The study was approved by the Medical Ethics Committee of the Radboud University Medical Center. No consent was obtained (data was analyzed anonymously).

The study was conducted from December 2017 till July 2018 at the Radboud University Medical Centre, a tertiary care, academic medical center in the Netherlands. Stool samples submitted for analyses of GI infection with routine PCR were parallel tested on the BioFire FilmArray. Data regarding infection control and patient management of included patients were recorded retrospectively. Only patients who were seen at our medical center and of whom fecal samples for routine diagnostics had been sent in, were eligible for inclusion. Patients where only testing for C. difficile toxin (CDT) was requested were not included, but patients where CDT was requested in combination with a different panel, were included.

No specific criteria for requesting a certain panel on the routine PCR were used. The physicians selected one or a combination of panels based on his or her assessment of the clinical symptoms together with the history of the patient. The available panels are: a (I) bacterial panel, with CDT on demand in addition, a (II) basic parasitic panel, a (III) parasitic panel for travelers, a (IV) parasitic panel for immunocompromised patients and a (V) viral panel (Table 1).

In case a parasitic panel was requested, a small quantity of the feces was transferred into a bead bottle containing beads and 1 mL phosphate-buffered saline (PBS). The sample was beated twice for 20 seconds. In case a bacterial or viral panel was requested, a small quantity was transferred into a tube containing 2,5 mL PBS. This was vortexed for 1 minute, let to rest for 10 minutes at room temperature and then vortexed again for another 1 minute. This was then centrifuged at 3000 rpm for 2 minutes. The suspension was pipetted into another tube. Samples for CDT were tested without any processing beforehand. All samples were either tested the same day, or frozen at -80°C until the test was run. A 195 μL sample for the routine PCR was used. All routine PCR’s were run on the Roche Aurora Flow in a monoplex or multiplex PCR assay with primers and probes as previously described.[10–18] The bacterial and parasitic panels were batch wise run once daily five days per week, the viral panel was run thrice weekly (Monday, Wednesday and Friday). When a test run was inhibited twice, the test was reported “not interpretable” and the clinician was asked to send in a new stool sample.

CDT testing was primarily done using the Techlab C.DIFF QUIK CHEK COMPLETE^® (Alere). Only in the case when Clostridium antigen was positive and the toxin test was negative, or vice versa, the sample was tested with the Cepheid GeneXpert^® XVI Xpert C. difficile for confirmation.

The BioFire FilmArray GI Panel detects thirteen bacteria, five viruses and four parasites (Table 1).[9] The assay was performed according to the manufacturer’s instructions. Test results of the FilmArray were not reported to the physician during the study.

Data collected from electronic patient charts included age, sex, comorbidity, length of hospital admission, use of antibiotics, additional diagnostics performed related to GI symptoms, isolation days, time of test ordering, arrival at the laboratory, initiation of the test, completion of the test and time of reporting to the physician. We used the time of test ordering as a proxy for the time the feces were collected because the exact moment of collection was unknown. Nosocomial diarrhea was defined as the onset of gastrointestinal symptoms 48 hours after admission to the hospital. We used a TAT of 1.5 hours for the FilmArray for calculations of the total TAT’s. Action taken by the clinician (e.g. patient removed from isolation) within 24 hours after a test result was defined as an action related to the test result.

For the statistical analyses we used IBM SPSS Statistics 23. All data are reported as mean ± 1 standard deviation (SD) or as median and range. Missing data were treated as missing. Statistical significance was assessed using the Wilcoxon test, the McNemar’s test, Fisher’s exact test or the Chi Square test where appropriate. A probability (p) value of ≤0.05 was considered statistically significant.

A total of 182 patients were included. All stool samples were tested simultaneously by routine PCR and on the FilmArray GI panel. Median age of the patients was 45 (0–97) years. 120 (65.9%) patients were hospitalized for a median of 8.5 days (1–206). Of the hospitalized patients, 98 (82.4%) were put in isolation with a median of 4 (1–57) days in isolation per patient (Table 2).

The most requested (combination) of panels were the basic parasitic panel (n = 24), viral panel (n = 42), bacterial panel combined with CDT (n = 17) and viral panel combined with CDT (n = 17) (S1 Fig). In 56% of all the patients more than one panel was requested by the clinician. Fig 1 shows the additional pathogens detected by the FilmArray when compared to the most frequently requested combination of panels. In each instance the FilmArray detected more pathogens. No inhibition was found using the FilmArray.

Routine PCR detected at least one pathogen in 52 (28.6%) patients, compared to 72 (39.6%) pathogens detected by the FilmArray (p = 0.001). Routine PCR detected in 48/52 (92.3%) patients one pathogen and in 4/52 (7.7%) patients two pathogens. The FilmArray detected in 50/72 (69.4%) patients one pathogen, in 19/72 (26.4%) patients two pathogens and in 3/72 (4.2%) patients three or more pathogens. FilmArray detected 29 pathogens that were either negative, not tested or “not detected because of inhibition” with the routine PCR. This included eight CDT, seven Noroviruses, ten EPEC and three Campylobacters.

The FilmArray detected 26 pathogens (mainly CDT and Norovirus) there were on a panel included in the routine PCR, but not ordered by the clinician (Table 3).

The FilmArray detected 19 pathogens that were not included in any of the routine PCR panels. Positive pathogens included 12 EPEC and six EAEC (Table 4).

The routine PCR detected 12 pathogens (mainly D. fragilis) that were not included in the FilmArray (Table 5).

The FilmArray detected nine pathogens that were on an ordered panel in the routine PCR but not detected by the routine PCR. This included six pathogens were the routine PCR was negative and three pathogens were the result in the routine PCR was “not interpretable” because of inhibition. The most common pathogens not detected or inhibited in the routine PCR, but positive in the FilmArray was Norovirus (n = 4, three times not interpretable in routine PCR). On the other hand, only one pathogen (CDT) was positive in the routine PCR and negative in the FilmArray. (Fig 2 and S2 Fig).

Fig 3 show the turnaround times. Time from collection to reporting was 53 hours (median) using routine PCR compared to 16 hours (median) for the FilmArray. Runtime for the routine PCR was 4 hours (median). Using the FilmArray testing was 29 hours (median) faster compared to routine PCR (Fig 3).

120 of the 182 patients (65.9%) were admitted to the hospital. The ordered routine PCR was negative in 86/120 patients (71.7%). Of the 86 admitted patients with a negative PCR, 66 (75.6%) were isolated. 31 of the patients put in isolation but with a negative PCR, were removed from isolation within 24 hours after the test result was reported to the clinician (Fig 4). In 26 of these 31 patients (83.9%) the FilmArray was also negative. In five patients the FilmArray found one or more pathogens (two Norovirus, one Campylobacter, one EAEC, one EPEC and one Cyclosporidium) and therefore should have been kept in isolation.

In 34 (28.3%) of the admitted patients the ordered routine PCR was positive. Of these patients, five were not yet isolated when the test result was reported to the clinician. Three patients were put in isolation within 24 hours after the test result became available and all three patients were also positive in the FilmArray (two rotavirus, two Campylobacters, one CDT). In two of the 34 admitted patients with a positive routine PCR the FilmArray tested negative, however the pathogens detected in the routine PCR were not included in the FilmArray (Entero-/Parechovirus and Adenovirus non-40/41). In addition, five patients with a negative routine PCR but with a positive FilmArray had undergone additional diagnostic procedures (two ultrasounds of the abdomen, one gastroscopy and three colonoscopies). The FilmArray detected one Sapovirus, one Campylobacter, one Plesiomonas, one C. difficile and three EPEC in these five patients.

If the FilmArray would have been implemented for routine molecular testing, additional diagnostic procedures and isolation days could have been prevented. In this study 26 patients were removed from isolation when the routine PCR result was reported negative to the clinician. If the FilmArray would have been used as a first line diagnostic system, for each of these patients a median of 29 hours would have been gained, thus saving 31.4 (range 1–187) days in isolation. Antibiotic therapy could have been stopped in three patients, this would have saved 3.6 (range 0.1–21.6) days on antibiotics. Five patients had undergone additional diagnostic procedures that could have been prevented. Five patients were wrongly removed from isolation due to a negative routine PCR.

In the outpatient population, we found no impact on either antibiotic therapy or the number of prevented diagnostic procedures (data not shown).