In this issue of Blood, Darzi et al have reported the results of a systematic review of the literature for prognostic factors for venous thromboembolism (VTE) and bleeding in acutely ill, critically ill, and chronically ill hospitalized medical patients.1 

The authors identified 14 studies of prognostic factors for VTE and 3 studies of prognostic factors for major or clinically relevant nonmajor bleeding in hospitalized patients. The authors report finding moderate-certainty evidence of association among 18 characteristics and increased risk of VTE, with variable strengths of association. Only 8 prognostic factors, including male sex, elevated D-dimer, elevated C-reactive protein (CRP), elevated heart rate, thrombocytosis, Barthel Index (a measure of functional independence) ≤9, immobility, paresis, previous VTE, thrombophilia, and active or past history of cancer had calculated odds ratios (ORs) of 2 or more.

There was moderate certainty of association between 15 characteristics and increased rates of bleeding, with similar variability in strength of association. Only 6 prognostic factors were calculated to increase the OR to 2 or greater, including morbid obesity, anemia, gastroduodenal ulcers, rehospitalization, and critical illness.

VTE is a common and morbid complication of hospitalized, medically ill patients and is responsible for up to 10% of hospital deaths. Although mechanical methods of prevention are available, the mainstay is pharmacologic prophylaxis, typically with either low molecular weight or unfractionated heparin. These agents, in turn, are associated with increased risk of bleeding, and thus the identification of those patients most likely to experience benefit and least likely to experience harm associated with these interventions is a long-standing clinical dilemma. To date, among medical inpatients, no clearly defined subpopulation for whom the benefits exceed the risks of pharmacologic prophylaxis has been identified.

A number of risk assessment models (RAMs) are available to help guide clinicians in selecting those patients at highest risk for VTE. Currently available models for medical patients include the Padua prediction score,2  and the International Medical Prevention Registry on Venous Thromboembolism (IMPROVE) VTE score.3  The Caprini risk assessment4  has been validated in surgical patients, a fundamentally different patient population for whom risk predictors and need for prophylaxis is better understood. These models use varying numbers of clinical characteristics to calculate a risk score which, in turn, is associated with a probability of VTE diagnosis within 90 days. However, these scores carry a number of limitations: they are both time- and labor-intensive because of the large number of parameters required (28 for Caprini), some of which are laboratory values not typically collected as part of standard of care. In addition, high-quality data to guide bleeding risk assessment with the use of pharmacologic prophylaxis are lacking. Ultimately this results in an incomplete picture for the clinician who is attempting to base patient management on these models, and decisions are often made based upon experience and/or judgment rather than objective risk scores. This in turn leads to considerable inconsistency in the use of VTE prophylaxis, which may or may not be superior to random assignment or to consulting my childhood Magic 8-Ball toy (“Outlook not so good”).

The findings of Darzi et al highlight one of the main challenges to developing useful risk prediction models to guide prophylaxis: the sheer number of factors potentially associated with increased VTE. The review by Darzi et al identified an additional 4 factors above and beyond those incorporated into existing RAMs. In this and other studies, many factors considered to be predictive by statistical standards contribute minimally to the overall risk, with ORs of less than 2.0 or even 1.5. Furthermore, many of these factors are not necessarily specific, but instead are very possibly markers of sicker patients who are more likely to have increased interventions, prolonged hospitalizations and, ultimately, more complications, which raises the question of how much value they add to prediction models or clinical decision-making.

Ultimately, there is a need for more specific and updated models that incorporate risk not only of VTE but also of bleeding, a call made more challenging by the fact that many risk factors overlap for both outcomes. Darzi et al found only 4 such factors, thus identifying a total of 14 potential specific prognostic markers for VTE and 9 for bleeding, which would be excellent candidates for incorporation into future RAMs.

Although these data are not yet ready for use in the clinic, Darzi et al have provided key data for developing an optimized RAM. Next steps will include careful development and validation of RAMs that will accurately predict actual risk of both VTE and bleeding with and without anticoagulant prophylaxis in medical patients. Although randomized controlled trials of the use of such a RAM in clinical practice will more than likely prove infeasible, given the ethics of withholding preventative therapy for highly morbid complications such as VTE, carefully performed validation studies that are focused on clinically significant outcomes such as symptomatic VTE and clinically relevant nonmajor and major bleeding will add much to our current understanding of these risks. Furthermore, comparison of such models to clinical judgement or a simpler approach (Magic 8-Ball excluded) would be welcome, given the burden of calculating risk score.

Ultimately, an effective RAM, derived from the data provided in this study, has the potential to reduce not only VTE rates but also those of bleeding and, if widely used, in-hospital and post-hospitalization mortality, which are worthy goals indeed. Will the findings of Darzi et al lead to development of such a RAM? I think the Magic 8-Ball must still advise us “Cannot predict now,” but this is certainly a step in the right direction.

Conflict-of-interest disclosure: The author declares no competing financial interests.

REFERENCES

1.
Darzi
AJ
,
Karam
SG
,
Charide
R
, et al
.
Prognostic factors for VTE and bleeding in hospitalized medical patients: a systematic review and meta-analysis
.
Blood
.
2020
;
135
(
20
):
1788
-
1810
.
2.
Barbar
S
,
Noventa
F
,
Rossetto
V
, et al
.
A risk assessment model for the identification of hospitalized medical patients at risk for venous thromboembolism: the Padua Prediction Score
.
J Thromb Haemost
.
2010
;
8
(
11
):
2450
-
2457
.
3.
Spyropoulos
AC
,
Anderson
FA
Jr.
,
FitzGerald
G
, et al;
IMPROVE Investigators
.
Predictive and associative models to identify hospitalized medical patients at risk for VTE
.
Chest
.
2011
;
140
(
3
):
706
-
714
.
4.
Caprini
JA
,
Arcelus
JI
,
Hasty
JH
,
Tamhane
AC
,
Fabrega
F
.
Clinical assessment of venous thromboembolic risk in surgical patients
.
Semin Thromb Hemost
.
1991
;
17
(
suppl 3
):
304
-
312
.