Information Overload and Health Decision-Making

With the push to improve decision-making with electronic health records and related health information technology, a key question to be answered is: How should we deal with information overload?

I’m defining information overload as a state of having more information available than one can readily assimilate, that is, people have difficulty absorbing the information into their base of knowledge. This hinders decision-making and judgment by causing stress and cognitive impediments such as confusion, uncertainty and distraction.

Information overload can adversely affect several types of data-intensive health-related decisions, including:
  • Decisions about wellness (preventing illness, maintaining health), which ought to take into account information such as a person’s behavioral and genetic risk factors, degree of physical activity/exercise, stress and emotional distress levels, use of vitamins and dietary supplements, etc.
  • Decisions about diagnoses (identifying an existing health problem), which ought to consider information such as a person’s physical and psychological symptoms, lab test results (of which there are over 4,000), medical history, allergies, demographics, psychosocial problems, genetics, the mind-body connection, etc.
  • Decisions about treatment selection and implementation (intervening to treat a medical and psychological health problem), which ought to be based on a person’s diagnostic information, evidence-based guidelines, personal preferences, social support network, available resources, etc.

Obtaining all this information requires the collection and analysis of a wealth of diverse data, including (but not limited to):
  • Physiological/biomedical problems and risk factors, e.g., body organ and system dysfunctions/disturbances; physical pain; energy and attentional excesses and deficits; eating, sleeping, and sexual disorders; mobility problems; allergies; etc.
  • Vital signs (e.g., heart beat, breathing rate, temperature, and blood pressure)
  • Lab test results (e.g., general blood & urine screenings, microbiology, virology, cytopathology, histopathology, cytogenetics)
  • Imaging studies
  • Medications being taken
  • Interventions being rendered
  • Dietary supplements being used
  • Medical/treatment history and personal demographics
  • Affective-motivation-characterological dysfunctions/problems, e.g., intensity, frequency, and duration of negative affect and emotional stability; maladaptive and dangerous behaviors including impulsivity, compulsions, and suicidality; personality and psychiatric disorders; etc.
  • Psychological vulnerabilities, e.g., sense of helplessness and hopelessness; ineffective coping strategies; low frustration tolerance; disturbing thoughts and negative emotions associated with them; traumatic experiences; self-image problems; etc.
    Psychosocial distress, e.g., occupational, educational, and social/interpersonal dysfunctions; current life-stressors; etc.
  • Psychoactive substance use, including alcohol & substance abuse, dependency, withdrawal
  • Psychological-physiological (mind-body) interactions, including (a) biomedical illnesses/traumas that may cause or exacerbate psychological symptoms, (b) medication side-effects that may cause or exacerbate psychological symptoms, and (c) psychological factors that may cause or exacerbate physical symptoms
  • Genetic markers
  • ICD and DSM diagnostic codes; CPT procedures codes
  • Intake and discharge/outcomes data
  • Healthcare utilization data
  • Consumer satisfaction
  • Motivation for self-care.

If a person has a health problem for which a substantial portion of this information would improve decisions, information overload becomes a real risk because there is simply too much information for a human mind to handle. So, shouldn’t we use computers to collect and analyze all the data that may be relevant to a person’s condition?

I bet most would say use of computers to collect volumes of data about a person'e health problems makes sense if they could : (a) obtain, organize, and analyze all the relevant data without great difficulty, inconvenience, and expense; (b) keep sensitive patient data secure; (c) allow the data to be shared with authorized persons; and (d) use artificially intelligent software programs to make sense of it all and help people make better decisions.

Unfortunately, this rational vision has not been realized. While computer power and artificial intelligence capabilities continue to increase exponentially (e.g., see Ray Kurzweil’s book “The Singularity is Near”), and while there are efficient and effective ways to collect, organize, analyze, and share all these data, humanity currently lacks the knowledge and understanding needed to develop a software system able to incorporate all this information to help guide health-related decisions.

So, what should we do? Focus on collecting “minimal standard data sets” that provides some useful information and avoids overload, but are not enough to improve health decisions substantially? Or should we begin collecting comprehensive data even though we lack the ability to use it all to support decisions, even at the risk of information overload?

If our goal is improve healthcare quality and control costs, I contend that we should collect all the relevant data humanly possible and turn it into useful information and knowledge that increases understanding for wise decision making. But how can this be done without creating information overload?

To answer this question, let’s re-examine the definition of information overload: It is a state of having more information available than one can readily assimilate, that is, people have difficulty absorbing the information into their base of knowledge. Well, what has to happen for people to increase people's ability to assimilate information?

I contend that people with more valid knowledge about a particular knowledge domain (i.e., field or branch of knowledge, such as diagnosing medical problems), and the more they understand that domain (e.g., the better able they are to use their knowledge to answer questions about prevention, diagnosis, and treatment), then the more they information they can absorb about that domain and use it to improve their decisions. In other words, the stronger one’s foundation of knowledge about something and ability to utilize that knowledge effectively, the more one can learn and integrate into one’s existing base of knowledge without experiencing information overload.

This means that a consequence of the knowledge gap in healthcare today is people’s susceptibility to information overload. This creates a viscous cycle of information input --> information overload --> information rejection --> inhibited knowledge growth. This results in a tendency to minimize information input, e.g., by focusing on minimal data sets rather than the collection and integration of comprehensive, multidisciplinary sets of data across patients’ lifetimes described in the previous post, including patient results (clinical outcomes & costs), provider characteristics and treatment methods/processes, and patient attributes.

Breaking out of this knowledge-inhibiting cycle requires a dramatic shift in the way we view and approach health information management.

Following are several methods for minimizing information overload:
  • Filtering. This involves defining what is useful (e.g., relevant and valid) and what isn’t, and then allowing only the useful information to be accessed. There are many different ways to filter information using software applications, which may include active or passive methods, and personal or social methods (including subject matter experts). See, for example, Collaborative Filtering, Information Filtering, and Intelligent Agent Filtering.
  • “Just-In-Time” (JIT) delivery. This involves delivering information in a “just-in-time” (JIT) manner, i.e., having the particular information you need “served to you” when you need, rather than having to search for it.
  • Competency-based instruction**. This involves tailoring the level of instruction to one’s ability to learn. Imagine an e-learning (distance learning) system that keeps track of your knowledge level about a particular topic (domain) in the curriculum using tests to evaluate what you’ve learned after receiving instruction. You do not receive instruction on subsequent topics until you’ve learned the preliminary information you need to know. And it makes sure you recognize what you still need to learn for a particular situation.
  • Personalized presentation. This involves presenting information in a manner tailored to a person’s preferences, i.e., customizing the way information is shown to minimize confusion and maximize clarity, and for maximum ease-of-use.
  • Using summary/aggregated data with “slicing, dicing and drill-down” capabilities. This involves combining lots of data into a few aggregate summaries and statistical analyses that give a bird's-eye view,” identify patterns and make predictions, test for statistical significance, and enables people to examine the data from different perspectives, as well as to see the data in “finer levels of granularity” (i.e., view the underlying details). OLAP (On-Line Analytical Processing) tools and spreadsheet pivot tables are technologies that do this through data mining. It is also common to “digital dashboards.”
  • Increase your level of knowledge and understanding. While the methods above rely on technology to avoid information overload, strengthening your mind by increasing what you know and understand about a topic/domain enables you to absorb (assimilate) more information in that area without becoming overloaded.

Here's an example of how these six methods can work together to help a healthcare practitioner become more knowledgeable and make better decisions without suffering information overload. Similar things can be done to benefit patients, payers, and others.

Imagine a person with a complex health problem being seen by his practitioner. A computerized diagnostic assessment tool such as the Problem Knowledge Couplerssoftware is used to obtain comprehensive information from the patient and practitioner. It then analyzes all the patient information, matches it with an extensive healthcare database, and presents specific recommendation concerning diagnosis and treatment, with links to relevant studies and other supporting documentation, thereby focusing attention on what’s most important (via information filtering). This information, along with any other relevant patient data stored in the practitioner’s EMR/EHR (electronic medical record/health record) and the patient’s PHR (personal health record), is then display in a patient profile tailored to his particular preferences (via personalized presentation).

Once an appropriate diagnosis and treatment approach are identified, a computerized clinical guidelines system is used to recommend particular evidence-based interventions, including specific protocols to follow. Upon the practitioner’s approval, the system uses this information to generate a targeted plan of care. If the practitioner needs instruction to assist in the delivery of the selected treatment regimen, the system determines what s/he has already leaned (via competency-based instructional methods) and what s/he needs to learn now in order to deliver quality care; it then serves him/her the additional information (via JIT delivery).

When the episode of care is completed and clinical outcomes data are collected, other software application analyzes all the data and presents summary data showing how well the patient responded to treatment compared to very similar patients (via a digital dashboard) on key measures.

By having these outcomes data de-identified and sent to a central data warehouse for research and analysis, they contribute to an evolving base of clinical information that increases knowledge and understanding, thereby enabling the assimilation of even more information, resulting in ever-improving guidelines and decision support processes.