Multimodal Ambulatory Monitoring of Daily Activity and Health-Related Symptoms in Community-Dwelling Survivors of Stroke: Feasibility, Acceptability, and Validity

Abstract

Objective:

To examine the feasibility, acceptability, and validity of multimodal ambulatory monitoring, which combines accelerometry with ecological momentary assessment (EMA), to assess daily activity and health-related symptoms among survivors of stroke.

Design:

Prospective cohort study involving 7 days of ambulatory monitoring; participants completed 8 daily EMA surveys about daily activity and symptoms (mood, cognitive complaints, fatigue, pain) while wearing an accelerometer. Participants also completed retrospective assessments and an acceptability questionnaire.

Setting:

Community.

Participants:

Forty survivors of stroke (N=40).

Interventions:

Not applicable.

Main Outcome Measures:

Feasibility was determined using attrition rate and compliance. Acceptability was reported using the acceptability questionnaire. Convergent and discriminant validity were determined by the correlations between ambulatory monitoring and retrospective self-reports. Criterion validity was determined by the concordance between accelerometer-measured and EMA-reported daily activity.

Results:

All participants completed the study (attrition rate=0%). EMA and accelerometer compliance were 93.6 % and 99.7%, respectively. Participants rated their experience with multimodal ambulatory monitoring positively. They were highly satisfied (mean, 4.8/5) and confident (mean, 4.7/5) in using ambulatory monitoring and preferred it over traditional retrospective assessments (mean, 4.7/5). Multimodal ambulatory monitoring estimates correlated with retrospective self-reports of the same and opposing constructs in the predicted directions (r=−0.66 to 0.72, P<.05). More intense accelerometer-measured physical activity was observed when participants reported doing more physically demanding activities and vice versa.

Conclusions:

Findings support the feasibility, acceptability, and validity of multimodal ambulatory monitoring in survivors of mild stroke. Multimodal ambulatory monitoring has potential to provide a more complete understanding of survivors’ daily activity in the context of everyday life.

Stroke remains a leading cause of long-term disability in the United States, producing a variety of symptoms that interfere with daily life experiences of survivors of stroke.^1-3 Increasing daily physical activity intensity and minimizing sedentary behavior are recommended by the American Stroke Association for mitigating stroke symptoms and risks for stroke recurrence.^4,5 The American Stroke Association statements^4,5 have underlined the benefits of physical activity for survivors of stroke, including improved mood,⁶ cognitive function,⁷ and quality of life.⁸ Despite the extensive benefits, survivors of stroke are more sedentary than populations with other chronic diseases.^9,10

Activity patterns fluctuate during the day because of changes in activity type and biopsychological factors (eg, pain, depression),¹¹ and yet most stroke studies have relied on retrospective questionnaires and laboratory-based performance tests to study daily activity participation.^12-14 Retrospective questionnaires require recalling past experiences and are thus prone to recall biases.^15,16 Cognitive impairment after stroke further increases the odds of recall errors, thereby limiting the accuracy of retrospective measures.¹⁷ Psychosocial factors (eg, self-awareness and social desirability) also contribute to under- and overestimation of physical activity.^18,19 Laboratory-based tests, often conducted in a controlled or artificial environment, lack ecological validity and only provide a snapshot of functioning at specific times.^20,21

With technological advancement, ambulatory assessments such as accelerometry and ecological momentary assessment (EMA) have emerged to overcome these limitations. Accelerometry provides continuous and objective recording of activity patterns throughout the course of everyday life^22.23; however, it alone is unable to capture the contextual and biopsychological information about activity behavior. This information is important for understanding why people are active and inactive. EMA involves real-time, repeated sampling of a person’s mood, health-related symptoms and other experiences via electronic surveys in natural environments.^24,25 Compared with traditional survey methods, EMA enhances ecological validity, minimizes recall bias, and allows for investigations of dynamic patterns of behavioral change over time.^24,25 EMA overcomes the geographic and time barriers that prevail in traditional in-person assessments, reducing administrative costs and enhancing flexibility.²⁶ Thus, multimodal ambulatory monitoring, in which passive, objective accelerometry is coupled with active, subjective EMA, has potential to elucidate daily activities.^27,28

The promise of multimodal ambulatory monitoring is based on the presumption that it is feasible, valid, and acceptable. Thus far, we are aware of 1 study that validated the use of accelerometry for monitoring physical activity during inpatient stroke rehabilitation,²⁹ and another study validated the use of EMA for assessing symptoms in hospitalized patients with stroke.³⁰ A recent study supported the combined use of accelerometry and EMA in community-dwelling survivors of stroke, but it focused on the paretic arm and/or hand use instead of everyday activity participation.³¹

The primary aim of the study was to determine the feasibility, acceptability, and validity of multimodal ambulatory monitoring to assess daily activity and health-related symptoms among community-dwelling survivors of stroke. Secondarily, we sought to inform the potential for the research and clinical application of multimodal ambulatory monitoring in the population with stroke.

Methods

Participants

Survivors of stroke were recruited from a hospital database of individuals from an urban population in St. Louis living at home in their own community. Inclusion criteria included age 18-65 years, mild to moderate stroke (National Institutes of Health Stroke Scale≤15),³² at least 3 months post stroke, premorbid modified Rankin scale ≤2,³³ owning a smartphone, and being able to provide reliable yes/no responses.³⁴ Exclusion criteria were previous neurologic or psychiatric disorder, severe apraxia (Apraxia Screen of Test for Upper-Limb Apraxia <6),³⁵ evidence of neglect (Star Cancellation Test ≤44),³⁶ impaired vision (Lighthouse Near Visual Acuity Test ≤20/100 corrected vision),³⁷ conditions that preclude moderate to vigorous physical activity participation, and hearing impairment (sound repetition screens >4).³⁸ We excluded users of wheeled mobility devices because of using a thigh-worn accelerometer. The Institutional Review Board approved the study, and all participants provided written informed consent.

Study procedure

During the preambulatory monitoring visit, participants completed screening and assessments as part of a larger investigation. Participants installed the Participation in Everyday Life (PIEL) Survey App³⁹ on their smartphones for EMA and were fitted with an accelerometer on the anterior unaffected thigh using a waterproof dressing. Participants were instructed to wear the accelerometer continuously and received logs to document any accelerometer nonwear periods and daily wake and sleep times. Participants received training and an instruction manual about the PIEL Surveys App and the accelerometer and practiced under supervision until mastery was achieved.

Seven-day ambulatory monitoring began the following day. Participants were prompted with auditory signals and vibrations to complete 8 EMA surveys, randomly distributed at 2-hour intervals from 8 AM to 10 PM, for 7 consecutive days. Each survey took less than 2 minutes and stayed open for 30 minutes after the initial prompt, with up to 3 reminders every 10 minutes if no response was received. This sampling schedule aligns with the current literature^21,40-43 and allows a comprehensive capture of daily activities while meeting the recommendations for monitoring free-living physical activity in adults.^44-47 Participants were contacted on the second and fifth day to check in and encourage compliance. A helpline was available for reporting problems or making inquiries.

Upon completion, participants returned to the laboratory to return study equipment, transfer data, and complete additional assessments.

Measures

Accelerometry

The activPAL^a provides physical activity monitoring and has been used and validated in the population with stroke.^48-50 It provides an accurate classification of physical activity intensity in healthy adults.^51,52 It is a small and lightweight thigh-worn accelerometer with a dynamic range of ±2 gravitational units. It responds to gravitational accelerations from segmental movement⁵³ and captures acceleration at a sampling rate of 20 Hz. Proprietary algorithms in the activPAL software integrate the static and acceleration data to estimate energy expenditure (metabolic equivalent [MET]), with which the intensity of activity can be categorized as sedentary (1≤MET<2), light (2≤MET<3), or moderate to vigorous (MET≥3). The METs of EMA-reported daily activities were estimated based on the average METs captured in the 20-minute window (±10 minutes) around the EMA prompt when the activity was reported.^51,54,55 The activPAL has been shown to provide an accurate and valid estimation of physical activity intensity against direct observation in the free-living environment.⁵¹ To identify nonwear time, accelerometer data were screened for periods of nonwear based on participants’ wear logs, which were cross-checked against the heat map visualization of the individuals’ activPAL data.

EMA survey

The EMA survey captures daily activity and health-related symptoms. The daily activity item (“What are you doing?”) captures participants’ current activities from a list of 44 daily activities, developed based on validated EMA surveys^21,56 and the Activity Card Sort.⁵⁷ According to the Activity Card Sort, activities were categorized into 4 domains (instrumental activities of daily living [eg, shopping], high-physical-demand leisure activities [eg, gardening], low-physical-demand leisure activities [eg, watching TV], and social activities [eg, visiting family or friends]), and an additional domain of activities of daily living (eg, toileting) was included.

Six single-item questions assessed depressed mood (“Right now, I feel depressed”), cognitive complaints (“Right now, my thinking is slow”), cheerfulness (“Right now, I feel cheerful”), physical (“Physically, I feel exhausted now”) and mental fatigue (“Mentally, I feel exhausted now”), and pain (“What is your level of pain right now?”). These questions were selected from the Patient Health Questionnaire-9 (PHQ-9),⁵⁸ Quality of Life in Neurological Disorders (Neuro-QoL) Cognition Function,⁵⁹ Fatigue Assessment Scale,⁶⁰ and Patient-Reported Outcomes Measurement Information System Pain Intensity.⁶¹ Pain was rated on a slider scale from 0 (no pain) to 10 (worst pain); others were rated on a 5-point scale from 1 (not at all) to 5 (very much).

Acceptability questionnaire

At the postambulatory monitoring visit, participants completed a survey, adapted from a previous study,¹¹ assessing study acceptability (eg, overall experience as a participant), multimodal ambulatory monitoring acceptability (eg, ease of using mobile surveys/activity monitor), and reactivity (ie, alteration of behavior or feelings because of multimodal ambulatory monitoring).

Additional measures

Participants completed a battery of measures, including the PHQ-960 for assessing depression severity, the Neuro-QoL Cognition Function Short Form,⁵⁹ the Fatigue Assessment Scale,⁶⁰ the Patient-Reported Outcomes Measurement Information System Pain intensity,⁶¹ the International Physical Activity Questionnaire–Long Form,⁶² the Pittsburgh Sleep Quality Index,⁶³ and the Charlson Comorbidity Index.⁶⁴

Statistical analyses

Feasibility

Feasibility was evaluated based on the attrition rate and participants’ compliance with multimodal ambulatory monitoring. Accelerometer compliance was based on wear time and the number of valid days of accelerometer wearing over the entire monitoring period. A valid day of accelerometer wearing was defined as having at least 10 hours of wear time during waking hours⁶⁵; EMA compliance was based on the EMA completion rate; multimodality compliance was based on the percentage of completed EMA surveys accompanied by valid accelerometer data.

Acceptability

Acceptability was based on descriptive analyses of participants’ responses to the acceptability questionnaire.

Validity

Convergent and discriminant validity were based on the correlations between multimodal ambulatory monitoring and the laboratory-based measures of the same and opposing constructs. Direct correspondence was not anticipated given the differences in real-time vs retrospective; however, there should be some degree of concordance between measures of the same or opposing construct.⁶⁶ Criterion validity was based on the concordance between accelerometer-measured activity intensity and domain-specific daily activities assessed by EMA. To examine measurement reactivity, we compared the accelerometer data 10 minutes before and after the EMA prompt using independent t tests. A significant difference would suggest that the act of responding to EMA may have altered the ongoing activity.

Sample size consideration

A correlation coefficient of 0.50 or higher is expected between ambulatory monitoring and the laboratory-based measures of the same and opposing constructs.^21,56,67 Based on a power analysis using G*power 3.70,⁶⁸ a sample size of 29 was estimated to have 80% power to detect a correlation coefficient of 0.50. The study sample size of 40 had 91.5% power to detect a correlation coefficient of 0.5, with 2-sided α=0.05.

Results

Forty survivors of stroke were enrolled for the study. As shown in table 1, on average, participants were 62.8 years of age and mainly employed (n=25, 62.5%), had 14.7 years of education, and were distributed similarly across sex, race, and marital status. The majority were diagnosed with an ischemic stroke (n=28, 70.0%) and were 2083.3 days post stroke.

Table 1.

Participant characteristics

Characteristic	Total (N=40)	Range
Age, mean ± SD	52.80±7.48	35-65
Years of education, mean ± SD	14.73±2.21	10-20
Sex, n (%)
Female	17 (42.5)
Male	23 (57.5)
Race, n (%)
White	22 (55.0)
Black	18 (45.0)
Marital status, n (%)
Married	19 (47.5)
Unmarried	21 (52.5)
Employment, n (%)
Employed	25 (62.5)
Non-employed	15 (37.5)
Stroke type, n (%)
Ischemic	28 (70.0)
Hemorrhagic	12 (30.0)
CCI, mean ± SD*	4.15±2.96	1-14
Time post stroke (d), mean ± SD	2083.25±1168.00	399-4325
Premorbid mRS, mean ± SD	0.10±0.38	0-2
Admission NIHSS, n (%)
0-5 Mild stroke	33 (82.5)
6-15 Moderate stroke	7 (17.5)
Current NIHSS, n (%)*
0-5 Mild stroke	40 (100)

Abbreviations: CCI, Charlson Comorbidity Index;mRS, Modified Rankin Scale;NIHSS, National Institutes of Health Stroke Scale.

^*At the time of study.

Feasibility

Participants completed the study with no attrition. The average EMA completion rate ranged from 76.8%-100%, with a mean of 93.6%. Twenty percent of the participants completed 100% of the scheduled EMA surveys, and 70% completed ≥90%. All participants achieved 7 valid days of wear, with an accelerometer compliance rate of 99.7%. Seventy-seven percent of the participants never removed the accelerometer during the study protocol. Regarding multimodality compliance, 99.8% of completed EMA surveys were accompanied by valid accelerometer data.

Acceptability

Figure 1 presents the results of the acceptability questionnaire (for descriptive statistics, see supplemental table S1). Nearly all participants (97.5%) rated their overall study experience as positive/very positive. Most participants were highly satisfied with and confident in using the study technology (mean satisfaction, 4.8/5; mean confidence, 4.7/5) and preferred it over traditional methods (ie, retrospective self-reports) (mean preference, 4.7/5). All participants rated the accelerometer as easy/very easy to use, and 97.5% rated the EMA as easy/very easy to use. Participants reported low interference of EMA (mean interference, 1.9/5) and accelerometer (mean interference, 1.3/5) with their daily routine.

Fig 1.

Acceptability ratings of multimodal ambulatory monitoring.

Validity

Table 2 reports the correlations between ambulatory and laboratory-based measures. Significant positive and negative correlations were observed in measures of the same and opposing constructs, respectively. For instance, the moderate to vigorous physical activity time captured by the accelerometer was positively correlated with the moderate to vigorous physical activity time assessed by the International Physical Activity Questionnaire-Long Form (r=0.51, P=.001) but negatively correlated with sitting time (r=−0.53, P<.001). Likewise, EMA-reported depressed mood (r=0.56, P<.001) was positively correlated with the PHQ-9, whereas cheerfulness was negatively correlated (r=−0.38, P=.016). The scatter plot matrixes of correlations are available in supplemental figs S1 and S2, and the descriptive statistics of the variables are available in supplemental table S2.

Table 2.

Correlations between multimodal ambulatory monitoring and traditional retrospective measures

Measure Traditional Measure	Accelerometer-Measured Physical Activity
	MVPA (min/wk)	Light PA (min/wk)	Sedentary Behavior (min/wk)	Total MET (min/wk)
IPAQ MVPA (min/wk)	0.51*	0.43*	−0.11	0.47*
IPAQ Sitting (min/wk)	−0.53†	−0.39‡	0.45*	−0.29
IPAQ total MET (min/wk)	0.47*	0.50*	−0.06	0.49*

Traditional measure	EMA-Reported Health-Related Symptom
	Depressed mood	Cheerfulness	Physical Fatigue	Mental Fatigue	Pain	Cognitive Complaint	Sleep Quality
PHQ-9	0.56†	−0.38‡	0.52*	0.53*	0.60†	0.41*	−0.56†
FAS (physical)	0.28	−0.24	0.40‡	0.32‡	0.39‡	0.37‡	−0.35‡
FAS (mental)	0.34‡	−0.29	0.52*	0.54†	0.37‡	0.33‡	−0.34‡
PROMIS Pain intensity	0.38‡	−0.14	0.51*	0.50*	0.72†	0.45*	−0.44*
NQ Cognitive function	−0.43*	0.27	−0.66†	−0.68†	−0.48*	−0.66†	0.27
PSQI Global score	0.39‡	−0.30	0.43*	0.40‡	0.58†	0.22	−0.52*

NOTE. For EMA items, mean scores across the monitoring period were aggregated for each individual.

Abbreviations: FAS, Fatigue Assessment Scale; IPAQ, International Physical Activity Questionnaire; MVPA, moderate to vigorous physical activity; NQ, Quality of Life in Neurological Disorders; PA, physical activity; PROMIS, Patient-Reported Outcomes Measurement Information System; PSQI, Pittsburgh Sleep Quality Index.

^*P<.01.

^†P<.001.

^‡P<.05.

Table 3 presents the accelerometer-measured physical activity intensity of EMA-reported daily activities. A higher portion of light (40%) and moderate to vigorous(10%) physical activity was recorded when participants reported doing more physically demanding activities (ie, high-physical-demand leisure activities), whereas a higher proportion of sedentary behavior (93%) was captured by the accelerometer when participants were engaged in less physically demanding activities (ie, low-physical demand leisure activities). Follow-up independent t tests with Bonferroni adjustment (α=0.0125 [0.05/4]) indicated that high-physical-demand leisure activities (1.79±0.79METs) had significantly higher METs than other activity domains, whereas low-physical-demand leisure activities(1.33±0.13METs) had significantly lower METs than the others(P<.001).

Table 3.

Physical activity intensity of domain-specific daily activities

EMA-Reported Daily Activity	Accelerometer-Measured Physical Activity Intensity
	Sedentary, n (%)	Light, n (%)	Moderate to Vigorous, n (%)	MET Mean ± SD
Activities of daily living	210 (78)	60 (22)	0 (0)	1.41±0.16
Instrumental activities of daily living	475 (65)	258 (35)	0 (0)	1.47±0.22
Low-physical-demand leisure activities	629 (93)	51 (8)	0 (0)	1.33±0.13
High-physical-demand leisure activities	53 (50)	42 (40)	10 (10)	1.79±0.79
Social activities	234 (77)	71 (23)	0 (0)	1.40±0.16

Independent t tests found no significant changes in the accelerometer data recorded before vs after responding to an EMA survey (t[4163]=1.04, P=.299).

Discussion

Advancing the use of technology in rehabilitation assessment is a priority goal of the National Institutes of Health Research Plan on Rehabilitation.⁶⁹ Mobile health technologies, including telecommunication follow-up, permit the study of individuals’ lived experience in their natural environments (urban and rural)^70-72 and thus inform the development of interventions to support continued recovery after discharge. Multimodal ambulatory monitoring combining 2 mobile technologies, accelerometry and EMA, has potential to expand our understanding of changes in activity patterns in relation to contextual and biopsychological factors. Our findings suggest that multimodal ambulatory monitoring is feasible, valid, and acceptable to monitor activity behavior and health-related symptoms in community-dwelling survivors of stroke.

All participants completed the study and displayed high compliance with EMA (93.6%) and accelerometry (99.7%), and 99.8% of the EMA surveys were completed in conjunction with valid accelerometer recordings. The high compliance rate is consistent with studies using similar combined methodology^31,73,74 and can be attributed to a number of facilitating factors inherent in the study design, including the training session for the study technology, user instruction manual, helpline, short EMA survey, check-in calls, and study compensation. Different from most EMA studies, the iOS- and Android-compatible PIEL Survey App allowed participants to use their smartphones for the study, which provided familiarity and convenience and eliminated the burden of carrying an extra assessment device. Accelerometer compliance was enhanced by the waterproof dressings, minimizing the need to take off the accelerometer for water activities. Moreover, nearly all participants rated the number of days to wear the accelerometer and the number of daily EMA surveys as not too many or too few. Taken together, the high retention and compliance suggest that the 7-day multimodal ambulatory monitoring was feasible and not overly burdensome for survivors of stroke.

Concurrent and discriminant validity were supported by the significant associations between multimodal ambulatory monitoring and the measures of the same and opposing constructs in the expected directions. As expected, we observed a trend of stronger associations between pairs measuring the same construct. For instance, EMA-reported cognitive complaints were correlated with Neuro-QoL Cognitive Function (r=−0.66, P<.001) more than other measures (r=0.22-0.45). Criterion validity was supported by the concordance between EMA-reported daily activities and the accelerometer data. Moderate to vigorous physical activity was captured by the accelerometer only when participants reported doing more physically demanding activities (ie, high-physical-demand leisure activities). In contrast, sedentary activity (93%) was captured when participants reported doing less physically demanding activities (ie, low-physical-demand leisure activities). The concordance between EMA-reported and accelerometer-measured daily activity suggests that survivors of stroke were capable of and accurate in reporting their current activity via EMA and that accelerometry was sensitive to differences in activity patterns associated with different daily activities. We also found no significant changes in accelerometer data before and after the EMA prompt, suggesting that the act of answering EMA did not influence the participants’ ongoing activity.

Acceptability was supported by positive feedback obtained from the acceptability questionnaire. Despite slight interference with daily routine, participants reported a very positive study experience and preferred ambulatory monitoring over other types of measures. Participants rated both EMA and the accelerometer as easy to use, which might explain their high satisfaction and confidence in using the multimodal ambulatory monitoring as well as their high compliance. Both the PIEL Survey App and the activPAL used in the current study were simple and required minimal handling: the PIEL Survey App runs automatically in the background and requires only one tap to enter the survey, whereas the activPAL was rated by the participants as very comfortable and has a battery that does not require charging during the 7 days. Regarding reactivity, all participants indicated that responding to EMA about symptoms did not influence their original symptoms. The majority (72.5%) indicated that wearing the accelerometer did not make them increase or decrease their physical activity; however, 10 participants (25%) indicated that wearing it made them aware of their physical activity and encouraged them to be more physically active.

Multimodal ambulatory monitoring can be of great value to research and clinical practice. It enables a more complete understanding of physical activity exerted during everyday life activities by supplementing movement data with contextual and biopsychological information that describes peoples’ behavior.²⁸ The high sampling rate and ecological validity of multimodal ambulatory monitoring provide the abundance, precision, and temporal resolution of data needed for conducting advanced analyses (eg, multilevel modeling, machine learning, network analysis) to unravel the dynamics of behavior change within a person. Such information ultimately informs the development of mobile or just-in-time adaptive interventions.⁷⁵

Study limitations

This study has limitations. Our sample only included adults with mild stroke and those younger than 65 years, limiting the generalizability of our findings to those with more severe stroke and elderly persons, who might face greater difficulty using the study technology. Future research should include and examine a broader population with stroke. Future studies should include users of wheeled devices and explore the use of wrist-worn accelerometers among them. The validity of the activPAL for classifying physical activity intensity has not been examined in the population with stroke and may result in over- or underestimation. Future studies should validate the activPAL because it offers a simple and usable tool to monitor the active and sedentary behaviors of survivors of stroke who need to increase their activity level to promote health. Moreover, we used interval-contingent EMA, which might fail to capture rare behaviors and the peaks of physical activity. An interactive approach, in which an EMA prompt is triggered when the phenomenon of interest (eg, physical activity) exceeds a certain threshold can be used to address this limitation and enhance the variation of data.²⁷ Additionally, all participants were smartphone users, whose preexisting experience likely facilitated their use of the study technology compared with someone who does not have a smartphone.

Conclusions

Multimodal ambulatory monitoring that combines accelerometry with EMA is a feasible, valid, and acceptable approach for monitoring daily activity and health-related symptoms among community-dwelling survivors of mild stroke. Multimodal ambulatory monitoring has potential to provide a more complete understanding of daily activity. The findings of this study provide the framework for the future development of mobile interventions to promote healthy behavior change and support continued recovery in the community.

Supplier

a. activPAL; PAL Technologies Ltd, Glasgow, Scotland.

List of abbreviations:

EMA

ecological momentary assessment

MET

metabolic equivalent

Neuro-QoL

Quality of Life in Neurological Disorders

PHQ-9

Patient Health Questionnaire-9

PIEL

Participation in Everyday Life

PSQI

Pittsburgh Sleep Quality Index