|This chapter provides guidance to bring the VTE prevention protocol effectively to bear at the point of care and to build out the infrastructure for monitoring and measuring this work.|
The Importance of Effective Implementation
After reviewing the evidence and selecting a venous thromboembolism (VTE) risk assessment model, the improvement team will begin gearing up for the all-important implementation phase. If performed skillfully, implementation of the VTE prevention protocol into orders at the point of care will lift adequate prophylaxis rates to 80 percent or more and set the stage for effective measurement, monitoring, and other interventions to eventually reach Level 5 on the Hierarchy of Reliability (Table 1.1).
Skilled implementation can overcome the weakness of a suboptimal VTE prevention protocol; similarly, flawed implementation of an excellent VTE prevention protocol will result in mediocrity and failure to reach the goal of reducing hospital-associated VTE (HA-VTE). Effective implementation of the VTE prevention protocol addresses the first four failure modes discussed in Chapter 2:
- No standardized protocols or order sets for VTE prevention exist.
- Order sets and prompts that reference VTE prevention are in place, but they provide inadequate guidance.
- Order sets with guidance are in place, but the order set is bypassed or not used.
- Order sets with guidance are in place and used, but used incorrectly.
With implementation, the improvement team will want to add more granular detail to the general VTE risk assessment models depicted in Chapter 4. For example, dosing of unfractionated heparin and low-molecular-weight heparin (LMWH) needs to be spelled out, and the mechanisms and responsibility for dosing adjustments for renal failure, obesity, and other conditions have to be defined. In addition, the team will want to engage with different services to determine which ones will need a variation from the general VTE prevention protocol. The potential pitfalls in these steps are numerous, and adding more layers of guidance for special populations can lead to complexity and poor efficiency of ordering.
Well-developed and effective clinical decision support (CDS) involves getting the right information, to the right people, in the right intervention formats, through the right channels, at the right points in workflow.1 A clinical decision template that outlines different desired functionality at each stage may help an implementation team think about building optimal CDS and measurement into different steps in the process of delivering optimal prophylaxis to the patient.
It is helpful at this point to identify the principles for effective implementation of VTE protocols in CDS. These principles bring the protocol guidance effectively to bear at the point of care and build the infrastructure for other interventions and monitoring. This can save effort and time down the road.
Five Principles for Effective Implementation in Clinical Decision Support
Principle 1: Keep It Simple for the End User
Improvement teams must strike a fine balance between providing a good risk assessment for the majority of the inpatient population and keeping the process simple and efficient for the end user. Almost always, simpler is better and less is more. Usability is immensely important, and success or failure may hinge on it.2
It is far more effective to provide less guidance in the time and space where prophylaxis is ordered. For substantial minority populations with special needs (e.g., OB/GYN, spinal surgery, or cardiovascular surgery patients), a dedicated order set tailored to them is likely a better approach than inserting details about these populations into a general medicine or surgery VTE prevention order set.
It is important to involve frontline ordering providers to make sure the VTE protocol is easy to use. Without their input, implementation will not go smoothly.
It is also important to minimize the calculations and data entry end users have to make and to automate the process for them. Even ticking off multiple risk factors for VTE in a point-based model becomes a tiresome task many providers will skip, particularly if it is already evident to them what prophylaxis is needed. For some risk factors or contraindications, it may help to auto-populate data elements from elsewhere in the record. Age, body mass index, creatinine clearance, already prescribed antiplatelet or anticoagulant agents, and platelet counts are a few examples of discrete data elements that could be auto-populated.
There are at times several acceptable options for prophylaxis, and there are often multiple choices for a given LMWH or oral anticoagulant. Improvement teams can simplify the work for the end user and reinforce standardization by streamlining the choices. For example, while the 9th edition of the American College of Chest Physicians on Antithrombotic Therapy and Prevention of Thrombosis (AT9) allows a wide variety of prophylaxis options for major orthopedic surgery, the protocol might only list the preferred institutional choices.
Principle 2: Do Not Interrupt Workflow
In general, an intervention that interrupts workflow will be rejected. This has several implications for design and implementation of VTE prevention order sets.
VTE prevention order sets enjoy the highest utilization when they simply appear as a module that is fully integrated into admission and transfer order sets that are already in use, rather than as a standalone order set clinicians must go out of their way to identify and choose. For example, confusion and workflow interruption can occur if nurses and physicians on the floor are not in sync on how the risk assessment is managed.
The VTE risk assessment and bleeding risk assessment are ideally performed quickly and concurrently when the choices for that combination of risks are presented directly to the provider, without interruption by intervals in time or space. In some computerized physician order entry (CPOE) systems, after a VTE risk level is determined, the appropriate prophylaxis options for the chosen level of VTE risk emerge from their nested position under the risk designation. In other CPOE systems, the risk assessment data entered on the first screen trigger the appearance of a second screen that contains only the choices appropriate for that level of risk in an algorithmic fashion. In these cases, the ordering provider is not asked to remember the risk designation from a previous screen, add up points, and so forth. These tasks are either done for the provider or are eliminated from the process to provide a smooth and uninterrupted workflow.
Principle 3: Design Reliability Into the Process
Part of the improvement team's job is to engineer higher reliability into the process of preventing HA-VTE. To achieve breakthrough improvement, the team must move beyond traditional methods (e.g., personal checklists, working harder next time, and education) to design order sets and reinforcing interventions that use at least one of the following high-reliability strategies:
- The desired action has a forcing function. Completion of a VTE prevention order set can be made mandatory by a forcing function. An electronic or human forcing function ensures that every patient being admitted or transferred in the hospital undergoes a VTE risk assessment.
- The desired action is the default action (i.e., not doing the desired action requires opting out). Only the protocol-preferred choices can be presented to ordering providers for any given combination of VTE and bleeding risk they designate. Choices other than those on the preferred list can be made, but a clinician must first explicitly opt out. For example, a progressive ambulation/mobility protocol can be made the default mode for physical therapy and nursing to pursue unless the physician provides guidance and opts out of that pathway.
- The desired action is prompted by a reminder or a decision aid. A daily reminder to reassess and certify the need for a central venous catheter is an example that can reduce upper extremity deep vein thrombosis (DVT) and line-associated infections.
- The desired action is standardized into a process (i.e., it takes advantage of work habits or patterns of behavior so that deviation feels weird). Standardized order sets with embedded risk assessment are an obvious example. Surveying existing order sets impacting VTE prophylaxis as part of the initial needs assessment, and replacing outdated VTE prevention order sets with new standardized ones, can help to discourage physicians from making up their own personal order sets and bypassing the standardized pathways.
- The desired action is scheduled to occur at known intervals (e.g., integrating DVT prophylaxis assessments into a larger quality and safety checklist to be reviewed daily).
- Responsibilities for a desired action are redundant. If nurses were to focus on patients who were not already on prophylaxis, for example, and to use the same protocol that physicians were using, their redundant check of VTE prophylaxis could be efficient and useful. An electronic alert might provide some of the same functionality.
If designed well, the VTE protocol will be an intervention that invokes several of these high-reliability strategies.
Principle 4: Pilot Interventions on a Small Scale
Piloting on a small scale creates opportunities to iron out glitches before implementing more broadly. Small-scale pilots can be as simple as a 5-minute focus group where five physicians give feedback on several versions of the protocol. Taking an order set out for a "test drive" is far more effective, however, when the pilot risk assessment and order set are applied to patient case scenarios, as ease of use and issues of ambiguity become much more apparent. Piloting measurement and monitoring techniques with early assessment is also highly recommended.
Principle 5: Monitor Use of the Protocol (and Plan for Measurement)
Rolling out the protocol is only the beginning. The improvement team must have a plan that ensures the VTE protocol is part of the completed admission orders for every patient who enters the facility.
A central challenge of standardization is constructing protocols that work for the great majority of patients while allowing for individualization of treatment. It is reasonable to anticipate variations from the protocol, but the team should capture these instances, learn from them, and take steps to reduce them. When providers bypass the protocol, their reasons might derive from logistics and deviations from normal workflow rather than resistance to the concept of standardization. Questions the team can ask include:
- Why is the order set not used in some areas?
- Can it be integrated into other heavily used order sets?
- Which types of admissions are inadvertently bypassing the protocol?
- Which patients do not fit the protocol?
- Can the protocol be changed so it fits more patients and situations?
- Which providers would benefit from focused educational efforts?
- Is the protocol stocked and restocked (if on paper) and in the workflow in all the key areas in the hospital?
The team will also want to plan for measurement. Automating measures is easier if planned into the process at inception. Meeting with the CPOE and/or the information technology team early and often about order set design and how to make measurement an integral part of the process can help. Some examples to consider:
- Storing information as discrete data elements as they can be recalled and organized into meaningful reports more easily than free text.
- Capturing all data element choices in the ordering process, including the declared DVT risk level and any contraindications to anticoagulant prophylaxis.
- Making graduated compression stockings or intermittent pneumatic compression devices (IPCDs) orders into discrete data elements, as well as the documentation for whether they are in place and turned on. A nurse could, for example, be asked to chart each shift whether IPCD was on or off and, if off, a pull-down menu could capture the reason.
- Capturing ambulation status as a discrete data element in monitoring adherence to protocols. Agree on an operational definition of full versus impaired mobility and structure documentation to routinely capture whether the patient is meeting that standard.
Many centers have adapted definitions such as "ambulates independently outside of room twice daily" or "ambulates 50 feet or more independently."
It may help to also think ahead about how to audit patients and determine whether they are on protocol-directed, adequate prophylaxis. In general, complexity of risk assessment in the ordering process will lead to similar complexity in monitoring whether patients are on appropriate prophylaxis. The importance of ease of use applies to both the ordering process and the measurement tools the team will need to deploy.
A properly designed order set, when well positioned and implemented, will prevent errors and get most patients on the correct prophylaxis. Monitoring order set use, and designing an ongoing process to identify patients who have fallen through the cracks, can spur mitigation of lapses in care concurrently. Finally, redesign of the process and order sets should continue to improve the system.3,4
Three Examples of Effective Implementation and Clinical Decision Support
The following are examples of effective order set design and implementation. They illustrate the central importance of implementation and clinical decision support techniques across disparate hospital settings and VTE risk assessment models.
The Johns Hopkins collaborative team used the "translating research into practice" (TRIP) model to implement mandatory VTE risk assessment and risk-appropriate prophylaxis.5 The TRIP model is consistent with the principles presented throughout this guide. Important steps included summarizing the evidence from a centralized steering group; identifying barriers through pilot testing, good measurement, and feedback; and reinforcing appropriate prophylaxis through staff engagement, education, regular evaluation, good clinical decision support in order sets, and layered interventions to reinforce the protocol.6
The Johns Hopkins team created VTE prevention decision algorithms for 16 distinct service groups (medicine, general surgery, trauma, and so forth) and integrated those algorithms into "smart" order sets. Order set implementation in medicine inpatient populations resulted in an increase in risk-appropriate prophylaxis and a large reduction in documented symptomatic VTE detected during or within 90 days after hospital discharge without any change in major bleeding or all-cause mortality.7 Implementation on trauma and surgery services enjoyed similar results.7,8
The risk assessment models this team used had some drawbacks. They were complex to use and not well used in paper form, and the threshold for prophylaxis in some models was much lower than the threshold supported by the AT9 guidelines in many cases (e.g., medical patients qualify simply by having a "major" risk factor of age >60 years).
The Johns Hopkins model is presented here for its effective approach to implementation, rather than to highlight the risk assessment model itself. The keys to successful implementation included advanced CDS integrated into CPOE versions of the tool.
Johns Hopkins Internal Medicine used several of the key principles of CDS. A forcing function made risk assessment mandatory, and the order sets were embedded in medicine admission orders. The tool was made easier to use by displaying relevant clinical data for risk assessment, automatically pulling in some data elements from the EHR, and by displaying default choices for prophylaxis corresponding to the VTE and bleeding risk factors chosen by the provider. Completion of the VTE risk assessment, risk level, and alignment with protocol guidance was explicitly captured to raise situational awareness at the point of care and for monitoring and feedback. Careful order set design therefore reached Level 3 on the Hierarchy of Reliability and set the stage for further progress and interventions.
The University of Michigan overcame the inherent complexities of the Caprini VTE risk assessment model with skillfully deployed CDS in CPOE. The improvement team adjusted its approach out of necessity and added more of the key principles of CDS outlined in this chapter after earlier attempts failed to achieve the desired results. Key strategies for success, arrived at over time, included targeting all adult inpatients, adding forcing functions with hard stops to guarantee a risk assessment was done, using algorithmic logic, grouping risk factors for the convenience of providers, and auto-populating some risk factors.
Importantly, the addition of risk score points is performed behind the scenes, with options appropriate for the point total displayed as the default prophylaxis choice. Note also the capture of VTE risk level, the dosing guidance for renal insufficiency, and the mandatory documentation of anticoagulation contraindications in those who defer risk-appropriate anticoagulant prophylaxis. In addition, a full suite of educational and faculty engagement techniques were used. The end result of this implementation effort was significant reductions in surgical and medical VTE rates.
Banner Good Samaritan Regional Medical Center participated in a Society of Hospital Medicine-sponsored mentored implementation collaborative and enjoyed a 59 percent reduction in total HA-VTE events, a 65 percent reduction in pulmonary embolism, and a 57 percent reduction in DVT. The comprehensive implementation effort included deployment of the CDS principles in reinforcing the medical center's VTE prevention protocol.
Certain key elements, such as weight and creatinine clearance, were pulled into the order set and made available to the ordering provider at the point of care. Mandatory selection of high, moderate, or low risk was mandated on admission and transfer. Risk-appropriate prophylaxis options were presented on declaration of VTE risk level, with dosing guidance for different situations and indications. Opting out of anticoagulant prophylaxis for moderate- or high-risk patients led to capture of anticoagulant contraindications and default choices for mechanical prophylaxis. Standardized timing of perioperative prophylaxis doses were offered as a default by designating the patient's status as surgical pre-op or surgical post-op.
The VTE risk level, orders, contraindications, and other data elements were captured and presented on the medication administration record and in transfers, and were also used to assist in monitoring prophylaxis patterns. Banner Good Samaritan used measure-vention, multiple methods to engage nurses and physicians, and audits to monitor and improve adherence to IPCDs.
Select for the Banner model slide presentation (PDF, 764 KB).