accuracy. If an analysis was already run and questioned, Independent Statistical Review re-examines it.
Can observational data support a causal conclusion?
Only under stated assumptions, and a regression does not supply them. Among observational study methods, propensity scores address measured confounders, an instrumental variable needs a valid instrument, and regression discontinuity needs a clear treatment cutoff. We state the assumptions and use associational wording when the design cannot support more.
How do you protect my research data and confidentiality?
Patient-level files, analysis scripts and study protocols are handled confidentially, and project information is not published. An NDA is available where required and can be arranged before sensitive materials are shared. Read more about confidentiality and data security.
What happens when I request an assessment, and how long does it take?
You share your research question, design and relevant materials, and a statistically trained Research Coordinator reviews the request and responds the same day. The analysis, deliverables, timeline and one fixed project fee are then confirmed before any work begins, so the timeline follows from the agreed scope.
Still have a question? Find more answers in our full FAQ.
NEXT STEP
Need Support Choosing the Right
Clinical Method?
Share your research question, study design and follow-up data. We review the outcome, the effect measure and the design, and confirm the appropriate method before any work begins.
No Commitment Required
Strictly Confidential
Fixed Project Fee
In-House PhD-Level Statisticians
Same-Day Response
Serving Researchers Since 2014