Evaluation of Machine Learning to Detect Influenza Using Wearable Sensor Data and Patient-Reported Symptoms: Cohort Study.

BACKGROUND: Machine learning offers quantitative pattern recognition analysis of wearable device data and has the potential to detect illness onset and monitor influenza-like illness (ILI) in patients who are infected. OBJECTIVE: This study aims to evaluate the ability of machine-learning algorithms to distinguish between participants who are influenza positive and influenza negative in a cohort of symptomatic patients with ILI using wearable sensor (activity) data and self-reported symptom data during the latent and early symptomatic periods of ILI. METHODS: This prospective observational cohort study used the extreme gradient boosting (XGBoost) classifier to determine whether a participant was influenza positive or negative based on 3 models using symptom-only data, activity-only data, and combined symptom and activity data. Data were collected from the Home Testing of Respiratory Illness (HTRI) study and FluStudy2020, both conducted between December 2019 and October 2020. The model was developed using the FluStudy2020 data and tested on the HTRI data. Analyses included participants in these studies with an at-home influenza diagnostic test result. Fitbit (Google LLC) devices were used to measure participants' steps, heart rate, and sleep parameters. Participants detailed their ILI symptoms, health care-seeking behaviors, and quality of life. Model performance was assessed by area under the curve (AUC), balanced accuracy, recall (sensitivity), specificity, precision (positive predictive value), negative predictive value, and weighted harmonic mean of precision and recall (F2) score. RESULTS: An influenza diagnostic test result was available for 953 and 925 participants in HTRI and FluStudy2020, respectively, of whom 848 (89%) and 840 (90.8%) had activity data. For the training and validation sets, the highest performing model was trained on the combined symptom and activity data (training AUC=0.77; validation AUC=0.74) versus symptom-only (training AUC=0.73; validation AUC=0.72) and activity-only (training AUC=0.68; validation AUC=0.65) data. For the FluStudy2020 test set, the performance of the model trained on combined symptom and activity data was closely aligned with that of the symptom-only model (combined symptom and activity test AUC=0.74; symptom-only test AUC=0.74). These results were validated using independent HTRI data (combined symptom and activity evaluation AUC=0.75; symptom-only evaluation AUC=0.74). The top features guiding influenza detection were cough; mean resting heart rate during main sleep; fever; total minutes in bed for the combined model; and fever, cough, and sore throat for the symptom-only model. CONCLUSIONS: Machine-learning algorithms had moderate accuracy in detecting influenza, suggesting that previous findings from research-grade sensors tested in highly controlled experimental settings may not easily translate to scalable commercial-grade sensors. In the future, more advanced wearable sensors may improve their performance in the early detection and discrimination of viral respiratory infections.

A Phase 2 Randomized Trial Evaluating the Antiviral Activity and Safety of the Direct-Acting Antiviral Bemnifosbuvir in Ambulatory Patients with Mild or Moderate COVID-19 (MOONSONG Study).

Bemnifosbuvir is an oral antiviral drug with a dual mechanism of action targeting viral RNA polymerase, with in vitro activity against SARS-CoV-2. We conducted a phase 2, double-blind study evaluating the antiviral activity, safety, efficacy, and pharmacokinetics of bemnifosbuvir in ambulatory patients with mild/moderate COVID-19. Patients were randomized 1:1 to bemnifosbuvir 550 mg or placebo (cohort A) and 3:1 to bemnifosbuvir 1,100 mg or placebo (cohort B); all doses were given twice daily for 5 days. The primary endpoint was a change from baseline in the amount of nasopharyngeal SARS-CoV-2 viral RNA by reverse transcription PCR (RT-PCR). The modified intent-to-treat infected population comprised 100 patients (bemnifosbuvir 550 mg, n = 30; bemnifosbuvir 1,100 mg, n = 30; cohort A placebo, n = 30; cohort B placebo, n = 10). The primary endpoint was not met: the difference in viral RNA adjusted means at day 7 was -0.25 log10 copies/mL between bemnifosbuvir 550 mg and cohort A placebo (80% confidence interval [CI], -0.66 to 0.16; P = 0.4260), and -0.08 log10 copies/mL between bemnifosbuvir 1,100 mg and pooled placebo (80% CI, -0.48 to 0.33; P = 0.8083). Bemnifosbuvir 550 mg was well tolerated. Incidence of nausea and vomiting was higher with bemnifosbuvir 1,100 mg (10.0% and 16.7% of patients, respectively) than pooled placebo (2.5% nausea, 2.5% vomiting). In the primary analysis, bemnifosbuvir did not show meaningful antiviral activity on nasopharyngeal viral load as measured by RT-PCR compared with placebo in patients with mild/moderate COVID-19. The trial is registered at ClinicalTrials.gov under registration number NCT04709835. IMPORTANCE COVID-19 continues to be a major global public health challenge, and there remains a need for effective and convenient direct-acting antivirals that can be administered outside health care settings. Bemnifosbuvir is an oral antiviral with a dual mechanism of action and potent in vitro activity against SARS-CoV-2. In this study, we evaluated the antiviral activity, safety, efficacy, and pharmacokinetics of bemnifosbuvir in ambulatory patients with mild/moderate COVID-19. In the primary analysis, bemnifosbuvir did not show meaningful antiviral activity compared with placebo as assessed by nasopharyngeal viral loads. The negative predictive value of nasopharyngeal viral load reduction for clinical outcomes in COVID-19 is currently unclear, and further evaluation of bemnifosbuvir for COVID-19 may be warranted despite the findings observed in this study.

Characterization of Influenza-Like Illness Burden Using Commercial Wearable Sensor Data and Patient-Reported Outcomes: Mixed Methods Cohort Study.

BACKGROUND: The burden of influenza-like illness (ILI) is typically estimated via hospitalizations and deaths. However, ILI-associated morbidity that does not require hospitalization remains poorly characterized. OBJECTIVE: The main objective of this study was to characterize ILI burden using commercial wearable sensor data and investigate the extent to which these data correlate with self-reported illness severity and duration. Furthermore, we aimed to determine whether ILI-associated changes in wearable sensor data differed between care-seeking and non-care-seeking populations as well as between those with confirmed influenza infection and those with ILI symptoms only. METHODS: This study comprised participants enrolled in either the FluStudy2020 or the Home Testing of Respiratory Illness (HTRI) study; both studies were similar in design and conducted between December 2019 and October 2020 in the United States. The participants self-reported ILI-related symptoms and health care-seeking behaviors via daily, biweekly, and monthly surveys. Wearable sensor data were recorded for 120 and 150 days for FluStudy2020 and HTRI, respectively. The following features were assessed: total daily steps, active time (time spent with >50 steps per minute), sleep duration, sleep efficiency, and resting heart rate. ILI-related changes in wearable sensor data were compared between the participants who sought health care and those who did not and between the participants who tested positive for influenza and those with symptoms only. Correlative analyses were performed between wearable sensor data and patient-reported outcomes. RESULTS: After combining the FluStudy2020 and HTRI data sets, the final ILI population comprised 2435 participants. Compared with healthy days (baseline), the participants with ILI exhibited significantly reduced total daily steps, active time, and sleep efficiency as well as increased sleep duration and resting heart rate. Deviations from baseline typically began before symptom onset and were greater in the participants who sought health care than in those who did not and greater in the participants who tested positive for influenza than in those with symptoms only. During an ILI event, changes in wearable sensor data consistently varied with those in patient-reported outcomes. CONCLUSIONS: Our results underscore the potential of wearable sensors to discriminate not only between individuals with and without influenza infections but also between care-seeking and non-care-seeking populations, which may have future application in health care resource planning. TRIAL REGISTRATION: Clinicaltrials.gov NCT04245800; https://clinicaltrials.gov/ct2/show/NCT04245800.

Predictors of Seeking Care for Influenza-Like Illness in a Novel Digital Study.

Background: Previous research has estimated that >50% of individuals experiencing influenza-like illness (ILI) do not seek health care. Understanding factors influencing care-seeking behavior for viral respiratory infections may help inform policies to improve access to care and protect public health. We used person-generated health data (PGHD) to identify factors associated with seeking care for ILI. Methods: Two observational studies (FluStudy2020, ISP) were conducted during the United States 2019-2020 influenza season. Participants self-reported ILI symptoms using the online Evidation platform. A log-binomial regression model was used to identify factors associated with seeking care. Results: Of 1667 participants in FluStudy2020 and 47 480 participants in ISP eligible for analysis, 518 (31.1%) and 11 426 (24.1%), respectively, sought health care. Participants were mostly female (92.2% FluStudy2020, 80.6% ISP) and aged 18-49 years (89.6% FluStudy2020, 89.8% ISP). In FluStudy2020, factors associated with seeking care included having health insurance (risk ratio [RR], 2.14; 95% CI, 1.30-3.54), more severe respiratory symptoms (RR, 1.53; 95% CI, 1.37-1.71), and comorbidities (RR, 1.37; 95% CI, 1.20-1.58). In ISP, the strongest predictor of seeking care was high symptom number (RR for 6/7 symptoms, 2.14; 95% CI, 1.93-2.38). Conclusions: Using PGHD, we confirmed low rates of health care-seeking behavior for ILI and show that having health insurance, comorbidities, and a high symptom burden were associated with seeking health care. Reducing barriers in access to care for viral respiratory infections may lead to better disease management and contribute to protecting public health.

Development and validation of prediction models to estimate risk of primary total hip and knee replacements using data from the UK: two prospective open cohorts using the UK Clinical Practice Research Datalink.

OBJECTIVES: The ability to efficiently and accurately predict future risk of primary total hip and knee replacement (THR/TKR) in earlier stages of osteoarthritis (OA) has potentially important applications. We aimed to develop and validate two models to estimate an individual's risk of primary THR and TKR in patients newly presenting to primary care. METHODS: We identified two cohorts of patients aged ≥40 years newly consulting hip pain/OA and knee pain/OA in the Clinical Practice Research Datalink. Candidate predictors were identified by systematic review, novel hypothesis-free 'Record-Wide Association Study' with replication, and panel consensus. Cox proportional hazards models accounting for competing risk of death were applied to derive risk algorithms for THR and TKR. Internal-external cross-validation (IECV) was then applied over geographical regions to validate two models. RESULTS: 45 predictors for THR and 53 for TKR were identified, reviewed and selected by the panel. 301 052 and 416 030 patients newly consulting between 1992 and 2015 were identified in the hip and knee cohorts, respectively (median follow-up 6 years). The resultant model C-statistics is 0.73 (0.72, 0.73) and 0.79 (0.78, 0.79) for THR (with 20 predictors) and TKR model (with 24 predictors), respectively. The IECV C-statistics ranged between 0.70-0.74 (THR model) and 0.76-0.82 (TKR model); the IECV calibration slope ranged between 0.93-1.07 (THR model) and 0.92-1.12 (TKR model). CONCLUSIONS: Two prediction models with good discrimination and calibration that estimate individuals' risk of THR and TKR have been developed and validated in large-scale, nationally representative data, and are readily automated in electronic patient records.

Clinical Significance of Medial Versus Lateral Compartment Patellofemoral Osteoarthritis: Cross-Sectional Analyses in an Adult Population With Knee Pain.

OBJECTIVE: To determine the comparative prevalence, associations with selected patient characteristics, and clinical outcomes of medial and lateral compartment patellofemoral (PF) joint osteoarthritis (OA). METHODS: Information was collected by questionnaires, clinical assessment, and radiographs from 745 eligible community-dwelling symptomatic adults age ≥50 years. PF joint space narrowing (JSN) and osteophytes were scored from skyline radiographs using the Osteoarthritis Research Society International atlas. Multilevel models were used to assess associations of compartmental PF joint OA with age, sex, body mass index (BMI) and varus-valgus malalignment, while median regression was used to examine associations with clinical outcomes (current pain intensity on a numeric rating scale [0-10] and the function subscale of the Western Ontario and McMaster Universities Osteoarthritis Index [0-68]). RESULTS: Isolated lateral PF joint OA was more common than isolated medial PF joint OA, particularly at higher severity thresholds. Irrespective of severity threshold, age (≥2 odds ratio [OR] 1.19 [95% confidence interval (95% CI) 1.12, 1.26]), BMI (≥2 OR 1.15 [95% CI 1.07, 1.24]), and valgus malalignment (≥2 OR 2.58 [95% CI 1.09, 6.07]) were associated with increased odds of isolated lateral JSN, but isolated medial JSN was only associated with age (≥2 OR 1.20 [95% CI 1.14, 1.27]). The pattern of association was less clear for PF joint osteophytes. Isolated lateral PF joint OA, defined by JSN or osteophytes, was associated with higher pain scores than isolated medial PF joint OA, but these differences were modest and were not significant. A similar pattern of association was seen for functional limitation but only when PF joint OA was defined by JSN. CONCLUSION: Isolated lateral PF joint OA is more common than isolated medial PF joint OA, and it is more consistently associated with established OA risk factors. It is also associated with higher, but clinically nonsignificant, pain and function scores than isolated medial PF joint OA, particularly when PF joint OA is defined using JSN.

Acute myocardial infarction in a young patient with hyperhomocysteinaemia.

Homocysteinuria is a rare inborn error of metabolism known to be associated with an increased risk of vascular events. A 36-year-old Caucasian man presented with a 2 day history of epigastric discomfort associated with nausea and sweating. He has a history of homocysteinuria and had been poorly compliant with treatment. An ECG showed ST-segment elevation and Q-waves in anterior leads. Blood tests showed markedly elevated high-sensitivity troponin and high homocysteine levels. He had a failed primary percutaneous coronary intervention due to extensive thrombus in the left anterior descending artery, which was aspirated and he received integrelin infusion for 48 h. Echocardiogram showed mild-to-moderate impairment of left ventricular function with apical akinesis extending to the mid-portion of anteroseptal walls consistent with anterior myocardial infarction. He was started on homocysteine-lowering treatment with betaine and folic acid. He is now on follow-up with clinical chemistry and cardiac rehabilitation.

Imported malaria in Scotland--an overview of surveillance, reporting and trends.

BACKGROUND: Imported malaria cases continue to occur and are often underreported. This study assessed reporting of malaria cases and their characteristics in Scotland. METHODS: Cases were identified at the study sites of Aberdeen, Edinburgh, Glasgow and Inverness. The number of cases identified in the period 2003-2008 was compared to surveillance databases from Health Protection Scotland (HPS) and the Malaria Reference Laboratory (MRL). Case characteristics were recorded and analysed. RESULTS: Of 252 cases of malaria diagnosed and treated, an estimated 235 (93.3%) were reported to the MRL. Between 2006 and 2008, 114 of 126 cases (90.5%) were reported to HPS. Plasmodium falciparum caused 173 cases (68.7%). Business and professional travel accounted for 35.3% of cases (higher in Aberdeen), followed by visiting friends and relatives (33.1%) and holiday makers (25.5%). The majority of infections were imported from West Africa and 65.7% of patients for whom data on prophylaxis was available had taken no or inappropriate prophylaxis. CONCLUSIONS: Reporting of malaria in Scotland can be improved. There is a continued need to optimise preventive measures and adherence to chemoprophylaxis amongst business travellers, those visiting friends and relatives, and holiday makers in endemic countries in order to reduce imported malaria cases.