Electronic health record-enabled routine assessment of medication adherence after solid organ transplantation: the time is now

Abstract

Medication nonadherence after solid organ transplantation is recognized as an important impediment to long-term graft survival. Yet, assessment of adherence is often not part of routine care. In this Personal Viewpoint, we call for the transplant community to consider implementing a systematic process to screen and assess medication adherence. We believe acceptable tools are available to support integrating adherence assessments into the electronic health record. Creating a standard assessment can be done efficiently and cost-effectively if we come together as a community. More importantly, such monitoring can improve outcomes and strengthen provider-patient relationships. We further discuss the practical challenges and potential rebuttals to our position.

1. Current state of adherence assessment

Nonadherence to immunosuppressive medication can have significant negative consequences for transplant recipients. Nonadherence to immunosuppression is common after transplantation, with estimates from 30% to 70%.¹ Nonadherence is recognized as a major modifiable risk factor for reduced long-term graft survival across organ types, age groups, and countries.^1,2 While the average transplant outcomes are excellent, graft loss due to nonadherence is unacceptable, given the increased mortality of those who cannot be retransplanted and those who die on the waitlist. We must strive to improve our management of patients until we achieve the goal of one transplant for life. Herein, we offer our perspective on why the time is now to take a substantial step toward improving transplant recipients’ medication adherence by implementing routine screening and assessment.

Surveys of transplant providers reveal that less than 30% of US transplant centers proactively screen for nonadherence after transplant.³ This is an improvement from approximately 6% 5 years earlier but remains woefully inadequate.⁴ Even more worrisome is that only 20% of centers routinely conduct evaluations after 6 months posttransplant when nonadherence is most likely to develop.⁵ Yet, the same respondents endorse that immunosuppression nonadherence is (1) a major problem at their center, (2) a primary contributor to graft loss, and (3) inadequately addressed compared to other posttransplant complications. The primary barriers to adherence assessment were the lack of an “unobtrusive, practical, and objective measure” and limited time with a desire for a low-cost, evidence-based solution that would not interfere with the workflow.³

We contend these views reflect an unrealistic expectation for adherence assessment tools. We agree that the ideal measure of nonadherence would be objective and direct while posing little burden to the patient and provider. However, it is unlikely that a perfect tool will become available. Repeated calls by experts over the last 2 decades, including international guidelines, have yet to spur a concerted effort to incorporate medication adherence assessment into the standard of care.^2,6,7 The reports recommend combining assessment tools into a composite standard because individual measures have low sensitivity and specificity, regardless of the degree of objectivity.^8,9

We must recognize that assessing adherence will require an upfront investment of time and resources. It is unlikely that any adherence assessment process will be cost-free or lack an impact on the current workflow. However, continuing to allow this to be a barrier is short-sighted and fails to recognize the impact of graft failure on quality of life and the cost to the health care system, payers, and society. Over $100 billion in avoidable US healthcare costs were attributed to medication nonadherence in 2012.¹⁰ Multiple economic evaluations demonstrate that reducing nonadherence can be cost-effective.^11,12 Enacting a combination of currently available tools and recognizing their limitations is likely better than the status quo.

2. Desired future state

We imagine health care professionals and patients want a program striving to achieve the highest level of care in which a dedicated team focuses on preventing, identifying, and managing medication nonadherence. Transplant centers should aim to create a culture in which adherence assessments are routine and normalized. Patients and caregivers should be comfortable voicing concerns without facing judgment and feel confident that they will receive support to overcome adherence barriers. Such an environment will facilitate patient trust, which subsequently has the potential to increase adherence.¹³

A standard, systematized process to identify the risk of immunosuppression nonadherence is a potential first step to achieving this optimal outcome. The specific details of how centers use this information to act can be tailored to individual circumstances and resources. A feasible initial pathway is to leverage the electronic health record (EHR). We envision an evidence-based module or dashboard with multiple variables measured longitudinally for adherence assessment. We propose incorporating (1) responses from a brief electronic self-report questionnaire provided by the patient during the check-in process or texted to the patient during appointment reminders, (2) rates of lab and clinic visit adherence (or appointment no-show rates), (3) estimated proportion of days covered for immunosuppressive drugs at stable doses, and (4) immunosuppression trough concentration variability or time in the therapeutic range. We have selected these measures based on the available supporting literature and the potential feasibility—much of this data can or is already calculated by commonly utilized EHR systems. For example, the calculation of proportion of days covered and appointment no-show rates is common, but the information is scattered throughout the chart, and not all providers may know where to find it.¹⁴ Even more complex items like tacrolimus variability have evidence of possibility, as time in the therapeutic range is available for warfarin.¹⁵ A medication adherence dashboard can combine these various tools to allow adherence assessments to occur in an effective and efficient manner.

A request by the authors to transplant pharmacist listservs for examples of EHR adherence tools in March 2021 yielded no additions beyond what the authors have created at their centers. These tools represent an ad hoc, imperfect, nontransferable collection of flowsheets, note templates, and reports, each digitizing a single piece of medication adherence information without integration. Even if an adherence screening or assessment occurs, it is most likely based on data fragmented across various steps, systems, and processes. In our experience, the likely culprit is that center-level development is difficult to have approved and created by busy hospital information technology teams with limited ability to alter the base program. The waste of resources by having each transplant center develop tools independently is too big an opportunity to ignore. The items we have selected are based on our assessment of the available evidence and are briefly described in the following sections. We encourage those considering our position to explore the cited literature to understand how the data may be best used at their center.

2.1. Self-report

Self-report is often cited as the most feasible medication adherence assessment tool.⁶ Despite concerns for recall and social acceptability bias, self-report often captures the highest nonadherence rates.¹ The questionnaire can be as brief as 2 questions about missed and late doses in the last month or more complex, expanding to cover latent predictors.¹⁶ The Basel Assessment of Adherence to Immunosuppressive Medications Scale (BAASIS) has been developed specifically for transplant recipients and follows an evidence-based taxonomy for defining adherence. A recent meta-analysis of the BAASIS provides strengthened evidence supporting its psychometric properties, suggesting it may be the preferred tool to implement.¹⁷ The meta-analysis provides evidence for the validity and reliability of the questionnaire. An additional benefit of the BAASIS is the availability of 13 language translations, with more under development. Permission for use can be obtained from the University of Basel (https://baasis.nursing.unibas.ch/how-to-obtain/). A limitation of the BAASIS questionnaire is that it does not evaluate variables that may contribute to nonadherence, such as cost, adverse effects, or medication beliefs.^1,6 Inquiring about these items may further identify barriers to medications that can be targeted with adherence interventions.¹⁸

2.2. Laboratory and clinic adherence

Rates of nonadherence to appointments have been associated with poor outcomes, including rejection and graft loss.^19,20 Since these discrete fields are available in many EHR systems, they should be an easy data point to incorporate. Monitoring laboratory and clinical adherence also has the added benefit of avoiding the streetlight effect, which may result from only assessing patients who present to the clinic (a sample of convenience rather than sampling the true at-risk population).²¹ A notation of the frequency and cause of hospitalization may provide further benefits by highlighting readmissions coded as nonadherence-related.

2.3. Medication refill records

Several adherence measures can be calculated using medication refill data, including the proportion of days covered and medication possession ratio. Automated calculations are becoming increasingly common in EHR. Easy, direct access to this information for the entire transplant care team (nurse co-ordinators, physicians, social workers, pharmacists, etc.) can bypass the time-consuming step of reaching out to the patient’s pharmacy, a historically reported barrier to the routine use of refill records.⁶ A limitation is that the accuracy of these automated results has not been thoroughly assessed. The most likely problem is that not all prescription records are reported to private third-party systems (eg, SureScripts) used by EHR vendors due to variations in insurance, cash pay, or patient assistance programs. However, a brief conversation with the patient could uncover if such a situation is present. To our knowledge, the automated calculation does not account for hospitalizations, and if transplant programs are not diligent in updating prescriptions, dose changes will not be accurately reflected. Nonetheless, even the automated calculation could be made more relevant by including the notation of hospitalizations and hold periods in the dashboard. Regardless, identifying gaps in medication possession can serve as a conversation starter and an opportunity to establish regular dialog about adherence.

2.4. Tacrolimus trough level variability

Tacrolimus trough level variability may be the most controversial approach discussed, but this method could include intrapatient variability (standard deviation, coefficient of variation, or time in therapeutic range) or undetectable levels. The sensitivity and specificity of the measure may be a concern but is lessened by the multimodal approach of such a dashboard.^9,22 While evidence regarding associations of high trough-level variation with nonadherence may be conflicting, the relationship with poor graft outcomes is well-established.²³ As such, exploring the cause of variability may be beneficial regardless of whether nonadherence is identified as the source. The relationship between nonadherence and levels of other drugs that undergo therapeutic drug monitoring, including cyclosporine, everolimus, and sirolimus, is even less well established. However, in theory, high variability or undetectable levels may signify a need for evaluation.

An EHR dashboard with this combination of strategies would fulfill the criteria frequently identified as necessary for acceptance—easily integrated into the workflow, inexpensive, and multimodal. Studies evaluating such a combined strategy have demonstrated the highest sensitivity in detecting medication nonadherence.⁸ The relative ease with which this information could be viewed (once the dashboard is created) would facilitate longitudinal assessment. Single centers that have taken a quality improvement (QI) approach to implement adherence promotion in all patients rather than a research approach have demonstrated a 50% reduction in the incidence of acute rejection.^18,24 Recent studies in other chronic illnesses suggest this QI approach is feasible and acceptable to health care providers.²⁵ Similar to a QI mindset, implementing a given adherence assessment does not stop us from conducting research and refining the process to address the limitations. In fact, the answers generated may be more robust because the dashboard can easily use discrete fields to facilitate gathering large, real-world datasets with adequate power and follow-up. An additional advantage of such a strategy would be including all recipients to avoid the selection bias of including only those who agree to participate in research (a generally more adherent population). Such action is consistent with developing a learning health system, and the ethics of using the EHR for such research have previously been described.²⁶

A single, unified source of medication adherence-related data increases its power and the value to be gained. Therefore, we propose partnerships or formal collaborations with EHR vendors that may be particularly fruitful and result in a more effective, comprehensive adherence assessment technique that is significantly less expensive. The vendors have already implemented many of these items but would benefit from a conversation with frontline clinicians to create a usable, centralized dashboard interface. We hope the market-leading companies can recognize the value such collaborations may have for patient care and satisfying the needs of consumers. Although improved data and technology do not necessarily equate to improved results, they are enablers. New technologies can greatly influence policy and practices, most notably through process standardization.

As with all proposals, there are disadvantages to this approach. The primary disadvantage is the unclear threshold for defining clinically meaningful immunosuppression non-adherence. None of the available assessment tools have ideal sensitivity and specificity, creating the risk of false negatives and false positives. However, a false negative assessment that does not identify nonadherence when present is no different than the current practice that misses nonadherence until it becomes clinically significant in the form of rejection or graft loss. The risk of a false positive may be more concerning because of the potential risk to the provider-patient relationship if inaccurate accusations of nonadherence are made. However, if the dashboard item(s) suggesting nonadherence are explored in a nonjudgmental fashion encompassing all the possible contributing factors, this risk is minimal. There is also the argument that assessment may increase immunosuppression adherence in patients already “adherent enough” with the negative consequences of toxicity or diversion of resources from those more in need.²⁷ Finally, multiple measures complicate the adherence assessment without empiric support as they are not consistently well correlated. We argue that inadequate correlation is not necessarily a negative and might be a sign the measures are identifying different patterns of nonadherence. In transplantation, we make decisions based on incomplete and evolving evidence with potentially bigger implications than adherence screening. No evidence is all-encompassing; transplant providers must routinely integrate information with varying degrees of uncertainty to exercise clinical judgment.

Recognizing the limitations of the tools, each center can make its own decisions on interpreting the available data, incorporating all measures, or prioritizing those that the center feels best represent medication adherence in their population. For example, if refill records are inaccurate due to the high use of a local pharmacy that does not report to the third-party vendor, the focus could be on the alternatives. The objective is to take the information presented by the dashboard coupled with clinical judgment to identify patients who may benefit from further exploration of medication adherence or barriers to adherence. We suggest the data serve as a starting point to tailor patient conversations and explore alternative explanations. Improved communication with patients around medication adherence and self-management can be a powerful tool for maintaining adherence. Further, evidence shows that adherence may improve when providers are simply aware of adherence status, even without a protocolized response.²⁸ The limitations of the tools can be acknowledged and used to inform our interpretation, not only to serve as an excuse to prevent action.

Finally, there is also value in sharing adherence information with the patient. Allowing patients to view and track their adherence information in the electronic health portal aligns with the 21^st Century Cures Act. As currently available, access to greater data via the portal, such as clinician notes, increases patients’ perceptions of their involvement in care, understanding of medications, and empowers them to be adherent.²⁹ Evidence also suggests access to such information may be particularly useful for those at the highest risk of nonadherence. Further, patients may identify errors in calculations or documentation, allowing for more accurate adherence assessments.³⁰ Capturing data in discrete fields increases how this information can be shared. For example, it could be added to the after-visit summary. Imagine the possibilities if steps were undertaken to further enhance the information in the portal with the thoughtful intention for it to be used as an adherence tool to maximize benefit and minimize harm.

3. Conclusion

We believe that medication adherence assessment tools do not need to be deemed “perfect” to be transformative. We should not wait for large-scale clinical trials that are unlikely to materialize or are too complicated to retain fidelity in the real world. The challenges and constraints of the adherence strategies have been highlighted above. However, we should resist the temptation to assign black-and-white adherence thresholds and instead recognize these approaches as tools for further patient-specific exploration and intervention. We thus believe that a standardized EHR infrastructure can provide a feasible solution for medication adherence screening, assessment, and monitoring at the current moment.

Identifying medication nonadherence can impact the recipient, waitlisted transplant candidates, and public health on a large scale. Missed opportunities to evaluate nonadherence should not be taken lightly. A good outcome in this context does not need to be identifying all nonadherence episodes, merely more than we capture now. It stands to reason that simply having more centers capable of conducting any level of prospective preventative adherence screening can achieve this goal. Once we have systematized the assessment of adherence, the focus can turn to optimizing strategies to manage nonadherence when it is identified. We hope this viewpoint will initiate a coordinated national and international discussion among community stakeholders to address the remaining questions systematically. We present one potential idea for improving the status quo: highlighting what we can do now to generate workable solutions that will accelerate the implementation of a routine adherence assessment in transplant recipients, transplant candidates, and beyond.

Abbreviations:

BAASIS

Basel Assessment of Adherence to Immunosuppressive Medications Scale

EHR

electronic health record

QI

quality improvement