The Impact of Local Environmental Health Capacity on Foodborne Illness Morbidity in Maryland

Abstract

Objectives. We evaluated the relationship between local food protection capacity and service provision in Maryland's 24 local food protection programs (FPPs) and incidence of foodborne illness at the county level.

Methods. We conducted regression analyses to determine the relationship between foodborne illness and local FPP characteristics. We used the Centers for Disease Control and Prevention's FoodNet and Maryland Department of Health and Mental Hygiene outbreak data set, along with data on Maryland's local FPP capacity (workforce size and experience levels, budget) and service provision (food service facility inspections, public notification programs).

Results. Counties with higher capacity, such as larger workforce, higher budget, and greater employee experience, had fewer foodborne illnesses. Counties with better performance and county-level regulations, such as high food service facility inspection rates and requiring certified food manager programs, respectively, had lower rates of illness.

Conclusions. Counties with strong local food protection capacity and services can protect the public from foodborne illness. Research on public health services can enhance our understanding of the food protection infrastructure, and the effectiveness of food protection programs in preventing foodborne illness.

Protecting the food supply requires diligence from farm to fork—from ensuring that our produce is grown in sanitary conditions to inspecting restaurants to ensure that food service workers are using proper hand-washing techniques. However, foodborne outbreaks continue to dominate the media headlines. Approximately 48 million cases of foodborne illness (FBI) occur annually,¹ with 66% of foodborne outbreaks associated with restaurants and 9% with catered events.² Numerous studies demonstrate that a large percentage of outbreaks are related to poor food-handling procedures.^2,3,4 Shigella, hepatitis A, and norovirus, among many other infections, can all be readily transmitted to restaurant patrons through improper hand washing by infected food handlers.⁵ In Maryland, where restaurant sales were projected to reach $8.7 billion in 2010⁶ and nearly 55% of residents eat in sit-down restaurants on a weekly basis,⁷ 63% of foodborne outbreaks reported to the Maryland Department of Health and Mental Hygiene (DHMH) occurred in restaurants.⁸

To prevent these outbreaks from occurring, a strong public health infrastructure is essential. In Maryland, ensuring that restaurants provide safe meals to consumers is the primary role of the state's 24 county-level food protection programs (FPPs).⁹ Housed in the environmental health division of the county-level health department, these programs conduct routine inspections of restaurants (hereafter referred to as food service facilities [FSFs]), public notification programs (such as posting FSF closures in local media outlets), educational programs for both FSF workers and county residents, and collaboration with county-level legislators to develop and enforce food protection regulations, such as certified food manager programs. The ability of local FPPs to conduct these tasks and provide services is contingent on a robust infrastructure and strong internal capacity—that is, structural inputs, such as workforce size and internal budgets, that allow the FPP to deliver services such as FSF inspections.

In light of new data indicating that FBI costs $156 billion a year,¹⁰ health departments are even more accountable to the public to reduce illnesses and their significant human and financial costs. With the resurgence of performance measurement at all levels of government, the need to measure local FPP capacity to protect the food supply and demonstrate the effectiveness of food protection programs, through measuring the impact of food protection programs on key public health outcomes, is more essential than ever to ensure sustained financial and political support for local programs. However, despite the longstanding importance of these food protection activities, in Maryland the relationship between county-level food protection capacity and services and FBI cases and outbreaks has not been evaluated. Using public health services and systems research methods, we sought to evaluate this relationship.

METHODS

We collected data on both the capacity of Maryland's 24 FPPs and the services and programs they provided in the fall of 2007 using the Maryland Environmental Health Capacity Survey, a 16-question written survey. This survey, developed with input and review by local environmental health practitioners and Johns Hopkins researchers, garnered a 92% response rate (22 out of 24 counties) and included data from 2001 through 2006. The survey contained questions on structural inputs, such as the size of the FPP workforce, employee experience, and annual budgets as well as the number of FSFs, the total number of required and completed FSF inspections, the existence of public notification programs for FSF closures and poor inspection scores, and certified food manager programs.

We acquired data on cases and outbreaks of FBI from 2 surveillance databases: the Foodborne Diseases Active Surveillance Network (FoodNet) and the Maryland DHMH outbreak data set. Run by the Centers for Disease Control and Prevention (CDC), FoodNet conducts active surveillance for 7 bacterial and 2 parasitic foodborne diseases in 10 states, including Maryland. The DHMH outbreak data set is an internal, passive surveillance database maintained by the state health department; outbreaks are reported to the DHMH by county environmental health divisions, physicians, and laboratories. The dependent variable throughout this analysis was FBI. To improve statistical power, we aggregated all FBI for all years into 1 total illness count per county.

To examine the relationship between capacity and cases of FBI at the county level from the 2 data sets, we used negative binomial regression, controlling for county population size as an offset in regression analyses. Results from a continuous analysis (e.g., compliance rates on a continuous scale) are presented in Table 1. The categorical analysis (e.g., compliance rates subdivided into 3 categories, with counties ranked as having low, moderate, or high compliance rates) are presented in Table 2. We obtained unadjusted bivariate incidence rate ratios for each analysis and subsequently converted them into estimated percentage reductions in incidence, which are presented in Tables 1 and 2. The estimated percentage reductions in incidence for each of the independent variables are described in the Results section. Finally, we conducted sensitivity analyses that examined different time periods in the FoodNet database, restricted larger counties from the data set, and restructured certain categorical variables to examine the robustness of the results; these analyses yielded no significant differences in reductions. We used the statistical package Stata 9.0 (StataCorp LP, College Station, TX) for this analysis.

TABLE 1.

Estimated Reductions in Incidence of Foodborne Illness, by Dichotomous and Continuous Food Program Capacity Variables: Maryland, 2001–2006

	FoodNet Data Set		DHMH Outbreak Data Set
Variables	IRR (95% CI)	Reduction in Incidence, %	IRR (95% CI)	Reduction in Incidence, %
Dichotomous variables
Certified food manager	0.92 (0.54, 1.56)	8.3	0.61 (0.26, 1.43)	39.3
Inspections published	0.66 (0.42, 1.06)	33.6	0.59 (0.27, 1.34)	40.2
Closures published	0.84 (0.53, 1.32)	16.3	0.73 (0.34, 1.54)	27.1
Continuous variables
FSF compliance rate	0.88 (0.33, 2.35)	12.0	0.17* (0.04, 0.65)	83.0
Food FTE	0.99 (0.96, 1.02)	1.0	0.98 (0.93, 1.02)	2.0
Experience food FTE	0.99 (0.97, 1.02)	1.0	0.96 (0.92, 1.01)	4.0
Food budget	1.00 (0.99, 1.00)	0.0	1.00 (0.99, 1.00)	0.0

Note. CI = confidence interval; DHMH = Maryland Department of Health and Mental Hygiene; FoodNet = Center for Disease Control and Prevention active foodborne illness database; FSF = food service facility; FTE = full-time equivalent; IRR = incidence rate ratio. Variables are defined as follows: “certified food manager,” existence of certified food manager program in county; “inspections published,” county publication of results of FSF inspections; “closures published,” county publication of FSF closures; “FSF compliance rate,” number of FSFs inspected divided by the number that should have been inspected at county level; “food FTE,” total number of food program full-time equivalent positions in county; “experience food FTE,” total years of experience of food program full-time equivalent positions in county; “food budget,” total annual budget for the county food program.

*P < .05.

TABLE 2.

Estimated Reductions in Incidence of Foodborne Illness, by Categorical Food Program Capacity Variables: Maryland, 2001–2006

	FoodNet Data Set		DHMH Outbreak Data Set
Variables	IRR (95% CI)	Reduction in Incidence, %	IRR (95% CI)	Reduction in Incidence, %
FSF compliance ratea
Low (Ref)	1.00		1.00
Moderate	0.90 (0.53, 1.51)	10.5	0.45* (0.21,0.97)	55.0
High	1.0 (0.60, 1.66)	−0.2	0.45* (0.22,0.95)	54.7
Food FTEb
Low (Ref)	1.00		1.00
Moderate	0.50** (0.33, 0.77)	49.6	0.88 (0.38, 2.01)	12.5
High	0.57** (0.38, 0.86)	43.2	0.64 (0.29, 1.43)	36.0
Experience food FTEc
Low (Ref)	1.00		1.00
Moderate	1.19 (0.71, 2.01)	−19.4	0.51* (0.26, 0.98)	49.5
High	0.91 (0.54, 1.54)	9.0	0.24*** (0.12, 0.47)	76.2
Food budgetd
Low (Ref)	1.00		1.00
Moderate	0.47** (0.30, 0.74)	52.6	0.49 (0.21, 1.14)	50.7
High	0.61* (0.40, 0.93)	38.9	0.52 (0.23, 1.16)	48.1

Note. CI = confidence interval; DHMH = Maryland Department of Health and Mental Hygiene; FoodNet = Center for Disease Control and Prevention active foodborne illness database; FSF = food service facility; FTE = full-time equivalent; IRR = incidence rate ratio. Variables are defined as follows: “certified food manager,” existence of certified food manager program in county; “inspections published,” county publication of results of FSF inspections; “closures published,” county publication of FSF closures; “FSF compliance rate,” number of FSFs inspected divided by the number that should have been inspected at county level; “food FTE,” total number of food program full-time equivalent positions in county; “experience food FTE,” total years of experience of food program full-time equivalent positions in county; “food budget,” total annual budget for the county food program. Categorical rankings use the lowest category as the comparison rank.

^aCategorical rankings for the FSF compliance rate variable are as follows: low, < 58.5%; moderate, 58.5%–80%; high, > 80%.

^bCategorical rankings for the food FTE variable are as follows, in number of food FTE employees: low, < 1.45; moderate, 1.45–4.79; high, > 4.79.

^cCategorical rankings for the experience food FTE variable are as follows, in years of experience per food FTE in 2006: low, < 5.73; moderate, 5.73–16.8; high, > 16.8.

^dCategorical rankings for the food budget variable are as follows, in actual dollars: low, < $161 612; moderate, $161 612–$331 598; High, > $331 598.

*P < .05; **P < .01; ***P < .001.

RESULTS

Table 1 contains unadjusted, bivariate dichotomous and continuous analyses and Table 2 contains unadjusted, bivariate categorical analyses. The categorical rankings themselves are presented in footnotes to Table 2.

Certified Food Managers

Research has demonstrated that requiring certified food managers to be present in FSFs reduces the occurrence of critical violations during restaurant inspections and the rate of FBI outbreaks.^11,12 In Maryland, 5 counties had laws requiring certified food managers. When analyzing FoodNet data, counties with a certified food manager requirement had a lower incidence of FBI (8.3% reduction) than did counties without this requirement. Identical analyses with the DHMH outbreak data set indicated that counties with a certified food manager requirement had a 39% reduction in FBI compared with counties without this requirement.

Public Notification

Twenty-three percent of Maryland's FPPs notified the public, through posting information in town, city, or county newspapers or on the county health department Web page, when an FSF in their county has a poor inspection report; 27% notify the public when FSFs are closed for failing an inspection, which has been shown to be effective in reducing FBI hospitalizations.^13,14 For both the FoodNet and DHMH outbreak data sets, counties that notified the public of poorly performing FSFs and FSF closures had lower rates of FBI (33.6%–40.2% and 16.3%–27.1% lower, respectively) than did counties that did not provide this service.

Food Service Facility Inspection Compliance Rate

All counties in Maryland are required by law to conduct inspections of FSFs in their jurisdiction at a set frequency. The ability of a county to meet this requirement can be calculated as the number of facilities inspected divided by the number of facilities that should have been inspected, defined here as the compliance rate. Analyses of FoodNet data using a continuous compliance rate indicated that with every percentage increase in the compliance rate, the FBI incidence decreased by 12%. With categorical analysis, counties with a moderate rate of compliance had a 10% lower rate of FBI than did counties with a low rate of compliance, although counties with high compliance had a higher rate of FBI than did counties with low compliance.

Continuous analyses with the DHMH outbreak data set yielded similar results, with a statistically significant 83% decrease in incidence with every 1% increase in compliance. Categorical data revealed a statistically significant reduction in FBI (P < .05) among counties reporting moderate and high rates of compliance (55.0% and 54.7% reduction, respectively) compared with counties with the lowest rate of compliance.

Workforce

FPPs are staffed by registered sanitarians, who provide the services offered by the program. On average, there were 5.5 full-time equivalent registered sanitarian positions per county, with a minimum of less than 1 full-time equivalent position per county and a maximum of 26 per county. Continuous analysis of the FoodNet data revealed that for every 1 full-time equivalent increase in a county's FPP staff, a mildly protective effect was observed. This protective effect was strengthened with categorical analysis. Counties with a moderate number of FPP full-time equivalent positions, as well as counties with a high number of such positions, had a statistically significantly lower risk of FBI (P < .01) than did counties with a low number of employees, although counties with a moderate staffing level had lower rates of FBI than did counties with a high staffing level.

Data from the continuous analysis of the DHMH outbreak data set were again closely aligned with FoodNet results. Further, categorization revealed that counties with a moderate number of full-time equivalent positions had a 12% lower rate of FBI than did counties with a low number of employees, and counties with the most full-time equivalent positions had a 36% lower rate of FBI than counties with the fewest full-time equivalent positions.

Experience

Along with the number of employees working in an FPP, their years of experience working in the field can potentially influence the capacity of the program, with more experienced employees often being more efficient and effective at their jobs. Statewide, county FPP employees had an average of 12.8 years of experience per full-time equivalent position, with a minimum of 1 year of experience per full-time equivalent position and a maximum of 42 years of experience. Analysis of this measure as a continuous variable yielded similar results across all data sets; for every year increase in experience per full-time equivalent position, the incidence of FBI decreased slightly. When this variable was categorized, however, differences emerged from the 2 data sets.

In the categorical analysis of FoodNet data, counties with moderately experienced employees had a 19% higher risk of FBI than did counties with the least experienced employees. Counties with highly experienced employees had a 9% lower risk of FBI than did counties with the least experienced employees. Analysis of the DHMH outbreak data set revealed that counties with employees with a moderate level of experience had a 49% lower rate of FBI than did counties with the least experienced employees, and counties with employees with the most experience had a 76% lower rate of FBI than did counties with employees with the least experience.

Budget

On average, counties had an annual operating FPP budget of $432 376, with a minimum of $32 514 per county and a maximum of $1 527 980 per county. For these and all analyses, because population size was used as an offset for the regression analyses, we used gross budget numbers. Continuous analysis of the FoodNet data indicated that food budgets did not affect the incidence of FBI at the county level. When we categorized the budget data, however, the impact of budget on FBI changed. Counties with moderate budgets had a statistically significant 53% lower incidence of FBI (P < .01) than did counties with small budgets, whereas counties with the largest budgets had a statistically significant 39% lower rate of FBI (P < .05) than did counties with small budgets. As with the compliance rate, counties with a moderate level of capacity (moderate budget ranking) performed better (i.e., had a larger percentage reduction in incidence of FBI) than did counties with a higher level of capacity (high budget ranking).

Categorization of the budget data using the DHMH outbreak data set indicated that counties with moderate budgets had a 51% decrease in the incidence of FBI compared with counties with a small budget, whereas counties with the highest budget had a 48% decrease in the incidence of FBI compared with counties with small budgets.

In summary, county-level capacity has the potential to affect the incidence of FBI. Further, for both data sets, counties with moderate FSF compliance rates and moderate food program budgets (moderate capacity) had larger estimated reductions in incidence than did counties with high levels of capacity. By contrast, in both data sets counties with high levels of experience per full-time equivalent position (high capacity) had larger estimated reductions in incidence than did counties with moderate levels of capacity.

Although some of these reductions in incidence are striking, they can be difficult to conceptualize, as reported cases of FBI are still relatively rare in the general population. Consequently, to illustrate this point, we evaluated the impact of 2 FPP dichotomous capacity variables (county publication of results of FSF inspections and existence of certified food manager program) on the potential estimated percentage reduction in incidence (displayed in Tables 1 and 2) using baseline rates of FBI (Table 3).

TABLE 3.

Estimated Impact of 2 Food Program Capacity Variables on Cases of Foodborne Illness, by Region and Statewide: Maryland, 2001–2006

Region	FoodNet Data Set			DHMH Outbreak Data Set
	Baseline Cases of Foodborne Illness per 100 000	Estimated Cases of Foodborne Illness per 100 000		Baseline Cases of Foodborne Illness	Estimated Cases of Foodborne Illness per 100 000
		Inspections Published	CFM		Inspections Published	CFM
Capital	34.6	23.0	31.7	15.0	9.0	9.1
Central	65.8	43.7	60.3	58.0	34.7	35.2
Eastern Shore	45.8	30.4	42.0	134.8	80.6	81.8
Southern	30.0	19.9	27.5	60.4	36.1	36.7
Western	24.6	16.3	22.6	50.3	30.1	30.5
Statewide	35.5	23.6	32.6	31.1	18.6	18.9

Note. CFM = certified food manager; DHMH = Maryland Department of Health and Mental Hygiene; FoodNet = Center for Disease Control and Prevention active foodborne illness database. “Inspections published” means that county publishes results of food service facility inspections. The baseline case rates are incidence rates calculated with foodborne illness counts from 2001 through 2006 from the FoodNet database and the DHMH database and a county-level population estimate from the 2000 US Census. The estimated cases of foodborne illness were calculated by multiplying the baseline case rate by the estimated reduction in incidence and subtracting that from the original baseline rate.

For Table 3, we randomly selected 1 county from each region of the state and calculated the baseline rate of FBI (for the FoodNet and DHMH outbreak data set separately). Then, using results from the estimated reductions in FBI incidence for these 2 independent capacity variables, we estimated the total number of cases of FBI. For example, there were 34.6 estimated cases of FBI annually at baseline in the Capital Region in the FoodNet data set, but if a county in that region had or were to implement a public notification program for poor FSF inspections, it could expect an estimated 33.6% reduction in the incidence of FBI, which would result in the number of cases experienced by the county annually to drop to 23 cases per 100 000, or a drop of 11.6 cases per year.

Figure 1 illustrates the estimated reductions in cases of FBI when both the dichotomous and categorical variables are applied. When appropriate, we used the high capacity ranking for each variable. For example, for the food program full-time equivalent variable, which measured the total number of full-time equivalent positions that inspected FSF, we included only counties that had the highest number of FPP employees in the figure. For the FoodNet data set, counties with the largest number of food program full-time equivalent positions had the largest estimated reduction in cases. On the other hand, for the DHMH outbreak data set, counties with the most FPP employee experience had the greatest reduction in FBI incidence.

FIGURE 1.

The impact of food program variables on reductions (per 100 000) in cases of foodborne illness: Maryland, 2001–2006.

Note. CFM = certified food manager; FoodNet = Centers for Disease Control and Prevention active foodborne illness database; FTE = full-time equivalent; MD DHMH = Maryland Department of Health and Mental Hygiene outbreak data set. For the “Food FTE,” “Food budget,” “Experience/food FTE,” and “Compliance rate” variables, high capacity rankings were included. For the “Inspections Published,” “Closures Published,” and “CFM” variables, only counties with these programs were included.

Through use of this technique, some FPP capacity variables were more influential in reducing the incidence of FBI than were others. Although these figures are simply estimates of possible reductions in FBI incidence, they illustrate the potential impact these services (such as providing the public with information about “bad actor” FSFs in their county) can have on the burden of FBI in a particular county.

DISCUSSION

This study indicates that high levels of capacity within local environmental health divisions and FPPs have the potential to positively affect the public's health. Counties with greater capacity, in the form of larger workforces and budgets, had reduced rates of FBI compared with counties with less capacity. Additionally, this study addressed multiple facets of the food protection system, from FSF inspections to the impact of certified food manager regulations to the effect of public notification programs. Further, it highlights the importance of measuring inputs, such as capacity, into the public health system, evaluating the effectiveness of the services provided, and determining the impact of public health services and programs on the public's health.

We performed this study using data from 2 FBI surveillance databases and 1 population survey, all of which had inherent limitations. FoodNet contains only bacterial and parasitic sporadic cases of FBI; viral illnesses are lacking, and no data are collected on where the illness was acquired (at an FSF or at home), making it impossible to fully assess the relationship between FSF inspection capacity and all cases of FBI at regulated facilities at the local level. The DHMH outbreak data set is a passive system and, although data are audited, they are not audited as routinely as are the FoodNet data set. Additionally, in this analysis, we included all cases identified as potentially foodborne and related to activities and services provided by local environmental health divisions, although these inclusion criteria might upwardly bias the results of this analysis.

In general, timely and accurate detection of FBI is a challenge, as most persons in the United States do not report mild gastrointestinal symptoms to local, state, and federal health agencies, which results in cases not being included in estimates of illness. As a result, the total number of cases reported in both data sets probably grossly underestimates the true disease burden in the population. Additionally, the reporting of FBI to the health department (or to a physician) is likely biased—if FSF patrons believe food eaten outside of the home made them ill, they may be more likely to report the illness than if a meal prepared at home made them ill. Further, Jones et al. cite

delayed reporting, limited resources and competing priorities in health departments, limited collection or testing of specimens, ill persons that do not seek health care, and lack of cooperation from clinicians and laboratories

as issues that impede the full collection of FBI data.^2(ps297) All of these issues affect the results of this analysis, although it is presumed that increased reporting of FBI in Maryland and nationwide would support this analysis and make the findings more universally statistically significant.

Data from the Maryland Environmental Health Capacity Survey was self-reported by local environmental health division directors and FPP staff, which raises the possibility of recall bias and reporting bias. Additionally, although the survey response rate was 92%, there are only 24 counties in the state. Consequently, the 22-county sample size rendered most regression outputs statistically insignificant. Exploratory multivariate analyses to explore the relationship between local FPP capacity and cases of FBI were also conducted but did not reveal universal reductions in foodborne illness incidence, again because of the small sample sizes. However, most exploratory models did demonstrate a reduction in incidence of FBI for counties with moderate and high levels of capacity.

Findings from this study suggest that better-resourced counties might be more effective in preventing illness than are their counterparts with less capacity. Further, we obtained significant findings for several of the analyses, but level of significance, as well as the size of the estimated reductions, differed across the 2 data sets; reductions in incidence were more pronounced in the DHMH outbreak data set than in FoodNet. There are several potential reasons for these differences. First, viruses, such as norovirus, are the most commonly reported outbreak-associated pathogens and are often tied to poor food-handling practices in FSFs, but they are included only in the DHMH outbreak data set.^1,15 Second, the DHMH outbreak data set captures cases of FBI within outbreaks, and most outbreaks occurred in regulated FSFs. It is therefore possible that the DHMH outbreak data set, which captures viruses and FSF-associated outbreaks, more accurately reflects the impact of environmental health divisions on the disease burden in the population than does FoodNet, which contains only sporadic bacterial and parasitic cases of FBI. Finally, it is important to note that the FoodNet data set is maintained and audited by the CDC and contains only laboratory-confirmed cases, whereas the DHMH outbreak data set is a passive system and includes both laboratory-confirmed and suspected outbreak-associated cases.

Another notable trend emerged in this analysis. In the FoodNet analyses, counties with moderate compliance rates, finances, and FPP staffing often had a greater reduction in FBI incidence than did the higher-capacity counties, and this relationship was statistically significant in several analyses. Again, this finding runs contrary to the hypothesis that counties with the greatest capacity are the most effective in reducing FBI. One possible explanation is that higher-capacity environmental health divisions in Maryland are, on the whole, in the wealthier counties and consequently have greater resources in terms of raw numbers of employees and financial support. In terms of population size, however, these counties had smaller workforces and budgets than did counties with fewer raw resources. Therefore, the impact those resources have on the burden of FBI might be masked by the sheer number of people living in those counties.

As we move even further into an age of accountability and performance measurement, quantifying the relationship between public health services and health outcomes will become more important, particularly as funds for public health services continue to diminish at the state and local levels. This first-of-its-kind study quantified the potential impact of local food protection capacity on an important and costly public health outcome, FBI. Additionally, it provides the groundwork for further exploration of FoodNet data coupled with state-level surveillance data to examine the relationship between local capacity and FBI.

This analysis not only highlights a method to measure public health services, but also demonstrates that strong local food protection programs in Maryland appear to protect the food supply more effectively than do less robust programs. It is therefore vitally important that local environmental health and food protection programs have sufficient resources to inspect FSFs, investigate outbreaks, and protect the public from foodborne threats. Further, these findings can serve as a foundation for future research to examine the relationship between public health structure inputs and services and public health outcomes in a multivariate context with a larger population. Finally, public health services and systems research methods, as described in this study, are the key to quantifying the importance of public health activities; they should therefore be supported and serve as a foundation for future research linking public health services to health outcomes.