An observational study was conducted to describe the prevalence of probiotic use in the inpatient setting. The study was divided into 2 parts: a cross-sectional study of prevalence of probiotic use in 2012 and a longitudinal study of probiotic use among the subset of hospitals reporting yearly from 2006-2012, inclusive. Because the data were deidentified at the patient and hospital levels, this work was determined not to involve human subjects and therefore was exempt from the regulations governing the protection of human subjects in research under 45 CFR 46.101(b)(5).¹² This work was conducted under the provisions of the Centers for Disease Control and Prevention–MarketScan Data Use Agreement.

The Truven Health MarketScan Hospital Drug Database (HDD) from the years 2006-2012, inclusive, was used to estimate probiotic use in the inpatient setting. The HDD is a relational database developed from hospital charge detail master data, containing all charges accumulated during the hospitalization, including room and board, supplies, procedures, laboratory testing, and pharmacy products. The drug data are derived from free-form text fields, which are then mapped to a drug classification system by a clinical coder. Codes of interest are obtained through text string searches of the generic drug name in the description field of the corresponding drug reference table. The database also includes standard administrative elements, such as patient demographics, hospitalization diagnosis and procedure codes, and facility characteristics.

To facilitate an informal comparison with a nationally representative sample, the Healthcare Utilization Project’s National Inpatient Sample (NIS) estimates from 2012 were compared with study sample estimates whenever possible (Supplemental Tables S1 and S2).

Data were restricted to those of hospitals reporting directly to Truven Health. Within these data, the study population consisted of all discharges, unless otherwise noted. Individual patients may have been present multiple times in the data as a result of multiple hospitalizations. Prior to database release, any discharges identified as having critical errors were removed. Critical errors include patient age <0 or >124 years, missing or invalid primary diagnosis or procedure codes, and diagnoses or procedures not corresponding to age or sex of the patient.

To identify probiotic use, text strings were searched in the HDD reference tables consisting of terms at the genus, species, and strain level and terms indicative of probiotics that were identified from several sources.^13-16 These terms included the following: probiotic, Lactobacillus, Bifidobacterium, Saccharomyces, Streptococcus, rhamnosus, plantarum, acidophilus, casei, johnsonii, boulardii, helveticus, bulgaricus, infantis, and reuteri (see Supplemental Table S3 for a longer list). In addition to these root terms, different spelling permutations were searched using the Perl Regular Expression function PRXMATCH (SAS Institute, Cary, NC).¹⁷ Identified codes and corresponding descriptions were reviewed by hand. Codes with terms or phrases in the descriptions inconsistent with a probiotic were removed (examples are shown in Supplemental Table S4). To ensure identification of all possible codes specific to probiotics, sections of the reference tables were also hand checked. Ultimately, 8 unique generic probiotic formulations consisting of ≥1 species were identified in the HDD drug reference tables (see Supplemental Table S5 for additional details). Dose was not considered in the identification process because relevant information (eg, number of colony forming units per dose) was not available.

The process for identifying antimicrobial use was conducted in a similar manner to that for probiotics. A previously developed list of term¹⁸ was used, which included antibacterial, antifungal, antiviral, and antiparasitic agents. Route of administration was restricted to inhalation, oral, and parenteral.

Prevalence of probiotic use was defined as the number of patients receiving a probiotic during hospitalization divided by the total number of patients discharged in 2012. Distributions of patient-, facility-, and hospitalization-level characteristics were tallied by probiotic group. For categorical variables, the denominator consisted of the number of discharges unless otherwise noted. For continuous variables, the mean, 95% confidence interval (CI), and median were presented. Unadjusted comparisons of patient- and hospital-level characteristics between patients with and without probiotic use were conducted using the independent samples t test for continuous variables and χ² test for categorical variables.

To describe trends in the prevalence of probiotic use over time, annual prevalence of probiotic use from 2006-2012 was calculated among the subset of hospitals reporting data during each of these years. The need to adjust for facility-level effects was confirmed using the covtest option in SAS’s PROC GLIMMIX (SAS Institute). The final model was adjusted for within-facility residual correlations using PROC GENMOD with a first-order autoregressive correlation structure and assuming a gamma distribution. Robust SEMs were used to safeguard against misspecification of correlation structure. The annual and overall change in prevalence was estimated using this model.

In 2012, 145 hospitals with 1,976,167 pediatric and adult hospital discharges submitted data directly to Truven Health. Probiotics were used in 51,723 (2.6%) of the hospitalizations and in 139 (96%) hospitals. At the hospital level, probiotic use ranged from 0.3%-8.5% (5th-95th percentile, 139 hospitals) of discharges.

Patients with any probiotic use were older (median age, 70 vs 57 years; P < .0001), had a longer mean length of stay (8.8 vs 4.4 days, P < .0001), and incurred higher charges ($63,732 vs $34,130, P < .0001) than patients without probiotic use. The probiotic group was also 21 times more likely to have a discharge diagnosis of C difficile infection (95% CI, 20.4-21.7; P < .0001), almost 9 times as likely to have used antimicrobials (odds ratio [OR], 8.6; 95% CI, 8.3-8.8; P < .0001), more likely to be admitted from another inpatient health care facility (OR, 1.4; 95% CI, 1.4-1.5; P < .0001), and more likely to be transferred to another health care facility at discharge (OR, 2.9; 95% CI, 2.8-2.9; P < .0001) than the nonprobiotic group (Table 1).

A total of 54,242 probiotic courses were used; 95% of patients in the probiotic group received just 1 probiotic formulation during their hospitalization, and the remaining 5% received 2-3 formulations. Each formulation contained between 1 and 4 organisms identified at the species level. As detailed in Table 2, the most commonly used probiotic formulations were S boulardii (32% of discharges in the probiotic group), L acidophilus and L bulgaricus (30%), L acidophilus (28%), and Lactobacillus rhamnosus (11%), not mutually exclusive.

Infectious, inflammatory, and gastrointestinal conditions comprised the bulk of the most common primary diagnostic categories in the probiotic group (Table 3). Obstetric and cardiovascular conditions were more common in both the nonprobiotic group (Table 3), the overall HDD sample, and the NIS (Supplemental Table S6).

Sixty hospitals submitted annually from 2006-2012 and therefore were included in the trend analysis. In 2006, 1.0% of discharges received at least 1 course of probiotics. Usage increased annually 1.2-fold (95% CI, 1.1-1.3; P < .0001) through 2012, in which 2.9% of discharges received at least 1 course of probiotics for an overall 2.9-fold (95% CI, 1.8-4.5; P < .0001) increase between 2006 and 2012 (Fig 1, Supplemental Table S7).

In this sample of 145 U.S. hospitals, a sizable number of patients received probiotics as part of their care. Although the proportion of probiotic use among discharges was small, the magnitude was considerable because almost 52,000 hospitalizations included probiotic use as part of prescribed medical care in 2012. Projected nationally, probiotics may have been utilized in the care of approximately 850,000 patients discharged from US hospitals (based on weights provided by Truven Health; additional details in Supplemental Appendix). In addition, the consistent increase in probiotic use from year to year between 2006 and 2012 suggests probiotic use is continuing to increase. The most common categories of diagnosis codes among the patients with probiotic use were associated with infectious diseases, antibiotic use, increased severity of illness, or exposure to health care.

These findings are similar to survey results on the increasing use of probiotic use among nonhospitalized persons. A recent study of probiotic use among a sample of noninstitutionalized adults in the United States also showed an increase in probiotic use over time. Recorded probiotic use in the prior 30 days grew from 0.4% (projected national estimate of 865,000) of adults in 2007 to 1.6% (3,857,000) in 2012.¹⁹ In the same time span, use among children was 0.3% (number not estimated) in 2007 and 0.5% (294,000) in 2012.²⁰

We identified 7 probiotic formulations in use across 96% of the 145 hospitals, with 71% of hospitals using 1-2 formulations (mean, 2.1; range, 1-6) per hospital in 2012. This finding is comparable with the aforementioned survey of probiotic use among U.S. academic medical centers in which 10 products were identified in stock or in use across 87% of the 114 responding centers, with 79% of the centers reported stocking or using only 1 product.¹⁰ At the species level, the top 4 probiotic products most frequently reported across the centers¹⁰ were also similar to those used in the current study, consisting of L rhamnosus, Lactobacillus gasseri and L bulgaricus, Lactobacillus acidophilus, and S boulardii.

A limitation of this study is the utilization of a nonprobability (convenience) sample of hospitals, which increases the potential for selection bias,²¹ The Northeast region was under-represented at 0.3% of discharges in the HDD sample compared with 19.1% in the Healthcare Utilization Project’s NIS, a nationally representative sample of hospital discharge data. Despite this regional difference, more clinically relevant measures, including top hospital discharge diagnoses, average length of stay, and discharge disposition, were similar to national estimates (Supplemental Table S6). Therefore, our findings may reflect an estimate of U.S. inpatient hospital probiotic use, although this extrapolation must be interpreted with caution. A second limitation is that this study relied on administrative data collected primarily for billing purposes. Administrative data lack clinical detail beyond what is needed for the purposes of tracking costs and charges in particular. Third, although the probiotic formulations were identified at the species level, strain identity (the unit at which probiotic activity is determined), product potency (eg, number of colony forming units at the time of expiration), and product source (eg, name of brand or manufacturer) were generally absent. These details would have been helpful in verifying the product identities. Fourth, because estimation of probiotic use was based on drug charge data, the prevalence may be underestimated. Unaccounted for sources of probiotics include foods containing probiotics (eg, yogurt) provided by the hospital and dietary supplements and foods containing probiotics brought to the hospital by the patient or patient’s family.

Manipulation or remediation of the microbiome may be an important strategy for prevention of health care–associated infections in the future. Whether probiotics are effective in preserving or restoring a healthy microbiome remains unknown, but the high prevalence of probiotic use among hospitalized patients may indicate a growing belief among clinicians that these agents may be an effective strategy for doing so. Currently, there is not enough evidence to support use of currently available probiotics in this clinical setting. Correspondingly, several professional societies directly state that they do not yet recommend probiotics for the prevention of CDI because of the need for further research in the areas of efficacy and safety.^22-25

Recently, the Agency for Health Research and Quality sponsored the development of an evidence report to compile safety assessments of probiotic agents containing specific microorganisms used in research studies for risk reduction, prevention, and treatment of disease. As a result of issues, including lack of safety assessments, inconsistent reporting, and poor documentation of interventions, one of the major conclusions of the report was that “the current literature is not well equipped to answer questions on the safety of probiotic interventions with confidence.”¹³ Adverse events among probiotic users that have been reported in case reports and clinical trials include bacteremia, fungemia, endocarditis, functional ileus, bowel distension, bowel ischemia, diarrhea, and increased risk of death.^26,27 Risk may be increased among patients with central venous catheterization, bacterial translocation, immunosuppression, critical illness, pancreatitis, and organ transplants.²⁶ Patients with these conditions are likely to be prescribed antibiotics, which places them at increased risk for microbiome disruption. Because the patients most likely to benefit are also most at risk for an adverse event, preclinical research focused on the selection of likely probiotics and carefully designed clinical trials with systematic assessment of safety is particularly important.

In the planning process of trials of probiotic efficacy and safety, issues to be mindful of include product labeling inaccuracies at the species and even genus level for reasons such as difficulty in identification, desire for consumer recognition,²⁸ and cross-contamination during manufacturing²⁹; varying susceptibility of strains to antibiotics³⁰; and product viability at the time of ingestion. The general questions needing to be answered include the following: which strain-specific organisms, which patient populations, at what doses, and in what time frames (related to antibiotic use in particular) are both safe and effective in the prevention or treatment of which diseases?