Variation in adverse event incidence rates between hospitals and hospital departments. The Dutch adverse events study.

Objective: To assess the variation in adverse event (AE) rates at the hospital and hospital department level, in order to gain insight in room for improvement of patient safety at each level. Methods: Randomly selected records of 7926 hospital admissions of 2004 from 4 university and 17 general hospitals were reviewed by trained nurses and physicians between August 2005 and October 2006. The physicians identified AEs, degree of preventability, most responsible medical specialty and causes of AEs, and patient and admission characteristics. Additional patient information, including coded discharge diagnoses (ICD-9), Charlson comorbidity index and coded interventions, was received from the Dutch registration of hospital information. We applied multivariable logistic multilevel analysis (MLwiN version 2.02), using AE incidence rates in university and general hospitals as an outcome, to apportion variance to hospital and hospital department level while correcting for differences in patient mix (age, sex, admission urgency, surgical or non-surgical admission, diagnostic group and comorbidity). Covariates were entered stepwise into the model and were centred to reference values for all Dutch hospital admissions. Results: The unadjusted AE incidence rates were higher in university hospitals (8.2, 95%CI 6.8 to 9.9) than in general hospitals (5.1, 95%CI 4.5 to 5.9). After adjusting for patient mix, this difference was no longer statistically significant (6.2, 95%CI 4.0-9.3 versus 3.7, 95%CI 2.9 to 4.7). The adjusted incidence of AEs per hospital ranged from 2.0% to 6.5%. Preventable AEs occurred less often in university hospitals (1.5, 95%CI 1.0 to 2.2) than in general hospitals (2.2, 95%CI 1.8 to 2.8), but this difference was not statistically significant and further adjustment for patient mix slightly decreased the estimated incidence rates. The adjusted incidence of preventable AEs per hospital ranged from 0.9% to 2.7%. When apportioning variance in the incidence of AEs, without additional correction for patient mix, differences between hospital departments had more impact (intraclass correlation (ICC) 10.3%) than differences between hospitals (ICC 3.4%). After correction for age, sex and admission urgency, the ICC for hospital departments decreased to 8.3%, and after further correction for admission to surgical or non-surgical department, diagnostic group and comorbidity to 5.9% while the ICC for hospital variation stayed stable. Age, admission urgency, surgical admission, and comorbidity showed a statistically significant association with the incidence of AEs. Increasing age, more comorbidity and admissions to surgical departments were associated with more AEs; whereas urgent admission was associated with less adverse events compared to planned admissions. Since patients are often treated by different medical specialties, the admission department not always reflects the specialty responsible for the AEs. Especially, in patients admitted to non-surgical departments 13 to 19% of AEs occur under responsibility of surgical specialities. In patients admitted to surgical departments, 98 to 100% of AEs are caused by surgical specialties. Limiting the analyses to surgical patients alone did not materially change the results and showed larger variance at the hospital department level compared to the hospital level. Discussion: In university hospitals more AEs but not more preventable AEs occurred than in general hospitals. When apportioning variance in incidence of AEs to hospitals and hospital departments, differences between hospital departments had more impact than differences between hospitals. Thus, when improving patient safety we should focus on unsafe hospital departments rather than on unsafe hospitals. (aut. ref.