We found both the number and rate of US short-stay hospital discharges with a diagnosis of CDAD increased from 2000 through 2003. The overall rate and increase in the number of discharges were most prominent among patients >65 years of age. Although the overall rate was highest in the Northeast, a significant linear increase was found only in the Midwest and South. Rates increased similarly in middle- and large-sized hospitals according to number of beds, and overall rates were similar in these 2 strata; in contrast, small hospitals had a lower overall rate and did not experience a significant increase.
Our study had several limitations. We analyzed only hospital discharge data and the sensitivity and specificity of hospital discharge coding for CDAD are largely unknown (16,17
). Although ICD-9-CM code for "an intestinal infection with C
" contains the organism name, the code may be used on the basis of the clinical suspicion of CDAD alone, in some instances without a positive laboratory test result. In a single-institution study, when the number of positive C
laboratory results were compared to the number of ICD-9-CM coded diagnoses, ICD-9-CM diagnoses overestimated by 32% the number of cases predicted by positive laboratory results (16
). However, a month-to-month correlation was found between cases detected by a positive C
laboratory test result and ICD-9-CM coded diagnosis; when patient medical records of these disparate cases (i.e., agreement with ICD-9-CM code and negative for C
laboratory results or not tested) were reviewed in detail, most had a history of CDAD during a previous admission or had a C
laboratory test ordered that either had a negative result or was cancelled before specimen collection. Another recent comparison of laboratory and ICD-9-CM data from all US Veterans Affairs hospitals demonstrated that ICD-9-CM coded diagnoses of CDAD underestimated by approximately half the number of CDAD patient discharges determined by positive laboratory results (17
). These investigators also confirmed that ICD-9-CM coded diagnoses of CDAD correlated with the number of CDAD patient discharges determined by positive laboratory results, both among different hospitals as well as over time within individual hospitals.
Along with the potential insensitivity of coding, limitations involve the sensitivity of commonly used diagnostic tests for CDAD. Many hospital laboratories have migrated away from performing culture for C
(sensitive but nonspecific for toxin-producing strains) or tissue culture cytotoxin assays toward the performance of less time-consuming, but generally less sensitive, toxin immunoassays. Data are available on diagnostic methods other than bacterial culture collected as part of proficiency surveys conducted by the College of American Pathologists (CAP) between 1999 and 2003 (18
). Responses from at least 2,250 North American clinical laboratories surveyed annually suggest that, from 1999 through 2003, <5% laboratories overall performed tissue culture cytotoxin assays. Although this proportion is small, even if it were much larger, sensitivity could still be an issue as some evidence has shown that even a tissue culture cytotoxin assay performed directly on stool misses a large proportion of patients with diarrhea and low numbers of cytotoxin-producing C
. difficile (19)
One possible interpretation of our findings is that the observed increase in CDAD reflects a migration away from immunoassays that detect only toxin A toward use of assays that detect toxin A and B after reports of fatal cases associated with toxin A–negative and toxin B–positive strains (20
). Indeed, the data from CAP mentioned above indicate such a trend: 78% of laboratories in 1999 performed toxin A immunoassays only and 7% performed toxin A and B immunoassays, whereas by 2003, 55% performed toxin A only immunoassays, and 38% performed toxin A and B immunoassays (18
). Although this change in testing practices was gradual, even if it had been more sudden, the increased sensitivity of combined toxin A and B immunoassays could not likely explain the observed increase in rates between 2000 and 2003 because toxin A-B–positive strains account for no more than 5% of C
). Nonetheless, to address this issue as well as the possibility that more CDAD cases were being diagnosed due to increased C
testing, we analyzed rates of diarrhea using ICD-9-CM codes for nonspecific causes and found no significant decrease. This suggests that the observed increases in CDAD were not simply due to diagnosing more disease among patients with diarrhea of nonspecific causes (data not shown).
Our findings suggest a large impact of excess illness and costs posed by CDAD on the US healthcare system; an impact that may approach or exceed that caused by other more widely recognized nosocomial pathogens. For example, the number of US short-stay hospital discharge diagnoses of CDAD during either 2001, 2002, or 2003 exceeded the estimated annual number (120,000) of methicillin-resistant Staphylococcus aureus
(MRSA) infections for 1999–2000 (22
). Kyne et al. recently estimated that each case of CDAD in their hospital was associated with $3,699 in excess healthcare costs and 3.6 extra days of hospitalization (7
). Based upon their hospital's rate of CDAD in 0.7% of discharges, they estimated that the total excess in US healthcare costs attributable to CDAD was likely >$1.1 billion. Even using our lower estimate for the total US cases in 2003 (i.e., 0.51% or 178,000), CDAD can be estimated to have resulted in >$600 million in excess healthcare costs and >600,000 excess hospital days in nonfederal facilities. However, such estimates only account for resource use in short-stay hospitals. Even though some residents infected in long-term care facilities receive their treatment in short-stay facilities, these estimates of excess healthcare costs do not account for the infection control and medication costs incurred within long-term care facilities.
Although our findings are consistent with a recent analysis of CDAD rates in NNIS hospitals (12
), our results highlight 2 new and unique developments in the epidemiology of this disease. Archibald et al. described a gradual increasing trend between 1987 and 2001 in ICU rates from hospitals with >500 beds and in hospitalwide rates from hospitals with <250 beds (12
). We found a sharp increase in hospitalwide rates during 2001–2003 following steady rates from 1996 to 2000; this increase was similar in medium and large hospitals. In contrast to early results from NNIS during the spread of MRSA (23
), when rates were highest in hospitals with the greatest number of beds, we found that overall rates of CDAD were similar in medium and large hospitals.
Several possible explanations may account for the increasing national rates of CDAD. One includes potentially new and evolving patterns of antimicrobial drug use, for example, use of the fluoroquinolones that have recently been implicated in outbreaks of CDAD (9,24,25
). Another potential contributing factor is the promotion of alcohol-based, waterless, hand sanitizers as the primary means of hand hygiene over soap and water. Because alcohol is not sporicidal, alcohol-based, waterless hand sanitizers may not be as effective as soap and water in removing C
; this factor has led to the recommendation that "during outbreaks of CDAD, washing hands with a nonantimicrobial [agent] or antimicrobial soap and water after removing gloves is prudent" (26
One important possibility is the emergence of strains of C
that are more fit and capable of causing transmission and disease. This emergence would not be unprecedented; in the early 1990s a strain of C
that was clindamycin resistant, the so-called "J strain," caused outbreaks in at least 5 geographically diverse hospitals (5,27
). Indeed, a recent report suggests that an emerging fluoroquinolone-resistant, epidemic strain of C
has been responsible for hospital outbreaks in at least 6 US states (Georgia, Illinois, New Jersey, Maine, Oregon, Pennsylvania) since 2001 (28
). This epidemic strain has continued to spread among additional US states (Connecticut, Florida, Massachusetts, Ohio, Texas), Canada, and Europe (29
). Another recent report suggests this strain produces 16- and 23-fold more toxins A and B, respectively, than current nonepidemic strains (30
). Depending on where the earliest outbreaks caused by this epidemic strain were reported, the higher overall rates in the northeastern United States may reflect the early spread of this strain (9,10,28,29
). If this hypothesis proves correct, it suggests other regions of the United States are likely to observe continued increases as the strain continues to spread geographically.
One striking finding we report is the marked variation in CDAD rates among different age groups, with rates in persons >65 years of age several fold higher than rates in the next younger age group (45–64 years). Although the importance of advanced age as a risk for CDAD is not a new idea (1,3
), this is the first report of national CDAD rates according to age group. Several possible reasons may explain this association between CDAD and age, not the least of which is increased exposure to healthcare facilities (including both acute and long-term facilities) and antimicrobial drugs. In addition, older persons may have decreased host defenses to protect them from CDAD. These conditions include decreased stomach acidity resulting from achlorhydria or a possibly increased use of medications such as histamine-2 receptor blockers or proton-pump inhibitors, medications that are becoming increasingly recognized in association with CDAD (31,32
). Recent evidence also suggests the importance of a humoral immune response in protecting against CDAD after colonization (33–37
). Thus, the decreased immune responsiveness commonly observed in older groups may be important in the development of CDAD in patients >65 years of age.
Contact precautions are recommended to prevent transmission of C
in the healthcare setting (38
). These consist of placing patients with CDAD in private rooms or cohorting CDAD patients together, using gloves and gowns for all patient contact, and either using disposable patient care equipment or cleaning such equipment between use with different patients. In addition, removing certain potential fomites, such as reusable electronic thermometers, from use in the general hospital patient population is important for controlling outbreaks (1
). Limited data support enhanced environmental cleaning, especially of heavily contaminated patient care equipment. Clinicians should be aware of the importance of adhering to these precautions for containing transmission of CDAD in healthcare facilities. However, because antimicrobial drug use is the single most important patient risk factor, the clinician's primary responsibility in the control of CDAD lies in the area of judicious antimicrobial drug use (1
In conclusion, the overall scope and magnitude of CDAD are great and may exceed those of other important hospital pathogens (e.g., MRSA), which suggests that C. difficile is one of the most common nosocomial pathogens. In addition, the financial costs and patient illness caused by CDAD in US short-stay hospitals appear substantial. Patients >65 years of age and those in intermediate- or larger-sized hospitals appear disproportionately affected. Because rates appear to have markedly increased during the first 3 years of this decade, new initiatives in the areas of surveillance, prevention, and control of CDAD are urgently needed. In the meantime, clinicians should be aware of the risk posed by CDAD in their hospitalized patients, remain cognizant of the importance of judicious antimicrobial drug use, and support infection control efforts for CDAD in the healthcare settings where they practice.