The Problem
The Electronic Health Record (EHR) problem list (PL) supports communication of the patient’s problems across a wide range of clinical environments and patient caregivers. An accurate problem list serves as a foundation for clinical care, population health management and multiple secondary processes including research and severity of illness risk scoring. Maintaining an up-to-date problem list is essential to patient-centered care but is often a secondary process compared to the needs of direct patient care.
The PL suffers from an accumulation of diagnoses that can be either outdated, no longer accurate, duplicative, similar or conflicting.(1) A recent study on problem list completeness found wide variations of for diagnoses that have been used as visit diagnoses for inclusion in problems such as diabetes or asthma.(2) Another large cohort study found 22% incidence of the common diagnosis hypertension to not be included in the problem list and multiple duplications of diagnoses like asthma, Crohn’s disease and diabetes.(3)
In addition to the problem of volume and accuracy of the PL content, PLs are also not organized in a format that supports easy identification of related or outdated diagnoses, which also hinders cleanup. In our investigations, we found anywhere from 20% to 50% of diagnoses on a patient’s PL are not current and up to date for the patient. Because EHR data for the PL is at the individual level, broader population-level reviews of the inaccuracies have been not well identified until this project was undertaken.
This summary seeks to describe the work accomplished to improve the organization and cleanup of the Duke Health problem list in our EHR.
Our Solution
We developed 21 system/condition-based groupers using SNOMED-CT hierarchal concepts refined with Boolean logic (Figure 1). System groupers included traditional medical specialty categories and clinically relevant care coordination and procedure-based groupings. Specialty organization of diagnoses are an improvement over foundation default PL organization in Epic that organizes problem lists primarily in an alphabetized format. While there are other options for PL display such as priority organization, these require the user to a
pply a prioritization level one by one and are not used by a majority of our clinical users.
SNOMED-CT codes are translated in a one-to-many format for ICD-10 codes, allowing for more complete ICD10 groupers out of the potential more than 100,000 ICD10 codes available. For example, the Neurology specialty grouper with 167 SNOMED-CT concepts mapped to 9,243 ICD10 codes.
Outcomes
The 21 specialty groupers were iteratively built by the primary study lead in the summer/fall of 2021 and implemented to be available for all Duke users in Epic in Dec 2021 and updated again in July 2022. All new users and most ancillary staff are now defaulted to new System Level organization of the PL for Duke Health users. Feedback has been very positive from all users but especially for staff who do chart reviews. To better understand the potential opportunity for automated cleanup of the PL, we analyzed a representative sample of PL diagnoses across 79 patients identified for specific focus on Oncology, Cardiology, Neurology and Orthopedics diagnoses. This format supported analysis of PL content and disorganization at more of an aggregate level.
Across the 79 patient cohort we identified 2,835 diagnoses ranging from 112 to 5 diagnoses with a mean of 36 diagnoses per patient. We found 1,508 (53.2%) singular concept diagnoses (i.e., diagnoses without duplicate, conflicting or similar diagnoses). In comparison, we found 1,327 (46.8%) diagnoses across 634 concepts with duplicate, related, conflicting, similar, lapsed or no longer active diagnoses. Notably this review was of the PL diagnoses alone and did not include chart review. The goal of this evaluation was to identify diagnoses that could be potentially automatically resolved or placed on the past medical history section.
In characterizing the state of the PL diagnoses with potential opportunities for cleanup, we found 7.7% of diagnoses were lapsed and potential targets for automatic conversion to history of diagnoses. The majority of these being for acute myocardial infarctions or acute stroke diagnoses where the date of onset of the problem was clearly more than 30 days old. We additionally found 124 exact duplicate diagnoses across 55 concepts identifying another significant opportunity for automated clean-up. Overall, 12.1% of the diagnoses from our evaluation subset had the potential opportunity for an automated clean-up.
Impact
We continue to work to implement an automated solution to clean-up the problem lists for all Duke Health patients in the Chronicles database in late 2022 to early 2023. This solution will continue to be applied to our problem lists in an ongoing format as a means to keep the PLs more accurate for our patients. We anticipate that this cleanup will have long-lasting effects across direct patient care and secondary processes that rely on problem lists diagnoses as a means to under the severity of illness of our patients.
Next Steps
We continue to work to implement an automated solution to clean-up the problem lists for all Duke Health patients in the Chronicles database in late 2022 to early 2023. This solution will continue to be applied to our problem lists in an ongoing format as a means to keep the PLs more accurate for our patients. We anticipate that this cleanup will have long-lasting effects across direct patient care and secondary processes that rely on problem lists diagnoses as a means to under the severity of illness of our patients.
References
- Kreuzthaler M, Pfeifer B, Ramos JA, Kramer D, Grogger V, Bredenfeldt S, Pedevilla M, Krisper P, Schulz S. EHR problem list clustering for improved topic-space navigation. BMC Medical Informatics and Decision Making. 2019 Apr 1;19(3):72.
- Wright A, McCoy AB, Hickman TT, Hilaire DS, Borbolla D, Bowes III WA, Dixon WG, Dorr DA, Krall M, Malholtra S, Bates DW. Problem list completeness in electronic health records: a multi-site study and assessment of success factors. International journal of medical informatics. 2015 Oct 1;84(10):784-90.
- Wang EC, Wright A. Characterizing outpatient problem list completeness and duplications in the electronic health record. Journal of the American Medical Informatics Association. 2020 Aug;27(8):1190-7
