Designing the Ideal Electronic Health Record

Yesterday, I keynoted a Veterans Administration meeting via teleconference (part of my effort to reduce travel, improve my carbon footprint, and be increasingly virtual) on the topic of designing the ideal electronic health record.

I was posed a simple question - If I had infinite resources, infinite time, and no legacy compatibility issues, how would I design the electronic health record of the future?

Here's my answer:

General
The web is the way. Given the 24x7 nature of healthcare, the need for physicians to be in many physical locations, and the multitude of clinician computing devices, the ideal EHR should be web-based, browser neutral and run flawlessly on every operating system. I highly recommend the use of AJAX techniques to give physicians a more real time interactive experience. Client/Server may have some user interface advantages, but it's just too challenging to install thick clients on every clinician computing device. Citrix is an expensive and sometimes slow remote access solution. Native web works.

Data in medicine is stored hierarchically i.e. a patient has multiple visits with multiple labs, with multiple results. This is a tree of data with the patient as the root and the lab values as the leaves. Using a hierarchical database such as Intersystems' Cache ensures that data for clinical care is stored in this tree format and thus can be very rapidly retrieved, ensuring fast response times for clinician users. For population health, clinical research, and performance reporting, relational databases work very well. Thus, I recommend a hierarchical database for the clinical care applications and relational data marts for the research applications.

The ideal EHR should incorporate decision support in laboratory, medication, and radiology ordering. EHRs should include "Event Driven Medicine" alerts about critical clinical issues and patient specific reminders about preventative/wellness care. Event Driven Medicine is the transformation of data into information, knowledge and wisdom based on decision support, business rules and timely notification of clinicians.

The EHR should include an easy to read clinical summary of all active patient problems, medications, visits, and labs and should be able to export this summary to personal health records such as Google Health, Microsoft Health Vault and Dossia.

Problem Lists
Problems should be entered via an electronic pick list of vocabulary controlled terms using SNOMED CT. The community of caregivers - PCPs, specialsts, ED physicians and hospitalists should be able maintain this problem list collectively, using social networking type tools. Call this Wikipedia for the patient. All caregivers should be able to associate notes and medications with entries on the problem list, making it easy to filter notes by problem and discontinue medications that are problem-specific when problems are resolved.

Medications
Medication Management features should include e-Prescribing for new medications, automatically linked to payer-specific formularies, electronic real time pre-auth/eligibility for high cost therapies, links to lifetime medication history from retail pharmacy and payer databases, and safety checking for drug/drug and drug/allergy interactions. Pharmacy initiated renewal workflow would reduce calls to the physician's office to refill medications.

Ideally, medication reconciliation features should include pre-population of the medication list based on the lifetime medication history from retail pharmacy, payer databases and personal health record applications. Using the same social networking type approach as mentioned with problem lists, all caregivers should be able to update/change/edit/comment on patient medications to keep them current. One click quick picks of commonly used medications should be available to make ordering Tylenol as easy as ordering books on Amazon.

Allergies
Allergies should be recorded by caregivers using vocabulary controlled entries for therapeutics, foods and environmental substances. Reaction type and severity should be codified as well as the identity of the allergy observer/documentation source i.e. did the patient self report that their Mom saw a rash to pencillin 30 years ago or did an ICU nurse watch the patient anaphylax to pencillin?

Visits
Each visit should be documented with a reason for visit (symptoms or problem), a pre-existing condition flag if the patient has had this before, a diagnosis, a list of therapies given, and the followup arranged.

Notes
Notes should be entered via structured and unstructured electronic forms. All text data should be searchable, so that physicians can easily locate old notes. Templates that are disease specific and macros that are specialty specific should be available to make documenting easier. Voice recognition for automated entry of free text should be available. Workflow for signing notes and forwarding notes to other providers should be easy to use.

Laboratory results
Laboratory results should be displayable in several ways - by date, by class of lab, by single result trended over time and in screening sheet format. Screening sheets are lists of disease specific lab results combined with decision support. For example, a diabetic screening sheet would include glucose, hemoglobin a1c, lipids, recent eye exam results, podiatry consults, and urinalysis. Alerts and reminders should be generated based on disease state, lab value, and trends.

As results are delivered, especially important results, clinicians should electronically sign an acknowledgement of lab result notification, ensuring that appropriate next steps are taken for patient care.

Radiology results
As mentioned in my recent blog on image management, all "ologies" should be stored in one place in the EHR and all should be viewable with a single electronic viewer. Radiology, Cardiology, GI, Pulmonology, Echo, Vascular, and Gynecology images should be easily viewable and these images should be managed according to business rules i.e. retained as required for medical record compliance, archived when no longer relevant etc.

Orders
Electronic ordering should include medications, Oncology Management, Laboratories, Radiology and general care (i.e. ordering home care supplies, wheelchairs etc). Orders should automatically be routed to the department and staff responsible for executing them.

Health Information Exchange
The EHR should be able to retrieve medication lists and clinical summaries from outside institutions as part of local/regional healthcare information exchange. EHRs should be able to send data to personal health records and receive patient entered data, especially telemetry data from home devices like glucometers, from personal health records.

Data Marts
Every night, data from the EHR should be exported to data marts for appropriate use with IRB approval for clinical trials, clinical research, population health analysis, performance measurement, and quality improvement.

At BIDMC, we're continuously improving our systems and we're well on the road to achieving much of this functionality. Of course, we'll never be done because the goal of the ultimate electronic health record is a continuously evolving target.