ISPOR Retrospective Database Analysis Good Research Practices — Part III (Analytic Methods)
This guideline provides best practices for analyzing retrospective database studies to estimate treatment effects. It focuses on methods to reduce bias when comparing treatments using real-world data, especially how to handle confounding factors that could distort results.
At a glance
Use when
Conducting nonrandomized studies of treatment effects using secondary databases, especially when estimating causal effects or adjusting for confounding in real-world evidence research.
Avoid when
When randomized controlled trial data are available and sufficient; when key confounders are unmeasured or poorly captured in the data; or when analysts lack expertise in advanced observational study methods.
Inputs
Secondary healthcare data (e.g., claims, electronic health records), research question on treatment effect, definition of exposure and outcome, covariate data for confounding adjustment
Outputs
Guidance on selecting and applying analytic methods, recommendations for bias reduction, best practices for statistical analysis and reporting in retrospective database studies
How it works
Part III of the ISPOR Retrospective Database Task Force report outlines good research practices for analytic methods in nonrandomized studies of treatment effects using secondary healthcare data. It emphasizes causal inference frameworks, appropriate selection and application of statistical methods for confounding control (e.g., propensity score methods, regression adjustments), and transparent reporting. The guideline addresses study design choices, model validation, sensitivity analyses, and limitations inherent in observational data.
- HTA domains
- Clinical Effectiveness, Costs & Economic Evaluation, Safety
- Assumptions
- That observational databases can yield valid causal inferences if appropriate analytic methods are applied; that researchers have access to sufficient covariate data to adjust for confounding; and that transparency in methodology improves study credibility.
- Strengths
- Developed by an international expert task force; based on current methodological evidence; provides practical, actionable guidance for analysts; promotes methodological rigor and reproducibility in real-world evidence studies.
- Limitations
- Does not eliminate inherent biases in observational data; requires skilled application; may not cover every specific database or clinical context; implementation depends on data quality and availability.
- Also known as
- ISPOR Good Research Practices for Retrospective Database Analysis - Part III, ISPOR RDA Good Research Practices - Analytic Methods
Questions this answers
- › How can causal effects be reliably estimated from nonrandomized studies using secondary databases?
- › What methods are recommended for controlling confounding in retrospective database analyses?
- › How should propensity scores be correctly applied and reported in observational studies?
- › What are appropriate sensitivity analyses to assess robustness of findings in database studies?
- › Which statistical models are suitable for different types of outcomes and data structures in observational research?
- › How can researchers minimize bias and improve validity when using real-world data for treatment comparisons?
References & sources
Similar by meaning
- ISPOR Good Research Practices for Retrospective Database Analysis Checklist (Part I)
- ISPOR Prospective Observational Studies to Assess Comparative Effectiveness Good Research Practices Task Force Report
- Good Practices for Real-World Data Studies of Treatment Effectiveness
- Guidelines for Good Database Selection and Use in Pharmacoepidemiology Research
- ISPOR-AMCP-NPC Questionnaire for Indirect Treatment Comparisons/Network Meta-Analysis
Beta record. Based on the original catalogue summary; primary-source enrichment pending.