[Evaluation of activity in intensive care. A comparison of administrative and epidemiologic data]

This paper reports on results of a formal comparison of data excerpted from 3 different data sources regarding patients cared for in Italian Intensive Care Units (ICU) in 1992. The analysis was carried out in order to assess whether the administrative data routinely collected for reimbursement and policy issues are valid when compared with data from epidemiologic studies. First, using data from the Central Service for Health Planning (SCPS) a general description of the whole national scenario is shown. All the Italian data from 265 Italian hospitals having at least one ICU were analyzed. In these hospitals there were 2,357 ICU beds (4.3 beds/1000 inhabitants, with large regional variability). About 100,000 cases were identified (1% of all the hospital admissions received ICU care at some moment during the stay), with an overall hospital mortality rate of 17%. This estimate was largely different from those obtained from GiViTI data base (5100 patients from a national sample of 114 ICUs), where mortality rates were always higher (25 and 30%, for ICU and hospital mortality). Second, a more detailed analyses and comparison was carried out in the subsample of cases admitted in the Hospitals of the Regione Lombardia. In this case, beside the data from SCPS (No = 20580) and from GiViTI (No = 1121), the Regional data-base formed by all the discharge from 41 hospitals having at least one ICU were available (NO = 16674), allowing a formal comparison between estimates obtained from 3 different sources. In this sub-sample, although administrative data showed a good capability and satisfactory accuracy to describe the volume and socio-demographic and clinical characteristics of the cases when compared with GiViTI estimates, still a systematic underestimate of the overall ICU and hospital mortality was present. Moreover, an unexpected high rate of cases were apparently discharged alive from ICU (16.5 versus 32.6% in GiViTI). Tentative explanations and implications of these phenomena are discussed in the text. Finally, in the Lombardia sub-sample a formal description of the case-mix using the Diagnosis Related Group (DRG) system was carried out in order to identify strengths and limitations of this approach when adopted in the ICU setting. Overall, when data are analyzed according to the kind of Major Diagnostic Category (MDC), 75% of all cases were concentrated in 5 alone, with very peculiar concentration in some specific categories. For example, 45% of all the hospital admissions related to the multiple trauma were actually admitted in the ICU setting. At DRG level, fifty of the 493 DRGs available explained 65% of all the cases, showing an unexpected capability of this system to detect the ICU case-mix. The most represented surgical DRG was the number 107 (Coronary Arterial Bypass Graft, without catheterization) and the corresponding medical was the 28 (Cerebro Vascular Accident, excluding TIA).