Clinical background
Peripherally inserted central catheters (PICCs) are associated with symptomatic deep vein thrombosis (DVT) of the upper limb, with incidences ranging from about 1% to over 10% depending on the patient population and catheter size [4]. The risk is influenced both by patient-related factors (cancer, previous venous thromboembolism, leucocytosis) and by device-related factors (number of lumens, catheter diameter, concurrent catheters). Decisions about which type of PICC to choose, how many lumens are needed and how closely the patient should be monitored have traditionally been informal. The Michigan risk score was developed to quantify this risk and thereby support a structured choice of device and intensity of monitoring.
Calculating the Michigan risk score
The score is the sum of five variables:
where
| Variable | Value | Points |
|---|---|---|
| : Number of PICC lumens | Single | 0 |
| Double | 1 | |
| Triple or quadruple | 2 | |
| 5 or more | 3 | |
| : Previous venous thromboembolism | None | 0 |
| More than 30 days before insertion | 2 | |
| Within 30 days of insertion | 3 | |
| : Active cancer | No | 0 |
| Yes | 3 | |
| : Another central venous catheter present at PICC insertion | No | 0 |
| Yes | 1 | |
| : White cell count >12 × 10⁹/L at insertion | No | 0 |
| Yes | 1 |
The risk classes are:
- Class I: 0 points
- Class II: 1–2 points
- Class III: 3–4 points
- Class IV: ≥5 points
The derivation cohort consisted of 23,010 adult patients who had PICCs inserted at hospitals belonging to the Michigan Hospital Medicine Safety Consortium between 2009 and 2012. The outcome was symptomatic, imaging-confirmed upper-limb DVT. In all, 475 patients (2.1%) developed PICC-related DVT [1]. A logistic mixed-effects model with hospital-specific random intercepts identified five variables independently associated with the outcome. Points were assigned on the basis of the estimated odds ratios, and the model was internally validated by bootstrapping to assess discrimination and calibration [1].
The precursor to the Michigan risk score was a smaller retrospective cohort study by the same group, comprising 966 PICC insertions, in which a cancer diagnosis and catheter gauge (5 and 6 French) were the factors that remained significant after multivariable adjustment [5]. That study laid the basis for including the number of lumens and catheter size in the final score.
Interpretation in practice
In the derivation cohort, the risk classes corresponded to the following observed thrombosis rates and odds ratios [1]:
| Risk class | Score | Thrombosis rate | Odds ratio (vs class I) |
|---|---|---|---|
| I | 0 | 0.9% | Reference |
| II | 1–2 | 1.6% | 1.68 (95% CI 1.19–2.37) |
| III | 3–4 | 2.7% | 2.90 (95% CI 2.09–4.01) |
| IV | ≥5 | 4.7% | 5.20 (95% CI 3.65–7.42) |
Class I (0 points): Low risk. A single-lumen PICC is preferable if clinically feasible. Routine ultrasound surveillance or pharmacological prophylaxis is not justified by current guidelines.
Class II (1–2 points): Moderately increased risk. Consider whether a double lumen is genuinely required, or whether a midline catheter might be an alternative. If a double lumen is necessary, choose the smallest possible French size. Clinically observe the arm during the dwell time.
Class III (3–4 points): Increased risk. Here the choice of device is decisive. Use a single lumen if possible, avoid concurrent central catheters, and consider ultrasound assessment if symptoms occur. In cancer patients, alternative vascular access (a port catheter) should be considered if long-term treatment is anticipated.
Class IV (≥5 points): High risk. The thrombosis risk is more than fivefold that of class I. If a PICC is the only realistic access, use a single lumen, avoid concurrent catheters, and inform the patient about symptoms of DVT. Close monitoring and a low threshold for ultrasound examination if thrombosis is suspected.
Validation and performance
In the derivation cohort the model showed good calibration and discrimination on internal validation by bootstrapping [1]. External validation has, however, given mixed results.
A Brazilian multicentre study at 16 hospitals included 12,725 PICC insertions in 11,135 patients (mean age 66.4 years; 51% women). DVT occurred in 129 patients (1.0%). The AUC was 0.70 for the multivariable model and 0.67 for the risk classification. Only the number of PICC lumens and previous VTE were significantly associated with DVT in this cohort. The odds ratios for classes III and IV compared with class I were 2.83 (95% CI 1.51–5.30) and 3.01 (95% CI 1.41–6.41) respectively [2]. Discrimination was thus lower than in the derivation cohort, but the direction of the risk grouping was confirmed.
A Chinese study from Peking Union Medical College Hospital included 2,163 patients with a total of 206,132 catheter days. Here the outcome was symptomatic upper-limb thrombosis, which occurred in 56 patients (2.6%). The AUC for the Michigan risk score was only 0.405 (95% CI 0.303–0.508), indicating worse than chance discrimination [3]. The thrombosis incidence was paradoxical in relation to the risk classes: 4.9% in class I, 7.5% in class II, 2.2% in class III and 0% in class IV. This study questions the score's generalisability to populations with different epidemiology, possibly influenced by differences in insertion technique, patient selection and care setting.
A systematic review and meta-analysis of 40 studies confirmed the strong relationship between catheter diameter and symptomatic DVT: pooled rates were 0.89% for 3 French, 3.26% for 4 French, 5.46% for 5 French and 10.66% for 6 French, with a significant difference between 4 and 5 French [4]. This supports the number of lumens, which correlates with catheter diameter, as a central and modifiable risk factor.
Limitations
The Michigan risk score was derived from a North American hospital cohort and has not been validated in European practice. The Brazilian validation showed moderate discrimination and the Chinese one showed inadequate discrimination, suggesting that the score's performance is context-dependent [2, 3].
The score does not capture several factors known to influence thrombosis risk: catheter diameter in French (only the number of lumens is included), the catheter-to-vein ratio, insertion technique (ultrasound-guided versus blind), tip position, infusion of irritant solutions, and impaired renal function. An Italian retrospective study of 1,431 patients found that the number of lumens was not independently associated with catheter-related thrombosis after adjustment for other factors, further questioning the score's choice of variables [6].
The score refers to symptomatic DVT. Asymptomatic thrombosis, which is commoner, is not captured. This matters because ultrasound screening can identify thromboses that the score does not predict.
The most modifiable factors in the score are the number of lumens and concurrent central catheters. Active cancer and previous VTE are patient-related and not modifiable at the time of insertion. A white cell count >12 × 10⁹/L may reflect infection or inflammation and is partly modifiable.
Routine pharmacological prophylaxis to prevent PICC-related thrombosis is not recommended in international guidelines, and the Michigan risk score should not be used to justify decisions about prophylaxis. An Italian retrospective cohort study did, however, find that both therapeutic and prophylactic anticoagulation were protective against catheter-related thrombosis (OR 0.007 and 0.328 respectively), but the evidence is not sufficient to change guidelines [6].
References
- Chopra V, Kaatz S, Conlon A, et al. The Michigan Risk Score to predict peripherally inserted central catheter-associated thrombosis. J Thromb Haemost 2017. PMID: 28796444
- Rabelo-Silva ER, Saffi MAL, Hirakata VN, et al. External Validation of the Michigan Risk Score for Predicting Peripherally Inserted Central Catheter-Related Deep Vein Thrombosis: A Multicenter Study in Brazil. J Infus Nurs 2026. PMID: 41494176
- Kang J, Sun W, Li H, et al. Validation of Michigan risk score and D-dimer to predict peripherally inserted central catheter-related thrombosis: A study of 206,132 catheter days. J Vasc Access 2022. PMID: 33860712
- Bahl A, Alsbrooks K, Gala S, et al. Symptomatic Deep Vein Thrombosis Associated With Peripherally Inserted Central Catheters of Different Diameters: A Systematic Review and Meta-Analysis. Clin Appl Thromb Hemost 2023. PMID: 37366542
- Chopra V, Ratz D, Kuhn L, et al. Peripherally inserted central catheter-related deep vein thrombosis: contemporary patterns and predictors. J Thromb Haemost 2014. PMID: 24612469
- La Cava L, Giustivi D, Bartoli A, et al. Risk Factors for Catheter-Related Thrombosis. J Clin Med 2026. PMID: 42194892
