Using linked hospitalisation data to detect nursing sensitive outcomes: A retrospective cohort study

Abstract

BackgroundNursing sensitive outcomes are adverse patient health outcomes that have been shown to be associated with nursing care. Researchers have developed specific algorithms to identify nursing sensitive outcomes using administrative data sources, although contention still surrounds the ability to adjust for pre-existing conditions. Existing nursing sensitive outcome detection methods could be improved by using look-back periods that incorporate relevant health information from patient's previous hospitalisations. Design and settingRetrospective cohort study at three tertiary metropolitan hospitals in Perth, Western Australia. ObjectivesThe objective of this research was to explore the effect of using linked hospitalisation data on estimated incidence rates of eleven adverse nursing sensitive outcomes by retrospectively extending the timeframe during which relevant patient disease information may be identified. The research also explored whether patient demographics and/or the characteristics of their hospitalisations were associated with nursing sensitive outcomes. ResultsDuring the 5 year study period there were 356,948 hospitalisation episodes involving 189,240 patients for a total of 2,493,654 inpatient days at the three tertiary metropolitan hospitals. There was a reduction in estimated rates for all nursing sensitive outcomes when a look-back period was applied to identify relevant health information from earlier hospitalisations within the preceding 2 years. Survival analysis demonstrates that the majority of relevant patient disease information is identified within approximately 2 years of the baseline nursing sensitive outcomes hospitalisation. Compared to patients without, patients with nursing sensitive outcomes were significantly more likely to be older (70 versus 58 years), female, have Charleson comorbidities, be direct transfers from another hospital, have a longer inpatient stay and spend time in intensive care units (p≤0.001). ConclusionsThe results of this research suggest that nursing sensitive outcome rates may be over-estimated using current detection methods. Linked hospitalisation data enables the use of look-back periods to identify clinically relevant diagnosis codes recorded prior to the hospitalisation in which a nursing sensitive outcome is detected. Using linked hospitalisation data to incorporate look-back periods offers an opportunity to increase the accuracy of nursing sensitive outcome detection when using administrative data sources.