Veterans Health Administration (VHA) has been strongly committed to reviewing and improving the quality of care provided to veterans. One example of this has been the VA Surgical Quality Improvement Program (VASQIP) (previously NSQIP/CICSP) database that developed reliable comparative measures of 30-day mortality and morbidity for surgical procedures leading to significant transformations in quality. However, 30-day clinical outcomes alone provide a limited picture of true surgical care delivery efficiency. Notably, for many low risk surgical procedures, the current VASQIP 30-day outcome measures mask important variation and differences in efficiency. This study was in part motivated by the intention to update VASQIP with the most state-of-science methodology and techniques as they become available.
The overall objective was to develop models for reliably profiling surgical quality of VA facilities that using mortality, morbidity, and costs and to examine facility-level variation.
Data Sources: VHA Corporate Data Warehouse (CDW), VIReC VHA Decision Support System, VA Vital Status File, VACCR, VA-CMS and VASQIP
Population: Patients who, during VA fiscal year 2007 - 2011, had at least one of the seven procedures, Transurethral Resection of the Prostate (TURP), Unilateral inguinal hernia, Colon resection for malignancy, Prostatectomy for malignancy, Hip replacement, Pulmonary lobectomy for malignancy, or Coronary Artery Bypass Graft Surgery (CABG).
Main Measures: The primary outcome variables were mortality, morbidity, and total costs from the date of surgery until 30 days post-surgery. Cost data were extracted for each patient using the MCA (formerly Decision Support System). The timeframe of 30 days following surgery was chosen for the cost outcome because VASQIP records complication outcomes within this same 30 day period. We included total costs during this period to ensure that we captured variation associated with complications, wound infection, readmissions and all potential costs that were attributable to the procedure among the national sample. We inflated all costs from earlier years to 2011 dollars, based on the Consumer Price Index from the Bureau of Labor Statistics. Morbidity was defined by aggregating VASQIP 30 day complication variables to form a logical variable that was 'true' if a patient experienced any of the included VASQIP complications and 'false' if the patient did not record a complication. Mortality was included via the VA Vital Status file which is the gold standard for determining death within the VA. Death within 30 days was determined by finding if the date of death occurred within the 30 day window following surgery.
The unit of analysis was the patient and the hierarchical regression models used in CMS were extended to fit hierarchical models to our data to accommodate multivariate outcome measures. That is, multivariate outcome measures can be considered as one level in our hierarchical model.
We examined the cost associated with hernia surgeries. The incidence of mortality and morbidity was low with low variability; however, cost of hernia repair varied widely across facilities. The median 30-day unadjusted cost for hernia surgery in 2011 dollars was $7,908 (IQR: $6,137-$10,416) and the mean was $8,856. There was substantial variation in 30-day costs. The highest mean 30-day cost for a site was $15,296 while the lowest mean 30-day cost was $3,985. Similar patterns were observed for median costs at each site. There were 44 hospitals above and 30 hospitals below the risk standardized median cost, with risk-standardized ratios (O/E) ranging from 0.51 to 1.69, showing a large variation in costs across facilities, even after controlling for patient case-mixes and regional labor cost differentials.
Examining the relative cost of unilateral inguinal hernia repair across individual facilities within the VA system, we see a wide variation that is not explained by morbidity, mortality or other clinical factors. Given that cost was determined using the same system (MCA) for all the institutions, this suggests that the differences in processes and structures (e.g. material purchasing contracts, lab and radiologic study ordering, etc.) could explain the wide variation in cost.
For prostatectomy, cost varied greatly across VA facilities. The expected cost of prostatectomy varied from around $1,000 to beyond $10,000 for a 60 year old White veteran without any pre-existing VASQIP risk factors. Such variations couldn't be explained by facility volume and label cost. With the adjustment to volume and labor cost, some of the small facilities are now properly identified as bad performers.
This project provided more comprehensive comparisons of VA surgical care quality among VA facilities using VASQIP data and recently developed statistical models. The analysis will help VA to identify specific VA facilities that have achieved low complication rates, while maintaining lower costs and those VA facilities that have both high complication rates and high costs . By further studying these identified VA facilities, we may provide the VA leadership with new information on how to improve quality of surgical care within the VA.
We plan on sharing our results with National Surgery Office and present a cyberseminar to VA researchers and the broader healthcare community. Newly developed Bayesian hierarchical models will help VA investigators to study quality variation among VA facilities . The use of these new methods can contribute to the statistical literature and improve the quality of research in VA.
None at this time.
Computational Modeling, Cost-Effectiveness, Efficiency, Statistical Methods