FilmArray® Meningoencephalitis panel in the diagnosis of central nervous system infections: stewardship and cost analysis in a paediatric hospital in Chile

Central nervous system (CNS) infections have been an ongoing concern in paediatrics due to the burden of disease, the risk of severe neurological sequelae, and the high mortality rate associated mainly with bacterial aetiologies. For this reason, the challenge over decades has been to improve the aetiological diagnosis of CNS infections, to shorten the window of opportunity during which this is performed, and to optimize the clinical management and outcomes of patients.

A few decades ago, bacterial aetiologies were the most common, especially H. influenzae and S. pneumoniae [1, 2]. However, since the introduction of the H. influenzae b and pneumococcal conjugate vaccine, the epidemiology of meningitis has changed [3]. Currently, most cases of meningitis are caused by viruses (aseptic meningitis) [4‐6]. Nigrovic et al., in a multicentre retrospective cohort (n = 3295) of patients between 1 month and 18 years of age, found that 96.3% of cases were aseptic meningitis (culture or latex negative for bacteria), while only 3.7% of cases were caused by a bacterial agent [7]. Among the most common bacterial pathogens in infants under 3 months of age are Streptococcus agalactiae and gram-negative rods, while in older infants, bacterial meningitis is caused mainly by S. pneumoniae of non-vaccine serotypes and N. meningitidis [4, 8]. In Chile, the vaccine for H. influenzae type b was incorporated into the National Vaccine Program in 1996, that for S. pneumoniae was incorporated in 2010, and that for N. meningitidis serotypes A, C, W, and Y was incorporated in 2012; together these measures sought to reduce the incidence of acute bacterial meningitis (ABM) due to these pathogens.

In 2011, in Roberto del Río Children’s Hospital, 73 cerebrospinal fluid (CSF) samples with altered cytochemical characteristics were classified as meningitis or meningoencephalitis; of these, 28.7% were suggestive of a bacterial pattern and the rest were consistent with aseptic meningitis (unpublished data).

Molecular diagnostic techniques have progressed rapidly in recent years, leading to modern techniques of diagnosing CNS infectious disease [9]. At present, the diagnosis of viral CNS infections is made using polymerase chain reaction (PCR) techniques, resulting in the aetiological confirmation of infection source in up to 45% of cases [10]. FilmArray® Meningoencephalitis panel (FAME) technology analyses CSF and is based on the specific amplification of certain regions of the pathogen genome using multiple nested PCRs. It presents a good correlation with conventional PCR, in addition to having a greater sensitivity in the detection of co-infections and a greater likelihood of identifying the aetiology of the infection [11‐13]. This technology may also impact the cost effectiveness of treatment interventions due to the better stewardship of resources secondary to accurate aetiological identification of infectious agents [14]. Faced with this, we posed the following question: Does the use of FAME improve the aetiological identification of infectious agents and reduce hospitalization costs among patients with suspected CNS infection treated at the Roberto Del Río Children’s Hospital? The main objective of the study was to compare the aetiology of the infection source among patients with suspected CNS infection with their associated hospitalization costs both before and after the clinical use of FAME.

During the study period, 607 CSF samples were obtained from children under 15 years of age at the Roberto del Río Children’s Hospital. A total of 198 were excluded, and 409 CSF samples were finally analysed for CNS infection (Fig. 1). Of these samples, 297 were collected during the pre-intervention period and 112 were collected during the post-intervention period. Of these, 39.7 and 46.4%, respectively, were female. Median of age of the subjects studied was 1.6 (IQR: 0.8-12.7) and 2.4 (IQR: 1.0-28.7) months for the retrospective and prospective cohorts, respectively. Age was stratified into 4 groups: < 6 months, 6-23 months, 24-71 months, and > 72 months (Table 1).

Of the included patients, 85.5% required hospitalization in the pre-intervention period, compared with 92.7% in the post-intervention period (p < 0.05). Of these, hospitalization in the ICU was more frequent in the post-intervention period (14.8% vs. 28.6% (p < 0.001)). However, median of ICU bed-days used was significantly lower in the post-intervention period (2 days) than in the pre-intervention period (3.5 days). ICU admission criteria were based on the Paediatric Index of Mortality (PIM), which is an indicator of paediatric severity validated for children under 16 years of age, which is measured when patients are admitted to hospital critical units and whose score ranges from 0 (no risk of mortality) to 100 (predicts certain mortality) [21]. In both groups, the median PIM score was 1.2 (IQR: 0.6-2.8), with no statistically significant differences between the two groups (p = 0.266, for the Mann-Whitney test, data not shown). This complements the results of the comparison of ICU bed days between both periods, making them comparable in terms of severity parameters.

There was no significant difference in the use of antibiotics upon admission to the hospital, but there was a difference in the use of antivirals, with a greater use recorded in the post-intervention period (p < 0.001) (Table 1). This is probably related to an increase in the aetiological identification of viral infectious agents, but establishing this association was not one of the objectives of this study.

The complementary diagnosis in patients who were hospitalized was significantly higher in the post-intervention period. However, the most frequent discharge diagnoses in the post-intervention period were CNS infections, seizures and epilepsy (all related to the CNS), unlike the diagnoses from the pre-intervention period, which were most frequently fever without an identified source and infection; both were statistically significant (Table 2).

The use of FAME significantly increases the aetiological identification of CNS infection sources [14, 22]. Even among patients with normal CSF, clinical symptoms suggestive of CNS infection were observed. Especially among patients younger than 6 months, the differences in aetiological identification of infectious agents was greater than those among older patients. Of all normal CSF collected from inpatients less than 6 months and those older than 6 months of age during the FAME period, almost 18% of infection aetiologies were identified, probably due to the ability of the FAME to recognize some viral agents, such as parechovirus, that can cause an infection without altering the CSF [23, 24]. The main aetiology of CNS infections was viral, as described in study results by S.E. Park [25]. However, the overall positivity rate in this study was 46.4%, almost double that observed in our study (24.1%), which may be explained by the difference in clinical criteria between medical teams performing lumbar punctures in children.

The increase in positive identification of the aetiology in CNS infections was the result only of the introduction of the new molecular tool, since neither the Gram stain, CSF culture nor blood cultures showed a significant difference between either period. Additionally, it is important to consider the difference in turnaround time. The use of FAME was available 24 h a day during business and non-business hours, and the processing of samples took only 65 min. Routine real-time PCR is processed only during business hours, and the response time is up to 48 h. FAME makes it possible to make timely clinical decisions [22]. During the second period evaluated, there was an increase in more complex exams.

The increase observed in imaging tests, considered an uncontrolled variable, may have been affected by events such as access to magnetic resonance imaging and tomography, as well as turnover in the team of neurology specialists who order these exams. The discharge diagnoses and aetiologies identified during the second period increased for neurologic disorders and decreased for febrile syndrome without focus. This could be due to a better diagnosis in CNS pathologies.

Goodlet et al. [26] showed 10 studies that measured the impact of FAMEs and antimicrobial use. However, there are few studies analysing the patient care costs with suspected CNS infection [27‐29]. In our study, the use of FAME was associated with a significant decrease in the occupancy rate of ICU bed-days, which are the costliest beds in public hospitals and so translate into direct cost savings. This phenomenon was also observed in the study by S. Duff [27], where an overall savings per patient of USD $3481 was estimated when FAME was performed on all CSF samples obtained from patients with suspected CNS infection. This is similar to the cost savings found in our study—only we calculated this cost using the ICU bed-day concept, considering the value of public hospitals The difference between both studies lies mainly in methodology, since Duff’s study was estimated according to a mathematical model and ours was the result of the analysis of real data from the comparison of 2 cohorts, without and with the use of FAME. However, our study was conducted with a child population and thus differs from Duff’s study, which was performed using an adult population. Indirect costs should also be considered in the decreased use of ICU beds-days: decreases in children‘s exposure to the risks of being hospitalized, such as health care associated infections or adverse events, as well as biological, psychological and social factors for children and their families, also do not represent valorised savings, nor does the decrease in productivity of parents and guardians, this could represent a limitation and undervaluation of costs savings of our study. Finally, the alternative use of beds improves and is relevant to the productivity of any health institution.

Another limitation of this study is due to the retrospective design of the evaluation of the pre-intervention period, which was constructed from databases of medical care records that had already been collected. As described above, the microbiological data registry was not initially constructed with the aim of comparing etiological identification and hospitalization costs among patients with suspected CNS infection, and it is likely that not all relevant factors have been identified and recorded as antimicrobial use. Furthermore, an additional disadvantage is that many different healthcare professionals will have been involved in the patient’s care, so the measurement of variables across the entire database would likely be less accurate and consistent than that achieved in the post-intervention period that was prospective. However, the use of records that had already been collected and stored in an electronic database meant that this study was relatively inexpensive, quick and easy to perform.

Finally, molecular laboratory diagnostic tools should not be seen only as an expense or an investment of the cost centre that uses them; their value should be analysed as part of the overall cost of the institution, taking into consideration the entire care process. In times of high demand for beds in our country, the public health system buys bed-days from the private system; a diagnostic tool like FAME enables the optimisation of resources in health care.

FAME use significantly increases the aetiological identification of infectious agents in children with suspected CNS infections, in both those with or without altered cytochemical CSF. This increase is greater among children under 6 months of age. FAME use enables more accurate aetiological identification and timely diagnosis, and it results in better opportunities for adequate treatment to be provided and for better clinical outcomes to be achieved [30]. The cost/benefits ratio between FAME test value and ICU bed-day cost is favourable to the use of FAME because the investment in the protocoled implementation of this test effectively implies a significant cost savings for the institution, at least in terms of ICU bed-days as direct cost. It is important to emphasize that there are indirect savings associated with patient safety, as well, including the reduction of psychosocial risks for the child and a reduced impact on the economic productivity of parents or guardians. Additionally, the alternative use of beds thus enabled is relevant to the management of the health care institution. We recommend the use of this diagnostic tool for the study of CSF samples from the age group of under 6 months or from those requiring admission to intensive care. Implementation of FAME in Chilean public hospitals in the diagnosis of CNS infection saves public resources and improves the opportunity for effective treatment based on accurate aetiological diagnoses of infectious agents.