Risk-Adjusted Case Series
Also known as: risk-stratified case series, adjusted case series, risk-corrected case series
A risk-adjusted case series is an observational study design that reports outcomes for a consecutive or defined group of patients undergoing the same procedure or sharing a condition, while statistically correcting for differences in patient-level baseline risk. Rather than presenting raw complication or mortality rates, it compares observed outcomes against expected rates derived from a validated scoring model (e.g., POSSUM, APACHE, ASA grade), enabling fairer evaluation of clinical performance across institutions or over time.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use a risk-adjusted case series when you have a single-centre or single-team dataset of consecutive patients and want to benchmark outcomes against expected rates given patient severity — particularly in surgical audit, quality improvement, or procedure safety reporting. It is appropriate when a randomised trial or comparative control group is unavailable or unethical, and when a validated risk model exists for the outcome of interest. Do NOT use it as a substitute for a controlled study when estimating treatment effects: the absence of a concurrent control group means causal conclusions cannot be drawn. Avoid it when the available risk model was calibrated in a very different patient population, as miscalibration will render the O/E ratio misleading. It is also unsuitable as a primary efficacy design for regulatory submissions.
Strengths & limitations
- Accounts for case-mix variation, enabling fairer performance comparison across centres, surgeons, or time periods than raw rate reporting.
- Applicable to routinely collected clinical data without requiring a randomised control group.
- Transparent benchmarking framework: the O/E ratio and its confidence interval are interpretable by clinicians and audit committees.
- Supports quality improvement cycles by identifying unexpectedly high or low outcome rates that warrant further investigation.
- Compatible with national registry infrastructure and surgical audit programmes (e.g., UK SCTS, STS database).
- No concurrent control group: causal inference about treatments or interventions is not possible.
- Results depend entirely on the validity and calibration of the chosen risk model; a poorly calibrated model produces misleading O/E ratios.
- Residual confounding from unmeasured or poorly measured patient characteristics cannot be eliminated.
- Retrospective data collection is common, introducing potential for incomplete or inconsistent covariate ascertainment.
- Small series have wide confidence intervals around O/E, making it difficult to detect true departures from expected performance.
Frequently asked
How is a risk-adjusted case series different from a standard case series?
A standard case series simply reports how many patients experienced a given outcome. A risk-adjusted case series goes further by applying a validated risk model to estimate how many outcomes were expected given the baseline severity of the specific patients treated, then expressing performance as an O/E ratio. This makes the results meaningful for benchmarking and quality assessment in a way that raw rates cannot.
Which risk model should I choose?
Choose the model validated for your clinical domain and outcome: POSSUM or P-POSSUM for general surgery morbidity and mortality, APACHE II/III for ICU mortality, EuroSCORE II for cardiac surgery, CHADS2-VASc for atrial fibrillation stroke risk, and so on. The model must have been calibrated in a population similar to yours, and ideally you should verify calibration in your own dataset using a Hosmer-Lemeshow test or calibration curve.
Can I use this design to show that my intervention works?
No. A risk-adjusted case series cannot establish causality because there is no control group. A favourable O/E ratio means outcomes were better than predicted for patients of that risk profile — it does not prove the intervention caused the improvement. Confounding by unmeasured factors remains. For efficacy evidence, a randomised controlled trial or, at minimum, a controlled observational study with propensity-score adjustment is needed.
How many patients do I need?
There is no universal minimum, but small series (fewer than 50 events) will produce O/E confidence intervals too wide to support meaningful conclusions. A common rule of thumb in regression modelling is at least 10 events per predictor variable in the risk model; for audit reporting, series large enough to generate stable O/E ratios — typically several hundred cases per reporting period — are preferred.
What is a funnel plot and why is it used here?
A funnel plot displays O/E ratios (or standardised mortality ratios) on the y-axis against the expected number of events on the x-axis, with control limits derived from a Poisson distribution. Units with small expected counts will have wider limits; large-volume units have narrower limits. Points outside the limits are statistical outliers warranting investigation. This approach, formalised by Spiegelhalter (2005), corrects for the tendency to over-interpret variation in small series.
Sources
- Copeland, G. P., Jones, D., & Walters, M. (1991). POSSUM: a scoring system for surgical audit. British Journal of Surgery, 78(3), 355–360. DOI: 10.1002/bjs.1800780327 ↗
- Mayer, E. K., Bottle, A., Darzi, A. W., & Aylin, P. (2004). Case volume and outcome in the surgical treatment of colorectal cancer. British Journal of Surgery, 91(9), 1104–1110. link ↗
How to cite this page
ScholarGate. (2026, June 3). Risk-Adjusted Case Series. ScholarGate. https://scholargate.app/en/epidemiology/risk-adjusted-case-series
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Case seriesEpidemiology↔ compare
- Diagnostic Accuracy Study DesignClinical Research↔ compare
- Propensity Score MatchingResearch Statistics↔ compare
- Risk-adjusted cohort studyEpidemiology↔ compare
- Survival AnalysisResearch Statistics↔ compare