Harnessing the Power of the Electronic Health Record for ALS Research and Quality Improvement: CReATe CAPTURE-ALS and the ALS Toolkit

Abstract

The electronic health record (EHR) is designed principally to support the provision and documentation of clinical care, as well as billing and insurance claims. Broad implementation of the EHR, however, also yields an opportunity to use EHR data for other purposes, including research and quality improvement. Indeed, effective use of clinical data for research purposes has been a longstanding goal of physicians who provide care for patients with ALS, but the quality and completeness of clinical data, as well as the burden of double data entry into the EHR and into a research database, have been persistent barriers. These factors provided motivation for the development of the ALS Toolkit, a set of interactive digital forms within the EHR that enable easy, consistent, and structured capture of information relevant to ALS patient care (as well as research and quality improvement) during clinical encounters. Routine use of the ALS Toolkit within the context of the CReATe Consortium’s IRB-approved Clinical Procedures to Support Research in ALS (CAPTURE-ALS) study protocol, permits aggregation of structured ALS patient data, with the goals of empowering research and driving quality improvement. Widespread use of the ALS Toolkit through the CAPTURE-ALS protocol will help to ensure that ALS clinics become a driving force for collecting and aggregating clinical data in a way that reflects the true diversity of the populations affected by this disease, rather than the restricted subset of patients that currently participate in dedicated research studies.

Introduction

Amyotrophic lateral sclerosis (ALS) is an inexorably progressive disease that remains without effective therapy despite a multitude of clinical trials over many decades ¹. While the reasons for this therapeutic vacuum are manifold ², the relatively low rates of patient participation in clinical research studies has been identified as one important contributing factor ³. The fact that only a small minority of patients with ALS are included in clinical studies may, in part, be explained by time limitations faced by both patient and caregiver(s) as well as physical and financial barriers ³. An efficient process for collecting research data at the point of care through multidisciplinary ALS clinics, where most patients receive their clinical care, may afford an opportunity to overcome these challenges.

Multidisciplinary clinic visits usually take several hours, requiring multiple assessments by numerous providers. Research visits are often similarly time consuming, and there is significant overlap between research procedures and clinical assessments such as measurement of vital capacity and administration of the Revised ALS Functional Rating Scale (ALSFRS-R)⁴. This “redundancy” is not only taxing to patients, but also time consuming to the staff, who may need to perform repetitive assessments. Ideally, the extensive phenotypic data routinely collected in the multidisciplinary clinic could be used both to support the provision of clinical care and also to advance research. This has increasingly become feasible since the introduction of the electronic health record (EHR). Indeed, over the past two decades, multi-institutional collaborative efforts to utilize EHRs in research have already yielded significant new insights into the understanding of disease phenotypes and their relationship to underlying genotypes ^5,6.

Although effective use of clinical data for research purposes has been a longstanding goal of physicians who provide care for patients with ALS, the burden of double data entry – into clinical notes and a research database – has significantly impeded these efforts. Now, however, with broad implementation of the EHR in clinical care, including ALS clinics, we have a unique opportunity to optimize our use of the EHR to facilitate both patient care and research. Although the rigor of traditional research studies may not easily be replaced by data collected through the EHR, efficient and effective use of EHR data for research purposes has the potential to dramatically increase patient participation in research ⁷.

These considerations served as an impetus for the Clinical Research in ALS and Related Disorders for Therapeutic Development (CReATe) Consortium to develop and implement the ALS Toolkit and the ClinicAl Procedures To sUpport REsearch in ALS (CAPTURE-ALS, clinicaltrials.gov identifier [NCT03489278]) study protocol, both of which are described below.

Challenges to Collecting Research-Quality Data During Routine Clinical Care

The EHR is designed principally to support the provision and documentation of clinical care, as well as billing and insurance claims ⁸. In order to also utilize the EHR for research purposes, three main challenges must first be overcome to generate usable data. First, clinical data might not be collected with the same rigor as research data. Second, clinical data are often not collected in a systematic or structured manner. Relevant clinical information, for example, may be found in free text notes. Third, and most challenging, clinical practice and style of documentation can vary enormously among providers. These factors result in significant variability in the completeness and quality of healthcare data, which complicates aggregation of data across organizations.

These challenges have been partially addressed in the past through strategies such as the use of natural language processing to extract specific information from text fields ⁹, although this approach is not without significant limitations¹⁰. Alternatively, administrative data such as ICD (international classification of diseases) codes can be used for more structured data collection, but the ICD codes alone are insufficient when the goal is to collect detailed phenotypic data. Further, ICD codes are frequently inaccurate and incomplete¹¹. Manual abstraction of EHR data into a dedicated research database is also a common approach, though this requires significant additional time and effort from research staff and introduces the potential for transcription errors. Furthermore, the variability in documentation requires those abstracting from the medical record to make estimates and judgment calls in order to complete research forms that require quantifiable information. These are issues that the Muscular Dystrophy Association’s (MDA) NeuroMuscular ObserVational Research (MOVR) data hub ¹², which is attempting to collect and aggregate multicenter data for a range of neuromuscular disorders, will need to address. MOVR also seeks integration with the EHR, and strategies such as the ALS Toolkit (see below) that we describe in this manuscript offer a potential mechanism to accomplish this goal, at least for patients with ALS.

The ALS Toolkit: Structure and Process

The ALS Toolkit was conceived to leverage the capabilities of an EHR to address the aforementioned challenges. In developing the ALS Toolkit, our goals were threefold: (a) to facilitate the systematic and structured collection of data during routine clinical care, without increasing the burden on healthcare providers; (b) to improve access to analyzable, high-quality research data that could be readily aggregated across organizations; and (c) to utilize the information collected to assess and improve the quality of the care provided based on professional organization recommendations such as the American Academy of Neurology (AAN) ¹³.

The ALS Toolkit was developed through a collaboration between three ALS neurologists (JK, MB, and DW) at major ALS Centers, and Epic (Verona, WI, USA), the developer of a widely used EHR system in the United States¹⁴. The neurologists identified data elements that are relevant to the clinical care of patients with ALS, many of which may also be used for research purposes. The selection of data elements was informed by knowledge of the AAN quality measures in the care of patients with ALS ¹³ with the goal of quantifying the frequency with which these quality measures are met, as well as the expectation that sharing data about adherence to these quality metrics with providers will empower quality improvement.

The ALS Toolkit comprises a set of SmartForms™ (Table 1, Figures 1 and 2) that enable easy, consistent, and structured capture of the clinical information collected during patient encounters. (Although EHR software also provides tools for end-users to generate standardized lists of texts, these tools are typically used to facilitate rapid documentation, but not to extract specific data elements). The interactive forms that comprise the ALS Toolkit are embedded in a navigator (i.e. a table of contents for components of clinical workflow) to guide physicians and allied healthcare professionals to systematically address important aspects of the disease. Importantly, however, none of these data elements are required, leaving the clinician with maximum flexibility to document only the information that is relevant to the clinical encounter. During the development of the Toolkit, we carefully considered, but decided against, “hard-stops” for missing data, requiring the clinician to enter something (even if incorrect) before proceeding. We also elected not to introduce informational alerts that simply serve as reminders to the clinician to enter a particular data element. While these can be dismissed, they do require additional clicks to do so. Our concern was that both approaches might render the ALS Toolkit less user-friendly, which might in turn hinder broad implementation.

Table 1.

List of ALS Toolkit interactive forms

Form	Information Collected	Encounter vs patient level
Additional Demographics a	Highest education level, veteran status, dominant hand, dominant foot/leg, Global Unique Identifier (GUID), caregiver information.	Patient
Onset Phenotype	Anatomical site, distribution and timing of onset of earliest symptoms, and date of diagnosis.	Patient
Diagnosis Phenotype	Date of initial symptoms, date of initial weakness, date of ALS diagnosis, premorbid body weight, date of first documented ALSFRS-R, score of first documented ALSFRS-R, ΔFRS since initial symptoms, ΔFRS since initial weakness.	Patient
Family Genetics	Family history of ALS, FTD and other neurodegenerative diseases, as well as results of genetic testing in the individual and in family members with neurodegenerative disease.	Patient
Revised ALS Functional Rating Scale (ALSFRS-R)	The 12 items in ALSFRS-R, a validated instrument for quantifying functional decline as disease progresses.	Encounter
Office Pulmonary Function Tests	Oxygen saturation, vital capacity (forced or slow), negative inspiratory force, sniff nasal inspiratory pressure (SNIP).	Encounter
ALS Cognitive Behavioral Screen (ALS-CBS)	The ALS-CBS score is a validated brief screen for cognitive and behavioral impairment	Encounter
General Neurological Examination	Basic, non-neuromuscular examination relevant to ALS.	Encounter
Neuromuscular Examination	Detailed neuromuscular examination, focused on the motor system.	Encounter
Clinical Assessment	Current symptoms, medical needs and treatments employed (respiratory, bulbar, mobility, ADLs, pain, psychosocial, cognitive/behavioral, active medical treatments for ALS (such as Riluzole), advanced care discussion, research related discussions (e.g. CAPTURE-ALS consent, brain or tissue donation, National ALS registry referral, check box to confirm that research opportunities were discussed). This form also captures key clinical events (e.g. gastrostomy tube placement, use of permanent assisted ventilation) and the dates on which these events took place.	Encounter
EMG Summary	Summary of EMG findings around the time of diagnosis	Encounter

Abbreviations: ΔFRS = Estimated progression rate: (48 – baseline ALSFRS-R) / months from symptom onset to baseline; ADLs, Activities of Daily Living; ALS-CBS, ALS Cognitive Behavioral Screen; ALSFRS-R, ALS Functional Rating Scale—Revised; EMG, electromyographic; FTD, frontotemporal dementia; GUID, global unique identifier; SNIP, sniff nasal inspiratory pressure.

^aIn addition to basic demographic data, such as gender, date of birth, race, and ethnicity, which are collected elsewhere in electronic health record.

Figure 1:

Clinical Assessment Form (partial screenshot) showing the ALS Clinical Assessment section (that captures key events since the last clinical visit and principal concerns the patient would like to address), as well as details of each domain: Respiratory, Sleep and Energy; and Bulbar, Gastrointestinal and Genitourinary. (Domains not shown: Gross Motor, Mobility and ADLs; Pain; Psychosocial; Cognitive, Behavioral, Mood and Pseudobulbar Affect; Treatments; Life Planning Assistance; and Research.) This form utilizes skip logic. For example, when the gastrostomy placed button is selected (shown in navy blue), the user is prompted to enter the date on which the gastrostomy tube was placed. The form also utilizes tristate buttons that are colored green (with a negative sign) when absent, colored red (with a positive sign) when present, and light-blue when information was not collected. The ALS Toolkit navigator is also shown in the left-hand margin.

© 2021 Epic Systems Corporation. Used with permission.

Figure 2.

A screenshot shows the completed cranial neuromuscular exam form, documenting the presence of atrophy and fasciculations in the lower face and tongue (but not the upper face), along with mild facial weakness, slow tongue movements, and a moderate degree of tongue weakness. The jaw jerk and gag reflexes are both present but not pathologically brisk, and speech is dysarthric. Pseudobulbar affect is present. Additional clinical notes are captured in the Comment box.

© 2021 Epic Systems Corporation. Used with permission.

Structurally, the ALS Toolkit records data at “patient” and “encounter” levels. Patient-level data are elements that only assume a single value or record for a given patient. Examples include the date and anatomical site of symptom onset, date of diagnosis, and date of tracheostomy. Once entered into the ALS Toolkit, these data elements carry forward from one encounter to the next. Importantly, these data elements also remain editable and may be updated if the provider identifies an error. By contrast, encounter-level data are unique to each clinical encounter. Examples include the ALSFRS-R, vital capacity, and elements of the neurological examination. Encounter-level data are entered at each clinical visit and are tracked over time, and do not auto populate from prior encounters – this is to minimize the potential error that might result from simply “cloning” prior encounter data. The SmartForms™ are dynamic; sections of forms are automatically shown/not shown based on logic, guiding the efficient collection of information.

The ALS Toolkit is available to all institutions using the EHR system in which the Toolkit was designed, obviating the need for each ALS clinic to design their own local set of interactive forms and, thereby facilitating efficient implementation. The prebuilt forms also ensure consistent data collection across institutions and eliminate the need for localized data mapping, thereby lowering the barriers to multisite data aggregation. Importantly, the technical design of the ALS Toolkit specifically aims to minimize the burden on local (i.e. at each medical center) information technology groups for implementation within their EHR system. Information regarding the ALS Toolkit, including its structure and setup instructions for local implementation, can be found on the EHR’s customer documentation hub (ie, Epic UserWeb).

Clinical Experience Using the ALS Toolkit

The ALS Toolkit has been in active use in the multidisciplinary ALS Clinic at the University of Miami since January 2018; California Pacific Medical Center (June 2018), University of Minnesota (June 2018), and University of Texas San Antonio Health Sciences Center (August 2019); and more recently, Wake Forest School of Medicine (November 2020). Several new centers are now also joining the effort, rapidly expanding the use of the ALS Toolkit. Recognizing that the multidisciplinary clinic at each center may be structured differently, each clinic has the flexibility to adopt their own approach to using the ALS Toolkit. Indeed, the ALS Toolkit forms can be filled out by any provider with appropriate EHR access.

At the University of Miami, for example, in preparation for implementing ALS Toolkit in the multidisciplinary clinic, the clinic director met with allied health professionals who work in the clinic to ensure that all team members had a holistic view of the workflow. To improve the team’s efficiency during a clinic visit, each team member was appropriately trained to complete specific parts of the ALS Toolkit; for instance, pulmonary function tests are documented by the respiratory therapist, while the ALSFRS-R is administered and entered by the clinic nurse. Being the first clinic to pilot the use of the ALS Toolkit and to do so for all ALS patients, we expected transient delays as multidisciplinary team members acclimatized to the new workflow and might require additional time for documentation. In anticipation, we scaled back clinic volume (i.e., number of patients scheduled per full clinic day) by ~20% for a few weeks, but found that efficiency quickly returned to pre-Toolkit implementation levels. The benefit of insights and experience gained from ALS Toolkit roll-out and use at the University of Miami has obviated the need for other clinics to even temporarily scale back patient volume in clinic during ALS Toolkit implementation.

For a given patient, each clinic visit is structured as a single encounter within the EHR, with all providers in the multidisciplinary clinic accessing the same encounter - and, therefore, the same instance of the ALS Toolkit. While data entry for different sections of the ALS Toolkit is completed by different team members, at the end of the clinic all sections produce a single visit note. An additional advantage of structuring our multidisciplinary clinic as a single encounter in the EHR is that all allied health professionals bill incident to the treating neurologist. Consequently, patients are only responsible for a single copay. One constraint with the current ALS Toolkit is that the toolkit forms can only be edited by one authorized user at a time. This restriction serves to maintain data integrity and to permit utilization of the desired advanced scripting capabilities. We have not found this to be a major limitation in the multidisciplinary clinic; each provider has their own designated section of the ALS Toolkit to complete for each patient, which is done during their rotation with the patient. When the provider leaves the room to see the next patient, they also exit the ALS Toolkit encounter (akin to leaving the patient’s chart in the room for the next provider).

The ALS Toolkit guides clinical providers through the steps of a comprehensive ALS-related history and examination (see Table 1). The structure and organization of the Toolkit allow a more meticulous description of the patient’s status. The compiled information can then automatically be pulled into a well-structured progress note. Moreover, the captured structured data (e.g. ALSFRS-R or vital capacity) can easily be tracked and longitudinal data displayed in progress notes. (Table 2).

Table 2:

Example of longitudinal data auto-generated and imported into a progress note

ALSFRS-R Total Scores		Slow Vital Capacity Measurements (erect)
Date	Score	Date	SVC (liters)	SVC (% predicted)
08/01/2019	18	08/01/2019	1.10	20
04/01/2019	22	04/01/2019	2.94	50
12/01/2018	28	12/01/2018	3.15	54
09/01/2018	31	09/01/2018	4.02	72

^*dates and data are fictitious

ALSFRS-R: ALS Functional Rating Scale - Revised

The ALS Toolkit can also be used for documentation outside of the multi-disciplinary clinic setting. Many clinics also use the ALS Toolkit for documentation of the initial clinical encounter with a new ALS patient, even if outside of the multi-disciplinary clinic setting. Also, if a nurse calls the patient prior to a clinic to administer the ALSFRS-R, for example, the provider can choose whether to enter this information in its own instance of the ALS Toolkit, or to hold the information and enter it at the time of the multidisciplinary clinic visit.

Although the ALS Toolkit is available to all institutions using the developer’s EHR system, regardless of their participation in the CReATe Consortium’s CAPTURE-ALS study (see below), we encourage all users to join this effort to aggregate ALS patient data across ALS clinics nationwide.

Using the ALS Toolkit for Research: CReATe CAPTURE-ALS

Development of the ALS Toolkit was impelled by the CReATe Consortium, which, through one of its research aims, seeks to aggregate EHR data across participating sites for research purposes. To this end, CReATe has established the IRB-approved CAPTURE-ALS study protocol (NCT03489278), and developed electronic informed consent and electronic Health Insurance Portability and Accountability Act (HIPAA) authorization forms. These consent forms are readily available through an easy to use online portal, independent from the EHR, and accessible through any electronic handheld device. As part of the CAPTURE-ALS protocol, consent also permits collection of blood samples for research purposes (which could be used in future DNA extraction or biomarker studies, for example). The long term goals of the CAPTURE-ALS protocol include: (a) comparison of ALS phenotypes based on EHR data to the data collected through traditional research visits; (b) improving the quality of ALS patient care through iterative use of site-specific reports of defined quality metrics; and (c) national aggregation of data to permit large-scale research projects that incorporate information from the broadest possible swath of ALS patients. In the future, the ALS Toolkit might even be used to capture clinical outcomes (e.g. ALSFRS-R or survival) in pragmatic or phase 4 clinical trials.

Data collected through the ALS Toolkit are most useful if they can successfully be extracted from the EHR and aggregated across participating centers. In collaboration with the software developer, a standard set of Structured Query Language (SQL) codes were developed to streamline the process of obtaining a uniform dataset from each participating center’s EHR database. These SQL queries pull data exclusively from patients with an active consent on file. In providing consent, clinic patients agree to allow a limited dataset (LDS) (which includes personally identifiable information [PII]) of their EHR data to be aggregated centrally by the CReATe Consortium. The data elements included in the extraction are prespecified and include all data collected through the ALS Toolkit (including that entered by allied health professionals), as well as a restricted set of data from other parts of the EHR, such as medical history, medications and relevant laboratory results. Dates (e.g. date of birth, date of diagnosis, dates of clinic encounters) are the only PII included in the limited dataset as they are necessary for the calculation of age- and time-related variables such as age at diagnosis or duration between clinic encounters.

Due to the presence of PII, the CReATe Consortium has put in place a Data Use Agreement (DUA) with each CAPTURE-ALS participating site, which governs the rules regarding data ownership, HIPAA compliance, intellectual property, and indemnification. The LDS are then transferred to the University of Miami via a secure file transfer protocol (SFTP) and aggregated by the data team that forms part of the Administrative Core of the CReATe Consortium.

One of the goals of the CReATe CAPTURE-ALS study, as noted above, is to validate the EHR data (e.g. ALSFRS-R rate of decline) against the gold-standard – namely, data collected from the same individuals through traditional research visits. To this end, date of birth is used as a secondary identifier to match EHR and research visit data, with NIH’s Rare Disease Registry Program Global Unique Identifier (RaDaR GUID) being the primary linkage variable. Once these steps are completed, dates are deleted and only fully deidentified data are retained for statistical analysis.

Conclusions

The ALS Toolkit is a long-awaited resource that facilitates structured collection and entry of data at the point of care during a multidisciplinary ALS clinic visit. While providing an easy to use workflow for ALS care providers, it also facilitates systematic collection of data that we anticipate will be of sufficient quality to facilitate research. Its ability to semi-automatically generate a well-structured progress note is a particularly welcome feature. By pulling clinically relevant information, captured via the ALS Toolkit, into the progress note, it increases provider efficiency in clinical documentation. Widespread use of the ALS Toolkit through the CAPTURE-ALS protocol will help to ensure that ALS clinics become a driving force for collecting and aggregating clinical data in a way that reflects the true diversity of the populations affected by this disease, rather than the restricted subset of patients that currently participate in dedicated research studies.

Abbreviations

ALS

Amyotrophic Lateral Sclerosis

EHR

Electronic Health Record

CReATe

Clinical Research in ALS and Related Disorders for Therapeutic Development

CAPTURE

Clinical Procedures to Support Research

ALSFRS-R

ALS Functional Rating Scale – Revised

ICD

International Classification of Diseases

MDA

Muscular Dystrophy Association

MOVR

NeuroMuscular ObserVational Research

AAN

American Academy of Neurology

HIPAA

Health Insurance Portability and Accountability Act

SQL

Structured Query Language

LDS

Limited Data Set

PII

Personally Identifiable Information

DUA

Data Use Agreement

SFTP

Secure File Transfer Protocol

RaDaR

Rare Disease Registry Program

GUID

Global Unique Identifier

Appendix 1. Authors

Name	Location	Contribution
Volkan Granit	University of Miami	Drafting of the initial manuscript; data acquisition; critical review and revision of the manuscript.
Anne-Laure Grignon	University of Miami	Patient consent and enrollment; data acquisition; critical review and revision of manuscript.
Joanne Wuu	University of Miami	Study conceptualization and design; study oversight; drafting of the manuscript for intellectual content; critical review and revision of manuscript.
Jonathan Katz	California Pacific Medical Center	ALS Toolkit development; critical review and revision of manuscript.
David Walk	University of Minnesota	ALS Toolkit development; critical review and revision of manuscript.
Sumaira Hussain	University of Miami	CAPTURE-ALS protocol management; critical review and revision of manuscript
Jessica Hernandez	University of Miami	Data acquisition; patient consent and enrollment; critical review and revision of manuscript
Carlayne Jackson	University of Texas Health Science Center at San Antonio	Patient consent and enrollment; data acquisition; critical review and revision of manuscript
James Caress	Wake Forest School of Medicine	Patient consent and enrollment; data acquisition; critical review and revision of manuscript
Tom Yosick	Epic	ALS Toolkit development; critical review and revision of manuscript.
Nancy Smider	Epic	ALS Toolkit development; critical review and revision of manuscript.
Michael Benatar	University of Miami	Study conceptualization and design; ALS Toolkit development; data acquisition; study oversight; drafting of the manuscript for intellectual content; critical review and revision of manuscript.