Developing a risk prediction model for 30-Day unplanned hospitalisation in patients receiving outpatient parenteral antimicrobial therapy

Objectives Outpatient parenteral antimicrobial therapy (OPAT) is increasingly used to treat a wide range of infections. However, there is risk of hospital readmissions. The study aim was to develop a prediction model for the risk of 30-day unplanned hospitalisation in patients receiving OPAT.

Methods Using a retrospective cohort design, we retrieved data on 1073 patients who received OPAT over two years (01/2015 - 01/2017) at a large teaching hospital in Sheffield, UK. We developed a multivariable logistic regression model for 30-day unplanned hospitalisation and assessed its discrimination and calibration abilities, and internally validated using bootstrap resampling.

Results The 30-day unplanned hospitalisation rate was 11% (123/1073). The main indication for hospitalisation was worsening or non-response of infection (42%; 52/123). The final regression model consisted of age (adjusted odds ratio [aOR], 1.18 per decade; 95% confidence interval [CI], 1.04-1.34), Charlson comorbidity score (aOR, 1.11 per unit increase; 95%CI, 1.00-1.23), prior hospitalisations in past 12 months (aOR, 1.30 per admission; 95%CI, 1.17-1.45), concurrent intravenous antimicrobial therapy (aOR, 1.89; 95%CI, 1.03-3.47), and endovascular infection (aOR, 3.51; 95%CI, 1.49-8.28). Mode of OPAT treatment was retained in the model as a confounder. The model had adequate concordance (c-statistic 0.72; 95%CI 0.67-0.77) and calibration (Hosmer-Lemeshow P=0.546; calibration slope 0.99; 95%CI 0.78-1.21) and low degree of optimism (bootstrap optimism corrected c-statistic, 0.70).

Conclusions We identified a set of six important predictors of unplanned hospitalisation based on readily available data. The prediction model may help improve OPAT outcomes through better identification of high-risk patients and provision of tailored care.