Identifying paediatric nursing-sensitive outcomes in linked administrative health data

BACKGROUND: There is increasing interest in the contribution of the quality of nursing care to patient outcomes. Due to different casemix and risk profiles, algorithms for administrative health data that identify nursing-sensitive outcomes in adult hospitalised patients may not be applicable to paediatric patients. The study purpose was to test adult algorithms in a paediatric hospital population and make amendments to increase the accuracy of identification of hospital acquired events. The study also aimed to determine whether the use of linked hospital records improved the likelihood of correctly identifying patient outcomes as nursing sensitive rather than being related to their pre-morbid conditions. METHODS: Using algorithms developed by Needleman et al. (2001), proportions and rates of records that identified nursing-sensitive outcomes for pressure ulcers, pneumonia and surgical wound infections were determined from administrative hospitalisation data for all paediatric patients discharged from a tertiary paediatric hospital in Western Australia between July 1999 and June 2009. The effects of changes to inclusion and exclusion criteria for each algorithm on the calculated proportion or rate in the paediatric population were explored. Linked records were used to identify comorbid conditions that increased nursing-sensitive outcome risk. Rates were calculated using algorithms revised for paediatric patients. RESULTS: Linked records of 129,719 hospital separations for 79,016 children were analysed. Identification of comorbid conditions was enhanced through access to prior and/or subsequent hospitalisation records (43% of children with pressure ulcers had a form of paralysis recorded only on a previous admission). Readmissions with a surgical wound infection were identified for 103 (4.8/1,000) surgical separations using linked data. After amendment of each algorithm for paediatric patients, rates of pressure ulcers and pneumonia reduced by 53% and 15% (from 1.3 to 0.6 and from 9.1 to 7.7 per 10,000 patient days) respectively, and an 84% increase in the proportion of surgical wound infection (from 5.7 to 10.4 per 1,000 separations). CONCLUSIONS: Algorithms for nursing-sensitive outcomes used in adult populations have to be amended before application to paediatric populations. Using unlinked individual hospitalisation records to estimate rates of nursing-sensitive outcomes is likely to result in inaccurate rates.