Observational studies on SNDS, our EDS databases, PMSI or eSND. Phenotyping algorithms, survival analyses, causal modelling and retrospective cohorts.
We design the study most suited to your scientific question, following the HAS methodological guidelines on real-world studies (2021) and ISPE GPP best practices.
A population indexed on an event (diagnosis, treatment initiation, surgical procedure) with longitudinal follow-up of exposures and clinical outcomes. Exposed/unexposed cohorts with active or historical comparator. This is the most common design on SNDS and PMSI.
Selection of cases (event of interest) and matching to controls within a defined cohort. Particularly suited to the study of rare events (adverse effects, cancers) with statistical power optimised by matching on age, sex and index date.
Each patient serves as their own control: comparison of exposed and unexposed periods in the same individuals. Eliminates time-invariant confounders. Ideal for pharmacovigilance and vaccine-event studies.
A snapshot of a population at a given point in time: point prevalence, description of clinical and therapeutic characteristics, sub-population segmentation. Often used for disease burden studies.
Leveraging the historical depth of SNDS (up to 10 years) to build cohorts with long follow-up and measure the long-term effectiveness of treatments or complex care trajectories.
HAS Guide — Real-world studies (2021)
Protocols compliant with the HAS methodological guide for the evaluation of medicines and medical devices using real-world data.
Adherence to the Guidelines for Good Pharmacoepidemiology Practices for rigour in study design and analysis.
From classical biostatistics to machine learning approaches, we deploy the methods best suited to the complexity of your real-world data.
Kaplan-Meier, Cox proportional and time-dependent hazard models, Fine & Gray competing risks models. Handling of informative censoring and recurrent events. Frailty models for clustered data.
Cox · Fine & Gray · FrailtyPropensity Score Matching (PSM), Inverse Probability of Treatment Weighting (IPTW), propensity score stratification. Overlap assessment and pre/post-matching balance diagnostics. Doubly robust estimation to limit residual bias.
PSM · IPTW · Doubly RobustSupervised algorithms for phenotyping (Random Forest, Gradient Boosting, XGBoost), variable selection via LASSO and Elastic Net, unsupervised clustering (k-means, DBSCAN) for sub-population identification. Cross-validation and model calibration.
XGBoost · LASSO · ClusteringConstruction and validation of disease detection algorithms using ICD-10 codes, CCAM procedures, medications (ATC) and GHM codes. Computation of sensitivity, specificity, PPV and NPV against reference cohorts or cross-validation.
ICD-10 · CCAM · ATCStandardised estimation of epidemiological indicators adjusted for age and sex. Crude and standardised rates (direct and indirect methods). Target population projection, estimation of treated patients and therapeutic coverage rates.
Standardisation · Target populationQuantification of residual bias via E-value and Quantitative Bias Analysis (QBA) sensitivity analyses. Intention-to-treat, per-protocol and as-treated analyses. Sensitivity analyses on criterion definitions, time windows and pre-specified subgroups.
E-value · QBA · RobustnessLogistic regressions (conditional, multinomial), Poisson, negative binomial, log-binomial. Mixed models and GEE for longitudinal and repeated data. Multi-state models for complex care trajectories with transitions between clinical states.
GLM · GEE · Multi-stateValuation of healthcare consumption from PMSI and SNDS: hospital costs (GHS), outpatient and retroceded drugs, CCAM procedures and biology, daily allowances, patient transport. Total and incremental cost models.
GHS · costs · PMSIWe leverage different real-world health data sources, each with its own specificities, depth and regulatory authorisations.
Medico-administrative data from 65M beneficiaries: outpatient care reimbursements, PMSI hospitalisations, chronic conditions (ALD), deaths. Cross-scheme linkage over 10 years of depth. Ideal for large-scale cohorts and comparative effectiveness studies.
65M beneficiaries · 10 yearsProgramme de Médicalisation des Systèmes d'Information: exhaustive hospital data covering MCO, SSR, HAD and psychiatry (RIM-P). Accessible without specific CNIL authorisation (general sample or MR-005/MR-006). Ideal for studies on hospital conditions, surgical procedures (CCAM) and stays (GHM/GHS).
MCO · SSR · HAD · RIM-PHealth data warehouses combining clinical data (registries, cohorts) with SNDS data. CardioHub (interventional cardiology), UroCCR (urology/kidney cancer), Colibri (pulmonology), DataMesh (gynaecology-obstetrics). SNDS linkage for long-term patient follow-up.
Registries · SNDS linkageGeneral practitioner data: diagnoses, prescriptions, biological results, clinical measurements (BMI, BP). 3,050 physicians representative of the French population. Pseudonymised data available in France without CNIL authorisation delays.
3,050 GPs · France31 pharmacy observatories covering 85% of pharmacies and 13 therapeutic areas. Real-time dispensing data: volumes, market shares, switching, generics, biosimilars. Longitudinal trend monitoring by condition.
85% pharmacies · 13 areasOur Concepción platform enables secure cross-referencing of multiple sources within a single project: registry + SNDS, PMSI + clinical data, NIS + SNDS. Deterministic or probabilistic matching depending on available identifiers.
Concepción · MatchingOur epidemiological and biostatistical expertise covers all the questions faced by pharmaceutical companies, institutions and learned societies.
Prevalence, incidence, mortality, comorbidities and associated costs in the general population. Estimation of the target population and the number of patients eligible for a treatment. Segmentation by clinical subgroups.
Comparison of real-world efficacy between treatments. Propensity score, IPTW or exact matching to control for indication bias. Analyses in pre-specified subgroups, multiple sensitivity analyses.
Active pharmacovigilance on SNDS: detection of rare adverse events, SCCS, disproportionality analysis. Post-authorisation safety signals for PSUR and RMP dossiers. Post-Authorisation Safety Studies (PASS).
Mapping of therapeutic trajectories: lines of treatment, switching, drop-out, reinitiation. Multi-state models for transitions between care phases. Analysis of inter-event delays and progression factors.
Quantification of the delay between first symptoms and confirmed diagnosis. Identification of diagnostic delay factors (geography, specialty, comorbidities). Characterisation of the diagnostic odyssey by analysis of pre-diagnosis pathways.
Valuation of direct costs (hospitalisations, drugs, procedures, biology, transport, daily allowances) from PMSI and SNDS. Average cost per patient, cost comparison between strategies, budget impact modelling with real-world data.
Our teams have designed and conducted epidemiological studies published in peer-reviewed journals, covering a wide range of conditions and mobilising large-scale cohorts.
Retrospective cohort study on SNDS evaluating the long-term outcomes of transcatheter aortic valve replacement (TAVI). Follow-up of survival, complications and reinterventions.
Observational study on SNDS analysing recurrent hospitalisations for heart failure, determinants of rehospitalisation and the impact of therapeutic strategies.
Large SNDS cohort on chronic heart failure: prevalence, incidence, care pathways, lines of treatment and mortality. Survival analyses and propensity scores for strategy comparison.
Study on idiopathic pulmonary fibrosis: disease burden, diagnostic delay, care pathways and impact of antifibrotic treatments on real-world survival.
Study of care trajectories in oncology. Analysis of lines of treatment, therapeutic sequences and predictive factors for switching or treatment discontinuation.
Large-scale cohort on rheumatological conditions: rheumatoid arthritis, spondyloarthritis. Effectiveness of biologics, switching and therapeutic persistence under real-world conditions.
Every epidemiological project follows a rigorous and transparent process, from protocol design to submission for publication in a peer-reviewed journal.
Protocol compliant with the HAS guide, pre-specified and locked SAP, regulatory dossier (CESREES, CNIL, MR-004/MR-006), scientific committee formation, feasibility on a sample.
Extraction via Victoria Pipeline, data quality control, primary and sensitivity analyses, interim report to the scientific committee, manuscript writing, peer-review submission and conference presentation.
Discuss your epidemiological study project with our experts.
We will guide you from protocol design to publication.