The aim of the study was to compare the impact of 3-day routine resite, with clinically indicated resite, on peripheral IVD complications.
Open (nonblinded), parallel group RCT.
The study was approved by the Tasmanian State Human Research Ethics Committee. Written informed consent was obtained prospectively from all participants.
Setting and sample
The study was undertaken at a large regional teaching hospital in Australia which functions as the tertiary referral centre for the northern half of the State of Tasmania. The hospital has more than 32,000 separations per annum, with a spectrum of medical and surgical specialties. Eligible patients were at least 18 years of age and scheduled or expected to have a peripheral IVD indwelling for at least 4 days, and they gave written informed consent. Exclusion criteria were immunosuppression, current bloodstream infection or an IVD already in situ for >48 hours. IVDs were inserted and cared for by the general nursing and medical staff; there was no special IV team or service.
Sample size calculations were performed using PASS 2008 (Version 8.0.8; Kaysville, UT) to detect a change in rates by 30% (from 36% to 25%, two-tailed α = 0.05, 90% power) on the basis of the complication rates of routinely resited IVs in a previous study . Although this indicated that n = 378 per group (total 756) were required, the study was ceased early (total n = 606 IVs) because all investigators left the employment of the institution. Consequently, study power was reduced, but remained over 80% (required minimum n = 282 per group).
All adult patients admitted to the inpatient acute medical and surgical wards of the study hospital were screened by a full-time research nurse. This excluded paediatric, day-surgery, mental health, obstetric, critical care and dialysis units.
Patients were randomly assigned (computer generated) in a 1:1 allocation ratio to either the 'routine replacement' (control) or 'clinically indicated replacement' (intervention) group. Assignment was concealed until randomisation by use of a telephone service. A tag was placed on the insertion site indicating the study group. All devices for the patient were managed as per randomised group. The intervention group did not have their IVD resited unless clinically indicated. This decision was made by the treating clinician (not the investigators), who ordered IVD resite if the device failed or phlebitis occurred and ongoing IV treatment was required. The control group had a new device relocated to a different site by the general medical or nursing staff every 3 days. Control devices could also be removed at any stage by the clinical staff if they were not required or if complications occurred. Clinical nursing and medical staff undertook insertion and follow-up care of all IVDs as per the CDC Guidelines .
Laboratory staff undertaking microbiological culture assessments were blinded to the study group. Due to the nature of the intervention, patients, research, and clinical staff were unable to be blinded. However, the investigators had no involvement in assessing or documenting complications.
IVDs were assessed by the clinical nursing staff on each nursing shift for complications as part of standard clinical practice in the hospital. Times and dates of device insertion and removal were recorded along with the reason for device removal and any protocol deviations. A full-time research nurse collected data from the hospital chart and sought clarification from patients and clinical staff if necessary. Microbiological investigations (device tip, blood cultures and site cultures) were performed by the clinical staff on clinical suspicion of infection by the treating clinician. Demographic and clinical data were collected on age, sex, diagnosis at hospital admission, phlebitis risk based on Tagar et al.'s classification (low/medium/high risk) , past history of phlebitis, any comorbidities requiring active medical treatment (e.g., type 2 diabetes or congestive heart failure), haemoglobin, concurrent infection at other sites, antibiotic therapy, type of surgery, type of infusate and any additives (and their level of irritability), vein and skin quality assessment, size of device, insertion site, health professional inserting the device, and setting for insertion, presence of other vascular devices, wound drains and urinary catheters. Vein quality was assessed as good (vein easy to visualise and easy to palpate with tourniquet on), fair (not easily visible but can palpate with tourniquet), or poor (veins small, scarred or difficult to palpate with tourniquet; may require heat pack to aid vasodilation). Skin quality was assessed as good (healthy, well hydrated, elastic), fair (mildly dehydrated, reduced elasticity), or poor (papery, dehydrated, or reduced elasticity).
The primary outcome was a composite measure of any complication causing unplanned cannula removal prior to completion of IV treatment. The composite included phlebitis, infiltration, occlusion, accidental removal, local infection, and IV device-related bloodstream infection (IVD-BSI). These were also analysed individually as secondary endpoints. A composite measure was chosen due to the low rates of these conditions individually and to the assumption that they are comparable measures of 'infusion failure'; that is, the device can no longer be used to deliver treatment. This approach has been used in previous studies on the topic [15–17]. Phlebitis was defined as two or more of pain, erythema, purulence, streak formation, or a palpable venous cord . Local infection IVD-BSI (bacteremia/fungemia with at least one positive blood culture obtained from a peripheral vein, clinical manifestations of infection, and no apparent source for the bloodstream infection (BSI) except the device with or without positive tip or entry site swab culture) were defined using CDC criteria . Other secondary outcomes were time in situ (hours of catheterisation from insertion to removal, both per patient and per device) ; IVDs per patient (number of peripheral devices inserted to complete the course of treatment) ; costs (calculations based on 20 minutes nursing or medical insertion time at relevant rates , plus the cost of the required equipment (cannula, insertion pack including dressing and solution, gloves, saline, syringe, extension tubing and starter pack for all plus fluid bag, tubing and secondary tubing for medication administration for those patients requiring this) for insertions, nursing time and equipment to routinely remove IVDs that were otherwise functional, and the costs of treating any complications that occurred (e.g., IVD-BSI). Cost calculations were undertaken from the viewpoint of the hospital using negotiated wage costs and purchasing agreements for government hospitals in the State of Tasmania. Costs would be similar for other Australian hospitals.
All randomised patients were analysed by intention to treat. Each patient was randomised once and could have multiple IVDs, with subsequent IVD resites managed as per the randomised group. Relative incidence complication rates per 1,000 IVD days and 95% confidence intervals were calculated to summarise the impact of clinically indicated replacement relative to 3-day replacement. Kaplan-Meier survival curves were drawn to compare time to first IVD complication between patients in the two study groups. To assess for any potential impact of protocol violations, a per protocol analysis was also undertaken. All clinical and demographic variables were subjected to univariate testing against the primary endpoint to guide selection of possible covariates for the multivariable model. Cox proportional hazards regression modelling was used to examine the impact of age, gender, oncology status, number of comorbidities (nil, one, two, or more than two), IV gauge, site, vein quality, skin quality, oral antibiotics, IV antibiotics, wound drain, inserter occupation, initial versus subsequent IVDs, phlebitis in a preceeding IVD, haemoglobin level, parenteral nutrition, continuous versus intermittent infusion, patient risk category and study group on the outcome of time to complication events using an additive model [3, 5, 7, 20–25]. In addition, to adjust for any inherent correlations or codependencies in the failure times of IVDs (i.e., same patient multiple failure-time data) within the Cox model, we also used the Prentice-Williams-Peterson conditional risk-set method . The Mann-Whitney test was used to compare various secondary outcomes between study groups. Cost differences were calculated using arithmetic means and the t-test . P values <0.05 were considered significant. All statistical data were entered and analysed using SPSS (Version 15.0; Chicago, IL) and Stata (Version 8.2; College Station, TX).