Clinical Impact of a Diagnostic Gastrointestinal Panel in Children

We performed a cross-sectional cohort study of children ≤18 years of age who had stool testing performed from 2013 to 2017 and analyzed in the Children’s Hospital Colorado microbiology laboratory. This included patient encounters in the inpatient, emergency department and urgent care (ED/UC), and ambulatory settings associated with a large, freestanding, quaternary-care, urban, academic children’s hospital and 4 affiliated community-based children’s hospitals.

Patient encounters with stool testing were identified from the electronic medical record (EMR) (Epic, Verona, WI). Patient demographics, clinical characteristics, diagnostic workup, and stool tests were obtained from the EMR. For hospitalized children with a bacterial and/or parasitic stool pathogen, we determined the use of antimicrobial agents via chart review. We obtained hospital charges from the finance department. To adjust for patient volume changes over the study period, we calculated the number of patient encounters for all inpatient, ambulatory, and ED/UC sites that sent stool testing over the study period. This study was approved by the Colorado Multiple Institutional Review Board.

The explanatory variable was the type of stool testing available. The GIP was introduced in October 2015, and the GIP era included the 24 months thereafter (November 1, 2015–October 31, 2017). Outcomes were compared between the GIP era and historic controls in the conventional diagnostic era, which included the 24 months previous (October 1, 2013–September 30, 2015). A 1-month period after introducing the GIP was excluded to allow for uptake of new testing.

Both eras tested for bacteria, parasites, and viruses but had different testing modalities, with varying sensitivities and pathogens that could be identified (Table 1). In the conventional diagnostic era, testing included Giardia and Cryptosporidium direct fluorescent antigen (Meridian Bioscience, Cincinnati, OH), ova and parasites, Clostridiodes difficile PCR (Xpert C difficile; Cepheid, Sunnyvale, CA), bacterial culture, and viral electron microscopy (Table 1). In the GIP era, GIP replaced the need for most other testing. The FilmArray GIP (BioFire Diagnostics) detects 22 enteric pathogens, including bacteria, viruses, and parasites (Table 1). To provide antibiotic susceptibilities, stool cultures were additionally performed when the GIP was positive for Shigella or Salmonella. Other stool tests available in the GIP era included ova and parasites and C difficile PCR, which were offered in both eras.

Testing for gastrointestinal pathogens was part of routine clinical care; there were no official guidelines for testing. The laboratory did not perform tests on formed stools, and there were no testing policy changes during the study period. In our analysis, we only included stool results that were released in the EMR and available to providers. Detection of enteropathogenic, enteroaggregative, and enterotoxigenic Escherichia coli on the GIP were not released unless specifically requested, given the unknown clinical significance of these pathogens in children in the United States.^17–19  Similarly, C difficile results from GIPs were not released for children <1 year of age because of concern for asymptomatic colonization.²⁰ 

The primary outcomes were LOS and resource use. These outcomes were evaluated in patients hospitalized with community-acquired diarrhea, defined as patients with stool testing collected within the first 72 hours of hospitalization and LOS of <14 days. By evaluating this subset of patients, we focused on patients presenting with diarrheal symptoms and excluded those with prolonged admissions for reasons other than gastroenteritis. Resource use included the mean number of ancillary laboratory tests performed, percent of patients with at least 1 abdominal imaging study, and total hospital charges. Ancillary laboratory testing included tests for electrolytes, inflammatory markers, complete blood counts, and bacterial blood culture. Ancillary imaging included abdominal studies (Supplemental Table 6). To focus on the impact of stool results on ancillary testing, we only included testing performed 1 to 7 days after stool collection. To account for inflation, hospital charges were adjusted to 2018 prices, with a 6% per year increase per unit charge.

Secondary outcomes included stool test use, pathogen detection, time to result, and time to treatment. All secondary outcomes, other than time to treatment, were evaluated in the overall population, including tests performed in ambulatory and ED/UC settings. Stool test use was defined as the number of patients who received at least 1 stool test and the total number of stool tests performed as raw numbers and proportions over patient volume per era. Pathogen detection included the number of positive results and percent positivity, defined as the percent of tests positive for a given pathogen out of the total number of tests performed that were capable of detecting that pathogen. Time to result was defined as time from stool collection to the release of results in the EMR. Time to treatment was defined as the time from stool collection to administration of the first antimicrobial agent that was given after stool collection and had an order indication (required for all antibiotics) for “acute gastroenteritis/sepsis.”²¹  We evaluated this outcome among children without complex chronic conditions (CCCs) who were hospitalized with community-acquired diarrhea and had test results positive for a bacterial and/or parasitic stool pathogen.²²  We did not compare time to treatment between the eras for tests positive for C difficile because the testing modality for C difficile did not change between eras and impact of the GIP on C difficile has been previously described.²³  Encounters were organized into 3 hospital settings on the basis of the highest level of care that patients required: inpatient, ED/UC, or ambulatory.

For the descriptive analyses, categorical variables were displayed as frequencies and percentages, and continuous variables were displayed as medians and interquartile ranges (IQRs). We used bivariate analyses to compare outcome variables between the 2 eras using Pearson’s χ² test (for categorical variables) or the Wilcoxon rank test (for medians). LOS and resource use outcomes were evaluated in hospitalized patients with community-acquired diarrhea as well as non-CCC subgroups identified a priori on the basis of pathogen type and those who received treatment with antimicrobial agents. We did not adjust for seasonal variability because both eras spanned the same calendar months. All analyses were performed by using SAS version 9.4 (SAS Institute, Inc, Cary, NC), and P values of <.05 were considered statistically significant.

There were 12 222 stool tests performed in 8720 patient encounters among 6733 unique patients. A total of 40% of patients were in the ambulatory setting, and 60% did not have a CCC (Table 2). There were similar patient characteristics between the 2 eras. In the GIP compared with conventional diagnostics era, the proportion of stool tests performed per patient volume decreased by 23% (total stool tests: 5402 from 6820), but the proportion of encounters that had stool testing per patient volume increased by 21% (total encounters with stool testing: 4830 from 3890; P values <.001). A higher percentage of encounters had only 1 stool test in the GIP, compared with the conventional diagnostic era (92% vs 54%; P < .001).

In the GIP era, there was more testing performed capable of detecting bacteria, viruses, and parasites, and more positive results for each pathogen type (Table 3). There was a higher percent positivity overall in the GIP era (40% vs 11%; P < .001), with an increased percent of positivity for viruses but decrease for bacteria. The most common pathogens detected in each era included C difficile, Salmonella, Campylobacter, Shiga toxin–producing E coli, and G lamblia in the conventional diagnostic era and C difficile, norovirus, sapovirus, rotavirus, and adenovirus in the GIP era (Table 3). In the GIP era, a higher percentage of patients had >1 organism detected (8.6% vs 0.4%; P < .001). For patients with a positive GIP result, most had a virus (54%), with a lower frequency of C difficile (38%), other bacteria (18%), and parasites (5%) detected.

In the GIP era, compared with the conventional diagnostic era, the median time to result decreased by over 24 hours (4 vs 31 hours; P < .001). Across both eras, there were 58 children hospitalized with community-acquired diarrhea who had enteric bacteria (excluding C difficile) and/or parasites and received antimicrobial agents. This represented 2.9% of the tested patients hospitalized with community-acquired diarrhea, with a higher percentage in the GIP, compared with the conventional diagnostic era (3.7% vs 1.9%; P = .02). For these patients, the time to treatment decreased by 25 hours in the GIP era (11 vs 36 hours; P < .001).

LOS and resource use were evaluated among 1986 children hospitalized with community-acquired diarrhea (885 in the conventional diagnostic era and 1101 in the GIP era) and in non-CCC subgroups on the basis of stool test results and antimicrobial administration (numbers per group are displayed in Fig 1). For hospitalized patients who received antimicrobial agents for a bacterial and/or parasitic pathogen, LOS decreased by 2 days (3.1 vs 5.1; P < .001; Table 4). However, there was no difference in LOS for all patients, regardless of stool test results (Table 4).

There was no difference in ancillary laboratory tests or imaging studies between eras (Table 5). For patients with test results positive for a bacterial and/or parasitic pathogen who received antimicrobial agents, there was a significant decrease in median hospital charges in the GIP era by nearly $20 000 (P < .001). There was also a decrease in charges for children without CCCs by ∼$3000 (P = .04), but there were no differences in charges for all patients (Table 5).

In a large pediatric study evaluating the impact of the GIP, we found more patients who underwent stool testing but a reduction in the number of individual stool tests performed, increased pathogen detection, and more rapid results. In patients with test results positive for bacterial or parasitic enteric organisms, faster results led to quicker initiation of antimicrobial agents, and those who received therapy had a decreased LOS and lower hospital charges. However, for the majority of patients, the GIP had no impact on LOS, ancillary testing, or charges. This highlights the critical need for diagnostic stewardship to better define and implement testing criteria to optimize the value of multiplex stool testing.

We identified several benefits of multiplex PCR stool testing, compared with conventional diagnostics. First, it was convenient for providers to order and the microbiology laboratory to perform 1 test per patient, instead of multiple. Second, GIP testing led to improved pathogen detection, with over 3 times the number of positive results compared with conventional diagnostics. This increase is likely due to improved sensitivity, increased testing, and improved case-finding with multiple pathogens on the panel, thereby eliminating the need for selective ordering.^11,14  Although the percent of positivity for bacterial pathogens decreased with the GIP, compared with bacterial culture in the conventional diagnostic era, this is likely because of reduced pretest probability as other studies have demonstrated high sensitivity of the GIP.¹¹  Third, GIP results were available to clinicians in under 5 hours, which was more than 24 hours earlier than with conventional diagnostics. These improvements in pathogen detection and result time are similar to findings in adult studies.^13,14,16  Our GIP result time was faster than the 9 to 41 hours reported in the literature, likely because our laboratory runs GIPs 24 hours per day and 7 days per week.^13,14,16  The rapid result time maximizes the potential for clinically relevant results to optimally influence patient care.¹² 

Although increased pathogen detection and faster result times are beneficial attributes of a diagnostic test, these benefits should ideally translate to improved patient outcomes or a reduction in health care use. We found that, in a small subset of patients hospitalized with community-acquired diarrhea who had test results positive for bacterial or parasitic pathogens and received antimicrobial therapy, use of the GIP led to quicker initiation of therapy (more than 24 hours earlier), decreased LOS (by 2 days), and lower hospital charges (nearly $20 000 per patient). Similarly, in adult studies, researchers have found that the GIP facilitated faster time to treatment.¹⁴  We hypothesize that faster results allowed clinicians to start treatment sooner, which resulted in more rapid symptom resolution. However, these improved outcomes were only found in 3% of tested patients because of the low frequency of patients with bacterial and parasitic gastroenteritis who received treatment.⁷  Although there was a statistically significant decrease in charges for children without CCCs in the GIP era, likely related to a decrease in LOS, this is relatively small and likely not financially significant.

For the overall population, the GIP’s improved detection and rapid turnaround time did not impact outcomes. The only other pediatric study to evaluate the impact of GIP testing on LOS did so in a small cohort of 61 patients and found a similar LOS, also with no difference based on testing type.¹³  We hypothesize that this is because discharge readiness and LOS are closely tied to symptom improvement. Given that a majority (60%) of GIP results were negative, suggesting a noninfectious etiology or untested pathogen, and most of the positive results were viral, management with supportive care was likely not modified in response to these results. Knowing that the patient had an infectious cause of their symptoms did not appear to facilitate earlier discharge, supported by the findings of another study.⁸  Additionally, providers may have limited experience with some viruses detected on the GIP (eg, sapovirus) and thus may be unsure what this additional information means for patients and their management. The use of ancillary testing was not different between the 2 cohorts, likely because ancillary testing is relatively uncommon and dictated by continued symptoms. Finally, we found no financially significant differences in charges. Although, in some adult studies, researchers note decreased abdominal studies and cost saving with GIPs, these findings are likely not generalizable to children, given the self-limited nature of most pediatric diarrhea, relatively low occurrence of imaging studies, and fact that pediatric infection control practices are more commonly driven by symptoms in addition to testing results.^13,15,16,24 

Our findings highlight the critical need for diagnostic stewardship to optimize the value of the GIP. We found high test use, particularly in children not sick enough to warrant hospitalization and in those without medical complexity, in which the risk for treatable bacterial or parasitic infections is less common.²⁵  This may be because of simplified clinician ordering, readily available testing, GIP novelty, and the perception that faster turnaround time would be more clinically beneficial. Given the high cost of GIPs and small percentage of tested patients with improved outcomes, further studies are needed to better predict those who are likely to benefit from stool testing. This will enable hospitals to develop testing guidelines to inform providers and optimize the value of the GIP. Implementation studies to understand the optimal diagnostic stewardship approach and ways to balance these interventions with physician autonomy are also needed.

These findings are similar to studies on other emerging rapid diagnostics, such as the meningitis encephalitis and respiratory panels, highlighting a systemic challenge with multiplex diagnostics.^12,26  Although rapid diagnostics are revolutionizing the way we test for and detect infectious pathogens, there are significant practical challenges that exist.¹²  As new tests are introduced into clinical practice, it is critical that diagnostic stewardship is rolled out in parallel to combat the natural tendency to overuse and overestimate their benefits.¹²  Given the added costs of these rapid diagnostics, it is also important to evaluate the cost-effectiveness and clinical impact of novel diagnostics to inform testing strategies and optimize high-value care.^12,27,28 

There are several limitations of this study. First, our findings are limited to 4 hospitals within a single health care system, which may limit generalizability. However, by including quaternary-care teaching hospital and satellite community hospitals, our cohort included both complex and previously healthy children. Second, we used historical controls and therefore there may have been institutional changes between the eras that unknowingly influenced our outcomes.

In this large pediatric study, we demonstrated that the introduction of the GIP led to increased pathogen detection and faster results. This translated into improved outcomes for only a small minority of patients, suggesting that unrestricted use of the GIP leads to low-value care. Similar to other novel rapid diagnostic panels, there is a critical need for diagnostic stewardship to optimize the value of the GIP.

We thank Ben Brown for his support collecting financial information and adjusting for inflation. We also thank Jennifer Sadlowski for her help collecting information on patients with CCCs.

CCC

complex chronic condition

ED/UC

emergency department and urgent care

EMR

electronic medical record

GIP

gastrointestinal panel

IQR

interquartile range

LOS

length of stay

PCR

polymerase chain reaction