Using Pandemic-Era Longitudinal Surveillance Data to Improve Investigations of Carbapenem-Resistant Acinetobacter baumannii Clusters

Abstract

Objectives:

Epidemiologic investigations of multidrug-resistant organism (MDRO) clusters depend on a thorough history of health care exposures for case patients; however, histories are often incomplete. We describe how the robust influx of longitudinal infectious disease surveillance data from the COVID-19 pandemic improved whole-genome sequence–related cluster investigations and validated single nucleotide polymorphism (SNP) cluster definition thresholds for carbapenem-resistant Acinetobacter baumannii (CRAB) in Massachusetts.

Methods:

We used data from infectious disease laboratory test results reported through an integrated person-based surveillance system. We extracted all results from January 1, 2019, through March 19, 2024, for each CRAB case patient in a cluster to identify health care exposures. In addition, we extracted data from epidemiologic investigations. We used data to identify spatial links between cases. We combined timelines with whole-genome sequence data to determine whether genetically related cases were spatially linked.

Results:

We extracted 2354 test results for 69 CRAB case patients across 9 clusters; 2007 test results were from COVID-19 events. Three-quarters (n = 1775) of test results indicated a health care exposure not identified through standard investigation. Fifty-five case patients had a spatial link with at least 1 other case patient in their cluster. The median number of SNP differences between all spatially linked cases and all non–spatially linked cases was 4 SNPs and 10 SNPs, respectively.

Practice Implications:

Using longitudinal surveillance data identified possible CRAB transmission events undetected through standard investigation. Comparing the median SNP differences between spatially linked and non–spatially linked cases confirmed that the 10-SNP threshold was appropriate for defining CRAB clusters. Health departments can apply this method to enhance infectious disease investigations and determine appropriate SNP thresholds for infectious disease clusters.

Multidrug-resistant organisms (MDROs) are an urgent public health threat, with carbapenem-resistant Acinetobacter baumannii (CRAB) near the top of the 2024 World Health Organization Bacterial Priority Pathogens list. ¹ It is imperative that public health departments conduct thorough investigations of these pathogens to identify health care facilities for intervention and halt transmission. However, because patients with MDROs often have complicated medical histories and admissions to multiple health care facilities,^2-4 and information collected during traditional epidemiologic investigations can be limited in quality and quantity, determining where and when cases became infected or colonized and understanding chains of transmission can be difficult.

A baumannii is a gram-negative opportunistic bacterium that primarily causes infection in health care settings. ⁵ Transmission in these settings can occur through contact with contaminated environmental surfaces or shared medical equipment, as well as person-to-person contact. ⁶ A baumannii can cause infections in the blood, urinary tract, lungs, and wounds. Patients most at risk for infection or colonization with A baumannii are those with indwelling medical devices (including ventilators), with open wounds, who are severely immunocompromised, who have been admitted to an intensive care unit, or who had a long hospital admission. ⁷ Health care facilities implement infection prevention and control measures such as contact precautions, hand hygiene compliance, and frequent cleaning and disinfection of patient spaces and shared medical equipment to prevent transmission.

A baumannii can develop resistance to carbapenems, a class of broad-spectrum beta-lactam antibiotics. ⁸ In 2017, CRAB caused approximately 8500 infections in hospitalized patients in the United States. ⁹ Due to limited treatment options for CRAB and its ability to easily spread in health care settings, the Centers for Disease Control and Prevention (CDC) deemed CRAB an urgent public health threat in US health care facilities in 2021. ⁹ In Massachusetts, inter- and intrafacility transmission of CRAB has been detected.

While CRAB is not a reportable pathogen in Massachusetts, the Massachusetts State Public Health Laboratory increased surveillance in January 2020 by requesting submission of all CRAB isolates. The laboratory performed genetic sequencing on the isolates, and epidemiologists conducted investigations for these cases. In addition to investigating MDROs, these staff members investigate and have access to data on a broad array of infectious diseases reportable to the Massachusetts Department of Public Health (DPH). ¹⁰ Data from these investigations are stored in an integrated person-based surveillance system. ¹¹ The alignment of the start of this isolate submission and testing with the COVID-19 pandemic provided unique insight into patient contact with the health care system through integrated infectious disease surveillance data, allowing for a better assessment of health care exposures and cluster definition validity.

The objective of this analysis was to describe how the Healthcare Associated Infections and Antimicrobial Resistance (HAI/AR) Program in the Division of Epidemiology and the Sequencing and Bioinformatics Core Division at the Massachusetts DPH used a novel method of integrating longitudinal infectious disease surveillance data, including robust amounts collected during the COVID-19 pandemic, into standard investigation practices for whole-genome sequence (WGS)–related CRAB clusters and used these data to validate single nucleotide polymorphism (SNP) thresholds for defining these clusters.

Methods

The Massachusetts DPH initially investigated cases and clusters by using standard processes for MDROs. Novel to CRAB investigations outlined here, we used longitudinal surveillance data across all reportable infectious disease conditions (excluding tuberculosis and sexually transmitted infections, which are handled outside the Division of Epidemiology per state regulations ¹⁰ ) and sequencing data to build timeline visualizations. We examined timelines to identify potential transmission events missed by standard investigations.

From January 1, 2020, through March 19, 2024, the Massachusetts DPH received reports of 139 CRAB cases. A “case” was the initial instance of a CRAB isolate being identified in a person. Epidemiologists in the HAI/AR Program investigated CRAB cases in alignment with CDC’s MDRO Containment Strategy Guidelines. ¹² Staff members at health care facilities where the case patient was receiving care collected epidemiologic and clinical information when the isolate was collected to determine possible opportunities for transmission. Epidemiologists made infection prevention and control recommendations to health care facilities based on investigation findings. The Massachusetts Virtual Epidemiologic Network (MAVEN) captured data obtained through investigations. MAVEN is the Massachusetts DPH electronic infectious disease surveillance and case management system. This system is described in detail elsewhere. ¹¹ Specific to this methodology, MAVEN is a person-based system, so infectious disease surveillance data are linked to an individual person over time. Public health staff have limited access to data, and access is defined by the staff member’s role in the health department.

Internal guidelines and national guidance dictate that CRAB cases are investigated at only initial identification of infection or colonization for each patient, because a person is subsequently considered colonized indefinitely. ¹³ Public health staff collected standard epidemiologic information at the time of the first reported positive test result. Staff members did not routinely collect information about health care exposures that occurred after the initial case investigation.

Health care facility and commercial laboratories submitted CRAB isolates to the Massachusetts State Public Health Laboratory for sequencing. DNA was extracted by using the DNeasy Blood and Tissue Kit (Qiagen) and then qualified and quantified by using the NanoDrop One (Thermo Fisher Scientific Inc) and Qubit 4.0 Fluorometer (Thermo Fisher Scientific Inc). DNA libraries were prepared by using the DNA Prep kit (Illumina, Inc) and sequenced on the MiSeq Sequencing System (Illumina, Inc) by using v2 500 cycle cartridges according to CDC PulseNet protocol. ¹⁴

We performed bioinformatics analyses as previously described. ¹⁵ We used SeqyClean version 1.10.09 (Ilya Y. Zhbannikov) to perform raw paired-end read quality control and adapter removal; CG-Pipeline (Lee Katz) to perform read processing and quality evaluation; Mash version 2.1 (Maryland Bioinformatics Labs) to perform organism prediction; Shovill version 1.0.4 (SPAdes version 3.12.0) and QUAST version 5.0.2 (Gurevich Lab at the Helmholtz Institute for Pharmaceutical Research Saarland) to perform genome assembly and evaluate assembly quality; Prokka version 1.14.0 (Torsten Seemann) to perform genome annotation; Roary version 3.12.0 (Andrew Page, Carla A. Cummins) and IQ-TREE version 1.6.12 (IQ-TREE Development Team) to perform pan genome generation, core genome alignment, and phylogenetic analysis; lyve-SET version 1.1.4f (Lee Katz) to perform high-quality single nucleotide polymorphism (hqSNP) analysis; ABRicate version 1.0.1 (Torsten Seemann) and AMRFinderPlus version 3.11.4 (National Center for Biotechnology Information) and the National Center for Biotechnology Information AMRFinderPlus database ¹⁶ to identify antimicrobial resistance genes; and multilocus sequence typing (MLST) using MLST v2.23.0 (Torsten Seemann) to obtain sequence type. ¹⁷ All sequencing data were uploaded to the National Center for Biotechnology Information CDC Division of Healthcare Quality Promotion’s HAI-Seq Gram-negative bacteria BioProject. ¹⁸ We used MicrobeTrace ¹⁹ to create transmission network visualizations for each CRAB sequence type.

The Sequencing and Bioinformatics Core Division alerted epidemiologists in the HAI/AR Program when a CRAB cluster was identified. The HAI/AR Program epidemiologists and Massachusetts DPH bioinformaticians defined a cluster as ≥2 CRAB isolates within ≤10 SNP differences of each other. Cluster identification typically occurred after initial case investigation, because the WGS process could take up to 1 month after isolate submission. Massachusetts DPH bioinformaticians added additional isolates to clusters if the WGS results determined they were within ≤10 SNP differences of any other isolate in the cluster.

Epidemiologists conducted standard cluster investigations at the time of cluster identification and as new cases were added. Standard cluster investigations focused on identifying commonalities between cases, including shared exposures at health care facilities. This type of cluster investigation, which considered only case investigation data, occurred for all MDRO clusters in the HAI/AR Program. Enhanced investigation of CRAB clusters using longitudinal infectious disease data (described hereinafter) identified additional spatial links and determined whether cases within these clusters were epidemiologically linked or whether genetic clustering resulted from a lack of genetic variability among CRAB isolates from epidemiologically unrelated cases.

For each case patient included in a cluster, we extracted additional laboratory test results from MAVEN for all reportable infectious disease events investigated by the Division of Epidemiology (including COVID-19) from January 1, 2019, through March 19, 2024. We grouped the specimen collection date and ordering health care facility for each test result by a unique case and cluster identification number. We defined a “health care exposure” as an instance in which a health care facility ordered a laboratory test for any reportable infectious disease.

We combined health care exposure data and created diagrams of cluster-specific timelines by using the “vistime” package in R studio. ²⁰ We grouped cases along the y-axis in accordance with facilities that had a health care exposure, and we plotted the date of exposure along the x-axis. We defined spatial links as an event in which ≥2 cases were at the same facility within 14 days of each other. We identified these links by examining overlaps in facility-specific exposures between case patients along the x-axis. We used the date of the first positive CRAB result for a case to determine whether a transmission event could have occurred at the time of a spatial link, because a case patient must be colonized or infected to transmit the organism. Some cases had multiple spatial links with multiple cases over time. We considered all identified spatial links when creating possible transmission pathways.

We also created diagrams informed by the number of SNP differences between cases from WGS data to illustrate possible transmission pathways. If cases had a spatial link and were more closely genetically related to each other than to other cases within the cluster, we hypothesized that a transmission event may have occurred between them.

Due to a lack of national guidance on the number of SNP differences that constitute clustered cases, the cluster threshold of 10 SNP differences was previously set on the basis of an internal review of the genetic diversity of CRAB isolates in Massachusetts. To validate the appropriateness of this threshold, we calculated the median SNP difference between (1) all cases in each cluster, (2) cases within each cluster that were spatially linked to another case, (3) cases within each cluster that had no spatial links to another case, (4) all cases among all clusters, (5) cases that were spatially linked to another case among all clusters, and (6) cases that were not spatially linked to another case among all clusters.

Ethical Considerations

This project did not require ethical approval or informed consent of participants. The Massachusetts DPH Institutional Review Board (IRB) follows federal regulations regarding ethical review. All data for this study were collected through routine public health surveillance activities per Federal Regulation 45 CFR 164.512(b). These regulations authorize public health authorities to collect data for disease control, including reporting disease, vital events, and conducting public health investigations. According to the Common Rule (Federal Regulation 45 CFR 46 subpart A §46.104), data collected for public health purposes are exempt from IRB review, as detailed in paragraph D5. Research conducted by federal agencies using government-collected information for nonresearch activities is also exempt, provided that identifiable private information is maintained in compliance with the E-Government Act of 2002, the Privacy Act of 1974, and, when applicable, the Paperwork Reduction Act of 1995. This ensures the data remain protected while serving public health objectives.

Results

All 139 CRAB cases reported in the investigation period underwent epidemiologic investigation. The median age of case patients at the time of first positive CRAB result was 66 (range, 0.23-91; IQR, 53-73) years. Nearly two-thirds (65%; n = 90) of case patients were male (Table 1).

Table 1.

Description of carbapenem-resistant Acinetobacter baumannii case patients in Massachusetts, 2020-2024 ^a

Characteristic	Overall (N = 139)	Unique cases in whole-genome sequence clusters (n = 69)
Age, median (range) [IQR], y	66 (0.23-91) [53-73]	68 (24-91) [60-74]
Sex, no. (%)
Female	49 (35)	33 (48)
Male	90 (65)	36 (52)

^aData source: Massachusetts Virtual Epidemiologic Network; Troppy et al. ¹¹

The Sequencing and Bioinformatics Core Division performed WGS on all 139 submitted CRAB isolates. Ninety-four genomes (68%) were identified as sequence type (ST) 2, and 30 genomes (22%) were identified as ST1142 by using the Pasteur MLST scheme. Other sequence types observed were ST3(1), ST25(1), ST78(1), ST79(1), ST164(1), ST203(1), ST380(2), ST499(2), and ST667(1). Three genomes (2%) did not result in a sequence type. By using within-sequence–type core-genome phylogenetic trees and hqSNP analyses, we identified 7 SNP clusters among ST1142 (1) and ST2 (6) genomes. One ST2 cluster was further resolved into 3 highly related subclusters for a total of 9 SNP clusters (Figure 1).

Figure 1.

Visualization of (a) ST1142 and (b) ST2 carbapenem-resistant Acinetobacter baumannii (CRAB), created by using MicrobeTrace. ¹⁹ Clusters are defined as ≥2 isolates within 10 single nucleotide polymorphism (SNP) differences. Nodes are colored by cluster. Links are sized by SNP distance: the thicker the link, the smaller the distance. Different node shapes within an SNP cluster indicate isolates linked to a facility-specific subcluster.

By using WGS, we found that 69 unique cases (50%) were highly genetically related to other cases and were included in at least 1 of the 9 clusters identified. The median age of case patients in a cluster was 68 (range, 24-91; IQR, 60-74) years. More than half were male (52%; n = 36) (Table 1).

We extracted data from 2354 test results across 13 infectious disease event types for the 69 cases included in a CRAB cluster during the investigation period. Of these test results, 2007 (85%) were from COVID-19 events. Three-quarters (n = 1775) of test results indicated a health care exposure that was not identified during the initial case investigation. The median number of exposures per case identified by examining data from test results but not during case investigation was 19 (IQR, 10-37). Three cases were highly genetically related to 2 clusters and, thus, were included in the analyses for both clusters, resulting in a total of 72 cases across all clusters.

In an example of a cluster timeline to identify spatial links (Figure 2a), in this cluster of 3 cases, we identified a link between Case Patient 1 and Case Patient 2 at Acute Care Hospital J before the first positive CRAB test result for Case Patient 1. After this potential transmission event, Case Patients 2 and 3 had a link at Long-Term Care Facility D approximately 2 weeks before the first positive CRAB test result for Case Patient 2. The data demonstrate a link between these 2 cases at Long-Term Care Facility D for 2 weeks after Case Patient 2 had tested positive for CRAB. Approximately 5 months after the last link between Case Patients 2 and 3, Case Patient 3 tested positive for CRAB. The WGS and timeline results were used to construct a potential transmission chain (Figure 2b). Although the directionality of transmission between Case Patients 1 and 2 cannot be deduced from this information alone, a transmission event likely occurred because these cases were related by 4 SNP differences. It was also likely that Case Patient 2 transmitted to Case Patient 3 while at Long-Term Care Facility D, with 5 SNP differences between the 2 cases. Based on a standard epidemiologic investigation alone, the link between Case Patients 1 and 2 was not identified. Case Patients 2 and 3 were known to be residents of Long-Term Care Facility D; however, the dates of admission were not known based on the investigation. Of the 72 total cases, 55 (76%) had a spatial link with at least 1 other case in their cluster (Table 2).

Figure 2.

Example timeline and transmission diagram for a whole-genome sequence–related carbapenem-resistant Acinetobacter baumannii (CRAB) cluster. (a) Health care exposures at various facilities plotted over time, grouped by case. Vertical colored lines correspond to the specimen collection date for the first positive CRAB result of each case. (b) Proposed chain of transmission for this cluster. The transmission chains are informed by spatial links, the date each case first tested positive for CRAB, and single nucleotide polymorphism (SNP) differences between each case. The italicized numbers represent the number of SNP differences between each case. The double-headed arrow between Case 1 and Case 2 indicates that the directionality of the transmission event between the cases was not able to be deduced based on the available information.

Table 2.

Description of carbapenem-resistant Acinetobacter baumannii (CRAB) clusters in Massachusetts, 2020-2024 ^a

Cluster	No. of cases			Median no. of SNP differences			Specimen date range, days b
	All	Spatially linked	Not spatially linked	All	Spatially linked	Not spatially linked
1	3	3	0	5	5	NA c	295
2	14 d	12	2	6	2	4	1108
3	10	10	0	11	5	NA c	850
4	12 d	11	1	8	2	10	488
5	13	10	3	13	9	14	467
6	3	0	3	5	NA c	5	323
7	8	6	2	7	9	6	96
8	5	3	2	5	2	6	435
9	4	0	4	10	NA c	10	1088
All	72	55	17	9	4	10	NA

Abbreviations: NA, not applicable; SNP, single nucleotide polymorphism.

^aData source: Surveillance data are from the Massachusetts Virtual Epidemiologic Network; Troppy et al. ¹¹ Genomic data are from the National Center for Biotechnology Information. ¹⁸

^bNumber of days between the specimen collection dates for the earliest detected and latest detected CRAB cases within a cluster.

^cCluster did not have either spatially linked or nonspatially linked cases.

^dThree CRAB cases were highly genetically related to clusters 2 and 4. The case counts for these clusters include these overlapping cases. There were 69 unique CRAB cases.

The median SNP difference between all cluster isolates was 9 SNPs (IQR, 5-13). The median SNP difference between isolates within a cluster that also shared a spatial link was 4 SNPs (IQR, 1-8). The median SNP difference between isolates within a cluster that did not have a spatial link with any cases in the cluster was 10 SNPs (IQR, 5-13) (Table 2). The median SNP difference between all ST2 clusters was 43 SNPs (range, 11-3225).

Discussion

With 76% of cases having a spatial link with at least 1 other case in the same cluster, transmission events likely occurred at the facilities identified in the cluster timelines. These transmission events were not identified during initial epidemiologic investigations. Initial case investigations were limited by the amount and quality of information provided by infection preventionists and clinical staff at the health care facilities. Most facilities did not have a complete record of encounters for a patient, particularly outside their own system, leaving a gap in knowledge of where to look for possible spatial links with other known cases. Although it is well understood that A baumannii spreads easily in health care settings,^5-7 literature is lacking on the methods for identifying inter- and intrafacility transmission across long periods. By using longitudinal surveillance data collected according to state infectious disease reporting guidelines, epidemiologists identified 1775 more health care exposures among cases than was provided during initial investigation. Supported by sequencing data, epidemiologists more easily pinpointed times when cases were at the same health care facility and assessed the possibility of a transmission event occurring during that time based on the genetic relatedness of the cases. Although our analysis was retrospective, our methods could be useful for real-time cluster investigations and to help identify facilities for infection prevention and control recommendations.

In addition to improving epidemiologic investigations, our data validated that the 10-SNP threshold used to define a CRAB cluster was appropriate. Setting the SNP difference threshold at 10 SNPs resulted in 9 clusters for investigation, with a median of 9 SNP differences between all isolates within a cluster. Of these, 7 clusters had at least 1 pair of spatially linked cases. The median SNP distance between cases that had a spatial link was 4 SNP differences. While these data suggest that the threshold could be lowered to decrease the burden of investigation on epidemiologists, retaining a threshold of 10 SNPs ensures a broad scope for identifying cases that may be genetically related and spatially linked. Providing epidemiologic context to laboratory data may be useful when determining appropriate cluster thresholds for other pathogens. Ensuring an appropriate cluster threshold leads to accurate and efficient cluster investigations and, thus, timely recommendations and implementation of interventions.

Mandates on testing and reporting ²¹ for COVID-19 provided an influx of data on health care contact during our 50-month study period. Millions of positive and negative laboratory and point-of-care test results were shared daily with health departments during the pandemic. ²² Requirements and numbers of available test results varied substantially during our study period. Although the COVID-19 pandemic provided uniquely detailed reporting of how patients moved through health care systems, other infectious disease reporting may fill gaps in engagement with health care. Infection control practitioners and epidemiologists can use these data to implement targeted infection control measures and contain the transmission of pathogens.

Our methods had multiple limitations. First, CRAB is not a reportable condition in Massachusetts; unreported cases were likely missing from our analysis. Second, because we used only a subset of surveillance data available in MAVEN (January 2019–March 2024), the amount of retrospective and prospective surveillance data available for each case varied according to the date of the first positive CRAB test result. Third, most of the surveillance data used to create the timelines were from COVID-19 events. COVID-19 is reported much less frequently in Massachusetts now than it was during the height of the pandemic, in part due to widely available at-home test kits, limiting the amount of data available for future analyses. Fourth, the date of the first positive CRAB test result for a case did not necessarily indicate that the case patient was not colonized or infected before that date. Patients colonized with CRAB could have intermittently positive results while shedding bacteria in the environment, which can spread to other patients or the hands and clothing of health care workers.^6,7 Thus, determining whether a case is infectious during a spatial overlap is challenging. Fifth, tests for reportable conditions are not ordered at every health care admission or encounter, meaning this method would not capture health care exposures in which a test result for a reportable condition was not collected. Sixth, because not all reported conditions required epidemiologic investigation, we categorized the location of a health care exposure for a case at the facility level, not the unit level (eg, intensive care unit). In large facilities, case patients admitted to the same facility on the same date may have been in entirely separate units, with no true overlap, lessening the chance that the identified spatial link was a true transmission event. Taking these limitations into account, our methods are most useful when used in tandem with traditional case investigation to bridge gaps in the understanding of a case patient’s movement between health care facilities.

Practice Implications

Using longitudinal surveillance data with WGS, the HAI/AR Program identified possible transmission events of CRAB in health care facilities in Massachusetts that were not identified through initial case investigations. These data validated that a 10-SNP threshold was appropriate for identifying CRAB clusters. Our results highlight the benefit of using other infectious disease surveillance data to understand health care exposure. Public health departments may find our methodology useful in identifying transmission events during cluster investigations or evaluating SNP cluster thresholds in their jurisdictions. WGS can be used to limit the number of cases requiring surveillance data review.