In the correspondence concerning these glimmers from the National Burn Repository (NBR), a number one request is to have an illustration of how to actually preen data when writing an NBR paper. This month's installment aims to do just that by looking at the internal inconsistencies in the reporting of burn mortality. Generally speaking, there are three data fields that pertain to burn mortality: 1) HOSPDISP, 2) HOSPDISP_Cat, and 3) DEAD. As users of the NBR database ourselves, we recently tried to determine the stated mortality in NBR data for all years reported. When examining all three of these data fields, we found that there were patients who had conflicting entries about mortality. In the NBR, the primary column to use in discerning mortality is DEAD. This column designates each patient as ALIVE or DEAD. The other columns, HOSPDISP and HOSPDISP_Cat, refer to the means by which the patients are discharged from the hospital. Entries in these columns include discharged home, morgue/funeral home, expired, nursing home and rehab, and died. As Figure 1 illustrates, using only one of the three pertinent columns to determine mortality has the potential to lead one astray. A simple query of the DEAD column reveals a total NBR mortality count of 11,277. The total number of records with any mention of death in any of the three fields is 15,099. Although most deaths are accurately accounted for and consistently entered across all three columns (10699), there are obvious mortalities that possibly go uncounted.
No takes yet. Share an insight, caveat, or question.
Pavlovich et al. (2009) studied this question.