This proposal aims to improve process measurement in order to better assess hospital quality of care. Discrete process measures like those used by CMS for Hospital Compare (HC) are commonly used by healthcare evaluation organizations because they are: (1) quick to collect (require less time in abstraction than traditional global measures);(2) easy to understand and point to potentially improvable provider actions;and (3) purportedly require no severity adjustment. In fact, a common motivation for utilizing process measurement is the belief that such measures do not require severity adjustment, beyond coarse inclusion and exclusion criteria. This application examines two fundamental questions regarding process measurement: (1) Do discrete process measures such as those utilized in CMS Hospital Compare need severity adjustment above and beyond the selection criteria commonly used;and (2) Are there other sampling schemes that could be utilized rather than typical random sampling (as is done by CMS) that could better (i.e., more efficiently and with less bias) sample processes at hospitals and therefore allow for the collection of more global process measures that could have stronger associations with outcomes than the discrete measures utilized in Hospital Compare. We present a conceptual model to study the need for severity adjustment for process measures and to compare various process measurement schemes based on reducing overall mean square error (MSE): This study will propose and present preliminary data on a potentially better method to sample charts for process measurement based on a multivariate matching algorithm we call """"""""Multivariate Template Matching"""""""" which produces directly standardized matches of patients in order to better select patients inside hospitals to compare process of care across hospitals. We will also examine process measures using a Medicare data set collected as part of the CMS Hospital Compare initiative to (a) establish that we can achieve excellent matches using the Template Matching algorithm;and (b) test whether the bias observed with the present sampling schemes used by CMS for Hospital Compare is reduced using Template Matching. In summary, working with a data set to which CMS has given us special permission to analyze, we will formally test for bias in Hospital Compare, and formally test a more optimal scheme for conducting process measurement through Multivariate Template Matching. If successful, Multivariate Template Matching would allow for the collection of more detailed global process measures since the required sample size for abstraction (and therefore cost) could be reduced.
This application seeks to improve process measurement by (1) testing whether unadjusted process measures are biased because patient factors are associated with process adherence;and (2) developing a new methodology, Multivariate Template Matching, for more efficiently selecting patient charts in which to follow and compare process adherence. The application seeks to utilize a large database of process measure abstractions by CMS through the Hospital Compare project. If successful, results from this application could be utilized to implement template matching when assessing process compliance for Hospital Compare and other programs which study process as a quality of care indicator.
| Silber, Jeffrey H; Rosenbaum, Paul R; McHugh, Matthew D et al. (2016) Comparison of the Value of Nursing Work Environments in Hospitals Across Different Levels of Patient Risk. JAMA Surg 151:527-36 |
| Silber, Jeffrey H; Rosenbaum, Paul R; Ross, Richard N et al. (2014) A hospital-specific template for benchmarking its cost and quality. Health Serv Res 49:1475-97 |
| Silber, Jeffrey H; Rosenbaum, Paul R; Ross, Richard N et al. (2014) Template matching for auditing hospital cost and quality. Health Serv Res 49:1446-74 |
