Bystander selection in the outpatient setting

Author: Christine Tedijanto
Email: ctedijanto@g.harvard.edu
Last update: January 10, 2019
Publication: Estimating the proportion of bystander selection for antibiotic resistance among potentially pathogenic bacterial flora

Description:

Antibiotic use creates a selective pressure for resistance. In this paper, we estimate the extent of bystander selection, which we define as selective pressures experienced by an organism due to antibiotics that were intended to treat another pathogen. We estimate a metric termed the "proportion of bystander exposures" using existing data from the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS), the Human Microbiome Project (HMP), and published carriage and etiology studies. This repository containss the analysis code and data needed to estimate the proportion of bystander exposures for 9 commonly carried bacterial species and 17 commonly used antibiotics in the United States.

Data:

• The .Rdata file contains the relevant Human Microbiome Project, ICD-9 code, and etiology data that are required for this project. The included dataframes are detailed below:

  1. ICD9v28: ICD9 diagnosis codes (v28) with long and short description from NBER.
  2. conditionCodes: ICD9 diagnosis codes to include or exclude in condition categories. Based on conditions in Fleming-Dutra et al. 2016. Codes are written using regular expressions (regex) for use in grepl function.
  3. etiologies: Estimated etiologies by condition, species, and age group. See Supporting Information in paper for sources.
  4. microbiome.adults: Carriage prevalence by species and visit number among adults participating in the Human Microbiome Project. Estimates for several species were based on additional carriage studies; see Supporting Information in paper for sources.
  5. microbiome.kids0: Carriage prevalence by species, body site, and study among children under 1yo. See Supporting Information in paper for sources.
  6. microbiome.kids1to5: Carriage prevalence by species, body site, and study measured in children 1-5yo. See Supporting Information in paper for sources.

* Data from the National Ambulatory Medical Care Survey (NAMCS) and the National Hospital Ambulatory Medical Care Survey (NHAMCS) are made publicly available by the CDC [here](https://www.cdc.gov/nchs/ahcd/datasets_documentation_related.htm). This analysis uses the Stata files from the 2010 and 2011 surveys.

Instructions:

Place outpatient_bystander_analysis.R, outpatient_bystander_analysis.Rdata, and NAMCS/NHAMCS Stata files in the same directory. For baseline analysis, code chunk B2(ix) ALTERNATE viii creating tiered diagnosis (from Fleming-Dutra et al. 2016) should be commented out. Run entire file. Dataframes bystander.df and bystanderbyclass.df contain bystander proportions for all species across individual antibiotics and antibiotic classes, respectively.

In order the run the sensitivity analysis using the tiered diagnosis, comment out code chunk B2(viii) and comment in code chunk B2(ix). Rerun all code.

Checks:

The baseline analysis may take up to 30 minutes to run (majority of runtime due to code chunk B3).

1. Microbiome file

Using the code and data included above, the age group-specific weighted prevalences for the first 5 rows of the microbiome dataframe should be as follows:

Species.strain.Key	wtprev_adults	wtprev_kids0	wtprev_kids1to5
Escherichia_coli	0.66349810	0.94870000	1.00000000
Haemophilus_influenzae	0.68631179	1.00000000	0.95896469
Klebsiella_pneumoniae	0.07414449	0.39097000	0.15000000
Moraxella_catarrhalis	0.02281369	0.45485113	0.50790000
Pseudomonas_aeruginosa	0.01901141	0.01359456	0.01359456

2. NAMCS/NHAMCS summary file

In the baseline analysis, using the code and data included above, the first 5 rows of the NAMCS.summary dataframe should be as follows:

antibiotic	dataset	condition	agegroup	wtVisits	wtVisits.se	drugclass
AMOXICILLIN	namcs2010	acuteSinusitis	adults	2356812	565377.953	PENICILLINS
AMOXICILLIN	namcs2010	acuteSinusitis	kids0	70355	70355.000	PENICILLINS
AMOXICILLIN	namcs2010	acuteSinusitis	kids1to5	158253	67000.831	PENICILLINS
AMOXICILLIN	namcs2010	chronicSinusitis	adults	1769606	337174.213	PENICILLINS
AMOXICILLIN	namcs2010	chronicSinusitis	kids0	166430	106789.037	PENICILLINS

In the baseline analysis, using the code and data included above, the first 5 rows of the NAMCS.summary.byclass dataframe should be as follows:

dataset	condition	agegroup	drugclass	wtVisits	wtVisits.se
namcs2010	acuteSinusitis	adults	PENICILLINS	3448543	681417.82
namcs2010	acuteSinusitis	kids0	PENICILLINS	70355	70355.00
namcs2010	acuteSinusitis	kids1to5	PENICILLINS	158253	67000.83
namcs2010	chronicSinusitis	adults	PENICILLINS	3650426	562887.03
namcs2010	chronicSinusitis	kids0	PENICILLINS	166430	106789.04

3. Bystander output

In the baseline analysis, using the code and data included above, the first 3 columns of the first 5 rows of the bystander.df dataframe should be as follows:

species	antibiotic	bystander_prop
Streptococcus_pneumoniae	AMOXICILLIN	0.8185041
Streptococcus_pneumoniae	AMOXICILLIN-CLAVULANATE	0.8576846
Streptococcus_pneumoniae	PENICILLIN	0.9867820
Streptococcus_pneumoniae	AZITHROMYCIN	0.9160512
Streptococcus_pneumoniae	CLINDAMYCIN	0.9825318