Implications for patients, clinicians and policy-makers
The updated systematic review within workstream 1 highlights the lack of supporting evidence for RAMs in this area and clarifies the limitations of previous work. In particular, our systematic review provides contemporary evidence that no method of risk assessment has been extensively validated, or methodically studied by implementation research.17 This work supports recommendations within recent current UK NICE guidance, which highlight the lack of evidence for an ‘optimal RAM’ and support use of any RAM published by a national UK body, professional network or peer reviewed-journal to aid risk assessment in medical and surgical inpatients.8 However, our cohort study within workstream 2 draws attention to the potential impact on practice of such a broad recommendation. Applying the RAMs identified by systematic review to a cohort of medical and surgical NHS patients within a single centre resulted in overall recommendations for pharmacological prophylaxis ranging from 91% (95% CI 88.6% to 95.2%) to as low as 13% (95% CI 9.4% to 17.7%). Such potential variation has clinical, educational and medico-legal implications for clinicians and policy-makers. This variation would also clearly have a significant impact on patients. Multiple thrombosis societies and patient-facing organisations have worked hard to raise awareness on VTE risks over the last decade and empower patients to self-advocate for risk assessment and appropriately challenge medical decision-making.169 If the strategy of risk assessment is variable and the supporting evidence limited, participation in shared decision-making may be difficult.
Furthermore, decision-analytic modelling suggests that routine VTE prophylaxis is likely to be more cost-effective for medical and surgical patients requiring hospital admission than a strategy based on any currently available RAM. When taken in context with the lack of supporting evidence for available RAMs and ongoing errors/omissions leading to preventable cases of hospital-acquired VTE,150 these findings have significant implications for policy-makers. There is reasonable evidence for the clinical effectiveness of pharmacological prophylaxis (subject to contraindications) for prevention of VTE in all medical and surgical patients admitted to hospital.4,13 If pharmacological prophylaxis for all is likely to also be the most cost-effective strategy, despite drug costs and potential harms, recommendation of unvalidated RAMs may be inappropriate. Based on our findings, it may be preferable for policy-makers to consider evaluating a new paradigm of ‘opt-out’ VTE prevention, in which all eligible patients are offered pharmacological prophylaxis on hospital admission without complex risk assessment. In this circumstance, patients who are ineligible for pharmacological prophylaxis due to contraindications, high bleeding risk or personal choice could still utilise mechanical thromboprophylaxis in accordance with current evidence.
In addition, during the course of this project we identified several pragmatic issues with RAM use and application which can lead to repeated system errors or disengagement. Several RAMs have subjective components with poor inter-rater reliability;170,171 RAMs may not be completed accurately (or at all) by busy clinical teams if they are deemed unhelpful to the encounter; clinicians may complete RAMs in a biased fashion due to experience (or inexperience) based on seniority; RAM completion and identification of high VTE risk may not translate to reliable prescribing; multifactorial RAMs applied to every patient accrue a significant work burden; and some multifactorial and complex RAMs will actually accrue decision-making points from only a small number of key variables. Although these issues do not feature widely in the current literature due to lack of implementation research, several studies have highlighted the numerous pragmatic challenges of RAM use.51,145,153 These cumulative issues also potentially support a paradigm shift towards an ‘opt-out’ strategy.
There are limitations in translating these research findings to practice. Our systematic review focused on RAMs which evaluate thrombosis risk (rather than bleeding risk). As such, our economic modelling work is also based on estimations in a population without an increased risk of bleeding. In addition, workstream 1 focused on population-level estimates. In reality, each individual patient encounter is likely to have a different risk profile regarding VTE and bleeding; practicalities of implementation may be different from the model and it may still be necessary to perform a degree of risk assessment, or to evaluate the contraindications for pharmacological prophylaxis in detail.
Finally, we highlight the limitations of using routine data methods in the context of future research. Such systems run the risk of identifying both false-negative and false-positive VTE cases, through failure to identify ambulatory (uncoded) cases and overdiagnosis of previous disease. Many of these issues reflect the changing approach to modern VTE management and would be challenging to mitigate within any future research studies. Our findings were similar with regard to bleeding events, showing poor sensitivity of ICD-10 coding data and multiple false-positive events identified across four NHS sites. These findings have implications for funders looking to support further work in this area and suggest large studies reliant on routine data collection methods for outcomes are likely to be inaccurate and therefore unfeasible.
Our PPI group highlighted at several points during the project that risk assessment in the UK setting still seems an uncertain process, with many clinicians underestimating the consequences of VTE and the importance of prophylaxis. In addition, several respondents in our survey of national leads (see Chapter 7) highlighted the ‘tick box’ nature of VTE risk assessment; many clinicians feel they have already made the decision to prescribe prophylaxis prior to any risk assessment, and therefore consider the assessment itself to be a futile exercise. These findings also potentially support a more implicit use of prophylaxis at the point of hospital admission, with a clinical focus on reliable and standardised implementation, summarised by the ‘opt out’ approach described above, in contrast to current practice (complex risk assessment followed by opting in). However, the PPI group was also clear that lack of information and discussion about VTE risk during hospital admission or in the periprocedural period, was an ongoing concern. It was further highlighted by the steering committee that risk assessment, for all its faults, offers a clear opportunity to educate junior clinicians and discuss the risk of VTE with patients and service users, raising awareness of the condition and improving dialogue on the topic. These softer benefits may be potentially lost through any process of routine and unqualified thromboprophylaxis prescribing.
Suggested research priorities
Similar to our previous work, we identified that the current evidence base for RAMs looking to select appropriate hospitalised patients for pharmacological thromboprophylaxis is very limited and estimates of prognostic accuracy are subject to substantial uncertainty.74 However, any future research on RAMs would be subject to the same challenges faced by this project, of VTE risk assessment as a contract standard within NHS practice and subsequent high rates of prophylaxis, supported by clear NICE guidance.8 It would therefore be unethical to expose patients to implementation research on RAMs incorporating any control arm without risk assessment, or thromboprophylaxis.
Alternatively, given our findings on cost-effectiveness it may be reasonable to consider implementation research focusing on assessment of a novel opt-out strategy, regarding pharmacological prophylaxis. Such work could be conducted in any NHS setting and allocate patients to an intervention arm (default prophylaxis subject to contraindications or high bleeding risk only) compared to standard care (formal risk assessment and tailored prophylaxis as per national guidance). Patients ineligible for pharmacological prophylaxis in either group could also be offered alternative mechanical prophylaxis using evidence-based strategies, to further evaluate impact. Our survey work would suggest such an approach to be of interest to the national VTE community, with more support for a cluster randomised design (rather than individual patient randomisation) at hospital level. Such work would need to consider the challenges of reliable follow-up and outcome ascertainment given our findings from workstream 2. Proposals could include adjudicated co-primary outcome definitions for VTE and major bleeding, evaluated using a hybrid approach of digital patient follow-up, routine data sources and local VTE identification methods.
Within the context of such an implementation study (including high numbers of patients receiving pharmacological thromboprophylaxis), a large observational cohort could also be evaluated to determine which risk predictors are associated with potential thromboprophylaxis failure. Given the challenges identified by this project regarding the completion methods and accuracy of risk assessment, risk predictor variables would likely require reliable, real-time collection by research teams on admission, outside of clinical assessment. Serial evaluation over the first 48 hours would allow evaluation of temporal RAM trends and novel biomarkers (D-dimer, CRP) on risk profile, building on recent systematic review work.172 Indeed, recently derived RAMs are heavily focused on objective biomarker variables at an early admission stage and echo concerns regarding inter-rater reliability with subjective RAM components.173 Multivariate analysis could subsequently identify which of these features during the first 48 hours of admission best identify likely thromboprophylaxis failure, allowing further risk stratification during hospital admission. If validated, this approach could lead to future trials of enhanced thromboprophylaxis strategies for higher-risk populations, in keeping with the ‘individualised approach’ relevant to specific subgroups, already proposed by several topic experts.174–176
The same logic could be applied to patients at the lower end of the risk spectrum. Within the context of an implementation study, patients who do not receive pharmacological thromboprophylaxis during routine care could be followed in a similar fashion to evaluate risk characteristics, temporal trends and outcomes. These data could facilitate multivariate analysis and subsequent identification of patients at a particularly low risk of VTE, who could be considered for early cessation of prophylaxis.
Taken together, the above projects could evaluate a paradigm shift in clinical VTE risk management; evaluation of a novel opt-out approach on hospital admission with the majority of medical and surgical patients immediately commenced on a weight-adjusted standard dose of pharmacological thromboprophylaxis without detailed risk assessment; a further evaluation at 48 hours using enhanced risk prediction, temporal trends in clinical progress and biomarker data; and a subsequent ‘rationalisation’ of prophylaxis at this time point – to increase, to continue or to cease. Such an approach would obviate complex risk assessment at the front door, when information, time and accuracy of diagnosis are limited and the chances of error are high, and facilitate early reliable use of prophylaxis followed by considered transition.
We found limited evidence on the accuracy or validity of bleeding risk assessment tools for hospitalised medical and surgical patients. Given we found bleeding risk to be a key factor in determining cost-effectiveness, further studies of bleeding risk and outcomes in patients receiving pharmacological thromboprophylaxis are urgently required. The high number of patients receiving prophylaxis in routine care could facilitate the rapid development of large observational cohort studies in this area, designed to accurately capture risk variables at presentation and evaluate their association with standardised major and CRNMB risk during hospital admission. Our findings in workstream 2 suggest that such a study could not rely on routine data collection methods to determine outcomes of relevance and may require significant research support to ensure diligent and accurate follow-up.
Last, our work is restricted primarily to medical, surgical and orthopaedic patients. We did not evaluate bespoke RAMs in specific patient subgroups, such as cancer or neurosurgery. Further studies of prophylaxis in specific patient populations are warranted, in particular where risk profile may differ significantly due to higher risk of critical site bleeding. A recent commissioned call for research into prophylaxis for patients with traumatic brain injury exemplifies this point.177 Future studies in bespoke populations are likely best delivered through RCTs, informed by our work on the accuracy of routine data regarding relevant outcomes in thrombosis research.