Comparing Self-Administered Outpatient Parenteral Antimicrobial Therapy (S-OPAT) with Alternative Delivery Models: Risk Factors for Treatment Failure and Complications in a UK Cohort
Our latest article was published as open access in Pathogens on 15 September 2026.
Comparing Self-Administered Outpatient Parenteral Antimicrobial Therapy (S-OPAT) with Alternative Delivery Models: Risk Factors for Treatment Failure and Complications in a UK Cohort
by
Oyewole Christopher Durojaiye, Charlotte Fiori, Kamille Jaff Maulion, and Evangelos I. Kritsotakis
Abstract: Self-administered outpatient parenteral antimicrobial therapy (S-OPAT) is an increasingly recognised antimicrobial delivery model, yet comparative evidence on its safety and effectiveness remains limited. We compared clinical outcomes across OPAT delivery models and examined predictors of adverse outcomes in S-OPAT. Adult OPAT episodes at a large UK teaching hospital (January 2023–January 2026) were categorised as S-OPAT, healthcare professional home-administered (H-OPAT), or clinic-based (C-OPAT). Primary outcomes were treatment failure, complications, and 30-day unplanned hospitalisation. Outcome risks were compared using robust multivariable Poisson regression with generalised estimating equations and Bonferroni correction. Among 1064 OPAT episodes, 286 (26.9%) were S-OPAT, and cure/clinical improvement occurred in 912/1064 (85.7%). Treatment failure occurred in 26/286 (9.1%) S-OPAT episodes, with no significantly higher risk than H-OPAT (adjusted relative risk [aRR], 0.72; 98.3% CI, 0.43–1.19) or C-OPAT (aRR, 0.94; CI 0.48–1.84). Rates of unplanned hospitalisation and OPAT-related complications were broadly comparable, except potentially vascular access-related events. Within S-OPAT, older age, higher comorbidity burden, diabetes, peripheral vascular disease, and central venous access were associated with treatment failure. Overall, these findings suggest that S-OPAT can be delivered safely and effectively to appropriately selected patients, while enhanced monitoring may be warranted for older individuals and those with greater comorbidity burden or central venous access devices.
Keywords: complications; outpatient parenteral antimicrobial therapy; S-OPAT; self-administration; treatment outcome
Access the full paper via:
Statistical methods
The effects of the OPAT delivery models on the risk of each binary outcome endpoint of interest were compared using risk ratios (RRs) estimated from a modified Poisson regression model with a log-link function and a robust ("sandwich") variance estimator [15,16]. OPAT episodes were the unit of analysis. Generalised estimating equations (GEE) with an exchangeable correlation structure were used to account for within-patient correlation among individuals with multiple OPAT episodes. To control the family-wise error rate across three comparisons (S-OPAT vs. H-OPAT, S-OPAT vs. C-OPAT, and C-OPAT vs. H-OPAT), pairwise p-values were adjusted using the Bonferroni method [17], applying an adjusted alpha level of 0.017 and calculating 98.3% confidence intervals (CIs) for the RRs. Omnibus p-values were also reported to provide an overall assessment of group differences.
A priori baseline covariates selected for inclusion as potential confounders or predictors in the regression models were patient characteristics (age, sex, comorbidities, and CCI) and treatment-related variables (infection type, antimicrobial class, concomitant IV antimicrobial therapy, type of vascular access device, and OPAT duration). Multicollinearity was assessed using variance inflation factors (Supplementary Table S2). To maintain parsimonious regression models and avoid multicollinearity, the CCI score was included as a covariate rather than each of its 19 individual component comorbidities. Antimicrobial class was also excluded due to high correlation with infection type. OPAT duration was considered an intermediate variable in the causal pathway between baseline covariates and outcomes and was therefore excluded from the regression models. Log-linearity of quantitative variables was examined using restricted cubic splines (Supplementary Figure S1). Age was retained as a continuous variable, whereas the CCI score was categorised into quartiles (0–1, 2, 3–4, and 5+).
The modified robust Poisson model was also used to explore risk factors for treatment failure, OPAT-related AEs, and 30-day unplanned hospitalisation among S-OPAT patients. In addition to the composite CCI score, the most common comorbid conditions—those present in at least 30 patients—were examined. Owing to the small number of outcome events in this subgroup, RRs were adjusted only for age and sex.
There were no missing data for any study variables. Stata 19 (StataCorp LLC, College Station, TX, USA) was used for data management and analysis. A two-tailed p-value < 0.05 was considered statistically significant.
Figure. Comparison of clinical outcomes across different OPAT delivery models.
Adj. p, adjusted p-value; Adj. RR, adjusted risk ratio; C-OPAT, clinic-based OPAT; CI, confidence interval; H-OPAT, healthcare-administered OPAT at home; OPAT, outpatient parenteral antimicrobial therapy; S-OPAT, self/carer-administered OPAT.
Table. Exploratory analysis of risk factors for adverse clinical outcomes among 286 self-/carer-administered OPAT episodes.
Graphical abstract
