Study design
We conducted a single centre, prospective cohort study to determine the
efficacy of two models at predicting unplanned readmissions in those
aged 75 and older who were initially admitted with an acute medical
condition. Data collection took place from February to April 2019 at a
large teaching hospital in central London. All patients 75 years of age
and older admitted to the acute admissions ward during a continuous
30-day period were included in the study. Exclusion criteria included
patients transferred from the acute admissions ward to another inpatient
ward and those who died prior to 30-days post-discharge. Data collection
was carried out Monday to Friday and as such patients both admitted and
discharged within the same Saturday-Sunday period may have been omitted
from collection. The primary outcome measure was the area under the
receiver operating characteristic curve (AUROC), also termed the
c-statistic, for the BOOST and LACE scoring systems. A power calculation
was completed using the R-based web tool easyROC [12] with sample
size determined using a type I error of 0.05, a power of 0.8, a
c-statistic of 0.7, and an allocation ratio of 6. The suggested sample
size was 152 with 19 positive and 133 negative cases. The allocation
ratio was predicted from literature suggesting a readmission rate of
14% in the elderly [13]. Secondary outcomes included the
significance of individual predictor model components, and the
sensitivity, specificity, and odds of high and low risk LACE and BOOST
patient groups.