Bacterial Infections Associated With Substance Use Disorders, Large Cohort of United States Hospitals, 2012–2017

Abstract

Background.

Rises in the incidence of bacterial infections, such as infective endocarditis (IE), have been reported in conjunction with the opioid crisis. However, recent trends for IE and other serious infections among persons with substance use disorders (SUDs) are unknown.

Methods.

Using the Premier Healthcare Database, we identified hospitalizations from 2012 through 2017 among adults with primary discharge diagnoses of bacterial infections and secondary SUD diagnoses, using International Classification of Diseases, Clinical Modification Ninth and Tenth Revision codes. We calculated annual rates of infections with SUD diagnoses and evaluated temporal trends. Blood and cardiac tissue specimens were identified from IE hospitalizations to describe the microbiology distribution and temporal trends among hospitalizations with and without SUDs.

Results.

Among 72 481 weighted IE admissions recorded, SUD diagnoses increased from 19.9% in 2012 to 39.4% in 2017 (P < .0001). Hospitalizations with SUDs increased from 1.1 to 2.1 per 100 000 persons for IE, 1.4 to 2.4 per 100 000 persons for osteomyelitis, 0.5 to 0.9 per 100 000 persons for central nervous system abscesses, and 24.4 to 32.9 per 100 000 persons for skin and soft tissue infections. For adults aged 18–44 years, IE-SUD hospitalizations more than doubled, from 1.6 in 2012 to 3.6 in 2017 per 100 000 persons. Among all IE-SUD hospitalizations, 50.3% had a Staphylococcus aureus infection, compared with 19.4% of IE hospitalizations without SUDs.

Conclusions.

Rates of hospitalization for serious infections among persons with SUDs are increasing, driven primarily by younger age groups. The differences in the microbiology of IE hospitalizations suggest that SUDs are changing the epidemiology of these infections.

Drug overdose deaths have increased considerably across the United States, primarily driven by an epidemic of opioid-involved overdose deaths [1]. While the risks of bloodborne virus transmission are well known, people with substance use disorders (SUDs), particularly those who inject drugs, are also at increased risk for serious bacterial infections associated with significant morbidity and mortality [2–6]. For example, a recent study estimated that people who inject drugs were 16 times more likely to develop invasive methicillin-resistant Staphylococcus aureus (MRSA) infections [7]. While use of sterile injection equipment (including needles, water, cottons) is an effective means of reducing transmission of bloodborne pathogens, even sterile needles can be contaminated with skin or environmental flora during injection and cause local or systemic bacterial or fungal infections [8].

The spectrum of infections caused by SUDs range from skin and soft tissue infections (SSTIs), the most common reason for hospital admission and emergency department visits among this population [9, 10], to infective endocarditis (IE) [11]. Recent studies have reported an increase in incidence of IE among persons with SUDs, especially in younger populations [12–14]. Other serious complications of injection drug use include osteomyelitis and abscesses of the central nervous system (CNS) and adjacent structures, such as epidural abscesses [5, 6, 15].

The incidence of drug overdose deaths has increased in recent years [1]. A previous multicenter study suggested that the incidence of serious infections related to SUDs has also increased from 2009 to 2013 [6]. However, an analysis of more recent national trends has not been conducted. In this study, we describe trends in hospitalizations due to serious bacterial infections among persons with SUDs and characterize the microbiology of these infections from electronic health records.

METHODS

We conducted a retrospective cohort study using the Premier Healthcare Database (PHD), a comprehensive electronic healthcare database from approximately 800 private and academic hospitals, representing approximately 20% of US inpatient discharges. The PHD contains data on sociodemographics, comorbidities, procedures, medications, patient charges and costs, and outcomes that are based on International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] and International Classification of Diseases, Tenth Revision, Clinical Modification [ICD-10-CM]) codes for diagnoses and procedures. Additionally, the PHD contains microbiology laboratory data from approximately 200 hospitals, including specimen identification, test name, test day of service and time, and result and sensitivity data [16].

We identified hospitalizations among persons aged ≥18 years from 2012 through 2017 with primary discharge diagnoses of selected bacterial infections using ICD-9-CM and ICD-10-CM diagnostic codes (Supplementary Table 1). We analyzed 4 diagnosis groups separately: IE, CNS abscess (including intracranial, intraspinal, and epidural abscesses), osteomyelitis, and SSTIs.

We described characteristics of patients with selected infections, including age, sex, race, insurance, region, length of stay, death, secondary diagnoses of interest, and total hospital costs to treat patients during hospital encounters. Total costs represent the sum of fixed costs and variable costs, where fixed costs included all overhead costs while variable costs included in-hospital services including procedures, room and board, services provided by hospital staff, and pharmacy costs. Deaths included deaths that occurred in the inpatient setting and discharges to hospice. Secondary diagnoses of interest included human immunodeficiency virus infection; viral hepatitis B, C, or D virus infection; and SUDs. We identified secondary SUD diagnoses among hospitalizations of serious infections using ICD-9 and ICD-10 diagnosis codes derived from the Clinical Classifications Software category 661 for “substance-related disorders” [17] (Supplementary Table 2), excluding alcohol, nicotine, and cannabis. Among hospitalizations for patients with SUDs, substances were classified as opioids, cocaine, amphetamines, hallucinogens, or other/unspecified drugs. SUD hospitalizations may include diagnosis codes for more than 1 substance type.

We used sampling weights provided by PHD to extrapolate results to nationwide hospitalizations with primary diagnosis codes for selected infections. Annual rates of hospitalizations were calculated by dividing the weighted number of hospitalizations by US Census population estimates for each year. Among hospitalizations with SUD diagnoses, we described the proportion with at least 1 opioid-related diagnosis code and also characterized length of stay, deaths, and hospitalization costs. To detect any discrepancies in coding due to the ICD-9 to ICD-10 transition, we initially analyzed temporal trends within 2 discrete time periods: the beginning of the study period (1 January 2012) up to the ICD-9/ICD-10 transition (30 September 2015) and from 1 October 2015 through 31 December 2017 (Supplementary Table 3). Descriptive statistics included means for continuous variables and frequencies for categorical variables as appropriate. We evaluated trends in the rates of primary IE, CNS abscess, osteomyelitis, and SSTI hospitalizations with secondary diagnoses of SUDs (IE-SUD, CNS-SUD, osteomyelitis-SUD, and SSTI-SUD, respectively) from 2012 through 2017. We tested for the presence of temporal trends via weighted regression using generalized estimating equations, including clustering of the data within hospital provider. To determine whether increased awareness and coding of SUDs could be responsible for an increase in the incidence of SUD diagnoses, we examined rates of hospitalizations with 2 separate control outcomes, leukemia or thyroid disorders, not expected to be associated with SUDs.

To assess whether microbiology of infections changed, we evaluated IE hospitalizations with available laboratory data. We described the distribution of positive cultures from blood and cardiac tissue specimens, compared the distributions among IE-SUD and non-SUD–related IE hospitalizations, and described temporal trends of positive cultures among all IE hospitalizations. For IE hospitalizations with all negative results among blood and cardiac tissue specimens, we characterized patients who had been transferred from another healthcare facility using the point-of-origin code provided by PHD. This code has been previously verified and is comparable to physician-documented history [18]. These data do not contain direct personal identifiers, and institutional review board approval was not required. All statistical analyses were conducted using SAS version 9.4 (SAS Institute Inc, Cary, NC).

RESULTS

The annual rates and outcomes of patients hospitalized with each group of selected infections are presented in Table 1. There were 72 481 weighted admissions with a primary diagnosis of IE recorded; of those, 20 784 (28.7%) included SUD diagnoses. From 2012 through 2017, the overall rate of IE and CNS abscess hospitalizations did not change (P = .0958 and P = .3315, respectively) and the overall rates of osteomyelitis and SSTI hospitalizations decreased (P = .0139 and P = .0022, respectively). The proportion of IE hospitalizations with a diagnosis of SUD increased from 19.9% to 39.4% (P < .0001). Among 242 651 osteomyelitis hospitalizations, prevalence of SUDs increased from 8.2% to 16.6% (P < .0001). Prevalence of SUDs among 48 742 CNS abscess admissions increased from 15.4% in 2012 to 23.8% in 2017 (P < .0001), and prevalence of SUDs among patients hospitalized for SSTIs increased from 6.8% to 9.4% (P < .0001). Among all infection hospitalizations with SUDs, 58.6% had ≥1 discharge diagnosis code related to opioids, while 40.8% had ≥1 code related to cocaine, 18.0% related to amphetamines, 0.2% related to hallucinogens, and 55.1% related to other drugs or unspecified drug use. Some hospitalizations had discharge diagnosis codes for more than 1 substance type, and thus categories are not mutually exclusive.

Table 1.

Annual Rates and Outcomes of Patients Hospitalized With Selected Bacterial and Fungal Infections as Primary Diagnosis, Premier Healthcare Database, 2012–2017

	2012	2013	2014	2015	2016	2017
Infective endocarditis: all hospitalizations
Weighted N	12 415	12 081	11 600	11 686	11 760	12 939
Rate per 100 000 persons	5.3	5.1	4.8	4.8	4.8	5.3
Substance use disordersa N (%)	19.9	23.8	26.4	29.7	32.6	39.4
Infective endocarditis: hospitalizations with substance use diagnoses
Opioidb (%)	55.6	58.2	54.9	62.4	67.4	72.4
Length of stay, mean, days	14.3	15.3	15.9	14.9	15.3	15.1
Diedc (%)	5.8	3.7	3.5	3.6	3.5	2.1
Hospitalization costs, mean	US$30 515	US$32 038	US$36 765	US$32 695	US$35 860	US$33 610
Osteomyelitis: all hospitalizations
Weighted N	39 246	39 623	40 537	40 014	47 458	35 773
Rate per 100 000 persons	16.7	16.7	16.9	16.5	19.4	14.5
Substance use disordersa N (%)	8.2	9.0	11.1	11.2	10.5	16.6
Osteomyelitis: hospitalizations with substance use diagnoses
Opioidb (%)	46.4	48.8	47.1	52.5	56.4	60.4
Length of stay, mean, days	11.5	11.9	10.8	10.5	12.3	11.6
Diedc (%)	0.9	1.0	0.8	0.3	0.8	0.4
Hospitalization costs, mean	US$20 706	US$21 432	US$19 366	US$20 551	US$24 815	US$22 607
CNS abscess: all hospitalizations
Weighted N	7801	7722	7480	7831	8764	9144
Rate per 100 000 persons	3.3	3.3	3.1	3.2	3.6	3.7
Substance use disordersa N (%)	15.4	15.9	17.5	18.9	19.6	23.8
CNS abscess: hospitalizations with substance use diagnoses
Opioidb (%)	50.5	56.7	47.8	52.4	60.8	59.1
Length of stay, mean, days	17.2	13.8	15.2	14.6	15.2	14.7
Diedc (%)	2.3	1.6	1.3	0.8	2.3	1.8
Hospitalization costs, mean	US$36 675	US$28 785	$33 190	US$31 515	US$31 151	US$30 064
SSTI: all hospitalizations
Weighted N	845 300	822 250	804 036	797 066	862 065	860 161
Rate per 100 000 persons	360.1	347.0	335.9	329.0	352.1	349.2
Substance use disordersa N (%)	6.8	7.8	8.5	9.0	8.7	9.4
SSTI: hospitalizations with substance use diagnoses
Opioidb (%)	48.9	47.7	47.6	49.1	54.7	53.8
Length of stay, mean, days	4.7	4.7	4.8	4.7	4.8	4.8
Diedc (%)	0.4	0.4	0.5	0.4	0.5	0.5
Hospitalization costs	US$8433	US$8212	US$8964	US$9137	US$9946	US$10 254

Abbreviation: CNS, central nervous system; SSTI, skin and soft tissue infection.

^aIncludes opioids, cocaine, amphetamines, hallucinogens, or other/unspecified drugs (Supplementary Materials).

^bProportion of hospitalizations of selected infections with at least 1 opioid-related diagnosis code (Supplementary Materials).

^cIn-hospital death or discharge to hospice.

Table 2 describes rates of hospitalizations for selected infections with SUD diagnoses by age group. Rates for each group of infections increased across the study period, primarily in younger age groups. Among all age groups, the rate of hospitalizations with diagnoses of SUDs increased from 1.1 to 2.1 per 100 000 persons for IE hospitalizations, 1.4 to 2.4 per 100 000 persons for osteomyelitis hospitalizations, 0.5 to 0.9 hospitalizations per 100 000 persons for CNS abscess hospitalizations, and 24.4 to 32.9 per 100 000 persons for SSTI hospitalizations. The rates of hospitalizations for persons with thyroid disease or leukemia and who also had SUD diagnoses were unchanged (0.3 and 0.2 hospitalizations per 100 000 persons, respectively).

Table 2.

Rates of Hospitalizations for Selected Infections, Thyroid Disorders, and Leukemia With Substance Use Disorder Diagnoses, Premier Healthcare Database, 2012–2017

		Year
		2012	2013	2014	2015	2016	2017
Infective endocarditis
18–44 years	Weighted N	1803	2048	2328	2579	3002	4089
	Rate per 100 000 persons	1.6	1.8	2.1	2.3	2.6	3.6
45–64 years	Weighted N	614	755	696	830	783	972
	Rate per 100 000 persons	0.7	0.9	0.8	1.0	0.9	1.2
65+ years	Weighted N	60	39	43	62	45	38
	Rate per 100 000 persons	0.1	0.1	0.1	0.1	0.1	0.1
Osteomyelitis
18–44 years	Weighted N	1332	1590	2076	2141	2229	2884
	Rate per 100 000 persons	1.2	1.4	1.8	1.9	2.0	2.5
45–64 years	Weighted N	1639	1691	2104	1983	2361	2750
	Rate per 100 000 persons	2.0	2.1	2.5	2.4	2.8	3.3
65+ years	Weighted N	252	269	330	359	396	308
	Rate per 100 000 persons	0.6	0.6	0.7	0.8	0.8	0.6
Central nervous system abscess
18–44 years	Weighted N	525	537	605	703	784	1036
	Rate per 100 000 persons	0.5	0.5	0.5	0.6	0.7	0.9
45–64 years	Weighted N	604	566	682	717	834	1047
	Rate per 100 000 persons	0.7	0.7	0.8	0.9	1.0	1.3
65+ years	Weighted N	72	84	84	62	96	97
	Rate per 100 000 persons	0.2	0.2	0.2	0.1	0.2	0.2
Skin and soft tissue infection
18–44 years	Weighted N	32 620	36 859	38 270	41 093	43 061	47 111
	Rate per 100 000 persons	29.3	33.0	34.1	36.4	37.9	41.4
45–64 years	Weighted N	22 276	24 072	26 405	26 708	28 011	29 079
	Rate per 100 000 persons	27.2	29.4	32.0	32.1	33.5	34.9
65+ years	Weighted N	2466	3183	3691	4101	4211	4743
	Rate per 100 000 persons	5.9	7.4	8.3	8.9	8.9	9.6
Thyroid disease
18–44 years	Weighted N	292	331	371	389	361	343
	Rate per 100 000 persons	0.3	0.3	0.3	0.3	0.3	0.3
45–64 years	Weighted N	245	402	355	368	305	310
	Rate per 100 000 persons	0.3	0.5	0.4	0.4	0.4	0.4
65+ years	Weighted N	115	78	79	103	56	91
	Rate per 100 000 persons	0.3	0.2	0.2	0.2	0.1	0.2
Leukemia
18–44 years	Weighted N	153	167	179	167	243	192
	Rate per 100 000 persons	0.1	0.1	0.2	0.1	0.2	0.2
45–64 years	Weighted N	210	227	267	253	289	235
	Rate per 100 000 persons	0.3	0.3	0.3	0.3	0.3	0.3
65+ years	Weighted N	125	118	140	91	87	130
	Rate per 100 000 persons	0.3	0.3	0.3	0.2	0.2	0.3

Includes opioids, cocaine, amphetamines, hallucinogens, or other/unspecified drugs (Supplementary Materials).

Overall, the rate of non-SUD–related IE hospitalizations without SUD diagnoses decreased from 4.2 hospitalizations per 100 000 persons in 2012 to 3.2 hospitalizations per 100 000 persons in 2017. Figure 1 compares temporal trends of IE-SUD hospitalizations by age group. Among patients aged 18–44 years, the rate of IE-SUD hospitalizations more than doubled from 1.6 per 100 000 persons in 2012 to 3.6 per 100 000 persons in 2017, resulting in approximately 4089 hospitalizations in 2017. In contrast, non-SUD–related IE hospitalizations for patients aged 18–44 years decreased slightly (Figure 1). Rates of non-SUD hospitalizations for CNS abscess, osteomyelitis, and SSTIs were also stable or decreased slightly over the study period (data not shown).

Figure 1.

Weighted infective endocarditis hospitalizations with and without substance use disorder diagnoses, Premier Healthcare Database, 2012–2017. Includes opioids, cocaine, amphetamines, hallucinogens, or other/unspecified drugs (see Supplementary Materials). Abbreviations: IE-SUD, infective endocarditis hospitalization with substance use disorder diagnoses; IE-Non-SUD, infective endocarditis hospitalization without substance use disorder diagnoses.

Comprehensive tables that include characteristics of hospitalizations for each group of serious infections by year are available in Supplementary Table 3. The proportion of white patients and Medicaid patients among all hospitalizations increased over the study period. Among all IE, OM, and CNS abscess hospitalizations, the percentage of patients aged 18–44 years increased over the study period, while the percentage of SSTI hospitalizations increased in patients aged ≥65 years. Among all infection hospitalizations except for SSTI, mean length of stay and median costs were higher in hospitalizations with SUD diagnoses: 15.2 vs 11.0 days and $19 320 vs $17 125 for IE hospitalizations, 11.5 vs 7.6 days and $12 910 vs $10 824 for osteomyelitis hospitalizations, and 15.1 vs 10.9 days and $22 073 vs $20 067 for CNS abscess hospitalizations. IE, OM, and CNS abscess patients with SUDs were also more likely to have a length of stay >20 days (26% vs 12%, 15% vs 4%, and 22% vs 11%, respectively). Among SSTI SUD and non-SUD hospitalizations, the mean length of stay was 4.8 days compared with 5.2 days, and median costs were $9232 compared with $10 054, respectively.

Among 17 063 unweighted IE hospitalizations, 4120 hospitalizations (24%) occurred in facilities with available microbiology test result data. Of these, there were 3149 hospitalizations with culture results from blood or cardiac tissue specimens, 1826 hospitalizations (58%) with at least 1 positive culture, and 1323 hospitalizations (42%) with negative culture results from blood or cardiac tissue specimens. Of the hospitalizations with negative culture results, 522 (40%) were transferred from another healthcare facility or clinic, 771 (58%) were transferred from a nonhealthcare facility (including possible clinic referral or transfer from an ambulatory surgery center), and 30 (2%) had an unknown point of origin.

The distribution of microorganisms identified from IE hospitalizations with positive cultures are described in Table 3. Approximately half of IE-SUD hospitalizations had a positive culture for S. aureus compared with 19.4% of non-SUD–related IE hospitalizations. Viridans streptococci were most common among non-SUD–related IE hospitalizations (29.4%). While gram-negative and fungal/mycobacterial infections were less common overall, significantly more were found in IE-SUD hospitalizations (21.6%) compared with non-SUD–related IE hospitalizations (8.6%). Among all positive S. aureus cultures, 45% were methicillin resistant (MRSA). Figure 2 describes the temporal trends of microorganisms identified among IE hospitalizations. Overall, the prevalence of S. aureus increased from approximately 24% of IE hospitalizations with positive cultures in 2012 to 39% in 2017. Conversely, prevalence of viridans streptococci decreased from 27% of IE hospitalizations with positive cultures in 2012 to 20% in 2017. For both IE-SUD and non-SUD–related IE hospitalizations, prevalence of S. aureus increased and prevalence of viridans streptococci decreased (data not shown).

Table 3.

Distribution of Microorganisms Among Infective Endocarditis Hospitalizations With Positive Cultures, Premier Healthcare Database, 2012-2017

Microorganism	Infective Endocarditis Hospitalization With SUD Diagnoses (n)	%a	Infective Endocarditis Hospitalization Without SUD Diagnoses (n)	%a
Gram-positive cultures
Staphylococcus aureus	235	50.3	263	19.4
Other staphylococci	31	6.6	188	13.8
Viridans group streptococci	84	18.0	400	29.4
Other streptococci	13	2.8	107	7.9
Enterococcus faecalis	48	10.3	246	18.1
Enterococcus faecium	1	0.2	20	1.5
Other enterococci	4	0.9	21	1.5
Other gram-positive	23	4.9	84	6.2
Gram-negative cultures
Pseudomonas spp.	20	4.3	9	0.7
Serratia spp.	18	3.9	3	0.2
Stenotrophomonas maltophilia	7	1.5	4	0.3
Enterobacter spp.	6	1.3	5	0.4
Escherichia coli	0	0.0	13	1.0
Klebsiella spp.	0	0.0	13	1.0
Other gram-negative	24	5.1	42	3.1
Other
Candida spp.	23	4.9	21	1.5
Other fungal/mycobacterial infection	3	0.6	7	0.5
Other positive cultureb	10	2.1	47	3.5

Abbreviation: SUD, substance use disorder.

^aPercentage of hospitalizations with at least 1 positive culture (N = 1826). Hospitalizations may have more than 1 positive culture resulting in the total exceeding 100%.

^bRare organisms or insufficient information to classify.

Figure 2.

Trends of microorganisms among infective endocarditis hospitalizations with positive cultures, Premier Healthcare Database, 2012–2017 (N = 1826). Abbreviation: S. aureus, Staphylococcus aureus.

DISCUSSION

While the overall rates of hospitalizations for IE, CNS abscess, osteomyelitis, and SSTIs remained stable or decreased between 2012 and 2017, the proportion of admissions with SUD diagnoses increased significantly among all serious infection hospitalizations. Most striking were increases in rates of hospitalizations with SUD diagnoses among patients aged 18–44 years, for whom the estimated rate of IE-SUD hospitalizations more than doubled from 1.6 in 2012 to 3.6 in 2017 per 100 000 people. These data also provide insight into the microbiology of IE among hospitalizations with and without SUD diagnoses using laboratory data from more than 200 US hospitals. The differences in IE-SUD and non-SUD–related IE hospitalizations, including length of stay, microbiology, as well as the significant increase in IE-SUD hospitalizations over the study period, suggest that SUDs are changing the epidemiology of these infections.

The incidence of drug overdose deaths has been rising over the past few years [1], and communities are reporting shifts in behavior toward an increase in the use of injection as the route of drug delivery [19, 20]. In parallel, the incidence of infections such as endocarditis and invasive MRSA in patients with SUDs or who inject drugs has been increasing [7]. The finding that S. aureus and viridans streptococci were the most common pathogens among IE-SUD patients suggests that additional emphasis on strategies to prevent infections caused by inadvertent injection of skin and mouth flora could be important for prevention of endocarditis [8].

Current efforts to prevent infections among SUD patients are focused on prevention and treatment of underlying SUDs and injection drug use in the community [21]. Increased awareness of harm reduction principles and education regarding the risks related to substance use during clinical treatment for infections are complementary strategies for addressing the substantial and rising burden of these acutely life-threatening diseases. A recent report from the Centers for Disease Control and Prevention advocates for routine care of persons who inject drugs to include advice on hand hygiene, not injecting into skin that has not been cleaned, and not using any equipment contaminated by reuse, saliva, soil, or water [22]. Proposals to provide comprehensive care that addresses both substance use treatment and prevention of infectious complications have included reducing barriers to medication-assisted treatment prescribing, conducting demonstration projects that focus on cotreatment in locations that routinely have SUD clients, and developing a subspecialty that combines infectious diseases and addiction treatment expertise [23–25].

This analysis has many strengths. Approximately 800 US hospitals were included for each year of the study, including both community and teaching hospitals from both rural and urban areas. Using PHD weights, we extrapolated our results to produce national estimates for primary diagnoses. We included 6 years of data, allowing for trend analyses, and provided updated recent national estimates up to 2017. Our ability to use the PHD to approximate results by Collier et al [6], who estimated serious infection hospitalization rates using the Nationwide Inpatient Sample, adds to the validity of this study (data not shown).

Several limitations should be noted. First, we relied on ICD-9-CM and ICD-10-CM discharge codes to identify diagnoses. These administrative data are collected primarily for billing purposes and adapted for research, and there may be misclassification in SUD codes. For example, ICD-10 codes that relate to “use” of opioids might be used for patients without an opioid use disorder but who are using opioids as prescribed. However, ICD-10 “use” codes were only found in approximately 1% of hospitalizations categorized as having SUDs in this study. Additionally, causal inferences cannot be obtained solely from diagnosis codes given during the same encounter. Infections could have been acquired from nondrug use-related exposures.

Coding practices may vary by facility and may have changed over the study period, particularly during the transition to ICD-10 [26] and as awareness of the opioid crisis has continued to increase in recent years. However, in a qualitative assessment, we did not observe substantial shifts in trends following the ICD-10 transition (see Supplementary Table 3). Additionally, our secondary analysis of non-SUD–related IE hospitalizations among patients aged 18–44 years showed no inverse trend. If increased coding of SUDs was entirely driving these results, we would expect a decrease in the rate of non-SUD-related IE of approximately half to account for the doubled rates of SUD-related hospitalizations in this age group, but the rate declined only slightly (1.4 to 1.2 per 100 000 persons). In addition, we found rates of leukemia and thyroid diseases hospitalizations, which are not likely to be associated with SUDs, to be relatively stable over this time, which also suggests that there is not a general trend to increase SUD coding during the study period. A further complication of reliance on administrative coding is underreporting of illicit drug use, with studies showing that 30%−50% of SUDs are not appropriately coded [27–29]. In a recent study, drug use was not recorded at the time of infection, but within 6 months, for more than half of SUD-related infections identified [30]. By including only inpatient diagnoses recorded during the same encounter, our study likely underestimates the burden of SUD-related infections. Finally, we could not determine routes of drug administration due to the lack of specific ICD-CM codes for injection drug use. We conducted a secondary analysis that was limited to substances more likely to be injected (opioids, cocaine, and amphetamines) and found a 2.5-fold increase in IE-SUD hospitalizations over the study period compared with a 2-fold increase when our more sensitive definition was used.

We used only the primary diagnosis code to identify patients within the 4 disease groups, and these syndromes may represent different facets of the same infection. Although this increases our positive predictive value of the coded definition, and validation studies have suggested good positive predictive value specifically for SSTI and osteomyelitis [31], it would likely underestimate infection burden in the United States. Therefore, our estimates, while similar to those from a previous study that used only primary codes [6], likely represent a minimum burden for these disease groups. In a sensitivity analysis that included diagnosis codes present on admission in any position for IE hospitalizations, we found an almost 4-fold increase in hospitalizations; however, increases in associated SUDs were comparable across IE definitions (data not shown).

Our analyses only included deaths that occurred in a hospital or hospice, which might not capture all deaths due to these infections. Finally, our microbiology analyses were limited to data available in PHD, which may not be representative of all IE hospitalizations in this study. Additionally, among the 43% of hospitalizations with microbiology data that showed negative culture results, 40% were transferred from another healthcare facility and may not be new IE hospitalizations. Some of these patients may have previously received antimicrobial medications, resulting in negative cultures (eg, treated at another facility if transferred or self-treated at home).

CONCLUSIONS

Our analyses have important clinical, economic, and public health implications. National rates of hospitalizations for serious bacterial infections among persons with SUDs continue to increase, driven primarily by younger age groups. These data highlight an important correlate of the ongoing opioid crisis and indicate the threat posed to recent progress in reducing the impact of invasive bacterial diseases such as S. aureus infection [32]. Interventions and health communications aimed at reducing the risk of bacterial infections as a complication of substance misuse should be added to preventive care for persons with SUDs [22].