Stool Cultures and Antimicrobial Prescriptions Related to Infectious Diarrhea

Acute diarrheal illness has been estimated to affect >200 million persons annually in the United States, resulting in ∼75 million visits to health care providers [1]. Antimicrobial therapy is contraindicated when treating certain infectious diarrheal illnesses, and published guidelines recommend the use of diagnostic stool cultures when prescribing antimicrobials [2]. Stool culture is also important in public health surveillance to determine the bacterial causes of diarrheal disease, identify outbreaks, and implement appropriate control measures.

The extent to which providers fail to use stool cultures to guide treatment of diarrhea is not well documented, nor is the extent of antimicrobial use. Here, we quantified stool culturing and antimicrobial prescribing in connection with episodes of acute diarrhea among a large Medicaid population and identified factors associated with these practices.

Methods. We conducted a retrospective cohort study of all persons aged <65 years enrolled in TennCare from 1995 through 2004. TennCare is the managed health care program covering Tennessee residents who are eligible for Medicaid. In 2005, ∼1.22 million persons were enrolled, representing 21% of the state's population. Information was obtained from TennCare outpatient visit records and pharmacy data. All outpatient episodes of diarrheal illness were included, along with associated antimicrobial prescriptions and laboratory stool studies. An episode was defined as an outpatient clinic or emergency department visit for diarrhea or a series of visits for diarrhea separated by ⩽30 days. A new episode was defined as a visit for diarrhea with no other visits for diarrhea during the previous 30 days; a single patient could have been responsible for >1 episode per year.

On the basis of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), a visit for diarrhea was defined as any visit with a physician-assigned code for diarrhea (001, 002, 004, 005, 007–009, 003.0, 006.0, 006.1, 006.2, or 787.91) in the first of 9 possible diagnostic fields or a code for dehydration (276.5) or for vomiting or nausea (787.0) in the first field accompanied by a diarrhea code in any subsequent diagnostic field. For each diarrheal episode, we identified prescriptions for antimicrobials that are potentially prescribed for diarrhea within 7 days after a visit; antimicrobials defined as being potentially prescribed for diarrhea included chloramphenicol, ampicillin, amoxicillin, erythromycin, clarithromycin, azithromycin, tetracycline, doxycycline, trimethroprim-sulfamethoxazole, metronidazole, vancomycin, clindamycin, and antihelminthics. All procedure codes indicating a completed order for stool culture within 5 days before and 7 days after any visit during an episode were identified. For all visits, we collected data on patient age, sex, race/ethnicity, residence location (urban, suburban, or rural), month and year of visit, and setting of visit (outpatient or emergency department). Diarrheal episodes with a concurrent diagnosis of acute respiratory tract infection (ICD-9-CM codes 460–466, 480–487, 490, 786.07, 786.1, and 786.2) were identified to determine the associations between respiratory tract infection, antimicrobial prescribing, and stool culturing. For comparison purposes, within the cohort of all enrollees we identified urine cultures and complete blood counts (CBCs) by use of procedure codes.

Multivariate modeling was used to analyze the effects of demographic variables and concurrent diagnosis of acute respiratory tract infection on the prescription of antimicrobials and the probability of stool culture. Generalized estimating equations with a Poisson link and a working independence correlation structure were used to account for the correlation induced by multiple visits for diarrhea per enrollee [3, 4]. Data were analyzed using SAS software (version 9.1; SAS Institute). The project was reviewed and approved by the institutional review boards of the Tennessee Department of Health, Vanderbilt University, and the Centers for Disease Control and Prevention. Approval to use deidentified data was provided by the Bureau of TennCare.

Results. During the 10-year study period, a mean of 1,164,000 persons (range, 1,085,349–1,290,951 persons) aged <65 years were enrolled in TennCare per year. Of the TennCare enrollees, 74% were white and 18% were black, 44% were male, 45% were children aged 0–17 years, and 41% were urban residents. According to the 2000 census, 80% of Tennessee's population were white and 16% were black, 49% were male, 25% were children aged 0–17 years, and 44% were urban residents.

A total of 11.64 million person-years of enrollment data were analyzed. Among these, 315,828 diarrheal episodes were identified (272 episodes/10,000 person-years) (figure 1). The rate was higher for white persons than black persons (316 vs. 185 episodes/10,000 person-years), for females than males (283 vs. 256 episodes/10,000 person-years), and for rural and suburban residents than urban residents (324, 307, and 206 episodes/10,000 person-years, respectively). Most patients (⩾92%) experienced only a single diarrheal episode, and only a single visit was made for 290,379 (91.9%) of the 315,828 diarrheal episodes.

Stool culture was performed for 15,820 (5.0%) of the 315,828 episodes. The proportion of diarrheal episodes associated with a stool culture was highest among children aged <3 months (11.3%) and was lowest among adolescents aged 11–18 years (2.7%) (figure 2A). The proportion of all diarrheal episodes associated with stool culture rose from 2.1% in 1995 to 5.8% in 2004 (figure 2B). By comparison, from 1995 through 2004, the proportion of TennCare enrollees for whom at least 1 CBC was performed increased from 14.3% to 33.7%, and the proportion for whom 1 or more urine cultures were performed increased from 2.6% to 6.4%.

Antimicrobials were prescribed for 32,949 diarrheal episodes (10.4%), increasing from 8.1% in 1995 to 11.2% in 2004. Of 9448 diarrheal episodes among children aged <3 months, 126 (1.3%) resulted in antimicrobial prescription, whereas 24,567 (15.6%) of 124,867 diarrheal episodes among adults aged ⩾19 years resulted in antimicrobial prescription.

Of the diarrheal episodes for which antimicrobials were prescribed, 3504 (10.6%) had a stool culture performed (figure 1), increasing from 5.6% in 1995 to 11.5% in 2004. Thirty-five (28%) of 126 children aged <3 months receiving antimicrobials had a stool culture performed, compared with 376 (7.6%) of 4943 persons aged 19–29 years.

Of all diarrheal episodes, 17,191 (5.4%) had a concurrent diagnosis of acute respiratory tract infection. Of these, antimicrobials were prescribed for 3193 (18.6%). Antimicrobials were prescribed for 29,755 episodes (10.0%) without a concurrent diagnosis of acute respiratory tract infection. Stool culture was performed for only 601 diarrheal episodes (3.5%) with a concurrent respiratory tract infection.

In a multiple regression model, white race (adjusted relative risk [ARR], 1.32 [95% confidence interval {CI}, 1.26–1.38]), urban residence (ARR, 1.65 [95% CI, 1.59–1.72]), antimicrobial prescription (ARR, 3.08 [95% CI, 2.95–3.21]), and not receiving a diagnosis of a concurrent acute respiratory tract infection (ARR, 1.75 [85% CI, 1.61–1.92]) were associated with higher rates of stool culture. In a separate model, white race (ARR, 1.08 [95% CI, 1.04–1.12]), female sex (ARR, 1.08 [95% CI, 1.05–1.11]), rural or suburban residence (ARR, 1.04 [95% CI, 1.02–1.08]), having a stool culture performed (ARR, 3.08 [95% CI, 2.95–3.21]), and receiving a diagnosis of a concurrent acute respiratory tract infection (ARR, 2.52 [95% CI, 2.41–2.63]) were associated with higher rates of antimicrobial prescription.

Discussion. In this large outpatient study, stool culture was rarely used for patients with diarrhea. Antimicrobials were more commonly prescribed and were accompanied by a stool culture only a tenth of the time. Guidelines from physician groups and public health agencies promote the use of diagnostic stool cultures when prescribing antimicrobials for diarrhea [2]. The limited proportion of antimicrobial prescriptions accompanied by stool culture and the fact that only a small minority of stool cultures eventually identify a pathogen suggest that few antimicrobial prescriptions for diarrheal disease are guided by a microbiologic diagnosis [5]. Antimicrobials are frequently used inappropriately for the treatment of acute enteric infections. Inappropriate antimicrobial use can lead to adverse outcomes, increased costs, and antimicrobial resistance [6].

The rate of diarrhea found in the present study was somewhat lower than that reported in older studies (e.g., Roy et al. [7]), although this is unsurprising given that our study relied on the documentation of clinical visits rather than on self-report; in addition, there are well-recognized limitations to comparing results among diarrheal prevalence surveys [7].

Health care provider and patient surveys have demonstrated infrequent stool culturing, even when antimicrobials are prescribed. A North Carolina survey reported that ⩾80% of 284 physicians did not request stool sampling for 75%–100% of patients with acute diarrheal illness [8]. In a multistate population-based telephone survey, 5.1% of respondents reported having had an acute diarrheal illness during the previous month [9]. Of those, only 3.7% had a stool culture ordered, and 7.8% were treated with antimicrobials. Although that large study relied on self-report, the results are similar to our findings. In another study based on the chart review of pediatric emergency and urgent care visits in one hospital, 24% of visits for acute diarrheal illness resulted in antimicrobial prescriptions [10]. Our study, using information from actual clinical visits, adds validity to these various survey results.

Stool culture has been perceived as being a diagnostic tool that causes expense without often yielding useful results. In one large hospital, only 2.4% of stool cultures identified a pathogen, and in <5% of the cases did the culture results lead to a change in therapy [11]. Among outpatients, positive culture results have been reported in 6%–15% of cultures [5, 12]. Nevertheless, diagnostic stool cultures can be important to both clinical and public health practice. Certain diarrheal illnesses respond well to properly selected antimicrobial treatment [2]. In addition, a microbial diagnosis can be the critical first step in detecting foodborne or waterborne outbreaks. Diarrheal disease outbreaks for which stool specimens were collected were 10 times more likely to have the outbreak source identified than were those for which specimens were not collected [13].

Ordering of stool cultures increased modestly during the 10-year study period. Concurrently, the proportion of TennCare enrollees for whom a CBC or urine culture was performed doubled, indicating a general trend toward increased diagnostic testing not specific for diarrheal disease. The observed increases in rates of testing and antimicrobial use are not unique to this study, and the reasons are not well understood but could include increasing patient expectations, malpractice concerns, and reimbursement issues.

In the present study, higher rates of stool culture were associated with white race and urban residence, both of which are typically associated with greater access to health care. White persons were also more likely to be prescribed antimicrobials for diarrhea, consistent with the results of studies of treatment for respiratory tract infections [14]. Compared with young children, adolescents and adults had lower rates of stool culture and higher rates of antimicrobial prescription.

A small proportion of patients with diarrhea in our study (<6%) received a diagnosis of a concurrent acute respiratory tract infection for which drugs on our list of antimicrobials could have been prescribed. A concurrent diagnosis of a respiratory tract infection was associated with a lower rate of stool culture and a higher rate of antimicrobial prescription. Diarrheal and respiratory tract illnesses frequently have a common viral etiology. This may have led providers to a presumptive diagnosis of a viral syndrome and to view stool culture as being less useful, although the concomitant increase in antimicrobial use is a worrisome contradiction to this assumption.

Twenty-one percent of the Tennessee population is enrolled in TennCare, and the demographics of the enrolled population closely mirror those of the state's population; however, a disproportionate number of enrollees are children, and some results might not be generalizable to other populations. Because the investigation was conducted using an administrative database, we were unable to assess disease severity or microbiological results related to provider decisions in the treatment of patients with diarrhea.

In this large, multiyear outpatient study, stool culture was rarely used to guide antimicrobial therapy among the patients visiting a health care provider for diarrhea. Antimicrobials were prescribed for >10% of diarrheal episodes. The potentially imprudent use of antimicrobials and a lack of information from diagnostic stool cultures have implications for both clinical and public health practice. Optimization of stool culture for public health surveillance, as a diagnostic tool, and as a guide to antimicrobial therapy warrants further attention.

• Received September 26, 2007.
• Accepted December 31, 2007.

• © 2008 by the Infectious Diseases Society of America