Seasonality and weather dependance of Acinetobacter baumannii complex bloodstream infections in different climates in Brazil

Recent studies report seasonality in healthcare-associated infections, especially those caused by Acinetobacter baumannii complex. We conducted an ecologic study aimed at analyzing the impact of seasons, weather parameters and climate control on the incidence and carbapenem-resistance in A. baumannii complex bloodstream infections (ABBSI) in hospitals from regions with different climates in Brazil. We studied monthly incidence rates (years 2006–2015) of ABBSI from hospitals in cities from different macro-regions in Brazil: Fortaleza (Ceará State, Northeast region), Goiânia (Goiás State, Middle-west) and Botucatu (São Paulo State, Southeast). Box-Jenkins models were fitted to assess seasonality, and the impact of weather parameters was analyzed in Poisson Regression models. Separate analyses were performed for carbapenem-resistant versus carbapenem-susceptible isolates, as well as for infections occurring in climate-controlled intensive care units (ICUs) versus non-climate-controlled wards. Seasonality was identified for ABSSI ICUs in the Hospitals from Botucatu and Goiânia. In the Botucatu hospital, where there was overall seasonality for both resistance groups, as well as for wards without climate control. In that hospital, the overall incidence was associated with higher temperature (incidence rate ratio for each Celsius degree, 1.05; 95% Confidence Interval, 1.01–1.09; P = 0.006). Weather parameters were not associated with ABBSI in the hospitals from Goiânia and Fortaleza. In conclusion, seasonality was found in the hospitals with higher ABBSI incidence and located in regions with greater thermal amplitude. Strict temperature control may be a tool for prevention of A. baumanii infections in healthcare settings.


Authors' affiliations Introduction
Despite their specific characteristics, healthcare-associated infections (HCAIs) can share some epidemiological determinants with those that occur in the community.
Seasonality, increasingly identified in recent studies, is one example [1]. It is prominent in bloodstream infections caused by Gram-negative bacilli (GNB), which have been linked to proximity to the equator [2], summer season [3] and high environmental temperatures measured either within [4] or outside hospitals [5].
This latter aspect is one of the gaps in our current understanding of HCAIs seasonality, since GNB incidence increases during warm periods even within units that are climate-controlled (and thus expected not to present relevant temperature variations) [1,5,6,7]. Some authors have theorized an influx from reservoirs outside healthcare settings, on the basis of greater seasonality of multidrug-susceptible (as opposed to multidrug resistant, supposedly "hospital-borne") GNB [8] or on molecular heterogeneity of summer strains (which suggests multiple sources) [9].
However, those findings were not supported by other studies, especially those pointing to relevant "summer peaks" of multidrug-resistant GNB infections [6,7,10].
Acinetobacter baumannii complex bloodstream infections (ABBSI) stand out among GNB HCAIs for their striking seasonal pattern [11]. They pose therefore unique opportunities to assess HCAIs seasonality. With that in mind, we conducted an ecological study aimed at analyzing how that seasonality of ABBSI varies in different climates, as well as its association with antimicrobial resistance and climate control in hospital units.

Ethical statement
This study adheres to the Helsinki declaration guidelines and was approved by the institutions Committees for Ethics in Research (CAAE # 81985517.0.1001.5411).

Study settings.
This study was conducted in three hospitals from cities located in different macro-

Time series analysis
Incidence rate series were fitted to stochastic, Box-Jenkins models (Seasonal Autoregressive Integrated Moving Average [SARIMA]) [13]. Those models allowed us to assess statistically the correlations both with data from immediately preceding

Models of weather dependence
Multivariable Poisson regression models were analyzed with average monthly temperature and humidity, as well as aggregate monthly rainfall as independent variables and monthly rates as outcomes. Both Poisson regression an time series analyses were conducted using STATA 14 software (StataCorp, College Station, TX).

Results
The aggregate incidence of ABSSI in the study period (for each type of hospital unit and pattern of carbapenem-resistance) is presented in Table 2. Briefly, the Botucatu hospital presented higher overall incidence, but also a greatest proportion of carbapenem-susceptible isolates (especially in years before 2011), when compared to the other two hospitals. Results from the Box-Jenkins models are presented in Table 3. Briefly seasonality was identified for ABSSI in ICUs in the Hospitals from Botucatu and Goiânia. In the Botucatu hospital, there was also overall seasonality for both resistance groups, as well as for wards without climate control. In the Poisson regression analysis ( Table 4), we found significant association of temperature with carbapenem-resistant and overall ABBSI in non-climatecontrolled units in the Botucatu hospital. There was also a similar association when we used overall hospital incidence of ABBSI as outcome. No association with weather was found for the other study hospitals.

Discussion
Brazil is a huge country with an area of 8.5 million square kilometers and 200 million inhabitants, distributed 27 states located in five macro-regions. Climates range from equatorial in the Amazon basin to temperate in the South region. There are great socioeconomic differences, with poorer areas in the North/Northeast and more developed states in the South/Southeast [14]. Some community-associated infectious diseases are restricted to a climate or biome (e.g., malaria in Amazon), while other occur seasonally in most areas (e.g., Dengue fever) [15]. Among all this complexity, one may wonder how geography and climate impact on the epidemiology of HCAIs occurring in circa 6,000 hospitals in Brazil.
A. baumannii complex infections are hyperendemic in Brazilian hospitals [16]. As in other countries, they preferentially affect critically ill patients, undergoing invasive procedures or carrying invasive devices [11,17]. Our study aimed, therefore, at comparing the impact of season and weather on ABBSI incidence in areas that were both geographically distant and presented different climates.
Counterintuitively, ABBSI incidence was higher in the hospital located in the area presenting lower temperatures [2]. Seasonality results did not exactly match findings from Poisson regression analysis, using weather parameters as predictors. Some aspects may account for those findings. First, one must notice that weather data were obtained from meteorological stations, located in the same city of (but not inside or nearby) the study hospitals. Therefore, they do not necessarily reflect, for instance, temperature and humidity inside hospital wards. This difference is obviously more relevant for the climate-controlled ICUs, though previous studies from our group have found association of temperature outside hospitals with ABBSI incidence in those units [5,6].
Another confounding aspect is the overall difference in the incidence of ABBSI among study hospitals. As one may infer from the Box-Jenkins models presented in Finally, there are potential drivers of seasonality which are not directly associated with weather, such summer understaffing [1], the presence of new students and resident doctors or even changes in diseases leading to hospital admission. All those factors require further investigation, though a previous study found no impact of patients' severity-of-illness or comorbidities on the seasonality of Gram-negative (including A. baumannii) bloodstream infections [7].
Our study was limited for not individually assessing the average severity-of-illness of patients admitted to different hospitals, as well as their structure for infection control and the quality of microbiology laboratory. However, the three hospitals were included in a recent multistate survey, with similar performance [18]. The same study found minor differences on overall HCAIs incidence among different regions in Brazil [19]. Other limitation was not assessing the incidence of pathogens (e.g., MRSA and GNB) that may compete ecologically for human and inanimate reservoirs inside hospitals [20]. It is therefore possible that the greater incidence of GNB such as Klebsiella spp in hospitals close to the Equator line 2 can impact on the lower incidence of A. baumannii found in our study.
In conclusion, we found seasonality of A. baumannii infections in hospitals located in areas with climates ranging from tropical to temperate. Both incidence, seasonality and association with weather were greater in the hospital located in the area with fresher climate and greater temperature range. Besides reinforcing the seasonal nature of that pathogen, this study points to air conditioning in hospital wards (which is uncommon in low-to-middle income countries [21]) as a tool for infection control and prevention.