1. MEASUREMENT OF THE VARIATION

 

1.2. ECHO HOSPITAL-SPECIFIC METHODS

 

Magnitude of an event: Crude incidence, adjusted incidence

Incidence, either crude or standardised, is representing the magnitude of a phenomenon; in this case, the risk for a patient to experience an event in the hospital where he or she has been treated, in a period of time. Events in ECHO are generally undesirable outcomes- case-fatalities and patient-safety indicators.

While the crude incidence is straightforwardly calculated, the numerator is represented by the number of events and the denominator by the number of patients at risk (usually per 100 or 1,000 patients), the case of the adjusted-incidence requires specific estimation using risk adjustment within logistic multivariate modelling.

In ECHO, incidence is generally adjusted using as co-variables patients’ age, sex and comorbidities, namely Elixhauser comorbidities. In a few specific indicators, ECHO includes additional factors for adjustment –e.g. major valve interventions concurrent with CABG or STEMI vs. NSTEMI AMI in mortality after CABG.

When ECHO models incidence assumes that risks tend to be more similar among the patients assisted within a specific hospital than among patients from different hospitals, hypothesizing the existence of a cluster effect. As a consequence, adjusted incidence of a specific event is estimated using 2-level hierarchical logistic modeling, where patients are nested into hospitals.

 

Interpretation note: Depending on the objective, ECHO may use national benchmarks or international benchmark. In the first case, logistic models will only use, in-country patients and hospitals; in the second case, incidence will be modelled using all patients at risk in all the ECHO countries.

 

 

The magnitude of the variation attributed to the hospital

Besides the adjusted incidence, logistic models provide two more statistics: a) the cluster effect - the fraction of variation explained by the hospital, beyond the variation explained by patients; and b) the magnitude of the hospital effect – as compared to patients’ effect.

Hospital-specific analyses are graphically represented by funnel and caterpillar plots. While funnels are generally used to represent hospitals’ incidence of events, caterpillar are meant to position hospitals with regard to the expected incidence.

 

1. Cluster effect and Magnitude of the hopital effect

2. Graphical representation: Funnel plots and Caterpillar plot

 

 

 


Please cite this publication as:

European Collaboration for Healthcare Optimization (ECHO) www.echo-health.eu. Zaragoza (Spain): Instituto Aragonés de Ciencias de la Salud-Instituto Investigación Sanitaria Aragón; c2011. Bernal-Delgado E, Thygesen LC, Martínez-Lizaga N, Comendeiro M on behalf of the ECHO consortium. Handbook on methodology: Measurement of the variation; 2014 Apr 27 [ accessed: date ]; Available from: http://www.echo-health.eu/ handbook/measuring_variation.html


 

 First page  Previous page: Geographic methods  Next Page: Measuring inequalities  Last Page

Page 10


This project is funded by the Framework 7
Programme of the European Union

ECHO Project   -   Disclaimer   |   Project enquiries