Abstract
In response to societal pressure to reduce expenditures and increase quality, we sought to develop a methodology to predict hospital charges related to congenital heart surgery.
Patients undergoing congenital heart surgery at Boston Children's Hospital in fiscal years 2007 to 2009 comprised the derivation cohort. Clinical data, including Current Procedural Terminology coding of the primary surgical intervention, were collected prospectively and linked to total hospital charges for an episode of care. Surgical charge categories were developed to group surgical procedure types using empiric data and expert consensus. A multivariable model was built using surgical charge categories and additional patient and procedural characteristics to predict the outcome, total hospital charges. A contemporary cohort for fiscal years 2010 to 2012 was used to validate surgical charge categories and the multivariable model.
In the derivation cohort, 2,105 cases met inclusion criteria. One hundred three surgical procedure types were categorized into seven surgical charge categories, yielding a grouper variable with an R(2) explanatory value of 47.3%. Explanatory value increased with consideration of patient age, admission status, and preoperative ventilator dependence (R(2) = 59.4%), as well as weight category, noncardiac abnormality, and genetic syndrome other than trisomy 21 (R(2) = 61.5%). Additional variability in charge was explained when extracorporeal membrane oxygenation utilization and greater than one operating room visit during the episode of care were added (R(2) = 74.3%). The contemporary cohort yielded an R(2) explanatory value of 67.7%.
The combination of clinical data with resource utilization information resulted in a statistically valid predictive model for total hospital charges in congenital heart surgery.
