| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
| ||||||||||||||||||||||||
FUTURE HISTORY |
1 Geriatrics Research, Education, and Clinical Center, Veterans Administration Salt Lake City Health Care System, Utah.
2 Department of Internal Medicine
3 Department of Medical Informatics, University of Utah Health Sciences Center, Salt Lake City.
Abstract
With deference to Isaac Asimov's The Foundation, which is the inspiration for this series, we briefly describe the "present history" of medical informatics (the application of information technology in medicine) in geriatrics, and then project a "future history" of this same endeavor. The older patient often has multiple acute and chronic problems that require management by a variety of medical professionals in a variety of settings. Proper care necessitates efficient gathering, integration, and management of information by each professional in each setting. As medical informatics evolves, we project that barriers to information exchange (both between providers and between providers and patients) will continue to decrease while the quality and relevance of exchanged information will continue to increase. The nexus of care will be the electronic medical record (EMR), which will shed its current paper chart metaphor and adopt an industrial process metaphor based on tasks and tolerances or goals. The multidisciplinary management of geriatric patients will strike a new balance: doctors, nurses, allied health professionals, family, and patients will all participate in the management of the patient's care. The EMR will coordinate data from a variety of novel sources, including wearable sensors monitoring physiologic parameters, falls, diet, ambulation, and medication compliance. The highly organized data in the EMR will allow explicit decision support for computer-facilitated, evidence-based care; will empower midlevel providers and patients with an increased role in the care plan; and will promote the realignment of care from hospitals/clinics to the patient's home.
Predicting the future, even in Foundation, is based on trend analysis. For our framework here we will focus on three trends: trends in provider-provider interaction, trends in provider-patient interaction, and trends in provider decision support. The first two are of special importance in geriatrics. The older patient is treated by a large number of different health workers in diverse settings, complicating communication at all levels. Understanding changes in clinical communication is critical to an understanding of the future of geriatric care. The third, trends in provider decision support, is a theme common to informatics generally but, as we shall see, offers special promise in the care of elderly people.
PAST AND PRESENT HISTORY
Provider–Provider Interaction
Today's geriatric patient receives care provided by a variety of medical professionals in a multitude of settings. It is not uncommon in the same year for an older patient to receive care in several clinics, a hospital, a skilled nursing facility, and at home (1). This care might be provided by a primary care physician, a hospitalist team, any of several medical or surgical specialist physicians, a physical or occupational therapist, an advance practice nurse, a registered nurse, a social worker, a dietician, and others. For most patients, a record of their care is stored in paper or electronic form, both at care facilities such as clinics and hospitals and also by the different providers themselves. The resulting barriers to information access are high. Such barriers are far more than a mere inconvenience. As the patient moves between providers and locations, hazardous lapses can occur in the care plan due to missing or out-of-date information (2).
The development of the electronic medical record (EMR) was motivated, in part, by the desire to allow multiple providers to access patient information from multiple locations. EMRs usually are not shared across institutions, but the goal of easy access by multiple providers within an institution or care network has been largely realized. That access, however, is usually limited to providing a static view of patient documentation, without regard to its integration and processing.
The implementation of most EMRs is based on the metaphor of the paper chart (3). The user interface of these systems typically uses tabs that mimic the tabs of paper charts (e.g., a tab for laboratory tests, a tab for consults, a tab for medications). Information about patient diagnoses and care plans are kept under the notes tab. Each encounter note is a stand-alone record that may or may not reference a previous or following note. Thus each note is a snapshot of the patient from the perspective of a given provider at a given time. Understanding the patient's history and care plan requires a careful read (and reread) of each note. At its best, the EMR should present an orderly, coherent, and rapidly assimilated view of the patient's progress (a fanciful analogy would be a child's flip pad displaying a galloping horse as one quickly flips the pages). Instead, perusing the succession of observations by different providers in the present EMR presents a jarring and disjointed depiction of the patient.
By design, EMRs facilitate entering and reading computerized notes. This feature has the adverse effect of decreasing the signal-to-noise ratio of pertinent information in a patient's record. For example, in the largest EMR in the world, the Veterans Administration's Computerized Patient Record System (CPRS), all notes by all providers are displayed on a scrolling selection list. Every note title on the list claims equal real estate, from an important admission history-and-physical note to a minor patient phone call note for a pharmacy refill. The plethora of choices makes it difficult to home-in quickly on essentials. Providers attempt to increase the signal and to decrease the noise by scanning down note titles to find their own last note, ignoring potentially pertinent notes of other providers (4).
Provider–Patient Interaction
In the past, a patient usually had to meet physically with a doctor to receive medical information, but the information received was highly pertinent. Barriers to access were high, but the signal-to-noise ratio was also high. The past contrasts sharply with the present, where thousands of medical sites are readily accessible on the Internet. Many sites do little to assist a reader in filtering and interpreting medical information, so barriers to access are low but the signal-to-noise ratio is also very low. Patients, aided by family members, can easily misinterpret information they find on the Internet. Documented cases of harm are rare (5) but a recent study found that medical website coverage tends to be inconsistent and requires a high level of reading ability to comprehend (6).
Providers have benefited from easier access to patient data in the EMR, but patients have not. To see their own data, which now is a right guaranteed by the Health Insurance Portability and Accountability Act, patients typically request a printed copy of their records. Initiatives to allow patients varying degrees of participation in managing their own records through the Web have been advocated (7). The flip side of accessing information is entering information. Very few EMRs allow patients to enter symptoms directly into the record. However, technology has existed for many years that allows patients to interact with the computer and enter in their symptoms. A patient starts by entering a chief complaint and the software presents the patients with a series of yes/no questions. This approach has been used as a way of eliciting pertinent positives and negatives and generating a preliminary differential diagnosis (8). This technology is a good example of decreasing barriers to patient access to entering data while increasing the signal-to-noise ratio of information.
Devices exist that prompt communication with patients at home. These monitors include a wireless hand-held unit that communicates with a base station that in turn communicates over telephone lines with an EMR. These systems can be automatically programmed to prompt patients to complete therapies and request information on monitoring parameters such as weight, pulse, and breathing status. In some populations such as indigent diabetics, this technology has been demonstrated to improve patient outcomes and decrease costs of care (9).
There has also been improvement in technology for obtaining and communicating patient physiologic parameters to providers. In the past, patient monitoring only occurred in the clinic or hospital. Now devices such as automatic blood pressure cuffs and Holter monitors are routinely used in the patient's home. Whereas most such portable monitors are only used intermittently, wearable sensors are now available for continuous patient monitoring. Following deaths in basic training, the U.S. Army commissioned a miniature, wearable monitor that records pulse, 3-lead electrocardiogram, core temperature, and other information. This technology can be configured to communicate wirelessly to a base station in the patient's home and relay information along telephone lines to the EMR. Other types of sensors exist that have not yet been integrated into commercial, wearable patient monitoring systems. Such sensors could be integrated into wearable devices that could measure acceleration to determine when a patient falls, or that could measure shear forces on the buttocks of a nursing home patient to risk-stratify for decubitus ulcers. Though not wearable, smart medication bottle caps are available to track patient compliance. Studies have shown that the use of such caps improves medication compliance in certain populations (10).
Provider Decision Support
Computerized decision support, the great promise of medical informatics, transforms the chart metaphor. The data in paper charts are static, waiting for providers to seek them out. Once the data are made machine readable, the computer can set them in motion to support the provider at the point of care. Common styles of support are assisting providers in interpreting data and in planning care. Enhanced data display is one simple example of assisting data interpretation; laboratory blood values can be displayed in tabular or graphical format with abnormal ranges. Automatic clinical warnings are another example of decision support. Warnings, or alerts, are usually implemented through a simple, compact algorithm (a good example is a drug–drug interaction alert). These warnings can be effective but are often overused, because simple algorithms tend to be very sensitive but not very specific. When useful alerts are drowned out by nuisance alerts—a low signal-to-noise ratio—studies show that providers tend to ignore all alerts (11).
Examples of support for formulating care plans include order configuring systems, data-driven reminders, and computerized clinical guidelines. Order configuring systems attempt to speed up common ordering scenarios using preconstructed options, and are most often seen in medication ordering. They can vary in sophistication from a quick order (e.g., a common dose, route, and schedule for a common drug), to a precoordinated set of interrelated orders, to an actual dialog that navigates a complicated decision tree based on real-time clinician input. Studies have shown clinician adoption of order-configuring systems remains tentative; quick orders are the most palatable, while complicated order sets are the least (12,13). Yet order sets have been shown to be highly effective in many clinical scenarios (14,15).
Reminders, data-driven flags that are raised to focus attention on clinical events that should follow a schedule, may be as simple as a static calendar flag (e.g., influenza vaccination every October 1st) or may draw on a small constellation of clinical data and dates (e.g., if a patient is female and older than 50 years and has not had a mammogram, then schedule one at next visit). Surprisingly, there is evidence that computerized reminders may lose efficacy or fail to alter outcomes over time (16,17).
Guidelines are a more complicated type of decision support. They seek to standardize care and improve quality while restraining costs. If a given guideline can be written in a machine-interpretable form, then the computer can assist in compliance (e.g., by anticipating orders or by cross-checking collection of key data elements). Efforts to establish standards for exchanging machine-readable guidelines are under way, as is research into whether such formats are expressive enough to adequately capture guideline logic (18,19). It is beyond the scope of this article to review the controversy surrounding clinical guidelines generally, but geriatrics specifically has enjoyed a mixed success. Acute and chronic pain management has been well expressed in guideline format (20), while antibiotic treatment—specifically the treatment of nursing home-acquired pneumonia—remains controversial (21,22).
One especially noteworthy development in computerizing guidelines is the style of explicit decision support exemplified by the Acute Respiratory Distress Syndrome Network protocol (15). Intensive care unit (ICU) monitoring equipment as well as EMR data are monitored by computer, and real-time ventilation adjustments are suggested (along with their rationale) to providers. Although designed for use by ICU specialists, in practice mostly nurses and therapists review this output. Providers accept or reject the automated recommendations, and care continues. The collective experience at several ICUs across the country is that well over 90% of the recommendations suggested by the computer are followed as given.
FUTURE HISTORY
In the near future, the technology and trends mentioned above will be extended and integrated to enhance patient care. The nexus of this integration will occur in the EMR. Over the next few years, functionality will increasingly be patched onto our present EMR systems. These patches will improve information flows among providers and patients and will help providers make better decisions. However, the limits of medical informatics will be constrained by the metaphor of the patient chart. As described above, this metaphor has the important characteristics of disintegration of information into the various chart sections of notes, labs, and orders. The metaphor also implies that plans of care be organized around progress notes that are long runs of free text—instead of bits of coded data or shorter, structured runs of free text.
The next phase in medical informatics will be facilitated by the abandonment of the patient chart metaphor and the adoption of an industrial process metaphor for patient care. The relevant characteristics of this metaphor are tolerances for outcomes and goal-oriented task management to achieve care within those tolerances. The outcomes relate to patient problems, but the metaphor is not exactly that of the problem-oriented EMR. Outcomes may be organized along lines of patient problems and health maintenance. However, explicit tolerances or goals are assigned to each patient in accordance with available, pertinent evidence from the literature or the patient's own history. Tasks are then mapped out, with responsibilities assigned to bring the patient's problems within tolerances.
Consider the example of a man with hypertension. The EMR has preassigned tolerances for blood pressure to every patient. After collecting blood pressure readings from the patient at home and at the clinic, the EMR will diagnose hypertension. The EMR will then suggest, in light of the patient's other problems and history, an antihypertensive such as hydrochlorothiazide. The EMR will then assign tasks to the patient to provide blood pressure readings to assess the efficacy of the drug so that the patient's systolic blood pressure will fall within the limits of 100 mmHg and 135 mmHg. The EMR will also explicitly recommend monitoring a serum potassium and creatinine in 7 to 10 days to make sure that the patient does not develop renal failure or hypokalemia.
In contrast to manufacturing, the patient will not be discarded if tolerances are exceeded. However, similar to industrial process control, the EMR will apply statistical techniques to trigger inquiries as to why the treatment is not working. Once the patient's blood pressure had been brought under control, these techniques will help identify assignable variation in blood pressure measurements. Data points with assignable variation are those that lie outside of statistical ranges such as 99% confidence intervals. Assignable variation is defined as variation in patient response for which an explanation is likely to be found (23). When data points are identified outside tolerances, the EMR can use the Internet or handheld input devices to query the patient about salt intake. The EMR may also query the medication bottle to see if pills have been dispensed in a timely manner.
Of course, the EMR and its sophisticated, explicit decision support will not be able to automatically resolve all problems; there is still ample room for the art of medicine. Providers will still be trained to recognize inappropriate treatment suggestions made by the EMR. Also there will not be evidence for many treatments, and the particular social and medical conditions of the patient will require human judgment. Patients will also still require personal contact with providers, as human interaction is vital to patient care. Finally, physicians need not worry that the EMR will completely replace them. Midlevel providers will take over care for 80% to 90% of patient problems, but the experience and training of physicians will be required for complicated patients or problems that do not fit the patterns recognized by the EMR.
Conclusion
The abandonment of the paper chart metaphor for the EMR, and the adoption of an industry processes metaphor, will facilitate improved integration and synthesis of information for providers and patients. It will also allow for more seamless integration of sophisticated decision support systems to help providers manage the nearly exponentially increasing evidence for the practice of medicine. Care will increasingly move to the patient's home with the patient having a greater role in managing his or her care. Midlevel providers will also have a more important role, as the EMR will allow them to efficiently handle routine patient decisions.
SIDEBAR: THE FUTURE EMR IN ACTION
Mr. Jones, an 80-year old man, moves to a new city to live closer to his family because he can't get around as well as he used to. His daughter discovers that he coughs at night and gets short of breath when he walks across the room. She logs on to the website sponsored by her integrated delivery health system. The system asks her some questions about her father's condition and suggests that he might have a heart condition and should be evaluated soon. Mr. Jones agrees to be seen, and the system takes the daughter to another web page where she helps her father enter-in his health history. In addition to the standard information, the history interface asks questions that directly bear on the patient's conditions. The system concludes that the patient should be seen within 1 week, and it makes an appointment with a nurse practitioner 2 days later. The system also tentatively schedules stress cardiac testing in 5 days.
At the appointment, the nurse practitioner asks some additional questions and performs a physical exam recommended by the EMR. In addition to the recommended exam, she notes several other findings including a probable basal cell carcinoma. She takes a picture of the suspicious lesion, which is also entered into the EMR and will be evaluated the next day by a dermatologist. The EMR reiterates its recommendation, and the nurse practitioner agrees that the patient should have stress cardiac testing at the previously arranged time. In the meantime, the nurse practitioner agrees with the EMR that the patient should stop the lisinopril, which may have caused the cough, and take atorvastatin, valsartan, hydrochlorthiazide, aspirin, sprinoloactone, and nitroglycerine sublingual tablets. The patient is also scheduled for blood laboratory tests in 7 days from the appointment. The cardiac stress test reveals ischemic heart disease, and the EMR recommends and the nurse practitioner agrees with the addition of low-dose metoprolol and a gentle program of physical therapy.
Because of the patient's mild cognitive impairment confirmed by the nurse practitioner, the dramatic change in the patient's medication regimen, and the patient's relatively sick status, the EMR recommends that the patient be sent home with a special monitoring kit. This kit has a base station that plugs into the phone jack. It communicates via radio waves with a module for monitoring vials that that patient wears around his chest. Another module monitors the patient's medication use. The first day, the medication monitor sends an alarm that the patient is not taking his medication correctly. As previously agreed, his daughter receives this alarm via an automatically generated phone call. She is able to help her father take the medications correctly.
Several days later, the patient feels better and decides not to get his blood laboratory tests. That evening, the EMR sends another message to Mr. Jones' daughter that he has missed the appointment. She brings him in the next day and his values are fine.
The patient weighs himself daily on a scale that communicates his weights through the base station in the patient's home to the EMR. The patient fails to meet his goals for weight loss. The EMR calls the patient and his daughter and, according to an evaluated algorithm, instructs the patient to double his diuretics. The EMR also queries the patient directly via the hand-held communication unit at the patient's home. With a series of yes/no questions, the unit asks the patient whether he has been compliant with his low-salt diet. The patient has not and the system makes a recommendation and, as previously agreed, sends his daughter an e-mail with a diet plan tailored for the patient so the daughter can follow up.
Meanwhile, the motion and cardiac sensors that are in the patient's chest monitoring pack have been continuously analyzed, and the system confirms that the patient is not complying with his physical therapy program. As the patient previously agreed, a physical therapist calls the patient to determine why the patient is having trouble complying with the prescribed exercises.
Two days later, the patient falls and breaks his hip. The cardiac and motion sensors that the patient is wearing communicate with the EMR, and the system suggests that a trip is the most likely explanation for the fall. Because the patient did not get up after the fall, the EMR calls the patient who does not respond. It then calls the patient's daughter. The daughter goes to the patient's home where she finds her father in pain, unable to get up, but otherwise OK. Based on this information, the EMR suggests sending a non-advanced-life support ambulance to bring the patient to the emergency room.
In the hospital, the patient develops complications, and by the time of discharge is severely weakened and mildly delirious. He is transferred to a skilled nursing facility for rehabilitation and care. The nursing home does not use the same EMR but is able to receive and use most of the pertinent information from the integrated delivery network's EMR.
The facility also uses a similar type of wearable monitor that they place on the patient at admission. This monitor reminds the aids to reposition the patient every few hours to prevent pressure ulcers. After the second day, the nursing home EMR has analyzed the motion data and found that Mr. Jones is at high risk for pressure ulcers. According to protocol, the patient is placed on an alternating pressure mattress. After a week, the motion and shear sensors communicate data to the EMR that indicate that that patient is no longer at high risk for pressure ulcers and the mattress is removed. The patient continues to make good progress and is eventually discharged back to home.
|
Address correspondence to Jonathan R. Nebeker, MD, MS, VA Salt Lake City Health Care System (182), 500 Foothill Drive, Salt Lake City, UT 84148. E-mail: jonathan.nebeker{at}hsc.utah.edu
Received April 21, 2003
Accepted April 22, 2003
References
| ||||||||||||||||||||||||
| HOME | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|---|
