Epidemiologic analysis and risk management practices for reducing E. coli in irrigation source water supplies and distribution systems.

Our proposal will help the leafy greens (LG) produce industry identify risk management practices and remediation measures that reduce generic E. coli in irrigation water supplies. We will use statistical procedures and epidemiological methods to complete the objectives below. Objective 1: Working in close collaboration with the California and Arizona LG produce industry and allied organizations, finalize the master data file for statistical and epidemiological analyses of objectives 2 through 5. Objective 2: Determine environmental, geographical, structural and operational risk factors for the occurrence of generic E. coli in irrigation water supplies. We will also determine the influence of different diagnostic methods on measured E. coli levels. Objective 3: Identify predisposing environmental, structural, and operational risk factors associated with generic E. coli exceedances in irrigation water supplies. Objective 4: Determine the ability of different mitigation measures to reduce the reoccurrence of an E. coli exceedance in irrigation water supplies. Objective 5: Develop more efficient irrigation water sampling plans for low- to high-E. coli risk source water supplies. Completing these objectives will assist the produce industry comply with microbiological standards for generic E. coli in irrigation water supplies, avoid future E. coli exceedances, and develop more efficient irrigation water sampling plans.

We propose to conduct a detailed statistical and epidemiological analysis of irrigation water data being collected by the produce industry in order to identify risk management practices and remediation measures that reduce generic E. coli in irrigation water supplies. We will also determine the feasibility of changing the rate of water sample testing for low- or high-risk irrigation water sources. Our first task will be to develop an advisory committee comprised of members from organizations such as the Center for Produce Safety, United Fresh Produce Association, Western Growers Association, California LGMA, California Leafy Greens Research Program, Salinas Valley Grower-Shipper Association, Produce Marketing Association, and commercial laboratories. Our 2 year project will complete the following objectives: 

Objective 1. Finalize master data file. In collaboration with CPS, growers and commercial labs, the master data file will be updated, enhanced, and compared against the goals of objectives 2 through 5 to resolve data gaps. Data will include date of sample, generic E. coli value, water chemistry, type of water source, geographical region, sampling location along the irrigation distribution system, and diagnostic method. Additional environmental, operational and structural risk factors will be identified for inclusion in the data file under confidentiality agreements. We will also add precipitation, ambient temperature, and wind velocity for each geographical region. 

Objective 2. Risk factors for measurable E. coli. Using logistic regression (prevalence) and negative binomial regression (concentration), we will identify environmental, geographical, structural and operational conditions and diagnostic methods associated with the occurrence of generic E. coli in irrigation water. This will help the produce industry identify risk factors and diagnostic procedures associated with elevated prevalences and higher concentrations of generic E. coli in irrigation water supplies at different points along their distribution. 

Objective 3. Risk factors for E. coli exceedances. Using mixed effects logistic regression (prevalence), we will identify predisposing environmental, structural, and operational risk factors associated with generic E. coli exceedances in irrigation water supplies. 

Objective 4. Mitigation measures for E. coli. Using mixed effects logistic regression (prevalence), we will test the ability of different mitigation measures used by growers to reduce the reoccurrence of an E. coli exceedance in their irrigation water supplies. 

Objective 5. Efficient use of water testing resources. Using the binomial distribution to calculate the probability to detect E. coli occurrences as a function of water sample testing rate and risk-level of the water source, this analysis will determine if low- or high-E. coli risk source water supplies can be sampled using lower or higher rates of sampling in order to maximize the efficiency of a water testing plan. In addition, we will evaluate if seasonal data suggest that certain times of year can be sampled more efficiently using higher or lower rates of sampling.This project will assist the produce industry comply with current LGMA microbiological standards for generic E. coli in irrigation water supplies, avoid E. coli exceedances by identifying effective remediation measures, and develop efficient irrigation water sampling plans for low- to high-risk irrigation water sources.

1. Working in close collaboration with the California and Arizona LG produce industry and allied organizations, finalize the master data file for statistical and epidemiological analyses of objectives 2 through 5. 

2. Determine environmental, geographical, structural and operational risk factors for the occurrence of generic E. coli in irrigation water supplies. We will also determine the influence of different diagnostic methods on measured E. coli levels. 

3. Identify predisposing environmental, structural, and operational risk factors associated with generic E. coli exceedances in irrigation water supplies. 

4. Determine the ability of difference mitigation measures to reduce the reoccurrence of an E. coli exceedance in irrigation water supplies. 

5. Develop more efficient irrigation water sampling plans for low- to high-E. coli risk source water supplies.

Findings Two datasets were generated. For the first dataset (Set 1, n=44,249) we established three tiers of analysis based on the thoroughness of the data. Tier 1 included all water samples that had a minimum of basic information (date, city, water source). Tier 2 was a subset of tier 1 data, tier 3 was a subset of tier 2, with tier 3 having the most complete information about the water sample. A second dataset (Set 2, n=15,486) was generated from a different cross-section of the produce industry. Similar to Set 1, these data were from across California growing regions covering a wide range of farming operations, water sources and locations. For Tier 1 data from Set 1, the majority of water samples (79% (35,093/44,249)) contained no detectable E. coli, and only 0.86% (380/44,249) of water samples exceeded the single sample maximum (SSM) of >235 MPN/100mL for foliar application. Less than 0.43% of samples exceeded the SSM for non-foliar application of >576 MPN/100mL. This indicates that exceedances in California irrigation water supplies were rare between the years of February 2007 through September 2010. The prevalence of water samples with any level of detectable E. coli (MPN ≥1 /100mL) varied between water sources. About 8% of well samples had detectable E. coli compared to 86% and 48% of canal and reservoir samples, respectively. For Tier 2 data from Set 1, E. coli concentrations (MPN/100mL) varied significantly across season for most of the regions. Levels of E. coli were highest in the two Central Coast regions during the fall season (SeptNov), while predicted concentrations were highest in the Desert region during the summer (June-Aug). We did not detect significant seasonal differences in E. coli concentration in the Central Valley. The occurrence of SSM exceedances for foliar application ranged from 0.45% to 1.65% across season, with highest proportion of exceedances occurring in summer and especially during the fall. When stratified by water source, well and reservoir water samples had a higher odds of exceedance during summer and fall, respectively, compared to exceedances in winter. The odds ratios for the seasonal patterns were just the opposite for canal water, whereby the odds of an exceedance was lower in summer and fall compared to winter, but when mean air temperature was added to the calculation the overall risk of an exceedance was much higher in summer than winter. Adjusted for season, mean air temperature was negatively associated with the odds of an exceedance for reservoirs. In contrast, there was a positive association between mean air temperature and the odds of an exceedance for canal sources (from mostly desert region), which functioned to substantially increase the calculated risk of an exceedance during summer and fall season. Adjusted for season, wind speed was negatively associated with the odds of an exceedance for well water sources but not significant for reservoir and canal sources. For Tier 3 data from Set 1, exceedances were also rare, with 1.1% of all 17,788 samples having >235 MPN/100mL. Concentrations of E. coli were significantly greater in reservoirs compared to wells on the same property in the Central Coast regions. In addition, the odds of an exceedance for foliar application was about three times higher for water taken from a reservoir compared to water taken from a nearby well on the same ranch. Similarly the odds of an exceednace was three time higer for water sampled during summer and fall compared to winter. Twenty-four hour cummulative precipitation was positively associated with the likelihood of an exceedance in wells and reservoirs. For canals in the California desert, the odds of an E. coli exceedance (>235 MPN/100mL) was about 12 times greater in summer compared to winter. This higher risk of an exceedance occurred during the time of year when leafy green produce was for the most part not grown, hence, microbial water quality was at its best during the period when leafy green produce was being grown. For Set 2 data (n=15,486), 0.71% of the samples exceeded the SSM of >235 MPN/100mL for foliar application and only 0.19% of samples exceeded the SSM for non-foliar application of >576 MPN/100mL. Concentrations of E. coli were greatest in the summer and fall for the entire data set. While exceedances for foliar application were rare in well samples (~0.2%) and non-existent in reservoir samples, they were more frequent in canal samples (2.5%). Exceedances were more common in the summer and fall. The proportion of water samples that exceeded the SSM for foliar application ranged from 0% to 2.8% depending on the point of entry. Using simulation methods to evaluate the advantages and disadvantages of alternative irrigation sampling plans, we determined that when the amount of E. coli across a season had minimal variability for a location, a relatively small number of samples were needed to correctly estimate water quality with high probability. Conversely, in scenarios with occasional high concentrations (i.e. sporadic spikes of E. coli), it is more difficult to capture these rare events with a sampling plan and thus a higher number of samples are needed to correctly classify water quality for a location. Reducing the rate of water monitoring under these high variance conditions may substantially reduce the probability that a grower correctly classifies their location’s water quality. As an alternative, the grower can potentially increase the volume processed per assay and substantially increase the likelihood of detecting E. coli in irrigation water, especially for well samples. Volumes of one liter or more appear to capture most of the benefit of using a higher volume, hence, private biotechnology firms might consider partnering with the produce industry to develop the necessary platform that conveniently and cost-effectively processes one liter samples similar to how 100mL are processed today. Recommendations Based on two large datasets totaling about 60,000 data points that represent a very large number of produce growers from throughout California from various sources of irrigation water and for all seasons of the year, the rate of exceedances was uncommon for most locations and sources of irrigation water. Using this low rate of exceedance as a justification to reduce the rate of monitoring for locations with persistent high water quality should be an active discussion point between produce industry, allied associations, scientists familiar with epidemiological and risk-based approaches, and regulatory agencies. To inform such a discussion it would be prudent to better understand the linkages or lack thereof between generic E. coli and the various foodborne pathogens of concern to the produce industry. It is possible that under certain specific risk factor scenarios there are linkages between high levels of indicator E. coli and pathogens such as E. coli O157:H7 and Salmonella, but in the absence of such risk factors minimal correlation likely exists for generic E. coli and pathogens associated with foodborne illness. This project has begun to identify some of these potential risk factors, but prospective studies need to be conducted in close collaboration with individual growers to verify proximity, structural and environmental risk factors for not just generic E. coli, but produce food safety pathogens. In our opinion, key to developing more cost-effective water quality monitoring plans for produce food safety is to adopt the use of larger volumes of water per assay. For many regions and sources of water such as wells the use of 100mL volumes does not provide reliable estimates of generic E. coli. We propose that the U.S. produce industry thoroughly evaluate the use of a standardized 1-liter assay that can simultaneously increase the accuracy of monitoring data and possibly set the foundation for more site-specific rates of sampling based on risk levels for bacterial contamination. As the industry moves in this direction, market forces will likely drive commercial innovation to develop an inexpensive 1-liter assay. Lastly, sampling water at critical periods of the growing season and during periods of high vulnerability for microbial water quality, when taken together with higher volumes per assay, could form the basis of a more strategic and focused protocol for monitoring irrigation water quality and insuring produce food safety.