17.
Incorporating Research
and Advances
in the Automated Patient Medical Record System
17.1 Project Context: Incorporating Research and
Advances in a Project
Creation of an automated patient medical record system goes beyond development of a software and hardware system. It requires research in the areas of new technology and medicine. It requires the development of national (or international) standards for medicine and new technology. It may require new laws.
See figure 17.1. In a research phase, after the research is done, there is an evaluation that determines whether to continue the research. A decision might be made to continue the research step, discontinue the research, or to proceed with incorporating and implementing the research into the project systems.
Incorporating the research into the project produces a new phase. Like for other phases, as part of the Business Analysis step of the phase, a determination should be made as to whether the overall design needs to be redone to incorporate the research ideas into the overall project.
17.2 Hardware Technology
Hardware technology is changing at a rapid pace, with increases in speed, storage capacity, portability, battery life, and other capabilities. New technology is continually being introduced, becoming less expensive as it matures. Some new technology, which existed in the past but was impractical, now is practical (e.g., biometrics).
17.2.1 Universal Patient Record Network and Telecommunications
Chapter 6 speculates on the main external network connections required for a universal patient health record. Figure 6.4 identifies what an initial network for an HMO and alliance healthcare organizations or for different regions of an HMO might look like. Figure 6.5 identifies what a network might look like after the system evolved into one handling unaffiliated healthcare organizations also.
In both cases, it is assumed that broadband connections [1] would be required for a healthcare organization’s connections to its own CPR repository and source document repository, probably using dedicated lines. For the extended network in figure 6.5 it is assumed that lower bandwidth connections, such as over an Internet, might suffice for connections to outside CPR repositories and source document repositories, as universal patient record information in outside healthcare organizations in the future could potentially be in many different possible locations, and thus dedicated network connections would not be feasible. In the future, with higher bandwidth public networks such as Sonet optical networks using WDM (wave length division multiplexing), dedicated lines may not be necessary [2].
In any case, network communications involving patient medical records are likely to require a larger amount of information than other types of network communications as
1. Some information may be voluminous, including digitized MRIs, x-rays and cat-scans.
2. Extra information is required in the communication to insure secure and reliable communications.
For example, for connections that require high security, encryption of information is needed, necessitating transmission of more information and thus higher bandwidth than non-encrypted data, and reliable communication requires two-way communication with the receiver of the information acknowledging correct receipt of the information.
In order to determine the network requirements for an automated patient medical record system, application analysts should work together with network analysts to identify the size and frequency of messages and the network bandwidth needed to handle these communications. Only through such an analysis can it be determined whether existing networks are adequate for the healthcare network required, or if network structures based upon future hardware, and software, advances are required.
Part of such a network likely will include the Internet or a similar such network. It should be noted that typical Internet communications are much slower than direct connections between computers. Figure 17.2(a) shows a possible connection over the Internet. A workstation might make a request for information from a Web server, establishing a path connecting these computers. The Web server sends back the requested information, through the same or another path. Usually, both the workstation and Web server are connected to the Internet through Internet service providers (ISPs) from whom they buy time. ISPs may buy time from other ISPs or be connected to computers or routers called network access points (NAPs) from whom the ISPs buy time. The NAPs interconnect, providing paths between ISPs.
Compare this Internet path with two computers that are directly connected. See figure 17.2(b). Such a bi-directional direct connection is referred to as a point-to-point connection, with equality of the capabilities of the computers it is referred to as peer-to-peer connection. Such a direct connection should be much faster than the Internet connection.
17.2.2 Computer Architecture
Reference [3] discusses strategic directions of computer architecture in the next 10 years. The author of the reference predicts that, like in the past 30 years, processor performance will continue to double every two to three years, memory densities will continue to quadruple every three years, and disk densities will quadruple every three years. This has an obvious good side to it, but also a less obvious bad side—computer systems quickly become obsolete, requiring constant upgrade.
Some specific possible changes in computer architecture that could have a very positive effect on the development of the automated patient medical record system, however, are the following:
· decreases in size and power consumption of computer systems, making for very lightweight, portable, but easy to use, computer systems that are mobile and can be carried around or easily picked up, and have a very long battery life (e.g., a thin, lightweight, clipboard size computerized equivalent to the patient chart)
· computer architectures that support multimedia processing, including speech, voice recognition, high density images, and video encode and decode and databases that support multimedia
· multiprocessor scalable computer (also known as parallel processors) with anywhere from 2 to 100 processors that can be used as servers and that can also support simultaneous multithreading. The multiprocessing capability could be used to execute programs for multiple workstations concurrently, while the multithreading capability could be used to execute the same program code (e.g., agents) concurrently. Multiprocessing capability is particularly important for very large data bases (sometimes called “data warehouses”) and by operational databases that can be used by a large number of users. Most multiprocessors fall into three categories: Symmetric Multiprocessors (SMP’s), clusters, and Massively Parallel Processors (MPP’s). This architectures are said to support scalability of large databases.
17.2.3 Storage I/O
Reference [4] discusses strategic directions of storage I/O for large-scale computing. The authors of the reference state that, in recent years, the amount of storage sold has been almost doubling each year; in the near future, it is expected to sustain an annual growth of about 60%.
The feasibility of CPR repositories and source document repositories is dependent upon the existence of high bandwidth network connections, but also upon high capacity storage media. Existing technologies include
· rigid or hard magnetic disks
· parallel disk arrays
· optical disks
· magnetic tape
· autochangers (racks of tape cartridges or optical disks that can be moved by robotics, the most common of which is a “jukebox” for optical disks).
Traditionally data outside mainframe computers has been stored on data servers on LANs. More and more data is being stored on its own network apart from the LAN, on “storage area networks”, SANs, using fiber channel technology. See reference [5].
A SAN links storage devices, including disks, disk arrays, and magnetic tape, to create a pool of storage that users can access directly. With SANs and fiber channels, data can be transmitted at higher speeds over greater distances than other methods. With the volume of medical information and need for speed of access to this information, SANs may be an appropriate way to pool storage devices.
17.2.4 “Portable” Computers and Wireless Networks
In order for a computer to be used at the “point of care”, it must not get in the way of communication between the caregiver and patient. Thus it must be portable and unobtrusive. Ideally, it would be connected to servers via a wireless network, so it would not be attached by wires and be carried from location to location. It must be light, especially so it could be carried around without strain. Such a very portable computer is especially necessity for home health care.
Many different types of “portable” computers currently exist. All have significant problems.
Laptop computers are increasingly becoming excellent substitutes for full-size PCs, with adequate screen sizes, full size keyboards, with 2 GigaByte+ disks and wireless network connections. But such computers are 6 pounds and over, still too heavy to carry around all day, and are not unobtrusive, with keyboard input.
Small computers are available, referred to as “palm-top” or “handheld” computers. But these are way too small to be useable for recording patient care and have a less powerful operating system than larger computers.
A type of “portable” computer that may be of use in the future is a tablet-size pen computer [6], perhaps the size of the current patient chart. With future technology, such a computer could possibly be made very lightweight; it could possibly be so lightweight, that it could be portable, or it could stay in the patient’s room, being treated identically to the current patient chart. (Note that some laptop computers convert into pen computers, sometimes removing keyboards or hard disks, making them lighter.)
The advantage and also possible liability of the pen computer is that there is either no keyboard, or the keyboard can be detached making the computer less obtrusive, and the computer can be used like a chart pad when being used in the care of a patient. This then requires that input be by pen, rather than be keyboard. Input then as described in section 12.2.4, more by selection than character input, is thus necessary to use a pen computer efficiently, with the pen computer being able to recognize the caregivers printing or handwriting.
Handwriting recognition software in pen computers is often not very good. In the future it may, hopefully, much improve.
Current versions of pen computers are made lightweight by not having a keyboard, and often not having a hard disk or even a floppy disk. For the current really lightweight pen computers, those without keyboards or off-line storage, everything input to the pen computer is saved in memory. Loss of power, for example, a battery going dead, could lose all information—perhaps, for such pen computers, data input could be immediately transmitted through a wireless network to the server, so the data would never be lost. For other pen computers having hard disks, information could be saved to the hard disk and not lost.
In any case, current battery technology is a limitation for all portable computers. For pen computers without permanent hard disks, battery technology needs to be improved so that loss of data could not occur. For pen computers with hard disks, battery technology needs to be improved so constant recharging becomes unnecessary, although if a pen computer is kept in a docking station when it is not in use, say in the inpatient’s room or in an outpatient exam room where it can be recharged, then this may not be a problem.
In all cases, to insure true portability, wireless networks should be used. In that regard, there is some concern about wireless networks that use radio waves interfering with medical instrumentation. Any possible interference could be avoided by use of infrared light communication rather than radio waves, but this requires line-of-sight transmission between the computer and wireless link. Infrared light cannot go through walls or turn corners.
For any portable computer, theft is another potential problem.
17.2.5 Diagnostic Imaging Storage and Monitors
Digitized computerized diagnostic image systems are sometimes called PACS (Picture Archiving and Communication System).
Diagnostic images must contain significant information to be of use. As mentioned earlier in this section, storage and transmission of such information requires very high capacity and quick storage and requires very high bandwidth network connections.
What is also needed is a monitor with a very high resolution, a high refresh rate, and which is probably quite large in physical size; each pixel (or dot) must be able to either display a great number of colors or a great number of gray scale values. Monitor resolutions are measured by a dot matrix array of points, expressed in terms of number of horizontal pixels (or dots) in length by the number of vertical pixels in height (e.g., 1048 x 1048). Refresh or scan rates must be high enough for the screen not to flicker (e.g., 60 Hz). Each pixel (or dot) may be a different color or gray scale value; stored for each pixel would be a binary number, or set of on and off switches, to identify the color or gray scale value (8 bits to represent colors or shades of gray).
Standard or typical values for VGA are 640 x 480 rasters with a refresh rate of 60 Hz supporting 16 different colors per pixel. Typical values for super VGA are 1024 x 768 rasters with 256 colors or 1280 x 1024 with 16 colors.
Although diagnostic images on a screen can be scrolled, full size display is much easier for a caregiver to use. Raster/bit requirements for full size display of diagnostic images for a representative PACS system, that by IBM [7], are the following:
· magnetic resonance tomography - 256 x 256 x 16
· computer tomography - 512 x 512 x 16
· digital subtraction angiography - 1024 x 1024 x 8
· digital luminescence radiography - 2048 x 2048 x 16
Reference [8] recommends workstations capable of displaying gray scale for 4 x 5 K for mammography. This is beyond the current capabilities of workstations.
There is a practical aspect of high-resolution monitors. High-resolution monitors cost lots of money, and further, such monitors may take up lots of space. Further, such monitors may not be able to also substitute for the lower resolution ones for other purposes; thus, space would be required both for the high resolution monitors and the lower resolution ones. Therefore due to cost and space considerations, it is unlikely that every physician ordering a diagnostic image would have such a monitor available to view the diagnostic images. Unless a copy of the diagnostic image is passed back to the ordering physician like it is in many cases now, the ordering caregiver would probably not be able to see the diagnostic image—he would only be able to read the interpretation. In such a case, a PACS system might not be of much benefit to most of a healthcare organization’s physicians.
There are other very significant barriers to PACS systems, both purely practical and psychological. On the practical side, regular analog methods using film are still capable of producing much more detailed images than digital ones. On the psychological side, radiologists and other caregivers must be comfortable with viewing diagnostic images on screens, rather than on film.
Digitization of diagnostic images is a necessary requirement for the up and coming field of computerized assistance in diagnosing diagnostic images. Only when diagnostic images are digitized is this feasible.
17.2.6 Security: Smart Cards, Biometrics, etc.
A smart card is a credit care size plastic card containing a microprocessor. A number of possible healthcare uses for smart cards were mentioned in this paper:
· to enable a caregiver to gain secure access to an automated healthcare system (e.g., a system used in Europe—see section 10.10)
· to enable a patient to gain secure access to his medical record
· to store a patient’s medical history and insurance information (unsuccessfully attempted before in the U. S—see section 10.10).
Many countries use smart cards for healthcare. Germany, with a form of socialized medicine, has issued a smart card to its citizens, mainly for health insurance reimbursement [9]. The Netherlands has a National Health Care containing a portable medical record, information on patient insurance, and information for medical research [10]. Austria has a Medcard used as a personal health card [10].
Smart cards suffer from the potential of loss or theft. Use of smart cards suffer from the potential of inaccurate information, especially medical information, and thus potential liability.
A far more secure method of providing security access for a caregiver to an automated patient medical record system and to a patient to see his medical record is biometrics, using a person’s physical characteristics to identify the person. Commercially available systems exist that identify caregivers by the following methods [11]:
· fingers
· hands
· eyes
· faces
· voices
· keystrokes
· signatures.
Smart card technology can also be combined with biometrics for extra security. Biometric data could be stored on a smart card and compared against that input from the caregiver user.
17.2.7 Caregiver Tracker
The author is unaware of whether or not such a device already exists or not, but here is a idea for a useful hardware device:
Each caregiver seeing patients within an outpatient clinic would wear a “tracker”. The tracker would be able to identify when a caregiver entered a room or left it. This tracker information could be fed to the automated patient medical record system which could use it to identify that a caregiver has seen the patient and what room he has seen the patient in, updating an “outpatient clinic room map” (see figure 12.21).
If there was a workstation in the room, the caregiver would be automatically logged on to the automated patient medical record system. And as he left he would automatically be logged off. This would insure that confidential patient information for the current patient was not available to a subsequent patient.
17.3 Software Technology
17.3.1 Mainframe Systems Versus Distributed Systems (Including the Intranet and Digital Libraries)
This book describes an automated patient medical record system that is distributed with interfaces with other clinical systems and interfaces through a TCP-IP, Internet type, network, to a patient registry, CPR and source document repositories.
IBM Corporation’s Customer Information Control System (CICS) is a transaction processing system on IBM mainframe systems and is IBM’s most popular mainframe system software system. The high use of mainframes using CICS systems, with continuing predictions that they will all be replaced by distributed systems, has shown that there are still many advantages of mainframe CICS systems over distributed systems:
· single log-on
· easy release of new and changed programs
· strong support for on-line transaction processing (OLTP) that splits processing into transactions, where a transaction (e.g., entry of an order) is a logical unit of work that usually involves a set of updates to a database that can either all be made permanent at once or all backed out if something goes wrong—Note that transactions are a necessary attribute for high volume systems [12,13,14]
· consistent security
· fail-safe backup of database information to any previous time
· no program failures due to program incompatibilities
· excellent use of multi-processing and multi-threading (multi-tasking), and also multi-programming (with respect to CICS and other mainframe programs running concurrently)
· much larger disk drive capacities
· spooling (queuing) of printouts on printers
· the ability to have central control of the entire system.
To be successful, distributed systems must also gain most of these capabilities.
Distributed systems in general are composed of many heterogeneous (i.e., different brand or model) computers with distributed databases and operating systems, with the computers connected by computer networks. Computers include servers that run the main software for multiple workstations (client computers), and may include mainframe computers. Each workstation serves a user with the workstation handling the GUI user interface for the particular user and passes information to the server to run the associated program on the server. (Distributed) databases are usually on the server or on connected mainframe (large scale) or minicomputer (medium scale) systems.
Distributed systems have the following advantages over CICS systems
· GUI user interfaces with windowing
· the ability to distribute processing among workstations, multiple server computers and legacy systems, and thus there is less of a chance of any component getting overloaded and causing poor response time
· independent control of independent systems.
As much as possible, the automated patient medical record system and universal patient record must possess all these positive attributes of CICS and distributed systems. Commercial products are available that provide CICS-like capabilities for heterogeneous distributed systems; some of these products are as follows:
· network communication management products (e.g., SNMP, Simple Network Management Protocol that is a protocol standard that has evolved as a support mechanism for TCP/IP based networks)
· remote control of remote servers or workstations
· single sign-on products with security control of multiple systems and all of the data
· OLTP (on-line transaction processing) products
· distributed database products, including data integrity features
· print service (to control and queue printouts)
· time service (to synchronize time among different systems)
· directory services to create a directory of files stored on multiple distributed computers
· software distribution products that allow software in remote computers to be installed, updated or removed without doing it manually
· change management software that keeps a history of changes so the system can be changed back to a previous version of the software, if necessary.
Together, these products support enterprise management of an organization’s distributed computer network. Together, these products, if available as a total package, are collectively referred to as an “enterprise management system (or package)”. In order for enterprise management system to work, the distributed network must follow standardized technology and structures for hardware, operating systems, data bases, fault tolerances, and network and communications transport, referred to as “open systems architecture” or simply “open architecture”. The Gartner Group has stated that of the enterprise management packages implemented so far, after 36 months only 30% have been fully implemented and an additional 10% are partially implemented [15].
(Note: A proposal for synchronizing the times on computers is to use the U.S. Atomic clock. The U.S. atomic clock, located outside Boulder, Colorado, sends out radion wave signals identifying the exact time. The U.S. atomic clock is operated by the U.S. Department of Commerce’s National Institute of Standards and Technology.)
A part of any automated patient medical record system, whether mainframe or distributed GUI system, should be the Internet, for medical reference and potentially for chart information. The Internet is the most successful “distributed system”. Its advantages are the following:
· composed of many independent systems
· communication occurs between systems, even when some network paths are down
· consistent network, naming and other standards that work despite systems being independent.
Although the Internet is probably the most successful distributed system, use of the Internet only for the patient medical record has the following problems:
· within the Internet, there is a lack of memory about a user and the previous screens he has seen (“statelessness”), thus making it difficult to develop secure, sophisticated, systems--see section 12.2.3
· the Internet is slow for an operational system
· there are potential major security risks with use of the Internet for the patient medical record; since the Internet is open to the general public, its use potentially opens up access to the patient medical record to the general public.
It should be noted that new standards are currently being developed for the Internet and TCP/IP-like (Internet-like) networks, internet protocol IP version 6 (IPv6) that replaces the current IP version 4 (IPv4) protocol [16]. IPv6 has the following advantages over IPv4:
· expanded addressing—allowing more locations to be uniquely identified within the network
· additional routing techniques—possibly allowing for somewhat higher speed networks and greater redundancy of connections
· greater support for transmission of multi-media
· additional security features.
A future network entity with similarities to the Internet is the “digital library”. From 1994 to 1998, the federal government through the NSF (National Science Foundation), DARPA (Defense Advanced Research Projects Agency), and NASA, funded the Digital Libraries Initiative (DLI) to do research on the study of a distributed library on-line [17,18]. Although the universal patient record is more homogeneous than such a digital library, it involves many of the same issues of organization, access, security and use of distributed information resources, and thus research on digital libraries may apply to the universal patient record.
17.3.2 Agents and Organizational Business Policies
The word “agent” has been used for many different abstract things, from a special way of writing programs to a program that specializes in a particular job function, similar to a very specialized human worker [19]. With my concept of “agent”, either of these definitions may indeed apply, but I have a far less abstract definition of an “agent”: a way of categorizing and separating out a set of independent code or tables implementing an organizational business policy, so that this policy could be easily implemented and as necessary later be changed by the people responsible for the business policy, rather than solely by computer people, without effecting the other code in the system. With this concept, a healthcare organization is not totally dependent upon programmer expertise to change business rules embedded within software systems.
Agents are best handled by distribution of agent code within the various computer systems making up the automated patient medical record system. Code or tables for regional agents should be in a regional computer. Code or tables for other agents might be in local computers. Distribution of this code might be best handled by use of CORBA, or other system software that enables distribution of this code.
Agents are a huge research area.
An agent can assist in incorporation of an organizational business policy in a system. An agent is only part of this incorporation. An organizational business policy should be documented and can implemented through a combination of code or tables (an agent), interfaces between systems, database information used by the agent, user interfaces to control input of the necessary information, and operational policies that users must follow to follow the business policy.
17.3.3 Security
Implementation of security within the automated patient medical record and universal patient record system must handle the many levels of security demanded by the system:
· federal regulations controlling access to a patient health record
· multiple state regulations controlling access to a patient health record, where the patient health record, in total, may exist in many different states of the U.S.
· potential international access to the patient health record
· caregiver access to patient clinical information based upon caregiver need to know
· patient access to his patient health record
· control of caregiver ordering
· control of caregiver ordering of controlled substances
· possible special security for some categories of patient clinical information, such as information related to psychiatry and genetics.
Because of these many levels of security demanded of the systems, security may be the “Achilles’ heal” that may make it difficult to fully implement the universal patient record.
17.3.4 Searching and Selecting from the Chart
Section 7.7.6 discusses information retrieval from an automated patient medical record and selection from an automated patient medical record from the point of view of the clinician who is using the automated medical record.
Within the last few years, significant research has been done by commercial companies on information retrieval of textual documents on the Internet via “search engines” and related software. Although this has direct applicability to retrieval of information for searching medical literature, and perhaps will have future applicability to anonymously searching through the universal patient record databases for patients for medical research, it may or may not have direct applicability to the information retrieval capabilities required by a clinician who will exam the patient. This is an area where significant additional research will be needed.
As discussed in section 7.7.6, creating a useful information retrieval (i.e., searching) method for clinicians to find information in the automated patient medical record and universal patient record is closely related to establishing a consistent medical vocabulary and thesaurus of medical terminology (see section 17.4.6). Clinicians must have a common medical vocabulary or the clinician who tries to find information will be searching for different information than the information that was entered by the previous clinicians.
17.3.5 Group Communication and Concurrency
Patient care is often a collaborative effort, involving a number of different caregivers all caring for the same patient. Because the automated patient medical record can be available to many different caregivers at the same time, whereas the paper chart is only available at a single location, the automated patient medical record has great potential for enhancing this group, collaborative, communication amongst caregivers in the care of a patient.
Group communication of caregivers may be
1. same time/same place: two or more caregivers, located at the same physical location, are looking at, and possibly adding to, the patient chart at the same time
2. same time/different place: the telemedicine situation where two or more caregivers, possibly located at different physical locations, are looking at, and possibly adding to, the patient chart at the same time.
3. different time/different place: the chart is accessed by a single caregiver at a time with communication possible via various means, including e-mail or messaging.
Case 3 is not a problem, as only a single caregiver is accessing the chart or other information at a time.
Cases 1 and 2 are nearly functionally equivalent and they are problematic. These cases must be handled by concurrency, and other, controls such as
· identification of “readers” who may, at a particular time, only view a part of the chart and “writers” who may update that part of the chart
· file sharing
· locking out of other users (possibly at the document or field level)
· back out procedures if things go wrong
· possible priority of assess of one caregiver over another
· cooperative editing systems, “multi-user systems where the actions of one user are instantaneously propagated to all the other participating users” [20]
· other controls.
In order to determine the required concurrency controls and other techniques for cases 1 and 2, the following questions must be answered: Can one caregiver be viewing a document in the chart, while another is adding to the document or changing it? If allowed or not, how does the system control this?
If one user is updating a document and another is viewing the document, will the viewer also see the update? If so, how can this be done?
Can two caregivers, or more, be adding to the same document in the chart at the same time? If so, how does the system control this? An example presented in this book was an emergency department triage document that included an H&P (see figures 12.47 and 12.50).
Can the two caregivers be changing the same fields at the same time? If so, what are the rules for this?
For case 2, same time/different place, where caregivers may be at vastly different geographic locations, the caregivers may work in two different facilities, each associated with a different distributed system. How is the chart concurrently accessed in this situation? What concurrency, and other, controls are required?
17.3.6 Voice Recognition
Voice recognition is currently practically used in medical care, especially in the radiology department for dictation of radiology results. Various companies have viable systems, including Kurzweil and IBM.
Considerations for picking voice recognition systems are the following:
· speaker dependent versus speaker independent systems: With speaker dependent systems, each user must go through sessions of inputting his voice so the system can adequately recognize his voice.
· discrete speech input versus continuous speech input: Discrete speech systems require the user to pause between words. Continuous speech systems do not.
· vocabulary supported by the system: Various available systems have vocabularies for internal medicine, radiology, orthopedic surgeons, emergency medicine, pathology, diagnostic imaging, cardiology.
· documents supported by the system: Some systems support particular kinds of documentation (e.g., SOAP notes, radiology reports).
· RAM memory required: because systems, especially the more sophisticated systems, require lots of RAM memory to run, the existing computer systems that can currently use them may be limited
· operating system required.
Various companies are in competition for the medical speech recognition market, including Dragon, Kurzweil, IBM and Kolvox. Any chosen system must have the capability of being integrated into the automated patient medical record system.
Voice recognition is also used in connection with call centers. Interactive voice recognition (IVR), also known as voice response units, or VRUs, is technology connected with a call center that interprets the caller’s voice and performs actions accordingly.
17.3.7 Handwriting Recognition
Handwriting recognition is important because, without it, pen computing is not feasible. I feel that pen computers are the only truly portable, unobtrusive, computers. Among all types of computers, a tablet pen computer, alone, can unobtrusively replace the current chart pad.
Handwriting recognition is part of some operating systems on PC’s also. For notebook computers and other PC’s using Microsoft Windows, Pen for Windows, is the extension generally used.
Considerations for handwriting recognition are the following:
· printed input versus cursive input: Can the user use handwriting or must he use printed input?
· controlled input versus writer dependent input: Must the user use a controlled input, inputting characters in a specific (“graffiti-type”)format, or will the system accept the user’s chosen character set?
· use of dictionary and word relationship techniques: Does the approach use a dictionary to improve accuracy? Does the approach user probabilities of certain words following others?
· training: Does the approach require training (i.e., the user to give controlled samples of input)?
· obsolescence: Is there a chance that the system will become obsolete? Many systems using handwriting recognition have disappeared.
· part of the computer’s operating system: So far, most handwriting recognition programs have been part of the operating system. If the handwriting recognition program is not part of the operating system, then consideration must be made about the capability of the program with specific applications and with the automated patient medical record system.
· RAM memory required: Because pen computers are usually limited in memory, but the quality of a handwriting recognition program might be dependent upon the size of the program, RAM memory requirements may be important.
A consortium of companies have developed a technical group UNIPEN [21] that has developed a benchmark for evaluation of handwriting software and that does research into better handwriting software.
17.4 Medical Area Research
17.4.1 Cases and Outcomes Research
Outcomes are the results of care given to a patient. The
purposes of treatment cases and chronic disease management cases are twofold,
1. continuity of care
2. evaluation of the outcomes of the medical care--with these outcomes based upon the patient, the medical condition of the patient, the treatments given and procedures given--so that different care, procedures and treatments for the same medication condition and similar patients can be compared and evaluated, and so that patient care can be improved.
How do you evaluate outcomes? How do you compare outcomes? These are research questions to be answered.
The outcomes of a treatment should be a final part of the treatment case and a part of chronic disease management case. In order to evaluate outcomes equitably for various providers, the condition, age, sex and other characteristics of the patient, available from the Patient’s Clinical Summary, should be taken into account.
One method to evaluate outcomes is a questionnaire given to the patient after a treatment, SF-36 or HSQ-12 (Health Status Questionnaire). Such questionnaires are currently available for a number of conditions, including knee replacement and hypertension. The questionnaire deals with gathering information on the patient’s perception of his quality of life after the medical care. The adequacy of such an outcomes evaluation is an open question, and methods of outcomes evaluations are likely to further evolve in the future.
Some outcomes can be more objective, such as measuring a patient’s blood pressure after a treatment to reduce the patient’s blood pressure. In fact, the patient may not even notice a difference in his health or quality of life, even if his blood pressure was successfully lowered.
The evaluation of general diagnostic testing for preventive healthcare based upon outcomes, such as sigmoidoscopy, can be evaluated on a statistical basis, comparing those who were given the test versus those who were not. For example, the rates of colorectal cancer and severity can be compared in those who had the sigmoids and those that did not.
The “evidence-based” evaluation [22] of medical tests, treatments and procedures based upon outcomes is crucial for the future improvement of medical care. Even older medical procedures are coming under scrutiny. Much more research is needed in this area.
Outcomes can be used to evaluate medical care in other ways also
· to evaluate physicians
· to determine any measures taken by a physician—such as how the physician explains a medication condition to a patient or whether the physician is the patient’s regular physician—that have a positive or negative effect upon the well-being of the patient and that increase or decrease a patient’s confidence in that physician.
How outcomes can be measured and their use will be continuing research areas.
Another research area is the proper use of treatment cases and chronic disease management cases in continuity of care. Users of treatment cases and chronic disease management cases require a new discipline in patient care. Sometimes a number of different caregivers may see the patient during the period of the treatment or management of the patient’s chronic disease. For purposes of continuity of care, the treatment, in general, should be continued according to the current treatment plan. Study should be done to determine how feasible this is and/or how to train physicians to have this discipline.
17.4.2 Determining Best Clinical Practice Guidelines
In association with public and private healthcare organizations, using treatment evaluation based upon outcomes, the Agency for Health Care Policy and Research (AHCPR), a government agency, has established clinical practice guidelines for treatment of 18 medical conditions based upon best scientific evidence [22]. See figure 5.4.
If an HMO requires clinical practice guidelines for other medical conditions, they must either establish their own from treatment cases, chronic disease management cases and outcomes, or get these guidelines from other organizations.
Research is required to determine how to best establish new clinical practice guidelines. Reference [23] states that patients’ needs and biological, social and environmental determinants should be factored in with the standard approach of determining clinical practice guidelines by best evidence as determined by randomized, controlled, clinical trials.
There is also indication that determination of best practices is also somewhat subjective. For example, in a study to determine whether clinical laboratory tests were appropriate or not, estimates of inappropriate laboratory use varied greatly (4.5% - 95%) [24].
When cost factors come into play, determination of a best practice may vary from year to year dependent upon changes in costs, even though the consideration of best medical practices stay the same. Because of an HMO’s and fee for service world’s differing view of costs, an HMO and the fee for service may make very different decisions on what is the best practice.
Once best practice guidelines are determined, they must be communicated to HMO caregivers. At least one healthcare organization has used the ASTM’s E1460 specification for “Defining and Sharing Modular Health Knowledge Bases (Arden Syntax for Medical Logic Modules)” as the specification for a clinical knowledge base used to implement clinical practice guidelines [25,26]. Further research needs to be done in the many different subject areas dealing with best clinical practice guidelines.
17.4.3 Predicting Diseases
Prediction of disease is perhaps similar to predicting the weather: Theoretically, the weather anywhere in the world could be predicted by analyzing the entire world—including its atmosphere, the history of the sun, the earth’s inhabitants, and the internals of the earth—and from this creating a detailed model of the earth and its future weather patterns throughout the world. Other more practical approaches to predicting the weather require much less information and are currently much more useful.
Likewise, there is a theoretic approach to disease prediction and more practical ones. The theoretic approach is identifying the total of genetic and environmental factors influencing disease in a person and predicting the person’s future diseases from this. This could potentially evolve into the most accurate approach to predicting disease in the far future, but is not practical at this time. Three other methods of disease prediction were proposed by this book in sections 5.4.1.7 and 8.6.
17.4.3.1 Theoretical Approach
As stated in section 5.4.1.7 from JAMA, “Individuals are born with a relatively fixed genetic status that in combination with environmental factors determines the propensities for a variety of disease states” [27]. With an automated patient medical record, there is the potential of having a much more complete social, family, environmental, and (with new information on the human genome) genetic history for a patient. Collectively, this information could potentially be used to identify the root causes of diseases. With this collective information, information for an individual patient could be used to identify diseases for which the patient has a propensity to develop.
At this point, this ability to predict diseases with any certainty for any individual is theoretical, but without a large-scale effort to collect and use this information, it will never be known if predicting diseases with a high degree of confidence is possible. For the information collected from a patient to be of use, it must be accurate, and thus it requires review by the patient to verify its accuracy. (Assuring accuracy of patient information is part of the reason for full disclosure to the patient discussed in the next section.)
17.4.3.2 Potential Practical Approaches
Sections 5.4.1.7 and 8.6 identify and discuss three other potential approaches to disease prediction: analytic disease prediction, disease progression analysis, and descriptive disease prediction.
Analytic disease prediction is predicting disease based upon previously identified risk factors for diseases (e.g., smoking as a predictor of lung cancer, anorexia nervosa as a predicator of osteoporosis) and upon protective factors. Analytic disease prediction requires clinical research to identify risk factors for, and protective factors against, different types of diseases. Since many of these risk factors are already known, then this form of disease prediction is already practical, although further research into risk factors will enhance this form of disease prediction including generation of probabilities for developing a disease based upon risk factors for the disease versus those for the total population or for a representative population (women over 50).
Disease progression analysis is recording a measurable indicator or indicators of a disease and from this predicting the future progression or manifestation of the disease in a patient. Disease progression analysis requires more clinical research in the progression of diseases before it becomes practical. Also, it requires caregivers to start using trend documents to track diseases in the automated patient medical record. Research that is needed includes determination of more measurable indicators of diseases, determination of how to work with multiple measurable indicators for a disease, determination of how to identify treatment decision points for each type of disease, and determination of feasible approaches to doing some very difficult measurements (e.g., how to measure the loss or degradation of cartilage in an injured knee over time).
Descriptive disease prediction is searching through patient medical records for patterns that are associated with the eventual realization of a disease. This could be done either by a computer or by a human looking for patterns within automated patient medical records, say through use of a data warehouse developed from patient medical records along with on-line analytical processing (OLAP). This form of disease prediction does not require clinical research (although the results of the disease prediction could suggest possible risk factors of diseases to later be studied by clinical research). However, it requires that there be an automated patient medical record for many individuals over an extended period of time to provide the information for this pattern matching.
17.4.4 Full Disclosure to the Patient
Proper evaluation of treatments and care is dependent upon the accuracy of information in the patient medical record. Accuracy of information in the patient medical record could be enhanced by patient review of his medical record after a visit. The possibility and practicality of patient review of chart information should be studied.
Consider a patient who has a consultation visit with a retinal specialist for previous blurriness in his left eye, which a previous retinal specialist diagnosed as macular degeneration. Figure 17.3 shows the consultation sheet that the patient later received.
From the patient’s recollection there are a number of inaccuracies in the consult:
1. Blurred vision 1 month ago was in the left eye, not the right.
2. The patient was never told to use an Amsler grid.
3. The consult never mentions that the patient was told that myopic degeneration usually behaves differently than age-related macular degeneration and is a different disease, and that having myopic degeneration does not mean the patient will get age-related macular degeneration.
Inaccuracy 1 could be catastrophic if there was an operation. Because the patient was given the consultation sheet, inaccuracy 2 is of no consequence, but would be of consequence if the patient has not received it. From the patient’s point of view, the most important statement to the patient, point 3, was left off the consultation sheet.
What this example shows is the following:
· Physicians may have inaccuracies in their reports, which the patient could catch (see 1 above).
· Physicians may think they have told the patient something when the physician hasn’t. Patient review of the visit information could convey to the patient this information.
· What a patient perceives as the most important statement to him may not be included in the consultation sheet. Such information could only be gathered by talking to the patient.
Thus disclosure of patient medical record information to a patient may protect against inaccuracies and physician misconceptions. Patient medical record information does not necessarily capture clinical information that the patient considers to be the most important.
Patient review of the patient medical record could also have negative consequences, including possible lawsuits and lesser patient regard for the physician. The usefulness of full disclosure of medical record information to a patient after a visit requires more research to weigh its benefits versus its possible disadvantages.
Now let me return to the point that that the medical record does not always capture what the patient perceives as the most important statements made to him in the interview by the physician. In the above case, there are no, at least immediate, medical consequences in the patient knowing that he has myopic degeneration rather than age-related macular degeneration, but consider another case where a physician’s statements, not captured in the chart, could have significant medical consequences.
In the early 1980s, a study was done where subjects were given two treatment alternatives [28]: