Background: The diagnostic process is dynamic – as additional information becomes available and diagnostic uncertainty is addressed, diagnosis evolves. An accurate and updated problem list in the electronic health record (EHR) should reflect modifications and refinement of the working diagnosis (1). Anecdotally, this does not happen: many providers do not the update problem list or favor free text entries of diagnoses in their notes (2). While typically used for billing and reimbursement, accurate Hospital Principal Problem (HPP) entries have utility for clear communication of key diagnoses during transitions (at discharge, during transfers between services); efficient identification of appropriate patient educational materials; as a variable in prediction algorithms (readmission risk scores); and informing patients who wish to access and understand their health data (3-6). The purpose of this study is to describe changes in the documentation of the admission diagnosis as reflected by the HPP entered in the EHR’s problem list.
Methods: We assembled a cohort of encounters of general medicine patients admitted to an academic medical center in Boston, MA during 2019. All problem list entries were retrieved from the EHR (Epic Systems, Inc.). We used descriptive statistics to calculate the frequency of HPP entry upon admission and determined whether these entries changed at discharge by querying the EHR’s problem list audit log. All diagnoses were categorized as either undifferentiated or etiologic using the ICD-10 coding system. Among cases with a change in the HPP from admission to discharge, we identified those with an undifferentiated diagnosis (ICD-10 “R” code) at admission. We then conducted chart review of all cases with a change from an undifferentiated (e.g., abdominal pain) to an etiologic (e.g., pancreatic adenocarcinoma) diagnosis, and an equal number of randomly sampled cases with a change from an undifferentiated to a different undifferentiated diagnosis. Two clinicians conducted a thematic analysis to identify common patterns that explained the change in diagnosis in these cases.
Results: Of 6,635 encounters (Figure 1), 5,527 (83%) had a diagnosis marked as the HPP at admission; of these, 331 (5%) were associated with a change in the diagnosis at discharge. In descending order of frequency, clinicians entering HPP upon admission were trainees (3,541; 64%), non-trainee physicians (1,413; 25.6%), and advanced practice providers (544, 9.8%). Undifferentiated diagnoses were the most frequent HPP type entered at admission (1,841; 33.3%). When the HPP changed from an undifferentiated to etiologic diagnosis, the most common explanatory theme was when diagnostic testing led to a specific diagnosis. The most common theme (Table 1) when the HPP changed from an undifferentiated to a different undifferentiated diagnosis was suboptimal documentation.
Conclusions: Most encounters had a diagnosis marked as the HPP at admission. Few encounters were associated with a change in the diagnosis at discharge, likely reflecting suboptimal documentation. Our findings suggest the need to develop quality improvement and educational initiatives targeted at both trainees and other hospital-based clinicians to improve the accuracy of HPP entry during the hospital encounter.