SM 04-2130 The impact of case definition on estimated incidence rates when using hospital discharges as reference statistic

Background The EU Injury Database IDB is designed to deliver public health indicators on various kinds of non-fatal injuries. Although IDB data should be recorded in every hospitalbased emergency department (ED), there are many countries which collect data only in samples of hospitals, which makes projections necessary. Previous analyses of estimates (2011–2013) unveiled that discharge figures are higher than the IDB based estimated numbers of admissions. It has been assumed, that these differences are caused by different case definitions in IDB samples and reference statistics (usually hospital discharge statistics). Objectives To assess the impact of varying case definitions, i.e. the inclusion/exclusion of non-residents, complications of medical interventions, sequelae of injuries, day care (zero nights) patients, follow-up (secondary) treatments, patients deceased in hospital, and the use of either ICD-10 chapter XIX (diagnoses) or XX (external causes). Methods Survey among national injury data administrators. Seven countries were able to provide the desired data on mentioned subgroups of hospitalized injury patients. Findings The average shares of complication of medical interventions and day-care patients are rather high (14% each), also follow-up treatments play a considerable role (8%). There are considerable differences between countries which indicates different coding routines: e.g. the percentage of complications ranges from 4% to26%. The number of admitted injury patients vary also considerable depending on what group of ICD-10 codes is used for defining ‘injury’. Conclusions When projections of national rates are necessary, it is essential to use coherent case definitions for the IDB sample as well as for the reference statistic. For comparable national rates, the same definitions need to be applied in all EU countries. However, when this is assured, the differences between IDB based estimates for admissions and HDS figures seem to diminish, which indicates, that IDB estimates are valid.