Friday, June 12, 2009

Toward a Meaningful Definition of Meaningful Use (part 2 of 2)

As I discussed in a prior post, the federal government's $20 billion stimulus programs for health IT (HIT) —called HITECH—will fund the development of innovative HIT and use a "carrot & stick" financial approach to encourage clinicians to use HIT in meaningful ways. Unfortunately, the government did not clearly define term “meaningful use,” which has led to an intense debate over its meaning.

The definition I proposed was “using HIT to increase care value (effectiveness and efficiency) by providing ever-better patient-centered cognitive support.” This definition raises the bar over other definitions being offered because it focuses realizing the benefits of ever-increasing care value (effectiveness and efficiency), which is something mainstream HIT does not do.

In this post, I do four things:
  1. Refine the patient-centered cognitive support (PCCS) definition
  2. Compare and contrast PCSS with clinical decision support (CDS)
  3. Clarify why PCCS capabilities in HIT tools should be a requirement of meaningful use
  4. Explain why radical innovation is essential.

Defining Patient-Centered Cognitive Support

As discussed in a recent report by the National Research Council of the National Academies, PCCS is a computerized process that improves decision making by fostering profound understanding through use of a "virtual patient" model.

According to their definition, the PCCS process employs a computerized model of a "virtual patient" that reflects (i.e., is an "abstraction of") an actual patient. An HIT tool would use this virtual patient to guide the selection and analysis of data. These targeted data would be:
…relevant to a specific patient and suggest their clinical implications…[This would] provide decision support…that helps clinicians decide on a course of action in response to an understanding of the patient's status…[These tools would take into account] patient utilities, values, and resource constraints…[and they would] support holistic plans [of care]…These virtual patient models are the computational counterparts of the clinician's conceptual model of a patient. They depict and simulate the clinician's working theory about interactions going on in the patient and enable patient-specific parameterization and multicomponent alerts. They build on submodels of biological and physiological systems and also exploit epidemiological models that take into account the local prevalence of diseases. The availability of these models would free clinicians from having to scan raw data, and thus they would have a much easier time defining, testing, and exploring their own working theories. What links the raw data to the abstract models might be called medical logic—that is, computer-based tools examine raw data relevant to a specific patient and suggest their clinical implications given the context of the models and abstractions. Computers can then provide decision support—that is, tools that help clinicians decide on a course of action in response to an understanding of the patient's status. At any time, clinicians have the ability to access the raw data as needed if they wish to explore the presented interpretations and abstractions in greater depth.
In other words, the virtual patient used in the PCCS process is a computer program with advanced computational algorithms (mathematical and logical operations/steps). The algorithms "…incorporate physics (such as mechanical and electrical properties of tissue) and biology (from physiological to biochemical information) into a platform so that responses to varied stimuli (biological, chemical, physical, and…psychological) can be predicted and results viewed" [Ref: Oak Ridge National Laboratory].

Furthermore, a HIT tool implementing the PCCS process takes "…observations of an individual patient and relates them to a vast dataset of observations of others with similar symptoms and known conditions. By processing all this information, the model can simulate the likely reaction of the individual patient to possible treatments or interventions. Such tools will not only improve the quality of treatment offered to patients who are already ill or injured, but could also be used in preventive medicine, to predict occurrence or worsening of specific diseases in people at risk, for example through family history [Ref: Europe's Information Society Portal]. These simulations and predictions are used to support decisions by identifying the treatment and preventive approaches most beneficial to the virtual patient model, which would then be most likely to benefit the actual patient upon which the virtual model is based.

The HIT-PCCS Gap

Unfortunately, today's mainstream HIT systems do not employ the PCCS process. This, according to same National Research Council report, is a most serious HIT gap. The reason is that PCCS-enabled HIT tools are essential for helping clinicians to understand their patients' problems and needs without having to:
…spend a great deal of time and energy searching and sifting through raw data about patients and trying to integrate the data with their general medical knowledge to form relevant mental abstractions and associations relevant to the patient's situation…[Unfortunately, today's HIT systems] squeeze all cognitive support for the clinician through the lens of health care transactions and the related raw data, without an underlying representation of a conceptual model for the patient showing how data fit together and which data are important or unimportant…As a result, an understanding of the patient can be lost amidst all the data, all the tests, and all the monitoring equipment. In the committee's vision of patient-centered cognitive support, the clinician interacts with models and abstractions of the patient that place the raw data into context and synthesize them with medical knowledge in ways that make clinical sense for that patient.
Since they do not use the PCCS process, mainstream HIT tools do not:
  • Help clinicians gain substantially greater understanding of their patients' situations (i.e., their strengths, weaknesses, risks, needs, and options)
  • Enable patients to understand their own situations better.
Decision-making suffers as a consequence.

Eliminating the HIT-PCCS gap would enhance understanding and promote better shared decision-making about treatment, prevention, health promotion, and self-maintenance (see this link and this link). Because both clinicians and patients would be better informed through the PCCS process, the decisions they make would be more likely result in better outcomes (higher quality and safety) at lower cost. This would translate into increased care value (effectiveness and efficiency). In other words, using HIT tools that implement the PCCS process would help realize important benefits to individuals and society. These benefits include achieving the goals of both the Federal HIT Strategic Plan and the Institutes for Healthcare Improvement's "Triple Aim."

Federal HIT Strategic Plan Goals

PCCS-enabled HIT would help achieve the goals of the Federal government's HIT strategy. According to the Office of the National Coordinator for Health Information Technology, the American Recovery and Reinvestment Act (ARRA) Implementation Plan:
American patients and their caretakers will be the ultimate beneficiaries of the following activities aimed at achieving the President's health IT initiative to accelerate the adoption of health IT and utilization of electronic health records. All of the activities discussed in this section support the current two Federal Health IT Strategic Plan goals:
  1. Inform Health Care Professionals: Provide critical information to health care professionals to improve the quality of care delivery, reduce errors, and decrease costs.
  2. Improve Population Health: Simplify collection, aggregation, and analysis of anonymized health information for use to improve public health and safety [Ref: ONC HIT] 

Institutes for Healthcare Improvement's "Triple Aim"

PCCS-enabled HIT also helps achieve the goals of the Institute for Healthcare Improvement (IHI) recently proposed healthcare improvement design—called the Triple Aim—which has these three critical objectives:
  • Improve the health of the population
  • Enhance the patient experience of care (including quality, access, and reliability)
  • Reduce, or at least control, the per capita cost of care [Ref: About the Triple Aim Initiative].
It is essential, therefore, that utilization of the PCCS process be included in the definition of meaningful use of HIT since sustainable healthcare reform benefits cannot be achieved without it!

PCCS and Meaningful Use of HIT

Based on the discussion to his point, it seems reasonable to conclude that HIT tools are used meaningfully if they employ the PCCS process in order to:
  • Save clinicians time and energy by automating searching and sifting through a patient's clinical details and related research guided by a virtual patient model.
  • Promote a deep and broad understanding of a patient's health status, including the interplay of biological, psychological, and social (i.e., biopsychosocial) influences—past, present, and future.
  • Provide effective, personalized decision support regarding diagnosis, treatment, prevention, and health promotion. And this decision support would:
    • Account for patient preferences, qualities, and circumstances
    • Help improve overall care value
    • Continually evolve.
The following section discusses how PCCS provides superior decision support.

PCCS and Decision Support

A key question concerning PCCS and decision support is: What HIT tools provide decision support and is this decision support based on the PCCS process? To answer this question, let's examine two classes of HIT tools that offering decision support: electronic health records (EHRs) and clinical decision support (CDS) systems.

Electronic Health Records

One type of HIT tool providing some decision support is the EHR (and its electronic medical record counterpart). According to the Concise Guide to CCHIT Certification Criteria, certified EHRs deliver the following decision support capabilities (note that I combined ambulatory and inpatient EHR decision support criteria in the following list):
  • Alerts and Warnings
    • Provide alerts/warnings when
      • There are abnormal test results
      • Patient's vital signs fall outside the normal range
      • Patient is allergic to a drug being ordered
      • Drug or food interactions may occur
      • A follow up test is recommended
      • Patient is currently on a drug for which an allergy has been newly entered
      • Drug side effects may occur based on diagnosis
      • More appropriate or cost-effective therapy could be substituted
      • Drug or food interactions may occur
      • Medication dose is out of recommended range
      • Patient is already on similar drug
      • Patient is currently on a drug for which an allergy has been newly entered
      • Order may be a duplicate
      • More appropriate or cost-effective therapy could be substituted
      • A follow-up or related order is recommended
      • Immunizations are due or overdue
    • Give the reasoning behind an alert, and allow override if appropriate
    • Allow adjusting alert severity based on the clinician's role
    • Report the effect of alerts on clinical decisions
    • Provide dosing guidance based on:
      • Patient weight
      • Lab results
      • Scientific reference material
    • Warn when a medication should not be given because of:
      • Patient age or weight
      • Pregnancy or mother who is nursing
    • Block ordering medications via the wrong route (such as oral vs I.V.)
  • Reminders
    • Provide reminders of recommended care that is due or overdue
    • Generate a list of patients for whom care is due or overdue
    • Generate letters to patients automatically for care that is due or overdue
  • Identify patients for disease and wellness management according to guidelines
    • Based on age, gender, diagnoses, medications, lab results
    • Allow physicians to personalize the care guidelines for individual patients
  • Generate patient education material for medications, diagnosis, procedures and tests
    • Allow tailoring for the patient
  • For inpatient nursing staff:
    • Display for the nurse at the time of administering medications:
      • Any previous alerts
      • Patient's test results and allergies
      • Allow the nurse to use bar-code technology to assure "5 rights" (right patient, drug, dose, time and route)
    • Require the nurse to complete tasks, such as allergy verifications, prior to giving medications.
This list of criteria defines EHR-based decision support as: (a) warnings and alerts about abnormal test results and vital signs, medication issues, duplicate orders, follow-ups, immunizations, and certain therapy substitutions; (b) reminders regarding care due dates; (c) assistance with selection of basic general guidelines in certain situations; (d) general patient education materials; and (e) basic information for hospital nursing staff.

Such EHR-based decision support can be helpful in certain ways. However, since they do not employ the PCCS process, conventional EHRs do not:
  • Search and sift through all of a patient's clinical data and the related research
  • Take into account all the relevant aspects of patient's particular combination of personal preferences, qualities, and circumstances
  • Examine the interactions between a patient's biopsychosocial health problems, threats, needs, and strengths
  • Do an adequate job reporting quality measures (as indicated in a draft report by the National Quality Forum).
And as a result, they do not:
  • Help emerge a deep understanding of a patient's particular health status and risks
  • Generate detailed, personalized, holistic plans of care.
So, even when EHRs provide decision support, their failure to employ the PCCS process severely limits their value in improving healthcare quality and controlling costs. The same can be said, by the way, for personal health records (PHRs).

Today's EHRs (and PHRs), therefore, fall far short of what is needed for "meaningful use" because they do not employ the PCCS process.

Let us now examine another type of HIT tool providing decision support: Clinical decision support (CDS) systems

Clinical Decision Support Systems

Clinical decision support (CDS) systems, not surprisingly, go well beyond the typical EHR in the area of decision support, and some may be add-ons to EHRs. These CDS systems offer:
  • Evidence-based diagnostic assistance
  • Personalized rather than generic information based on a patient's unique symptoms and background
  • In-depth evidence-based guidelines and clinical pathways.
Following are some examples of CDS systems:
Do such CDS systems employ the PCCS process? Well, things tend to get a bit blurry here. A CDS system does implement the PCCS process if it uses evolving virtual patient models to help increase care value by:
  • Automating data searching and sifting
  • Enabling a deep and broad understanding of a patient's biopsychosocial health status
  • Providing personalized decision support precise enough to account for an individual patient's preferences, qualities, and circumstances.
Even if certain CDS systems do utilize the PCCS process, this HIT class is not commonly used in clinical practice or by patients, which only adds to HIT-PCCS gap. 

Establishing Meaningful Use by Bridging the HIT-PCCS Gap

Bridging the HIT-PCCS gap means deploying mainstream HIT tools the implement the PCCS process. These tools would demonstrate a meaningful use of HIT, as discussed below.

Why Meaningful HIT Use Requires PCCS

The reason for making the PCCS process a requirement of meaningful HIT use is because it fosters profound understanding, supports evidence-based decisions, and promotes ever-greater care value by helping to answer questions such as:
  • What are the person's current health problems and risks, taking into account (a) all pertinent physiological and psychological signs and symptoms, (b) all relevant biomedical and psychosocial influences, (c) any related treatments and medications received, and (d) the outcomes of care already rendered? What are the metabolic, genetic, emotional, and behavioral factors affecting the person's health and wellbeing?
  • Is the person's health status being affected by a mind-body interaction and, if so, how is this interaction manifested (see this link for more)?
  • How does the person compare to other people having the same kind of problems, qualities, and circumstances? How are the person's similarities and differences associated with clinical outcomes?
  • What is the prognosis (likely outcome)—short-term and long-term, physically and psychologically—if the person makes no lifestyle changes?
  • What should the plan of care be for treating the person's problems, or for avoid his/her risks from becoming problems—taking into (a) account all relevant research (including conventional allopathic and complementary and alternative approaches), as well as (b) the person's preferences, qualities, and circumstances? What are the risks, benefits, and costs of different plan of care options according to the research?
  • When should certain tests, procedures, or prescriptions not be done/given because they were already done/given, or because they are unnecessary or inappropriate?
  • If an error is made, how can it be rectified with least adverse impact on the person?
  • When has a recommended test or treatment been missed or overlooked, and what should be done about it now?
  • How should the care be coordinated for efficient, effective continuity of care across the healthcare continuum? Who should be collaborating in delivering the person's care and why? What particular personal health information can and should be exchanged with each particular collaborator?
  • How effective is the care currently being rendered (refers to treatment process assessment)? When should a plan of care be modified, why should it be changed, and how should it be different?
  • What was the outcome of each episode of care? What positive and negative factors contributed to the outcome?
  • When does variance from (departure from, non-adherence to) a preferred practice guideline result in better outcomes for a certain types of patients than compliance to it? Why does the variance happen? Who is most likely to benefit from a particular guideline?
  • What patient education/training is required for people with a particular condition in order to promote good self-maintenance?
  • Is the patient adhering to the plan of care? If not, then what are his/her psychological blocks, economic and social obstacles, etc. and how can a patient become more motivated to follow the plan? When is it good that a patient does not adhere to a particular care plan and why?
  • What about a person's social relationships are likely to improve or worsen outcomes? How should one's plan of care be adjusted accordingly?
Unless questions such as these can be answered validly and reliably, there is little chance that HIT decision-support will increase care value and realize sustained improvements in care effectiveness and efficiency. This is why bridging the HIT-PCCS gap is essential to the meaningful use of HIT.

How HIT Tools Can Provide PCCS

Creating and evolving innovative HIT tools that provide PCCS can be a daunting challenge. Accomplishing this goal would require innovative PCCS-enabled HIT tools that:
  • Manage complete personal health information (PHI)
  • Develop and using virtual patient models
  • Support collaboration in loosely-coupled professional and social networks
  • Fit the HIT tools into existing clinical workflows. 

Managing complete protected health information (PHI)

Innovative HIT systems that employ PCCS should securely manage (obtain, analyze, and present) complete biopsychosocial protected health information (PHI) over people's entire lifetimes. To be useful, this PHI should:

Developing and using virtual patient models

It is important that these virtual patient models present decision-support information that is relevant to the specific patient (a) in the context of the current situation and (b) in relation to the whole patient and his/her predispositions. Following are examples of what the models should do.

The virtual patient models should obtain comprehensive PHI from any data streams, manual inputs, biometric sensors, and data stores (databases, files, etc.). In addition to patient status and health history, this information should encompass clinical process data, as well as results tracking, which includes outcomes data, guideline compliance rates, and the reasons for variance (departures) from the guideline recommendations.


The virtual patient models should use computational algorithms that analyze the data obtained in order to identify important patterns (e.g., trends, associations, clusters, and differences) useful for making predictions, linking diagnosis to cost-effective treatments, conducting health-related surveillance (biosurveillance and post-market drug & medical device surveillance), etc. And test the data for statistical relevance to determine which information provides reasonable explanations. The results of such analyses would help determine, for example:
  • Whether a person's risk factors and changes in lab test results or vital signs indicate an imminent or worsening health condition
  • How a person's attributes (e.g., gender, age, medical history, conditions, vital signs, symptoms, genetics, attitudes, etc.) compare to people in different diagnostic groups
  • What treatment options and self-management approaches are most likely to result in the best outcomes for a particular person by accounting for the individuals particular attributes
  • If a medication currently in the market is evidencing side-effects at a higher rate than found in clinical trials
  • If clusters of a particular illness are widespread and indicative of a pandemic, or if the clusters are localized and indicative of environmental toxin, etc.
The virtual patient models should also provide feedback (including suggestions and reminders) and guidance (e.g., diagnostic aids and evidence-based guidelines) presented in personalized views that facilitate decision making, care coordination, and competent care delivery. This would help:
  • Clinicians (a) make valid diagnostic decisions; (b) make evidence-based preventive and therapeutic determinations; (c) deliver appropriate care cost-effectively through efficient, safe and effective procedures; and (d) avoid under-testing, over-testing, under-treating, and over-treating their patients.
  • Patients understand their diagnoses, risks, and treatment options, as well as learn how to self-managements their own health wisely and responsibly.
These models would, therefore, provide PCCS through useful personalized information that increases the likelihood of positive outcomes.

Supporting collaboration in loosely-coupled professional and social networks

Loosely-coupled professional and social networks (as opposed to technical networks) consist of people from multiple locations—who have different roles, responsibilities and experiences—who collaborate to make decisions beyond the knowledge or skills of any individual. These loosely-coupled networks would enable clinicians, researchers, patients, and informal caregivers to pool their wide diversities of knowledge, ideas, and points of view, thereby providing a larger collection of intellectual resource and offering access to a greater variety of non-redundant information and knowledge on which to base decisions.

For example, collaborating researchers and clinicians would foster the emergence of health science knowledge by analyzing, discussing, and interpreting care process and outcome data in light of patients’ diagnoses and qualities. This would promote the development and evolution of virtual patient models by transforming this knowledge into evolving evidence-based guidelines aimed at the continuous improvement of care effectiveness and efficiency.
Another important thing these collaborative networks can do is share and "play seriously with" different virtual patient models. That is, they would compare models and test them for their ability to reflect reality accurately; they manipulate the models to represent different scenarios, such as "what if" scenarios about the probability of future occurrences; and they discuss the assumptions and results the models produce. When they find models that disagree or generate invalid results, they examine the fundamental assumptions built into the models, looking for logical flaws and inconsistencies and debating about the assumptions and practical value of the model. By challenging the model's assumptions, useful counterintuitive insights often emerge, innovative thought is sparked, new questions arise, and compelling and unexpected issues are discovered. This means that sharing and playing with models is an effective path to innovation and value creation.

As such, these loosely-coupled networks provide the greatest opportunities for emerging creative ways to develop, evolve, and use the virtual patient models that provide PCCS.
These loosely-coupled networks should be supported by a cyberinfrastructure that, as described by the National Science Foundation,"…combines computing, information management, networking and intelligent sensing systems into powerful tools for…collecting and analyzing large volumes of data, performing experiments with computer models and bringing together collaborators from many disciplines." [Ref: NSF]. The cyberinfrastructure should be secure, economical, easy-to-use, and convenient.

Fitting the HIT tools into clinical workflows

PCCS-enabled HIT tools should assist clinicians in making decisions during their natural course of work, rather than requiring major adjustments of their workflows. This would increase the likelihood that clinicians will take advantage of that PCCS. 

The Need for Radical Innovation

The National Research Council’s report calls for radical change this way:
Change in the health care system can be viewed through two equally important
lenses—those of evolutionary and of radical change. Evolutionary change means
continuous, iterative improvement of existing processes sustained over long
periods of time. Radical change means qualitatively new ways of conceptualizing
and solving health and health care problems and revolutionary ways of addressing those problems. Any approach to health care IT should enable and anticipate both types of change since they work together over time.
Unfortunately, the HIT industry is focused on the gradual evolutionary change of existing products in lieu of radical leaps of change through development of new breeds PCCS-enabled HIT tools. This is why an abundance of conventional EHR commodities now crowd the mainstream market (see, for example, this link and this link), yet there is little attention given to CDS systems and almost no attention to PCCS.

As a result, there is a tendency to define meaningful use simply in terms of conventional HIT commodities, instead of “raising the bar” to new heights by requiring disruptive (radical, discontinuous) PCCS-enabled innovations. For example, the HIMSS definition of meaningful use, which was developed by a HIT vendors’ association, calls for the immediate use of current day EHRs and related HIT commodities, but doesn’t require CDS systems to be used until 4-7 years from now. Furthermore, the degree of decision support to be delivered by the CDS systems is minimal and falls far short of delivering the benefits of the PCCS process.

The meaningful use definition, therefore, ought to balance these evolutionary changes with the requirement for dramatically different types of software applications—a new generation of radical innovations—that employ the kind of PCCS able to help transform our healthcare system for the better.

Conclusion

Any good definition of meaningful use of HIT ought to include the implementation of the PCCS process to drive ever-evolving clinical decision support.

Since mainstream HIT tools to not employ the PCCS process, realizing such meaningful use will require substantial long-term commitment by diverse groups of collaborators in the development, use, and evolution of virtual patient models. If increasing healthcare value is truly our nation's goal, then there is no good alternative!

No comments: