Surveillance for Shiga Toxin–producing Escherichia coli, Michigan, 2001–2005

A surveillance system used different detection methods to estimate prevalence of Shiga toxin–producing Escherichia coli during 2003–2005 and 2001–2002. More non-O157 serotypes were detected by enzyme immunoassay than by evaluation of non-sorbitol–fermenting E. coli isolates. We therefore recommend use of enzyme immunoassay and culture-based methods.


The Study
The Michigan Department of Community Health implemented a sentinel surveillance system to evaluate blood-containing stool samples from 20 laboratories during April 2003-October 2005 and all stool samples from 2 hospitals during July 2004-October 2005.All suspect non-sorbitol-fermenting E. coli from the remaining laboratories were also examined.
The samples, transported in C&S transport medium (Medical Chemical Corporation, Torrance, CA, USA), were screened for Shiga toxin (Stx) by enzyme immunoassay (EIA) (Meridian BioScience, Cincinnati, OH, USA) after enrichment with gram-negative broth (Remel, Lenexa, KS, USA).EIA is sensitive and specific but cannot detect the Stx2e variant (5), and Pseudomonas aeruginosa can produce false-positive results (6).Samples were cultured on SMAC (Remel) and cefixime-tellurite SMAC (7), and samples from the 2 hospitals were tested for occult blood (Beckman Coulter, Fullerton, CA, USA) before EIA testing.Serotyping (Statens Serum Institute, Copenhagen, Denmark; BD Difco, Franklin Lakes, NJ, USA) and realtime PCR for stx1,2 genes (8) were performed on strains that had positive EIA results, suspect non-sorbitol-fermenting E. coli, and multiple colonies of sorbitol-fermenting (SF) strains that had positive EIA results.For some samples, the EIA result was negative but NSF stx-positive colonies were detected on SMAC, which indicated a falsenegative EIA result.Epidemiologic data were obtained for STEC-positive patients.
To adjust for factors associated with hospitalization, we fit a logistic regression model that included age and symptom variables in the model.The adjusted associations were similar to the crude associations.Hospitalization was more frequent for persons with bloody diarrhea (adjusted odds ratio [OR] 1.8, 95% confidence interval [CI] 1.04-3.08)and HUS (adjusted OR 16.0, 95% CI 2.00-127.47).Also, persons 19-64 (adjusted OR 1.6, 95% CI 1.05-2.59)and >65 (adjusted OR 6.6, 95% CI 2.57-17.15)years of age were hospitalized more frequently than persons ≤18 years of age.

Conclusions
Enhanced detection methods did not significantly increase the year-to-year recovery of STEC.Overall, the observed STEC prevalence decreased slightly over time, similar to the national trend of an overall 42% decrease in STEC O157 incidence during 1996-2004 ( 9).This reduction is likely attributable to numerous factors, including heightened consumer awareness ( 9) and improved screening protocols during food production (10).
Enhanced surveillance did, however, enhance detection of non-O157 serotypes; 4.3% of EIA-positive stools were non-O157 compared with 0.5% of suspect NSF E. coli.Additionally, among the STEC found, 34 (48%) were non-O157 and 37 (52%) were O157 when EIA was used on suspect stools, compared with only 3 (1.6%)non-O157 and 177 (98.3%)O157 among NSF E. coli.Despite enhanced surveillance, STEC prevalence is probably still underestimated, particularly for non-O157 serotypes, because not all ill persons seek medical care and not all laboratories submit suspect stools for evaluation.Nevertheless, in 5 years, our surveillance identified 66 (15%) cases that would have been undetected by conventional methods; 31 (47%) were non-O157.Among those patients for whom data were available, 27 (42%) of 64 were <18 years of age, 22 (43%) of 51 were hospitalized, and 39 (76%) of 51 had bloody diarrhea.Although bloody stool and patient age are poor predictors of STEC infection (11), our analysis demonstrates that screening bloody stool samples improves detection of non-O157, and blood and older age are important predictors of more severe disease, which may be more costly if undetected.
Hospitalization of STEC patients with and without HUS costs an estimated US $30,307 and $4,061 per patient, respectively (12).Therefore, Michigan hospital costs associated with STEC infection likely exceeded $1,119,050 during 2001-2005, as 198 patients were hospitalized and 12 had HUS.Identification of each additional STEC case could have a substantial public health effect in that 1 case may lead to the recognition of an outbreak, which if detected early, could contribute to a cost savings as well as reduced STEC-associated illness.We estimated that the cost to detect each of the 66 additional cases using the EIA ($7 per test including labor) differed considerably when we evaluated screening of all stool samples ($2,426/per positive) versus suspect stool samples ($10/per positive).
No widely available test detects all STEC, and use of multiple methods is not cost-effective.Consequently, we recommend using EIA in conjunction with SMAC culture to recover isolates for molecular characterization and subsequent outbreak investigations.Although occult blood tests did not enhance the sensitivity of STEC recovery, patient data and accompanying epidemiologic information may help identify which samples to test, thereby prevent-ing future outbreaks.Because such epidemiologic information is often not available to laboratory personnel, we suggest that clinical laboratories work with medical administrations to use EIAs as their standard of practice and to facilitate routine availability of such information.Until more sensitive and cost-effective STEC screening methods are available, facilities that cannot implement EIAs should forward stool samples that are suspect, as well as those with positive screening results, to public health laboratories.These laboratories can easily evaluate suspect stools for STEC by EIA or PCR followed by culture of all positive samples to recover the isolate for further characterization.