As our smart phones become even smarter, our EHRs will likely need to follow suit, especially with a health care system in transformation.
Health care information technology adoption has taken a relatively sharp jump, last year in particular. According to a recent Office of the National Coordinator for Health Information Technology/American Hospital Association study, the number of U.S. providers using either a basic or a comprehensive electronic health record system is on the rise. The substantial increase between 2010 and 2011 probably is due in no small part to the federal incentives for meaningful use.
Still, the percentage of hospitals and physician practices with very sophisticated electronic health records remains relatively modest. Additionally, while the industry emphasizes the need to lessen the gap between the haves and the have-nots, small, rural providers still show a much lower rate of technology adoption than their larger counterparts.
As we encourage EHR use, shift the balance from basic to comprehensive EHRs, raise the bar on the definition of meaningful use, and narrow the gap between provider haves and have-nots, we must also focus on the shifting nature of EHRs — a shift that will be necessary if EHRs are to support accountable care effectively.
The Transaction-Based EHR
If you look back at the electronic health records implemented over the years, you’ll see that the focus has been on transactions such as writing a prescription, retrieving a result or documenting a visit. In automating these transactions, we had to address several different challenges.
For example, in the outpatient arena, every 15 minutes, there’s another patient to see, so transactions must be performed in a fairly narrow window of time — about five or seven minutes. In the inpatient arena, the core challenge is the coordination of care: ensuring that those responsible for a patient’s care are aware of the care tasks that need to be performed and the results of tests and procedures. In both inpatient and outpatient situations, the EHR has to support care in very diverse settings such as obstetrics, primary care, neurology, cardiology and the ICU.
As the early EHRs came on the market, we emphasized speed, efficiency, ease of use and the ability to cover the spectrum of inpatient and outpatient settings. We talked about the ROI of these systems in terms of the improvement over paper-based transactions. There were fewer legibility problems and fewer medication errors, documentation was more complete, data was more accessible, and transcription costs dropped. A variety of studies illustrated the strides we made as an industry in improving the clinical transactions associated with care delivery.
To be fair, the focus of these EHRs was not strictly on supporting transactions. We built in clinical decision support that provides health maintenance reminders and ensures order appropriateness. We added some basic analytics to assess provider performance, care costs and quality. We also ventured into patient engagement via personal health records and disease-specific monitoring. While these advances provided key stepping-stones, the core focus of today’s EHR remains the transaction and making it as efficient and as effective as possible.
Tackling the Accountable Care Challenge
There are several cross-cutting problems we have to solve if we are to provide the type of care that we expect and, frankly, need to deliver under this new era of reimbursement and of patient and consumer expectations:
Failure to follow the evidence. Any number of studies during the past decade have shown that U.S. health system performance continues to fall far short of what is attainable, especially given the enormous resources devoted to improvement efforts. For example, in 2003, Elizabeth McGlynn, Ph.D., and her colleagues at the RAND Corp.looked at some well-established treatment guidelines.
They attempted to determine to what degree patients receive care according to the protocol. They found that only a little more than 50 percent did, indicating a massive failure to follow the evidence for a good amount of the care. Now, obviously, given reimbursement pressures and expectations, this type of failure is going to be less and less tolerable.
The care delivery system has begun to respond to those dismal findings. More recent studies, such as the Commonwealth Fund’s 2011 National Scorecard on U.S. Health System Performance, show that adherence to treatment standards for heart attack, heart failure and pneumonia, for example, has experienced notable improvement. This likely is due to federal policy linking Medicare payment updates to hospitals’ agreement to publicly report their results. Nonetheless, a significant gap remains between leading and lagging hospitals.
There’s simply too much to know. The second problem we face is that there’s too much information for physicians to master. If you had leukemia 100 years ago, they called it blood disease and you likely died. Sixty years ago, five different forms of leukemia or a lymphoma had been identified; today there are upward of 90 leukemia types or lymphomas combined. Last year, 700,000 articles were added to the referring biomedical literature. Ten years ago, it was 400,000. It is anticipated that in the coming years it’ll be 1 million. There’s just too much to know.
Increasingly, we have to help those who deliver care be up-to date on what the most current (and rapidly growing) knowledge or practice is in terms of both diagnosis and treatment.
A lot of care processes just don’t work very well. We’ve all heard stories, or perhaps experienced, a problem with care in which we can’t help but wonder what went wrong to cause such poor communication, how a key piece of information could have been overlooked, or why something fell through the cracks.
Extrapolated data from various studies of outpatient care clinical processes show that for every 1,000 women with a marginally abnormal mammogram, there appear to be 360 women who will not receive appropriate follow-up care. Similarly, for every 1,000 patients who qualified for secondary prevention of high cholesterol, 380 will not have an LDL-C screening within three years on record. As there’s no argument about what constitutes an appropriate next step, the question is: Why is there such a remarkable failure of a process?
The answer varies. In the case of the mammogram, at times the patient refused. Or she went to a different organization and the care provider wasn’t made aware. Or she received follow-up care, but it wasn’t recorded. Or perhaps, with approximately 150 test results crossing the desk of the average primary care provider every day, a particular patient may have fallen through the cracks.
These cross-cutting challenges — failure to follow the evidence, too much to know and inadequate care processes — indicate a need to surround the transaction with intelligence. At a minimum, this intelligence will need to suggest steps required to ensure that care follows the evidence, to identify treatment options that result from the dazzling pace of medical discovery, and to alert providers that care processes have deviated from acceptable levels of performance. This intelligence must detect acts of commission (choosing an outdated treatment approach) and of omission (a patient has failed to keep an appointment to see a specialist).
Toward the Intelligence-Based EHR
These scenarios illustrate that while we’ve made some impressive gains in automating core health care transactions, much work remains to address both the process and knowledge issues that plague care delivery and to ensure that the best evidence is followed routinely. To accomplish this, we must make the shift from a transaction-oriented record to an intelligence-oriented record, capable of achieving the following core objectives:
- Guide clinical diagnostic and therapeutic decisions.
- Ensure that the sequence of care activities conforms to the evidence and performance contract requirements.
- Monitor the execution of core clinical processes.
- Capture, report and integrate quality and performance measures into EHRs.
- Expand the scope of clinical data gathered about a patient, including data from other providers in the region, data provided by patients and data that is captured from one’s genome.
- Support the interactions of the care team, as all intelligence is not necessarily machine intelligence focused on specific transactions. Often the intelligence is the care team working together (perhaps using social network tools) to come to a conclusion and a consensus about how best to treat the patient.
In order to achieve these core objectives, the EHRs we provide must have several key capabilities, including:
- foundational sets of templates, guidelines and order sets that reflect the best evidence or established best practice;
- a process-management infrastructure that supports basic transaction checking such as drug-drug interactions, as well as asynchronous alerting like panic lab reporting and process monitoring and guidance;
- team-based care support such as shared work lists, as well as tools for patient engagement and health information exchange;
- novel decision aids like predictive models that can tell us if a particular patient is likely to be readmitted because he or she is fragile or has a substandard social situation at home that may negatively impact healing;
- context-aware order sets and documentation templates that guide the physician and help infer what types of orders should be placed and what types of documentation should be done
- intelligent displays of data, intelligent correction and identification of data, and extraction of structure by going through free text and pulling out quality measures or problems that were not previously in a patient’s problem list, for example.
Lastly, there will be a new generation of real-time analytics that will measure quality and process performance and assess guideline adherence, financial performance, and provider treatment and outcome variations.
Smarter Tools to Manage Big Data
The shift to intelligent EHRs will require incorporating the capabilities described. In addition, tomorrow’s EHRs will be shaped by work, still early, that applies big data techniques and methods to health care data. The big part of big data will be the result of the growth in patient data captured in EHRs coupled with data from imaging, molecular medicine, patient-provided data and insurance claims.
Consider hypertension, which can be managed with diet and lifestyle modifications and with medication in some cases. Also consider that improperly managed hypertension often can lead to more significant issues such as heart disease, stroke or even heart failure.
Through a combination of advanced in vitro diagnostics, diagnostic imaging, and the data archiving and extraction capabilities afforded by an intelligence-based EHR, we can identify the individuals most at risk and start them on a regimen modifying diet and behavior — well in advance of the onset of hypertension, and certainly well before more serious issues develop.
Bringing together the clinical evidence and drawing meaningful conclusions will require sophisticated and refined IT tools to extract valuable information from volumes of data. These same tools also will enable us to refine our processes and significantly reduce the variation in care that long has plagued our health care system. Big data approaches will be used to support post-market surveillance, comparative effectiveness and clinical trial hypothesis framing.
While we must not lose sight of our nation’s goals of increasing the adoption and meaningful use of EHRs by all providers, we must also understand that the transition from the transaction-based EHR to the intelligence-based EHR may become one of the most critical undertakings in our journey toward more accountable, cost-effective care.
John Glaser, Ph.D., is the CEO, Health Services, at Siemens Healthcare in Malvern, Pa. He is also a regular contributor to H&HN Daily.