Background:
Accurate problem lists linked to the electronic medical record (EMR) can be a source of structured clinical data useful for communication among clinicians for patient care, clinical decision support design, and EMR phenotyping. Nevertheless, modifying the electronic problem list is disruptive to clinician workflow under traditional charting methods, resulting in incomplete, inaccurate, and outdated problem lists that do not reliably structure clinical data. In 11/2013, Stanford Hospital’s intensive care unit (ICU) implemented problem based charting (PBC), a system of clinical documentation that uses the problem list as an anchor for physician notes. Our study evaluates the effectiveness of PBC for improving problem list utilization and completeness.
Methods:
We conducted a pre-post study from a 30 consecutive month period prior to PBC implementation in the ICU (pre-PBC) from 5/1/2011 to 11/22/2013 and after PBC implementation (post-PBC) from 11/23/2013 to 5/1/2016. Frequencies of total, resolved, and deleted hospital problems on the problem list were ascertained per patient encounter. To assess problem list completeness, we assessed the recall for five common ICU conditions– sepsis, acute respiratory failure, acute renal failure, pneumonia, and venous thromboembolism – with the problem list, pre- and post-PBC. We used the billing code list as the reference standard. The recall for each clinical condition was calculated by dividing the total number of patient encounters that contained the corresponding ICD9 codes in both the billing code list and the problem list (true positive) by the number of patient encounters that contained the corresponding ICD9 codes on the billing code list. Precision was calculated by dividing the true positive frequency by the total number of encounters containing corresponding ICD9 codes on the problem list.
Results:
4669 and 5838 patient encounters were identified in the pre-PBC and post-PBC periods, respectively. There were no significant differences in the distributions of age, gender, race, % smokers, and length of stay between the two groups. Physicians in the post PBC period added a higher mean number of new problems per encounter (6.2 stdev 4.7 vs 3.6 stdev 3.2, p<0.001). There was also a higher proportion of problems per encounter that were resolved by the physician in the post-PBC period (11% stdev 2% vs 6% stdev 1.8%, p<0.001), but no difference in the proportion of deleted problems per encounter (2% stdev 1%). The encounter frequencies, recall, and precision for five common ICU conditions, pre- and post-PBC, are shown in table 1. There was a significant increase in the recall of all five queried ICU conditions with the problem list after PBC while maintaining high levels of precision: sepsis (8% vs 42%, p<0.001), acute respiratory failure (34% vs 59%, p<0.001), acute renal failure (35% vs 56%, p<0.001), pneumonia (28% vs 37%, p<0.001), venous thromboembolism (29% vs 43%, p<0.001).
Conclusions:
PBC is a readily available, albeit not widely used feature of existing EMR platforms, such as EPIC, that could integrate electronic problem list management into physician workflow. Our study, to our knowledge, is the first to demonstrate that PBC can improve both problem list utilization and completeness, suggesting that it may be a viable tool for leveraging physician expertise to structure clinical data at the time of the clinical encounter, which has applications for both direct patient care and big data analytics.