Steering Committee interviews revealed that the IOM domains were useful but varied in importance. The domain rated lowest in utility was “administrative processes” and the highest was “health information and data.” The Steering Committee added a number of additional features to the IT domains, with each domain acquiring several additions overall. For example, the IOM defines “health information and data” as, “electronic storage and immediate access to diagnoses, allergies, lab test results, medications, clinical notes, demographics.” Committee members added several items, including patient assessment and surveys, routine outcomes assessments, and results of ongoing assessments.
Results from the literature review have been published.19
Much of the literature is nonexperimental. In studies with experimental designs, components related to positive outcomes are connection to an EMR, computerized prompts, population management, specialized decision support, electronic scheduling, and patient health records. The only informatics component not associated with improved outcomes was access to treatment guidelines. This is consistent with some previous, broader reviews.21,32
The expert panel agreed that information systems should be patient-centered and emphasize patient preferences, functioning, and quality of life. The “recovery” model, which was originally developed for mental health,33
was seen as applicable to other chronic medical disorders. This recovery model emphasizes the importance of moving beyond disease alone to strategies for supporting individuals with chronic illness as they work toward life goals and aspirations.34
The panel agreed that the routine assessment of critical outcomes was necessary, using reliable, valid, and commonly used measures such as the nine-item scale of the patient health questionnaire for depression or blood pressure for hypertension.35
Choice of outcomes should be guided by quality improvement priorities. The panel recommended using automatic scoring algorithms to link problematic outcomes scores to appropriate treatment recommendations and agreed that these results should be easily accessible and interpretable by providers at the time of the clinical encounter. There were several ideas for displaying these results, such as a timeline graphing scores against medication trials, clinical encounters, and landmark events. There have been published informatics tools that graph blood pressure against antihypertensive medications on a timeline.36
Outcomes data should also be usable for population-based improvement of health care quality.
A key problem with current EMRs is information overload, and data presentation that does not efficiently meet the needs of providers in complex cases. In response, the panel stated that there should be an interactive, sequential, and comprehensive treatment plan in the EMR. Each medical condition in the patient’s problem list would be accompanied by a clinical care pathway, which could be used by clinicians to quickly identify the current stage of treatment for each of the patient’s problems in relation to best-practice algorithms. For example, in the management of depression, the care pathway could start with initiation of either antidepressant medication or psychotherapy, then proceed to intensifying the treatment, and trying alternative or combination treatments for patients failing to respond. Once depression has remitted, the patient would enter a maintenance phase of treatment. The treatment plan would clearly indicate which phase of treatment the patient is in, and how long they had been there. Outcome indicators, such as depression severity, could be graphically displayed in relation to treatment trials.
With regard to the diabetes use case, panel members agreed that communication, collaboration, and coordination of care among all providers is critical to improving care. A method such as an electronic flow sheet to track steps in care over time would be beneficial to the coordination of diabetes care. Panel members believed that there is a need for scheduling long-term care needs (such as eye exams and foot care for diabetes) and that a tickler alert system would be ideal for long-term, automated scheduling.
Another recommendation was structured, automated templates for progress notes, allowing providers to easily report and access critical information regarding treatment history and planned treatment steps. The panel agreed that templates with functionalities such as guideline-based checklists for providers could enhance diabetes care. Other suggestions that were endorsed by the panel included creating provider alerts for gaps in prescription refills to target issues with medication adherence, and creating a system of automated phone messages for patients, reminding them of appointments and medication refills.
The panel concluded with a discussion of comorbid disorders. Panel members emphasized that the treatment plan should include each of the patients’ medical conditions, whereas allowing for prioritization of problems. Accessing one medical condition, for example, diabetes, could trigger the display of other important and related comorbid conditions, such as hyperlipidemia, as well as the display of relevant labs and outcome measures. Furthermore, panel members believed that the treatment plan should list all providers and care managers associated with each condition and be shared across treatment sites. Because comorbidity is very common in diabetes care, the decision was made to merge the diabetes and comorbidity use cases. The use cases were revised based on results from the panel, and the final versions are available online (DOI http://dx.doi.org/10.1007/s11606-007-0303-04).