By definition Problem-based Learning and Improvement (PBLI) is dependent upon recognition of existing knowledge gaps. Recognition of knowledge gaps is hampered by individual clinician "Frame of Reference", i.e., the sum total - and limits - of one's knowledge, training, experience and bias. There is not a single global clinician frame of reference and the only certainty is that any individual physician's frame of reference is but a fraction of the sum total of all medical knowledge. Traditional attempts to identify knowledge gaps are often critically limited by dependence upon the clinician to recognize the presence of the gap in the first place and therefore seek remedial education. This only works in scenarios where the clinician's fund of knowledge is sufficient to allow him or her to recognize that knowledge is incomplete. It cannot address the more dangerous scenario where the demands of the assessment fall completely outside of the clinician's frame of reference. As example, a clinician cannot perform an appropriate evaluation of a patient with the acute onset of chest pain unless they are aware of the possibility, presentation history and signs of vascular dissection. Nor can the resulting documentation include a complete set of pertinent data except by extraordinary happenstance which still is of no help to the clinician without the knowledge of its significance. This limitation applies to all manner of decision support from second opinions, textbooks and journals and typical computerized decision support systems. Unless the clinician actively seeks assistance these sources are of limited utility.

What is needed is a method of recognizing knowledge gaps which is completely independent of the clinician's frame of reference. Such an approach would have a role in both PBLI and patient safety functions. This is the foundation of the development of the existing Lifecom software design: parallel data analysis by the knowledge engine as a means to seek out and identify knowledge gaps in real time.

At present, the healthcare system relies heavily on gross outcomes-based quality improvement strategies to identify knowledge gaps. Unfortunately this method is a 'failure mode' analysis which by definition requires that an untoward event has already occurred and has been identified. It is far better to identify erroneous trends in assessment before irreversible harm occurs. Though technically most demanding, the optimal time for identification of knowledge gaps is during the process of bedside patient assessment. This point has the highest impact to patient care and is most likely to result in clinician knowledge retention since problems and solutions are most tightly linked at this juncture. Real-time intervention theoretically allows CQI/PI strategies to be implemented before an untoward event unfolds rather than the traditional after action response. Early intervention and recognition during assessment also reduces the opportunity to reinforce suboptimal or erroneous assumptions and reduces their clinical impact by providing "intellectual course corrections" before events can spiral out of control. This may prevent error cascades of missed or delayed diagnosis and inappropriate work-up or treatment. This approach is highly instructive, effective and is prototyped by bedside instruction of students and residents by senior attending physicians and in sequential case study presentations within the medical literature.

A real-time approach is difficult to achieve throughout a clinician's career and it is hampered by our current documentation traditions. In the minds of many clinicians the documentation of clinical records has become a separate process from the clinical assessment of a patient. This is due to numerous factors including ever more complicated business practices associated with medical care. Documentation most often occurs after the completion of the assessment and the clinician's decisions are made and occasionally hours after the actual assessment occurs. This means that most documentation is retrospective and reflects not the raw observational details of the clinical situation nor the cognitive or chronological sequence by which a decision was made, but rather the resulting synthesis of data in support of the clinician s decision. Such documentation is inherently incomplete since confirmation biases in human cognition may lead the clinician to search for data in support of a pre-patterned response or expectation while ignoring the presence of negative data. In essence most documents do not record the intermediate decision processes leading to final diagnostic or management decisions. The fact that only gross measures of decision success (such as mortality) can easily be measured allows reinforcement of erroneous patterns in learning and decision-making in cases of either self-limiting or complex disease or when the final outcome of the error may be separated by significant periods of time making linkage difficult. Raw observations and far more complete details of intermediate decision analysis are needed to evaluate and intervene in the actual process of decision-making.

Incentives for clinicians to record real-time detailed assessment data are beyond the reach of traditional methods of medical documentation including EHRs. A regulatory approach might force clinicians to record more data in real-time but cannot address the issue of ensuring that the data set is not compromised by knowledge gaps and is certainly a negative approach to the problem. What is needed are new incentives for clinicians to link documentation and data gathering with clinical assessment thereby capturing a data set sufficient to support the minimum requirements needed by the integral knowledge engine to independently identify knowledge gaps.

Lifecom approaches these challenges through the use of a highly graphical portable user interface designed to support rapid situationally relevant data capture. Once a clinician enters data, the integral knowledge engine reviews all data (including the complete patient medical record and medical co-morbidity) and interacts with the clinician reporting its own differential diagnosis, management, triage or team recommendations, associated knowledge content, and directed questions to further elaborate more complete details of the assessment. The knowledge engine recommendations and prompts are transparent to the user but directed questions are presented in a blinded and unbiased fashion so as not to bias the user's response and observations.