Published on in Vol 5, No 10 (2021): October

Preprints (earlier versions) of this paper are available at, first published .
Impact of App-Delivered Mindfulness Meditation on Functional Connectivity, Mental Health, and Sleep Disturbances Among Physician Assistant Students: Randomized, Wait-list Controlled Pilot Study

Impact of App-Delivered Mindfulness Meditation on Functional Connectivity, Mental Health, and Sleep Disturbances Among Physician Assistant Students: Randomized, Wait-list Controlled Pilot Study

Impact of App-Delivered Mindfulness Meditation on Functional Connectivity, Mental Health, and Sleep Disturbances Among Physician Assistant Students: Randomized, Wait-list Controlled Pilot Study

Original Paper

1Department of Radiology and Imaging Sciences, Emory University, Atlanta, GA, United States

2Department of Neurology, Emory University, Atlanta, GA, United States

3Department of Surgery, Emory University, Atlanta, GA, United States

4Department of Family and Preventative Medicine, Emory University, Atlanta, GA, United States

5Physician Assistant Program, Department of Family and Preventative Medicine, Emory University, Atlanta, GA, United States

Corresponding Author:

Jeremy L Smith, MSDS, PhD

Department of Radiology and Imaging Sciences

Emory University

1364 Clifton Road Northeast

Atlanta, GA, 30322

United States

Phone: 1 404 989 0524


Background: Health care provider and trainee burnout results in substantial national and institutional costs and profound social effects. Identifying effective solutions and interventions to cultivate resilience among health care trainees is critical. Although less is known about the mental health needs of physician assistants (PAs) or PA students, accumulating research indicates that they experience similarly alarming rates of burnout, depression, and emotional exhaustion. Mobile app–delivered mindfulness meditation may be an effective part of salubrious programming to bolster long-term resilience and health among PA students.

Objective: This study aims to examine the impact of app-delivered mindfulness meditation on self-reported mental health symptoms among PA students. A secondary aim is to investigate changes in brain connectivity to identify neurobiological changes related to changes in mental health symptoms.

Methods: We recruited PA students enrolled in their third semester of PA school and used a longitudinal, randomized, wait-list–controlled design. Participants randomized to the mindfulness group were provided 1-year subscriptions to the 10% Happier app, a consumer-based meditation app, and asked to practice every day for 8 weeks. Before randomization and again after completion of the 8-week program, all participants completed resting-state functional magnetic resonance imaging as well as self-report assessments of burnout, depression, anxiety, and sleep impairment. App use was acquired as a measure of mindfulness practice time.

Results: PA students randomized to the mindfulness group reported improvements in sleep impairment compared with those randomized to the wait-list control group (ηp2=0.42; P=.01). Sleep impairment decreased significantly in the mindfulness group (19% reduction; P=.006) but not in the control group (1% reduction; P=.71). There were no other significant changes in mental health for those randomized to app-delivered mindfulness. Across all students, changes in sleep impairment were associated with increased resting-state functional connectivity between the medial prefrontal cortex (a component of the default mode network) and the superior temporal gyrus, as well as between areas important for working memory. Changes in connectivity predicted categorical conversion from impaired to nonimpaired sleep in the mindfulness group.

Conclusions: This pilot study is the first to examine app-based mindfulness for PA students’ mental health and investigate the impact of mindfulness on PA students’ brain function. These findings suggest that app-delivered mindfulness may be an effective tool to improve sleep dysfunction and that it may be an important part of the programming necessary to reduce the epidemic of suffering among health profession trainees.

JMIR Form Res 2021;5(10):e24208




Although the high prevalence of depression among medical students and residents is well characterized and increasingly appreciated [1,2], little is known about the mental health of physician assistant (PA) students [3,4]. The research that has been conducted indicates that practicing PAs report high levels of burnout [5] and depression [6], and this is despite the fact that PAs often report high levels of job satisfaction [6]. Although very few studies have evaluated burnout or depression among PA students, one recent study found that almost 80% of PA students report high levels of emotional exhaustion, with almost that many expressing interest in interventions to improve their well-being [7]. Overall, there is a critical need to examine the mental health needs of PA students and evaluate interventions to bolster PA student resilience.

A substantial and growing body of research indicates that mindfulness meditation enhances well-being [8]; reduces anxiety and depression [9]; and optimizes immune signaling, stress responsivity, and cognitive function [10-12]. Moreover, mindfulness meditation has shown great promise for improving sleep disruption and insomnia symptoms [13-15], which may be beneficial to health professional trainees who often experience high rates of sleep dysfunction [16]. With its apparent efficacy, several clinical trials have examined whether mindfulness training improves well-being among health care trainees, including medical and nursing students and medical residents [17]. Accumulating research indicates that mindfulness reduces anxiety and depression and enhances well-being among health profession trainees, and a recent meta-analysis highlighted the potential efficacy of mind-body wellness programs such as mindfulness meditation for improving trainee well-being [18,19]. However, most studies examining mindfulness among trainees have examined time-intensive interventions that are prohibitive for most trainees. Although a recent study found that incorporating mindfulness into PA coursework increased self-reported well-being [20], very few studies have examined the impact of mindfulness training on the mental health of PA students.

Even fewer studies have examined the impact of mindfulness on trainee brain function. Growing evidence indicates that the health-relevant effects of mindfulness are mediated by alterations to the default mode network (DMN), the salience network (SN), and the systems involved in executive control, often referred to as the central executive network [21,22]. In addition, at least 3 studies indicate that mindfulness training [23,24] or dispositional mindfulness [25] are related to increased functional connectivity between the prefrontal cortex and the amygdala, generally interpreted as augmented emotion regulation via top-down control of the amygdala. Although these studies indicate that benefits from mindfulness meditation are related to changes in functional connectivity within and among these brain regions, to date, no studies have examined changes in functional connectivity related to improvements in well-being among health profession trainees.


Here, we use a longitudinal, randomized wait-list–controlled design to examine the impact of app-delivered mindfulness meditation on self-reported mental health symptoms among PA students. A second aim is to examine whether changes in mental health are associated with changes in brain connectivity, indexed using a whole-brain connectome and multivariate pattern analysis approach to query within- and between-network connectivity across the entire brain before and after mindfulness training. On the basis of prior studies, we hypothesize that app-delivered mindfulness would reduce burnout, depression, anxiety, and sleep impairment; that changes in mental health outcomes would be associated with changes in DMN, SN, and central executive network connectivity; and that changes in mental health and connectivity would be positively associated with practice time, indexed as app use.


Students enrolled in their third semester of PA school were recruited for the study just before the commencement of their clinical rotations. Students were recruited via in-person presentations held after their classes, and 16 students were enrolled in the study. Upon providing informed consent in accordance with the university’s institutional review board, participants were randomly assigned (in Microsoft Excel, using the randbetween function) to either a mindfulness meditation intervention using the 10% Happier app (practitioners group) or to a wait-list (control group). Study personnel were blinded to group randomization, except for 1 researcher who corresponded with the participants and was not involved in data collection or analysis.

PA students randomized to the practitioner group were asked to practice app-guided mindfulness meditation for approximately 12 minutes per day for 8 weeks. Before randomization and again after completion of the 8-week program, all participants completed self-report assessments and functional magnetic resonance imaging (fMRI), both described in detail below. A total of 2 participants were removed from the analysis for excessive movement during fMRI, resulting in a cohort of 50% (7/14) practitioners (6/7, 86% women) and 50% (7/14) controls (5/7, 71% women; Table 1). Participants were compensated US $100 for completing both assessments. This study was part of a larger preregistered clinical trial (NCT03452670) that included the planned enrollment of several trainee and employee populations. For the larger trial, self-reported perceived incivility was the primary outcome measure and burnout, depression, and anxiety were secondary outcome measures. Enrollment for the larger study did not meet the planned recruitment goals, and this was an exploratory analysis of a subset of the enrolled participants. In addition to the secondary outcome measures, we included a measure of sleep impairment given the prevalence of sleep dysfunction among health profession trainees and the relationship between sleep dysfunction and depression [26].

Table 1. Demographic information (N=14).
CharacteristicsApp, n (%; n=7)Wait-list, n (%; n=7)Chi-square (df)P value

Female6 (86)5 (71)0.4 (1, 14).52

Male1 (14)2 (29)0.4 (1, 14).52
Relationship status

Single1 (14)3 (43)5.7 (3, 14).23

Divorced0 (0)1 (14)5.7 (3, 14).23

In a relationship4 (57)2 (29)5.7 (3, 14).23

Married2 (29)1 (14)5.7 (3, 14).23

White6 (86)5 (71)3.0 (3, 14).21

African American or Black individual1 (14)03.0 (3, 14).21

Asian0 (0)2 (29)3.0 (3, 14).21

Other0 (0)0 (0)3.0 (3, 14).21

Mindfulness Meditation Intervention

Participants randomized to the meditation group were provided with a 1-year subscription to the app 10% Happier [27,28]. We chose 10% Happier among a variety of meditation apps currently in the market given its marketing toward fidgety skeptics and its orientation toward practical applications of mindfulness [27], which we thought would appeal to PA students. For example, the app notes, “Just in case you’re worried, meditation does not require a lot of the things people think it might. For example, you don’t have to sit in a particular position. (Unless you want to, of course). You also don’t have to light incense, chant, or believe in anything in particular. There’s nothing to join, no special outfits to wear” [29]. 10% Happier provided subscriptions for the study participants, suggestions on recommended content, and anonymized app use data. App use was acquired as the elapsed time (in seconds) that the app meditation modules were used by each person.

Students randomized to the meditation group were given the following instructions:

We would like for you to try to practice every day for 8 weeks, even if it is only for one minute. Based on mindfulness research and on the suggestions of the app developers, we would like you to try the following programs: “The Basics” and “Emotional Agility.” If you are pressed for time and cannot do a module from these programs, please do the “One minute counts.”

The Basics program included 16 modules with didactic instructions and mindfulness practice time varying between 4:20 to 13:22 minutes (average meditation length: 9:44 minutes, SD 3:17 minutes), and it serves as an introduction to mindfulness meditation. The practices include mindfulness of the sensations of the breath, mindfulness toward sensations and experiences of the body, and mindfulness toward the contents of the mind. In addition, the meditations encourage the practitioner to use the skill of mental noting to label their mental contents. The Emotional Agility program included 15 modules with didactic instructions and mindfulness toward mental content (focusing on emotions); meditation practice in the Emotional Agility program varied from 11:00 to 13:00 minutes (average meditation length: 12:18 minutes, SD 0:40 minutes). The meditation practices in these modules included mindfulness toward the sensations of the body and breath, coupled with other practices aimed at cultivating awareness and understanding of emotions as mental contents and the nonjudgmental stance toward emotions, with a goal of optimizing the response to one’s emotions.

Self-reported Measures

We measured incivility using the Incivility in Nursing Education—Revised Survey [30], which contains 24 items that ask students about behaviors they exhibited or witnessed in the past 12 months (eg, “students made rude gestures or nonverbal behaviors towards others”). Participants indicated how often these behaviors occurred by selecting from 1=never, 2=rarely (once or twice), 3=sometimes (approximately once per month), or 4=often (more than once per month). Items were summed according to instances of low-level (15 items; eg, “Expressing disinterest, boredom, or apathy about course content or subject matter”) and high-level incivility (9 items; eg, “Making condescending or rude remarks toward others”). Scores were averaged, such that the range was 1-4, with higher scores indicating more incivility exhibited or witnessed.

We measured burnout using the School Burnout Inventory [31], a 9-item survey asking students about how much burnout they have felt in the past month (eg, “I feel overwhelmed by my schoolwork”). Respondents indicated the degree to which they agreed with each statement on a scale of 1-6, where 1=completely disagree and 6=completely agree. Total scores ranged from 9-54, with higher scores indicating more burnout.

The Depression Anxiety and Stress Scale [32] is a 42-item survey asking about feelings of depression, anxiety, and stress that the respondent has experienced in the past week. Participants indicated the degree to which they agreed with each statement on a scale of 0-3, where 0=does not apply to me at all and 3=applied to me very much or most of the time. Each of the three subscales included 14 items. The depression subscale assessed general dysphoria, anhedonia, self-contempt, and hopelessness. A score of 0-9 indicated no depression, 10-13 indicated mild depression, 14-20 indicated moderate depression, 21-27 indicated severe depression, and scores ≥28 indicated extremely severe depression. The anxiety subscale assessed symptoms and subjective feelings related to acute autonomic arousal. A score of 0-7 indicated normal levels of anxiety, 8-9 indicates mild anxiety, 10-14 indicated moderate anxiety, 15-19 indicated severe anxiety, and scores ≥20 indicated extremely severe anxiety.

Sleep dysfunction was measured using the 8-item PROMIS (Patient-Reported Outcomes Measurement Information System) sleep-related impairment (SI) short form 8b, which assessed the frequency with which participants experienced alertness, sleepiness, tiredness, and functional impairments associated with sleep problems during waking hours (eg, “I had difficulty falling asleep”) in the previous 7 days [33]. Items were scored on a 5-point scale, such that higher scores indicated more sleep impairment. Raw scores for each of the eight items were totaled and converted to standardized scores using conversion tables published on the PROMIS website [34].

To assess safety, participants were asked to report their agreement with a number of statements reflecting positive (eg, “I enjoyed using the app”) and negative experiences (eg, “I found it very difficult to do the meditations”) with the app. In addition, they were asked to report anything else good, bad, or neutral that they wanted us to know about the app.

Resting fMRI Image Preprocessing

Baseline and postprogram resting-state fMRI (rsfMRI) data were acquired on a 3T Siemens Prisma FIT (Siemens Healthineers; 8-minute multiband acquisition with 2 seconds repetition time, 2.97×2.97×2.00 mm voxels, 70° flip angle, and echo train length 37). All preprocessing and bivariate correlation (connectome) analyses were performed in the CONN Toolbox (v19c) under MATLAB (vR2019a) [35]. Standard preprocessing methods were applied to the rsfMRI and anatomical volumes in CONN, which wraps SPM8 [36,37] and aCompCor [38] noise source removal functions. It comprised slice timing, field map, and motion correction; coregistration and normalization between rsfMRI images, anatomical images, and Montreal Neurological Institute, standard stereotactic space; smoothing at a 5 mm filter width at half-maximum, which limits intersubject differences and increases signal-to-noise ratio), linear detrending, bandpass filtering at 8-90 mHz; and regression of the 6 motion parameters and their first-order derivatives, along with cerebrospinal fluid and white-matter signals [39-41], by a general linear model (GLM). Scans (repetition time intervals [TRs]) that exhibited motion or global signal change beyond a SD 1.5 IQR limit were marked as invalid scans and nulled for the purposes of the GLM (see the Quality Assurance and Quality Control document, Multimedia Appendix 1). Three subjects exhibited 1-8 TRs, out of 240 total TRs, with global signal change beyond this tolerance. These TRs were tagged as invalid and were not included in further analyses. We did not add the mean whole-brain signal as a regressor, as there is some evidence that doing so may artificially introduce negative correlations and that the aCompCor method, in combination with bandpass filtering and orthogonalization of motion parameters, is preferable to global signal regression [42]. All structural and denoized functional data, gray matter, white matter, and CSF masks were manually inspected to confirm registration validity. In addition, a Fisher (inverse hyperbolic tangent) transformation was applied to bivariate correlation measures before between-subjects analysis to ensure that the connectivity data conformed to the normality assumptions of the GLM (see the Quality Assurance and Quality Control document in the Multimedia Appendix 1).

Identification of Regions of Interest and Computation of Connectivity Matrices

Regions of interest (ROIs) were computed from preprocessed rsfMRI data using multivariate pattern analysis (Norman et al [43]). The first step of the multivariate pattern analysis procedure, as implemented [31] in the CONN Toolbox, comprised dimensionality reduction into 64 components by singular value decomposition for each subject and condition (group and visit). Subject and condition-specific correlation maps—of size (number of subjects × number of conditions × number of voxels in each data set) and comprising standardized (Fisher transformed) bivariate correlation coefficients between each pair of voxels—were then generated from each subject’s reduced-dimensionality data set. Next, a set of four principal components, capturing approximately 95% of the between-subjects or between-conditions variance, was obtained from the aggregate matrix (ie, with all subjects and conditions combined or stacked). Finally, the first two of these four principal components were selected for further analysis and subjected to a standard two-way analysis of variance (ANOVA; group × visit) to determine whether the components modulated with either condition. Application of a statistical threshold based on Gaussian random field theory [44-46], which estimates error fields within an fMRI statistical map after smoothing with the FWHM kernel (described above), yielded a map of 87 clusters that ostensibly represented any between-group or between-visit differences in voxel-to-voxel functional connectivity across the brain with a cluster growth threshold of P≤.05 (uncorrected) and a topological false discovery rate (FDR [47]) of PFDR≤.05. A total of 16 of these clusters were excluded because of their localization in white matter. Another 15 ROIs representing components of the default mode, sensorimotor, visual, salience, dorsal attention, frontoparietal, language, and cerebellar networks, predefined in the CONN Toolbox, were also included, for a total of 102 ROIs (Table 2 and Figure S1 in Multimedia Appendix 2).

To facilitate testing of our hypothesized relationships among meditation practice time, changes in mental health outcomes, and brain network connectivity, mean signals were extracted from the 71 a posteriori and 15 a priori ROIs for each subject and condition and aggregated into baseline (visit 1) and eighth-week (visit 2 follow-up) ROI-to-ROI connectivity matrices, each of size (number of subjects × number of ROIs × number of ROIs). As with the voxel-wise correlation maps, these ROI-ROI connectivity matrices comprised standardized (Fisher transformed) bivariate correlation coefficients between each pair of ROIs. To simplify the within-group analysis, a delta matrix, Δconn, was also computed as Δconn=conn8wks–connbaseline. The baseline, eighth-week, and delta ROI-ROI connectivity matrices (Figure S2 in Multimedia Appendix 2) were leveraged in lieu of voxel-to-voxel connectivity matrices for all further analyses.

Table 2. List of the 71 gray matter regions of interest derived from multivariate pattern analysis, with predefined atlas-based regions of interest. Coordinates for region of interest centers of mass and peak voxels (where applicable) are provided in MNI152 standard coordinate space.
Type and CLUSTER_IDVoxelsCenter of mass (MNIa)Peak (MNI)ATLAS ROIb

Brain stem

bs1_bstem33–1.2–29.3–45.60–30–46Brainstem gray matter
Basal ganglia

caud1_Rput12620.127.74.824184R putamen/globus pallidus

caud2_Rput110829.37.11.83446R middle insula/putamen

caud3_Rgp4522.3–9.4–3.122–14–4R globus pallidus/thalamus/motor thalamus

caud4_Rput3631.3–18.71.932–160R putamen

cb1_Rlob93644.2–61.5–45.712–62–54R cerebellum (VIII) lobule IX

cb2_Llob6269–21.7–49.8–46.7–28–48–38L cerebellum (VI)/prob. WM

cb3_Rlob8b9916.4–46.9–50.318–46–50R cerebellum (IX) lobule VIIIb

cb4_Rlob7b8232–74.9–54.128–76–54R cerebellum (VII) lobule VIIb

cb5_Rcrus25210.5–72.2–24.76–72–28R cerebellar crus 2/vermis

cb6_Lcrus239–48.7–56.8–42.6–56–56–44L cerebellar crus 2/lobule VIIa crus I

cing1_Rpcc42611–47.420.810–4818R precuneus/posterior cingulate/BA23v

cing2_Rmicg2118.3–33.349.818–3450R middle cingulate cortex/ventromedial BA5

f1_Lba9408–3.148.345–125442L superior frontal gyrus/mid BA9

f10_Lba6r33–50.94.316.5–52416L precentral gyrus/BA44/rostral BA6/caudoventrolateral BA6

f11_Rba6ba82818.931.760.1203262R superior frontal gyrus/superior BA6-BA8 transitional area

f12_Lba6ba827–31.1459.8–30460L middle frontal gyrus/inferior BA6-BA8 transitional area

f2_Rba10p27118.157.5–13.324622R posterior BA10/superior frontal gyrus

f3_Rba8v20747.610.644.5481248R ventral BA8A/caudal middle frontal gyrus/IFJ

f4_Rba8dl16328.718.549.2282054R middle frontal gyrus/dorsolateral BA8A

f5_Lba8dl133–2030.550.2–203048L dorsolateral BA8A/superior frontal gyrus

f6_Lsma124–15.27.953.7–101250L SMA/supplementary and cingulate eye area/medial frontal gyrus

f7_Rba24d901.9–24.646.12–2250R dorsal BA24/SMA/superiomedial BA4

f8_Rba10d88–17.562.67.9–226010L superior frontal gyrus/dorsal BA10/BA46/area Fp1

f9_Lba44v81–53.417.92.8–54184L ventral BA44/inferior frontal gyrus pars triangularis
Medial temporal

mtl1_Lphg1137–15.7–46.5–10.4–16–42–10L parahippocampal area 1/subiculum/lingual gyrus/area TH

mtl2_Rphg8927–35.9–13.428–44–6R parahippocampal gyrus/rostral lingual gyrus/ventromedial visual area 1

mtl3_Renrc8321.2–26.6–28.522–28–26R entorhinal cortex/presubiculum/caudal BA35/36

mtl4_Rprc5029.6–1.1–34.430–2–34R perirhinal/ectorhinal cortex/rostral BA36/BA35/parahippocampal gyrus

mtl5_Lamg34–25.4–4.1–21–24–4–16L piriform cortex/amygdala/laterobasal amygdala

mtl6_Lenrc33–19.2–9.4–33.7–22–8–34L entorhinal cortex/BA28/BA34/parahippocampal gyrus

mtl7_Rhc2330.3–16.2–1830–16–14R hippocampus/CA3 fields/CA3 fields

o1_Rba39rd28734.7–69.841.636–7238R intraparietal area1/rostro-dorsal BA39/middle occipital gyrus

o2_Llop208–16.2–101.6–2.1–14–106–8L lateral occipital pole

o3_Rv2v3d2923.5–98.910.724–9810R superior occipital gyrus/visual area V2/dorsal visual area V3

o4_Lv225–24.9–50.50–24–500L precuneus/prostriate area/lingual gyrus/visual area V2

o5_Rba37mv2022.3–42.3–17.222–44–18R ventral visual complex/area FG3/medioventral BA37

op1_Rro6053.92.42.95446R rolandic operculum/rostral BA6/BA43

orb1_Rba1120818.718.5–22.82030–16R superior orbital gyrus/lateral BA11/area Fo3

orb2_Lba10r14m94–5.341.3–12.7–644–16L rostral BA10/medial BA14/gyrus rectus

orb3_Rifgpo8932.222.9–31.43222–26R inferior frontal gyrus pars orbitalis/superior BA47/lateral BA11/area Fo3

orb4_L47l76–49.529.6–5.9–4628–6L lateral BA47/inferior frontal gyrus pars orbitalis

orb5_Rba10v494.259.6–11.3660–10R ventral BA10/medial BA11/middle orbital gyrus

orb6_Lpofc36–1414.1–14.4–1414–14L gyrus rectus/area Fo2/posterior orbitofrontal complex/subcallosal gyrus

orb7_Lba11l34–20.132.1–18.8–1834–22L superior orbital gyrus/area Fo3/lateral BA11

orb8_Rofc252.116.1–21.7216–22R orbitofrontal cortex/area Fo2

orb9_Rba14m204.628.5–13.2430–14R middle orbital gyrus/area s32/area 25/medial BA14

pfc1_Lba9pv50–37.533.822–402826L posterior ventral BA9/BA46

pfc2_Lba9av40––38540L anteroventral BA9/46/rostral BA47

pl1_Lipl280–35–67.625.1–36–6824L middle occipital gyrus/inferior parietal lobule/rostro-ventral BA39

pl2_Lba7m233–2.9–53.558.1–6–5266L medial BA7A/medial BA5

pl3_Lba7m5ml19044.620.410.346166L medial BA7A/mediolateral BA5

pl4_Lpos2114–4.2–65.942.6–2–7044L parieto-occipital sulcus area 2/medial BA7/precuneus

pl5_Rpos23216.9–59.528.220–6026R precuneus/parieto-occipital sulcus area 2

t1_Lffc95–46.6–56.9–18.9–44–60–16L fusiform gyrus/fusiform face complex/lateroventral BA37/area FG2

t10_Ravsts3156.3–7.2–21.656–6–20R anteroventral superior temporal sulcus/middle temporal gyrus

t11_Laud531–61.5–0.3–8.7–64–4–4L superior temporal gyrus/area TE3/auditory 5 complex

t12_Rba38l2847.221.4–26.34822–26R temporopolar cortex/lateral BA38/dorsal area TG

t13_Lpiri6–41.76–10.7–404–12CLUSTER5/L piriform cortex/insula/posterior insula area 2

t14_Lba37vl3–48.7–64–3.3–48–64–4CLUSTER10/middle temporal gyrus/area PH/ventrolateral BA37

t2_Lba38l68–35.34–23.2–386–20L dorsal area TG/lateral BA38/temporopolar cortex

t3_Rba226466.9–17.3066–140R auditory 4 complex/area TE3/caudal BA22

t4_Lba20il62–54.8–11.9–34.2–56–12–32L anterior area TE2/inferolateral BA20/inferior temporal gyrus

t5_Lba22c60–65.2–46.711.8–68–4812L superior temporal visual area/caudal BA22

t6_Rins4534.2–16.723.434–1422R insula/area OP2-3/VS

t7_Lpins237–35.4–6.4–6.5–36–8–4L posterior insular area 2/circular insula

t8_Lba38m36–45.824–25.4–4426–26L temporopolar cortex/dorsal area TG/medial BA38

t9_Raud53365.1–38.65.566–386R middle temporal gyrus/auditory 5 complex/ventral superior temporal sulcus

th1_Lpfthal82–5.7–14.4–3.5–6–14–2L thalamus/thalamic area IPF/prefrontal-directed thalamus

th2_Rpmthal7215.5–18.64.916–186R thalamus/premotor-directed thalamus

medial PFCc (DMNd)e155–3N/AfN/AN/Amedial prefrontal cortex, default mode network

L lateral parietal (DMN)–39–7733N/AN/AN/AL lateral parietal cortex/von Economo PG, default mode network

R lateral parietal (DMN)47–6729N/AN/AN/AR lateral parietal cortex/von Economo PG, default mode network

posterior cingulate (DMN)1–6138N/AN/AN/Aposterior cingulate cortex, default mode network

L lateral sensorimotor–55–1229N/AN/AN/AL lateral sensorimotor cortex

R lateral sensorimotor56–1029N/AN/AN/AR lateral sensorimotor cortex

superior sensorimotor0–3167N/AN/AN/Asuperior (mesial) sensorimotor cortex

medial visual2–7912N/AN/AN/Amedial visual cortex/Brodmann 18/calcarine gyrus/visual area V1

occipitopolar visual0–93–4N/AN/AN/Aoccipitopolar visual cortex/Brodmann 17/calcarine gyrus

L lateral visual–37–7910N/AN/AN/AL lateral visual cortex/visual area V4

R lateral visual38–7213N/AN/AN/AR lateral visual cortex/visual area V4

anterior cingulate (SNg)02235N/AN/AN/Aanterior cingulate cortex, salience network

L anterior insula (SN)–44131N/AN/AN/AL anterior insula, salience network

R anterior insula (SN)47140N/AN/AN/AR anterior insula, salience network

L rostral PFC (SN)–324527N/AN/AN/AL rostral prefrontal cortex, salience network

R rostral PFC (SN)324627N/AN/AN/AR rostral prefrontal cortex, salience network

L supramarginal g. (SN)–60–3931N/AN/AN/AL supramarginal gyrus/Brodmann 40, salience network

R supramarginal g. (SN)62–3532N/AN/AN/AR supramarginal gyrus/Brodmann 40, salience network

L FEF (DANh)–27–964N/AN/AN/AL frontal eye fields, dorsal attention network

R FEF (DAN)30–664N/AN/AN/AR frontal eye fields, dorsal attention network

L inferior parietal s. (DAN)–39–4352N/AN/AN/AL inferior parietal sulcus, dorsal attention network

R inferior parietal s. (DAN)39–4254N/AN/AN/AR inferior parietal sulcus, dorsal attention network

L lateral PFC (FPNi)–433328N/AN/AN/AL lateral prefrontal cortex/Brodmann 9/46, frontopolar network

L posterior parietal cortex (FPN)–46–5849N/AN/AN/AL posterior parietal cortex/von Economo PFm, frontopolar network

R lateral PFC (FPN)413830N/AN/AN/AR lateral prefrontal cortex/Brodmann 9/46, frontopolar network

R posterior parietal cortex (FPN)52–5245N/AN/AN/AR posterior parietal cortex/von Economo PFm, frontopolar network

L inferior frontal language area–51262N/AN/AN/AL inferior frontal gyrus

R inferior frontal language area54281N/AN/AN/AR inferior frontal gyrus

L posterior STGj language area–57–4715N/AN/AN/AL posterior superior temporal gyrus

R posterior STG language area59–4213N/AN/AN/AR posterior superior temporal gyrus

Anterior cerebellum0–63–30N/AN/AN/AAnterior cerebellum

Posterior cerebellum0–79–32N/AN/AN/APosterior cerebellum

aMNI: Montreal Neurological Institute stereotactic coordinate system.

bROI: regions of interest.

cPFC: prefrontal cortex.

dDMN: default mode network.

eNot available (for predefined ROIs, subject signals for correlation analyses were derived from the average signal over the entire region and cluster sizes are not available).

fN/A: not applicable (for predefined ROIs, subject signals for correlation analyses were derived from the average signal over the entire region and peak voxel locations are not available).

gSN: salience network.

hDAN: dorsal attention network.

iFPN: frontopolar network.

jSTG: superior temporal gyrus.

Statistical Analysis

Descriptive statistics (means, SDs, and SEs) were used to characterize baseline demographics and responses to pre- and postintervention psychometric assessments. Meditation practice time was the sum of the 8 weeks of practice, which was provided by the app company. Missing items in the psychometric scales were estimated with expectation maximization [48] (missing items never accounted for >5% of the total data) using other items within the scale as predictor variables. Baseline differences between the app and wait-list groups were assessed using independent t tests for continuous variables and chi-square tests for categorical variables (Tables 1 and 3). Post- versus preintervention differences in burnout, depression, anxiety, or sleep impairment in the mindfulness group were assessed by repeated-measures ANOVA. Given the exploratory nature of our analyses, we first used an α level of .05. Tests of the hypotheses were also conducted using Bonferroni-adjusted α levels of .008 (.05/6 outcome variables). To evaluate whether statistically significant outcomes were related to mindfulness practice time, we conducted Spearman correlation analyses between practice time and changes in all relevant outcomes.

Table 3. Group means and differences with respect to the Patient-Reported Outcomes Measurement Information System sleep-related impairment short form 8b (sleep disturbance), School Burnout Inventory (burnout), and The Depression Anxiety and Stress Scale (depression and anxiety) scores at baseline and >8 weeks. Sleep disturbance indices were lower in the app group than the wait-list group at >8 weeks.

Time 1Time 2F test (df)

App group, mean (SD)Wait-list group, mean (SD)t test (df)App group, mean (SD)Wait-list group, mean (SD)t test (df)
Sleep disturbance51.9 (9.99)54.1 (2.99)–0.57 (12)42.2 (7.45)53.4 (7.15)–2.93a (12)8.68a (1, 12)
Depression3.00 (2.23)4.86 (2.67)–1.41 (12)3.14 (2.41)4.00 (2.83)–0.61 (12)0.48 (1, 12)
Anxiety5.14 (3.08)8.00 (3.96)–1.51 (12)3.43 (2.82)4.57 (4.20)–0.60 (12)0.28 (1, 12)
Burnout30.7 (6.21)34.0 (5.48)–1.05 (12)29.9 (8.69)34.9 (7.67)–1.14 (12)0.58 (1, 12)

aStatistically significant at P<.05.

Differences in Connectivity by Group, Visit, and Practice Time Bin

All within- and between-group statistical analyses were performed in MATLAB v2019a. Within-group differences in connectivity distributions (mindfulness at 8 weeks vs mindfulness at baseline and wait-list at 8 weeks vs wait-list at baseline) were assessed against each ROI pair in the baseline and eighth-week ROI-ROI connectivity matrix, as appropriate, using a 2-tailed t test assuming equal variances (MATLAB ttest2() function). Correction for FDR (the expected proportion of false discoveries between all ROI-ROI pairs with similar or larger effects) was performed at a critical value of α=.05 via the Benjamini-Hochberg procedure (MATLAB mafdr() function [49]). Similarly, between-group differences in connectivity (mindfulness at 8 weeks vs mindfulness at baseline and wait-list at 8 weeks vs wait-list at baseline) were assessed against each ROI pair in the Δconn matrix via one-way ANOVA omnibus test (anova1() function) using group as a predictor, with post hoc comparisons by 2-tailed t test. In addition, mindfulness subjects were divided into low (practice time <53 minutes), moderate (53 to <225 minutes), and high (≥225 minutes) practice time bins, and a 1-way ANOVA was conducted against the Δconn matrix using practice time bin as the predictor and post hoc comparisons by 2-tailed t tests. In each of the above cases, P values were FDR corrected using the Benjamini-Hochberg procedure at α=.05.

Differences in Connectivity by Practice Time, Mental Health Scores, and Mental Health Scores

For any mental health outcomes that had a significant group-by-time interaction, we examined whether the change or changes were related to changes in functional connectivity. As our analyses indicated that the only significant impact of app-delivered mindfulness was on the PROMIS sleep impairment measure, we limited our analyses to this measure. Although connectivity data were normally distributed because of the application of the Fisher transform in preprocessing, the ordinal nature of PROMIS sleep impairment scores across all subjects, skewness of the practice time data within the mindfulness group, and the presence of outliers in SI and practice time data rendered parametric approaches inappropriate for analysis of these predictors relative to connectivity. Consequently, we used a rank-based method, Spearman rho, to investigate such relationships. Spearman rank-order correlations (ρ) are equivalent to the Pearson product-moment correlation coefficient but are applied to ranks rather than continuous values and are, therefore, less susceptible to strong outliers. Spearman correlations were calculated for practice time (mindfulness subjects only) and change in SI score (ΔSI=SI8wks–SIbaseline) versus Δconn using the MATLAB function corr(). FDR correction was performed as previously described, except that P value rankings were computed over all ROI-ROI pairs within the mindfulness group only for ρ (practice time and Δconn) analysis. All P values were FDR corrected, as previously described.

Machine Learning Classifiers for Sleep Impairment Based on Connectivity

Of the 7 subjects randomly assigned to the mindfulness group, 3 (43%) reported sleep impairment at baseline but no longer reported impairment at 8 weeks, whereas, of the 7 waitlisted subjects, 3 (43%) had sleep impairment at baseline and continued to exhibit impairment at 8 weeks (Multimedia Appendix 2, Figure S4). Therefore, we were interested in testing whether conversion between sleep impairment and nonimpairment might be predicted by a supervised learning algorithm based on the change in functional connectivity (Δconn) between baseline and >8 weeks. Accordingly, per-visit labels were assigned to each subject based on the PROMIS-derived classification of sleep impairment, such that a PROMIS SI score commensurate with sleep impairment was encoded as 1 and nonimpairment as 0. A ΔIMPAIRMENT score (IMPAIRMENT8wks–IMPAIRMENTbaseline) was then computed and used as target classes (0=no change, 11 subjects total; 1=improvement, 3 subjects total; and −1=decline, 0 subjects). A corpus of six binary classifiers, comprising support vector, gradient boosting, random forest, Gaussian naïve Bayes, linear discriminant, and multilayer perceptron estimators were implemented in scikit-learn, v0.22.1, under Python 3.7.7 (see Multimedia Appendix 2 for training and testing details). After excluding undefined values (NaNs) corresponding to the conn diagonal from the matrix, conn samples and their associated labels ϵ {−1, 0, +1} were split into training (30%) and testing (70%) sets. The samples were reshuffled into train/test sets over 5 tuning rounds. In each round, the classifier hyperparameters were tuned by random search (RandomizedSearchCV() in scikit-learn) over 5000 iterations.

A second series of classification runs were performed to test whether further dimensionality reduction improved the classifier performance. This second series leveraged the same classifiers used previously but the features (ROI-ROI connections) used for classification comprised 13 principal components—accounting for 99.99% of the data variance—derived from the conn matrix using decomposition. PCA() in scikit-learn. The contribution of each feature (connection) to each of the 13 components was ascertained by taking the magnitude of the dot product of the conn matrix, renormalized to the explained variance, and the principal components, yielding a collinearity metric between each connection and component. The results of the PCA decomposition and the connections assigned to each feature are shown in Figure 1, which also shows the contribution of each feature (connection in the conn matrix) to the 13 principal components recovered by computing the dot product between the Δconn matrix and the derived components. Hyperparameters were computed separately for this second series of classifications, following the strategy previously described.

Figure 1. Results of principal components decomposition of Δconn matrix. (a) explained variance and cumulative explained variance for the first 14 principal axes. 13 components accounted for 99.99% of the explained variance in the delta matrix (Δconn), suggesting a large amount of multicollinearity across ROI-ROI connections. The recomposition of the Δconn matrix from the 13 selected components is shown in the inset. (b) within-group correlations between per-subject Δconn values and the 13 selected components, averaged across subjects within each group. (c) Primary contributors to each component based on the dot product of the original Δconn matrix and the 13-PC decomposition. Note the substantive contributions of posterior superior temporal, fusiform (t14), orbitofrontal, and default mode, dorsal attention, and salience-network components to components PC1 and PC2. These two components represented nearly one-third of the explained variance in the Δconn matrix. DAN: dorsal attention network; DMN: default mode network; PFC: prefrontal cortex; SN: salience network; STG: superior temporal gyrus.
View this figure


At time 1, before randomization, the groups showed no differences in any self-report variables (all P values>.16) or demographic variables (all P values>.21). Participants randomized to the mindfulness group used the app between 0 and 466.2 (mean 182.8, SD 182.8) minutes. A total of 2 trainees randomized to the practitioner group did not use the app at all; however, these PA students were included in all analyses in an intent-to-treat design. None of the participants reported adverse events or adverse experiences with the app.

There was a significant group-by-time interaction for sleep impairment, such that participants randomized to mindfulness group reported a reduction in impairment compared with those randomized to the wait-list group (F12=8.68; P=.01; ηp2=0.42). No other self-reported outcomes were significant for the group-by-time interaction (low incivility: F12=0.27; P=.61; high incivility: F12=0.29; P=.60; burnout: F12=0.56; P=.47; depression: F12=0.53; P=.48; anxiety: F12=1.23; P=.28). A paired sample t test indicated that participants in the mindfulness group reported significant reductions in sleep impairment (t6=3.35; P=.02). Students randomized to the mindfulness group did not report any other significant changes, although there was a trend toward a reduction in burnout (t6 2.20; P=.07). Paired sample testing indicated that the wait-list control group reported a reduction in anxiety (t6=3.62; P=.01). Finally, independent sample t tests indicated that there was a significant difference between the mindfulness and wait-list groups in terms of sleep impairment at time 2 (t12=–2.93; P=.02). There were no other significant differences between the groups at time 2. Finally, changes in sleep impairment were not significantly correlated with mindfulness practice time within the mindfulness group (Spearman rs=–0.36; P=.43; Figure 2).

Figure 2. Differences in sleep impairment by group (mindfulness vs waitlist) and visit (time 1, baseline; Time 2, >8 weeks). An asterisk (*) indicates significant differences in sleep impairment at P≤.05. PROMIS: Patient-Reported Outcomes Measurement Information System.
View this figure

Connectivity by Group and Visit

A 2-way ANOVA was conducted to assess whether group (mindfulness or control) and/or visit (baseline and >8 weeks) were significant predictors of between- and within-group variance in the baseline and eighth-week connectivity matrices or the delta matrix (Δconn=connectivity at >8 weeks − connectivity at baseline). ANOVAs were conducted on each ROI-ROI pair separately, and P values were corrected for FDR using the Benjamini-Hochberg procedure. We failed to find significant effects of the group × visit interaction for any connection, and subsequent inspection of the main effects indicated that although group and visit each accounted for a statistically significant amount of variance across the connectivity matrices, they did not affect the same connections. Consequently, main-effects ANOVAs and subsequent post hoc analysis by a 2-tailed t test were conducted for group and visit separately.

The ANOVA on group conducted against the baseline and eighth-week connectivity matrices indicated significant differences in control subjects’ connectivity among the supplementary motor, middle temporal, inferior temporal, occipitopolar, and orbitofrontal cortices at baseline, and among the insular, cerebellar, lateral visual, and superior frontal cortices and thalamus at >8 weeks (Table 4). Post hoc t tests revealed that, at baseline, mindfulness participants exhibited stronger connectivity than controls with respect to left supplementary motor–left ventrolateral Brodmann area 37 (t12=5.37; PFDR=.02) and left inferolateral Brodmann area 20–right middle temporal gyrus (t12=4.73; PFDR=.05). The left BA37 ROI was localized to the ventrolateral aspect of the ipsilateral middle temporal gyrus corresponding to the area PH of Economo-Koskinas [50], and the left BA20 ROI was localized to the anterior aspect of visual area TE2. In contrast, controls exhibited stronger left lateral occipitopolar–ipsilateral superior temporal gyrus connectivity than the mindfulness subjects (t12=−4.98; P=.03) at baseline. The latter ROI was associated with area TE3 and/or auditory complex 5. At >8 weeks, mindfulness subjects exhibited stronger connectivity than control subjects with respect to the left dorsolateral Brodmann area 8–left posterior insula (t12=4.84; PFDR=.05), left cerebellar lobule VI–right lateral visual cortex (t12=4.65; PFDR=.05), and right premotor thalamus–right lateral visual cortex (t12=4.32; PFDR=.05).

Table 4. Results of t tests on group (mindfulness, or MF, vs wait-list, or CX) and visit (visit 1, v1, vs visit 2, v2). Connections surviving false discovery rate correction are presented as pairs of source and versus regions of interest.
Group and sourcesVersust test (df)aPFDR
MF v1>CX v1

f6_Lsmat14_Lba37vl5.3665 (12).02

o2_Llopt11_Laud5–4.977 (12).03

t4_Lba20ilt9_Raud54.7331 (12).049
MF v2>CX v2

f5_Lba8dlt7_Lpins24.7415 (12).048

cb2_Llob6R lateral visual4.6543 (12).05

th2_RpmthalR lateral visual4.3193 (12).05
MF v2>MF v1

posterior cing. (DMN)anterior cing. (SN)5.7184 (7).009

orb2_Lba10r14mt7_Lpins25.5659 (7).01

f11_Rba6ba8anterior cingulate (SN)4.9121 (7).02

R FEF (DAN)R inf. frontal lang. area4.8411 (7).04

posterior cing. (DMN)L anterior insula (SN)4.1621 (7).04

cing1_RpccL anterior insula (SN)4.1902 (7).04

orb5_Rba10vt7_Lpins24.1199 (7).047

t7_Lpins2medial PFC (DMN)4.6925 (7).05
CX v2>CX v1

o2_Llopmtl5_Lamg5.4641 (5).01

aTwo-tailed t test.

The ANOVA on visit conducted against the delta matrix Δconn indicated that baseline-to-eighth-week connectivity differed between mindfulness and control subjects with respect to connections between the left lateral occipitopolar cortex and the ipsilateral amygdala and in mindfulness subjects’ connectivity among the anterior and posterior cingulate, insular, orbitofrontal, and medial prefrontal cortices (Tables 4 and 5). Post hoc t tests subsequently revealed that control subjects exhibited higher connectivity between left lateral occipitopolar cortex and left amygdala at >8 weeks than at baseline (t=5.46; PFDR=.01). In contrast, mindfulness participants exhibited higher connectivity at >8 weeks, relative to baseline, between anterior cingulate (an SN component) and posterior cingulate (a DMN component; t=5.72; PFDR=.01), left rostral orbitofrontal Brodmann area 10 and ipsilateral posterior insula (t=5.57; PFDR=.01), anterior cingulate and right superior frontal gyrus (t=4.91; PFDR=.02), right frontal eye fields (a dorsal attention network component) and ipsilateral inferior frontal gyrus (t=4.84; PFDR=.04), posterior cingulate and left anterior insula (an SN component; t=4.16; PFDR=.04), the ventral aspect of the right orbitofrontal Brodmann area 10 and left posterior insula (t=4.12; PFDR=.05), and left posterior insula and medial prefrontal cortex (t=4.69; PFDR=.05). The superior frontal ROI was associated with an area denoted i6-8 (Assem et al [51]) located in the superior aspect of the transition area between the premotor cortex (Brodmann area 6) and frontal eye fields (Brodmann area 8; Table 2).

Table 5. Results of one-way analysis of variance on delta matrix with group as the predictor. Post hoc analysis by 2-tailed t test. P values false discovery rate-corrected by Benjamini-Hochberg procedure. Note that, as the connectivity coefficients were normalized, they are directly comparable with a Cohen d statistic and thus indicate effect sizes. The differences listed here fall in the range of 0.15-0.35 and should be regarded as small-to-moderate effects.
ID1ID2ROI1ROI2F test (df)MSEaModel P valuePost hocMean differenceP value
t1_Lffcorb2_Lba10r14mL fusiform face complexL r. BA10/med. BA1420.1465 (1, 12)0.016<.001Control<mindfulness–0.3031<.001
pl2_Lba7morb1_Rba11L med. BA7/med. BA5R lat. BA11/area Fo315.3067 (1, 12)0.0285.002Control<mindfulness–0.3531.002
Anterior cerebellumPosterior cerebellumAnterior cerebellumPosterior cerebellum8.8866 (1, 12)0.0179.01Control<mindfulness–0.2134.01
mtl3_Renrccb4_Rlob7bR entorhinal ctx./presubiR cerebellar lobule VIIb6.3201 (1, 12)0.0162.03Control<mindfulness–0.171.03
pl4_Lpos2caud2_Rput1L parieto-occip. s./medial BA7R mid. insula/putamen5.9626 (1, 12)0.0215.03Control<mindfulness–0.1914.03
pl3_Lba7m5mlf4_Rba8dL med. BA7/med. lat. BA5R mid. frontal g./BA 85.9356 (1, 12)0.051.03Control<mindfulness–0.294.03
cb2_Llob6pl2_Lba7mL cerebellar lobule VIL med. BA7/med. BA54.9141 (1, 12)0.018.046Control<mindfulness–0.1588.046
orb3_Rifgpomtl2_RphgR IFG orb./sup. BA47/lat. BA11R parahippocampal g4.889 (1, 12)0.0167.047Control<mindfulness–0.1527.047
mtl1_Lphg1f5_Lba8dlL parahippocampal g./subiL dors. lat. BA84.7757 (1, 12)0.027.049Mindfulness<control0.1919.049

aMSE: mean square error.

Changes in Connectivity With the SI Score and Practice Time

Nearly all subjects exhibited a decrease in SI scores between baseline (control: 54.1 [SD 2.99]; mindfulness: 51.89 [SD 9.99]) and +8 weeks [control 53.36 [SD 7.15]; mindfulness: 42.19 [SD 7.45]). Mean ΔSI scores, defined as SI score at >8 weeks minus SI score at baseline, were −0.76 (SD 5.13) and −9.70 (SD 6.18) for the control and mindfulness groups, respectively. Spearman rank correlation, assessed against the delta matrix (Δconn, defined as connectivity among the 102 ROIs at >8 weeks minus connectivity at baseline) across all subjects, indicated that a greater decrease in SI score was associated with increased connectivity between the right superior frontal gyrus (BA6/8 transition area, i6-8) and ipsilateral inferior parietal sulcus (ρ=0.82, PFDR=.03) and between the left superior temporal gyrus and medial prefrontal cortex (a component of the DMN; ρ=0.81; PFDR=.05). The superior temporal ROI was localized to the caudal aspect of Brodmann area 22 (Wernicke area) and the superior temporal visual cortex. Greater decreases in SI score were also associated with reduced connectivity between the left supplementary motor area and the ventrolateral aspect of the ipsilateral middle temporal gyrus/Economo-Koskinas area PH (ρ=0.89; PFDR<.01; Table 6).

Within the mindfulness group, Spearman correlations indicated an association between practice time and changes in the connectivity of four connections: between the right inferior parietal sulcus (a component of the dorsal attention network) and right lateral visual cortex, between the right inferior parietal sulcus and the occipitopolar visual cortex, between the right middle orbital gyrus and orbitofrontal Brodmann area 11, and between the right middle orbital gyrus and the left lateral sensorimotor cortex. The strengths of these connections increased linearly with practice time, with the exception of the left lateral sensorimotor–middle orbital gyrus connection, which decreased with practice time (Tables S1 and S2 and Figure S3 in Multimedia Appendix 2).

Table 6. Spearman correlations (rho) between the connectivity strength of selected connections and change in sleep impairment score (ΔSI, defined as sleep impairment at 8 weeks minus sleep impairment at baseline), taken across all subjects. Correlations are corrected for multiple comparisons using false discovery rate.
t5_Lba22cDMN medial PFC–0.8057.05
f11_Rba6ba8DAN R IPS–0.8234.03


Throughout 5 rounds of training or test group shuffling, each with 5000 iterations of hyperparameter tuning, the 6 estimators (support vector, gradient boosting, random forest, Gaussian naïve Bayes, linear discriminant, and multilayer perceptron) were equally proficient in classifying the change from sleep impairment at baseline to no sleep impairment at >8 weeks based on the full 102 × 102-ROI delta matrix (Δconn) for each subject. All estimators correctly classified true converters; conversely, they misclassified at least 1 nonconverter in each of the 5 rounds. Reducing the dimensionality of the Δconn matrix to 13 components (99.99% of the cumulative explained variance; Figure 1) via principal component analysis increased the accuracy of most classifiers by 3-5% by increasing estimator specificity (decreasing the false positive rate). Conversely, the estimators tended to misclassify true converters as nonconverters, increasing the false-negative rate, when predictions were based on the reduced-dimensionality delta matrix. Under this schema, Gaussian naïve Bayes and random forest classifiers slightly outperformed other estimators, with an accuracy rate of approximately 80%. We must, however, urge caution in overinterpreting these results, given the small number of true converters (n=3) to nonconverters (n=11) and the absence of subjects converting from nonsleep-impaired to sleep-impaired.

Principal Findings

Provider burnout and depression have profound national and institutional economic costs, as well as deep societal and social effects. Burnout costs an estimated US $150 billion per year or almost 5% of the nation’s health care expenditure [52]. Although these costs are generally estimated based on the effects of burnout and depression among physicians, PAs also report remarkably high levels of burnout [5] and depression [6]. Although little is known about the public health impact of PA burnout, physician burnout increases malpractice rates, exacerbates physician shortages, erodes both health care organization morale and patient experience [53], and reduces clinical effectiveness [54,55]. Given their overlapping roles and day-to-day activities, it is likely that burnout has a comparable effect on PAs.

Similar to the relative lack of data on burnout and well-being among practicing PAs, far less is known about the mental health needs of PA students than is known about medical residents and medical students. The PA profession is in a period of some flux, with a workforce population that is growing in overall number and that is increasingly young and female [4,56]. PA students may have unique needs, given the differences between PAs and physicians. PA students often report choosing their professional route based on concerns about debt load [57] and expectations of a healthier work-life balance [58]. Here, students reported low levels of depression (0% of students) and relatively low levels of anxiety. Approximately 21% (3/14) of students indicated moderate anxiety, and 7% (1/14) of students indicated severe anxiety at the outset of the study. More students (43%) reported sleep impairment; 29% (4/14) of students reported mild impairment, and 14% (2/14) of students reported moderate impairment before randomization. It is important to note that the students who enrolled in the study may not be representative of the entire PA student population.

Mindfulness meditation has shown great promise for improving sleep disruption and insomnia symptoms [13,14], and it may be beneficial to health professional trainees who often experience high rates of sleep dysfunction [16]. However, most studies examining mindfulness among health professionals and trainees have examined time-intensive interventions that are prohibitive for many. Previous studies have shown that app-delivered mindfulness may be effective in reducing anxiety among physicians [59]. Although we included a wait-list control group to control self-selection and the inevitable changes that occur during PA school but not attributable to the intervention, future studies should include an active control condition. Although this has been challenging in studies of app-delivered mindfulness to date, recent work has advanced in this area toward developing smartphone apps that can be used as active comparators (eg, Huberty et al [60]).

PA students randomized to mindfulness reported a significant reduction in sleep impairment compared with students randomized to the wait-list. Although 43% (3/7) of students randomized to mindfulness reported mild (t=55-59.9) or moderate (t=60-69.9) sleep impairment before randomization, none of the students randomized to mindfulness reported sleep impairment after training (t<55) [61]. These findings are consistent with previous research indicating that mindfulness-based interventions are effective in reducing sleep disturbance and for altering neurobiology related to the default mode in adults with sleep disturbances [62]. Our findings extend previous work to indicate that short, app-delivered mindfulness is beneficial for improving self-reported sleep impairment. A recent study found that most study participants using a popular mindfulness app (Calm) downloaded it to improve sleep impairment [63]; considering our data further highlights the potential importance of app-delivered mindfulness in the context of sleep and sleep dysfunction. This study also adds to what is known about the impact of mindfulness on medical trainees [64] and may be part of a critical solution for sleep dysfunction, which is associated with an increased risk of depression, burnout, and medical errors [65,66]. Although changes in sleep impairment were not correlated with practice time, the relationship was in the expected direction, and it is likely that we were not powered to detect this relationship.

These data also indicate that improvements in sleep are associated with connectivity changes between the DMN and regions important for emotion, attention, and social cognition. Previous studies have shown that disordered sleep is related to altered brain function in the DMN as well as the SN [67-69]. Here, using principal components decomposition, we found that a small number of connections among DMN, SN, and dorsal attention network components, and with superior temporal, fusiform, and orbitofrontal areas, are closely associated with one another and with the explained variance of the delta-connectivity matrix. This area, explicitly characterized as i6-8, is functionally distinct from regions involved in simple eye movements and is considered part of a core complex involved in working memory, along with the inferior parietal sulcus [51]. The finding that changes in this network of regions are related to changes in sleep impairment further bolsters the existing evidence that the DMN and SNs are affected by or involved in sleep impairment.

PA well-being is a complex and multifactorial issue. Isolation, sleep deprivation and disturbance, and feeling overwhelmed by the amount of material they need to master are risk factors for depression among health care trainees [65,70]. Moreover, lack of time for self-care and stigma toward treatment-seeking are barriers to addressing mental health crises among trainees [2,71]. For these reasons, it is unlikely that a short-term, app-delivered mindfulness meditation program will be a stand-alone solution. Rather, addressing trainee mental health must be comprehensive and should include structural and organizational solutions alongside individualized resilience programming. Moreover, wellness programs for trainees must be sustainable and preventive in nature rather than reactive [64,72]. Medical training programs must make wellness feasible within the lives of trainees, and app-delivered programming may be a feasible and sustainable piece to foster a culture of resilience among PA students.


This study had a small sample, and the findings may not be representative of all PA students. We were likely underpowered to detect small effects, and the changes in sleep impairment did not reach significance at alpha levels adjusted for multiple comparisons. Moreover, it is unclear whether improvements in sleep impairment reported by students randomized to mindfulness are enduring. There is some evidence that there are sex differences in how disordered sleep affects brain function [73]. The students in our sample were primarily women, and thus, the results may not be generalizable to male trainees. Despite these limitations, the methods used here are a novel approach to understanding sleep impairment and a mindfulness intervention that may improve it, and these data indicate that app-delivered mindfulness may be effective for PA students.


The authors would like to express their gratitude to the research assistants who helped collect the data reported here and the faculty in the Emory PA program for facilitating student recruitment. The authors are also grateful to Chris Larsen for his guidance and expertise in embarking on this research and 10% Happier for providing app subscriptions. 10% Happier provided input on recommended content during study design but did not have access to study data and was not involved in data analysis or interpretation.

Conflicts of Interest

None declared.

Multimedia Appendix 1

Quality control and assurance measures, including motion correction and signal change assessments, validity of subject-to-stereotactic space registrations, effects of denoizing on connectivity, and functional connectivity as a function of distance from the seed voxel.

PDF File (Adobe PDF File), 283 KB

Multimedia Appendix 2

Supplementary figures and tables, including results of multivoxel pattern analysis, connectivity matrices by group and visit, and additional regression and classification results.

PDF File (Adobe PDF File), 1002 KB

  1. Mata DA, Ramos MA, Bansal N, Khan R, Guille C, Di Angelantonio E, et al. Prevalence of depression and depressive symptoms among resident physicians: a systematic review and meta-analysis. JAMA 2015 Dec 08;314(22):2373-2383 [FREE Full text] [CrossRef] [Medline]
  2. Rotenstein LS, Ramos MA, Torre M, Segal JB, Peluso MJ, Guille C, et al. Prevalence of depression, depressive symptoms, and suicidal ideation among medical students: a systematic review and meta-analysis. JAMA 2016 Dec 06;316(21):2214-2236 [FREE Full text] [CrossRef] [Medline]
  3. Dyrbye LN, Shanafelt TD, Sinsky CA, Cipriano PF, Bhatt J, Ommaya A, et al. Burnout among health care professionals: a call to explore and address this underrecognized threat to safe, high-quality care. National Academy of Medicine. 2017.   URL: https:/​/nam.​edu/​burnout-among-health-care-professionals-a-call-to-explore-and-address-this-underrecognized-threat-to-safe-high-quality-care/​ [accessed 2021-09-10]
  4. Essary AC, Bernard KS, Coplan B, Dehn R, Forister JG, Smith NE, et al. Burnout and job and career satisfaction in the physician assistant profession: a review of the literature. National Academy of Medicine. 2018.   URL: https:/​/nam.​edu/​burnout-and-job-and-career-satisfaction-in-the-physician-assistant-profession-a-review-of-the-literature/​ [accessed 2021-09-10]
  5. Tetzlaff ED, Hylton HM, DeMora L, Ruth K, Wong Y. National study of burnout and career satisfaction among physician assistants in oncology: implications for team-based care. J Oncol Pract 2018 Jan;14(1):e11-e22 [FREE Full text] [CrossRef] [Medline]
  6. Varner D, Foutch B. Depression and burnout symptoms among Air Force family medicine providers. JAAPA 2014 May;27(5):42-46. [CrossRef] [Medline]
  7. Johnson AK, Blackstone SR, Simmons W, Skelly A. Assessing burnout and interest in wellness programs in physician assistant students. J Physician Assist Educ 2020 Jun;31(2):56-62. [CrossRef] [Medline]
  8. Demarzo MM, Montero-Marin J, Cuijpers P, Zabaleta-del-Olmo E, Mahtani KR, Vellinga A, et al. The efficacy of mindfulness-based interventions in primary care: a meta-analytic review. Ann Fam Med 2015 Nov;13(6):573-582 [FREE Full text] [CrossRef] [Medline]
  9. Goyal M, Singh S, Sibinga EM, Gould NF, Rowland-Seymour A, Sharma R, et al. Meditation programs for psychological stress and well-being: a systematic review and meta-analysis. JAMA Intern Med 2014 Mar;174(3):357-368 [FREE Full text] [CrossRef] [Medline]
  10. Creswell JD, Lindsay EK. How does mindfulness training affect health? A mindfulness stress buffering account. Curr Dir Psychol Sci 2014 Dec 16;23(6):401-407. [CrossRef]
  11. Black D, Slavich G. Mindfulness meditation and the immune system: a systematic review of randomized controlled trials. Ann N Y Acad Sci 2016 Jun;1373(1):13-24 [FREE Full text] [CrossRef] [Medline]
  12. Cásedas L, Pirruccio V, Vadillo MA, Lupiáñez J. Does mindfulness meditation training enhance executive control? A systematic review and meta-analysis of randomized controlled trials in adults. Mindfulness 2019 Dec 10;11(2):411-424. [CrossRef]
  13. Ong JC, Moore C. What do we really know about mindfulness and sleep health? Curr Opin Psychol 2020 Aug;34:18-22. [CrossRef] [Medline]
  14. Winbush NY, Gross CR, Kreitzer MJ. The effects of mindfulness-based stress reduction on sleep disturbance: a systematic review. Explore (NY) 2007;3(6):585-591. [CrossRef] [Medline]
  15. Neuendorf R, Wahbeh H, Chamine I, Yu J, Hutchison K, Oken BS. The effects of mind-body interventions on sleep quality: a systematic review. Evid Based Complement Alternat Med 2015;2015:902708 [FREE Full text] [CrossRef] [Medline]
  16. Sen S, Kranzler H, Didwania A, Schwartz A, Amarnath S, Kolars J, et al. Effects of the 2011 duty hour reforms on interns and their patients: a prospective longitudinal cohort study. JAMA Intern Med 2013 Apr 22;173(8):657-62; discussion 663 [FREE Full text] [CrossRef] [Medline]
  17. Dobkin PL, Hutchinson TA. Teaching mindfulness in medical school: where are we now and where are we going? Med Educ 2013 Aug;47(8):768-779. [CrossRef] [Medline]
  18. McConville J, McAleer R, Hahne A. Mindfulness training for health profession students-the effect of mindfulness training on psychological well-being, learning and clinical performance of health professional students: a systematic review of randomized and non-randomized controlled trials. Explore (NY) 2017;13(1):26-45. [CrossRef] [Medline]
  19. Lomas T, Medina JC, Ivtzan I, Rupprecht S, Eiroa-Orosa FJ. A systematic review and meta-analysis of the impact of mindfulness-based interventions on the well-being of healthcare professionals. Mindfulness 2018 Nov 29;10(7):1193-1216. [CrossRef]
  20. Hoover EB, Butaney B, Stoehr JD. Exploring the effectiveness of mindfulness and decentering training in a physician assistant curriculum. J Physician Assist Educ 2020 Mar;31(1):19-22. [CrossRef] [Medline]
  21. Creswell JD, Taren AA, Lindsay EK, Greco CM, Gianaros PJ, Fairgrieve A, et al. Alterations in resting-state functional connectivity link mindfulness meditation with reduced interleukin-6: a randomized controlled trial. Biol Psychiatry 2016 Jul 01;80(1):53-61. [CrossRef] [Medline]
  22. Mooneyham B, Mrazek M, Mrazek A, Schooler J. Signal or noise: brain network interactions underlying the experience and training of mindfulness. Ann N Y Acad Sci 2016 Apr;1369(1):240-256. [CrossRef] [Medline]
  23. Hölzel BK, Hoge EA, Greve DN, Gard T, Creswell JD, Brown KW, et al. Neural mechanisms of symptom improvements in generalized anxiety disorder following mindfulness training. Neuroimage Clin 2013 Mar 25;2:448-458 [FREE Full text] [CrossRef] [Medline]
  24. Doll A, Hölzel BK, Mulej Bratec S, Boucard CC, Xie X, Wohlschläger AM, et al. Mindful attention to breath regulates emotions via increased amygdala-prefrontal cortex connectivity. Neuroimage 2016 Jul 01;134:305-313. [CrossRef] [Medline]
  25. Modinos G, Ormel J, Aleman A. Individual differences in dispositional mindfulness and brain activity involved in reappraisal of emotion. Soc Cogn Affect Neurosci 2010 Dec;5(4):369-377 [FREE Full text] [CrossRef] [Medline]
  26. Fang H, Tu S, Sheng J, Shao A. Depression in sleep disturbance: a review on a bidirectional relationship, mechanisms and treatment. J Cell Mol Med 2019 Apr;23(4):2324-2332 [FREE Full text] [CrossRef] [Medline]
  27. Harris D, Warren J, Adler C. Meditation for Fidgety Skeptics: a 10% Happier How-To Book. New York: Harmony; 2018.
  28. Ten percent happier.   URL: [accessed 2021-09-10]
  29. The basic of mindfulness meditation. Ten Percent Happier.   URL: [accessed 2021-09-10]
  30. Clark CM, Barbosa-Leiker C, Gill LM, Nguyen D. Revision and psychometric testing of the Incivility in Nursing Education (INE) survey: introducing the INE-R. J Nurs Educ 2015 Jun;54(6):306-315. [CrossRef] [Medline]
  31. Salmela-Aro K, Kiuru N, Leskinen E, Nurmi J. School Burnout Inventory (SBI): reliability and validity. Eur J Psychol Assess 2009 Jan;25(1):48-57. [CrossRef]
  32. Lovibond P, Lovibond S. The structure of negative emotional states: comparison of the Depression Anxiety Stress Scales (DASS) with the Beck Depression and Anxiety Inventories. Behav Res Ther 1995 Mar;33(3):335-343. [CrossRef] [Medline]
  33. Hanish A, Lin-Dyken D, Han J. PROMIS sleep disturbance and sleep-related impairment in adolescents: examining psychometrics using self-report and actigraphy. Nurs Res 2017;66(3):246-251 [FREE Full text] [CrossRef] [Medline]
  34. Sleep-related impairment: a brief guide to the PROMIS® Sleep-Related Impairment instruments. PROMIS. 2020.   URL: https:/​/www.​​images/​PROMIS/​manuals/​PROMIS_Sleep-Related_Impairment_Scoring_Manual.​pdf [accessed 2021-09-10]
  35. Whitfield-Gabrieli S, Nieto-Castanon A. Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks. Brain Connect 2012;2(3):125-141. [CrossRef] [Medline]
  36. Ashburner J, Barnes G, Chen CC, Daunizeau J, Flandin G, Friston K, et al. SPM12 Manual. In: SPM12 Manual. London, UK: Wellcome Trust Centre for Neuroimaging; 2014.
  37. Penny W, Friston K, Ashburner J, Kiebel S, Nichols T, editors. Statistical Parametric Mapping: the Analysis of Functional Brain Images. London, UK: Elsevier; 2011.
  38. Behzadi Y, Restom K, Liau J, Liu TT. A component based noise correction method (CompCor) for BOLD and perfusion based fMRI. Neuroimage 2007 Aug 01;37(1):90-101 [FREE Full text] [CrossRef] [Medline]
  39. Fox MD, Snyder AZ, McAvoy MP, Barch DM, Raichle ME. The BOLD onset transient: identification of novel functional differences in schizophrenia. Neuroimage 2005 Apr 15;25(3):771-782. [CrossRef] [Medline]
  40. Fox MD, Zhang D, Snyder AZ, Raichle ME. The global signal and observed anticorrelated resting state brain networks. J Neurophysiol 2009 Jun;101(6):3270-3283 [FREE Full text] [CrossRef] [Medline]
  41. Fransson P. How default is the default mode of brain function? Further evidence from intrinsic BOLD signal fluctuations. Neuropsychologia 2006;44(14):2836-2845. [CrossRef] [Medline]
  42. Chai XJ, Castañón AN, Ongür D, Whitfield-Gabrieli S. Anticorrelations in resting state networks without global signal regression. Neuroimage 2012 Jan 16;59(2):1420-1428 [FREE Full text] [CrossRef] [Medline]
  43. Norman KA, Polyn SM, Detre GJ, Haxby JV. Beyond mind-reading: multi-voxel pattern analysis of fMRI data. Trends Cogn Sci 2006 Sep;10(9):424-430. [CrossRef] [Medline]
  44. Worsley KJ, Marrett S, Neelin P, Vandal AC, Friston KJ, Evans AC. A unified statistical approach for determining significant signals in images of cerebral activation. Hum Brain Mapp 1996;4(1):58-73. [CrossRef] [Medline]
  45. Friston KJ, Holmes AP, Worsley KJ, Poline J, Frith CD, Frackowiak RS. Statistical parametric maps in functional imaging: a general linear approach. Hum Brain Mapp 1994;2(4):189-210. [CrossRef]
  46. Poline J, Worsley KJ, Holmes AP, Frackowiak RS, Friston KJ. Estimating smoothness in statistical parametric maps: variability of P values. J Comput Assist Tomogr 1995;19(5):788-796. [CrossRef] [Medline]
  47. Chumbley J, Worsley K, Flandin G, Friston K. Topological FDR for neuroimaging. Neuroimage 2010 Feb 15;49(4):3057-3064 [FREE Full text] [CrossRef] [Medline]
  48. Graham JW. Missing data analysis: making it work in the real world. Annu Rev Psychol 2009;60(1):549-576. [CrossRef] [Medline]
  49. Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J R Stat Soc Series B Stat Methodol 1995;57(1):289-300. [CrossRef]
  50. von Economo EC, Freiherr C, Koskinas G, Triarhou LC. Atlas of Cytoarchitectonics of the Adult Human Cerebral Cortex. Basel: Karger; 2008.
  51. Assem M, Glasser M, Van Essen DC, Duncan J. A domain-general cognitive core defined in multimodally parcellated human cortex. Cereb Cortex 2020 Jun 30;30(8):4361-4380 [FREE Full text] [CrossRef] [Medline]
  52. The Hidden Epidemic: The Moral Imperative for Academic Health Centers to Address Health Professionals' Well-Being. Atlanta: Emory University Woodruff Health Sciences Center; Jun 2017.
  53. Rathert C, Williams ES, Linhart H. Evidence for the quadruple aim: a systematic review of the literature on physician burnout and patient outcomes. Med Care 2018 Dec;56(12):976-984. [CrossRef] [Medline]
  54. Del Canale S, Louis DZ, Maio V, Wang X, Rossi G, Hojat M, et al. The relationship between physician empathy and disease complications: an empirical study of primary care physicians and their diabetic patients in Parma, Italy. Acad Med 2012 Sep;87(9):1243-1249. [CrossRef] [Medline]
  55. Rakel DP, Hoeft TJ, Barrett BP, Chewning B, Craig B, Niu M. Practitioner empathy and the duration of the common cold. Fam Med 2009;41(7):494-501 [FREE Full text] [Medline]
  56. Cockerham W, Hinote B. PAs in a changing society: a sociologic perspective. JAAPA 2015 Aug;28(8):18-20. [CrossRef] [Medline]
  57. Essary A, Coplan B, Cawley J, Schneller E, Ohsfeldt R. Women, family medicine, and career choice: an opportunity cost analysis. JAAPA 2016 Sep;29(9):44-48. [CrossRef] [Medline]
  58. Hooker R, Robie S, Coombs J, Cawley J. The changing physician assistant profession: a gender shift. JAAPA 2013 Sep;26(9):36-44. [CrossRef] [Medline]
  59. Roy A, Druker S, Hoge EA, Brewer JA. Physician anxiety and burnout: symptom correlates and a prospective pilot study of app-delivered mindfulness training. JMIR Mhealth Uhealth 2020 Apr 01;8(4):e15608 [FREE Full text] [CrossRef] [Medline]
  60. Huberty J, Eckert R, Puzia M, Laird B, Larkey L, Mesa R. A novel educational control group mobile app for meditation interventions: single-group feasibility trial. JMIR Form Res 2020 Jul 21;4(7):e19364 [FREE Full text] [CrossRef] [Medline]
  61. Online Assessment Measures: LEVEL 2—Sleep Disturbance—Adult. American Psychiatric Association. 2012.   URL: [accessed 2021-09-10]
  62. Kurth F, Luders E, Wu B, Black DS. Brain gray matter changes associated with mindfulness meditation in older adults: an exploratory pilot study using voxel-based morphometry. Neuro 2014;1(1):23-26 [FREE Full text] [CrossRef] [Medline]
  63. Huberty J, Vranceanu A, Carney C, Breus M, Gordon M, Puzia ME. Characteristics and usage patterns among 12,151 paid subscribers of the calm meditation app: cross-sectional survey. JMIR Mhealth Uhealth 2019 Nov 03;7(11):e15648 [FREE Full text] [CrossRef] [Medline]
  64. Wasson LT, Cusmano A, Meli L, Louh I, Falzon L, Hampsey M, et al. Association between learning environment interventions and medical student well-being: a systematic review. JAMA 2016 Dec 06;316(21):2237-2252 [FREE Full text] [CrossRef] [Medline]
  65. Kalmbach D, Arnedt J, Song P, Guille C, Sen S. Sleep disturbance and short sleep as risk factors for depression and perceived medical errors in first-year residents. Sleep 2017 Mar 01;40(3):zsw073 [FREE Full text] [CrossRef] [Medline]
  66. Rosen IM, Gimotty PA, Shea JA, Bellini LM. Evolution of sleep quantity, sleep deprivation, mood disturbances, empathy, and burnout among interns. Acad Med 2006 Jan;81(1):82-85. [CrossRef] [Medline]
  67. Nie X, Shao Y, Liu SY, Li HJ, Wan AL, Nie S, et al. Functional connectivity of paired default mode network subregions in primary insomnia. Neuropsychiatr Dis Treat 2015 Dec 16;11:3085-3093 [FREE Full text] [CrossRef] [Medline]
  68. Khazaie H, Veronese M, Noori K, Emamian F, Zarei M, Ashkan K, et al. Functional reorganization in obstructive sleep apnoea and insomnia: a systematic review of the resting-state fMRI. Neurosci Biobehav Rev 2017 Jun;77:219-231 [FREE Full text] [CrossRef] [Medline]
  69. Pace-Schott EF, Zimmerman JP, Bottary RM, Lee EG, Milad MR, Camprodon JA. Resting state functional connectivity in primary insomnia, generalized anxiety disorder and controls. Psychiatry Res Neuroimaging 2017 Jul 30;265:26-34 [FREE Full text] [CrossRef] [Medline]
  70. Lapinski J, Yost M, Sexton P, LaBaere RJ. Factors modifying burnout in osteopathic medical students. Acad Psychiatry 2016 Feb 25;40(1):55-62. [CrossRef] [Medline]
  71. Williford ML, Scarlet S, Meyers MO, Luckett DJ, Fine JP, Goettler CE, et al. Multiple-institution comparison of resident and faculty perceptions of burnout and depression during surgical training. JAMA Surg 2018 Aug 01;153(8):705-711 [FREE Full text] [CrossRef] [Medline]
  72. Slavin SJ. Medical student mental health: culture, environment, and the need for change. JAMA 2016 Dec 06;316(21):2195-2196. [CrossRef] [Medline]
  73. Dai X, Nie X, Liu X, Pei L, Jiang J, Peng D, et al. Gender differences in regional brain activity in patients with chronic primary insomnia: evidence from a resting-state fMRI study. J Clin Sleep Med 2016 Mar 15;12(3):363-374 [FREE Full text] [CrossRef] [Medline]

ANOVA: analysis of variance
DMN: default mode network
FDR: false discovery rate
fMRI: functional magnetic resonance imaging
GLM: general linear model
PA: physician assistant
PROMIS: Patient-Reported Outcomes Measurement Information System
ROI: region of interest
rsfMRI: resting-state fMRI
SI: sleep-related impairment
SN: salience network
TR: repetition time interval

Edited by J Torous; submitted 09.09.20; peer-reviewed by J Huckins, H Cramer, K Sanada; comments to author 14.12.20; revised version received 23.04.21; accepted 23.07.21; published 19.10.21


©Jeremy L Smith, Jason W Allen, Carla I Haack, Kathryn L Wehrmeyer, Kayley G Alden, Maha B Lund, Jennifer S Mascaro. Originally published in JMIR Formative Research (, 19.10.2021.

This is an open-access article distributed under the terms of the Creative Commons Attribution License (, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Formative Research, is properly cited. The complete bibliographic information, a link to the original publication on, as well as this copyright and license information must be included.