A Perspective on a Quality Management System for AI/ML-Based Clinical Decision Support in Hospital Care.

Although many artificial intelligence (AI) and machine learning (ML) based algorithms are being developed by researchers, only a small fraction has been implemented in clinical-decision support (CDS) systems for clinical care. Healthcare organizations experience significant barriers implementing AI/ML models for diagnostic, prognostic, and monitoring purposes. In this perspective, we delve into the numerous and diverse quality control measures and responsibilities that emerge when moving from AI/ML-model development in a research environment to deployment in clinical care. The Sleep-Well Baby project, a ML-based monitoring system, currently being tested at the neonatal intensive care unit of the University Medical Center Utrecht, serves as a use-case illustrating our personal learning journey in this field. We argue that, in addition to quality assurance measures taken by the manufacturer, user responsibilities should be embedded in a quality management system (QMS) that is focused on life-cycle management of AI/ML-CDS models in a medical routine care environment. Furthermore, we highlight the strong similarities between AI/ML-CDS models and in vitro diagnostic devices and propose to use ISO15189, the quality guideline for medical laboratories, as inspiration when building a QMS for AI/ML-CDS usage in the clinic. We finally envision a future in which healthcare institutions run or have access to a medical AI-lab that provides the necessary expertise and quality assurance for AI/ML-CDS implementation and applies a QMS that mimics the ISO15189 used in medical laboratories.

Practice-Level Association between Antibiotic Prescribing and Resistance: An Observational Study in Primary Care.

A direct relation between antibiotic use and resistance has been shown at country level. We aim to investigate the association between antibiotic prescribing for patients from individual Dutch primary care practices and antibiotic resistance of bacterial isolates from routinely submitted urine samples from their patient populations. Practices' antibiotic prescribing data were obtained from the Julius Network and related to numbers of registered patients. Practices were classified as low-, middle- or high-prescribers and from each group size-matching practices were chosen. Culture and susceptibility data from submitted urine samples were obtained from the microbiology laboratory. Percentages of resistant isolates, and resistant isolates per 1000 registered patients per year (population resistance) were calculated and compared between the groups. The percentages of resistant Escherichia coli varied considerably between individual practices, but the three prescribing groups' means were very similar. However, as the higher-prescribing practices requested more urine cultures per 1000 registered patients, population resistance was markedly higher in the higher-prescribing groups. This study showed that the highly variable resistance percentages for individual practices were unrelated to antibiotic prescribing levels. However, population resistance (resistant strains per practice population) was related to antibiotic prescribing levels, which was shown to coincide with numbers of urine culture requests. Whether more urine culture requests in the higher-prescribing groups were related to treatment failures, more complex patient populations, or to general practitioners' testing behaviour needs further investigation.

Oseltamivir-resistant pandemic A(H1N1) 2009 influenza viruses detected through enhanced surveillance in the Netherlands, 2009-2010.

Enhanced surveillance of infections due to the pandemic A(H1N1) influenza virus, which included monitoring for antiviral resistance, was carried out in the Netherlands from late April 2009 through late May 2010. More than 1100 instances of infection with the pandemic A(H1N1) influenza virus from 2009 and 2010 [A(H1N1) 2009] distributed across this period were analyzed. Of these, 19 cases of oseltamivir-resistant virus harboring the H275Y mutation in the neuraminidase (NA) were detected. The mean 50% inhibitory concentration (IC50) levels for oseltamivir- and zanamivir-susceptible A(H1N1) 2009 viruses were 1.4-fold and 2-fold, respectively, lower than for the seasonal A(H1N1) influenza viruses from 2007/2008; for oseltamivir-resistant A(H1N1) 2009 virus the IC50 was 2.9-fold lower. Eighteen of the 19 patients with oseltamivir-resistant virus showed prolonged shedding of the virus and developed resistance while on oseltamivir therapy. Sixteen of these 18 patients had an immunodeficiency, of whom 11 had a hematologic disorder. The two other patients had another underlying disease. Six of the patients who had an underlying disease died; of these, five had received cytostatic or immunosuppressive therapy. No indications for onward transmission of resistant viruses were found. This study showed that the main association for the emergence of cases of oseltamivir-resistant A(H1N1) 2009 virus was receiving antiviral therapy and having drug-induced immunosuppression or an hematologic disorder. Except for a single case of a resistant virus not linked to oseltamivir therapy, the absence of detection of resistant variants in community specimens and in specimens from contacts of cases with resistant virus suggested that the spread of resistant A(H1N1) 2009 virus was limited. Containment may have been the cumulative result of impaired NA function, successful isolation of the patients, and prophylactic measures to limit exposure.

Modelling of early viral kinetics and pegylated interferon-alpha2b pharmacokinetics in patients with HBeag-positive chronic hepatitis B.

BACKGROUND: Pegylated interferon alpha2b (PEG-IFN-alpha(2b) is effective for the treatment of hepatitis B e antigen (HBeAg)-positive chronic hepatitis B, although its mechanism of action remains unclear. HBeAg loss is achieved in 36% of patients after one year of PEG-IFN-alpha2b treatment and combination therapy with lamivudine is not superior to PEG-IFN-alpha2b monotherapy. METHODS: Early pharmacokinetics and viral kinetics were analysed in patients treated for 52 weeks with PEG-IFN-alpha2b with or without lamivudine. RESULTS: After 4 weeks of treatment, there was a median viral decline of 2.94 log10 copies/ml in those treated with PEG-IFN-alpha2b and lamivudine and only 0.45 log10 copies/ml in the PEG-IFN-alpha2b monotherapy group. Peak PEG-IFN-alpha2b levels were reached approximately one day after administration and subsequently declined exponentially, consistent with a viral load rebound near to baseline levels at the end of the dosing period in most patients receiving PEG-IFN-alpha2b monotherapy. Modelling of pharmacokinetics and viral kinetics data in this group revealed that viral load was minimal 3.6 days after PEG-IFN-alpha2b administration, the mean maximal and mean antiviral effectiveness was 70% and 48% with a mean infected cell loss rate of 0.07 per day, while no significant biphasic decline was observed. CONCLUSIONS: PEG-IFN-alpha2b induces a sustained response in a considerable number of patients despite limited direct antiviral activity during the first weeks of antiviral therapy.

Pseudo-outbreak of hepatitis B virus infection associated with contamination of a semiautomatic cap remover.

A pseudo-outbreak of hepatitis B virus caused by cross-contamination from a semiautomatic cap remover for blood collection tubes is reported. The source of the outbreak was elucidated by using basic epidemiological methods. Laboratories should always be critical about their results in order to identify contamination problems.

Outbreak of infection with a multiresistant Klebsiella pneumoniae strain associated with contaminated roll boards in operating rooms.

An outbreak with a multiresistant Klebsiella pneumoniae (MRKP) strain among seven patients admitted to the adult intensive care unit (ICU) of a regional teaching hospital in The Netherlands was investigated. Epidemiologic investigations revealed a short delay between an operation and the acquisition of the MRKP strain. A case-control study comprising 7 cases and 14 controls was conducted to identify the risk factors associated with the acquisition of the MRKP strain. An operation at each of two operation rooms was strongly associated with the acquisition of the MRKP strain: odds ratio of 36 (95% confidence interval, 2.7 to 481.2; P=0.003, Fisher exact two-tailed test). Cultures of environmental specimens of the operation rooms revealed contamination of the roll boards used to transport patients from the bed to the operation table with the MRKP strains. Molecular genotyping of the isolates revealed clonal similarity between the isolates of the seven cases, isolates from environmental specimen cultures, and in addition, an MRKP isolate from a re-patriated ICU patient from earlier that year. The outbreak ended after cleaning and replacement of the roll boards in the operation rooms and implementation of additional barrier precautions for colonized or infected patients. It was concluded that two operation rooms played a significant role in the transmission of an MRKP strain between ICU patients during the presented outbreak.

High percentage of methicillin-resistant Staphylococcus aureus isolates with reduced susceptibility to glycopeptides in The Netherlands.

While testing the in vitro activities of 14 antimicrobial agents against 107 methicillin-susceptible Staphylococcus aureus (MSSA) and 250 methicillin-resistant S. aureus (MRSA) isolates collected in The Netherlands, we found to our surprise that 19 (7.6%) MRSA isolates were suspected of having reduced susceptibilities to the glycopeptides when the Etest system (AB Biodisk, Solna, Sweden) was used with a large inoculum (no. 2 McFarland standard) and an extended incubation time (48 h) on brain heart infusion agar for MIC testing. Eventually, 15 of these isolates were classified as heterogeneously resistant to glycopeptides (heterogeneously glycopeptide-intermediate S. aureus [hGISA] isolates) according to the population analysis profile-area under the curve analysis. The MICs at which 50 and 90% of isolates are inhibited obtained with the Etest system with the large inoculum were as follows: for MSSA isolates, 3.0 and 4.0 micro g/ml, respectively, for both teicoplanin and vancomycin; for MRSA isolates, 3.0 and 8.0 micro g/ml, respectively, for teicoplanin, and 3.0 and 4.0 micro g/ml, respectively, for vancomycin. This is the first report of hGISA isolates in The Netherlands.

Use of multienzyme multiplex PCR amplified fragment length polymorphism typing in analysis of outbreaks of multiresistant Klebsiella pneumoniae in an intensive care unit.

We developed and optimized a new modified amplified fragment length polymorphism (AFLP) typing method to obtain a multibanding fingerprint that can be separated by agarose gel electrophoresis. Both to maximize the discriminatory power and to facilitate the computer-assisted analysis, bacterial DNA was digested with four different restriction enzymes. After ligation of adaptors to the DNA fragments, PCR testing of various single primers was performed. Two single primers that gave optimal results with regard to band resolution and discriminatory power were selected and combined. The computer-assisted analysis of fingerprint patterns was performed with Pearson's product-moment correlation values of densitometric curves, without assigning bands to peaks. Thus, the analysis is not subject to human interpretation errors. With this method, we investigated two outbreaks of multiresistant Klebsiella pneumoniae in an intensive care unit and various sporadic isolates of K. pneumoniae and Klebsiella oxytoca. Cluster analysis of isolates analyzed in different experiments and on different gels showed that fingerprint patterns clustered correctly according to subspecies or to the outbreaks. Multienzyme multiplex PCR AFLP revealed that the first outbreak was caused by two different types of strains. Outbreak two was caused by yet another strain of K. pneumoniae. In conclusion, the typing method used here is easy to perform and highly reproducible, and due to generation of complex banding patterns, it has a higher discriminatory power. Furthermore, the multienzyme multiplex PCR fingerprints are easy to analyze, and a reliable database can be stored in the computer to facilitate comparison of future isolates of Klebsiella spp. The method can be performed in every clinical microbiology laboratory.