Evaluation of a prototype machine learning tool to semi-automate data extraction for systematic literature reviews.

BACKGROUND: Evidence-based medicine requires synthesis of research through rigorous and time-intensive systematic literature reviews (SLRs), with significant resource expenditure for data extraction from scientific publications. Machine learning may enable the timely completion of SLRs and reduce errors by automating data identification and extraction. METHODS: We evaluated the use of machine learning to extract data from publications related to SLRs in oncology (SLR 1) and Fabry disease (SLR 2). SLR 1 predominantly contained interventional studies and SLR 2 observational studies. Predefined key terms and data were manually annotated to train and test bidirectional encoder representations from transformers (BERT) and bidirectional long-short-term memory machine learning models. Using human annotation as a reference, we assessed the ability of the models to identify biomedical terms of interest (entities) and their relations. We also pretrained BERT on a corpus of 100,000 open access clinical publications and/or enhanced context-dependent entity classification with a conditional random field (CRF) model. Performance was measured using the F1 score, a metric that combines precision and recall. We defined successful matches as partial overlap of entities of the same type. RESULTS: For entity recognition, the pretrained BERT+CRF model had the best performance, with an F1 score of 73% in SLR 1 and 70% in SLR 2. Entity types identified with the highest accuracy were metrics for progression-free survival (SLR 1, F1 score 88%) or for patient age (SLR 2, F1 score 82%). Treatment arm dosage was identified less successfully (F1 scores 60% [SLR 1] and 49% [SLR 2]). The best-performing model for relation extraction, pretrained BERT relation classification, exhibited F1 scores higher than 90% in cases with at least 80 relation examples for a pair of related entity types. CONCLUSIONS: The performance of BERT is enhanced by pretraining with biomedical literature and by combining with a CRF model. With refinement, machine learning may assist with manual data extraction for SLRs.

Less improvement in knee function and higher rates of dissatisfaction in the short-term following total knee arthroplasty in people with mild radiographic arthritis.

INTRODUCTION: The purpose of this study was to assess if severity of radiographic changes of knee arthritis was associated with patient improvement after total knee arthroplasty (TKA). We hypothesised that patients with mild arthritis were more likely to report lower satisfaction, improvement in knee function and Oxford knee score (OKS) compared to patients with moderate or severe arthritis. MATERIALS AND METHODS: Secondary analysis of prospectively collected data from TKA patients of two arthroplasty centres with knee radiographs available for assessment of disease severity. Patients completed the Oxford knee score (OKS) and were asked to rate the global improvement in knee condition and their satisfaction at 6 months post-TKA. Bivariable analysis and multivariable regression models were used to test the association between disease severity and each outcome. RESULTS: 2226 patients underwent primary TKA and 3.6% had mild arthritis. Mean OKS improved from 17.0 (SD 18.0) to 38.0 (SD 8.1) 6 months after TKA. Two hundred and twenty-two patients (10%) reported 'Poor' or 'Fair' satisfaction, and 173 (8%) reported knee function was 'Much worse', 'A little worse' or 'About the same' 6 months post-TKA. Patients with mild arthritis showed improvement in OKS [mean improvement in OKS = 19 (SD 15)], but were significantly more likely to report dissatisfaction (OR = 3.10, 95% CI 1.62 to 5.91, p = 0.006), lack of improvement (OR = 4.49, 95% CI 2.38 to 8.47, p < 0.001) and lower OKS scores (- 3 points, 95% CI - 5.39 to - 0.85, p = 0.008) compared to patients with moderate to severe arthritis. CONCLUSIONS: While patients with mild radiographic arthritic changes improve after TKA, they were significantly more likely to report higher rates of dissatisfaction, less improvement in knee function and OKS compared to patients with moderate-severe grades of arthritis.

A smartphone intervention that enhances real-world memory and promotes differentiation of hippocampal activity in older adults.

The act of remembering an everyday experience influences how we interpret the world, how we think about the future, and how we perceive ourselves. It also enhances long-term retention of the recalled content, increasing the likelihood that it will be recalled again. Unfortunately, the ability to recollect event-specific details and reexperience the past tends to decline with age. This decline in recollection may reflect a corresponding decrease in the distinctiveness of hippocampal memory representations. Despite these well-established changes, there are few effective cognitive behavioral interventions that target real-world episodic memory. We addressed this gap by developing a smartphone-based application called HippoCamera that allows participants to record labeled videos of everyday events and subsequently replay, high-fidelity autobiographical memory cues. In two experiments, we found that older adults were able to easily integrate this noninvasive intervention into their daily lives. Using HippoCamera to repeatedly reactivate memories for real-world events improved episodic recollection and it evoked more positive autobiographical sentiment at the time of retrieval. In both experiments, these benefits were observed shortly after the intervention and again after a 3-mo delay. Moreover, more detailed recollection was associated with more differentiated memory signals in the hippocampus. Thus, using this smartphone application to systematically reactivate memories for recent real-world experiences can help to maintain a bridge between the present and past in older adults.

RWE Framework: An Interactive Visual Tool to Support a Real-World Evidence Study Design.

OBJECTIVES: Real-world evidence (RWE) is essential for the development of pharmaceutical and medical technologies and informs treatment-related decisions by regulatory agencies, payers, healthcare providers, and patients. Given that planning RWE studies present diverse challenges, we developed the RWE Framework, a concise, visual, interactive tool designed to align multidisciplinary stakeholders toward common goals and encourage a methodical approach to RWE study planning. METHODS: A search of published literature and internet-based resources was performed to identify guidance on RWE study planning with decision and/or visual aids. A conceptual framework for a study design tool was developed based on best practices for RWE studies, enhanced with an infographic design, and refined by multidisciplinary input from RWE researchers. RESULTS: The searches confirmed an unmet need for a concise tool to support a broad range of RWE study designs: only two sources with decision/visual aids were identified. The novel RWE Framework comprises sequential decision steps with instructions to guide users through consideration of research objectives, product approval status, study setting, outcomes of interest, data availability in routine practice, need for primary data collection and/or randomization, study type and methodology, and applicable regulatory standards. Pilot testing using case studies of pharmaceutical assets demonstrated the utility of RWE Framework and applicability for use in team environments. CONCLUSIONS: The RWE Framework is a novel, concise, visual, and interactive tool to inform RWE study planning. It addresses a broad range of real-world study types and research objectives and was found to enhance RWE decision-making by multidisciplinary teams. Further validation is warranted.

Evaluation of pain relief treatment and timelines in emergency care in six European countries and Australia.

PURPOSE: Inadequate relief of pain is common in prehospital and hospital emergency department (ED) settings. We investigated pain treatments and timelines in patients receiving pre-hospital and hospital ED care to provide insight into potential approaches to reduce the burden of trauma-related pain. PATIENTS AND METHODS: In this observational, retrospective chart review, patients had received emergency care for musculoskeletal trauma injuries and analgesic treatment for moderate-to-severe pain in Belgium, France, Germany, Italy, Spain or Sweden. As inhaled low-dose methoxyflurane (LDM) is used extensively in Australia but was not widely available in Europe at the time of this analysis, data from Australia were collated to provide insight into the potential utility of this analgesic in Europe. The primary endpoint was time to administration of first pain relief treatment following arrival of paramedic/ED care. RESULTS: Randomly selected physicians (n=189) collated data from 856 patients (Europe: n=585; Australia: n=271) via an online survey. Time to first pain relief treatment varied between countries and was significantly longer across Europe versus Australia (mean [SD] 38.1 [34.7] vs 29.9 [35.5] mins; P=0.0017). Patients from Australia who received LDM experience a shorter mean (SD) time to first pain treatment following arrival of emergency care versus patients who received other analgesics (propensity score matched [n=85] per group: 21.7 [24.2] vs 39.1 [43.0] mins; P=0.0013). Across all countries, mean (SD) time to first analgesic was shorter when treatment was administered by paramedics versus hospital ED staff (15.7 [14.7] vs 49.1 [38.4] mins). CONCLUSIONS: While there was a large variation in analgesia timelines across countries, mean times are shorter in Australia compared with Europe overall. In Australia, use of LDM was associated with a significantly shorter time from emergency assistance to first pain treatment compared with non-LDM treatments. Further studies are needed to investigate the utility of LDM in Europe.

Different targets of mechanical alignment do not improve knee outcomes after TKA.

BACKGROUND: Accurate alignment is a basic principle of TKA surgery, but achieving alignment within this target may not translate into superior outcomes after surgery. PURPOSE: To assess if neutral TKA mechanical alignment was associated with superior knee outcomes and to examine the effect of various aspects of pre-operative and post-operative alignment on knee function. METHODS: Analysis of a database of 444 TKA patients between June 2009 and October 2016. Knee outcomes (WOMAC, AKSS and knee range of motion) were collected before surgery and during follow-up at a minimum of six months. RESULTS: Analysis included 444 TKA patients (62% female, mean age 66 years, mean follow-up 23 months). Deformity varied from 21° varus (mean = 7.9, SD = 2.8) to 17° valgus deformity (mean = 7.7, SD = 2.8). Pre-operatively, 101 (23%) knees were in native neutral mechanical alignment, while 278 (63%) were in varus and 65 (15%) were in valgus. Post-operatively, a group of 365 (82%) TKA were found to be in neutral mechanical alignment and a group of 79 (18%) TKA were noted to be 'Outliers' (17 [4%] TKA > 3° varus and 62 [14%] TKA > 3° valgus alignment). Restoration of the target of alignment of 0 ±â€¯3° or 0 ±â€¯1°, did not have better functional outcomes scores, range of motion or prosthesis longevity than those in the outlier range. CONCLUSION: Neutral TKA alignment did not appear to be a significant contributing factor to the improvement in knee function in short-medium term follow-up.