A "one-size-fits-most" walking recognition method for smartphones, smartwatches, and wearable accelerometers.

The ubiquity of personal digital devices offers unprecedented opportunities to study human behavior. Current state-of-the-art methods quantify physical activity using "activity counts," a measure which overlooks specific types of physical activities. We propose a walking recognition method for sub-second tri-axial accelerometer data, in which activity classification is based on the inherent features of walking: intensity, periodicity, and duration. We validate our method against 20 publicly available, annotated datasets on walking activity data collected at various body locations (thigh, waist, chest, arm, wrist). We demonstrate that our method can estimate walking periods with high sensitivity and specificity: average sensitivity ranged between 0.92 and 0.97 across various body locations, and average specificity for common daily activities was typically above 0.95. We also assess the method's algorithmic fairness to demographic and anthropometric variables and measurement contexts (body location, environment). Finally, we release our method as open-source software in Python and MATLAB.

Smartphone-Based Activity Recognition Using Multistream Movelets Combining Accelerometer and Gyroscope Data.

Physical activity patterns can reveal information about one's health status. Built-in sensors in a smartphone, in comparison to a patient's self-report, can collect activity recognition data more objectively, unobtrusively, and continuously. A variety of data analysis approaches have been proposed in the literature. In this study, we applied the movelet method to classify the activities performed using smartphone accelerometer and gyroscope data, which measure a phone's acceleration and angular velocity, respectively. The movelet method constructs a personalized dictionary for each participant using training data and classifies activities in new data with the dictionary. Our results show that this method has the advantages of being interpretable and transparent. A unique aspect of our movelet application involves extracting unique information, optimally, from multiple sensors. In comparison to single-sensor applications, our approach jointly incorporates the accelerometer and gyroscope sensors with the movelet method. Our findings show that combining data from the two sensors can result in more accurate activity recognition than using each sensor alone. In particular, the joint-sensor method reduces errors of the gyroscope-only method in differentiating between standing and sitting. It also reduces errors in the accelerometer-only method when classifying vigorous activities.

Estimating longitudinal depressive symptoms from smartphone data in a transdiagnostic cohort.

BACKGROUND: Passive measures collected using smartphones have been suggested to represent efficient proxies for depression severity, but the performance of such measures across diagnoses has not been studied. METHODS: We enrolled a cohort of 45 individuals (11 with major depressive disorder, 11 with bipolar disorder, 11 with schizophrenia or schizoaffective disorder, and 12 individuals with no axis I psychiatric disorder). During the 8-week study period, participants were evaluated with a rater-administered Montgomery-Åsberg Depression Rating Scale (MADRS) biweekly, completed self-report PHQ-8 measures weekly on their smartphone, and consented to collection of smartphone-based GPS and accelerometer data in order to learn about their behaviors. We utilized linear mixed models to predict depression severity on the basis of phone-based PHQ-8 and passive measures. RESULTS: Among the 45 individuals, 38 (84%) completed the 8-week study. The average root-mean-squared error (RMSE) in predicting the MADRS score (scale 0-60) was 4.72 using passive data alone, 4.27 using self-report measures alone, and 4.30 using both. CONCLUSIONS: While passive measures did not improve MADRS score prediction in our cross-disorder study, they may capture behavioral phenotypes that cannot be measured objectively, granularly, or over long-term via self-report.

Smartphone-based Assessment of Preoperative Decision Conflict and Postoperative Physical Activity Among Patients Undergoing Cancer Surgery: A Prospective Cohort Study.

OBJECTIVE: To determine the prevalence of clinically significant decision conflict (CSDC) among patients undergoing cancer surgery and associations with postoperative physical activity, as measured through smartphone accelerometer data. BACKGROUND: Patients with cancer face challenging treatment decisions, which may lead to CSDC. CSDC negatively affects patient-provider relationships, psychosocial functioning, and health-related quality of life; however, physical manifestations of CSDC remain poorly characterized. METHODS: Adult smartphone-owners undergoing surgery for breast, skin-soft-tissue, head-and-neck, or abdominal cancer (July 2017-2019) were approached. Patients downloaded the Beiwe application that delivered the Decision Conflict Scale (DCS) preoperatively and collected smartphone accelerometer data continuously from enrollment through 6 months postop-eratively. Restricted-cubic-spline regression, adjusting for a priori potential confounders (age, type of surgery, support status, and postoperative complications) was used to determine trends in postoperative daily physical activity among patients with and without CSDC (DCS score >25/100). RESULTS: Among 99 patients who downloaded the application, 85 completed the DCS (86% participation rate). Twenty-three (27%) reported CSDC. These patients were younger (mean age 48.3 years [standard deviation 14.2]-vs-55.0 [13.3], P = 0.047) and more frequently lived alone (22%-vs-6%, P = 0.042). There were no differences in preoperative physical activity (115.4 minutes [95%CI 90.9, 139.9]-vs-110.8 [95%CI 95.7, 126.0], P = 0.753). Adjusted postoperative physical activity was lower among patients reporting CSDC at 30 days (difference 33.1 minutes [95%CI 5.93,60.2], P = 0.017), 60 days 35.5 [95%CI 8.50, 62.5], P = 0.010 and 90 days 31.8 [95%CI 5.44, 58.1], P = 0.018 postoperatively. CONCLUSIONS: CSDC was prevalent among patients who underwent cancer surgery and associated with lower postoperatively daily physical activity. These data highlight the importance of addressing modifiable decisional needs of patients through enhanced shared decision-making.

Correction: Huang, E.J., and Onnela, J.P. Augmented Movelet Method for Activity Classification Using Smartphone Gyroscope and Accelerometer Data. Sensors 2020, 20(13), 3706.

The authors wish to make the following corrections to this paper [...].

Augmented Movelet Method for Activity Classification Using Smartphone Gyroscope and Accelerometer Data.

Physical activity, such as walking and ascending stairs, is commonly used in biomedical settings as an outcome or covariate. Researchers have traditionally relied on surveys to quantify activity levels of subjects in both research and clinical settings, but surveys are subjective in nature and have known limitations, such as recall bias. Smartphones provide an opportunity for unobtrusive objective measurement of physical activity in naturalistic settings, but their data tends to be noisy and needs to be analyzed with care. We explored the potential of smartphone accelerometer and gyroscope data to distinguish between walking, sitting, standing, ascending stairs, and descending stairs. We conducted a study in which four participants followed a study protocol and performed a sequence of activities with one phone in their front pocket and another phone in their back pocket. The subjects were filmed throughout, and the obtained footage was annotated to establish moment-by-moment ground truth activity. We introduce a modified version of the so-called movelet method to classify activity type and to quantify the uncertainty present in that classification. Our results demonstrate the promise of smartphones for activity recognition in naturalistic settings, but they also highlight challenges in this field of research.

Using Smartphones to Capture Novel Recovery Metrics After Cancer Surgery.

Importance: Patient-generated health data captured from smartphone sensors have the potential to better quantify the physical outcomes of surgery. The ability of these data to discriminate between postoperative trends in physical activity remains unknown. Objective: To assess whether physical activity captured from smartphone accelerometer data can be used to describe postoperative recovery among patients undergoing cancer operations. Design, Setting, and Participants: This prospective observational cohort study was conducted from July 2017 to April 2019 in a single academic tertiary care hospital in the United States. Preoperatively, adults (age ≥18 years) who spoke English and were undergoing elective operations for skin, soft tissue, head, neck, and abdominal cancers were approached. Patients were excluded if they did not own a smartphone. Exposures: Study participants downloaded an application that collected smartphone accelerometer data continuously for 1 week preoperatively and 6 months postoperatively. Main Outcomes and Measures: The primary end points were trends in daily exertional activity and the ability to achieve at least 60 minutes of daily exertional activity after surgery among patients with vs without a clinically significant postoperative event. Postoperative events were defined as complications, emergency department presentations, readmissions, reoperations, and mortality. Results: A total of 139 individuals were approached. In the 62 enrolled patients, who were followed up for a median (interquartile range [IQR]) of 147 (77-179) days, there were no preprocedural differences between patients with vs without a postoperative event. Seventeen patients (27%) experienced a postoperative event. These patients had longer operations than those without a postoperative event (median [IQR], 225 [152-402] minutes vs 107 [68-174] minutes; P < .001), as well as greater blood loss (median [IQR], 200 [35-515] mL vs 25 [5-100] mL; P = .006) and more follow-up visits (median [IQR], 2 [2-4] visits vs 1 [1-2] visits; P = .002). Compared with mean baseline daily exertional activity, patients with a postoperative event had lower activity at week 1 (difference, -41.6 [95% CI, -75.1 to -8.0] minutes; P = .02), week 3 (difference, -40.0 [95% CI, -72.3 to -3.6] minutes; P = .03), week 5 (difference, -39.6 [95% CI, -69.1 to -10.1] minutes; P = .01), and week 6 (difference, -36.2 [95% CI, -64.5 to -7.8] minutes; P = .01) postoperatively. Fewer of these patients were able to achieve 60 minutes of daily exertional activity in the 6 weeks postoperatively (proportions: week 1, 0.40 [95% CI, 0.31-0.49]; P < .001; week 2, 0.49 [95% CI, 0.40-0.58]; P = .003; week 3, 0.39 [95% CI, 0.30-0.48]; P < .001; week 4, 0.47 [95% CI, 0.38-0.57]; P < .001; week 5, 0.51 [95% CI, 0.42-0.60]; P < .001; week 6, 0.73 [95% CI, 0.68-0.79] vs 0.43 [95% CI, 0.33-0.52]; P < .001). Conclusions and Relevance: Smartphone accelerometer data can describe differences in postoperative physical activity among patients with vs without a postoperative event. These data help objectively quantify patient-centered surgical recovery, which have the potential to improve and promote shared decision-making, recovery monitoring, and patient engagement.

Constructing a confidence interval for the fraction who benefit from treatment, using randomized trial data.

The fraction who benefit from treatment is the proportion of patients whose potential outcome under treatment is better than that under control. Inference on this parameter is challenging since it is only partially identifiable, even in our context of a randomized trial. We propose a new method for constructing a confidence interval for the fraction, when the outcome is ordinal or binary. Our confidence interval procedure is pointwise consistent. It does not require any assumptions about the joint distribution of the potential outcomes, although it has the flexibility to incorporate various user-defined assumptions. Our method is based on a stochastic optimization technique involving a second-order, asymptotic approximation that, to the best of our knowledge, has not been applied to biomedical studies. This approximation leads to statistics that are solutions to quadratic programs, which can be computed efficiently using optimization tools. In simulation, our method attains the nominal coverage probability or higher, and can have narrower average width than competitor methods. We apply it to a trial of a new intervention for stroke.

Modeling of Clinical Phenotypes Assessed at Discrete Study Visits.

In studies of clinical phenotypes, such as dementia, disability, and frailty, participants are typically assessed at in-person clinic visits. Thus, the precise time of onset for the phenotype is unknown. The discreteness of the clinic visits yields grouped event time data. We investigate how to perform a risk factor analysis in the case of grouped data. Since visits can be months to years apart, numbers of ties can be large, causing the exact tie-handling method of the Cox model to be computationally infeasible. We propose two, new, computationally efficient approximations to the exact method: Laplace approximation and an analytic approximation. Through extensive simulation studies, we compare these new methods to the Prentice-Gloeckler model and the Cox model using Efron's and Breslow's tie-handling methods. In addition, we compare the methods in an application to a large cohort study (N = 3,605) on the development of clinical frailty in older adults. In our simulations, the Laplace approximation has low bias in all settings, and the analytic approximation has low bias in settings where the regression coefficient is not large in magnitude. Their corresponding confidence intervals also have approximately the nominal coverage probability. In the data application, the results from the approximations are nearly identical to that of the Prentice-Gloeckler model.

Inequality in treatment benefits: Can we determine if a new treatment benefits the many or the few?

In many randomized controlled trials, the primary analysis focuses on the average treatment effect and does not address whether treatment benefits are widespread or limited to a select few. This problem affects many disease areas, since it stems from how randomized trials, often the gold standard for evaluating treatments, are designed and analyzed. Our goal is to learn about the fraction who benefit from a new treatment using randomized trial data. We consider the case where the outcome is ordinal, with binary outcomes as a special case. In general, the fraction who benefit is non-identifiable, and the best that can be obtained are sharp lower and upper bounds. Our contributions include (i) proving the plug-in estimator of the bounds can be inconsistent if support restrictions are made on the joint distribution of the potential outcomes; (ii) developing the first consistent estimator for this case; and (iii) applying this estimator to a randomized trial of a medical treatment to determine whether the estimates can be informative. Our estimator is computed using linear programming, allowing fast implementation. R code is provided.

Medications associated with clinical deterioration in hospitalized children.

BACKGROUND: Medical emergency teams have been shown to reduce mortality in children's hospitals, but there are many potential barriers to their activation. Surveillance tools using electronic health record data help identify children at risk of deterioration. Existing early warning scores primarily include vital signs, but may benefit from the incorporation of medications. OBJECTIVE: We aimed to identify the therapeutic classes of medications temporally associated with clinical deterioration that could be incorporated with vital signs into surveillance tools. DESIGN: Case-crossover study. SETTING: The Children's Hospital of Philadelphia. PATIENTS: Children with clinical deterioration, defined as cardiopulmonary arrest, acute respiratory compromise, or urgent intensive care unit transfer while hospitalized on pediatric wards (n = 141). EXPOSURES: Intravenous administrations of medications from therapeutic classes administered in ≥5% of control periods. RESULTS: Nine therapeutic classes were significantly associated with clinical deterioration: glycopeptide antibiotics, anaerobic antibiotics, third-generation and fourth-generation cephalosporins, aminoglycoside antibiotics, systemic corticosteroids, benzodiazepines, loop diuretics, narcotic analgesics (full opioid agonists), and antidotes to hypersensitivity reactions. CONCLUSIONS: We identified a set of therapeutic classes associated with increased risk of clinical deterioration. Future work should focus on evaluating whether including these therapeutic classes in multivariable models improves their accuracy in detecting early, evolving deterioration.