Anders Krifors

Objectives: To assess the performance of a reasoning large language model (LLM) in identifying medication errors in medical incident reports. Materials and Methods: OpenAI’s O4-mini LLM was adapted using prompt engineering on 75 000 anonymized incident reports from the Västmanland region of Sweden (2019-2024). To guide the prompt design, we used a subset of 2434 reports, which were manually reclassified by pharmacists as medication-related or not. For validation, 200 reports (January 2024-March 2024) were independently classified by 2 pharmacists to establish a reference classification. Moreover, the LLM performed binary classification, with concordance rates measured against the expert consensus. Results: The LLM achieved a concordance rate of 96.0% (192/200; 95% CI, 92.3-98.3) with expert classification. Eight cases (4.0%) showed disagreements, primarily due to linguistic ambiguity or context-dependent interpretation. Five cases involved pharmacists classifying reports as non-medication-related, while the LLM classified them as medication-related, with the reverse in 3 cases. Subcategorization accuracy was 76.5%. Discussion: The LLM showed expert-level performance, outperforming existing automated methods. Thus, its integration into incident reporting systems might improve the efficiency, accuracy, and consistency of patient safety monitoring. Conclusion: This validated AI-driven method can be integrated directly into clinical informatics workflows, enabling healthcare organizations to rapidly and consistently identify medication errors, ultimately enhancing patient safety outcomes.

Dalbavancin is a long-acting lipoglycopeptide approved for acute bacterial skin and soft-tissue infections. Its prolonged half-life enables outpatient treatment, reducing the burden of hospitalisation. Despite increasing off-label use for complex Gram-positive infections, real-world effectiveness data remain limited. This study aimed to evaluate the clinical effectiveness and safety of dalbavancin in a real-world tertiary care setting in Sweden. We retrospectively analysed the medical records of all patients (n = 66) who received dalbavancin in Region Västmanland, Sweden, from 2019 to 2023. Patient characteristics, source of infection, identified pathogens, treatment regimens, and outcomes were extracted from medical records. The primary outcome was clinical cure at 6 months; secondary outcomes included mortality, need for suppressive therapy, and adverse events. Sixty-six patients (median age 73 years; 47% female) received dalbavancin for orthopaedic/bone infections (56%), endocarditis (23%), vascular graft infections (6%), bacteraemia (6%), sacral ulcer infections (6%), and other infections (3%). The patients had significant comorbidities: diabetes (38%), malignancy (33%), chronic kidney disease (44%), and substance use disorders (17%)."Methicillin-susceptible" Staphylococcus aureus was the predominant pathogen (31% of isolates). Dalbavancin was prescribed to facilitate outpatient therapy (53%), address poor compliance (17%), or manage antibiotic intolerances (17%). Patients received a median of 2 doses (range 1-17). At 6-month follow-up, 62% achieved clinical cure, 18% remained on suppressive therapy, and 20% died, primarily from underlying conditions. Adverse events were infrequent (6%) and generally mild. Dalbavancin achieved a 62% cure rate despite significant comorbidities, offering a safe alternative to inpatient care for elderly, comorbid patients.

The purpose of this study was to prospectively investigate the incidence of influenza-associated pulmonary aspergillosis (IAPA) in influenza patients admitted to intensive care units in Sweden. The study included consecutive adult patients with PCR-verified influenza A or B in 12 Swedish intensive care units (ICUs) over four influenza seasons (2019-2023). Patients were screened using serum galactomannan and β-d-glucan tests and fungal culture of a respiratory sample at inclusion and weekly during the ICU stay. Bronchoalveolar lavage was performed if clinically feasible. IAPA was classified according to recently proposed case definitions. The cohort included 55 patients; 42% were female, and the median age was 59 (IQR 48-71) years. All patients had at least one galactomannan test, β-d-glucan test and respiratory culture performed. Bronchoalveolar lavage was performed in 24 (44%) of the patients. Five (9%, 95% CI 3.8% - 20.4%) patients were classified as probable IAPA, of which four lacked classical risk factors. The overall ICU mortality was significantly higher among IAPA patients than non-IAPA patients (60% vs 8%, p = 0.01). The study represents the first prospective investigation of IAPA incidence. The 9% incidence of IAPA confirms the increased risk of invasive pulmonary aspergillosis among influenza patients admitted to the ICU. Therefore, it appears reasonable to implement a screening protocol for the early diagnosis and treatment of IAPA in influenza patients receiving intensive care. ClinicalTrials.gov NCT04172610, registered November 21, 2019.

Purpose The purpose of this study was to prospectively investigate the incidence of influenza-associated pulmonary aspergillosis (IAPA) in influenza patients admitted to intensive care units in Sweden.Methods The study included consecutive adult patients with PCR-verified influenza A or B in 12 Swedish intensive care units (ICUs) over four influenza seasons. Patients were screened using serum galactomannan and β-d-glucan tests and fungal culture of a respiratory sample at inclusion and weekly during the ICU stay. Bronchoalveolar lavage was performed if clinically feasible. IAPA was classified according to recently proposed case definitions.Results The cohort included 55 patients; 42% were female, and the median age was 59 (IQR 48–71) years. All patients had at least one galactomannan test, β-d-glucan test and respiratory culture performed. Bronchoalveolar lavage was performed in 24 (44%) of the patients. Five (9%, 95% CI 3.8% − 20.4%) patients were classified as probable IAPA, of which four lacked classical risk factors. The overall ICU mortality was significantly higher among IAPA patients than non-IAPA patients (60% vs 8%, p = 0.01).Conclusion The study represents the first prospective investigation of IAPA incidence. The 9% incidence of IAPA confirms the increased risk of invasive pulmonary aspergillosis among influenza patients admitted to the ICU. Therefore, it appears reasonable to implement a screening protocol for the early diagnosis and treatment of IAPA in influenza patients receiving intensive care.Trial registration ClinicalTrials.gov NCT04172610, registered November 21, 2019

Detection of SARS-CoV-2 RNA in serum, viremia, has been linked to disease severity and outcome. The kinetics of viremia in patients receiving remdesivir has not been thoroughly studied and could help predict treatment response and outcome. We investigated the kinetics of SARS-CoV-2 viremia and factors associated with baseline viremia, viral clearance and 30-day mortality in patients receiving remdesivir. An observational study including 378 hospitalised patients (median age 67 years, 67% male) sampled with serum SARS-CoV-2 RT-PCR within ± 24 h of initiation of remdesivir treatment. Baseline viremia was present in 206 (54%) patients with a median Ct value of 35.3 (IQR = 33.3–37.1). In patients with baseline viremia, the estimated probability of viral clearance was 72% by day 5. Ct values decreased significantly during remdesivir treatment for viremic patients, indicating an increase in viral load. In total, 44 patients (12%) died within 30 days, and mortality was significantly associated with viremia at baseline (OR = 2.45, p = 0.01) and lack of viral clearance by day 5 (OR = 4.8, p = < 0.01). Viral clearance was not associated with any individual risk factor. Viremia appears to be a prognostic marker before and during remedesivir treatment. The resolution of viremia was similar to patients not receiving remdesivir in other studies, and the decrease in Ct values during treatment questions the antiviral capacity of remdesivir in vivo. Prospective studies are warranted to confirm our findings.

Background: Invasive candidiasis (IC) is a severe and often fatal fungal infection that affects critically ill patients. The development of animal models that mimic human disease is essential for advancing our understanding of IC pathophysiology and testing experimental or novel treatments. We aimed to develop a large animal model of IC that could provide a much-needed addition to the widely used murine models. Results: A total of 25 pigs (including one control), aged between 9 and 12 weeks, with a median weight of 25.1 kg (IQR 24.1–26.2), were used to develop the porcine IC model. We present the setup, the results of the experiments, and the justification for the changes made to the model. The experiments were conducted in an intensive care setting, using clinically relevant anaesthesia, monitoring and interventions. The final model used corticosteroids, repeated Candida inoculation, and continuous endotoxin. The model consistently demonstrated quantifiable growth of Candida in blood and organs. The registered physiological data supported the development of the sepsis-induced circulatory distress observed in IC patients in the ICU. Conclusions: Our proposed porcine model of IC offers a potential new tool in the research of IC.

Vaccination against SARS-CoV-2 reduces the risk of hospitalisation and death, but vaccine-induced IgG antibodies against the spike protein (IgG S) decline over time. Less is known about the nature ...

The T2Bacteria panel is a direct-from-blood assay that delivers rapid results, targeting E. coli, S. aureus, K. pneumoniae, A. baumanii, P. aeruginosa, and E. faecium (ESKAPE pathogens). In this study, T2Bacteria and T2Candida (targeting C. albicans/C. tropicalis, C. glabrata/C. krusei, and C. parapsilosis) were evaluated in parallel with blood cultures in 101 consecutive surgical patients with suspected intra-abdominal infection admitted to the intensive care unit or high dependency unit. Fifteen patients had bacteremia, with T2Bacteria correctly identifying all on-panel (n = 8) pathogens. T2Bacteria was positive in 19 additional patients, 11 of whom had supportive cultures from other normally sterile sites (newly inserted drains, perioperative cultures or blood cultures) within seven days. Six of these eleven patients (55%) received broad-spectrum antibiotics at the sampling time. T2Candida identified the two cases of blood-culture-positive candidemia and was positive in seven additional patients, three of whom were confirmed to have intra-abdominal candidiasis. Of four patients with concurrent T2Bacteria and T2Candida positivity, only one patient had positive blood cultures (candidemia), while three out of four patients had supporting microbiological evidence of a mixed infection. T2Bacteria and T2Candida were fast and accurate in diagnosing on-panel bloodstream infections, and T2Bacteria was able to detect culture-negative intra-abdominal infections.

The T2Candida magnetic resonance assay is a direct-from-blood pathogen detection assay that delivers a result within 3–5 h, targeting the most clinically relevant Candida species. Between February 2019 and March 2021, the study included consecutive patients aged >18 years admitted to an intensive care unit or surgical high-dependency unit due to gastrointestinal surgery or necrotizing pancreatitis and from whom diagnostic blood cultures were obtained. Blood samples were tested in parallel with T2Candida and 1,3-β-D-glucan. Of 134 evaluable patients, 13 (10%) were classified as having proven intraabdominal candidiasis (IAC) according to the EORTC/MSG criteria. Two of the thirteen patients (15%) had concurrent candidemia. The sensitivity, specificity, positive predictive value, and negative predictive value, respectively, were 46%, 97%, 61%, and 94% for T2Candida and 85%, 83%, 36%, and 98% for 1,3-β-D-glucan. All positive T2Candida results were consistent with the culture results at the species level, except for one case of dual infection. The performance of T2Candida was comparable with that of 1,3-β-D-glucan for candidemic IAC but had a lower sensitivity for non-candidemic IAC (36% vs. 82%). In conclusion, T2Candida may be a valuable complement to 1,3-β-D-glucan in the clinical management of high-risk surgical patients because of its rapid results and ease of use.

Background: Health-care workers are at risk of contracting and transmitting SARS-CoV-2. The aim of this study was to investigate the prevalence of SARS-CoV-2 IgG antibodies and the rate of seroconversion in an environment with high exposure to SARS-CoV-2. Methods: 131 health-care workers at the Department of Infectious Diseases in Västerås, Sweden, were included in the study. Abbott’s SARS-COV-2 IgG immunoassay was used with a signal cut-off ratio of ≥1.4. Every third week from the beginning of May, blood samples were drawn, and the participants completed a questionnaire regarding symptoms consistent with COVID-19 and the result of any SARS-CoV-2 PCR performed since the last sampling occasion. Participants with IgG antibodies against SARS-CoV-2 were re-sampled only on the sixth and last occasion. Results: At the start of the study, 18 (15%) participants had SARS-CoV-2 IgG antibodies. At the end, 25 (19%) of 131 participants were seropositive. One case of asymptomatic infection was detected, and two cases with PCR-confirmed COVID-19 did not develop IgG antibodies. Conclusion: The low rate of seroconversion during the study suggests that it is possible to prevent transmission of SARS-COV-2 in a high-exposure environment. Compliance with adequate infection control guidelines is the likely explanation of our findings.