Electronic Health Records and Heart Failure

Introduction

The use of electronic health records (EHR) is rapidly becoming pervasive in the United States, in part due to requirements established by the Health Information for Economic and Clinical Health (HITECH) Act. Prior to HITECH, fewer than 10% of physicians used EHRs,¹ whereas following HITECH, nearly 90% of hospitals and clinics have adopted at least a basic EHR.² The EHR market globally is now > $30 billion, includes hundreds of countries of all economic status, and ranges from small private practices to national EHR systems. Despite great enthusiasm for the potential of EHRs to improve clinical outcomes, EHR adoption overall has not been associated with improved outcomes and only minimal improvements in care quality in patients with heart failure (HF).^3–5 However, there is evidence that specific EHR-based tools can improve adherence to evidence-based guidelines of HF care, patient self-management, and clinical outcomes.⁶ Herein we will review EHR tools useful in managing HF, how they can be used to address specific clinical applications, and evidence supporting their efficacy in improving outcomes (Figure 1). Cited literature is primarily specific to care of patients with HF given the distinct needs of this population. Where no HF-specific data exists, studies involving other diseases or non-selected populations is referenced.

Figure 1 –

Electronic health record components and potential benefits to patients with heart failure and their clinicians

Background

EHR components

EHRs are enormously complicated, some costing hundreds of millions of US dollars to implement. Each EHR has specific strengths and weaknesses, which are often based on early priorities and design decisions. EHRs are often described as ‘digitized medical charts,’ thereby focusing on clinical documentation. However many contemporary EHRs are made up of 3 fundamental components: computerized physician order entry (CPOE), clinical documentation and data access, and billing/financial tracking functionality. (See Figure 2) EHRs increasingly have 2 additional important features: clinical decision support systems (CDSS) and data warehouse capacity. The value of electronic data access in routine clinical care is intuitive and widely apparent, and the addition of data warehouses and health information exchanges is increasing this value further. Billing and financial auditing are crucial capabilities to US institutions in particular and have been the foundation of several commercial EHRs, sometimes at the expense of utility and efficiency in clinical care. This discussion will focus on the EHR components that are commonly encountered at the point-of-care and are most useful in routine clinical care of patients with HF. Knowledge regarding the features discussed should be immediately useful to any HF clinician using EHRs.

Figure 2 –

Electronic health record schematic with major components important for management of heart failure

Commercial vs. Free/open-source software

A detailed comparison of advantages and disadvantages of commercial vs. free/open-source software (FOSS) EHRs is beyond the scope of this manuscript. The use of FOSS vs. commercial EHRs varies significantly worldwide based on regional constraints and requirements. In general, FOSS EHRs offer dramatically lower implementation costs and greater institution-specific customizability compared with superior development capacity, scalability, technical support including software maintenance, and security of commercial EHRs. Many of the features below, with the possible exception of machine learning (ML) analytics, are available or implementable in most EHRs. This review will not compare components of specific EHR vendors or platforms and will instead discuss features in an EHR-agnostic fashion. Clinicians are encouraged to investigate tools available at their respective institutions.

Tools

Dashboards

EHR dashboards summarize patient characteristics, interventions, and other features in order to monitor and manage groups of patients. (See Figure 3) Patient cohorts may represent a shared disease condition (e.g. HF, diabetes mellitus) or administrative designation (e.g. inpatient service, individual clinician panel). A wide range of clinicians (e.g. physician, nursing staff, case manager, pharmacist) may access dashboards to monitor clinical status, treatment adherence, healthcare encounters, test results, and other features, thereby improving efficiency in managing large numbers of patients. For example, HF dashboards can track use of guideline-directed medical therapy (GDMT) as well as whether the patient is on a target dose of each agent.⁷ Many dashboard tools allow a clinician to select a patient of interest to get more detail, open their individual chart, and make management changes as needed. The ability of dashboards to summarize and display diverse data from multiple patients efficiently has led them to become important components of population health management, particularly of chronic diseases.^7,8

Figure 3:

Patient cohort dashboard example Courtesy of Codex Health, Inc., Palo Alto, CA

Historically condition-specific dashboards were populated by hand (i.e. clinicians select individual patients one by one). More recently, methods for identifying disease cases such as HF are being used to automatically screen patients and even populate dashboards.⁹ As the diversity of digital patient data increases, dashboards are increasingly used to aggregate and present data from disparate sources such as remote device measurements, quality of life (QOL) inventories, pharmacy utilization, and test results.¹⁰ Given the complexity and interconnectedness of these new data streams, dashboards will undoubtedly be a critical part of incorporating them into clinician workflows efficiently.

Predictive Analytics

There has been a substantial amount of work using EHRs to develop and refine analytic models to guide HF management.^11–13 Methods used to create clinical models from EHR data include regression models, time-to-event or survival models, classification/regression trees, random forest analysis, neural networks, and others, although consistency and completeness of data is a major challenge. Automated HF identification and risk stratification have been the most widely-studied applications although there have some efforts to predict treatment response.¹⁴ Perhaps the most significant limitation in the use of EHR-based predictive analytics has been lack of EHR implementation and integration with clinical workflow. Most published EHR-based predictive models were trained on retrospective data extracted from the EHR or accompanying data warehouse, but they are not easily translated to real-time use for a number of reasons. These factors include complexity of methods, inability of EHRs to execute those methods (e.g. ML), uncertain effects on EHR stability, security considerations if using an external server, and operational resources required for implementation. Most importantly, change management in clinician acceptance, behavior and workflows is critical to translate even high-quality predictions to clinical action

Standardization frameworks

Applications such as order sets and clinical pathways are used increasingly to promote consistent care across provider groups. When constructed well, order sets and pathways make it easier and faster for clinicians to adhere to recommended management than not to (i.e. pre-specified orders rather than writing orders from scratch). Order sets and pathways were used long before CPOE was available and today are implemented in largely the same format in EHRs. Added benefits of using EHR-based order sets include the ability to update content and immediately disseminate them in a consistent manner across large healthcare systems, the potential for modular order sets (e.g. components of diabetes mellitus, HF and chronic obstructive pulmonary disease hospital admission order sets for a patient with multiple comorbidities) and automated presentation of order sets based on algorithmic ascertainment of disease conditions (e.g. suggest a using HF order set for a patient hospitalized with a BNP > 150). Order sets and pathways are frequently employed to optimize performance metrics as well. Importantly, consideration of human factors elements in design of HF order sets appears to improve utilization, and clinician participation is strongly encouraged.¹⁵

Remote patient monitoring

Remote patient monitoring (RPM) using devices such as CardioMEMs, implanted cardiac defibrillators (ICDs), pacemakers, and wearable devices are increasingly used to monitor patients’ clinical status outside of healthcare encounters. Internet of Things systems comprised of Wi-Fi and Bluetooth-enabled devices such as scales, blood pressure cuffs, activity monitors and mobile apps for symptom inventories are also being developed.^16,17 Although some technologies have shown promise, a significant limitation of most RPM strategies for HF is the general lack of robust EHR integration. This may be due to a number of reasons including proprietary vendor technology, complexity of developing an EHR interface, and lack of clarity on how and which data to store in the EHR. Consequently, clinicians must often log in to web-based dashboards outside of their EHR workflow. Generalizable strategies for generating meaningful signals from remote device data and determining the locus of that processing (i.e. internal vs. external to the EHR) remain nascent. Workflows to make timely and beneficial use of remotely captured data within healthcare systems via the EHR are not well-developed. Nevertheless, it is highly likely that integration of remotely captured data into the EHR will be critical for realizing their full potential.

Clinical decision support

EHR-based clinical decision support systems (CDSS) have become essential tools for supporting clinicians during routine care¹⁸ and have shown promise in improving HF outcomes.¹⁹ In reality, all EHR tools described above are forms of clinical decision support. For the purposes of this review, the term CDSS will be used primarily to discuss alert-based functionality. The most commonly encountered CDSS alerts are those identifying drug-drug and drug-allergy interactions, but CDSS are becoming increasingly complex in combination with innovative dashboards, advanced predictive analytics, and care standardization tools. The most general framework of CDSS alerts consists of a logical statement, or ‘trigger’, followed by an alert, either interruptive or passive, to inform the clinician of a recommendation. A simple example is “If the patient’s LVEF is < 40% and they are not on a beta-blocker (BB), then alert the clinician the patient should be on a BB.” (See Figure 4) Interruptive or active CDS alerts are more effective than passive alerts in modifying clinician behavior including in HF^20,21 but have the greater potential to cause alert fatigue. Careful consideration of details such as clinical impact, alert frequency, recommended action, and workflow disruption is critical to designing effective CDSS.^22,23 As with all tools discussed here, clinician input improves utilization of CDSS.^15,24

Figure 4:

Generic CDS example – evidence-based beta-blocker adherence.

Patient portals

EHR patient portals represent an important communication tool between patients and clinicians, and they provide an avenue for engaging patients with educational materials, self-advocacy and management tools, and decision aids. They can also be used to collecting RPM data such as symptom inventories and vital signs.^25,26 One important positive consequence of the COVID-19 pandemic has been the rapid maturation and increased utilization of telehealth and patient portals, including in elderly patients who previously were less likely to use eHealth tools.²⁷ Data regarding impact of patient portals on clinical outcomes in a general population is limited, and a recent Cochrane Review concluded that quality of available results was too low to discern any clinical benefits taken together.²⁸ However, there is evidence that patient portals may be particularly useful in patients with multiple chronic conditions.^29,30 Availability of a patient’s health record via a patient portal can be valuable tool in continuity of care compared with traditional requests for their records.

Thus far, sophisticated, multimodal mobile health (mHealth) applications are not widely integrated with EHRs, largely for technical reasons in variability of interface and EHR specifications between institutions. Efforts are underway to integrate patient-reported outcomes instruments with EHRs,³¹ although best practices for utilizing these data have not yet been defined. It is likely that in the future, features of the mHealth applications if not the applications themselves will be integrated with EHR patient portals, but the challenge of EHR interoperability will remain a significant barrier.

Applications

Specific clinical tasks are often undertaken using multiple EHR tools. (See Table 1) The choice of each tool is largely determined by the most effective way to impact workflows key to the desired improvements.

Table 1:

EHR tools and common use in applications for management of heart failure

	Application
Tool	HF case detection	Risk stratification	Care gaps	Resource allocation	Patient engagement
Dashboard		+	+	+
Predictive analytics	+	+		+
Remote monitoring		+		+	+
CDSS	+	+	+	+
Patient portal			+		+

CDSS = Clinical decision support systems

Automated diagnosis and risk stratification

Many EHR-based applications rely on identifying patients to whom HF-specific interventions should be applied. Historically, this has been a largely manual process wherein providers or create lists based on chart review or knowledge of the patient. This approach likely resulted in high specificity, but large numbers of undiagnosed patients may have been missed, especially in large healthcare systems. A variety of strategies have been used to automatically identify HF patients using EHR data including simple heuristic criteria,³² natural language processing,³³ and ML models.³⁴ One application of combined automated HF diagnosis and risk stratification is identification of patients who may qualify for advanced therapies including implanted devices and transplantation.^35–37 An evolving important application of EHRs is the support of end-of-life care planning, either by referral or presentation of relevant information such as risk scores and a patient’s expressed wishes regarding care goals.³⁸

There are many examples of risk stratification models to identify HF patients and assess their likelihood of adverse clinical events. Rehospitalization within 30 days of discharge is one of the most common endpoints that such models aim to reduce. An increasing proportion of these models are generated and/or validated using EHR data, whereas earlier models relied primarily on derivation and validation using clinical trial, observational cohort, or registry data. Training predictive models using real-world EHR data might result in better performance when finally implemented as clinical tools. However technical limitations of EHRs have until very recently required even models derived from EHR data to be accessed and return results to clinicians outside of the EHR and normal workflow. The convenience of automated risk stratification can therefore increase the likelihood that it will be used in clinical care and leverage more patient data than a clinician typically has time to review, although presence and accuracy of key variables (e.g. New York Heart Association class) can limit practical use.

Care gaps and standardization of management

EHRs appear useful in assessing adherence to HF management guidelines and identifying care gaps,³⁹ and HF management guidelines can be encoded within EHRs to improve adherence to evidence-based recommendations.⁴⁰ Dashboards, predictive analytics, CDSS, and patient portals have all been applied to closing care gaps.^{14,25,39,41,42} Although EHR reporting tools and business intelligence are not discussed extensively here, assessment of adherence to care guidelines and quality metrics is a core use of such systems. It should be noted that due to the lack of guidelines and, until recently, effective medications for improving outcomes in HF with preserved ejection fraction, studies and data on EHR-facilitated improvement in care gaps is limited to HF with reduced ejection fraction.

HF-specific care pathways have been demonstrated to improve some HF outcomes, particularly in hospitalized patients.^43–46 These pathways leverage HF-specific order sets but also include guidance regarding decision-making, for example regarding level of care on hospital admission. Pathways are often institution-specific and designed by staff based on local workflows. Historically these pathways were available only outside the EHR, in the form of either paper flowcharts or web-accessible PDFs. Recent EHR advances have made it possible to render interactive care pathways for a number of both acute and chronic conditions. In some implementations, patient-specific information is imported automatically, and the clinician can place orders directly from the displayed pathway. Data on interactive EHR-based pathways are limited but do show promise that this may be both an efficient and effective functionality for standardizing care.^47,48

Resource management

EHR-based predictive analytics may be used in combination with dashboards and CDSS to allocate scarce or costly resources to patients with the highest potential benefit. In doing so, it is possible to coordinate multiple services efficiently in a semi-automated fashion.⁴⁹ EHR tools such as order sets and alerts can also be used to optimize selection of e.g. radiology studies and reduce use of costly interventions.

Outcomes

The above-described applications implemented using EHR tools are also used in several combinations in an effort to improve clinical outcomes, care measures, and system efficiency (see Table 2). Data regarding the impact of EHRs generally and EHR tools individually is mixed and highly dependent upon the institution clinical actions taken based on the outputs of these tools. There is marked variation in efficacy between different disease conditions as well. For this discussion, data regarding associations between EHR tools and outcomes are presented almost entirely with respect to HF populations.

Table 2:

EHR tools and impact on heart failure outcomes

	Outcomes
Tool	Clinical events	Quality of care	Cost/ utilization	Patient experience
Dashboard	+
Predictive analytics	+		+
Remote monitoring				+
CDSS	+	+	+
Patient portal		+		+

CDSS = Clinical decision support systems

Care Quality

EHR-based interventions have been shown in some studies to increase use and optimization of certain forms of GDMT for HF, although other analyses have not demonstrated any benefit.⁵⁰ The best evidence of EHR efficacy thus far demonstrates increased use of GDMT, in particular BB, angiotensin converting enzyme inhibitor or angiotensin receptor blocker, mineralocorticoid receptor antagonists, and HF education.^21,51–53 EPIC-HF showed that an educational tool regarding GDMT recommendations sent via the EHR patient portal to patients prior to their clinic visit resulted in greater rates of GDMT intensification than control patients.²⁵ Intuitively one might expect EHRs to improve rates referral for implanted devices (CRT, LVAD), but data are mixed.⁵³ The latter may be due to difficulty in encoding the criteria for device therapy referral without resulting in an excess of premature prompts during optimization of GDMT. However there are examples of complex patient selection algorithms coupled with targeted CDSS embedded in the EHR that successfully identify and promote referral of appropriate patients to advanced heart failure services resulting in improved outcomes.^37,54

Clinical events

Reduction of 30-day hospital readmissions is one of the most widely studied applications of EHR technology in care of patients with HF. There is some evidence that 30-day readmissions following HF hospitalization may be reduced by using HF dashboards.⁵⁵ HF order sets have been associated with increased guideline adherence, decreased length of stay, and all-cause mortality.^56–58 RPM using implanted devices has shown promise in identifying HF decompensation^59,60 and in some cases, improving HF outcomes such as hospitalization.⁶¹ Availability of RPM data in the EHR is limited, but this is an active area of development and some subset of these data will likely be available in the EHR in the future.

Optimization of predictive analytics identifying patients at high risk for readmission or mortality are a cornerstone of EHR-based efforts to improve patient outcomes. A number of high-performing models predicting hospital readmission have been derived using EHR data, but translating these into reduced HF readmission rates has been complicated a) implementing models in the EHR and b) establishing effective accompanying workflows and interventions. Real-time EHR-based predictive analytics that have been combined successfully with clinical interventions have been shown to improve 30-day readmission rates. For example, Amarasingham et al. demonstrated that provision of intensive inpatient interventions based on an EHR-based predictive model for 30-day readmission resulted in a reduction in readmission rate from 26.2% to 21.2% (OR 0.73, p=0.01).⁴⁹ Using a different approach, the Stanford Heart Failure dashboard leveraged automated HF patient identification tools in combination with a dashboard to improve utilization of an established readmission reduction program reduced 30-day readmission rates from 14% to 10.1%.⁹ Data regarding reduction in clinical events using RPM and EHR patient portals are limited, although there is some evidence that patient engagement with these tools is associated with improved outcomes such as hospital readmission.⁶²

Patient experience

Access to EHR data has also been shown to improve engagement, communication with provider team, empowerment, self-education, and self-management behaviors in patients with HF.^63–65 These have been shown in some settings to translate into improved self-care behavior,⁶⁵ QOL,⁶⁵ GDMT escalation,²⁵ and accuracy of EHR data.⁶⁶ Electronic capture of patient reported outcomes such as Kansas City Cardiomyopathy Questionnaire via EHR is also perceived by patients with HF to facilitate communication with their providers and as beneficial to their care.⁶⁷

Cost and resource utilization

General cost savings associated with EHR implementation following the HITECH act is difficult to determine but for Medicare is estimated at ~ $1 billion/year between 2010 and 2013.⁶⁸ With respect to management of HF, EHRs can reduce costly or unnecessary interventions while increasing utilization of under-used interventions. EHR-based CDSS has been shown to reduce costs by decreasing redundant or inappropriate testing as well as discouraging use of potentially harmful medications.^69,70 Access to comprehensive clinical data on presentation to the emergency department can optimize acute care to improve inpatient outcomes including length of hospital stay.As stated earlier, CDSS directed by predictive analytics can also help allocate limited resources and interventions such as multidisciplinary care interventions to high risk patients.⁴⁹ Closing care gaps for invasive therapies can also increase use of interventions like ICD, CRT, LVAD, and transplant. Given the high cost of care for the HF population, incremental improvements in efficiency have potential to result in significant savings on institutional, regional, and national levels.

Conclusions

EHRs are becoming ubiquitous in modern clinical practice. They offer a wide array of applications that support clinical care including several that have been shown to improve the efficiency and quality of care in patients with HF. EHRs should not be seen as replacements for clinicians and clinical judgement. Clinician participation in design, education in optimal use, and integration of these tools with corresponding workflows can help clinicians to spend less time looking at the EHR and more time caring for the patient.

Key Points.

• Electronic health records provide an array of tools including dashboards, analytics, patient portals, and decision support that facilitate care of patients with congestive heart failure.

• Electronic health record tools can be used for several applications important in management of heart failure including risk stratification, and care standardization.

• Although mixed, there is evidence that electronic health record tools can be used to improve clinical outcomes, resource utilization, quality of care, and patient satisfaction.

• Electronic health record tools are more usable and useful with clinician design input.

Synopsis.

Increasing global adoption of electronic health records (EHRs) is transforming delivery of clinical care. EHRs offer tools that are useful in the care of heart failure ranging from individualized risk stratification and decision support to population management. EHR tools can be combined to target specific areas of need such as standardization of care, improved quality of care, and resource management. Leveraging EHR functionality has been shown to improve select outcomes including guideline-based therapies, reduction in adverse clinical outcomes, and improved cost efficiency. Central to success is participation by from clinicians and patients in design and feedback of EHR tools.