Biostatistics · Healthcare · Business · Dissertation research
50 time-to-event dissertation topics grouped by survival model, from Kaplan–Meier and Cox regression to competing risks, frailty, cure models and survival machine learning.
50 Research topics
10 Model groups
14 Data sources
8 Universities cited
2015–26 Thesis years
By Siddharth Gupta · Statssy
Survival analysis answers a practical question: how long does it take until a defined event occurs? That event can be death, relapse, readmission, staff turnover or a customer leaving.
The most important decisions happen before fitting the model. You need an event definition, a meaningful time origin, sufficient follow-up and a clear approach to censoring. The 50 topics below connect research questions with possible models, relevant datasets and related university theses.
Which survival model do you need? Select an approach based on the event structure and research question, then check assumptions, censoring, data granularity and available follow-up.
How to use this list Identify the event you want to model and decide when follow-up starts. Choose two or three questions from a relevant discipline. Open each source and confirm its event indicator, observation time and censoring rules. Read the linked university thesis, focusing on its handling of censoring, competing events and model diagnostics. Review your shortlist with a supervisor or book a free consultation . Jump to a survival model group 01. Kaplan-Meier and log-rank tests 02. Cox proportional hazards regression 03. Checking proportional hazards and time-varying effects 04. Competing risks 05. Parametric survival and AFT models 06. Frailty and clustered survival 07. Recurrent events and multi-state models 08. Cure models, interval censoring and joint models 09. Survival analysis in business and social science 10. Machine learning for survival data Survival model group 01 / 10
Kaplan-Meier and log-rank tests Start here. Kaplan-Meier curves show how many people are still event-free over time. The log-rank test checks if two or more groups differ.
TOPIC 01 Do survival curves for adult low-grade glioma differ by age, sex and race? Study design Retrospective cohort Survival model Kaplan-Meier curves with log-rank test Dataset SEER Cancer Data (NCI) Free login TOPIC 02 Did cancer survival fall for patients diagnosed during the COVID-19 pandemic? Study design Retrospective cohort Survival model Kaplan-Meier curves by year of diagnosis Dataset SEER Cancer Data (NCI) Free login TOPIC 03 Do married patients live longer after a cancer diagnosis? Study design Retrospective cohort Survival model Log-rank test with stratified Kaplan-Meier curves Dataset SEER Cancer Data (NCI) Free login TOPIC 04 Does performance status at diagnosis predict survival in advanced lung cancer? Study design Clinical cohort Survival model Kaplan-Meier curves with log-rank trend test Dataset R survival package datasets Open TOPIC 05 How long do women in India wait between births, and does it differ by education? Study design Cross-sectional survey with birth histories Survival model Kaplan-Meier curves for birth intervals Dataset DHS Program Data, with NFHS India Free login Survival model group 02 / 10
Cox proportional hazards regression Cox regression is the workhorse of survival analysis. It gives hazard ratios while adjusting for other factors.
TOPIC 06 Does a delay in starting treatment shorten survival in lung cancer? Study design Retrospective cohort Survival model Multivariable Cox regression Dataset SEER Cancer Data (NCI) Free login TOPIC 07 Does social isolation raise the risk of death in US adults? Study design Prospective cohort Survival model Cox regression with adjustment for health status Dataset NHANES Linked Mortality Files (CDC) Open TOPIC 08 Do adults with depression symptoms die earlier than those without? Study design Prospective cohort Survival model Cox regression Dataset NHANES Linked Mortality Files (CDC) Open TOPIC 09 Does chemotherapy improve survival in pancreatic cancer after adjusting for stage? Study design Retrospective cohort Survival model Cox regression with propensity score adjustment Dataset SEER Cancer Data (NCI) Free login TOPIC 10 Does household income predict all-cause mortality in the US? Study design Prospective cohort Survival model Cox regression with age as the time scale Dataset NHANES Linked Mortality Files (CDC) Open Survival model group 03 / 10
Checking proportional hazards and time-varying effects Cox models assume a factor has the same effect at every point in time. These topics test that rule and fix it when it breaks.
TOPIC 11 Does the effect of performance score on survival fade over time in the veterans lung cancer trial? Study design Randomised trial data Survival model Schoenfeld residual test with time-varying coefficients Dataset R survival package datasets Open TOPIC 12 Do lab values measured during an ICU stay predict death better as time-varying covariates? Study design Retrospective cohort Survival model Cox regression with time-varying covariates Dataset MIMIC-IV (PhysioNet) Apply TOPIC 13 Does the effect of stage on colorectal cancer survival weaken after five years? Study design Retrospective cohort Survival model Cox regression with stage by time interaction Dataset SEER Cancer Data (NCI) Free login TOPIC 14 Does time to antibiotics change survival in ICU sepsis patients? Study design Retrospective cohort Survival model Time-varying exposure Cox model to avoid immortal time bias Dataset MIMIC-IV (PhysioNet) Apply TOPIC 15 Does health insurance coverage change mortality in older US adults? Study design Prospective cohort Survival model Stratified Cox regression with a proportional hazards check Dataset NHANES Linked Mortality Files (CDC) Open Survival model group 04 / 10
Competing risks When people can experience more than one type of event, a standard Kaplan-Meier curve overstates risk. Competing risks models fix this.
TOPIC 16 Do older cancer patients die of their cancer or of other causes first? Study design Retrospective cohort Survival model Cumulative incidence functions with Fine-Gray model Dataset SEER Cancer Data (NCI) Free login TOPIC 17 Do adults with diabetes die more often from heart disease than from other causes? Study design Prospective cohort Survival model Cause-specific hazards models Dataset NHANES Linked Mortality Files (CDC) Open TOPIC 18 Do ICU patients leave hospital alive or die first, and what drives each outcome? Study design Retrospective cohort Survival model Competing risks regression for length of stay Dataset MIMIC-IV (PhysioNet) Apply TOPIC 19 Does radiotherapy for breast cancer raise the risk of later heart death? Study design Retrospective cohort Survival model Fine-Gray model with cancer death as a competing event Dataset SEER Cancer Data (NCI) Free login TOPIC 20 Do patients in addiction treatment complete it, drop out or get transferred first? Study design Retrospective cohort Survival model Cumulative incidence with competing discharge reasons Dataset TEDS Data Files (SAMHSA) Open Survival model group 05 / 10
Parametric survival and AFT models Parametric models like Weibull and log-normal assume a shape for survival time. Accelerated failure time models report how much a factor speeds up or slows down the event.
TOPIC 21 Does a Weibull model fit heart failure survival as well as a Cox model? Study design Clinical cohort Survival model Weibull regression compared with Cox using AIC Dataset Heart Failure Clinical Records (UCI) Open TOPIC 22 Do month-to-month contracts shorten how long customers stay? Study design Customer records Survival model Weibull accelerated failure time model Dataset Telco Customer Churn (Kaggle) Free login TOPIC 23 Does maternal education change how long mothers breastfeed in India? Study design Cross-sectional survey Survival model Log-normal accelerated failure time model Dataset DHS Program Data, with NFHS India Free login TOPIC 24 How long do new US firms survive in each sector? Study design Firm cohort Survival model Life tables with Weibull and log-logistic fits Dataset Business Dynamics Statistics (US Census) Open TOPIC 25 Can parametric models predict long-term survival beyond the end of a trial? Study design Randomised trial data Survival model Flexible parametric (Royston-Parmar) extrapolation Dataset R survival package datasets Open View the printable 50-topic guide
Survival model group 06 / 10
Frailty and clustered survival Patients in the same hospital, region or family share risks you cannot measure. Frailty models add a random effect for that shared risk.
TOPIC 26 Do cancer registries differ in patient survival after case mix is taken into account? Study design Retrospective cohort Survival model Shared gamma frailty Cox model by registry Dataset SEER Cancer Data (NCI) Free login TOPIC 27 Does county poverty explain differences in childhood cancer survival? Study design Retrospective cohort Survival model Frailty model with county random effects Dataset SEER Cancer Data (NCI) Free login TOPIC 28 Do repeat kidney infections in the same patient share a hidden risk? Study design Clinical cohort Survival model Shared frailty model for repeated events Dataset R survival package datasets Open TOPIC 29 Do some patients carry a hidden tendency for repeat ICU admissions? Study design Retrospective cohort Survival model Patient-level frailty model Dataset MIMIC-IV (PhysioNet) Apply TOPIC 30 Do employees in the same job role share attrition risk? Study design Employee records Survival model Shared frailty Cox model by job role Dataset IBM HR Analytics Attrition (Kaggle) Free login Survival model group 07 / 10
Recurrent events and multi-state models Some events happen more than once, and some people move through stages. These models follow every event and every move.
TOPIC 31 How often do bladder tumours come back after treatment? Study design Randomised trial data Survival model Andersen-Gill and PWP recurrent event models Dataset R survival package datasets Open TOPIC 32 Who gets readmitted again and again after an overdose? Study design Retrospective cohort Survival model Recurrent event Cox model with sandwich variance Dataset MIMIC-IV (PhysioNet) Apply TOPIC 33 Does a heart transplant change survival once waiting time is modelled as a stage? Study design Clinical cohort Survival model Illness-death multi-state model Dataset R survival package datasets Open TOPIC 34 How fast do people with MGUS move to myeloma, and how many die first? Study design Clinical cohort Survival model Multi-state model with transition hazards Dataset R survival package datasets Open TOPIC 35 Do repeat hospital stays explain part of the effect of a diagnosis on death? Study design Retrospective cohort Survival model Joint frailty model for recurrent events and death Dataset MIMIC-IV (PhysioNet) Apply Survival model group 08 / 10
Cure models, interval censoring and joint models These are the advanced methods. Use them when some patients never have the event, when events fall between visits, or when a biomarker changes over time.
TOPIC 36 What share of colon cancer patients are cured after surgery and chemotherapy? Study design Randomised trial data Survival model Mixture cure model Dataset R survival package datasets Open TOPIC 37 Has the cured share of melanoma patients risen since immunotherapy arrived? Study design Retrospective cohort Survival model Mixture cure model by period of diagnosis Dataset SEER Cancer Data (NCI) Free login TOPIC 38 Does tumour gene expression predict progression found between scheduled scans? Study design Retrospective cohort Survival model Interval-censored survival model Dataset NCI Genomic Data Commons (TCGA) Open TOPIC 39 Does rising bilirubin predict death in primary biliary cholangitis? Study design Clinical cohort Survival model Joint model for longitudinal and survival data Dataset R survival package datasets Open TOPIC 40 How many patients does a trial need to detect a hazard ratio of 0.75? Study design Simulation study Survival model Simulation-based power analysis for time-to-event outcomes Dataset R survival package datasets Open Survival model group 09 / 10
Survival analysis in business and social science Survival analysis works for any time-to-event question. Customers leave, staff resign, firms close and people return to prison.
TOPIC 41 Does overtime shorten how long employees stay? Study design Employee records Survival model Cox regression on years at company Dataset IBM HR Analytics Attrition (Kaggle) Free login TOPIC 42 Does finding a job after prison delay rearrest? Study design Administrative cohort Survival model Discrete-time survival model Dataset NIJ Recidivism Forecasting Challenge Open TOPIC 43 Do young firms exit faster during recessions? Study design Firm cohort Survival model Life tables with piecewise exponential model Dataset Business Dynamics Statistics (US Census) Open TOPIC 44 How long do unemployment spells last, and who stays out of work longest? Study design Household panel Survival model Discrete-time hazard model for spell duration Dataset PSID Free login TOPIC 45 Do first-generation students take longer to finish a degree? Study design Longitudinal cohort Survival model Discrete-time hazard model for time to degree Dataset NLSY97 Free login Survival model group 10 / 10
Machine learning for survival data Machine learning can handle many predictors and complex patterns. Always compare it with a Cox model using the C-index.
TOPIC 46 Do random survival forests predict ICU death better than a Cox model? Study design Retrospective cohort Survival model Random survival forest compared with Cox using the C-index Dataset MIMIC-IV (PhysioNet) Apply TOPIC 47 Which genes predict recurrence and survival in uterine cancer? Study design Retrospective cohort Survival model LASSO-penalised Cox regression Dataset NCI Genomic Data Commons (TCGA) Open TOPIC 48 Does expression of RNA methylation genes predict kidney cancer survival? Study design Retrospective cohort Survival model Cox regression with gene expression cut-points Dataset NCI Genomic Data Commons (TCGA) Open TOPIC 49 Can a deep learning survival model predict heart failure deaths better than Cox? Study design Clinical cohort Survival model DeepSurv compared with Cox using the C-index Dataset Heart Failure Clinical Records (UCI) Open TOPIC 50 Can gradient boosting predict 10-year mortality from a health survey? Study design Prospective cohort Survival model Gradient boosted survival model with calibration checks Dataset NHANES Linked Mortality Files (CDC) Open View the printable 50-topic guide
14 datasets for survival and time-to-event analysis Source access varies: some are direct downloads, some require registration and MIMIC-IV requires approval. Not all provide complete start dates, event times and censoring indicators; review documentation before selecting a survival model.
US cancer registry survival data; use SEER*Stat access and check site, year and treatment variable availability.
De-identified hospital and ICU records. Credentialed access, training and approved data-use conditions apply.
Survey mortality follow-up; check public-use restrictions, sample weights and eligibility.
Teaching and trial datasets such as lung, veteran, colon, bladder, kidney, heart and PBC.
Small clinical event-time dataset including follow-up duration and mortality.
Cancer genomic data and linked clinical information; some files or variables require controlled access.
Birth histories, child health, breastfeeding and reproductive measures; registration and approval of data requests may be required.
Customer tenure and churn labels, with limitations for reconstructing true event histories.
Synthetic employee attrition data; tenure and attrition do not establish individual exit-event histories.
Records of supervision and annual recidivism outcomes for discrete-time prediction.
Aggregate firm entry, exit, age and employment counts by sector and geography.
US longitudinal cohort with schooling, employment and family events.
PSID Free login US longitudinal household data for work, job duration and family outcomes.
Substance-use treatment admissions and discharges; verify the linkage and duration variables.
Printable survival analysis dissertation guide
Get all 50 topics in one PDF A printable guide with all 50 questions, study designs, statistical models, dataset URLs and university thesis references will be linked here after the PDF is created and uploaded to WordPress.
PDF download coming soon How to develop one of these into a dissertation topic Define the clock and event. Specify the time origin, event and censoring assumptions.Plan around the number of events. Base sample size and model complexity on the event fraction, anticipated effect sizes and precision—not a universal events-per-variable threshold.Explore survival curves. Kaplan–Meier plots are often helpful, though competing events require cumulative incidence estimates.Check assumptions. Assess proportional hazards, nonlinearity, informative censoring and competing events as appropriate.Evaluate and report transparently. Include confidence intervals, sensitivity analyses and calibration for prediction models.For related methods, see public health research topics , regression dissertation topics and data science dissertation topics .
Related topic collections Need a second pair of eyes?
Choose a survival model your data supports. Bring your dataset and event definition to a free 15-minute consultation. We can review follow-up time, censoring and the right method for your topic.
Book a free 15-minute call 