Risk factors for central line-associated blood stream infections (CLABSI) were investigated in a home parenteral nutrition (HPN) cohort.Methods
All patients receiving HPN between January 1st 2016 and July 31st 2017 were included. Data was analysed by individual central venous catheters (CVC) rather than by patients who may have had multiple CVCs. CLABSI (fever, positive blood cultures through line, no alternative infection source) was considered an end point even if the line was salvaged. Patient factors included underlying diagnosis, presence of end stoma, enterostomy, use of opioids/Tauralock and line type. Infection-free line survival was estimated by the Kaplan–Meier method. Univariate and multivariate Cox proportional-hazards models were used to determine the contribution of baseline variables.Results
251 CVCs were followed in 142 patients (67.6% female, age 17-92, median 56.5) over 94779 line days. Over the 18 month period 49 CLABSI were documented in 37 patients, equating to 0.52 infections per 1000 catheter days.Results
Cox proportional-hazards modelling revealed a number of individual factors associated with decreased CLABSI-free line survival. Patients with joint hypermobility and GI symptoms (OR 14.6, 95% CI 3.16-68.1, p<0.001); enteric dysmotility (OR 7.1, 95% CI 1.5-32.4, p=0.01) and use of a dual lumen line (OR 2.26, 95% CI 1.24-4.1, p=0.007) were also associated with increased risk of CLABSI. Other variables showing an association with increased risk of CLABSI included use of opioids (p<0.01), use of tauralock (p<0.001) and presence of a feeding or venting gastrostomy tube (p<0.01). Presence of an end stoma or a fistula was associated with decreased risk of CLABSI (p<0.01). Increasing age was associated with decreased risk of CLABSI (OR 0.96 for each additional year of life, 95% CI 0.94-0.97, p<0.001).Results
We built a multivariate hazards model incorporating variables showing a significant correlation with CLABSI in univariate analysis. Age and diagnosis remained significant; several of the variables showed strong co-linearity. An underlying diagnosis of joint hypermobility and GI symptoms was significantly associated with age in the lowest quartile. Since age and diagnosis were confounded, we tested the model with each in isolation and found that a model using diagnosis fitted the data better than a model using age (model concordance 0.815 for diagnosis vs 0.804 for age, p<0.0001 for both).Conclusion
We examined the effect of patient factors on CLABSI in our HPN cohort using a novel line-focussed approach. Our overall infection rates are comparable with contemporary case series. Opioid use was confirmed as a risk factor, but presence of an end stoma was not. Notably young patients with joint hypermobility and GI symptoms comprised a high risk group for line infection.