The National Surgical Quality Improvement Program in Non-Veterans Administration Hospitals
Key result
The VA NSQIP mortality model accurately predicted 30-day mortality in non-VA general and vascular surgery patients (C-index 0.934), demonstrating the methodology's applicability outside the VA.
Why the study?
Does the NSQIP methodology generate reasonable predictive models for 30-day mortality and morbidity in non-VA hospitals compared to VA hospitals?
Population
44,107 general and vascular surgery patients, comprising 2,747 patients from 3 non-VA academic medical…
Comparison
Application of the National Surgical Quality… vs Concurrent VA hospital data collected by the…
Design
Cohort
Follow-up
30-day
Authors
Discussion
Loading...
Member takes
Implication
May support NSQIP implementation in non-VA hospitals; extends VA methodology but remains hypothesis-generating pending prospective outcomes data.
Key Points
- To assess the feasibility and predictive accuracy of implementing the National Surgical Quality Improvement Program (NSQIP) risk-adjusted methodology in non-Veterans Administration (VA) hospitals.
- Conducted a multicenter pilot study across three non-VA academic medical centers using dedicated, trained nurse reviewers who abstracted clinical data from general and vascular surgery patients between October 1999 and September 2000.
- Transmitted patient records via a secure internet portal and developed logistic regression models comparing non-VA data to concurrent VA NSQIP data, assessing discrimination using C-indices.
- Evaluated 2,747 non-VA cases (2,251 general surgery; 496 vascular surgery) alongside 41,360 VA cases (31,393 general surgery; 9,967 vascular surgery), with bivariate risk-outcome relationships matching across >66% of risk variables.
- Application of VA NSQIP mortality models to non-VA data demonstrated high discrimination, yielding C-indices of 0.942 for general surgery, 0.915 for vascular surgery, and 0.934 overall.
- VA NSQIP morbidity models yielded lower discrimination in non-VA data (C-indices of 0.778 for general surgery, 0.638 for vascular surgery, and 0.760 overall), with no preliminary differences observed in risk-adjusted mortality between cohorts.
Study Design
Type
Observational (n=44,107)
Multicenter
Yes
Structured PICO
Does the NSQIP methodology generate reasonable predictive models for 30-day mortality and morbidity in non-VA hospitals compared to VA hospitals?
P
Population
44,107 general and vascular surgery patients, comprising 2,747 patients from 3 non-VA academic medical centers and 41,360 patients from VA hospitals.
I
Intervention
Application of the National Surgical Quality Improvement Program (NSQIP) methodology and predictive models.
C
Comparator
Concurrent VA hospital data collected by the NSQIP.
O
Outcome
Predictive value (C-indices) of logistic regression models for 30-day mortality or morbidity.
Main Result
Effect estimate: C-index 0.934
The NSQIP methodology can be successfully implemented in non-VA hospitals, generating highly predictive models for 30-day mortality.
Limitations
- Non-VA sample size was smaller than the VA cohort
- Non-VA sample size was smaller than the VA sample size
Cite This Study
Fink et al. (2002) conducted an observational in General and vascular surgery (n=44,107). National Surgical Quality Improvement Program (NSQIP) methodology vs. Veterans Administration (VA) hospitals was evaluated on Predictive value (C-index) of VA NSQIP mortality and morbidity models applied to non-VA data (C-index 0.934). The VA NSQIP mortality model accurately predicted 30-day mortality in non-VA general and vascular surgery patients (C-index 0.934), demonstrating the methodology's applicability outside the VA.