68 Application of the Modified Hospital Length of Stay Predictor

Abstract

Abstract Introduction Developing formulas to predict accurate length of stay (LOS) is challenging in burn patients as many complex variables can effect LOS. A recently published article in 2016 provided new formulas (NFs) to predict hospital LOS. Our goal is to compare our patients actual LOS (ALOS) to the NFs published, the traditional 1day/%TBSA rule, and to the hospital expected LOS (ELOS) based on patients CMI/DRGs. Methods Data was collected from Jan ‘16 to Dec ‘18 on all patients >18 years of age using NTRACS burn registry. We excluded deaths, non-burns, readmissions, any records with missing information, and any patients transferred to another acute care hospital. We performed multiple regression to examine the relationship between LOS with age, inhalational injury (INHINJ), and TBSA on all patients. We divided this population into three groups per the NFs recommendations: 1) INHINJ; 2) without INHINJ < 40 years old; 3) without INHINJ >40 years old. Using these three groups, we calculated the difference between ALOS to 1) the traditional LOS formula, 2) the NFs as established in the 2016 article, and to 3) the hospitals ELOS metrics. For acuity comparison, we also calculated average CMI in each study group. The NFs published are as follows: Results The multiple regression model with all three variables (Age, % TBSA, and INHINJ) produced R² = 0.321, F(3,883) = 139.23, p < .05. INHINJ, age, and TBSA had significant positive regression, indicating an increase or presence of all three variables will lengthen LOS. The regression equation is: LOS = (-3.93) +1.71(AGE) +1.11(TBSA) +15.81 (Inhalation). Every year increase in age increased LOS by 1.71 days. Every 1% increase in TBSA increased LOS by 1.11 days. Presence of INHINJ (0= no INHINJ; 1= INHINJ) increased LOS by 15.81 days. Table 1 depicts the mean ± SD of the difference in LOS (predicted - actual) for the three predictive models of LOS. Negative values indicate underestimation of LOS and positive values indicate overestimate of LOS. Conclusions Application of the predictive LOS formulas has underestimated LOS in this population, with the exception of the NFs to predict LOS on patients >40 years old without INHINJ. Furthermore, an R² value of 0.32 indicates that the variables utilized in this study do not account for 68% of the variation we see. This is further exemplified by the wide standard deviations, particularly in the INHINJ group. Applicability of Research to Practice Directly Applicable.