Knowledge, Standards and the Healthcare Crisis: Part 10

In my last post [click here for first in series], I discussed how Radical Transformers and Minimalists differ in their view of health knowledge needs, and how they differ in their motivation to change our healthcare system. Following is a brief summary.

On the one hand, Radical Transformers are motivated by a vision calling for fundamental changes in our current healthcare system. They have immense data needs since they focus on making knowledgeable diagnostic, preventive and treatment decisions that continually improve care outcomes and value. They use comprehensive, personalized information, which comes from extensive data about:

• The "whole-person" (mind-body-environment) over one's entire lifetime, including all key psychological, physiological, genetic, environmental and other factors that may affect diagnosis and treatment prescription
• Both sick-care (allopathic) and well-care (wellness & prevention) interventions[1]
• Both conventional and CAM (complementary & alternative medicine) approaches.[2]

Minimalists, on the other hand, are not motivated by a desire to change our healthcare system profoundly; instead, they are satisfied with slow incremental change that barely disturbs the status quo. They need much fewer data for making health and healthcare decisions because, unlike Radical Transformers, they focus on making diagnoses using information about relatively narrow set of signs & symptoms. They then prescribe a conventional, allopathic treatment regimen based on that diagnostic information, which does not require knowledge of the whole person, CAM, nor well-care options.

The issue of how much data, information and knowledge we need for making health-related decisions does not, however, end here. The amount of data diversity also affects the reliability (dependability) and validity (accuracy) of the information we use to make determinations about diagnoses, treatments, risk factors, and preventive care. Data that are more complete yield more valid and reliable information, resulting in better decisions and outcomes. I will discuss this and then begin answering the question: What has to happen for good data to become useful knowledge that leads to ever-better and more affordable care?

Reliability

Reliability is a statistical measure that indicates the degree to which information presents a trustworthy picture of a patient's condition and satisfaction (before, during, and after treatment). You need highly reliable data to have useful and dependable information with which to understand a person's problems and needs and make wise decisions.

A major reason for low reliability is the failure to use an adequate amount of data. In fact, the reliability of a health assessment "…increases when the number of [data] items …are increased and aggregated. It is a truism, but too often forgotten, that we cannot have either validity or utility without reliability [and] … the accuracy or reliability of measurement increases with the length of a test [italics added]. Since no single item is a perfect measure, adding items increases the chance that the test will elicit a more accurate sample and yield a better estimate of the person's [condition]."[3]

So, while you want to limit the amount of data collected in order to save time and effort, you have to be careful that the pool of data defining your data standard is not too limited! Consider this analogy: Pictures produced by most computer printers are comprised of patterns of tiny dots. Each dot is a piece of data that forms a part of the whole picture. In general, the greater the number of dots used to compose a picture, the more clarity and detail (resolution) it has, as shown below:
Figure A

Figure B

What makes B a clearer image than A is the number of dots. The more dots per square inch, the more dependable our interpretation of what we see because the image with greater detail provides more useful information. In the same way, the more data one can use to diagnose and treat a patient's condition, the better one's understanding of the person's problems and needs. Better decisions and outcomes are the likely result of using this comprehensive knowledge.

Validity
While reliability refers to the dependability of information, validity refers its accuracy. To yield valid information, an assessment must accurately measure all of the data required to support sound diagnostic and/or treatment-related decisions.

Note that there are many types of validity. One is diagnostic or discriminative validity , which measures the ability of an assessment instrument to diagnose a patient's disorder. A useful assessment instrument must classify patients into homogenous groupings using rigorous statistical analyses. That is, patients with similar characteristics in terms of their physical and psychological signs & symptoms, symptom etiologies (causes), functional impairment levels, life-stressors, demographics, etc. should be grouped into a single, precise, diagnostic category. Furthermore, patients within that diagnostic category should respond in a similar way to particular healthcare interventions.

For complex or multifaceted medical conditions, and for problems with a psychological component, a substantial amount of information is often necessary for making valid diagnoses. Consider, for example, evaluating depression. It is important to assess the nature, severity, and etiology (causes) of the depressive symptoms in light of a person's current life-events, past experiences, and personal demographics. This means using a vast data pool that measures:

• The intensity, frequency, duration, and cyclical time occurrences of the depression
• The etiology of the depression, including family history, current psychosocial and biomedical problems, medication side-effects, and psychoactive substance abuse
• The nature and degree of dysfunctional cognition associated with the depression such as thoughts of helplessness, hopelessness, suicidal ideation, self-deprecation, and existential/spiritual dilemmas, as well as cognitive slowing, rigidity, and focusing problems
• The nature and degree of concomitant (co-occuring) physiological symptoms such as lethargy versus agitation, changes in sleeping and eating patterns, and physical complaints
• The nature and degree of behavioral disruptions such as social alienation versus clinging dependence, and occupation or education dysfunction
• The nature and degree of coexisting emotional problems such as anger toward self, anxiety, guilt, and shame
• Demographics such as age, sex, ethnicity, and socioeconomic status.

After diagnosing a patient's problems, a particular treatment regimen is determined. Helping decide what treatments work best for a particular patient calls on another form of validity, i.e., predictive validity . This form of validity measures the ability to predict the specific treatments and levels of care that will produce the best outcomes for each type of patient. High predictive validity is difficult to achieve and, as is the case with diagnostic validity, it requires a pool of data comprehensive enough to relate to all patient populations and treatment modalities.

Another example of the need for comprehensive data comes from a recent article about the value of "meta-analysis," i.e., analyzing data combined from multiple clinical trials as a strategy for monitoring and assuring medication safety.[4] It reports how a researcher discovered a dangerous public health threat after stumbling upon data about Avandia, a medication for Type 2 diabetes, which may increase the risk of heart attacks. The report not only generated concerns about of Avandia's safety, but also resulted in considerable controversy about the validity of the conclusions from clinical trial about the risk-to-benefit tradeoffs.

By combining the data from multiple studies, meta-analysis is able to use more comprehensive information than typically available from a single study. This information comes from data about the types of patients enrolled in clinical trials, including demographic characteristics, disease severity, treatment regimens, and use of concomitant medications, among other factors.

The article concluded with a discussion of how the current day data standardization process, which aims to form consensus among multiple stakeholders, often sacrifices rich data "granularity" (i.e., diverse data containing fine details). This loss of detailed nuances can mean the loss of essential information required for making good decisions about the problems and needs of a particular patient in a particular situation.

A third example making the case for data comprehensiveness is that longitudinal data, which may span decades, reveals health problem trends (e.g., prostate cancer[5]) based on changes over time that cannot be determined accurately with an occasional data “snap-shot.”

And a final example, as presented in my previous post, is the need for adequate data variety to a valid evaluation of the mind-body connection. This includes factors such as medication side-effects that cause cognitive, emotional or behavioral problems; stress-related disorders that cause or magnify physical symptoms; and medical illnesses that present as a psychological condition.

Now that I've discussed how data comprehensiveness relates to the reliability and validity of information we use to make diagnostic and treatment decisions, let's return to the question: What has to happen for good data to become useful knowledge that leads to ever-better and more affordable care?

Good Data - Useful Knowledge

While a large, comprehensive pool of data is often essential for making good decisions about one's health and healthcare needs, it is important to make the data pool as concise and useful as possible by eliminating poor, unnecessary and redundant data using scientifically rigorous procedures such as reliability and validity analyses. Existing data pools ought to be continually updated and revised based on these analyses.

Accomplishing this requires, in part, flexible, dynamic information systems that evolve continually to (a) accommodate changes in a patient's condition over time, and (b) adjust to changing data standards by which health problems, treatments and outcomes are assessed.

I will continue this discussion in my next post.

References:

[1] Beller, S. and Sabatini, S. (2007). Integration of Sick-Care with Well-Care.
[2] National Institute of Health's National Center for Complementary and Alternative Medicine (NCCAM)
[3] Walter Mischel (1979). Distinguished Scientist Award Address. American Psychologist: 34(9); 742.
[4] Grasela, T.H. (2007). Data Standardization - Square Pegs in Round Holes?
[5] PSA Trends Predict Aggressive Prostate Cancer

Knowledge, Standards and the Healthcare Crisis: Part 9

In the previous eight posts [click here for first in series], I've discussed key issues concerning the healthcare crisis, data and technology standards, use of data-information-knowledge, and technological solutions. In my last post, I then posed these two questions:

1. How can we know if the data being collected are complete, appropriately complex, comprehendible, relevant and useful?
2. What has to happen for good data to become useful knowledge that leads to ever-better and more affordable care?

I will now begin to answer the first.

Answering this question requires that we clearly know what do we want to do with these data. That is, we have to determine our goals and objectives for using the data. I suggest that there are two general points of view: Use the data to promote profound and rapid incremental change vs. focusing on slow, minimalist change.

Some people and institutions want to use the data to generate information and knowledge for the radical transformation of our healthcare system through continuous improvement in care effectiveness and efficiency. They a have very different way of answering the first question than those who are content with the status quo or seek minimal change. So, let's compare and contrast the "Radical Transformers" from the "Minimalists."

Radical Transformers

The radical transformers want to start curing our healthcare crisis in meaningful ways by making profound changes now. [1] They demand data that ultimately helps consumers and healthcare providers to make valid and reliable decisions, and to reward individuals who implement those decisions in ways that result in higher quality, lower cost (i.e., high-value) outcomes. They want to know about the well-care interventions that help prevent illness and serious psychological distress, optimize well-being and quality of life, and avoid complications. They also want to know about the sick-care interventions that help patients recover more quickly, with fewer risks and side effects, and remain healthier longer. And they want to know how best to integrate sick-care with well-care. [2]

In other words, they want data that generate sufficient amounts of easily accessible, highly useful information. And they want this information to support the growth of an "evergreen" (continually growing and evolving) knowledge base of valid, reliable and relevant evidence-based guidelines, which are personalized to each patient/consumer's particular health and healthcare needs. This knowledge should enable consumers and providers to have a deep and complete understanding of the most efficient and effective ways to:

• Assess each person's physiological and psychological problems and risks
• Select and implement well-care and sick-care interventions best suited for that person.

This means continually making the correct risk assessment, preventive, diagnostic, and treatment decisions for each person.

The data they need to accomplish these admirable goals are extensive in both their "depth and breadth." That is, they need data from many individuals and a wide diversity of different types of data. Such data include details about:

• Initial signs (vital signs, professional observations, lab test results, diagnostic evaluations, etc.); symptoms (physical and psychological problems experienced and reported by patients); and diagnoses
• Changes in the signs and symptoms following care delivery, including changes in the degree of functionality, mobility, pain and discomfort, emotional distress, overall quality of life, etc.
• Medication side-effects, errors and omissions, mortality rates, etc.
• The specific care rendered, including prescriptions, procedures, therapies, lifestyle change recommendations, supplements used, etc.
• Any evidence-based guidelines used, including aspects of the guidelines that were not followed ("at variance") and why they weren't implemented
• Patient/consumer compliance (adherence) to the recommendations
• Patient satisfaction
• Cost (i.e., administrative/claims data).

In addition, this comprehensive data should enable understanding of important interactions (including the mind-body connection [3]), trends, exceptions, cause & effect relationships, etc.

Of course, the amount of data required is greater for more serious, complex and chronic problems than for simple, short-term problems (such a sprained ankle).

Anyway, these people understand the complexities of the human mind and body [4] and are willing to do what it takes to build a healthcare system of the future, including reforming current economic models [5] and redirecting competition. [6]

Minimalists

Unlike the Radical Transformers, Minimalists aren't motivated to fix our healthcare system in profound ways. They reject the claim that we need to collect and analyze more comprehensive data. Instead of being driven to build a healthcare system of the future, Minimalists tend to:

• Focus on gaining financially from the economic and competition models that plague the current healthcare system
• Perceive data collection as an onerous and expensive task to be avoided
• Fear that accountability and transparency will make them look bad
• Deceive themselves into believing that there is no knowledge gap problem [7]
• Reject or minimize the importance of the mind-body connection
• Be closed to complementary and alternative medicines (CAM)
• View guidelines as an infringement on their professional judgment.

An example of the Minimalist mind-set is today's P4P [8] programs that reward providers for implementing certain procedures for patients with particular diagnoses. The problem is that measuring performance by compliance to a handful of recommended processes does not necessarily improve outcomes. [9] Nevertheless, Minimalists continue to resist collection of comprehensive clinical data by claiming, for example, that it is simply too cumbersome a task. [10]

Minimalists also contend that we don't need comprehensive clinical data for evaluating outcomes since we could simply rely on administrative (claims) data that are routinely collected when submitting insurance claims. But this is clearly not the case for many reasons. [11]

And Minimalists tend to look for simple cause & effect relationships and easy explanations when trying to understand health problems, which would justify their minimal data requirements. But the quest for such simplicity often breeds ignorance, self-deception and faulty conclusions, and inhibits the healthcare community from knowing the best ways to prevent health problems and treat them cost-effectively for each patient/consumer.

To exemplify this issue of complexity, take genetic research. According to a recent NY Times article: "To their surprise, researchers found that the human genome might not be a 'tidy collection of independent genes' after all, with each sequence of DNA linked to a single function, such as a predisposition to diabetes or hearth disease. Instead, genes appear to operate in a complex network, and interact and overlap with one another and with other components in ways not yet fully understood." [12]

And what about the complexities of the mind-body connection? Here are two telling graphics:





As reflected in the images above, "psychosomatic disorders" add greatly to our country's healthcare costs. According to Thomas Pautler, M.D., a physician and lecturer specializing in psychosomatic disorders:

If we define psychosomatic illnesses as those involving a disturbance of physiology related in some way to situational conditions but without actual permanent end-organ damage (for example, migraines, functional bowel disease, and types of chronic pain), then we may account for as many as 25% of all outpatient visits. If we expand our definition of psychosomatic illness to include conditions such as hypertension, peptic-ulcer disease, hyperthyroidism, asthma, and chronic skin disorders where actual pathological changes are apparent as well as significant psychological factors, we can easily expand our ambulatory care percentage to the 50% range. Lastly, if we include serious physiological disorders such as disturbances in autoimmunity and the tendency for these disorders to appear or flare up with significant life changes and stress, we may continue to widen the magnitude of the psychosomatic problem . . . Even if we limit ourselves to the conditions that are purely psychosomatic - without demonstrable permanent end-organ changes - this 25% of illness may occupy a full 50% of the clinician's time in their management."[13]

An enormous amount of clinical data and related information is required before mental health disorders can be precisely diagnosed, appropriate treatments can be empirically determined, and interventions can be delivered with maximum efficiency and efficacy. The necessary data and information are lacking, however, partly due to the complicated and multifaceted nature of psychological problems. Mental disorders are extremely complex because every person and every disorder have their own unique set of symptoms and levels of dysfunction. [14] Furthermore, there are thousands of psychological and psychobiological symptoms, each of which can be associated with many types of functional impairments. Thus, certain mental disorders exhibit severe symptoms that are manifested in every aspect of a patient's life. They affect a patient's physical, behavioral, cognitive, emotional, interpersonal, and occupational functioning. Other mental disorders have fewer or less severe symptoms that cause less dysfunction. Nevertheless, all psychological difficulties are painful and create some degree of behavioral disruption, loss of productivity, somatic difficulties, and social conflicts. This complexity has made it very difficult to achieve a precise, detailed assessment of patients' symptoms and levels of dysfunction.

The data-gathering and analysis process is further strained by the wide range of possible underlying causes (i.e., etiologies) of each mental disorder symptom. [15] This situation has further complicated the information acquisition process. Assessing symptom etiology is difficult because psychological and psychobiological symptoms may be caused or exacerbated by many factors, including: current psychosocial stressors, childhood traumas, dysfunctional cognitive attributions and appraisals, erroneous beliefs, disturbing memories and mental images, psychological defenses, skill deficits, conditioned behaviors, neurotransmitter imbalances, and genetically predetermined temperament factors. In addition, a wide variety of biomedical illnesses and traumata, substance abuse, and medication side effects may present as or exacerbate patients' symptoms. Despite the complexities of symptom etiology, this information is often critical in making effective treatment decisions. For example, a patient experiencing depression due to an endocrine disorder should receive a different treatment than someone who is depressed due to an interpersonal problem. Understanding symptom etiology is also critical for effective prevention programs; one must know what causes mental health problems so they can be prevented. Thus, in addition to obtaining information about the nature and severity of patients' symptoms and levels of dysfunction, treatment-relevant information regarding symptom etiology must be objectively assessed.

In my next post, I will discuss how the validity (accuracy) and reliability (dependability) of information are tied directly to the completeness of the data upon which the information is built. I will also begin answering the second question: What has to happen for good data to become useful knowledge that leads to ever-better and more affordable care?

References:

[1] Curing Healthcare Blog: Do we need profound changes now?
[2] Wellness Wiki: Well-Care Sick-Care Integration
[3] Wellness Wiki: Biopsychosocial healthcare
[4] Curing Healthcare Blog: Making sense of the complexity and keeping perspective
[5] Wellness Wiki: Reforming Current Economic Models
[6] Wellness Wiki: Redirecting Competition
[7] Wellness Wiki: The Knowledge Gap
[8] Wellness Wiki: Pay for Performance
[9] The Health Care Blog: QUALITY: Performance measures only have a little of the answer
[10] Modern Healthcare Online: Quality reporting initiative may be too cumbersome
[11] Wellness Wiki: Using Claims Data
[12] Caruso, Denise. A Challenge to Gene Theory, a Tougher Look at Biotech. NY Times (7/1/07).
[13] Pautler, T. (1991). A Cost-Effective Mind-Body Approach to Psychosomatic Disorders. In Anchor. K. N. (Ed.), Handbook of Medical Psychotherapy: Cost-Effective Strategies in Mental Health. New York: Hogrefe & Huber.
[14] VandenBos, G. R. (1993). U.S. Mental Health Policy: Proactive Evolution in the Midst of Healthcare Reform. American Psychologist 48, 287.
[15] Coie, J. D., Watt, et al. (1993). The Science of Prevention: A Conceptual Framework and Some Directions for a National Research Program. American Psychologist, 48, 1013-1021.

Knowledge, Standards and the Healthcare Crisis: Part 8

I started this topic with the statement that our healthcare system needs radical transformation since:

• All patients "…are at risk for receiving poor health care, no matter where they live; why, where and from whom they seek care; or what their race, gender, or financial status is"[1]
• Healthcare is increasingly more expensive and less accessible[2], with more than 46 million uninsured in the U.S. from every age group and at every income level, 8 out of 10 being in working families[3]
• There is a "knowledge gap"-the healthcare community is drowning in oceans of information, yet doesn't know the best ways to prevent health problems and treat them cost-effectively.[4]

I then went on to explain how these daunting problems can be solved through creation of a knowledge-based healthcare system that drives continuous improvements in care safety, quality and affordability by enabling everyone to:

• Know the best ways to prevent illness, avoid complications of chronic diseases, and treat health problems (i.e., in the most effective and efficient manner)
• Use this knowledge to promote wellness, self-management, and recovery
• Participate in evolving this knowledge to make it ever-more useful and effective.

And then I discussed how data and technology standards are essential for obtaining, sharing and using knowledge effectively, and how such standards are a double-edge sword (i.e., there are serious problems with many standards in use today).

I concluded with a review of how a secure, economical, node-to-node architecture-with universal translation, composite reporting, and application integration-are essential components to the successful implementation of an intelligent and efficient quality improvement system.

I will now tie this all together as I present what might be considered the "holy grail" health-knowledge system.

Imagine patients and other healthcare consumers, along with clinicians and researchers, who collaborate to build an evergreen (i.e., continually growing and evolving) knowledge base of comprehensive information. This information comes from data obtained, which protect patients' privacy, via controlled clinical studies and real world outcomes research. These data are received around the globe every day and are analyzed on a regular basis in order to find associations between biological and psychological signs & symptoms, lab studies, diagnoses, genetic data, demographics, wellness interventions, sick care treatments, patient preferences, care costs, and clinical outcomes. The knowledge emerging from all this information is then used to create and validate evidence-based guidelines promoting high-value (i.e., safe & cost-effective) care alternatives that are matched to the particular needs of different consumers and providers. Understanding the relationships between health problems, care interventions and results enables both professionals and consumers to make wiser decisions that improve outcomes and control costs.

Underlying this information, knowledge and understanding are the data obtained from patients/consumers' PHAs (personal health applications) and providers' EHRs (electronic health records).

Clinicians and hospitals who collaborate in a patient's care, then share and view the patient data via next-generation CCRs (continuity of care records) tailored to each practitioner's particular needs. These CCRs go well beyond the ones being developed today by adding:

• Sophisticated clinical decision support capabilities that present warnings and alerts, as well as clinical guidelines (and pathways)
• Patient self-care information (i.e., "information therapy")
• Tools that track compliance to the guidelines, reasons for non-compliance (i.e., "variance"), and clinical & financial outcomes.

In addition, researchers (in universities, public health facilities, etc.) receive and study the de-identified health data on an ongoing basis-including details about patient health, medications, procedures and other interventions done, and the results of such care-which are used to develop evolving guidelines.

The data and guidelines are transmitted through networks of networks[5] using a simple, secure, low-cost node-based architecture and e-mail that require no build out of existing IT infrastructures. The nodes' "universal translation" function accommodates all data standards, as well as any non-standardized data sets and terminologies. Furthermore, since the nodes communicate asynchronously via publisher-subscriber process, and since they can present interactive reports through "desktop/standalone" applications (i.e., they are not limited to Internet browsers), critical information can therefore be accessed offline using rich, powerful tools. This means:

• There is no loss of data when a network connection drops out (i.e., unexpected disconnection), and there is no single point of failure to disrupt and entire network when a central server develops problems
• All the information can be accessed anywhere/anytime, even if there is no Internet or other network connections
• A great deal of data can be exchanged even when bandwidth is low and connectivity is intermittent (e.g., using dial-up)
• Each person controls their own data since they are stored locally (in their own computers) in encrypted files
• Total cost of ownership is minimized since there is no need to rely on expensive central servers and server administrators
• Performance is greatly increased when performing complex, intensive computations since all data processing is done quickly and easily using local computer resources, rather than waiting for a strained central server, or being restricted by the limitations of a browser
• You can integrate multiple desktop applications, which cannot be done securely using a browser. [6] [7]

These innovative technological solutions, however, are only part of the story. Other important issues include determining:

• What data to collect and exchange
• How to analyze, interpret and validate the data to generate useful information
• How to organize, access, share and discuss the information to emerge useful knowledge
• How to use the knowledge to improve care quality and control costs.

Dealing with these issues, I contend, requires valid & reliable data, information and knowledge (dik) that are comprehensive, complex and comprehendible.

• " Comprehensive dik is required for understanding the "big picture" clearly. This big picture reflects a person's physical and psychological risks, strengths, problems and preferences, as well as the evidence-based well care and sick care intervention options best suited to that individual.
• Complex dik provides crucial insights into care that are not possible using today's "minimum data set" standards. This higher level understanding comes from analyzing data that reveal such complexities as:
• Medication-related interactions (e.g., drug-drug, drug-supplement, drug-metabolism and drug-lab results interactions, as well as allergic reactions)
• Mind-body and mind-body-environment interactions[8] (e.g., the adverse affect of psychological stress and emotions on one's immune system, the affect of one's belief systems on one's health, etc.)
• Medication side-effects and biomedical conditions that present as psychological symptoms
• Trends, including changes in lab test results, functionality and signs & symptoms over a person's lifetime
• The correlation of treatments and outcomes for different patient populations and providers
• The reasons for not following particular recommended treatment processes and the results of such variance.
• Comprehendible dik:
• Are readily understandable (e.g., unambiguous, valid, reliable, relevant and useful)
• Maintain important nuances of meaning (e.g., uses the correct terminology standards)
• Do not overload people with irrelevancies or redundancies.

Two new questions now arise, which I will address in my next post; they are:

1. How can we know if the data being collected are complete, appropriately complex, comprehendible, relevant and useful?
2. What has to happen for good data to become useful knowledge that leads to ever-better and more affordable care?

References:

[1] The First National Report Card on Quality of Health Care in America by RAND Corp (2006)
[2] Health Care Coverage in America: Understanding the Issues and Proposed Solutions by The Alliance for Health Reform (March 2007)
[3] The Current Situation - WellnessWiki
[4] The Knowledge Gap - WellnessWiki
[5] Linking Providers Via Health Information Networks. Alliance for Health Reform. (Dec 2006).
[6] Is the Browser Singularly Capable of Everything? Software Development Times (June 15, 2007)
[7] New Google Tool Gets Offline Access in Gear. Eweek (June 11, 2007)

[8] Biopsychosocial Healthcare - WellnessWiki

Knowledge, Standards and the Healthcare Crisis: Part 7

In my last post, I began discussing I.T. solutions for dealing with the healthcare standards problem. I started by describing the benefits of a "publisher-subscriber node-to-node architecture with universal translation." I now continue with a discussion of two more important parts of the solution: "compositing reporting" and "application integration."

Composite Reporting

A node-to-node health information exchange architecture I've described has the added benefit of generating composite reports. These reports are comprised of information sent from multiple publisher nodes to a single subscriber node. The subscriber node takes all that information and combines it into a single integrated patient health profile report.

For example, let's say this report is a "continuity of care record" (CCR) [1] that a primary care physician (PCP) wants to use to help keep track of what's going on with the treatment a patient is receiving from several provider specialists. The PCP's node, which serves as the subscriber, would send a request for CCR data from all the patient's specialists. Upon receipt, the specialists' nodes, which serve as the publishers, retrieve the requested data from their different EHR (electronic health record) databases and send the data automatically to the PCP's node. The PCP's node then incorporates the data into a composite report tailored to the PCP's needs and preferences, and then presents it on screen for the PCP to view. The PCP's subscriber node could also be instructed to request data from the publisher node connected to the patient's PHR (personal health record) database and, upon receipt, include specificPHR data into the same CCR report as authorized by the patient.

Now, if the nodes also utilized universal translation (see the previous post), the data being sent by the subscriber nodes to the PCP's publisher node would be transformed appropriately, so the data always arrives in the right format (structure) and with the right terminologies (semantics). The result would be a useful, cohesive, and understandable CCR report containing information from disparate databases and data standards.

Application Integration

Still another way to make information useful, while avoiding data standards problems, is through application integration. This refers to an ideal way to present information residing in different software applications to support healthcare decisions.

Software vendors have been inventing different ways to display patient data that reside in disparate databases and are presented trough disparate applications.

• First, there was single sign-on (SSO) with which the user name and password are entered once. Multiple applications then display a patient's data through different windows on the computer screen. In its most basic form, SSO simply bypasses the login and user-validation screens of all but the first application accessed. The person still has to navigate through each application individually, however.
• Then came context management, which uses SSO, but goes a step further by synchronizing multiple applications. This enables a clinician to view a list all his/her patients, select one, and have multiple windows, corresponding to multiple applications, refresh automatically. This means the information presented through the windows of all the applications is related to the selected patient.
• Unified view takes context management one step further by presenting the data from different applications in a single window.
• Data integration involves synthesizing data across multiple applications and presents the data in a fully integrated manner. For example, when context management with a unified view is used to examine how a patient's medications affect his/her lab data, it simply presents one window with the lab results and another with the patient's active medications. That is, the data in each window are accessible only to the application that presents it; there is no interaction between the applications, so the data is static.
  
  A data integration system, however, recognizes the relationship between the data in the lab results window and the data in the meds window. This means the system can provide decision support by, for example, displaying a warning that lab result A may be affected by the patient's use of medication B. This goes beyond context management with a unified view since it applies analytic intelligence to the data being displayed by different applications. These data associations, however, are transient, i.e., the logic processes used for analytics and decision support must be run each time the same data are presented since there is no way to store the associations.

• Application integration goes beyond data integration since it retains the associations between data from multiple sources by creating and storing new forms of data that can be accessed repeatedly. Using the example above, it means that an application integration system would generate and store a new piece of data indicating that the problematic lab result A may be due to the patient's use of medication B. This warning information can be retrieved at any time in the future without having to startup the medication and lab results applications, and without having to analyze their data all over again. And the warning information can even be shared with other authorized persons, such as in a CCR report.[2]

In my next post, will conclude the topic of knowledge, standards and the healthcare crisis. I will draw upon the information I've presented to answer the questions posed in the first post of the series: What can we do to foster the widespread creation, use and evolution of healthcare knowledge without breaking the bank and without being hampered by the constraints imposed by data and technology standards?


References:
[1] American Academy of Family Physicians Center for Health Information Technology: ASTM Continuity of Care Record (2007)

[2] Holland, M. (Feb 2007). Improving Clinical Workflow with Unified Data Access and Management. IDC.

Knowledge, Standards and the Healthcare Crisis: Part 6

In the previous posts [click here for first in series], I described many of the problems facing the healthcare industry as it attempts to deal with data and technology standards. These daunting problems include high cost, complexity, difficulty accommodating changes, loss of meaning and nuance, trouble defining quality, inadequate measures, political influence, etc. I now discuss how an innovative approach to the use of health information technology would help solve these problems.

Solving the Problems with Healthcare Standards

From a technological perspective, what's needed to solve the problem with standards is a simple, low-cost, reliable, secure, hassle-free way to exchange and view structured & unstructured health information, anywhere and anytime, in a way that:

• Maintains the full meaning and nuance of the information being exchanged in order to maximize understanding and the information's usefulness, regardless of the data standards being used.


• Supports fluid connectivity between all IT systems, regardless of their technology standards.


• Gives all authorized consumers/patients, providers, suppliers, payers (insurers), and purchasers (employers and self-insured) the information they need, in the way they need it, to support decisions and guide actions.


• Supplies researchers with the information they need to evaluate clinical outcomes and care processes, so they can create, continually evolve, and widely disseminate evidence-based guidelines.

I contend that the best way to do this is through a secure node-to-node network architecture using template-driven software applications having "publisher-subscriber," data translation, and personalized reporting capabilities. Let me explain.

Why a Node-to-Node Architecture

In a node-to-node architecture, each node is a software application in a computer that sends and receives information from other nodes. This architecture supports "peer-to-peer" (P2P) networks in which each node stores its data files locally and shares them with other nodes without being controlled by a centralized server.[1] The telephone system and e-mail are good examples of node-to-node. Every phone and every computer are nodes. By picking up the phone and dial a number, or by typing in an e-mail address, you can communicate with whomever you want, and do it anytime and anywhere. Your call or e-mail is routed automatically to where you want it to go through a series of simple switches. This open network is quite different than a centralized system in which you must first sign on to a central server that determines who you are authorized to contact before sending them your message, i.e., all information must pass through a central authority that controls all communications. In addition, such centralized systems typically require the costly development and ongoing maintenance of a centralized patient record locator to know where to find patient data.

The following make the case for a node-to-node/peer-to-peer architecture for exchanging healthcare data:

• "The United States' National Health Information Network, or NHIN, will differ from the UK's project in a number of ways. Rather than having a single, closed network with a central database overseen by one government agency, the U.S. system will be decentralized, operating more like a peer-to-peer network, with records distributed across the system. Think Napster on steroids. …the NHIN will allow a doctor to quickly call up a patient's digital records from whatever databases they may reside in-at a hospital, at the family doctor's or dentist's office, at a clinical lab, wherever."[2]


• "After initial testing using a centralized patient index, [Massachusetts' MA-Share HIE determined that] the maintenance for that looked like it would be more than users would want to pay. So the exchange uses distributed peer-to-peer networking. The MA-Share exchange provides an appliance to let members push financial transactions, e-prescriptions, and clinical summaries-so a doctor can send a file to another doctor or provide prescription data to a pharmacy." [3]


• "To make significant gains in patient safety through the adoption of health IT, providers will need to adopt IT systems that can 'speak the same language' to each other. In computer terms, they should be 'interoperable.' But interoperability isn't enough. To communicate, different health IT systems must also be linked in some way. This is 'connectivity.' One model of connectivity, in a national health IT context, would be a non-proprietary 'network of networks.' …Several issues must be addressed if different health information systems are to communicate. …Some suggest that there should be one uniform national system with one central repository. This approach presents challenges: the sheer volume of data that would need to be handled, significant concerns about privacy and security threats, and likely disputes about governing and paying for a centralized system. Another option is a series of regional networks, as advocated by ONC [the Office of the National Coordinator of Health Information Technology, formally ONCHIT]. ONC's strategic frame- work suggests that a national network should be structured around regional health information organizations (RHIOs). RHIOs would store, organize and exchange patient health information within a defined geographic region, under local rather than national governance. These regional organizations would form a "network of networks" across the nation." [4]

Publisher-Scriber Communications Model

The nodes in these P2P networks employ a publisher-subscriber communications model in which a publisher node uses its communications software application to publish (send) information to one or more authorized subscriber (receiver) nodes. Once transmitted, the subscriber nodes use their subscriber applications to retrieve that information and present it as reports. In other words, the publisher-subscriber model uses an "application to application" transfer process in which each participating node uses a particular software application for exchanging (sending and receiving) information.

The publisher and subscriber applications support a particular operating system OS) and Internet connection using broadband or dial-up service. A node that uses an e-mail client (such as Microsoft Outlook on Windows OS) is one such example.

At one end of the connection, the publisher node must authorize the information transfer by authenticating that the subscriber node is allowed to receive the information. At the other end of the connection, each subscriber node must allow the publisher to deposit the information into a directory (i.e., a folder in computer's drive) as a file with a specific format (such as an MS Word, Excel, or "comma separated value" file).

Universal Translation

A node-to node architecture incorporating "universal translation" provides a means for modifying (transforming, translating) information as it passes between nodes, so that each subscriber node receives from a publisher node the right information, in the right format (structure), and with the right terminologies (semantics).

This is where data and technology standards are handled. That is, the universal translator makes the necessary transformations to the information sent by a publisher node, so different subscriber nodes can use that information to generate their particular reports and, if desired, to store the information received in the subscriber nodes' databases. It can accommodate any data standards and operate with systems using any technology standards.

Advantages and Benefits of the Node-to-Node Architecture

The advantages and benefits of this asynchronous, publisher-subscriber, node-to-node architecture are many, including the following:

• Is exceptionally flexible:
• Accommodates any data and technology standards, so everyone gets the information they need in the way they need it
• Allows anyone to communicate with anyone else in any way
• Can use multiple connectivity options, i.e., radio transmission, satellite transmission, wire transmission, wireless transmission.


• Has maximum reliability since it leverages the most reliable network in the world, i.e., the switched network (like the telephone system).


• Is inexpensive to deploy and operate because it doesn't require changes to existing I.T. infrastructures and keeps implementation costs low by eschewing additional equipment and system purchases.
• Is robust and resilient since there is no single point of failure; so, unlike centralized networks that are disrupted if a central server goes down, the node-to-node network is survivable in a disaster since it keeps going even if individual nodes are disabled.


• Makes scalability a non-issue, which means there's no need to purchase new equipment or redesign software as the network grows; this is unlike a centralized system in which there tends to be significant costs in time and money to meet the needs of a growing network.
• Is highly secure since there are no external database queries; firewalls are not crossed.

In my next post, I discuss other parts of the solution: Composite Reporting and Application Integration.

References:

[1] Wikipedia - Peer to Peer and WellnessWiki - Network Architectures (see Node Mesh Network)

[2] Charett, R.N. (2006). Dying for Data: A comprehensive system of electronic medical records promises to save lives and cut health care costs-but how do you build one? IEEE Spectrum Online (Oct 2006)

[3] Kolbasuk McGee, M. (May 28, 2007). Urgent Care. Informationweek.com

[4] Linking Providers Via Health Information Networks. Alliance for Health Reform. (Dec 2006).

Knowledge, Standards, and the Healthcare Crisis: Part 5

In the previous four posts [click here for first in series], I described the data and technology standards commonly used to enable the exchange of health information between disparate data sources. I also discussed why such information exchange is vital to the creation and use of knowledge leading to increased healthcare value. In addition, I mentioned several challenges to using standards effectively.

In this post, I delve into the problems faced by the healthcare industry when dealing with standards.

The Problems with Healthcare Standards

We confront one set of problems with data (terminology, care measurement and care process) standards, and another with technology (messaging) standards.

Problems with Terminology Data Standards

Problems associated with terminology standards are significant:

• According to William Hammond, professor emeritus of community and family medicine at Duke University, there's "been ongoing discussion about implementing health data standards harmonization and cooperation for 20 years, yet no one has defined all the standards needed to support a national health information network, and no one has identified what's missing." Just agreeing on medical terminology is a big issue. And, according to Michael Rozen, vice chairman of the IEEE-USA Medical Technology Policy Committee, "When you say 'gross profit,' everyone in finance knows what that means [but] in medicine, there are 126 ways to say 'high blood pressure.' "[1]


• While setting an arbitrary standard for health-related terms is a way to foster widespread communications between people from different regions, organizations and healthcare cultures/communities, there's also a downside to such standards, i.e., they lose information due to reduced "semantic precision and nuance." In other words, there's a good reason to have multiple ways of saying high blood pressure. For example, malignant hypertension refers to very high blood pressure with swelling of the optic nerve behind the eye, which is usually accompanied by other organ damage like heart failure, kidney failure, and hypertensive encephalopathy. Pregnancy-induced hypertension, on the other hand, is a pregnancy-induced form of high blood pressure (also called toxemia or preeclampsia). Referring to a patient's condition using the standard term "hypertension," while clearly conveying that the person has high blood pressure, looses these important details, which could very well affect treatment decisions and outcomes.


• Diagnostic code standards-including all versions of the ICD and DSM-have several serious limitations. These problems include the fact that (a) these standards are not detailed enough to describe the nuances of all diseases and conditions and (b) some diagnoses are not useful in making treatment decisions.[2] Since treatment selection is based (or should be based) on a patient's diagnosis, we need a diagnostic standards that have greater precision. This requirement is amplified with personalized care is, in which each patient's unique makeup (including genetics) and the mind-body connection are taken into account (not to mentions ones abilities and preferences).

Problems with Care Measurement and Process Data Standards

As I discussed in a previous post, care measurement and process standards relate to evaluating care quality and provider performance, and to establishing practice guidelines. Some of the problems associated with these standards, include the following:

• Achieving wide-ranging and meaningful quality standards requires many more years of dedicated effort by many people and substantial financial resources.[3]


• Standards should evolve continuously, changing as necessary to accommodate new knowledge. Unfortunately, it typically takes 17 years before clinical evidence is implemented in practice guidelines.[4] [5]


• Simply maintaining nation-wide data standards is a slow and costly process.


• And, as I discussed in an earlier post, there are many problems with practice guideline and quality measurement standards:


• It's difficult to determine when there is enough evidence supporting a practice guideline and there is no longer any need to spend time or money on its continuous evaluation.


• It's difficult to determine when a definition of quality is too narrow, which can happen, for example, when measuring quality based on cost or symptom reduction, without giving adequate consideration to prevention or the continuity of care.


• It's difficult to determine how best to measure quality when resources are scarce and optimal care for the community may require less than "the best" care for its individual members.


• It's difficult to determine how best to measure quality if outcomes are more strongly affected by patient compliance than by physician orders.


• It's difficult to determine if care quality is of poor when a provider follows the recommended practice guideline, but the patient is atypical and responds poorly.


• Using claims (administrative) data to measure care quality, as in often done today, is grossly inadequate.


• Assessing care quality using process data may not be valid since they do not necessarily reflect care outcomes.


• It's difficult to determine how to avoid political and ideological biases when determining what evidence to use as the basis for establishing the guidelines.


• Many areas of healthcare lack care process standards and useful quality measures. Different healthcare disciplines and specialties require different types of data to evaluate quality.

Problems with Technology Messaging Standards

The problems with standards aren't limited to data standards; they also plague technology messaging standards:

• When multiple information systems use the messaging standard to communicate, changing the standard cost huge sums as all the systems using them must be overhauled. A good real world example is the Year 2000 problem, where computer systems were built using a messaging standard that required only the last two digits of the year to be used when transmitting data containing dates. So, when 2000 rolled around, this data standard made it impossible to differentiate between years beginning with 19 and those beginning with 20 (i.e., 4/5/05 could be Apr 5, 1905 or 2005). This problem easily cost hundreds of billions of dollars to fix.


• The Healthcare Information Technology Standards Panel, which is setting technical standards for a nationwide record system, identified an initial set of 90 medical and technology standards, out of an original list of about 600. These standards specify such things as how lab reports are to be exchanged electronically and entered into a patient's electronic record, as well as how past lab results are to be requested. More than 190 organizations-representing consumers, providers, government agencies, and standards development organizations-participating in the panel. It's no wonder, therefore, that a consensus on medical standards is so difficult and fraught with politics as standards-setting involve intense negotiations and delicate compromises. And once such IT standards are set, software systems and databases must be designed to conform with those standards.[6]

Summary

While data and technology standards offer a way to handle information exchange challenges, they come with issues posing serious problems in terms of cost, effort, time, hassle, complexity, inefficiency, usability, reliability, information loss, political influence, etc.

In my next post, I will discuss ways to solve the daunting problems plaguing the use of healthcare standards.

References:
[1] Dying for Data: A comprehensive system of electronic medical records promises to save lives and cut health care costs—but how do you build one? IEEE Spectrum Online (Oct 2006)
[2] Current Diagnostic Codes are Inadequate – WellnessWiki
[3] U.S. Health Care Sector Moves Rapidly To Provide Consumer Information on Value. HHS (May 9, 2007)
[4] Balas, E. A., & Boren, S. A. (2000). Managing clinical knowledge for health care improvement. In J. Bemmel & A. T. McCray (Eds.), Yearbook of Medical Informatics (pp. 65-70). Stuttgart: Schattauer Verlagsgesellschaft mbH.
[5] Clancy, C. M., & Cronin, K. (2005). Evidence-based decision making: Global evidence, local decisions. Health Affairs, 24(1), 151-162.
[6]  Dying for Data: A comprehensive system of electronic medical records promises to save lives and cut health care costs—but how do you build one? IEEE Spectrum Online (Oct 2006)

Knowledge, Standards, and the Healthcare Crisis: Part 4

In the previous post, I discussed "care measurement and process standards," which are data standards dealing with diagnosing health problems, determining treatments, and assessing care quality and provider performance. [Click here for first post in series] In this post, I turn to technology standards, and to "messaging format" standards in particular. Whereas data standards focus on making information understandable and useful to humans, messaging format standards focus on enabling the exchange, or interoperability, of data and information (i.e., "transactions") across healthcare systems.

HL7 Messaging Standard

The Health Level Seven (HL7) Messaging Standard is the most recognized. It specifies the technical aspects of sending messages so that one software program can exchange information with another, so the information is "understood" by the receiving machine. This standard handles information related to order entry, scheduling, medical record and image management, patient administration, observation reporting, financial management, and patient care transactions.

For example, an HL7 formatted message delivering data about a patient's EKG would be written like this: "OBX2ST93000.1^VENTRICULAR RATE(EKG)91/MIN60-100". Here's what it means:

• OBX = The message is a report of an observation/result
• 2ST = The data value is a two character string
• 93000.1^VENTRICULAR RATE(EKG) = The code number and name of the EKG test
• 91/MIN = The observation value and units (beats/minute)
• 60-100 = The reference or normal range for this test. 

Note that this HL7 message standard just provides a message structure (syntax), i.e., the use of pipe symbols ("") to separate message elements, and the order in which the elements appear; none of the content (codes, terminologies, values) is defined by this HL7 standard. Also note that the next version of HL7 (version 3) will be tied to specific terminologies, thereby adding semantic capabilities that enable different data systems to communicate with each other.

HL7 Version 3 messages are XML documents, which use a very complex and verbose structure of "markup tags" to identify the data values. These tags are strings of characters surrounded by angle brackets, which are depicted in the figure below. The figure is a small section of an HL7 clinical document in XML, which includes the use of the SNOMED CT terminology standard in which the terms "Osteoarthritis", "finding site" and "right knee" are used to define the medical history note that the patient is "complaining of disabling osteoarthritis of the right knee."

The beauty of XML is that anything can be defined using the markup tags. A down side is that XML is very inefficient. For example, in the XML document above, it takes about 700 characters to record an observation that's only about 100 characters in length. Furthermore, such XML documents can be complex to write and difficult for humans to read. This concludes my description of standards used in healthcare. In the next post, I'll delve into the problems with today's standards and will the offer innovative strategies for solving those problems.

Knowledge, Standards, and the Healthcare Crisis: Part 3

In the previous post, I discussed "terminology standards," which deal with the meaning and use of words (terms). In this post, I continue with the discussion of data standards, focusing this time on standards for diagnosing health problems, determining treatments, and assessing care quality and provider performance. This all relates to care measurement and process standards. [Click here for first post in series]

Care Measurement and Process Standards

Care measurement and process standards focus on:

• Diagnosing health problems
• Selecting and delivering treatments
• Evaluating care performance and value.

Diagnosing Health Problems

Physiological (bodily) and psychological (mental-emotional-behavioral) measures are used to diagnose a patient’s health problems.

Physiological Measurement Standards

Physiological measurement standards include vital signs and lab test "reference ranges." For example, the standard measures for hypertension is systolic pressure consistently greater than 140 mm Hg, or diastolic pressure consistently 90 mm Hg or more and a standard measure for diabetes is fasting blood glucose level of 126 mg/dL or higher on two occasions. Genetic markers associated with illnesses may also be considered a type of biologic measurement standard. These standards not only help diagnose a patient's condition, but may also determine one's risk of developing a disease.

Psychological Measurement Standards

Probably the most common psychological measurement standard is the IQ test, which defines a score of 90-110 as being within the "normal" range of intelligence. There are also standardized tests that measure mental status (e.g., awareness, memory and other cognitive functions), as well as depression, anxiety, personality traits and other psychological factors.

Selecting and Delivering Treatments

The diagnostic measurement standards are useful if they help select a particular practice guideline identifying a particular treatment for a particular patient with a particular diagnosis. The guidelines provide recommendations for the prevention, treatment, and maintenance of many nontrivial illnesses, conditions, disorders and other healthcare problems. There are three thorny problems, however:

1. Today's diagnostic systems often fail to point to the best treatment options.[1]
2. Few guideline standards are specific enough to account for individual differences in patient with the same diagnosis. For example, a recent study found that a moderately high total cholesterol level is associated with higher survival in certain patients with heart failure.[2]
3. Constantly evaluating and revising guidelines based on new knowledge is very difficult. But if they do not continually evolve, the guidelines are just "a record of the past, and little more-they should have an expiration date."[3]

Evaluating Care Performance and Value

At least three standards are related to clinician performance and care value:

1. Process compliance standards
2. Clinical outcome standards
3. Care value standards.

Process Compliance Standards Process compliance standards measure provider's performance based on whether they followed prescribed guidelines reflecting preferred care processes. For example, typical Pay for Performance (P4P) programs reward providers who perform certain predefined procedures (processes), such as doing a Hemoglobin A1c test a certain number of times each year for patients with diabetes. These standards measure the degree of compliance to such established procedures.

Clinical Outcomes Standards

Outcomes standards define whether clinical goals are achieved for patients with particular conditions. For example, the Hemoglobin A1c test target goal for diabetic control of blood glucose is defined as less than 7.0%. Unlike process compliance standards, clinical outcomes do not focus on whether specific procedures were followed; instead, they measure the effectiveness of whatever treatments were delivered.

Care Value Standards

If our healthcare system was rational and guided by wisdom, a top priority of healthcare professionals and consumers would be:

• Gaining valid knowledge about healthy living, the causes and diagnosis of physical and mental health problems, and the highest value treatments.
• Understanding how to use this knowledge to maximize value by increasing the effectiveness and efficiency of care delivery and self-maintenance.
• Continuously evolving this knowledge and using it to improve care quality and lower costs continually.

So, what is care "value."

Care value can be measured by dividing the quality if that care by its cost, i.e., V = Q / C:

• Q (Quality) is defined as the degree to which care is delivered safely, effectively and equitably. The care may include conventional and alternative interventions for treating illness, as well as wellness intervention for prevention and health optimization. Quality can be measured based process compliance standards, clinical outcomes standards, or both.
• C (Cost) is defined as the degree to which the care is delivered efficiently and economically.
• V (Value), therefore, can be defined as cost-effectiveness ("bang for the buck").

If there is to be significant improvement in healthcare delivery, a useful and reliable quality standard must be established for every healthcare domain/discipline/field. Only then can care value be determined.

Potential Pitfalls of Care Quality Measurement

While costs can sometimes be tricky to calculate, measuring quality is the major challenge. The potential pitfalls of quality measurement are enormous! Consider the following:

• We have a long way to go. According to HHS Secretary Mike Leavitt, "Medical associations and others have begun the work of developing quality standards and cost measurement, but we have many years of work ahead of us to achieve the wide-ranging and meaningful quality standards we need."[4]
• No mater what quality measures are used, there are complex issues to be resolved, such as:
• At what point is there sufficient confidence in an evidence-based practice guideline that there is no longer any need to spend time or money on the continuous evaluation of its reliable and validity?
• When is a definition of quality too narrow, e.g., by focusing on cost or symptom reduction, but not considering prevention, recurrence, coordination and continuity of care, or the patient-physician relationship?
• How do you measure quality when resources are scarce and optimal care for the community may require less than "the best" care for its individual members (e.g., delegating office nurses to perform certain activities that physicians used to do)?
• What is the best way to measure quality if outcomes are more strongly affected by patient compliance than by physician orders? This may occur, for example, if certain providers have personalities that trigger greater patient compliance, and visa versa.
• Is it poor quality care if a provider follows the recommended practice guideline, but the patient is atypical and responds poorly? [5]
• Use of claims (administrative) data to measure care quality is grossly inadequate for many reasons.[6]
• Assessing care quality using process data may not be valid since they do not necessarily reflect care outcomes.[7]
• One thorny issue is how to avoid political and ideological biases when determining what evidence to use as the basis for establishing the guidelines. [8]
• Many areas of healthcare lack care process standards and/or quality measures. Different healthcare disciplines and specialties require different types of data to evaluate quality. For example, it's foolish to measure the quality of mental healthcare services with data appropriate for evaluating cardiologists' performance; and the same is true for a podiatrist, dentist, chiropractor, etc.-each need different measures for determining quality, but they are often lacking.[9]

To summarize this post, it is critical to have useful, reliable standards to assist with diagnosing patient problems, selecting and delivering the best treatment options, evaluating clinical performance and identifying care value. Unfortunately, we have a long way to do before such standards become a reality.

This concludes by review of data standards. In my next post, I'll examine "technology standards," which focus on enabling the exchange, or interoperability, of information across healthcare systems.

References:
[1] Current Diagnostic Codes are Inadequate - WellnessWiki
[2] Reuters (Sep 20, 2006). Elevated cholesterol may benefit failing hearts.
[3] Gawande, A (2004). The Bell Curve. The New Yorker.
[4] Bush's Value-Driven Health Care Plan Gains Steam as More Employers Step Up (May 10, 2007)
[5] Donabedian, A. (2005). Evaluating the Quality of Medical Care. The Milbank Quarterly 83, 691-729.
[6] Use of claims data is inadequate - WellnessWiki
[7] HealthDay (July 5, 2006). Hospital Ratings Don't Fully Reflect Patient Outcomes. [
8] Healy, B. (Sep. 2006).Who Says What's Best? U.S. News and World Report. [
9] Need for specialy measures - WellnessWiki

Knowledge, Standards, and the Healthcare Crisis: Part 2

In my previous post, I discussed how knowledge is the foundation of healthcare improvement, and how health information exchange is vital for creating and using knowledge. I then introduced the notion that standards are essential for sharing information and implementing knowledge in a way that improves patient care. I also mentioned that, while beneficial, there are substantial challenges to the effective implementation of standards.

Continuing on the topic of Knowledge, Standards and the Healthcare Crisis, I will now begin define what standards actually are.

What are standards?

Standards are models, principles, policies, or rules that provide an agreed-upon framework for doing and understanding things. There are many different types of standards. When it comes to health information exchange, both data and technology standards are important. These standards describe (a) how health data are to be categorized and defined and (b) how different software systems are to communicate with each other when exchanging data. I will now discuss each.

Data standards

Data standards can be divided into at least four categories: terminology, measurement, care process, and messaging format standards. In this post, I describe terminology standards.

Terminology Standards Defined

Health-related terminologies are sets of terms representing a system of concepts within a specified field (domain) of healthcare. In other words, a terminology standard refers to a "nomenclature," i.e., a systemic way of naming and categorizing things in a given category.

Terminology standards include classifications and vocabularies that group together related terms so they can be more easily and consistently understood. Classifications arrange related terms for easy retrieval. Vocabularies use sets of specialized terms to facilitate communication by reducing ambiguity.

Take, for example, the term "high blood pressure" -- the following terms are synonyms of high blood pressure or the names of conditions referring to it:

accelerated hypertension; arteriolar nephrosclerosis; benign hypertension; benign intracranial hypertension; chronic hypertension; essential hypertension; familial hypertension; familial primary pulmonary hypertension; genetic hypertension; hypertension-essential; hypertension-malignant; hypertension-renovascular; hypertensive crisis; idiopathic hypertension; idiopathic pulmonary hypertension; malignant hypertension; nephrosclerosis-arteriolar; pph; pregnancy-induced hypertension; primary obliterative pulmonary vascular disease; primary pulmonary hypertension; primary pulmonary hypertension (pph); primary pulmonary vascular disease; pulmonary arterial hypertension, secondary; pulmonary hypertension; renal hypertension; secondary pulmonary hypertension; severe hypertension; toxemia; toxemia of pregnancy[1], hyperpiesia, and hyperpiesis.

Now imagine two electronic health record systems attempting to exchange patient data. One system is able to recognize the term "high blood pressure" and the other the term "hypertension," but neither can recognize both terms. These two computers would be unable to share the data because they don't "understand" what each other is "saying." This is because computers cannot deal with synonyms (using different words to say the same thing) or homonyms (when the same terms or phrase means different things in different contexts). So, when multiple healthcare providers treat the same patient (who may have multiple health problems), exchanging patient data can be difficult, which is due to the issues of semantics and syntax.

Semantics and Syntax

Semantics and syntax are standards of language. Semantics refers to the meaning of words, expressions and sentences, i.e., how they are defined. Syntax, on the other hand, is the structural or grammatical rules that define how symbols in a language may be combined to form words, phrases, expressions, etc., which includes spelling and word order. For example, in the U.S., this pattern of numbers "###-##-####" could be the syntax for coding a Social Security Number and mmm/dd/yyyy the syntax for a date.

In the situation above, there is semantic confusion since the two software systems define excessive blood pressure using different terms (high blood pressure vs. hypertension). Syntax would be a problem if, for example, both systems used the term high blood pressure, but only one required that the three words be connected, i.e., "high_blood_pressure".

Classifications and vocabulary terminology standards attempt to address these issues.

Classifications

Terminology classification standards in healthcare use a hierarchical index. The ICD-9 diagnostic standards, for example, classifies high blood pressure using this hierarchical index: Diseases of the circulatory system > Hypertensive disease > Essential hypertension (which includes high blood pressure; hyperpiesia; hyperpiesis; arterial, essential, primary and systemic hypertension; and hypertensive vascular); and it gives it a classification code number of 401.[Update; The ICD-10 is not being used]

Vocabularies

Terminology vocabularies standards, on the other hand, often consist of "controlled vocabularies," which are similar to the Library of Congress Subject Headings used by most libraries cataloguing books. Another example is the Yellow Pages in the phone book where, for example, car dealerships are listed under "Automobiles" instead of "Cars" or "Dealerships." Automobiles is, therefore, the "controlled vocabulary" used by the yellow pages.

In healthcare, the MeSH thesaurus is a controlled vocabulary catalog for searching biomedical and health-related information and documents. Searching for "high blood pressure" in the MeSH database returns the heading "Hypertension" and defines it as "Persistently high systemic arterial BLOOD PRESSURE. Based on multiple readings (BLOOD PRESSURE DETERMINATION), hypertension is currently defined as when SYSTOLIC PRESSURE is consistently greater than 140 mm Hg or when DIASTOLIC PRESSURE is consistently 90 mm Hg or more." So, MeSH says "hypertension" should be the term everyone uses to define blood pressure readings within this range; and if they use a different term, it should be translated to "hypertension."

Examples of Existing Terminology Standards

Following are some of the healthcare terminology standards system in use today:

• International Classification of Diseases (ICD) codes for diagnosis disorders
• International Classification of Impairments, Disabilities and Handicaps (ICIDH) codes for diagnosis handicaps
• International Classification of Nursing Practice (ICNP) for class nursing vocabularies
• Diagnostic and Statistical Manual (DSM) codes for classification of mental disorders
• Logical Observations: Identifiers, Names, and Codes (LOINC) codes for representing laboratory tests and procedures
• Current Procedural Terminology (CPT) codes for identifying conventional treatment procedures
• Advanced Billing Concept (ABC) codes for identifying integrative medicine procedures
• Digital Imaging and Communications in Medicine (DICOM) for distributing and viewing any kind of medical image
• Health Care Financing Association (HCFA) that controls Medicare and Medicaid and supports standards for reimbursement
• Unified Medical Language System (UMLS), a system linking together various medical vocabularies
• Systematized Nomenclature of Medicine-Clinical Terms (SNOMED-CT), a system of standardized medical terminology
• The Medical Subject Headings (MeSH) thesaurus, a controlled vocabulary produced by the National Library of Medicine and used for indexing, cataloging, and searching for biomedical and health-related information and documents
• Health Plan Employer Data and Information Set (HEDIS) is a standardized set of 60 performance measures for managed care plans.

In my next post, I define measurement, care process and messaging standards.

Reference:

[1] ICON Health Publications Official Health Sourcebooks

Knowledge, Standards, and the Healthcare Crisis: Part 1

There is widespread acknowledgement that our healthcare system needs radical transformation since:

• All patients “…are at risk for receiving poor health care, no matter where they live; why, where and from whom they seek care; or what their race, gender, or financial status is”[1]
• Healthcare is increasingly more expensive and less accessible[2], with more than 46 million uninsured in the U.S. from every age group and at every income level, 8 out of 10 being in working families[3]
• There is a “knowledge gap”—the healthcare community is drowning in oceans of information, yet doesn't know the best ways to prevent health problems and treat them cost-effectively.[4]

In this next series of posts, I will offer an answer to this daunting question: What can be done to drive continuous improvements in care safety, quality and efficiency, which would enable people to remain healthy longer, manage chronic conditions more effectively, and receive the best possible healthcare delivered in the safest and most economical way?

My answer focuses on the creation, use and evolution of valid health knowledge. Why? Because, I contend, the quality of care would improve dramatically and costs would drop precipitously if everyone:

• Knew the best ways to prevent illness, to avoid complications of chronic diseases, and to treat health problems in the most effective and efficient manner
• Used this knowledge to promote wellness, self-management, and recovery
• Participated in evolving this knowledge to make it ever-more useful and effective.

So, what would it take to foster widespread knowledge creation, use and evolution in our healthcare system?

Well, since knowledge emerges from information,[5] it is essential that both consumers/patients and providers have access to useful health information, including patient health data, care outcomes, and evidence-based guidelines. Furthermore, the information must be presented in a way tailored to each person’s needs and be made available whenever it’s needed. Unfortunately, this is much easier said than done for many reasons.

One daunting core problem involves exchanging patient data between disparate electronic record systems. After all, knowledge can’t grow and care can’t improve unless patients share their health information with their providers, providers share patient information with each other, and researchers have access to this information to develop evidence-based guidelines. And this must be done in a convenient and secure manner that protects patient privacy.

With cost estimates for developing a national health record system enabling patient data exchange being between $100-276 billion,[6] the question is, why must it be so expensive? Aren’t there any easy, inexpensive ways to do it? Let’s examine these questions.

One way to reduce health information exchange costs is by developing and using standards that promote interoperability between disparate health record systems.

Standards are models, principles, policies, or rules that provide an agreed-upon framework for doing and understanding things. When it comes to health information exchange and knowledge growth, at least two types of standards come into play: data and technology standards. These standards describe how health data are categorized and defined, how health outcomes and healthcare performance are measured, how healthcare knowledge is used, and how different software systems communicate with each other when exchanging data.

In my next post, I examine this double-edged sword of standards, pointing out their benefits and the thorny problems they create.

References:

[1] The First National Report Card on Quality of Health Care in America by RAND Corp (2006)

[2] Health Care Coverage in America: Understanding the Issues and Proposed Solutions by The Alliance for Health Reform (March 2007)

[3] The Current Situation - WellnessWiki

[4] The Knowledge Gap - WellnessWiki

[5] DIKUW Blog

[6] Linking Providers Via Health Information Networks by The Alliance for Health Reform (2006) and Dying for Data by R.N. Charette (2006)

Personal Health Application

In this post, I propose the development of a Personal Health Application (PHA). It is a next generation consumer-centric information system that helps improve healthcare delivery, self-management and wellness by providing clear and complete information, which increases understanding, competence and awareness.

PHAs would:

• Incorporate sick-care data currently found in Electronic Health Records (EHRs) used by healthcare providers and Personal Health Records (PHRs), and add well-care data focusing on prevention, self-management, and emotional well-being


• Give a high-definition, big picture, whole-person view of a person's physiological & psychological risk factors, current health, health trends, and projected health status.


• Reveal the interventions that are effective for an individual by integrating and analyzing a lifetime of data about health status & quality of life, conventional and complementary & alternative medicine (CAM) treatments received, and the clinical outcomes of that care.


• Enable the exchange of patient data with providers' EHRs, as well as obtaining data directly from lab, pharmacy or hospital systems.
  

Whereas today's PHRs present narrow views of a person's general health information, PHAs would provide clear, comprehensive views of the whole person-mind, body, spirit and environment-showing risk factors, current health status, health trends, and projected one's future health status. Revealing such trends and predicting one's health condition under different scenarios can be powerful motivators for health living, as well as offering important clinical insights for healthcare providers.

Furthermore, PHRs do little to inform a person about treatment efficacy and the value of CAM approaches. PHAs, on the other hand, would provide this information by collecting and analyzing a lifetime of detailed health data to show what works for the person and what doesn't.

PHAs also bridge well-care and sick-care:

• Sick-care focuses on the treatment of diagnosed physical & psychological problems


• Well-care focuses on preventing serious illnesses and complications, and increasing people's well-being and quality of life through self-management and healthy lifestyles.

The objectives of a PHA are to inform, empower and enable consumers to make better decisions and act responsibly. This includes enabling consumers to:

• Be helpful and proactive in managing their health, rather than passive and reactive.


• Make wise decisions when agreeing to specific treatment options and living health lifestyles


• Carry out strategies for remaining healthy longer


• Comply with plans of care when ill to speed recovery, avoid complications, and achieve the best possible quality of life


• Deal effectively with personal problems and life stressor to maximize one's overall well-being.