# Prior culture-guided prediction of antibiotic susceptibility in recurrent respiratory tract infections: a retrospective cohort analysis

## TL;DR

This study shows that prior culture results can effectively predict antibiotic resistance in recurrent respiratory infections, helping choose better empirical treatments.

## Contribution

The study introduces a Bayesian method to quantify the predictive value of prior cultures for antibiotic susceptibility in recurrent respiratory tract infections.

## Key findings

- Prior cultures predicted future resistance to 22 antibiotics with >70% accuracy.
- Ciprofloxacin and levofloxacin showed ≥85% predictive value for resistance.
- Imipenem and amikacin had ≥85% predictive value for future susceptibility.

## Abstract

Recurrent respiratory tract infections (RRTIs) are a serious problem in older adults, necessitating appropriate early empiric antibiotic therapy. However, rising antimicrobial resistance complicates empiric treatment selection. Although prior culture results can offer guidance for empiric decisions, little data exist to quantify their predictive value in RRTIs. Therefore, we constructed a respiratory antibiogram and assessed Bayesian metrics of prior gram-negative isolates for predicting subsequent resistance or susceptibility in this population.

A retrospective cohort included hospitalized RRTI patients (defined as RTIs occurring ≥2 times in 6 months or ≥3 times in 12 months). Patient-specific antibiograms were constructed. A Bayesian prevalence-calibrated estimation was used to assess the predictive validity of prior respiratory cultures for subsequent antimicrobial susceptibility in paired isolates. Sensitivity (Sen), specificity (Spec), Bayesian positive predictive value (PPV), negative predictive value (NPV), and resistance prevalence were calculated for 24 antibiotics, with bootstrap 95% confidence intervals. Analyses were restricted to bacterial pathogens.

We included 463 visits from 160 unique patients, with a median age of 70. Pathogen distribution revealed Gram-negative bacteria (89.50%) and Gram-positive bacteria (10.50%). Antibiogram results showed that Pseudomonas aeruginosa exhibited >30% resistance to 10 antibiotics, while Klebsiella pneumoniae exhibited >35% resistance to 16 antibiotics. Prior cultures had good predictive value (>70%) for detecting future resistance to 22 antibiotics, with ciprofloxacin, levofloxacin, cefuroxime, cefotaxime, ampicillin, ampicillin/sulbactam, and ceftriaxone showing excellent predictive values (≥85%). Additionally, prior cultures had excellent predictive value (≥85%) for detecting future susceptibility to imipenem, amikacin, tobramycin, meropenem, minocycline, and tigecycline.

Considerable antibiotic resistance was detected among isolates in patients with RRTIs. Using a prior culture as a guide can enhance the probability of selecting an effective empirical agent.