BACKGROUND: Los Angeles County Department of Public Health (LACDPH) requires that general acute care hospitals (GACHs) report suspect Clostridium difficile infection (CDI) outbreaks. However, approaches to identify outbreaks vary by facility and often are subjective. To more objectively and systematically identify outbreaks, we assessed the ability of SaTScan, a statistically-based cluster detection software, to detect reported CDI outbreaks from GACHs using National Healthcare Safety Network (NHSN) data.
METHODS: All CDI outbreaks reported to LACDPH by GACHs from 2010 through 2016 were reviewed. For GACHs with a documented outbreak, healthcare facility-onset CDI cases reported to the laboratory-identified (LabID) module of NHSN were analyzed iteratively before and during the reported outbreak by month. A prospective, purely temporal discrete Poisson analysis at both the facility and unit (unique patient care area) level was performed in SaTScan. Statistically significant clusters at the p <0.10 level were compared to reported outbreaks for the date detected, number of cases, and duration.
RESULTS: Seven CDI outbreaks across six GACHs were reported to LACDPH from 2010 through 2016. Reported outbreaks had a median 12 cases (range: 3-38) over a median 49 days (range: 6 to 160). Facility-level SaTScan analysis identified three of the reported outbreaks with clusters including a median 15 cases (range: 3-136) over 30 to 457 days. One such outbreak would not have been detected via SaTScan until 50 days after it was reported. SaTScan would have signaled the others 68 and 216 days earlier than reported, before 17 and 58 subsequent cases occurred, respectively. In all seven outbreaks, SaTScan unit-level analysis identified at least one unit per facility with significantly increased CDI during the outbreak time period. Unit-level clusters had a median 5 cases (range: 2-17) over 61 days (range: 30-517). Five reported outbreaks were focused on a single unit; SaTScan analyses identified four of these units, in which three could have been identified 2 to 185 days earlier, before 2 to 10 subsequent cases occurred. In one facility, SaTScan identified clusters in multiple units adjacent to the unit reported, 98 days earlier than the reported outbreak, and identified 15 additional cases.
CONCLUSIONS: SaTScan is a free software that can systematically identify potential outbreaks, sometimes earlier than traditional reporting. Unit-level analysis is likely more sensitive than facility-level analysis. Health departments and GACHs may consider complementing traditional reporting methods with temporal scan statistics.