Applied logistic regression · Biostatistics · Research methods
50 logistic regression dissertation topics, from binary and adjusted models through multinomial, ordinal, conditional, multilevel and penalised logistic regression.
50 Topics
10 Model groups
12 Data sources
3 Universities
2018–26 Thesis years
By Siddharth Gupta · Statssy
Logistic regression models categorical outcomes. Binary, multinomial, ordinal and conditional logistic approaches answer different kinds of questions; survey design and clustering can also change model specifications.
Each topic below is accompanied by a proposed study design, statistical model, starting dataset and related university thesis. Check the data feasibility notes where additional variables or a different cohort may be required.
Which logistic model do you need? How to use this list Identify the outcome variable and decide whether it is binary, nominal or ordinal. Choose two or three topics and inspect the original data codebooks for eligibility and measured variables. Check event counts, sample size, missingness and the proposed covariates. Read the related university thesis for methodology and reporting examples. Review your options with your supervisor or book a free consultation . Jump to a model group 01. Binary logistic regression 02. Adjusting for confounders 03. Interaction and effect modification 04. Multinomial logistic regression 05. Ordinal logistic regression 06. Case-control studies and conditional logistic regression 07. Multilevel logistic regression and GEE 08. Rare events and penalised logistic regression 09. Prediction, classification and model evaluation 10. Survey weights, prevalence and propensity scores Model category 01 / 10
Binary logistic regression Start here. Binary logistic regression explains a yes or no outcome and reports odds ratios.
TOPIC 01 Is HPV vaccination lower in states that voted Republican? Study design Cross-sectional survey Statistical model Binary logistic regression with state predictors Suggested dataset BRFSS (CDC) Open Data feasibility: BRFSS needs state election results merged in, and HPV vaccination measures checked by survey year.
TOPIC 02 Does vaccine hesitancy differ by race and ethnicity? Study design Cross-sectional survey Statistical model Binary logistic regression Suggested dataset Pew Research Center Datasets Free login + BRFSS (CDC) Open Data feasibility: Pew vaccine hesitancy and BRFSS vaccination behaviour may not measure the same outcome; specify a compatible question.
TOPIC 03 Which children miss doses in multi-dose vaccine series? Study design Cross-sectional survey Statistical model Binary logistic regression with survey weights Suggested dataset NHANES (CDC) Open Data feasibility: NHANES does not offer the same immunisation-history coverage as the National Immunization Survey–Child for ages 19–35 months.
TOPIC 04 Are immigrants less likely to receive the HPV vaccine? Study design Cross-sectional survey Statistical model Binary logistic regression by nativity Suggested dataset NHANES (CDC) Open TOPIC 05 Does knowing about cervical cancer raise screening uptake among women? Study design Cross-sectional survey Statistical model Binary logistic regression Suggested dataset DHS Program Data, with NFHS India Free login Data feasibility: DHS may not jointly measure cervical cancer knowledge and screening in the selected survey.
Model category 02 / 10
Adjusting for confounders A crude odds ratio can mislead. Multiple logistic regression adjusts for other factors so the main effect is fair.
TOPIC 06 Is oral HPV infection linked to mouth and throat cancer risk factors? Study design Cross-sectional survey Statistical model Adjusted logistic regression Suggested dataset NHANES (CDC) Open TOPIC 07 Does having health insurance improve blood pressure control? Study design Cross-sectional survey Statistical model Multiple logistic regression with confounder adjustment Suggested dataset NHANES (CDC) Open TOPIC 08 Is serious psychological distress linked to heart disease after adjustment? Study design Cross-sectional survey Statistical model Multiple logistic regression Suggested dataset BRFSS (CDC) Open Data feasibility: BRFSS does not consistently include a validated serious psychological distress scale.
TOPIC 09 Is caesarean birth linked to early child development delays? Study design Cross-sectional survey Statistical model Adjusted logistic regression Suggested dataset DHS Program Data, with NFHS India Free login TOPIC 10 Do race and income predict whether prostate cancer gets guideline treatment? Study design Retrospective cohort Statistical model Multiple logistic regression Suggested dataset SEER Cancer Data (NCI) Free login Data feasibility: SEER generally lacks individual household income and a complete guideline-concordance measure.
Model category 03 / 10
Interaction and effect modification Some effects are stronger in one group than another. Interaction terms test this directly.
TOPIC 11 Do women’s gender attitudes change their risk of partner violence differently by region? Study design Cross-sectional survey Statistical model Logistic regression with interaction terms Suggested dataset DHS Program Data, with NFHS India Free login TOPIC 12 Is arthritis more strongly linked to depression in women than men? Study design Cross-sectional survey Statistical model Logistic regression with sex interaction Suggested dataset NHANES (CDC) Open TOPIC 13 Does discrimination make depression worse for postpartum health care use? Study design Cross-sectional survey Statistical model Logistic regression with product terms Suggested dataset BRFSS (CDC) Open Data feasibility: BRFSS does not routinely contain pregnancy discrimination alongside postpartum healthcare utilisation.
TOPIC 14 Is depression more strongly linked to high blood pressure in some racial groups? Study design Longitudinal cohort Statistical model Logistic regression with race interaction Suggested dataset Add Health Free login TOPIC 15 Does marital stress affect heart symptoms more in women than men? Study design Cross-sectional survey Statistical model Logistic regression with sex by stress interaction Suggested dataset NHANES (CDC) Open Data feasibility: NHANES does not generally measure marital stress and post-MI angina burden together.
Model category 04 / 10
Multinomial logistic regression When the outcome has three or more unordered categories, use multinomial logistic regression.
TOPIC 16 Do smoking, vaping and edibles differ by sexual identity among cannabis users? Study design Cross-sectional survey Statistical model Multinomial logistic regression Suggested dataset BRFSS (CDC) Open TOPIC 17 Who becomes housed, unstably housed or homeless in young adulthood? Study design Longitudinal cohort Statistical model Multinomial logistic regression Suggested dataset Add Health Free login TOPIC 18 Are cannabis users more likely to have anxiety, panic or both? Study design Cross-sectional survey Statistical model Multinomial logistic regression Suggested dataset NHANES (CDC) Open Data feasibility: NHANES may not provide appropriate clinical panic/anxiety categories; check the underlying instruments.
TOPIC 19 Which lifestyle changes do people with prediabetes make, if any? Study design Cross-sectional survey Statistical model Multinomial logistic regression on change type Suggested dataset NHANES (CDC) Open Data feasibility: NHANES does not follow the same people after prediabetes diagnosis for measured lifestyle changes.
TOPIC 20 Does race predict which type of weight-loss surgery patients receive? Study design Registry cohort Statistical model Multinomial logistic regression on surgery type Suggested dataset MIMIC-IV (PhysioNet) Apply Data feasibility: MIMIC-IV is not a bariatric surgery registry. Procedure-specific sample size and outcomes require checking.
Model category 05 / 10
Ordinal logistic regression When categories have an order, like mild, moderate and severe, ordinal logistic regression keeps that order.
TOPIC 21 Do patients and clinicians agree on care goals, and does disagreement predict return visits? Study design Clinical cohort Statistical model Ordinal logistic regression on agreement level Suggested dataset MIMIC-IV (PhysioNet) Apply Data feasibility: MIMIC-IV lacks routine paired patient–clinician goal-of-care agreement surveys.
TOPIC 22 Does feeling unsafe at school predict worse mental health ratings? Study design Cross-sectional survey Statistical model Proportional odds model Suggested dataset General Social Survey (GSS) Open + Add Health Free login TOPIC 23 What predicts higher anxiety levels before a medical procedure? Study design Cross-sectional survey Statistical model Ordinal logistic regression with proportional odds test Suggested dataset NHANES (CDC) Open Data feasibility: NHANES does not routinely assess pre-procedure anxiety.
TOPIC 24 Are health workers who face workplace violence more likely to drink at riskier levels? Study design Cross-sectional survey Statistical model Ordinal logistic regression on drinking risk levels Suggested dataset BRFSS (CDC) Open Data feasibility: BRFSS does not typically record workplace violence among nurses.
TOPIC 25 Which risk factors predict mild, severe or critical COVID-19? Study design Hospital cohort Statistical model Ordinal logistic regression on severity Suggested dataset MIMIC-IV (PhysioNet) Apply Model category 06 / 10
Case-control studies and conditional logistic regression Case-control studies compare people with and without a disease. Matched designs need conditional logistic regression.
TOPIC 26 Which birth and family factors raise the risk of childhood brain tumours? Study design Case-control study Statistical model Unconditional logistic regression Suggested dataset SEER Cancer Data (NCI) Free login Data feasibility: SEER has cancer cases, not the comparable population controls and birth exposures for this study.
TOPIC 27 Is high birth weight linked to early-onset lymphoma? Study design Population-based case-control study Statistical model Conditional logistic regression Suggested dataset SEER Cancer Data (NCI) Free login Data feasibility: SEER does not provide matched non-cancer controls and individual birth weights.
TOPIC 28 Is a past skin condition linked to later pancreatic cancer? Study design Matched case-control study Statistical model Conditional logistic regression Suggested dataset MIMIC-IV (PhysioNet) Apply TOPIC 29 Which antibiotics raise the risk of C. difficile infection in hospital? Study design Matched case-control study Statistical model Conditional logistic regression Suggested dataset MIMIC-IV (PhysioNet) Apply TOPIC 30 Do racial gaps exist in the timing of cleft lip repair? Study design Registry cohort Statistical model Logistic regression with matched comparison Suggested dataset MIMIC-IV (PhysioNet) Apply Data feasibility: MIMIC-IV is not a representative paediatric cleft-lip repair registry.
Model category 07 / 10
Multilevel logistic regression and GEE When people are grouped in households, villages or counties, multilevel models and GEE give correct standard errors.
TOPIC 31 Which county traits predict low school vaccination rates? Study design County panel Statistical model Multilevel logistic regression Suggested dataset BRFSS (CDC) Open Data feasibility: BRFSS adult surveys cannot yield kindergarten vaccination rates.
TOPIC 32 How much does worm infection vary between villages? Study design Cross-sectional survey Statistical model Multilevel logistic regression with community random effects Suggested dataset DHS Program Data, with NFHS India Free login Data feasibility: DHS/NFHS generally lacks laboratory-confirmed hookworm infections.
TOPIC 33 Which household traits raise malaria risk in children? Study design Cross-sectional survey Statistical model Two-level logistic regression Suggested dataset DHS Program Data, with NFHS India Free login TOPIC 34 How common is malaria without symptoms, and where is it concentrated? Study design Cross-sectional survey Statistical model Multilevel logistic regression with cluster effects Suggested dataset DHS Program Data, with NFHS India Free login Data feasibility: Asymptomatic malaria requires both test-confirmed infection and concurrent symptom measures.
TOPIC 35 Do teenage risk behaviours predict outcomes over several survey waves? Study design Longitudinal cohort Statistical model GEE logistic regression with time-varying covariates Suggested dataset Add Health Free login Model category 08 / 10
Rare events and penalised logistic regression Rare outcomes and many predictors break standard logistic regression. Firth and LASSO methods fix this.
TOPIC 36 Which environmental factors predict counties with new Lyme disease cases? Study design County panel Statistical model Firth logistic regression for rare events Suggested dataset BRFSS (CDC) Open Data feasibility: BRFSS does not include county Lyme disease incidence and environmental exposure measures.
TOPIC 37 Which patients receive intensive care at the end of life? Study design Retrospective cohort Statistical model Penalised logistic regression Suggested dataset MIMIC-IV (PhysioNet) Apply TOPIC 38 Which household contacts of TB patients miss their screening? Study design Retrospective cohort Statistical model Firth logistic regression Suggested dataset DHS Program Data, with NFHS India Free login Data feasibility: DHS does not follow household TB contacts through screening completion.
TOPIC 39 Does early-life weight predict infections in young children? Study design Longitudinal cohort Statistical model LASSO logistic regression Suggested dataset NHANES (CDC) Open Data feasibility: NHANES repeated cross-sections are not a longitudinal early-childhood cohort.
TOPIC 40 Can rare genetic variants be tested together for disease risk? Study design Genetic association study Statistical model Penalised logistic regression with burden tests Suggested dataset NHANES (CDC) Open Data feasibility: NHANES public data do not provide suitable individual-level rare variant genotypes.
Model category 09 / 10
Prediction, classification and model evaluation Logistic regression is also a prediction tool. These topics test discrimination, calibration and comparison with machine learning.
TOPIC 41 Can a logistic model predict company bankruptcy from financial ratios? Study design Firm-year panel Statistical model Logistic regression with ROC and AUC Suggested dataset Taiwanese Bankruptcy Prediction (UCI) Open Data feasibility: The UCI bankruptcy dataset is a labelled prediction sample, not a firm-year panel by default.
TOPIC 42 Does logistic regression predict credit default as well as machine learning? Study design Customer records Statistical model Logistic regression compared with random forest Suggested dataset Default of Credit Card Clients (UCI) Open TOPIC 43 Can a screening score predict which hospital patients have Legionnaires’ disease? Study design Retrospective cohort Statistical model Logistic prediction model with calibration plot Suggested dataset MIMIC-IV (PhysioNet) Apply Data feasibility: Confirm diagnostic case numbers and reference diagnosis before predicting Legionnaires’ disease in MIMIC-IV.
TOPIC 44 Does a simple logistic model diagnose heart disease as well as complex ones? Study design Clinical dataset Statistical model Logistic regression compared with gradient boosting Suggested dataset Heart Disease (UCI) Open TOPIC 45 Which patients complete addiction treatment, and can we predict it? Study design Treatment discharge records Statistical model Logistic regression with cross-validation Suggested dataset TEDS Data Files (SAMHSA) Open Model category 10 / 10
Survey weights, prevalence and propensity scores National surveys need weights. Logistic regression also builds propensity scores for causal questions.
TOPIC 46 How common is depression around pregnancy in the US? Study design Cross-sectional survey Statistical model Survey-weighted logistic regression Suggested dataset NHANES (CDC) Open Data feasibility: Pregnancy timing and depression measures may have inadequate joint sample coverage in NHANES.
TOPIC 47 Are some racial groups less likely to get cholesterol screening? Study design Cross-sectional survey Statistical model Survey-weighted logistic regression Suggested dataset NHANES (CDC) Open TOPIC 48 Did Medicaid waivers raise use of addiction medicine? Study design State-year panel Statistical model Logistic difference-in-differences Suggested dataset TEDS Data Files (SAMHSA) Open Data feasibility: TEDS does not provide direct Medicaid waiver dates or comprehensive medication claims.
TOPIC 49 Does wealth explain health gaps once income is accounted for? Study design Cross-sectional survey Statistical model Propensity score weighting with logistic models Suggested dataset BRFSS (CDC) Open Data feasibility: BRFSS does not ordinarily measure household net wealth.
TOPIC 50 Do toddlers in some neighbourhoods get less sleep? Study design Cross-sectional survey Statistical model Survey-weighted logistic regression with area effects Suggested dataset NHANES (CDC) Open Data feasibility: NHANES public data may lack the required toddler sleep and neighbourhood measures.
12 datasets for logistic regression research These sources include health surveys, clinical data, business outcomes and publicly available classification benchmarks. Access types vary and not every source contains all the fields a question requires.
US health survey with interviews, physical exams, blood tests and vaccination history.
Yearly phone survey of 400,000 US adults on health behaviours, screening, vaccination and chronic disease.
Household surveys in 90+ countries on maternal care, child health, malaria and women’s status.
US cancer registry data on diagnosis, stage, treatment and patient traits.
De-identified hospital and ICU records with labs, diagnoses and outcomes.
Follows US teenagers into adulthood with data on health, family and behaviour.
US attitude survey since 1972 with many ordered and yes or no questions.
US surveys on health attitudes, technology and social issues.
95 financial ratios for 6,800+ firms with a bankruptcy flag.
30,000 credit card clients with payment history and default status.
Classic clinical dataset with 13 risk factors and a heart disease diagnosis.
US admissions to and discharges from substance use treatment, with reasons for leaving.
Printable logistic regression topic guide
Get all 50 topics in one PDF A downloadable branded guide with all 50 topics, study designs, models, dataset sources and related thesis links will be added after the PDF is created and uploaded.
PDF download coming soon How to develop a logistic regression dissertation Define the outcome. Specify the event and reference category.Plan sample size. Assess event frequency and model complexity; the ten-events-per-predictor rule is only a rough heuristic.Pre-specify confounders. Choose important variables based on subject knowledge rather than p-values alone.Use survey design correctly. NHANES, BRFSS and DHS may need weights, strata and clustering.Check performance. Report odds ratios and confidence intervals, and examine calibration and discrimination for prediction models.Related topic collections Need a second pair of eyes?
Choose a topic your data can support. Bring your research question, available dataset and proposed model to a free 15-minute consultation.
Book a free 15-minute call 