To the Editor—The incidence of healthcare-associated infections (HCAIs) caused by gram-negative bacilli (GNB) increases in summerReference Perencevich, McGregor and Shardell ¹ and has been associated to outside environmental temperatureReference Schwab, Gastmeier and Meyer ² and low latitudes.Reference Fisman, Patrozou and Carmeli ³ Evidence has been consistent for temperateReference Richet ⁴ and, to a lesser extent, tropical climates.Reference Fortaleza, Caldeira and Moreira ⁵ The factors underlying this phenomenon and implications for infection control are not clear.

One of the causal hypotheses proposes that environmental reservoirs outside healthcare settings increase in warmer periods and that microorganisms are somehow carried into hospitals.Reference Richet ^{4 ,} Reference Eber, Shardell, Schweizer, Laxminarayan and Perencevich ⁶ This hypothesis is based on the following findings: (1) seasonality and meteorological dependence occur even in hospitals with complete climate controlReference Richet ⁴ ; (2) polyclonal increase of Acinetobacter baumannii during summer, suggesting sources other than cross transmissionReference Christie, Mazon, Hierholzer and Patterson ⁷ ; (3) seasonality for overall but not for multidrug-resistant A. baumannii, implying a possible community-associated source.Reference Fukuta, Clarke, Shields, Wagener, Pasculle and Doi ⁸

We designed an ecological study aimed at testing premises that are central to the hypothesis that there are differences in seasonal pattern and influence of weather on the incidence of A. baumannii HCAIs according to climatization of hospital units or to imipenem resistance. The study was conducted in a teaching hospital in inner Brazil (Hospital Estadual Bauru, 335 beds). In that hospital, intensive care units (ICUs) have climate control, but all the other hospital wards are not climatized.

We obtained data from records of patient with cultures positive for A. baumannii from 2006 through 2017. Clinical cultures collected after the day 3 of admission were included (ie, the 3-midnight rule).Reference Cohen, Calfee and Fridkin ⁹ We included only the first culture positive for A. baumannii for each subject. Monthly incidence was calculated for overall A. baumannii and for subgroups based on unit of admission, specimen, and resistance to imipenem.

Monthly meteorological parameters (ie, average temperature, average relative humidity, and aggregated rainfall) were collected from a nearby meteorological station (Institute of Meteorological Research, State University of São Paulo, City of Bauru, São Paulo State, Brazil).

We tested time series for seasonality using Box-Jenkins models and autocorrelation plots.Reference Box, Jenkins and Reinsel ¹⁰ The association with weather was tested using Poisson regression. Meteorological parameters were dichotomized, based on the 75th percentile of monthly values (temperature, 24.7°C [76.5°F]; relative humidity, 79.6%; rainfall, 163.6 mm). Analyses were performed using NCSS 9 software (LLC, Kaysville, UT) and SPSS version 20 software (IBM, Armonk, NY).

We identified 1,207 patients with HCAIs due to A. baumannii. The overall incidence was 12.99 per 10,000 patient days: 67.12 for ICUs and 5.11 for noncritical wards. The imipenem resistance rate was 78.9%.

Time series of incidence rates (either of overall A. baumannii or of subgroups) tested negative for seasonality. The results of our analysis of association with weather are presented in Table 1. We found a significant positive association with temperature for overall A. baumannii. The analyses of subgroups identified stronger associations with temperature for ICUs, isolates from blood cultures, and those that were imipenem resistant.

Table 1

Poisson Regression Model for Impact of Monthly Meteorological Parameters on the Incidence of Healthcare-Associated Infections Caused by Acinetobacter baumannii

Note. RR, rate ratio; CI, confidence interval; ICUs, intensive care units.

^a Incidence in cases per 10,000 patient days.

^b Meteorological values were dichotomized at the 75^th percentile of monthly values (temperature, 24.7°C [76.5°F]; relative humidity, 79.6%; rainfall, 163.6 mm). Statistically significant associations (P<.05) are presented in boldface.

The absence of seasonality in Box-Jenkins models may be due to the poorly defined pattern of seasons in the tropical climate. The area where the study hospital is located has a mostly warm and humid period (usually October through March, including summer and spring), with lower temperatures in the rest of the year. However, this pattern is irregular, and brief increases in environmental temperature often occur, even during fall and winter.

A different picture emerges when we consider associations of outside temperature and the incidence of A. baumannii. This association was stronger in the following conditions: positive blood cultures, ICUs, and imipenem-resistant strains. The recovery of A. baumannii from blood cultures may be interpreted either as a sign of “true infection” (as opposed to colonization) or as a marker of the invasiveness of strains. Either way, our findings reinforce the impact of temperature on the incidence of HCAIs caused by this agent.

While the severity of critical patients is an obvious risk factor for acquisition of A. baumannii, the causes underlying the impact of the outside temperature on its incidence in ICUs are not straightforward. We are even less certain why we found greater association of temperature with imipenem-resistant strains. Notably, this finding is contrary to the picture described by Fukuta et al.Reference Fukuta, Clarke, Shields, Wagener, Pasculle and Doi ⁸ Because infections caused by imipenem-resistant isolates (which are most likely healthcare-associated) increase in warmer months, we cannot infer that reservoirs in the outside environment are implicated in this phenomenon. On the other hand, increases in incidence according to outside temperature, even among patients admitted to units with complete climate control, are unlikely to be due to inanimate reservoirs within those units. Finally, because greater incidence of the pathogen of interest was not associated with any particular month or season, we can rule out the hypothesis that understaffing due to summer vacations would lead to greater transmission of nosocomial pathogens.Reference Richet ^{4 ,} Reference Eber, Shardell, Schweizer, Laxminarayan and Perencevich ⁶

We should approach our findings in terms of implications for research and implications for practice. Studies focusing on changes in healthcare work processes (eg, adherence to hand hygiene, conformities in isolation precautions) in different periods of the year or under different weather conditions should be conducted. Also, research including molecular strain typing could help determine whether there is increased cross transmission of A. baumannii (as well as other GNB) in periods of higher outside temperature. The obvious implication for practice is the requirement of intensifying infection control measures during warm months. Other recommendations may arise from continuing research.

The phenomenon of seasonality and meteorological determination of HCAIs caused by GNB remains puzzling. Still, from our perspective, it should not be regarded as mere curiosity. Instead, its elucidation may provide novel opportunities for infection prevention and control.