Estimated Incidence of Antimicrobial Drug–Resistant Nontyphoidal Salmonella Infections, United States, 2004–2012

Salmonella infections are a major cause of illness in the United States. The antimicrobial agents used to treat severe infections include ceftriaxone, ciprofloxacin, and ampicillin. Antimicrobial drug resistance has been associated with adverse clinical outcomes. To estimate the incidence of resistant culture-confirmed nontyphoidal Salmonella infections, we used Bayesian hierarchical models of 2004–2012 data from the Centers for Disease Control and Prevention National Antimicrobial Resistance Monitoring System and Laboratory-based Enteric Disease Surveillance. We based 3 mutually exclusive resistance categories on susceptibility testing: ceftriaxone and ampicillin resistant, ciprofloxacin nonsusceptible but ceftriaxone susceptible, and ampicillin resistant but ceftriaxone and ciprofloxacin susceptible. We estimated the overall incidence of resistant infections as 1.07/100,000 person-years for ampicillin-only resistance, 0.51/100,000 person-years for ceftriaxone and ampicillin resistance, and 0.35/100,000 person-years for ciprofloxacin nonsusceptibility, or ≈6,200 resistant culture-confirmed infections annually. These national estimates help define the magnitude of the resistance problem so that control measures can be appropriately targeted.


Background
We describe the use of a Bayesian hierarchical model (BHM) to estimate resistance incidence. We used data on isolations of Salmonella serotypes from the Laboratory-based Enteric Disease Surveillance (LEDS) and resistance proportions from the National Antimicrobial Resistance Monitoring System (NARMS). The yearly surveillance data of 48 states (excluding Alaska and Hawaii) from both LEDS and NARMS are volatile due to sampling variation and may be biased due to underreporting. For NARMS data, many states have small numbers of isolates due to the sampling scheme (1 in 20), particularly for Heidelberg and less common serotypes. The estimation of resistance proportions by state and year is unreliable due to the small sample size. BHM provides a framework to mitigate the issues based on partial pooling (borrowing strength) from structured data, e.g. neighboring states may exhibit similarity in incidence and resistance proportions. BHM reduces variability in estimates by spatial smoothing of geographically related surveillance data. It provides a flexible approach by accounting for structured and non-structured variances in the data.
Another advantage of BHM is its utility in handling missing data. Data were missing from both surveillance systems, especially for some combinations of serotypes and resistance types. For example, not all states reported or submitted isolates of the major serotypes every year, thus infection incidence rates and resistance proportions were not available for the states that did not report or submit isolates for the year. In Bayesian statistics, missing values are treated as unknown parameters and are estimated in the same manner as other parameters in the model, and Bayesian estimation of missing values takes into account the uncertainty of parameter estimation. We set the normal distribution variance parameter, equal to 2 to impose a temporal autocorrelation between the resistance proportion of a state in a given year and that of the preceding year; that of the first year is set to be normal variate of zero mean to anchor the posterior.
, in equation 2 is the structured state spatial random effect reflecting a time-varying neighborhood effect (2).
where u -s denotes states adjacent to state s. Adjacency is defined as sharing a border with the focal state s, , the number of neighboring states of state s. For τu ,we adopted a weak gamma prior proposed by Kelsall and Wakefield (1) This prior assumes that the spatial random effects for a single adjacent state has a standard deviation centered around 0.05 with 1% probability being smaller than 0.01 or larger than 2.5 (1).
Finally, φs,t is state-time interaction term of normal variate ,~ (0, ) After experimenting with different options, we settled with a fixed τφ equal to 2 to balance the amount of shrinkage from observed values across the various states and years. For missing Ts,t, we assumed them as either the mean of the known submission rates (estimated from submitted rates over the years when submission occurred) or as 1 if the former was not available.
In the latter case, the influence of the assumed values (one isolate) would be minimized.

LEDS model of Salmonella incidence:
The standard model for incidence based on count data is the Poisson distribution (3).
However, counts and incidence rates of different serotypes varied drastically from year to year ( Fig. 2). We found that use of a Poisson model was inadequate to capture the variability observed in the data and resulted in estimates of little, if any, shrinkage of observed values. To capture the observed variability in yearly observed incidence rates, we adopted a truncated normal distribution for the incidence rates (/100,000) Is,t (truncated for Is,t <0) ,~( , , 0.1) We adopted a similarly structured model as the NARMS model described above We set as 5 to impose a temporal autocorrelation of incidence rates of state s to be related to that of the preceding year; that of the first year was set to be normal variate of zero mean.

Adjustment for not fully serotyped LEDS data
We applied serotype-resistance data to all LEDS isolates, including not fully serotyped isolates, after adjustment for incomplete serotyping for all 48 states. For each state, we imputed serotypes for LEDS isolates that were not fully serotyped based on the observed proportions of five serotype categories (Enteritidis, Typhimurium, Newport, Heidelberg, and other) among fully serotyped isolates over the 9 years.

Adjustment for underreporting to LEDS by Florida
The reported Salmonella incidence rates in Florida were much lower than those from states in the region, indicating significant underreporting from the state. We only adjusted for underreporting by Florida for overall nontyphoidal Salmonella and the four major serotypes.
where denotes the state effect, the year effect, and , the state-year interaction.
The following priors were used Note, we used a large value 10 as the precision parameter for , to shrink Florida estimates more effectively toward the regional mean.
The adjusted estimates of incidence rates in Florida were closer to the means from the six southern states. We used the adjusted incidence rates in Florida (Appendix Table) to replace the observed values as inputs to run the BHM for estimating resistance incidence.

Summary posterior estimates of overall nontyphoidal Salmonella:
Posterior estimates of resistance proportion, incidence rates, and resistance incidence of Similarly, the posteriors of clinically important resistance for four serotype categories (Enteritidis, Typhimurium, Newport, Heidelberg) were derived from the aggregated joint distributions of the posteriors of the corresponding measures of the mutually exclusive resistance categories (i.e., resistance to ceftriaxone, nonsusceptibility to ciprofloxacin, and resistance to ampicillin).

Posterior estimates vs. observed values:
We assessed the shrinkage of posterior resistance proportions (predicted) vs. crude proportions (observed) related to the number of isolates tested. Appendix Figure 1 shows the shrinkage for ampicillin resistance among isolates of overall nontyphoidal Salmonella, isolates of the four major serotypes, and other fully serotyped isolates. As part of model fitting, we plotted predicted estimates and observed values of resistance proportion, Salmonella infection incidence,

Software
The models were run in R (4) with R2WinBUGS package (5)