KAKirstin Aschbacher

Kirstin Aschbacher

*This is an algorithmically generated author record based on publicly available information.

Kirstin Aschbacher has a total of 109 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Longitudinal Patterns of Engagement and Clinical Outcomes: Results From a Therapist-Supported Digital Mental Health Intervention

    by Kirstin Aschbacher, Luisa M. Rivera, Silvan Msc Hornstein, Benjamin W. Nelson, Valerie L. Forman-Hoffman, Nicholas C. Peiper

    Objective Digital mental health interventions (DMHI) are an effective treatment modality for common mental disorders like depression and anxiety; however, the role of intervention engagement as a longitudinal “dosing” factor is poorly understood in relation to clinical outcomes. Methods We studied 4,978 participants in a 12-week therapist-supported DMHI (June 2020- December 2021), applying a longitudinal agglomerative hierarchical cluster analysis (HCA) to the number of days per week of intervention engagement. The proportion of people demonstrating remission in depression and anxiety symptoms during the intervention was calculated for each cluster. Multivariable logistic regression models were fit to examine associations between the engagement clusters and symptom remission adjusting for demographic and clinical characteristics. Results Based upon clinical interpretability and stopping rules, four clusters were derived from the HCA (in descending order): 1) sustained high engagers (45.0%); 2) late disengagers (24.1%); 3) early disengagers (22.5%); and 4) immediate disengagers (8.4%). Bivariate and multivariate analyses supported a dose-response relationship between engagement and depression symptom remission, while the pattern was partially evident for anxiety symptom remission. In multivariable logistic regression models, older age groups, males, and Asians had increased odds of achieving depression and anxiety symptom remission, while higher odds of anxiety symptom remission were observed among gender expansive individuals. Conclusions Segmentation based on the frequency of engagement performs well in discerning timing of intervention disengagement and a dose-response relationship with clinical outcomes. The findings among the demographic subpopulations indicate therapist-supported DMHIs may be effective in addressing mental health problems among patients who disproportionately experience stigma and structural barriers to care. Machine learning models can enable precision care by delineating how heterogeneous patterns of engagement over time relate with clinical outcomes. This empirical identification may help clinicians personalize and optimize interventions to prevent premature disengagement.


  • Open Access

    Toward Digital Precision Care: Large-scale Age and Gender Norms for Heart Rate Variability Biofeedback Metrics from a Real-World Digital Mental Health Intervention

    by Kirstin Aschbacher, Mara Mather, Paul Lehrer, Richard Gevirtz, Elissa Epel, Nicholas C. Peiper

    Purpose: Heart rate variability biofeedback (HRVB) is an efficacious treatment for depression and anxiety. However, translation to digital mental health interventions (DMHI) is hindered by the lack of normative data for real-time HRVB metrics, without which, biofeedback risks reduced engagement and efficacy.Methods: We analyzed HRVB data from 5,158 participants in a therapist-supported DMHI incorporating slow-paced breathing to treat depression or anxious symptoms. A real-time metric of HRVB amplitude, shared with users, and a gold-standard research metric of low-frequency (LF) power were computed for each session, and averaged within-participant over 2 weeks. Descriptive statistics characterized age- and gender-based norms for HRVB metrics, while multivariate regression analyses confirmed the effects adjusting for demographic, psychological, and health factors.Results: Real-time HRVB amplitude correlated strongly (r=.93, p<.001) with the LF research gold-standard. Age was associated with a significant decline in HRVB across the lifespan (p<.001), which was steeper among men than women, adjusting for demographic, psychological and health factors. Moreover, resting high and low-frequency power, body mass index, hypertension, Asian race, depression symptoms, and trauma history were significantly associated with HRVB amplitude in multivariate analyses (p’s<.01). Conclusions: Real-time HRVB metrics correlate highly with research gold-standard metrics, enabling automated biofeedback delivery as a treatment component of DMHIs. However, without a precision care algorithm to tailor feedback difficulty, automated HRVB will likely result in suboptimal efficacy and heightened attrition, particularly among older adults. Moreover, we identify clinical and health factors (BMI, hypertension) relevant to building an equitable, accurate algorithm.


  • Open Access

    Author Correction: Feasibility of continuous fever monitoring using wearable devices

    by Benjamin L. Smarr, Kirstin Aschbacher, Sarah M. Fisher, Anoushka Chowdhary, Stephan Dilchert, Karena Puldon, Adam Rao, Frederick M. Hecht, Ashley E. Mason

  • Open Access

    Association of Depressive Symptoms with Resting Heart Rate Variability recorded from a Wearable Device under Naturalistic Conditions: A Machine Learning Study

    by Silvan Hornstein, Moritz Seiler, Valerie L Forman-Hoffman, Benjamin W Nelson, Kirstin Aschbacher, Kerstin Ritter, Kevin Hilbert

    Heart Rate Variability (HRV) is a putative biological mechanism and marker of depression, which can be measured with low costs and in naturalistic settings. Therefore, applying machine learning (ML) on HRV data, could be a promising path to predict and monitor depressive symptoms in patient populations. A sample of 2629 participants completed self-administered HRV resting state recordings with a wearable device as part of a therapist-supported digital mental health intervention for anxiety and depression. Findings revealed that higher baseline Patient Health Questionaire-9 (PHQ-9) scores were associated with significantly higher heart rate (r = 0.14, p <0.001) and lower root mean error of successive differences (r = -0.09, p <0.001). We contrasted multiple machine learning algorithms with repeated 5-fold cross validation on a separate training set of 1830 to classify participants as having moderate depressive symptoms, indexed as a PHQ-9 score of 10 or greater. The best performing algorithm reached an area under the receiver operating characteristic score (ROC AUC) of 0.56 (p = 0.003) on the previously unseen test set, which was less than expected. Sensitivity analyses showed that model performance was not substantially impacted by variations in PHQ cut-offs or data quality. In conclusion, using wearable-derived HRV for cross-sectional prediction of depressive symptoms is challenging in settings with myriad real-world, day-to-day sources of variation.


  • Open Access

    Evidence-Based Therapist-Supported Digital Mental Health Intervention for Patients Experiencing Medical Multimorbidity: A Retrospective Cohort Intent-to-Treat Study

    by Benjamin W Nelson, Nicholas C. Peiper, Kirstin Aschbacher, Valerie Forman-Hoffman

    Objective: Multimorbidity or the co-occurrence of multiple health conditions is increasing globally and associated with significant psychological complications. It is unclear whether digital mental health (DMH) interventions for patients experiencing medical multimorbidity (MMB) are effective, particularly given that this patient population faces more treatment resistance. Therefore, there is a need for initial preliminary research to examine whether DMH interventions are associated with reductions on depressive and anxiety symptoms in the context of MMB.Methods: This preregistered retrospective cohort intent-to-treat study with 2,819 patients enrolled in a therapist-supported DMH (TS-DMH) intervention examined the associations between MMB and mental health. Results: Results indicated that more MMB was significantly associated with greater presenting mental health symptom severity. MMB did not have a deleterious influence on depressive symptom trajectories across treatment, although having one medical condition was associated with a steeper decrease in anxiety symptoms compared to patients with no medical conditions. Finally, MMB was not associated with time to dropout, but was associated with higher dropout and was differentially associated with fewer beneficial treatment outcomes, although this is likely attributable to higher presenting symptom severity, rather than lesser symptom reductions during treatment. Conclusions: The MHP showed preliminary effectiveness as an evidence-based treatment for patients experiencing depression and anxiety in the context of MMB. Future DMH treatments and research might investigate tailored barrier reduction and extended treatment lengths for patients experiencing MMB to allow for greater treatment dose to reduce symptoms below clinical outcome thresholds.

  • Open Access

    Adversity in early life and pregnancy are immunologically distinct from total life adversity: macrophage-associated phenotypes in women exposed to interpersonal violence

    by Kirstin Aschbacher, Melissa Hagan, Iris M. Steine, Luisa Rivera, Steve Cole, Alyssa Baccarella, Elissa S. Epel, Alicia Lieberman, Nicole R. Bush

    Early childhood and pregnancy are two sensitive periods of heightened immune plasticity, when exposure to adversity may disproportionately increase health risks. However, we need deeper phenotyping to disentangle the impact of adversity during sensitive periods from that across the total lifespan. This study examined whether retrospective reports of adversity during childhood or pregnancy were associated with inflammatory imbalance, in an ethnically diverse cohort of 53 low-income women seeking family-based trauma treatment following exposure to interpersonal violence. Structured interviews assessed early life adversity (trauma exposure ≤ age 5), pregnancy adversity, and total lifetime adversity. Blood serum was assayed for pro-inflammatory (TNF-a, IL-1ß, IL-6, and CRP) and anti-inflammatory (IL-1RA, IL-4, and IL-10) cytokines. CD14+ monocytes were isolated in a subsample (n = 42) and gene expression assayed by RNA sequencing (Illumina HiSeq 4000; TruSeq cDNA library). The primary outcome was a macrophage-associated M1/M2 gene expression phenotype. To evaluate sensitivity and specificity, we contrasted M1/M2 gene expression with a second, clinically-validated macrophage-associated immunosuppressive phenotype (endotoxin tolerance) and with pro-inflammatory and anti-inflammatory cytokine levels. Adjusting for demographics, socioeconomic status, and psychopathology, higher adversity in early life (ß = .337, p = 0.029) and pregnancy (ß = .332, p = 0.032) were each associated with higher M1/M2 gene expression, whereas higher lifetime adversity (ß = −.341, p = 0.031) was associated with lower immunosuppressive gene expression. Adversity during sensitive periods was uniquely associated with M1/M2 imbalance, among low-income women with interpersonal violence exposure. Given that M1/M2 imbalance is found in sepsis, severe COVID-19 and myriad chronic diseases, these findings implicate novel immune mechanisms underlying the impact of adversity on health.


  • Open Access

    Machine learning prediction of blood alcohol concentration: a digital signature of smart-breathalyzer behavior

    by Kirstin Aschbacher, Christian S. Hendershot, Geoffrey Tison, Judith A. Hahn, Robert Avram, Jeffrey E. Olgin, Gregory M. Marcus

    Excess alcohol use is an important determinant of death and disability. Machine learning (ML)-driven interventions leveraging smart-breathalyzer data may help reduce these harms. We developed a digital phenotype of long-term smart-breathalyzer behavior to predict individuals’ breath alcohol concentration (BrAC) levels trained on data from a smart breathalyzer. We analyzed roughly one million datapoints from 33,452 users of a commercial smart-breathalyzer device, collected between 2013 and 2017. For validation, we analyzed the associations between state-level observed smart-breathalyzer BrAC levels and impaired-driving motor vehicle death rates. Behavioral, geolocation-based, and time-series-derived features were fed to an ML algorithm using training (70% of the cohort), development (10% of the cohort), and test (20% of the cohort) sets to predict the likelihood of a BrAC exceeding the legal driving limit (0.08 g/dL). States with higher average BrAC levels had significantly higher alcohol-related driving death rates, adjusted for the number of users per state B (SE) = 91.38 (15.16), p < 0.01. In the independent test set, the ML algorithm predicted the likelihood of a given user-initiated BrAC sample exceeding BrAC ≥ 0.08 g/dL, with an area under the curve (AUC) of 85%. Highly predictive features included users’ prior BrAC trends, subjective estimation of their BrAC (or AUC = 82% without the self-estimate), engagement and self-monitoring, time since the last measure, and hour of the day. In conclusion, an ML algorithm successfully quantified a digital phenotype of behavior, predicting naturalistic BrAC levels exceeding 0.08 g/dL (a threshold associated with alcohol-related harm) with good discrimination capability. This result establishes a foundation for future research on precision behavioral medicine digital health interventions using smart breathalyzers and passive monitoring approaches.


  • Open Access

    Comparison of the Physical Activity Measured by a Consumer Wearable Activity Tracker and That Measured by Self-Report: Cross-Sectional Analysis of the Health eHeart Study

    by Alexander J Beagle, Geoffrey H Tison, Kirstin Aschbacher, Jeffrey E Olgin, Gregory M Marcus, Mark J Pletcher

    Background: Commercially acquired wearable activity trackers such as the Fitbit provide objective, accurate measurements of physically active time and step counts, but it is unclear whether these measurements are more clinically meaningful than self-reported physical activity. Objective: The aim of this study was to compare self-reported physical activity to Fitbit-measured step counts and then determine which is a stronger predictor of BMI by using data collected over the same period reflecting comparable physical activities. Methods: We performed a cross-sectional analysis of data collected by the Health eHeart Study, a large mobile health study of cardiovascular health and disease. Adults who linked commercially acquired Fitbits used in free-living conditions with the Health eHeart Study and completed an International Physical Activity Questionnaire (IPAQ) between 2013 and 2019 were enrolled (N=1498). Fitbit step counts were used to quantify time by activity intensity in a manner comparable to the IPAQ classifications of total active time and time spent being sedentary, walking, or doing moderate activities or vigorous activities. Fitbit steps per day were computed as a measure of the overall activity for exploratory comparisons with IPAQ-measured overall activity (metabolic equivalent of task [MET]-h/wk). Measurements of physical activity were directly compared by Spearman rank correlation. Strengths of associations with BMI for Fitbit versus IPAQ measurements were compared using multivariable robust regression in the subset of participants with BMI and covariates measured. Results: Correlations between synchronous paired measurements from Fitbits and the IPAQ ranged in strength from weak to moderate (0.09-0.48). In the subset with BMI and covariates measured (n=586), Fitbit-derived predictors were generally stronger predictors of BMI than self-reported predictors. For example, an additional hour of Fitbit-measured vigorous activity per week was associated with nearly a full point reduction in BMI (–0.84 kg/m2, 95% CI –1.35 to –0.32) in adjusted analyses, whereas the association between self-reported vigorous activity measured by IPAQ and BMI was substantially smaller in magnitude (–0.17 kg/m2, 95% CI –0.34 to –0.00; P<.001 versus Fitbit) and was dominated by the Fitbit-derived predictor when compared head-to-head in a single adjusted multivariable model. Similar patterns of associations with BMI, with Fitbit dominating self-report, were seen for moderate activity and total active time and in comparisons between overall Fitbit steps per day and IPAQ MET-h/wk on standardized scales. Conclusions: Fitbit-measured physical activity was more strongly associated with BMI than self-reported physical activity, particularly for moderate activity, vigorous activity, and summary measures of total activity. Consumer-marketed wearable activity trackers such as the Fitbit may be useful for measuring health-relevant physical activity in clinical practice and research.


  • A digital biomarker of diabetes from smartphone-based vascular signals

    by Robert Avram, Jeffrey E. Olgin, Peter Kuhar, J. Weston Hughes, Gregory M. Marcus, Mark J. Pletcher, Kirstin Aschbacher, Geoffrey H. Tison

    The global burden of diabetes is rapidly increasing, from 451 million people in 2019 to 693 million by 20451. The insidious onset of type 2 diabetes delays diagnosis and increases morbidity2. Given the multifactorial vascular effects of diabetes, we hypothesized that smartphone-based photoplethysmography could provide a widely accessible digital biomarker for diabetes. Here we developed a deep neural network (DNN) to detect prevalent diabetes using smartphone-based photoplethysmography from an initial cohort of 53,870 individuals (the ‘primary cohort’), which we then validated in a separate cohort of 7,806 individuals (the ‘contemporary cohort’) and a cohort of 181 prospectively enrolled individuals from three clinics (the ‘clinic cohort’). The DNN achieved an area under the curve for prevalent diabetes of 0.766 in the primary cohort (95% confidence interval: 0.750–0.782; sensitivity 75%, specificity 65%) and 0.740 in the contemporary cohort (95% confidence interval: 0.723–0.758; sensitivity 81%, specificity 54%). When the output of the DNN, called the DNN score, was included in a regression analysis alongside age, gender, race/ethnicity and body mass index, the area under the curve was 0.830 and the DNN score remained independently predictive of diabetes. The performance of the DNN in the clinic cohort was similar to that in other validation datasets. There was a significant and positive association between the continuous DNN score and hemoglobin A1c (P ≤ 0.001) among those with hemoglobin A1c data. These findings demonstrate that smartphone-based photoplethysmography provides a readily attainable, non-invasive digital biomarker of prevalent diabetes.


  • Open Access

    Comparison of the Physical Activity Measured by a Consumer Wearable Activity Tracker and That Measured by Self-Report: Cross-Sectional Analysis of the Health eHeart Study (Preprint)

    by Alexander J Beagle, Geoffrey H Tison, Kirstin Aschbacher, Jeffrey E Olgin, Gregory M Marcus, Mark J Pletcher

    BACKGROUND Commercially acquired wearable activity trackers such as the Fitbit provide objective, accurate measurements of physically active time and step counts, but it is unclear whether these measurements are more clinically meaningful than self-reported physical activity. OBJECTIVE The aim of this study was to compare self-reported physical activity to Fitbit-measured step counts and then determine which is a stronger predictor of BMI by using data collected over the same period reflecting comparable physical activities. METHODS We performed a cross-sectional analysis of data collected by the Health eHeart Study, a large mobile health study of cardiovascular health and disease. Adults who linked commercially acquired Fitbits used in free-living conditions with the Health eHeart Study and completed an International Physical Activity Questionnaire (IPAQ) between 2013 and 2019 were enrolled (N=1498). Fitbit step counts were used to quantify time by activity intensity in a manner comparable to the IPAQ classifications of total active time and time spent being sedentary, walking, or doing moderate activities or vigorous activities. Fitbit steps per day were computed as a measure of the overall activity for exploratory comparisons with IPAQ-measured overall activity (metabolic equivalent of task [MET]-h/wk). Measurements of physical activity were directly compared by Spearman rank correlation. Strengths of associations with BMI for Fitbit versus IPAQ measurements were compared using multivariable robust regression in the subset of participants with BMI and covariates measured. RESULTS Correlations between synchronous paired measurements from Fitbits and the IPAQ ranged in strength from weak to moderate (0.09-0.48). In the subset with BMI and covariates measured (n=586), Fitbit-derived predictors were generally stronger predictors of BMI than self-reported predictors. For example, an additional hour of Fitbit-measured vigorous activity per week was associated with nearly a full point reduction in BMI (–0.84 kg/m2, 95% CI –1.35 to –0.32) in adjusted analyses, whereas the association between self-reported vigorous activity measured by IPAQ and BMI was substantially smaller in magnitude (–0.17 kg/m2, 95% CI –0.34 to –0.00; P<.001 versus Fitbit) and was dominated by the Fitbit-derived predictor when compared head-to-head in a single adjusted multivariable model. Similar patterns of associations with BMI, with Fitbit dominating self-report, were seen for moderate activity and total active time and in comparisons between overall Fitbit steps per day and IPAQ MET-h/wk on standardized scales. CONCLUSIONS Fitbit-measured physical activity was more strongly associated with BMI than self-reported physical activity, particularly for moderate activity, vigorous activity, and summary measures of total activity. Consumer-marketed wearable activity trackers such as the Fitbit may be useful for measuring health-relevant physical activity in clinical practice and research.