Development of a Machine Learning-Based Model for Predicting the Incidence of Peripheral Intravenous Catheter-Associated Phlebitis
, , , , , , , , , , , , , , , oraz
31 lip 2024
O artykule
Kategoria artykułu: Research Article
Data publikacji: 31 lip 2024
Zakres stron: 232 - 244
Otrzymano: 07 lut 2024
Przyjęty: 15 cze 2024
DOI: https://doi.org/10.2478/jccm-2024-0028
Słowa kluczowe
© 2024 Hideto Yasuda et al., published by Sciendo
This work is licensed under the Creative Commons Attribution 4.0 International License.
Fig. 1.
Fig. 2.
Fig. 3.
Fig. 4.
Patient characteristics of the development and validation cohorts at ICU admission
| Age, mean (SD), years | 68.1 (15.2) | 67.9 (15.0) |
| Sex, male (n, %) | 1485 (61.9) | 637 (61.9) |
| Body height, mean (SD), cm | 161 (9.6) | 161 (9.6) |
| Body weight, mean (SD), kg | 59.9 (15.4) | 60.1 (14.9) |
| BMI, mean (SD) | 23.0 (4.7) | 23.1 (4.5) |
| APACHE II, mean (SD) | 19.2 (8.3) | 19.2 (8.2) |
| SAPS II, mean (SD) | 44.0 (19.5) | 44.3 (19.0) |
| SOFA, mean (SD) | 6.8 (3.7) | 6.7 (3.7) |
| Charlson comorbidity index, mean (SD) | 4.2 (2.6) | 4.2 (2.6) |
| ICU admission from (n, %) | ||
| Operation room | 921 (38.4) | 427 (41.5) |
| Emergency room | 965 (40.2) | 404 (39.3) |
| General ward | 363 (15.1) | 142 (13.8) |
| Outpatients | 18 (0.8) | 2 (0.2) |
| Transfer from other hospital | 133 (5.5) | 54 (5.3) |
| Type of admission to the ICU (n, %) | ||
| Elective surgical | 478 (19.9) | 214 (20.8) |
| Emergency surgical | 443 (18.5) | 213 (20.7) |
| Medical | 1479 (61.6) | 602 (58.5) |
| ICU admission category (n, %) | ||
| Cardiology | 860 (35.8) | 341 (33.1) |
| Pulmonary | 350 (14.6) | 160 (15.6) |
| Gastrointestinal | 243 (10.1) | 100 (9.7) |
| Neurology | 455 (19.0) | 212 (20.6) |
| Trauma | 95 (4.0) | 41 (4.0) |
| Urology | 21 (0.9) | 10 (1.0) |
| Gynaecology | 16 (0.7) | 8 (0.8) |
| Skin/tissue | 33 (1.4) | 17 (1.7) |
| Others | 327 (13.6) | 140 (13.6) |
| Sepsis at ICU admission (n, %) | 495 (20.6) | 209 (20.3) |
| Mechanical ventilation within 24 hours after admission to ICU (n, %) | 1433 (59.7) | 631 (61.3) |
PIVC characteristics during the insertion of the development and validation cohorts
| Catheter inserted by (n,%) | ||
| Doctor | 203/1,879 (10.8) | 81/801 (10.1) |
| Nurse | 1,673/1,879 (89.0) | 720/801 (89.9) |
| Inserted Site (n, %) | ||
| Upper arm | 245/2,378 (10.3) | 111/1,021 (10.9) |
| Forearm | 1,303/2,378 (54.8) | 546/1,021 (53.5) |
| Elbow | 113/2,378 4.8) | 50/1,021 (4.9) |
| Wrist | 118/2,378 5.0) | 44/1,021 (4.3) |
| Hand | 341/2,378 (14.3) | 166/1,021 (16.3) |
| Lower leg | 152/2,378 (6.4) | 73/1,021 (7.1) |
| Dorsal foot | 106/2,378 (4.5) | 31/1,021 (3.0) |
| Catheter material | ||
| PEU-Vialon |
777/2,400 (32.4) | 310/1,029 (30.1) |
| Polyurethane | 658/2,400 (27.4) | 320/1,029 (31.1) |
| Polyethylene | 0/2,400 (0) | 0/1,029 (0) |
| Tetrafluoroethylene | 910/2,400 (37.9) | 382/1,029 (37.1) |
| Others | 55/2,400 (2.3) | 17/1,029 (1.7) |
| Catheter gauge (n,%) | ||
| 14G | 1/2,357 (0.04%) | 0/1,011 (0) |
| 16G | 51/2,357 (2.2) | 22/1,011 (2.2) |
| 18G | 56/2,357 (2.4) | 33/1,011 (3.3) |
| 20G | 612/2,357 (26.0) | 276/1,011 (27.3) |
| 22G | 1,592/2,357 (67.5) | 662/1,011 (65.5) |
| 24G | 45/2,357 (1.9) | 17/1,011 (1.7) |
| Antiseptic solution before catheterisation (n,%) | ||
| None | 5/1,863 (0.3) | 3/802 (0.4) |
| Alcohol | 1,817/1,863 (97.5) | 782/802 (97.5) |
| 0.2% chlorhexidine alcohol | 14/1,863 (0.8) | 7/802 (0.9) |
| 0.5% chlorhexidine alcohol | 10/1,863 (0.5) | 5/802 (0.6) |
| 1.0% chlorhexidine alcohol | 12/1,863 (0.6) | 5/802 (0.6) |
| 10% povidone iodine | 2/1,863 (0.1) | 0/802 (0) |
| other | 3/1,863 (0.2) | 0/802 (0) |
| Use of ultrasonography (n,%) | 36/1,844 (1.9) | 22/792 (2.8) |
| Number of trials for insertion (n,%) | ||
| 1 | 1,501/1,834 (81.8) | 618/785 (79.7) |
| 2 | 221/1,834 (12.1) | 92/785 (11.7) |
| ≥3 | 112/1,834 (6.1) | 75/785 (9.6) |
| Difficulties with the insertions (n, %) | ||
| Easy | 882/1,811 (48.7) | 350/783 (44.7) |
| Slightly easy | 535/1,811 (29.5) | 237/783 (30.3) |
| Slightly difficult | 306/1,811 (16.9) | 150/783 (19.2) |
| Difficult | 88/1,811 (4.9) | 46/783 (5.9) |
| Glove (n,%) | ||
| Sterile | 14/1,836 (0.8) | 5/794 (0.6) |
| Non-sterile | 1,738/1,836 (94.7) | 758/794 (95.5) |
| Nothing | 84/1,836 (4.6) | 31/794 (3.9) |
| Dressing (n,%) | ||
| Chlorhexidine-impregnated dressing chrolehexidne | 0/2,377 (0) | 0/1,019 (0) |
| Sterile polyurethane dressing | 2,338/2,377 (98.4) | 989/1,019 (97.1) |
| Non-sterile polyurethane dressing polyuretane | 35/2,377 (1.5) | 25/1,019 (2.5) |
| Gauze dressing | 0/2,377 (0) | 1/1,019 (0.1) |
| Tape dressing | 4/2,377 (0.2) | 4/1,019 (0.4) |
| Any infection during catheter dwell (n, %) |
550/2,400 (22.9) | 253/1,029 (24.6) |
| Duration of catheter dwell, median (IQR), hour | 44.7 (20.7–79.1) | 41.5 (21.0–76.5) |
| Phlebitis (n,%) | 208/2,400 (8.7) | 105/1,029 (10.2) |
Difference of c-statistics in each model in the validation cohort
| Binary outcome models | |
| LASSO | 0.680 (0.625–0.735) |
| Random forest | 0.677 (0.622–0.731) |
| Gradient boosting tree | 0.646 (0.587–0.706) |
| Logistic regression model | 0.633 (0.575–0.691) |
| Survival outcome models | |
| Random survival forest | 0.655 (0.603–0.708) |
| Cox proportional hazards model | 0.516 (0.454–0.578) |
Difference of c-statistics in each model in the development cohort
| Binary outcome models | |
| LASSO | 0.699 (0.662–0.736) |
| Random forest | 0.980 (0.973–0.986) |
| Gradient boosting tree | 0.892 (0.870–0.914) |
| Logistic regression model | 0.725 (0.688–0.762) |
| Survival outcome models | |
| Random survival forest | 0.645 (0.606–0.684) |
| Cox proportional hazards model | 0.581 (0.542–0.621) |