Accountability in Research

The growing interest in leveraging artificial intelligence (AI) tools for healthcare decision-making extends to improving antibiotic prescribing. Large language models (LLMs), a type of AI trained on extensive datasets from diverse sources, can process and generate contextually relevant text. While their potential to enhance patient outcomes is significant, implementing LLM-based support for antibiotic prescribing is complex. Here, we specifically expand the discussion on this crucial topic by introducing three interconnected perspectives: (1) the distinctive commonalities, but also the crucial conceptual differences, between the use of LLMs as assistants in scientific writing and in supporting antibiotic prescribing in real-world practice; (2) the possibility and nuances of the expertise paradox; and (3) the peculiarities of the risk of error when considering LLMs to support complex tasks such as antibiotic prescribing.

Among people with HIV (PWH), real-world data on the long-term renal function evolution while receiving tenofovir alafenamide (TAF)-containing antiretroviral therapy (ART) remain scarce. PWH who initiated or were switched to TAF-containing ART and controls who received non-tenofovir-containing ART were included for follow-up. We retrospectively collected demographic, clinical, and laboratory data, including estimated glomerular filtration rate (eGFR), urine β-2 microglobulin, and urine protein-to-creatinine ratio (UPCR). The association between the duration of ART exposure and change of eGFR was compared in locally estimated scatterplot smoothing (LOESS) regression. Factors associated with an excess decline of eGFR (defined as a decline > 2.5 ml/min/1.73 m2 per year; or > 25% throughout the observation) among TAF-receiving PWH were also evaluated. Overall, 2422 PWH receiving TAF-containing regimens and 252 controls were included, with the median follow-up duration being 4.8 and 5.4 years, respectively. In the LOESS regression, the predicted change of eGFR at weeks 240 was − 8.0 (95% CI, − 9.1 to − 6.8) ml/min/1.73 m2 for TAF group, compared to − 11.1 ml/min/1.73 m2 (95% CI, − 15.4 to − 6.7) for non-TAF group. In the TAF group, 183 (7.6%) experienced an excessive renal function decline. Furthermore, the levels of urine β-2 microglobulin and UPCR remained stable throughout the observation. A higher plasma HIV RNA level, old age, presence of clinically significant proteinuria throughout observation, and having a higher eGFR at baseline were associated with an excessive decline of eGFR among TAF-receiving PWH. Our study suggests that long-term exposure to TAF-containing ART was not associated with augmented eGFR declines among PWH.

Pneumococcal conjugate vaccines (PCVs) were first introduced in the paediatric United Kingdom (UK) immunisation programme in 2006 which led to significant declines in invasive pneumococcal disease (IPD) caused by targeted serotypes. Although paediatric PCVs provide some indirect protection to adults, a significant IPD burden remains in older adults. Here, we compared three adult (65+ years-old) and risk group (2–64-year-old) vaccination scenarios, namely a continuation of the status quo with PPSV23 vaccination, using the recently licensed-in-adults PCV20, or using the new adult-focused 21-valent PCV, V116. A population-level compartmental dynamic transmission model (DTM) was adapted to the UK setting. The model described Streptococcus pneumoniae carriage transmission dynamics and disease progression in the presence of age- and serotype-specific pneumococcal vaccines. We calibrated the DTM to age- and serotype-specific IPD data in the UK and used the model to make projections under the different adult vaccination scenarios while keeping PCV13 immunisation in children. The calibrated model yielded reasonable parameter values and model fits that closely matched observed IPD dynamics. Among 65+ year-olds, 10-year model projections predicted that the routine use of V116 would reduce IPD incidence by 15.5%, while PCV20 would reduce IPD incidence by 8.9% and the continued use of PPSV23 would increase incidence by 3.83%. There was a notable decrease in IPD incidence in the serotypes unique to V116. In the serotypes included in PCV20 but not V116, the model did not predict a resurgence of IPD. Projections revealed that in adults, V116 led to significantly greater reductions in IPD in the 65+ year-old population compared with PCV20 or PPSV23.

Invasive Escherichia coli disease (IED) incidence has increased over recent years among aging populations and has rising antimicrobial resistance. Here, we report on a comparative, cross-sectional, retrospective analysis of US patients with IED to quantify IED-related healthcare resource utilization (HCRU), costs, and impact on health-related quality of life (HRQoL). This study included Kaiser Permanente Northwest (KPNW) members aged ≥ 60 years enrolled between July 2019 and January 2020. Patients were divided into three groups: Group 1 had experienced a recent IED episode (≤ 3 weeks before enrollment); Group 2 had experienced a former IED episode (13–18 months before enrollment); Group 3 was at risk with no prior history of IED. Data were collected from electronic hospital records, a patient survey, and the EuroQol 5-Dimension 5-Level (EQ-5D-5L) questionnaire. Mean costs were adjusted according to individual follow-up. Patient characteristics were generally consistent across Groups 1 (n = 289), 2 (n = 319), and 3 (n = 340). Inpatient hospitalization was observed in 84%, 44%, and 15% of patients in Groups 1, 2 and 3, respectively. Mean direct costs per patient (per 30-day follow-up) were $17,168, $2530, and $1094 in Groups 1, 2, and 3, respectively. Mean total costs per patient in the year following an IED episode (Group 2) were $35,034 vs. $16,163 in the at-risk Group 3. HRQoL was poor for patients with recent IED, with a mean EQ-5D-5L utility index value of 0.25 on the worst day of illness. During a 12-month follow-up period, rehospitalization rates and mean number of antibiotic prescriptions were ~ threefold higher for patients who recovered from IED vs. those at risk. These data demonstrate substantial short- and long-term impacts of IED on HCRU, IED-related costs, and HRQoL. Additional research is needed to further value the impact of novel IED prevention strategies.