New Risk Model Flags Skin Damage Around PICC Lines Before It Strikes

For millions of patients who depend on peripherally inserted central catheters, known as PICCs, the thin plastic tubes threaded into the upper arm are lifelines that deliver chemotherapy, nutrition and antibiotics over weeks or months. Yet the very dressings and adhesives that keep these catheters secure can quietly wage war on the skin beneath them. A condition called catheter-associated skin injury, or CASI, can leave the insertion site red, blistered, eroded or torn, forcing treatment interruptions and sometimes the premature removal of the catheter itself. Now, a large prospective study from a tertiary hospital in China has taken a major step toward predicting which patients will develop this painful complication before the first sign of damage appears, and the results carry both promise and important caveats for clinical practice.

The research, published in Nursing Open, followed 1,877 adults who received PICCs at a single comprehensive hospital between June 2022 and September 2023. Trained investigators assessed the skin around each catheter insertion site every week until the catheter was removed or six months had passed, whichever came first. The team defined CASI according to the Infusion Therapy Standards of Practice, which describe it as any abnormality such as erythema, vesicles, bullae, erosion or tearing observed in the area covered by the dressing or securement device for thirty minutes or longer after removal. Suspected lesions were photographed, coded and archived, and uncertain cases were jointly reviewed by the principal investigator and a dermatologist to reach a final diagnosis.

Overall, 156 patients, or 8.31 percent of the cohort, developed PICC-related CASI. The incidence was 8.86 percent in the development cohort of 1,309 patients and 7.04 percent in a temporal validation cohort of 568 patients enrolled later in the study period. Most injuries were mild, accounting for roughly 82 percent of cases in both groups, while moderate and severe injuries made up the remainder. The average catheter dwell time at the moment of injury was around 65 to 66 days, suggesting that the risk accumulates gradually over the weeks a catheter remains in place. These rates are lower than figures reported in earlier studies, which the authors attribute partly to their inclusion of both cancer and non-cancer patients and partly to the hospital’s specialist intravenous therapy team, which managed every catheter insertion and maintenance session with standardized procedures.

To build the prediction model, the researchers assembled 38 candidate predictors spanning six domains: demographic characteristics, disease-related information, nutritional status, treatment-related factors, catheter-insertion characteristics and catheter-maintenance practices. Rather than screening variables one by one, they entered all 38 candidates directly into a least absolute shrinkage and selection operator, or LASSO, logistic regression with ten-fold cross-validation. This machine-learning technique penalizes weak predictors, shrinking their coefficients toward zero and effectively discarding them, which helps guard against overfitting when the number of events is limited. The penalty parameter was chosen using the conservative one-standard-error criterion, yielding a parsimonious set of seven variables with non-zero coefficients.

Those seven candidates were then refitted in an ordinary multivariable logistic regression, and one of them, catheter diameter, was dropped because its rarest category contained only six participants and no injury events, producing unstable estimates. The final model settled on six predictors: a primary tumour diagnosis, a history of previous catheterization, use of iodophor disinfectant, use of a functional dressing, cytotoxic drug exposure and hormonal drug exposure. Three of these factors raised risk while two appeared protective. A history of catheterization was the strongest predictor, multiplying the odds of skin injury by just over five. Cytotoxic chemotherapy more than quadrupled the odds, and a cancer diagnosis nearly quadrupled them. Hormonal therapy, which included corticosteroids and sex hormones, roughly doubled the odds. In contrast, iodophor disinfectant reduced the odds by nearly 90 percent, and functional dressings such as thin foam or hydrocolloid materials reduced them by about 84 percent.

The biological logic behind these associations is plausible. Repeated catheter placements leave patients with fewer suitable veins and expose the skin to cycles of adhesive application and removal, which mechanically strips the protective barrier. Cytotoxic agents are notorious for causing erythema, dryness and blistering, and they can also weaken immunity and nutrition, compounding vulnerability. Corticosteroids thin the skin, impair collagen synthesis and delay repair. Cancer itself, along with radiotherapy and nutritional impairment, undermines skin integrity from multiple directions. On the protective side, povidone-iodine may spare the skin the irritant effects of alcohol-based chlorhexidine preparations, a pattern consistent with the CLEAN trial, which found more severe skin reactions with chlorhexidine-alcohol in patients with short-term central venous catheters. Foam and hydrocolloid dressings, meanwhile, manage local moisture, reduce friction and shear, and cushion fragile skin better than standard transparent polyurethane films.

Performance testing revealed a sobering story about transportability. In the development cohort, bootstrap internal validation with 1,000 resamples produced an optimism-corrected area under the receiver operating characteristic curve of 0.811, a level generally considered acceptable discrimination, and the calibration slope of 0.943 sat close to the ideal value of one. But when the model was applied unchanged to the temporal validation cohort of 568 later-enrolled patients, discrimination fell to 0.681. The calibration intercept of minus 0.668 and slope of 0.447 indicated systematic overprediction: the model tended to overestimate risk, particularly at the higher end of the predicted-probability range, where relatively few events supported the estimates. Decision-curve analysis suggested potential net benefit over treating everyone or no one, but only in parts of the threshold range and only convincingly in the development data.

The authors are candid about why performance degraded and what it means. Seasonal differences in recruitment, shifts in case mix, variation in treatment exposure and local antiseptic and dressing practices may all have contributed, since temperature and humidity are known to influence skin-barrier function and adhesive irritation. They also acknowledge methodological limitations: no formal sample-size calculation was performed before recruitment, the LASSO selection step was not repeated inside the bootstrap loop, and the subsequent unpenalized refit introduces additional selection uncertainty. The study was conducted at a single centre with convenience sampling, and several potentially relevant factors, including dressing-change frequency, product type, baseline skin condition and cumulative medication exposure, were not captured. For these reasons, the team classifies the tool as a development-stage model that requires recalibration and independent multicentre validation before any routine clinical use.

Even so, the findings offer nurses and clinicians a structured way of thinking about who needs the closest surveillance. Patients with cancer, previous catheterizations and intensive anticancer regimens emerge as the clear high-risk group, and the study highlights that two modifiable elements of catheter care, the choice of antiseptic and the choice of dressing, may meaningfully shift the odds. The authors suggest that povidone-iodine and functional dressings could be considered for patients at elevated risk of skin injury, while stressing that these observational associations need prospective confirmation before practice changes. Future work should define clinically meaningful risk thresholds, link them to specific preventive actions such as enhanced skin assessment or alternative antiseptic selection, and test whether embedding the model into nursing pathways actually reduces the burden of catheter-associated skin injury. Until then, the study stands as a careful, transparently reported reminder that the skin around a lifesaving catheter deserves the same vigilance as the device itself.

Subject of Research: Development and temporal validation of a clinical risk prediction model for PICC-related catheter-associated skin injury

Article Title: Development and Temporal Validation of a Prediction Model for PICC‐Related Catheter‐Associated Skin Injury

Article References: Development and Temporal Validation of a Prediction Model for PICC‐Related Catheter‐Associated Skin Injury. (n.d.). https://doi.org/10.1002/nop2.70895

Image Credits: AI Generated

DOI: 10.1002/nop2.70895

Keywords: PICC, catheter-associated skin injury, risk prediction model, LASSO regression, nursing, vascular access devices, chemotherapy, iodophor disinfectant, functional dressings, temporal validation, calibration, oncology