We appreciate the interest by Chen and associates in our recent study of predictive risk factors for endogenous endophthalmitis among hospitalized patients with hematogenous infections. They raise several important points that should certainly be considered in evaluating patients, and that merit discussion. In itemized response:
First, as Dr Chen and colleagues note, patients with more severe disease frequently require longer hospitalizations and longer treatment with intravenous antimicrobials. Since treatment with intravenous antimicrobials may sometimes be continued in an outpatient setting (home health care or visits to an infusion center), we considered longer hospitalizations to be an indicator of more severe infection or worse systemic health. Compared to fungemia, bacteremia (including sepsis) was associated with markedly lower rates of presumed endogenous endophthalmitis. Among patients with bacterial infections, 0.04% of patients with bacteremia and 0.09% of patients with septicemia (defined as ICD-9 codes 995.91 and 995.92) developed endophthalmitis. However, among patients with bacteremia, sepsis was associated with a nonsignificant odds ratio of 0.71 for development of endogenous endophthalmitis (CI 0.48–1.05, P = .09). We suspect this may reflect low numbers, as well as overlap and unreliability/inconsistency in coding of sepsis vs bacteremia.
Second, regarding urinary tract infection, we are not aware of routine positive urine cultures among patients with bacteremia or fungemia; however, for all of the reasons stated, we agree that patients with urinary tract infection should be considered individually.
Third, owing to the limitations inherent in code-based data, we lack sufficient granularity to evaluate patients with renal disease from diabetes vs other cause. Although there is an ICD-9 diagnosis code for diabetes with renal manifestations, it is separate from the code for chronic renal failure/chronic kidney disease (for which etiology is unspecified), and we cannot reliably elaborate on clinical details for the 17 patients with chronic renal disease and endogenous endophthalmitis. We included only first-hospitalization data to avoid over-representing patient factors. Since odds were calculated for patient hospitalizations, the sample would have been subject to bias had we included patients with multiple hospitalizations.
Finally, we did not report risk factors by age subgroup owing to the small n for the study—with only 8 pediatric patients with endogenous endophthalmitis, we could not run a reliable or accurate regression. However, we agree that the points raised by Dr Chen and colleagues are important to consider in a clinical setting.