Historically, fatal injury monitoring and surveillance have relied on mortality data derived from death certificates (DC). However, problems associated with utilizing DC have been well documented. Recently, access to and utilization of hospital discharge data (HDD) have offered a new and important secondary source of data regarding in-hospital deaths. However, studies have shown that discrepancies between the HDD and the corresponding DC often exist. This discrepancy was especially evident when comparing HDD to the vital statistics data (VSD) for deaths by falls among those aged 65 and over in 19 states. This was a retrospective forensic review of elderly (age 65 and over) fall-associated fatalities (E880-E888) identified from HDD and VSD in Allegheny County, Pennsylvania, between 1997 and 1998. Seventy-seven cases were identified, with the original manner of death listed as natural (34), suicide (1), and accidental (42) on the DC. Following a forensic review of the cases, the manner of the death on the DC should have been changed from natural to accidental in 28% (n = 12) of the cases, representing an undercount in the VSD. Undercounts were due to a failure of clinicians to account for the significance of a fall event that contributed to subsequent pathology and death. In addition, in that 22% (n = 17) of the HDD fall-associated deaths, the fall did not contribute directly or sequentially to the underlying cause of death, thereby representing an overcount in the HDD. Based on these findings we recommend (1) elderly fall surveillance systems should only count HDD E-coded falls that demonstrate a serious traumatic injury which directly or subsequently results in death, (2) all in-hospital fall-associated deaths should be reported to and reviewed by coroner/Medical Examiner offices for determination of the cause and manner of death, and (3) physicians should be better educated in properly completing death certificates.