PURPOSE: In a previous study we observed that hospital coders frequently underestimate the incidence of sepsis compared to that diagnosed by critical care physicians. We noted that a significant numbers of patients with sepsis were not identified by hospital coders with lack of proper documentation by physicians to be the most contributory factor.
METHODS: With IRB approval, all suspected sepsis patients admitted to ICTU at our facility from Oct 2009-Dec 2010 were screened and enrolled if informed consent was obtained. During the first year we noted discordance between physician and coders diagnosed sepsis. To enhance coder recognition of sepsis we instituted a nurse-driven protocol. Presence or absence of SIRS criteria was noted on a form by a nurse who then had the physician answer the questions 1) infection present? 2) Sepsis present? The completed form was then placed in the medical record available to the hospital coders. Final diagnosis of sepsis or no sepsis was determined by experienced critical care physicians based on ACCP/SCCM consensus definitions after review of clinical data, including culture results, and discharge diagnoses by hospital coders. Protocol was introduced in July 2010 for 3 months then held for a 3 month period.
RESULTS: Prior to July 2010, only 55% of 71 patients with sepsis were correctly coded. During the period July - Dec 2010, 117 patients were enrolled, with 88 (75%) meeting consensus criteria for sepsis. 51 were in the intervention period and 37 in the post. During the intervention period, coders correctly identified 47/51 (92%) and 33/37 (89%) in the post. Mortality was (33%) in the group of patients that were not coded as sepsis compared to (42%) that were coded. Coders noted infections in 6/8 patients not coded as sepsis and missed the infection in 2 others.
CONCLUSIONS: This protocol improved coder identification of septic patients and led to improved documentation by the physicians during both periods.
CLINICAL IMPLICATIONS: Epidemiological and outcome study performed using coding data may not have a representative patient population.
DISCLOSURE: The following authors have nothing to disclose: Gary Kinasewitz
Improved coding techniques regarding sepsis.