Tetracycline Resistance Genes and Respiratory Pathogens
Tetracycline Resistance Genes and Respiratory Pathogens
The Apriori algorithm created a list of the 25 most frequent item sets found throughout the respiratory results database (Table 1). The largest measurements of support included TRGs (40.5%), S. pneumoniae (36.3%), H. influenzae (28.7%), and M. catarrhalis (25.1%). S. pneumoniae and H. influenzae were co-detected with TRGs 19.8% and 15.1% of all positive tests results. MRSA and S. aureus were detected 12.4% and 10.0% of all positive results, respectively. 6.5% of all positive tests included co-detection for MRSA and TRGs and 4.1% included simultaneous detection between S. aureus and TRGs.
From the frequent item set results, the Apriori algorithm identified the 25 association rules with the highest levels of support (Table 2). The top of the table includes the rules {S. pneumoniae} = > {Tetracycline}, {H. influenzae} = > {Tetracycline}, and {H. influenzae, M. catarrhalis} = > {S. pneumoniae} with 19.8%, 15.1%, and 7.2% support, respectively. Each of the highlighted association rules indicates a pathogen of interest or TRG that all have confidence measurements over 50.9% and lift values over 1.3. This confidence threshold establishes that, given the antecedent of a rule, the consequent is found in at least 50.9% of the cases. Since each rule had a lift measurement over 1, they were all considered to be rules of interest. From these results, and from literature associating TRGs with specific gram-positive bacteria, we sought to elucidate temporal co-detection patterns to better understand how patient age and geography relate to TRG co-detection.
Figure 1 displays a scatterplot including the frequencies of respiratory pathogens co-detected with TRGs by year. It is clear that the frequency of co-detection between each pathogen and TRG increased with time. From January 2010- August 2013, the percentage of gram-positive bacteria S. aureus, MRSA, and S. pneumoniae co-detected with TRGs rose 43%, 26%, and 27%, respectively. From January 2010- August 2013, the frequency of gram-negative bacteria H. influenzae and M. catarrhalis co-detected with TRGs rose 32% and 24%, respectively. Over each year, the rate of TRGs co-detected with these five potential pathogens rose a minimum of 3.9%.
(Enlarge Image)
Figure 1.
The frequency of TRGs Detected with Respiratory Pathogens Increased from 2009–2013. The scatterplot shows the rise in co-detection between TRGs and respiratory pathogens in recent years. The data is aggregated by week. Loess smoothing curves pass through the data points and are shaded by a 95% confidence interval.
Figure 2 displays the time-series co-detection frequency of TRGs with each pathogen segmented by 4 age groups. The patient sample size and total number of positive tests for each pathogen are included in Additional file 1: Tables S1-S7 http://www.biomedcentral.com/1471-2334/14/460/additional. In each age group, the rates of co-detection between TRGs and each pathogen increasd with time. Older patients showed a larger co-detection rate than younger patients. For H. influenzae, M. catarrhalis, and S. pneumoniae, patients (0–2] had the lowest rate of co-detection, followed by patients (2–13] years of age, then patients (50–100] years of age, concluded by patients (13–50] years of age, which had the highest rate of co-detection. In 2013, there was an average 33%, 22%, and 26% difference in co-detection rate from children aged (0–2] to patients age (13–50] for H. influenzae, M. catarrhalis, and S. pneumoniae, respectively. In 2013, an alarming 82% patients of age (13–50] had TRGs co-detected with S. pneumoniae. For S. aureus and MRSA, the age group (13–50] had the highest rate of co-detection between TRGs and the designated pathogen, but all other age groups had similar and co-detection frequencies over time.
(Enlarge Image)
Figure 2.
The Frequency of TRGs Detected with Respiratory Pathogens Increased in Each Age Group over Time. Each graph displays the co-detection frequency for each age group and pathogen. The data is aggregated by month. A loess smoothing line passes through data points and are shaded by 95% confidence intervals.
Figure 3 includes a heatmap indicating the co-detection frequencies of TRGs and each of the 5 respiratory pathogens separated by state from November 2009- August 2013. Additional file 1: Table S9 http://www.biomedcentral.com/1471-2334/14/460/additional displays the number of positive tests recorded from each state and the frequency that the pathogen was co-detected with a TRG. The highest frequency of co-detection occurred in the Northeast, especially Maryland and New Jersey. In the Southeast, Florida, Alabama, and Georgia accounted for the largest percentage of TRG co-detection. Along a westward gradient, the percentages of TRG co-detection dropped dramatically. Louisiana, Oklahoma, Texas, and Colorado all had co-detection frequencies under 53%. The level of TRGs co-detected from patients in New Jersey for the five pathogens was between 54.4% and 86.9%. Moving westward, the frequencies of TRGs co-detected was lower in Tennessee with levels between 42.7% and 55.3% for the five pathogens. In Texas, the rate of co-detection decreased even further to between 33.3% and 10.7%. Since Eastern states reported a much higher frequency of co-detection over Western states, a longitudinal trend was established.
(Enlarge Image)
Figure 3.
Heatmap Displaying the Longitudinal Effect of TRG Co-detection across the United States. The heat map is ordered by state on the bottom where the eastern states are on the right and the western states are on the left side of the map. The circles which are most red are the states that have the highest levels of tetracycline resistance co-detection regarding each pathogen, while states that are the most blue have the lowest rates. Size of the circles correspond to sample size of the data.
Results
The Apriori algorithm created a list of the 25 most frequent item sets found throughout the respiratory results database (Table 1). The largest measurements of support included TRGs (40.5%), S. pneumoniae (36.3%), H. influenzae (28.7%), and M. catarrhalis (25.1%). S. pneumoniae and H. influenzae were co-detected with TRGs 19.8% and 15.1% of all positive tests results. MRSA and S. aureus were detected 12.4% and 10.0% of all positive results, respectively. 6.5% of all positive tests included co-detection for MRSA and TRGs and 4.1% included simultaneous detection between S. aureus and TRGs.
From the frequent item set results, the Apriori algorithm identified the 25 association rules with the highest levels of support (Table 2). The top of the table includes the rules {S. pneumoniae} = > {Tetracycline}, {H. influenzae} = > {Tetracycline}, and {H. influenzae, M. catarrhalis} = > {S. pneumoniae} with 19.8%, 15.1%, and 7.2% support, respectively. Each of the highlighted association rules indicates a pathogen of interest or TRG that all have confidence measurements over 50.9% and lift values over 1.3. This confidence threshold establishes that, given the antecedent of a rule, the consequent is found in at least 50.9% of the cases. Since each rule had a lift measurement over 1, they were all considered to be rules of interest. From these results, and from literature associating TRGs with specific gram-positive bacteria, we sought to elucidate temporal co-detection patterns to better understand how patient age and geography relate to TRG co-detection.
Figure 1 displays a scatterplot including the frequencies of respiratory pathogens co-detected with TRGs by year. It is clear that the frequency of co-detection between each pathogen and TRG increased with time. From January 2010- August 2013, the percentage of gram-positive bacteria S. aureus, MRSA, and S. pneumoniae co-detected with TRGs rose 43%, 26%, and 27%, respectively. From January 2010- August 2013, the frequency of gram-negative bacteria H. influenzae and M. catarrhalis co-detected with TRGs rose 32% and 24%, respectively. Over each year, the rate of TRGs co-detected with these five potential pathogens rose a minimum of 3.9%.
(Enlarge Image)
Figure 1.
The frequency of TRGs Detected with Respiratory Pathogens Increased from 2009–2013. The scatterplot shows the rise in co-detection between TRGs and respiratory pathogens in recent years. The data is aggregated by week. Loess smoothing curves pass through the data points and are shaded by a 95% confidence interval.
Figure 2 displays the time-series co-detection frequency of TRGs with each pathogen segmented by 4 age groups. The patient sample size and total number of positive tests for each pathogen are included in Additional file 1: Tables S1-S7 http://www.biomedcentral.com/1471-2334/14/460/additional. In each age group, the rates of co-detection between TRGs and each pathogen increasd with time. Older patients showed a larger co-detection rate than younger patients. For H. influenzae, M. catarrhalis, and S. pneumoniae, patients (0–2] had the lowest rate of co-detection, followed by patients (2–13] years of age, then patients (50–100] years of age, concluded by patients (13–50] years of age, which had the highest rate of co-detection. In 2013, there was an average 33%, 22%, and 26% difference in co-detection rate from children aged (0–2] to patients age (13–50] for H. influenzae, M. catarrhalis, and S. pneumoniae, respectively. In 2013, an alarming 82% patients of age (13–50] had TRGs co-detected with S. pneumoniae. For S. aureus and MRSA, the age group (13–50] had the highest rate of co-detection between TRGs and the designated pathogen, but all other age groups had similar and co-detection frequencies over time.
(Enlarge Image)
Figure 2.
The Frequency of TRGs Detected with Respiratory Pathogens Increased in Each Age Group over Time. Each graph displays the co-detection frequency for each age group and pathogen. The data is aggregated by month. A loess smoothing line passes through data points and are shaded by 95% confidence intervals.
Figure 3 includes a heatmap indicating the co-detection frequencies of TRGs and each of the 5 respiratory pathogens separated by state from November 2009- August 2013. Additional file 1: Table S9 http://www.biomedcentral.com/1471-2334/14/460/additional displays the number of positive tests recorded from each state and the frequency that the pathogen was co-detected with a TRG. The highest frequency of co-detection occurred in the Northeast, especially Maryland and New Jersey. In the Southeast, Florida, Alabama, and Georgia accounted for the largest percentage of TRG co-detection. Along a westward gradient, the percentages of TRG co-detection dropped dramatically. Louisiana, Oklahoma, Texas, and Colorado all had co-detection frequencies under 53%. The level of TRGs co-detected from patients in New Jersey for the five pathogens was between 54.4% and 86.9%. Moving westward, the frequencies of TRGs co-detected was lower in Tennessee with levels between 42.7% and 55.3% for the five pathogens. In Texas, the rate of co-detection decreased even further to between 33.3% and 10.7%. Since Eastern states reported a much higher frequency of co-detection over Western states, a longitudinal trend was established.
(Enlarge Image)
Figure 3.
Heatmap Displaying the Longitudinal Effect of TRG Co-detection across the United States. The heat map is ordered by state on the bottom where the eastern states are on the right and the western states are on the left side of the map. The circles which are most red are the states that have the highest levels of tetracycline resistance co-detection regarding each pathogen, while states that are the most blue have the lowest rates. Size of the circles correspond to sample size of the data.
Source...