Epidemiology of Healthcare Facility-Associated Nontuberculous Mycobacteria From 2012 Through 2020 in a 10-Hospital Network in the United States

Abstract

Background

Data on the epidemiology of healthcare facility-associated (HCFA) nontuberculous mycobacteria (NTM) are sparse. We performed a multicenter longitudinal cohort study of HCFA NTM epidemiology.

Methods

We retrospectively analyzed positive cultures for NTM performed from 2012 through 2020 within a network of 10 US academic hospitals and associated clinics. A unique NTM episode was defined as a patient's first positive culture for a particular NTM species and specimen source category (pulmonary vs extrapulmonary). Episodes linked to specimens obtained on day 3 or later of hospitalization were classified as hospital-onset (HO). Seven hospitals contributed at least 12 months of data prior to January 2014. Within this closed cohort, incidence rate ratios (IRRs) and trends in incidence from 2014 through 2020 were estimated, assuming the number of episodes followed the Poisson distribution.

Results

A total of 12 855 unique NTM episodes occurred from 2012 through 2020 during 19 248 137 patient-days of surveillance. Of these episodes, 3045 (24%) were HO. HO incidence rates were highly variable among hospitals, with a median hospital rate of 1.06 episodes per 10 000 patient-days (range, 0.35–5.48). Within the 7-hospital closed cohort from 2014 through 2020, the incidence rate of HO episodes decreased from 2.29 to 1.42 episodes per 10 000 patient-days (IRR, 0.62; 95% confidence interval, .53–.73; P < .0001).

Conclusions

Incidence rates of HO NTM episodes decreased from 2014 through 2020, but rates varied substantially among hospitals. These results provide comprehensive data on HO NTM isolation, including benchmark rates that can be used to improve hospital-based NTM surveillance.

Incidence rates of hospital-onset nontuberculous mycobacteria (NTM) episodes decreased from 2014 through 2020 within a hospital network, but rates varied substantially among hospitals. Results provide benchmark rates of healthcare facility-associated NTM that can be used to improve hospital-based NTM surveillance.

Nontuberculous mycobacteria (NTM) are pervasive in community and healthcare environments [1, 2]. Once acquired from either setting, these organisms can cause invasive pulmonary and extrapulmonary infections that are challenging to eradicate and associated with substantial healthcare costs, morbidity, and mortality [3–5].

Over the past decade, healthcare facility-associated (HCFA) NTM infections and outbreaks have received increased scrutiny [6, 7]. In fact, the Centers for Disease Control and Prevention reported that NTM were the most common pathogens implicated in consultations they performed from 2014 through 2017 for non-Legionella water-related healthcare-associated infections (HAIs) [8]. Numerous investigations have found that NTM acquisition occurred due to NTM colonization of healthcare facility water [9, 10], ice [11], or equipment, such as bronchoscopes [12] and heater–cooler units used during cardiac bypass surgeries [13, 14].

Despite increasing concern over NTM acquired at healthcare facilities, data on the epidemiology and incidence of HCFA NTM remain scarce. While numerous studies have reported that the prevalence or incidence of NTM cases is increasing worldwide [15–18], investigators in the United States have largely conducted population-based NTM surveillance (eg, cases per 100 000 population) for selected counties or states [19–22]. These data estimate the community burden of NTM disease over time within certain regions but offer limited utility to health systems attempting to interpret their HCFA NTM rates and trends. In addition, aside from outbreak investigations, few epidemiologic studies have reported hospitalization status or days of hospitalization accrued at the time of NTM diagnosis. Finally, most US states do not require reporting of NTM cases [23], and, in contrast to other HAIs [24], hospitals do not submit rates of incident NTM episodes per patient-day. Therefore, while hospitals can evaluate trends in their own NTM incidence rates to detect acute increases [25], external benchmarks for longitudinal rate comparisons do not exist.

We performed this study to retrospectively analyze the epidemiology of HCFA NTM that occurred over 9 years within a 10-hospital network. We hypothesized that this analysis would address important shortcomings in existing HCFA NTM surveillance and demonstrate the potential utility of standardized HCFA NTM surveillance definitions. Furthermore, we anticipated that these findings would provide a framework for hospitals to better understand their center's NTM epidemiology, compare their NTM burden to that of other hospitals, and identify center-specific targets for preventing HCFA NTM.

METHODS

Setting

The principal investigator contacted hospital epidemiologists, infectious diseases physicians, and microbiologists at 13 US academic hospitals with high-volume thoracic transplant programs. Of these centers, 10 agreed to participate. The hospital network represented 8 states, including Texas, North Carolina (n = 2), Pennsylvania (n = 2), Missouri, Washington, Ohio, Michigan, and Tennessee (Table 1, Supplementary Figure 1). The Duke University Health System Institutional Review Board (IRB) served as the single IRB of record and waived informed consent.

Table 1.

Characteristics of a 10-Hospital Network That Performed Retrospective Culture-Based Surveillance for Nontuberculous Mycobacteria From 2012 Through 2020

Hospital No.	Hospital Location (State)	No. of Months of Surveillance (Dates)		Patient-Daysa	HO Episodes	HO Episodes per 10 000 Patient-Days
Hospital cohort included in 2014–2020 longitudinal rate analyses
1	Texas	102	(July 2012–Dec 2020)	1 061 328	582	5.48
2	North Carolina	108	(Jan 2012–Dec 2020)	2 661 791	898	3.37
3	Pennsylvania	108	(Jan 2012–Dec 2020)	2 181 919	471	2.16
4	Missouri	108	(Jan 2012–Dec 2020)	2 908 853	335	1.15
5	Pennsylvania	108	(Jan 2012–Dec 2020)	1 876 399	182	0.97
6	Washington	108	(Jan 2012–Dec 2020)	1 146 185	63	0.55
7	Ohio	96	(Jan 2013–Dec 2020)	2 889 956	102	0.35
Hospitals that contributed data beginning after January 2013
8	Michigan	90	(July 2013–Dec 2020)	2 309 444	265	1.15
9	North Carolina	54	(July 2016–Dec 2020)	1 155 212	84	0.73
10	Tennessee	36	(Jan 2018–Dec 2020)	1 057 050	63	0.60
10 hospital totals		918	(Jan 2012–Dec 2020)	19 248 137	3045	1.58

Abbreviation: HO, hospital-onset.

^aExact counts of monthly patient-days at hospital 6 during 2012 were not available. These counts were estimated based on patient-days accrued at this hospital in 2013. Similarly, patient-day counts were missing for hospital 7 for 2 months of 2013, and these counts were estimated based on patient-days reported for 2014.

Data Collection

We retrospectively identified clinical cultures that isolated NTM and were performed within the hospital network from January 2012 through December 2020. Five centers provided data on all cultures positive for NTM beginning in January 2012. The remaining 5 centers provided culture data beginning on the earliest date that complete study data were available, ranging from July 2012 through January 2018 (Table 1). All centers provided continuous NTM data from the center's specific study start date through December 2020. NTM isolates were included if specimens were obtained at the primary hospital or outpatient clinics associated with each site. All cultures were performed at the discretion of treating clinicians and per standard operating procedures of clinical microbiology laboratories.

Each center extracted data on cultures that recovered NTM from microbiology and electronic health record databases. Study variables consisted of a unique coded patient identifier, patient age, culture date, specimen source (ie, body site of sample collection), NTM species, hospital admission status (ie, admitted or outpatient), date of admission, and location where the specimen was obtained (eg, specific hospital ward, clinic, or procedural suite). Data were extracted without manual chart review, except for rare instances of missing or nonspecific data. Each hospital additionally reported monthly facility-wide inpatient patient-days [26]. Hospitals also summarized NTM outbreaks and water management procedures used during the study period.

Definitions

The culture date was the date that the patient specimen was obtained, which was also considered the date of diagnosis. Specimen sources recorded as sputum, bronchoalveolar lavage (BAL) or bronchial wash, other airway specimens, and bronchial or lung tissue were categorized as pulmonary sources. All other sources, including pleural and sinus specimens, were considered extrapulmonary sources. An NTM episode was defined as a patient's first positive culture for a particular NTM species and specimen source category (pulmonary vs extrapulmonary). Based on prior NTM surveillance data and investigations [10, 25, 27–29], episodes linked to specimens obtained on day 3 or later of hospital admission were classified as hospital-onset (HO) NTM. For specimens obtained on day 1 or 2 of admission, associated NTM episodes were classified as community-onset (CO). Episodes linked to specimens obtained in the outpatient setting were termed outpatient NTM.

Analyses Plan

NTM episodes were stratified by specimen source, species, admission status (HO, CO, or outpatient), culture year, and hospital. Incidence rates of episodes for admitted patients were calculated per 10 000 patient-days, and outpatient rates were expressed as episodes per month of surveillance at individual hospitals.

Time-trended analyses of incidence rates were performed for episodes that occurred from January 2014 through December 2020 at the 7 hospitals that provided data on cultures performed in January 2013 or earlier. Requiring 12 months of hospital-specific baseline data prior to performing time-trended analyses allowed duplicate episodes for the same patient, NTM species, and specimen source category to be excluded. For episodes among admitted patients within this 7-hospital closed cohort, incidence rate ratios (IRRs) and annual trends in incidence from 2014 through 2020 were estimated using maximum likelihood, assuming the number of episodes followed the Poisson distribution. Wald 95% confidence intervals (CIs) were calculated, and likelihood ratio χ² tests were used to compare incidence rates.

We performed a sensitivity analysis to evaluate the impact of requiring more days of hospitalization prior to the culture date to meet HO criteria. We also performed a sensitivity analysis to evaluate the trends in incidence rates from 2014 through 2019, excluding 2020 episodes, which may have been affected by the coronavirus disease 2019 (COVID-19) pandemic [30].

We estimated the proportion of episodes that represented clinical NTM infection. From the pulmonary episodes, we excluded Mycobacterium gordonae, which typically reflects contamination or colonization and has been excluded from other NTM epidemiologic analyses [3, 31, 32]. We then excluded pulmonary episodes that did not meet microbiologic criteria for NTM-pulmonary disease (NTM-PD) based on 2020 international guidelines [33]. Per the findings of prior studies, we estimated that between 61% [32] and 86% [34] of pulmonary episodes that met microbiologic criteria also met clinical and radiologic criteria. Finally, we assumed that extrapulmonary episodes represented infection [3].

Calculations were performed in SAS, version 9.4 (SAS Institute, Cary, NC).

RESULTS

NTM Epidemiology in a 10-Hospital Network From 2012 Through 2020

Across the 10-hospital network, 24 376 total NTM isolates were identified from 2012 through 2020 during 19 248 137 patient-days of surveillance (Table 1). Among these isolates, 12 855 (53%) represented unique NTM episodes that occurred among 10 535 patients.

Of the 12 855 NTM episodes, 11 075 (86%) were pulmonary and 1780 (14%) were extrapulmonary (Table 2). Pulmonary episodes were predominantly diagnosed via sputum (n = 6939; 63%) or BAL/bronchial wash (n = 3557; 32%) cultures (Supplementary Table 1). The initial specimen source for extrapulmonary episodes was most commonly soft tissue (n = 291; 16%), blood (n = 248; 14%), skin (n = 224; 13%), or wound (n = 163; 9%; Supplementary Table 2). Among the 10 535 patients, 215 (2%) developed pulmonary and extrapulmonary episodes, including 180 cases where both episodes were caused by the same species. For these 180 cases, the time between initial NTM isolation from different sources was a median of 31 days (interquartile range [IQR], 2–196).

Table 2.

Characteristics of 12 855 Episodes of Nontuberculous Mycobacteria Isolation That Occurred From 2012 Through 2020 in a 10-Hospital Network

Characteristic	Hospital-Onset Episodes (n = 3045)	Community-Onset Episodes (n = 1865)	Outpatient Episodes (n = 7945)	All Episodes (n = 12 855)
Patient age, median (interquartile range), y	60 (47–69)	59 (43–69)	64 (49–73)	62 (47–71)
Pulmonary sourcea	2444 (80)	1365 (73)	7266 (91)	11 075 (86)
Sputum, endotracheal suction, or other airway specimen	1423 (58)	956 (70)	4914 (68)	7293 (66)
Bronchoalveolar lavage or bronchial wash	992 (41)	324 (24)	2241 (31)	3557 (32)
Lung tissue, bronchial tissue, or bronchial brushing	29 (1)	85 (6)	111 (2)	225 (2)
Extrapulmonary sourcea,b	601 (20)	500 (27)	679 (9)	1780 (14)
Soft tissue, skin, or wound	152 (25)	177 (35)	349 (51)	678 (38)
Blood	116 (19)	87 (17)	45 (7)	248 (14)
Intraabdominal	83 (14)	37 (7)	24 (4)	144 (8)
Pleural fluid/tissue	73 (12)	34 (7)	16 (2)	123 (7)
Sinus	5 (1)	11 (2)	76 (11)	92 (5)
Joint	15 (2)	46 (9)	29 (4)	90 (5)
Cardiothoracicc	33 (5)	19 (4)	11 (2)	63 (4)
Lymph node	15 (2)	10 (2)	18 (3)	43 (2)
Stool	25 (4)	12 (2)	6 (1)	43 (2)
Bone	13 (2)	18 (4)	10 (1)	41 (2)
Eye	0 (0)	4 (1)	33 (5)	37 (2)
Otherd	71 (12)	45 (9)	62 (9)	178 (10)
Species
Mycobacterium avium complex	1465 (48)	933 (50)	3792 (48)	6190 (48)
Mycobacterium abscessus complex	399 (13)	230 (12)	1082 (14)	1711 (13)
Mycobacterium gordonae	222 (7)	113 (6)	847 (11)	1182 (9)
Mycobacterium chelonae–Mycobacterium immunogenum	348 (11)	168 (9)	612 (8)	1128 (9)
Mycobacterium fortuitum	142 (5)	96 (5)	396 (5)	634 (5)
Mycobacterium mucogenicum	87 (3)	64 (3)	137 (2)	288 (2)
Mycobacterium kansasii	68 (2)	42 (2)	132 (2)	242 (2)
Mycobacterium arupense	55 (2)	45 (2)	79 (1)	179 (1)
Mycobacterium lentiflavum	23 (1)	15 (1)	110 (1)	148 (1)
Other speciese	236 (8)	159 (9)	758 (10)	1153 (9)

Data are presented as number (%) unless otherwise indicated.

^aDenominators for indented pulmonary and extrapulmonary subcategories are the total for the associated category.

^bSpecimens described in more specific subcategories were excluded from soft tissue, wound, and bone subcategories.

^cLymph node specimens were excluded from the cardiothoracic subcategory and listed separately.

^dOther specimen source subcategories for all episodes included extrapulmonary abscess, tissue, or fluid, not otherwise specified (n = 74); urine (n = 33); spine (n = 13); nose or nasal cavity (n = 11); liver/bile (n = 10); bone marrow (n = 9); vascular catheter (n = 7); cerebrospinal fluid (n = 5); ear (n = 4); oral cavity (n = 4); other vascular (n = 2); retropharyngeal abscess/tonsil (n = 2); cervical fluid (n = 1); hernia mesh (n = 1); larynx (n = 1); and an unspecified fistula (n = 1). Additional details of extrapulmonary sources are provided in Supplementary Table 2 and Supplementary Table 9.

^eFor 586 (5%) total episodes caused by other species, a nontuberculous mycobacterium was isolated, but the species could not be identified. Other species with ≥10 total episodes included Mycobacterium xenopi (n = 78), Mycobacterium terrae (n = 56), Mycobacterium marinum (n = 39), Mycobacterium szulgai (n = 38), Mycobacterium simiae (n = 29), Mycobacterium porcinum (n = 28), Mycobacterium neoaurum (n = 23), Mycobacterium scrofulaceum (n = 21), Mycobacterium nebraskense (n = 19), Mycobacterium peregrinum (n = 19), Mycobacterium paraffinicum (n = 18), Mycobacterium haemophilum (n = 17), Mycobacterium smegmatis (n = 15), Mycobacterium conceptionense (n = 11), Mycobacterium asiaticum (n = 10), and Mycobacterium obuense (n = 10). Less common species are detailed in Supplementary Table 10.

Mycobacterium avium complex (MAC) was the most common species isolated, accounting for 6190 (48%) of the 12 855 total episodes, followed by Mycobacterium abscessus complex (MABC; n = 1711; 13%), M. gordonae (n = 1182; 9%), and Mycobacterium chelonae– Mycobacterium immunogenum (n = 1128; 9%; Table 2). Compared with pulmonary episodes, species of rapidly growing mycobacteria accounted for a greater proportion of the 1780 extrapulmonary episodes, most notably for MABC (n = 394; 22%), M. chelonae–M. immunogenum (n = 341; 19%), and Mycobacterium fortuitum (n = 160; 9%; Supplementary Table 1, Supplementary Table 2). Of the 10 535 patients, 1676 (16%) had discrete episodes caused by different NTM species.

A total of 4910 (38%) episodes were diagnosed via cultures performed during hospital admissions, and 3045 (24%) episodes were classified as HO (Table 2). Specimens linked to HO episodes were obtained at a median of 6 days of hospitalization (IQR, 4–11; Supplementary Figure 2). The overall network incidence rate of HO episodes was 1.58 per 10 000 patient-days, ranging from 0.35 to 5.48 episodes per 10 000 patient-days across the 10 hospitals. The median hospital incidence rate was 1.06 episodes per 10 000 patient-days.

NTM Incidence Rates Within the 7-Hospital Closed Cohort From 2014 Through 2020

Within the 7-hospital closed cohort from 2014 through 2020, the incidence rate of NTM episodes among admitted patients decreased from 3.27 to 2.13 episodes per 10 000 days (IRR, 0.65; 95% CI, .57–.74; P < .0001; Figure 1, Supplementary Table 3, Supplementary 4). This decrease was largely explained by the decline in the HO incidence rate from 2.29 to 1.42 episodes per 10 000 days (IRR, 0.62; 95% CI, .53–.73; P < .0001; Figure 2, Table 3). Trend analysis using an unadjusted Poisson regression model estimated that the HO incidence rate decreased by 10% per year (95% CI, 8%–12%; P for trend < .0001). The trends in HO incidence rates did not meaningfully change after HO episodes were reclassified to begin on day 5 or day 7 of hospitalization rather than day 3 (Supplementary Table 5). Similarly, after excluding 2020 episodes, trends in HO rates remained consistent (Supplementary Table 6, Supplementary Figure 8).

Figure 1.

Incidence rates of episodes of nontuberculous mycobacteria that occurred from 2014 through 2020 within a 7-hospital cohort. Incidence rates for hospital-onset and community-onset admitted episodes were calculated per 10 000 patient-days. Outpatient incidence rates were calculated per month of surveillance at individual hospitals.

Figure 2.

Poisson regression model of HO episodes of nontuberculous mycobacteria that occurred from 2014 through 2020 within a 7-hospital cohort. Abbreviation: HO, hospital-onset.

Table 3.

Incidence Rates of Hospital-Onset Nontuberculous Mycobacteria Episodes that Occurred From 2014 Through 2020 Within a 7-Hospital Cohort

Characteristic	HO IRR for 2020 Versus 2014 (95% CI)	P Value	Estimated Percent Change in HO IR per Year (95% CI)	P Value for Trend
All HO episodes	.62 (.53 to .73)	<.0001	−10 (–12 to –8)	<.0001
Hospital number
1	.81 (.55 to 1.21)	.30	–7 (–11 to –3)	.001
2	.40 (.30 to .52)	<.0001	–18 (–21 to –14)	<.0001
3	.56 (.36 to .85)	.01	–12 (–17 to –7)	<.0001
4	.60 (.39 to .93)	.02	–10 (–16 to –4)	.001
5	1.07 (.63 to 1.82)	.81	0 (–8 to 9)	.99
6	.50 (.17 to 1.46)	.20	–14 (–26 to –1)	.04
7	1.55 (.77 to 3.15)	.22	8 (–3 to 20)	.15
NTM species
Mycobacterium avium complex	.96 (.75 to 1.23)	.75	–4 (–7 to –1)	.005
Mycobacterium abscessus complex	.41 (.26 to .62)	<.0001	–12 (–17 to –7)	<.0001
Mycobacterium chelonae–Mycobacterium immunogenum	.48 (.30 to .77)	.002	–15 (–21 to –8)	<.0001
Mycobacterium gordonae	.32 (.18 to .56)	<.0001	–23 (–29 to –17)	<.0001
Mycobacterium fortuitum	1.08 (.48 to 2.45)	.85	–3 (–13 to 7)	.51
Other NTM species	.47 (.31 to .71)	.0004	–14 (–19 to –9)	<.0001
Specimen source category
Pulmonary	.60 (.50 to .72)	<.0001	–10 (–12 to –8)	<.0001
Extrapulmonary	.72 (.50 to 1.05)	.09	–8 (–12 to –3)	.001

Incidence rates for 2020 were compared to 2014 incidence rates, and Poisson regression models were used to estimate the relative change in hospital-onset incidence rates per year.

Abbreviations: CI, confidence interval; HO, hospital-onset; IR, incidence rate; IRR, incidence rate ratio; NTM, nontuberculous mycobacteria.

The CO incidence rate in 2020 was lower than the 2014 rate due to the decrease in episodes that occurred in 2020. Sensitivity analysis revealed that the CO incidence rate was not significantly different in 2019 compared with 2014 (IRR, 1.06; 95% CI, .86–1.31; P = .60; Figure 1, Supplementary Table 6). The rate of outpatient episodes fluctuated during the study and was 7.89 episodes per month in 2020 compared with 8.46 episodes per month in 2014 (Supplementary Table 3).

HO incidence rates varied substantially by hospital (Figure 3). Of the 7 hospitals, 5 (71%) experienced significant decreases in annual HO incidence rates, which declined between 7% and 18% per year at these 5 hospitals (Table 3, Supplementary Figure 3). Hospitals 1 and 2 had the highest HO incidence rates and largest rate fluctuations. Hospital 1 experienced a large NTM outbreak from 2015 through 2017 caused predominately by MAC, and hospital 2 investigated a major clonal outbreak of MABC that occurred from 2013 through 2015. Both outbreaks were associated with the opening of a new inpatient hospital building [10] and were mitigated by using sterile water for patient care [27] and numerous additional hospital water safety improvements (Supplementary Table 7). Hospital 2 additionally mitigated the extrapulmonary phase of the MABC outbreak by using enhanced maintenance and disinfection protocols for heater–cooler units used during cardiac surgeries [35, 36]. Most other hospitals also implemented policies designed to improve water safety, particularly among vulnerable patients.

Figure 3.

Incidence rates of HO episodes of nontuberculous mycobacteria that occurred from 2014 through 2020 within a 7-hospital cohort, stratified by hospital. Abbreviation: HO, hospital-onset.

HO incidence rates also decreased over time when stratified by NTM species and specimen source category. Each of the 4 most common species exhibited significant reductions in annual HO incidence rates, which decreased between 4% and 23% per year, respectively (Table 3, Supplementary Figure 4, Supplementary Figure 5). HO incidence rates for pulmonary episodes decreased an estimated 10% per year (95% CI, 8%–12%; P for trend < .0001), and HO incidence rates for extrapulmonary episodes decreased 8% per year (95% CI, 3%–12%; P for trend = .001; Supplementary Figure 6, Supplementary Figure 7).

Estimates of Clinical NTM Infection

Of the 11 075 pulmonary episodes, 5912 (53%) were caused by species other than M. gordonae and met microbiologic criteria for NTM-PD [33] (Supplementary Table 8). We estimated that between 3606 (61%) [32] and 5084 (86%) [34] of these 5912 episodes also met clinical and radiologic criteria for NTM-PD. Assuming that all 1780 extrapulmonary episodes represented infection, we estimated that between 5386 (42%) and 6864 (53%) total episodes represented infection. Using the same approach to characterize the 3045 HO episodes, we estimated that between 1387 (46%) and 1710 (56%) HO episodes represented infection.

DISCUSSION

We performed a detailed analysis of the epidemiology of more than 12 000 incident NTM episodes that occurred among more than 10 000 patients over 9 years within a network of 10 US academic hospitals. Nearly 40% of episodes were diagnosed during hospitalizations, and most of these episodes, termed HO episodes, were linked to specimens obtained on day 3 or later of admission. Incidence rates of HO episodes varied considerably by hospital, with a 15-fold difference between the highest and lowest rates.

Within a closed cohort of 7 hospitals, the HO incidence rate decreased 38% from 2014 through 2020, with an estimated annual decrease of 10%. Five of the 7 hospitals experienced a significant downward trend in HO incidence, and a downtrend in incidence was also significant for each of the 4 most common NTM species, as well as for pulmonary and extrapulmonary HO episodes.

Multiple factors likely contributed to the impressive variability in HO incidence rates among hospitals. Municipal water supplies that serve hospitals have different disinfection protocols and NTM colonization burdens [37, 38]. In addition, heterogeneous hospital ages, plumbing infrastructure, and water management programs may have affected risk of NTM acquisition [39]. While all hospitals were large academic centers with thoracic transplant programs, differences in patient populations and thresholds for performing clinical mycobacterial cultures may also have influenced HO rates. Detailed evaluation of climate and geography [40] was outside the scope of this study, but we did not observe overt associations based on hospital locations.

The substantial decrease in HO incidence rates that occurred throughout the network over the course of this study contrasts with the increase in NTM rates that numerous population-based US studies have reported [19–22]. We believe that HO incidence rates decreased in our study, at least in part, due to improved hospital water safety practices throughout the network. Several hospitals limited tap water exposure to vulnerable patients, installed point-of-use water filters, or enhanced water disinfection procedures. The decrease in extrapulmonary HO episodes was likely partially due to widespread improvements in the maintenance and disinfection of heater–cooler units that began in 2015 in response to invasive infections in cardiac surgery patients caused by Mycobacterium chimaera [13] and other NTM species [14].

While the incidence rates of CO and outpatient NTM episodes were also lower in 2020 compared with 2014, these differences were primarily due to rate decreases that occurred in 2020. We suspect that the notably decreased isolation of NTM in these categories during 2020 was due to fewer elective admissions and outpatient diagnostics amidst the onset of the COVID-19 pandemic [30]. In contrast, the number of HO episodes was stable in 2020, despite an overall decrease in patient-days, suggesting that HO NTM episodes tend to affect patients with nonelective, more complex hospitalizations [7].

This study was unique because it characterized a large number of pulmonary and extrapulmonary NTM episodes that occurred among hospitalized patients and reported detailed incidence rates of HO episodes, stratified by hospital, species, and specimen source category. Simple definitions of episodes and HO criteria were applied to culture data via a streamlined approach without manual chart review. As a result, hospitals can now efficiently compare their rates of NTM isolation to past performance metrics at the same hospital and to external benchmarks, a critical tool that has not previously existed for calibration of hospital NTM surveillance. Use of standardized definitions and benchmark rates has the potential to dramatically improve HCFA NTM surveillance and help infection prevention teams identify the highest-yield targets for detailed, resource-intensive investigation and improved NTM prevention. For example, rate comparisons could instruct hospital water management and use [39], the potential utility of screening hospital water sources for NTM, and use of sequencing of clinical and environmental isolates to elucidate mechanisms of NTM acquisition [41, 42].

This study had 3 primary limitations. First, we used standardized NTM surveillance definitions intended to sensitively capture important HO cases. As a result, some NTM episodes categorized as HO were undoubtedly present on admission and diagnosed later. Likewise, other episodes categorized as CO or outpatient could have been acquired during prior hospitalizations. Importantly, we captured only incident diagnoses by excluding cultures indicative of persistent colonization or infection. In past investigations, these definitions accurately classified notable hospital-acquired NTM cases and outbreaks [10, 25, 27, 28]. Furthermore, surveillance definitions based on the number of days between hospital admission and microbiologic diagnosis have been used successfully to monitor other HAIs [26]. Therefore, hospitals may find these definitions useful and efficient for surveillance activities, including signaling the potential need for detailed investigations of HO cases. However, definitive adjudication of specific NTM cases concerning for healthcare acquisition is complex and often requires numerous steps, including careful clinical review, environmental sampling, and genomic analysis [41, 42]. Second, we did not differentiate NTM colonization from infection. While we estimated that approximately 50% of episodes represented clinical infection, we aimed to sensitively detect NTM acquisition that occurred at hospitals, regardless of progression to invasive infection. Broad inclusion of incident NTM episodes provides hospitals with a comprehensive picture of NTM transmission, facilitating early investigation and infection prevention, especially for patients at increased risk of developing clinical disease. Even for pseudo-outbreaks and NTM species that rarely cause infection, improved recognition and prevention of NTM acquisition has the potential to prevent use of unnecessary antibiotics and medical procedures [12, 29, 33]. Third, all centers in this study were large academic transplant hospitals. Generalizability of our findings is uncertain for hospitals that care for few immunosuppressed or critically ill patients or perform mycobacterial cultures less frequently. To improve understanding of HCFA NTM in other settings and promote prospective evaluation of standardized NTM surveillance definitions, public health authorities should consider making NTM a reportable condition in all states.

In conclusion, incidence rates of HO NTM varied substantially among study hospitals. Despite this variability, network-wide HO incidence rates consistently declined from 2014 through 2020. This analysis of HCFA NTM isolation used standardized surveillance definitions and provides external benchmarks that can be used to improve NTM surveillance. Prospective evaluation of this surveillance approach is warranted.

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Note

Financial support. A. W. B. was supported by the National Institute of Allergy and Infectious Diseases (NIAID) of the National Institutes of Health (NIH; grant K08-AI163462). J. A. A. was supported by NIAID of NIH (grant K01-AI137317).