Rethinking PICO in the Machine Learning Era: ML-PICO

Background  Machine learning (ML) has captured the attention of many clinicians who may not have formal training in this area but are otherwise increasingly exposed to ML literature that may be relevant to their clinical specialties. ML papers that follow an outcomes-based research format can be assessed using clinical research appraisal frameworks such as PICO (Population, Intervention, Comparison, Outcome). However, the PICO frameworks strain when applied to ML papers that create new ML models, which are akin to diagnostic tests. There is a need for a new framework to help assess such papers. Objective  We propose a new framework to help clinicians systematically read and evaluate medical ML papers whose aim is to create a new ML model: ML-PICO (Machine Learning, Population, Identification, Crosscheck, Outcomes). We describe how the ML-PICO framework can be applied toward appraising literature describing ML models for health care. Conclusion  The relevance of ML to practitioners of clinical medicine is steadily increasing with a growing body of literature. Therefore, it is increasingly important for clinicians to be familiar with how to assess and best utilize these tools. In this paper we have described a practical framework on how to read ML papers that create a new ML model (or diagnostic test): ML-PICO. We hope that this can be used by clinicians to better evaluate the quality and utility of ML papers.     

Tumor vascular signals in renal masses: detection with Doppler US

The vascularity of 49 renal masses (26 malignant and 23 benign lesions) was investigated with duplex Doppler ultrasound. Doppler signals obtained at the margins of renal masses were defined as "tumor signals" when the Doppler-shifted frequency of the lesion exceeded the frequency shift in the ipsilateral main renal artery. These exceeded 2.5 kHz with a 3-MHz insonating frequency. Among the 26 renal masses that subsequently proved to be malignant, tumor signals were obtained in 15 of 18 (83%) untreated renal cell carcinomas, in three of four Wilms tumors, and in two patients with metastases to the kidney, but not in the one patient with lymphoma. None of the 23 benign renal masses demonstrated tumor signals. Tumor vascularity in malignant lesions gives rise to abnormal, high-velocity, Doppler-shifted signals that can help in the differential diagnosis of renal masses.     

Age group differences in psychological distress: the role of psychosocial risk factors that vary with age

BACKGROUND: There is continuing controversy about how age affects depression and anxiety, with a lack of consistent results across studies. Two reasons for this inconsistency are age bias in measures and different patterns of exposure to risk factors across age groups in various studies. METHOD: Data on anxiety and depression symptoms were collected in a community survey of 7485 persons aged 20-24, 40-44 or 60-64 years. These measures were investigated for factorial invariance across age groups. Data were also collected on a wide range of potential risk factors, including social, physical health and personal factors, with the aim of determining whether these factors might partly or wholly account for age group differences. RESULTS: The invariance of correlated latent factors representing anxiety and depression was examined across age groups, and a generalized measure of psychological distress was computed. Depression, anxiety and psychological distress showed a decline across age groups for females and a decline from 40-44 to 60-64 years for males. Some of these age differences were accounted for by other risk factors, with the most important being recent crises at work and negative social relationships with family and friends. CONCLUSION: Psychological distress generally declined across the age range 20-64 years and this was not attributable to measurement bias. Differential exposure to risk factors explained some, but not all, of the age group difference. Therefore other mechanisms that explain the lower level of distress in older age groups remain to be identified.     

Detection of reactive oxygen species (ROS) and apoptosis in human fragmented embryos

In human in-vitro fertilization (IVF)-embryo transfer, the in-vitro culture environment differs from in-vivo conditions in that the oxygen concentration is higher, and in such conditions the mouse embryos show a higher concentration of reactive oxygen species (ROS) in simple culture media. ROS are believed to cause damage to cell membranes and DNA fragmentation in somatic cells. This study was conducted to ascertain the level of H2O2 concentration within embryos and the morphological features of cell damage induced by H2O2. A total of 62 human oocytes and embryos (31 fragmented, 15 non-fragmented embryos, 16 unfertilized oocytes) was obtained from the IVF-embryo transfer programme. The relative intensity of H2O2 concentrations within embryos was measured using 2',7'-dichlorodihydrofluorescein diacetate by Quanti cell 500 fluorescence imaging and DNA fragmentation was observed with transmission electron microscopy and an in-situ apoptosis detection kit. The H2O2 concentrations were significantly higher in fragmented embryos (72.21 +/- 9.62, mean +/- SEM) compared to non-fragmented embryos (31.30 +/- 3.50, P < 0.05) and unfertilized oocytes (30.75 +/- 2.67, P < 0.05). Apoptosis was observed only in fragmented embryos, and was absent in non-fragmented embryos. Electron microscopic findings confirmed apoptotic bodies and cytoplasmic condensation in the fragmented blastomeres. We conclude that there is a direct relationship between increased H2O2 concentration and apoptosis, and that further studies should be undertaken to confirm these findings.      

On-Road Behavior in Older Drivers With Mild Cognitive Impairment

ObjectivesDementia increases the risk of unsafe driving, but this is less apparent in preclinical stages such as mild cognitive impairment (MCI). There is, however, limited detailed data on the patterns of driving errors associated with MCI. Here, we examined whether drivers with MCI exhibited different on-road error profiles compared with cognitively normal (CN) older drivers.DesignObservational.Setting and ParticipantsA total of 296 licensed older drivers [mean age 75.5 (SD = 6.2) years, 120 (40.5%) women] recruited from the community.MethodParticipants completed a health and driving history survey, a neuropsychological test battery, and an on-road driving assessment including driver-instructed and self-navigation components. Driving assessors were blind to participant cognitive status. Participants were categorized as safe or unsafe based on a validated on-road safety scale, and as having MCI based on International Working Group diagnostic criteria. Proportion of errors incurred as a function of error type and traffic context were compared across safe and unsafe MCI and CN drivers.ResultsCompared with safe CN drivers (n = 225), safe MCI drivers (n = 45) showed a similar pattern of errors in different traffic contexts. Compared with safe CN drivers, unsafe CN drivers (n = 17) were more likely to make errors in observation, speed control, lane position, and approach, and at stop/give-way signs, lane changes, and curved driving. Unsafe MCI drivers (n = 9) had additional difficulties at intersections, roundabouts, parking, straight driving, and under self-navigation conditions. A higher proportion of unsafe MCI drivers had multidomain subtype [n = 6 (67%)] than safe MCI drivers [n = 11 (25%)], odds ratio 6.2 (95% confidence interval, 1.4–29.6).Conclusion and ImplicationsAmong safe drivers, MCI and CN drivers exhibit similar on-road error profiles, suggesting driver restrictions based on MCI status alone are unwarranted. However, formal evaluation is recommended in such cases, as there is evidence drivers with multiple domains of cognitive impairment may require additional interventions to support safe driving.     

Open-label titration study of the safety of RMP-7 in patients with the acquired immune deficiency syndrome

RMP-7, a nine-amino acid bradykinin analogue, has been shown in animals to temporarily increase the permeability of the blood brain barrier to small molecules including amphotericin B, when administered intravenously. We sought to evaluate the safety of escalating doses of RMP-7 administered to human volunteers with the acquired immune deficiency syndrome (AIDS). Six HIV antibody-positive adults with CD4+ cell counts <50/mm³ received three increasing doses of RMP-7 on successive days: 30 ng/kg, 100 ng/kg and 300 ng/kg infused over 2, 2 and 10 min, respectively. Adverse experiences were dose-related, mild-moderate in intensity, primarily related to vasodilation and resolved rapidly without sequelae. Mean maximum increases in pulse rate at 30 ng/kg, 100 ng/kg and 300 ng/kg were 4.0, 7.8 and 28.2 beats per min, respectively. The maximum changes in average mean arterial pressure were +7.7, +5.6 and −0.2 mmHg from baseline, respectively. Minor increases in liver enzymes were noted in three patients, all with pre-existing enzyme elevations. Despite the high frequency of both occult and overt cardiovascular abnormalities in advanced HIV infection, RMP-7 is shown to be safe in this group of AIDS patients at all dosage levels tested, with adverse effects similar to previous experience in healthy humans.