Predictors of inappropriate PICC placement, estimated using multivariable regression
| Predictors | OR (95% CI) | P value |
| Patient factors | ||
| Age | 1.01 (1.01 to 1.02) | <0.001 |
| Female gender | 1.41 (1.13 to 1.77) | 0.003 |
| Charlson Comorbidity Index score 1 | 3.33 (2.48 to 4.48) | <0.001 |
| Charlson Comorbidity Index score 2 | 1.71 (1.27 to 2.30) | <0.001 |
| Charlson Comorbidity Index score 3 or greater | 2.98 (2.35 to 3.78) | <0.001 |
| Laboratory-based Acute Physiology Score | 1.01 (1.01 to 1.02) | <0.001 |
| PICC placed in ICU | 1.97 (1.52 to 2.54) | <0.001 |
| Physician factors | ||
| Proportion of patients receiving a PICC | 1.00 (0.97 to 1.02) | 0.71 |
| Male gender | 0.84 (0.65 to 1.09) | 0.19 |
| Years of independent practice | 1.01 (1.00 to 1.01) | 0.28 |
| Average annual case load | 1.00 (1.00 to 1.00) | 0.79 |
| Hospital site | ||
| Hospital A | 1.46 (1.05 to 2.03) | 0.024 |
| Hospital B | 0.92 (0.64 to 1.31) | 0.63 |
| Hospital C | 0.71 (0.47 to 1.09) | 0.12 |
| Hospital D | 1.25 (0.87 to 1.80) | 0.23 |
A multivariable generalised estimating equation model was fit to explore the association of patient, physician and hospital predictors of inappropriate PICC use. For multilevel categorical variables, the following categories were used as referents: Charlson Comorbidity Index score 0, Hospital E.
ICU, intensive care unit; PICC, peripherally inserted central catheter.