Appendix 1

Risk adjustment variables utilized in multivariate regression analysis in individual studies.

Jones J 1995[28] APACHEII, patients’ age, severity of disease, comorbidity. Sociodemographic, physiological and diagnostic data collected for APACHE II scoring.
Iapichino G 2004[29] Logistic regression model based on backward regression approach – to assess relation ship between overall and high risk volume and hospital mortality.
Patients demographicand clinical characteristics (age, provenance, surgical conditions, diagnosis, prediction of hospital mortality by SAPSII score and destination, process of care data (length of stay, cumulative NEMS score, level of care on admission, changes in level of care, macro geographic area/hospital ICU characteristics, university affiliation type, size and nurse: patient ratio.
Durairaj L 2005[30]cox regression using fraility models. Variables used included- age, gender, APACHE III score, admitting diagnoses and admission source. Median LOS, median ICU LOS, and % of inter hospital transfers. Hospital teaching status not included as there was a high degree of correlation between volume and teaching status.
Glance L 2006[31]Two analyses – 1stanalysis comparing volume association controlling only for patient risk factors, second volume association with patient risk factors and ICU characteristics. Age, SAPSII score used for patient risk factors, ICU characteristics – type, closed/optional, physician coverage during day/night, nurse to patient ratio included. Hospital characteristics – medical school affiliation, geographical location, association with a primary critical care fellowship, trauma centre designation.
Kahn J 2006[32]APACHE III score, length of hospital stay before admission to ICU, primary diagnosis as assessed by admitting nurse, preadmission location of the patient, academic status of the hospital, type of ICU and geographic location, presence or absence of intensivists in ICU.
Needham D 2006[34] year of mechanical ventilation, patient sex and age group, urgency of admission, individual Charlson index co-morbidities, most, responsible diagnosis, hospital region, Generalized estimating equations used to account for clustering. Separate multivariate analyses for medical and surgical patients.
Peelen L 2007[33]first analysis consisted of case mix correction with age, sex, number of dysfunctioning organ systems and SAPS II score. Second analysis using ICU organization (number of intensivists/bed and number of nurses/ ICU bed) as covariate along with patient risk factors, third analysis included each co-variate with p < 0.10 in the multivariate regression analysis.
Carr BG 2009[10] age, Glasgow coma scale, mechanical ventilation, Acute physiology score (APS). Hospital level variables- region, teaching status and annual volume.
Kahn J 2009[38]age, gender, admission source, comorbidities, severity of illness on admission, hospital teaching status and hospital technological capacity. Comorbidities measured in two ways- using Charlson index and individual Elixhauser co morbidities. Severity of illness measured using MediQual atlas probabilities.
Metnitz B 2009[35]age, sex, chronic conditions, reason for admission, severity of illness (SAPS II score), number and severity of organ dysfunction using the logistic organ dysfunction score (LOD), level of provided care, length of ICU and hospital stay. Patient to nurse ratio, work – utilization ratio (workload to available staff).
Reinikainen M 2010[27]age, SAPSII score, simplified acute physiology score, therapeutic interventional scoring system (TISS) score, type of hospital – university, non-university central hospitals and regional hospitals.
Gopal S 2010[36]investigator blinding, hospital and ICU characteristics, ICU type, patient sex, age, APACHE II score, year of ventilation, length of ICU stay and urgency status.
Darmon M 2010[37] patient characteristics, age, gender, diagnostic category, organ failure, length of hospital stay, type of patient- medical/surgical, SAPS II score,