Evaluation of health-related quality of life in patients receiving outpatient parenteral antimicrobial therapy (OPAT) in a UK setting

2024-03-21

Our latest research article just appeared in Expert Review of Anti-infective Therapy, 21 March 2024.

Evaluation of health-related quality of life in patients receiving outpatient parenteral antimicrobial therapy (OPAT) in a UK setting

Oyewole Christopher Durojaiye and Evangelos I. Kritsotakis

Access the full paper via:

Abstract

Background: Studies assessing the benefits of outpatient parenteral antimicrobial therapy (OPAT) have paid less attention to patient-centered factors such as patients' experiences and their health-related quality of life (HRQoL).

Research design and methods: Prospective before-and-after quasi-experimental study enrolled adult patients receiving OPAT at a tertiary hospital in Derbyshire, UK, between October 2022 and October 2023. Consenting patients completed paired EQ-5D-3 L questionnaires before OPAT initiation and upon completion of therapy or 30 days after its commencement (whichever occurred first). Changes and predictors of change in HRQoL indicators and associations with clinical outcomes (treatment failure, adverse events, and 30-day unplanned readmission) were examined.

Results: Health state index and visual analogue scale (EQ VAS) scores of 162 enrolled patients were significantly lower than the UK population averages, but the patients experienced significant improvements in both scores and in four EQ-5D dimensions (mobility, self-care, usual activities, and pain/discomfort). Baseline health index and EQ VAS scores were significant independent predictors of positive changes in HRQoL scores.

Conclusions: OPAT is associated with improved patient-reported quality of life and facilitates early return to work or school. Nevertheless, it is crucial to closely monitor patients with a lower baseline quality of life to optimize their overall OPAT experience.

Keywords: EQ-5D-3L; OPAT; Outpatient parenteral antimicrobial therapy; Patient-centered outcomes research; Quality of life.

Statistical methods

A minimally important difference (MID) of 0.074 units was defined a priori as a clinically relevant change of the EQ-5D index score [Citation19]. With an expected standard deviation (SD) of 0.29 [Citation 19], an MID of 0.074 corresponds to an effect size of 0.25, which is regarded as a small to medium effect. We calculated that approximately 160 patients were required to detect a change of 0.074 units in mean EQ-5D index scores from baseline to the final assessment, with 90% power and a 5% significance level in a paired t-test. This sample size allows for reliable estimation of 16 predictors in a multiple linear regression model, based on the ''rule-of-thumb'' of 10 subjects per variable [Citation 20].

Data analysis

Descriptive statistics were used to summarize sociodemographic, clinical, and OPAT outcome data. To present the EQ-5D-3 L descriptive system, we calculated the numbers and percentages of patients reporting each level of problems on each of the dimensions at both the baseline and final assessments. Additionally, we determined the means (with SDs) of EQ VAS and EQ-5D-3 L index scores at these two time points. The observed baseline mean scores were compared with published mean scores for the UK general population using one-sample t-tests. The mean health state index and EQ VAS scores for the UK general population have previously been reported as 0.86 and 82.5, respectively [Citations 22, 23].

To assess the magnitude and statistical significance of changes between the two time points, we calculated 95% confidence intervals (CIs) for the differences in the proportion of patients reporting improvements in each dimension and performed McNemar's test. For the summary EQ-5D-3 L index and EQ VAS scores, we calculated 95% CIs for mean differences and paired t-test statistics. To aid in the interpretation of the estimated changes, we computed standardized effect sizes using Cohen's g and d statistics for proportion and mean differences, respectively [Citation 25]. Following Cohen's benchmarks, we identified effect sizes as negligible (g < 0.05, d < 0.2), small (0.05 ≤ g < 0.15, 0.2 ≤ d < 0.5), medium (0.15 ≤ g < 0.25, 0.5 ≤ d < 0.8), or large (g ≥ 0.25, d ≥ 0.8) [Citation 25].

Changes in EQ-5D-3 L were also analyzed using the Paretian Classification of Health Change (PCHC) approach [Citation 26]. Patients were classified as 'improved' if they demonstrated improvement in at least one dimension without worsening in any other dimension of the EQ-5D-3 L system; 'worsened' if they worsened on at least one dimension and did not show improvement in any other dimension; and 'no change' if they exhibited the same response in each dimension at baseline and the final assessment. Patients who improved in one or more dimensions and worsened in others were classified as 'mixed change.' PCHC summarizes overall changes in patients' self-reported health without relying on preference weights (utility scores) from the general public, thereby avoiding potential inference bias [Citation 27]. We estimated the proportion of patients in each PCHC class by calculating 95% CIs using Wilson's score method.

Analysis of covariance with ordinary least squares estimation was employed to identify groups of patients that derived heterogeneous effects from OPAT treatment in terms of improvements in HRQoL. EQ VAS and EQ-5D-3 L health index scores at the follow up assessment were examined as response variables in two separate multiple linear regression models. A set of a priori selected baseline covariates were examined, including patient characteristics (age, sex, CCI, and clinical frailty score) and therapy-related variables (combination antimicrobial therapy, indication for OPAT, mode of delivery, and type of vascular access device). In these models, we controlled for imbalances in baseline EQ VAS and EQ-5D-3 L scores to estimate the direct causal effects of baseline exposures [Citation28]. Covariate-specific effect sizes were assessed using Cohen's f2 statistic and were interpreted as small (f2 ≥ 0.02), medium (f2 ≥ 0.15), or large (f2 ≥0.35) effects, respectively. Multicollinearity of predictor variables was ruled out by examining variance inflation factors (Supplementary Table S2). Graphical inspection of residuals indicated deviations from normality and homoscedasticity (Supplementary Figures S2 and S3). Therefore, bootstrapping was performed (1,000 replications) to estimate the standard errors and P-values. CIs were constructed using the bias-corrected and accelerated bootstrap method.

Multivariable Poisson regression with a log-link function and robust variance estimation was performed to estimate relative risks (RR) with 95% CIs [Citation 29]. This analysis aimed to quantify the associations between self-reported HRQoL indices at baseline (EQ-5D-3 L dimensions, full health state, EQ-5D-3 L index score, and EQ VAS score) and subsequent patient outcomes (complications during OPAT, treatment failure, and 30-day unplanned hospitalizations). The models were adjusted for sex, age, and CCI. The log-linearity of continuous variables was examined using restricted cubic splines (Supplementary Figures S4 and S5).

No missing values were observed for any of the study variables. Two-sided p-values were reported for all analyses, and statistical significance was considered at p < 0.05. All analyses were performed using Stata version 18 (Stata Corporation, College Station, Texas, U.S.A.).

Figure 1: Paretian Classification of Health Change (PCHC) from baseline to final assessment for N = 162 patients. Vertical lines are 95% confidence intervals.

Figure 2: EQ-5D-3L descriptive system for patients (N = 162) at both baseline and at final assessments.

Table . Changes in self-reported health-related quality of life indication of patients (N = 162), from baseline to final assessment.