|Home | About | Journals | Submit | Contact Us | Français|
Infective endocarditis is a potentially fatal consequence of illicit injection drug use. We estimate that the number of hospitalization for injection drug use–related infective endocarditis increased by 38%–66% in the United States between 2000–2001 and 2002–2003, a period during which the number of at-risk persons (i.e., injection drug users) remained stable. Increasing methamphetamine use and/or drug injection frequency may have increased the incidence of infective endocarditis among active injection drug users.
Infective endocarditis (IE) is a potentially fatal consequence of injecting illicit drugs, such as heroin, cocaine, and methamphetamine . Injection drug use (IDU) can increase IE risk through a variety of mechanisms. Drug solutions may contain particulate matter (e.g., talc) that damage cardiac valves if injected intravenously . In addition, poor injection hygiene (e.g., lack of skin cleaning before injecting), injecting with unsterile equipment, and injecting contaminated drug solutions can introduce high circulating bacterial loads . Finally, the physiologic effects of specific drugs may increase IE risk. For example, the prevalence of IE is high among persons who inject cocaine, and researchers and clinicians have partially attributed this elevated prevalence to the vasospasm and cardiac tissue damage that cocaine induces . We report on a nationwide increase in the number of hospitalizations for IDU-related IE during the period 2000–2003 in the United States and propose possible causes of this trend that merit investigation.
We identified cases of IDU-related IE that occurred from 1996 through 2003 in the Centers for Disease Control and Prevention National Hospital Discharge Survey database (NHDS) . The Centers for Disease Control and Prevention designed NHDS to produce nationally representative annual data on hospitalizations in non-Federal, short-term care facilities with 6 beds . Annually, NHDS captures ~270,000 inpatient stays in 500 hospitals located in the 50 states and Washington, DC . For each case, NHDS records patient sociodemographic characteristics, medical data (including up to 7 diagnoses and 4 procedures, coded using the International Classification of Disease, Ninth Revision, Clinical Modification system), and administrative items .
The International Classification of Disease, Ninth Revision, Clinical Modification system does not specify whether cases of IE are IDU related. Drawing on past efforts to identify cases of IDU-related IE in hospitalization data , we developed 2 algorithms to identify cases of interest in NHDS, using data on patient diagnoses, procedures, and sociodemographic characteristics; we designed one algorithm to be specific and the other to be sensitive. The specific algorithm classified cases in the database as IDU-related IE if the individual met each of the following criteria: (1) the patient received a diagnosis of bacterial endocarditis or endocarditis that was not otherwise specified; (2) the patient’s medical record included a diagnosis or procedure indicating the use of a commonly injected illicit drug (e.g., cocaine, heroin, or methamphetamine); diagnoses included opioid-, cocaine-, and amphetamine-related dependence, use, poisoning, or accidental death; and procedures included drug addiction counseling, detoxification, and rehabilitation; (3) the patient was 16–64 years of age when hospitalized (an age category that captures the majority of IDUs and excludes seniors, a group that is particularly susceptible to bacterial infections irrespective of injection status ); and (4) the patient had no documented conditions or procedures performed that predispose to IE (e.g., prosthetic valve or device, rheumatic or congenital heart disease, cardiac cathetherization, or dialysis).
Notably, this algorithm will misclassify cases of IDU-related IE as noncases in instances in which patients conceal their drug use from hospital staff. Therefore, we constructed a second, more sensitive algorithm to capture these hidden users; as with all sensitive measures, this algorithm will capture more cases than the specific algorithm at the expense of misclassifying some noncases as cases. On the basis of the epidemiology of IE , we designed the sensitive algorithm to capture individuals who received diagnoses of bacterial endocarditis and met criteria 3 and 4 but did not necessarily meet criterion 2. However, patients who received a diagnosis of endocarditis that was not otherwise specified had to meet criteria 2, 3, and 4 to be classified as having a case of IDU-related IE in the sensitive algorithm. (We considered including a diagnosis of hepatitis C virus infection in both the sensitive and the specific algorithms as another means of identifying IDUs, because hepatitis C virus infection is common among IDUs and is relatively rare among non-IDUs. The Centers for Disease Control and Prevention’s hepatitis C virus testing guidelines, however, broadened substantially in 1998 ; therefore, we decided against this modification. Given increased testing over time, classifying hepatitis C virus–positive individuals as IDUs would have artificially inflated temporal trends in the number of IDU-related IE cases detected.)
The number of cases of IDU-related IE that were identified by each algorithm was weighted to reflect the US population, using NHDS survey weights. We aggregated results into 2-year periods to ease interpretation (period 1, 1996–1997; period 2, 1998–1999; period 3, 2000–2001; and period 4, 2002–2003). The χ2 test was used to identify temporal changes in the number of IDU-related IE hospital discharges. Specifically, for every pair of consecutive periods, we tested whether the observed number of hospital discharges occurring during each period differed from the expected number; the expected number for each period was calculated as one-half of the sum of the number of IDU-related IE hospital discharges during 2 consecutive periods. Because HIV-related immunosuppression increases IE risk  and new therapies reducing HIV-related immunosuppression emerged during the study period , we present results stratified by the HIV infection or AIDS status of the patients.
Both the specific and sensitive algorithms suggest that the number of IDU-related IE hospital discharges nationwide was substantially higher during period 4 (2002–2003) (table 1), with slight to moderate fluctuations in the number of these discharges during earlier periods. According to the specific algorithm, there were 3858–5490 IDU-related IE hospital discharges nationwide during periods 1–3. In contrast, during period 4, the specific algorithm captured 9113 such discharges—an increase of 66.0% over period 3. According to the sensitive algorithm, there were 17,361–18,266 IDU-related IE hospital discharges nationwide during periods 1–3. The sensitive algorithm captured 24,025 such discharges during period 4—an increase of 38.4% over period 3. The χ2 test indicates that the number of discharges varied between each consecutive period, with 1 exception (the number of cases captured by the sensitive algorithm was statically identical for periods 1 and 2). A qualitative review of these data indicates that the change of greatest magnitude occurred between periods 3 and 4 for both algorithms. Notably, the number of IDU-related IE hospital discharges among people with HIV infection or AIDS decreased substantially during the study period (table 1).
Both algorithms indicate substantial increases in the number of hospital discharges for IDU-related IE nationwide between periods 3 and 4. To our knowledge, this is the first time such a trend in IDU-related IE has been reported nationwide, although another study reported an increase in such hospitalizations in New York City during the period 1996–1999 .
Possible causes of the observed nationwide increase merit consideration. First, it is unlikely that the observed increase is an artifact of changing IE diagnostic criteria. The Duke criteria have been the standard diagnostic tool for IE since 1994 . Modifications proposed to the Duke criteria in 2000 likely produced greater diagnostic specificity with little change in sensitivity among IDUs .
Second, available data indicate that the increase in the number of IDU-related IE hospitalizations was not produced by an increase in the size of the at-risk population (i.e., IDUs). An ongoing National Institute on Drug Abuse–funded study suggests that the number of past-year IDUs nationwide remained relatively constant during the study period (unpublished data from the Community Vulnerability and Response to IDU-related HIV project 5R01DA013336).
Third, the observed increase in the number of IDU-related IE does not appear to be a function of increasing prevalence of 2 factors associated with IE among active IDUs: HIV-related immunosuppression and cocaine injection. We found that cases of IDU-related IE among patients with HIV infection or AIDS decreased markedly during the study period—a finding consonant with past research that may reflect the advent of HAART and HIV-positive individuals’ improved access to addiction treatment . Likewise, drug treatment admissions data, which are often used as a marker of population trends in substance use, indicate that the number of treatment admissions of persons who injected cocaine was relatively stable during periods 3 and 4 .
Given the rising prevalence of community-acquired methicillin-resistant Staphylococcus aureus infection in the United States, we investigated whether this bacterium might have contributed to the observed increase in the number of IDU-related IE hospital discharges. Our analyses indicate that, although codiagnoses of community-acquired methicillin-resistant S. aureus infection among IDU-related IE hospital discharges indeed increased during the study period, this increase did not coincide with the increase in IDU-related IE hospital discharges observed here.
Possibly, the increase in the number of IDU-related IE is part of a broader increase in the number of IE overall, particularly given recent increases in the number of people who are vulnerable to IE (e.g., elderly individuals and people with prosthetic valves) . Our algorithms, however, excluded such vulnerable populations, and we suspect that, because of distinct causal mechanisms, the incidence of IDU-related IE follows a different temporal trend from that of non-IDU-related IE .
By reviewing (1) the known causes of IDU-related IE and (2) documented changes occurring during the study period in the contexts in which IDUs use drugs, we identified the mechanisms that may have recently increased the incidence of IE among active IDUs. First, increasing methamphetamine use may be such a mechanism, because drug treatment admissions of persons who currently use methamphetamine increased nationwide during the period 1996–2003, with the steepest increase occurring during 2001–2002 (admissions increased by 23% during these 2 years, from 133,971 in 2001 to 164,916 in 2002, and then increased again by 7% in 2003, to 176,764) . Methamphetamine use (regardless of the mode of administration) induces a host of cardiac complications that are similar to those induced by cocaine, including vasospasm and injury to the myocardial surface . Hypothetically, methamphetamine injection may increase IE risk through pathophysiologic mechanisms similar to those proposed for cocaine. A second mechanism may be increasing injection frequency among heroin injectors. The mean price of heroin decreased by 40%, from $403 per pure gram in 1995 to $241 per pure gram in 2002 . Dramatic price decreases may have allowed increased injection frequency among individuals who were already injecting heroin. Increased injection frequency provides more opportunity to introduce skin flora into the circulatory system.
The primary limitation of our study was the absence of codes denoting IDU in the International Classification of Disease, Ninth Revision, Clinical Modification system. We addressed this limitation by constructing 2 algorithms to identify cases of interest; one was designed to be specific, and the other was designed to be sensitive. Both algorithms suggest that IDU-related IE hospital discharges increased during the period 2002–2003—a concordance that reveals convergent validity.
Research regarding whether the 2 mechanisms proposed above or other mechanisms (acting alone or in combination) have produced the observed increase in the number of IDU-related IE cases is needed so that effective interventions can be established to stem IE among IDUs. Meanwhile, organizations serving IDUs could educate IDUs and health care providers about IE symptoms, so that affected individuals could obtain medical care as early as possible.
We thank Drs. David Perlman and Holly Hagan and 3 anonymous reviewers for their insightful comments on this manuscript.
Financial support. The Community Vulnerability and Responses to IDU-Related HIV (5R01DA013336 to H.L.F.C., S.R.F., B.T., J.E.B., and K.G.) and National Institute on Drug Abuse career development award (K23DA16165 to D.C.).
Potential conflicts of interest. All authors: no conflicts.