Automated Detection of Infectious Disease Outbreaks in Hospitals: A Retrospective Cohort Study

Division of Infectious Diseases and Health Policy Research Institute, University of California Irvine School of Medicine, Irvine, California, United States of America.
PLoS Medicine (Impact Factor: 14.43). 02/2010; 7(2):e1000238. DOI: 10.1371/journal.pmed.1000238
Source: PubMed


Detection of outbreaks of hospital-acquired infections is often based on simple rules, such as the occurrence of three new cases of a single pathogen in two weeks on the same ward. These rules typically focus on only a few pathogens, and they do not account for the pathogens' underlying prevalence, the normal random variation in rates, and clusters that may occur beyond a single ward, such as those associated with specialty services. Ideally, outbreak detection programs should evaluate many pathogens, using a wide array of data sources.

Download full-text


Available from: Ken P Kleinman
  • Source
    • "Furthermore, even well-established infection prevention strategies can be often disregarded [3] at various levels. Overall, these factors may lead to the transmission of pathogens within the hospital, a missed opportunity to investigate a disease cluster, or false ascertainment of a cluster resulting in the misuse of hospital resources for investigational purposes [2]. Statistical methods, such as the scan statistic, may enhance the identification of disease clusters and/or outbreaks in the hospital setting [2,4]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background In healthcare facilities, conventional surveillance techniques using rule-based guidelines may result in under- or over-reporting of methicillin-resistant Staphylococcus aureus (MRSA) outbreaks, as these guidelines are generally unvalidated. The objectives of this study were to investigate the utility of the temporal scan statistic for detecting MRSA clusters, validate clusters using molecular techniques and hospital records, and determine significant differences in the rate of MRSA cases using regression models. Methods Patients admitted to a community hospital between August 2006 and February 2011, and identified with MRSA > 48 hours following hospital admission, were included in this study. Between March 2010 and February 2011, MRSA specimens were obtained for spa typing. MRSA clusters were investigated using a retrospective temporal scan statistic. Tests were conducted on a monthly scale and significant clusters were compared to MRSA outbreaks identified by hospital personnel. Associations between the rate of MRSA cases and the variables year, month, and season were investigated using a negative binomial regression model. Results During the study period, 735 MRSA cases were identified and 167 MRSA isolates were spa typed. Nine different spa types were identified with spa type 2/t002 (88.6%) the most prevalent. The temporal scan statistic identified significant MRSA clusters at the hospital (n = 2), service (n = 16), and ward (n = 10) levels (P ≤ 0.05). Seven clusters were concordant with nine MRSA outbreaks identified by hospital staff. For the remaining clusters, seven events may have been equivalent to true outbreaks and six clusters demonstrated possible transmission events. The regression analysis indicated years 2009–2011, compared to 2006, and months March and April, compared to January, were associated with an increase in the rate of MRSA cases (P ≤ 0.05). Conclusions The application of the temporal scan statistic identified several MRSA clusters that were not detected by hospital personnel. The identification of specific years and months with increased MRSA rates may be attributable to several hospital level factors including the presence of other pathogens. Within hospitals, the incorporation of the temporal scan statistic to standard surveillance techniques is a valuable tool for healthcare workers to evaluate surveillance strategies and aid in the identification of MRSA clusters.
    Full-text · Article · Jul 2014 · BMC Infectious Diseases
  • Source
    • "Furthermore, within the hospital setting, rule-based criteria are often applied to ascertain if an outbreak has occurred [2]. Consequently, the above approaches may result in false outbreak alarms or outbreaks that are overlooked, and subsequently the misuse of hospital resources or a missed opportunity for further case investigation [2] or prompt intervention, prevention, and control. Therefore, studies evaluating the incorporation of various statistical methods to complement traditional surveillance techniques within the hospital setting are being performed. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background In hospitals, Clostridium difficile infection (CDI) surveillance relies on unvalidated guidelines or threshold criteria to identify outbreaks. This can result in false-positive and -negative cluster alarms. The application of statistical methods to identify and understand CDI clusters may be a useful alternative or complement to standard surveillance techniques. The objectives of this study were to investigate the utility of the temporal scan statistic for detecting CDI clusters and determine if there are significant differences in the rate of CDI cases by month, season, and year in a community hospital. Methods Bacteriology reports of patients identified with a CDI from August 2006 to February 2011 were collected. For patients detected with CDI from March 2010 to February 2011, stool specimens were obtained. Clostridium difficile isolates were characterized by ribotyping and investigated for the presence of toxin genes by PCR. CDI clusters were investigated using a retrospective temporal scan test statistic. Statistically significant clusters were compared to known CDI outbreaks within the hospital. A negative binomial regression model was used to identify associations between year, season, month and the rate of CDI cases. Results Overall, 86 CDI cases were identified. Eighteen specimens were analyzed and nine ribotypes were classified with ribotype 027 (n = 6) the most prevalent. The temporal scan statistic identified significant CDI clusters at the hospital (n = 5), service (n = 6), and ward (n = 4) levels (P ≤ 0.05). Three clusters were concordant with the one C. difficile outbreak identified by hospital personnel. Two clusters were identified as potential outbreaks. The negative binomial model indicated years 2007–2010 (P ≤ 0.05) had decreased CDI rates compared to 2006 and spring had an increased CDI rate compared to the fall (P = 0.023). Conclusions Application of the temporal scan statistic identified several clusters, including potential outbreaks not detected by hospital personnel. The identification of time periods with decreased or increased CDI rates may have been a result of specific hospital events. Understanding the clustering of CDIs can aid in the interpretation of surveillance data and lead to the development of better early detection systems.
    Full-text · Article · May 2014 · BMC Infectious Diseases
  • Source
    • "For example, drug-resistant pathogens found in hospitals either are due to nosocomial transmission or are brought into the hospital by the patient. The former can be identified using surveillance data by assessing the number of cases in a fixed time period [1]; this identification is essential in optimizing hospital control measures. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Surveillance systems of contagious diseases record information on cases to monitor incidence of disease and to evaluate effectiveness of interventions. These systems focus on a well-defined population; a key question is whether observed cases are infected through local transmission within the population or whether cases are the result of importation of infection into the population. Local spread of infection calls for different intervention measures than importation of infection. Besides standardized information on time of symptom onset and location of cases, pathogen genotyping or sequencing offers essential information to address this question. Here we introduce a method that takes full advantage of both the genetic and epidemiological data to distinguish local transmission from importation of infection, by comparing inter-case distances in temporal, spatial and genetic data. Cases that are part of a local transmission chain will have shorter distances between their geographical locations, shorter durations between their times of symptom onset and shorter genetic distances between their pathogen sequences as compared to cases that are due to importation. In contrast to generic clustering algorithms, the proposed method explicitly accounts for the fact that during local transmission of a contagious disease the cases are caused by other cases. No pathogen-specific assumptions are needed due to the use of ordinal distances, which allow for direct comparison between the disparate data types. Using simulations, we test the performance of the method in identifying local transmission of disease in large datasets, and assess how sensitivity and specificity change with varying size of local transmission chains and varying overall disease incidence.
    Full-text · Article · Jul 2013 · PLoS ONE
Show more