HHS Public Access Author manuscript Author Manuscript
J Neuropsychiatry Clin Neurosci. Author manuscript; available in PMC 2017 July 01. Published in final edited form as:
J Neuropsychiatry Clin Neurosci. 2016 ; 28(3): 199–204. doi:10.1176/appi.neuropsych.15090235.
Quantifying post-stroke apathy with actimeters Andrew M. Goldfine, M.D., M.Sc., Dept. of Neurology, Stony Brook Medicine Behdad Dehbandi, M.Sc., Burke Medical Research Institute, White Plains, NY and Weill Cornell Medical College of Cornell University, New York, NY
Author Manuscript
Juliana M. Kennedy, B.S., Burke Medical Research Institute, White Plains, NY Briana Sabot, Burke Medical Research Institute, White Plains, NY Cory Semper, M.E., and Burke Medical Research Institute, White Plains, NY David Putrino, P.T., Ph.D. Burke Medical Research Institute, White Plains, NY and Weill Cornell Medical College of Cornell University, New York, NY
Author Manuscript
Abstract The authors tested the hypothesis that wrist-worn actimeters can quantify the severity of poststroke apathy. The authors studied 57 patients admitted to an acute rehabilitation unit for ischemic or hemorrhagic stroke. After accounting for motor deficit of the affected arm and age, each increment of Apathy Inventory correlated with 5.6 fewer minutes moving per hour. The overall statistical model had an R2 of only 0.341, suggesting unexplained factors for total movement time. Wrist-worn actimeters may serve as an objective, quantifiable measure of post-stroke apathy in patients with an intact upper extremity, but cannot alone be used to diagnose apathy.
Keywords Stroke; rehabilitation; geriatric neuropsychiatry
Author Manuscript
Introduction Apathy occurs in approximately 35% of patients after stroke and is associated with more disability and lower quality of life {1–4}. Apathy is defined as a reduction in goal-directed behavior and flattening of emotions in comparison with the patient’s prior degree of
Address for Reprints: Andrew M. Goldfine, MD, Health Sciences Center T12-020, Department of Neurology, Stony Brook Medicine, Stony Brook, New York 11794-8121, [email protected]. Location of work: Burke Rehabilitation Hospital, 785 Mamaroneck Ave, White Plains, NY
Goldfine et al.
Page 2
Author Manuscript
function, not due to an impairment of arousal or medication side effect {5}. It is typically defined using scales, based on patient report or clinician observation, all of which involve subjective interpretation {6}. As in most behavioral disorders, there is currently no objective, quantifiable definition of apathy. This serves as a limitation for the conduct of observational and clinical trial research, as subjective measures limit inter-rater reliability and the interpretation of outcomes (e.g., quantifying change over time).
Author Manuscript
One approach to quantify human behavior is the use of actimeters. These are small, typically wireless, accelerometers or gyroscopes that can be secured to subjects’ bodies to record movement, typically for many days at a time. They have been used to quantify total movement, inter-limb patterns, and body position with the potential for secondary analyses including sleep duration, circadian rhythms, and steps taken per day (reviewed in {7}; recent examples in stroke recovery are {8; 9}). Actimeters are a promising tool for clinical research because they are affordable, user-friendly, non-invasive, and produce outputs that are ecologically valid and linearly quantifiable, unlike many standard clinical research tools. Actimeters have already been used in studies investigating apathy in order to determine if there is an association between apathy and the amount of time that subjects move {10; 11}. These studies were performed in patients with a variety of brain injuries, as well as in Alzheimer disease, and found that apathetic patients generally move less during the day. These findings may not be extrapolated to stroke survivors, as post-stroke motor dysfunction directly affects movement ability. Thus, metrics associated with quantity of movement may reflect stroke severity rather than apathy severity.
Author Manuscript
Here, we asked whether amount of movement measured by a wrist-watch style actimeter on the unaffected wrist correlates with severity of apathy in patients with stroke. We studied patients undergoing acute rehabilitation as this setting provides a controlled environment with similar daily demands placed on all patients. This study aimed to determine the potential utility of actimeters as a measure of apathy severity, and therefore their potential utility in future clinical trials. We hypothesized that subjects with apathy would move less throughout the day. We also performed exploratory analyses on the correlation of amount of movement with recovery during rehabilitation.
Methods Subjects
Author Manuscript
From July 2014 to May 2015 we studied patients admitted to an acute rehabilitation hospital for deficits due to an ischemic or hemorrhagic stroke. To ensure that the actimeter values were not affected by disorders of movement, we did not study patients with bilateral upper extremity weakness (one arm had to have a Fugl-Meyer score of 66 out of 66), or movement disorders such as Parkinson disease. We also excluded patients where we could not diagnose apathy severity including those on sedating or anti-psychotic medications during the day, hypoarousal such as from an infection or sleep disturbance, or active psychiatric disease other than depression. The study was approved by the Burke Rehabilitation Hospital Committee for Human Rights in Research, and all subjects or their legally authorized representatives gave signed informed consent.
J Neuropsychiatry Clin Neurosci. Author manuscript; available in PMC 2017 July 01.
Goldfine et al.
Page 3
Behavioral outcome measures
Author Manuscript
Apathy was quantified using the Apathy Inventory Clinician Version (AI-C) {12}, scored weekly by the subjects’ treating speech therapist as part of standard of care. The AI-C asks the clinician to report on: (1) emotional blunting; (2) loss of initiative (i.e., goal-directed behavior); and (3) loss of interest in activities and other people. Each question is scored 0 = no problem to 4 = major problem, with a total score ranging 0–12. Scores of 4 and above are considered clinically significant apathy. Scores were based on behaviors observed over the previous week. We report the AI-C score closest to the time of the actimetry recording. To quantify depressive symptoms, we interviewed patients with the Montgomery-Asberg Depression Rating Scale (MADRS; 0–60 with higher scores representing worse depression) {13}. Note that this was only performed on the subset of patients with intact communication ability (mild or no aphasia as judged by treating speech pathologist).
Author Manuscript
Motor dysfunction was assessed with the upper extremity Fugl-Meyer motor scale by subjects’ treating occupational therapists {14; 15}. This scale assesses reflexes, strength, and dexterity with scores ranging 0 (flaccid paralysis) to 66 (full strength). Disability was assessed at admission and discharge by treating therapists using the FIM™ (UB Foundation Activities, Inc.). Scores range 18–126 with lower scores representing more disability.
Author Manuscript
For the multiple regression analysis, the Fugl-Meyer upper limb assessment was used as a measure of stroke severity. The Fugl-Meyer was chosen in place of the FIM because FuglMeyer is an impairment measure, which means it is less affected by other factors. For example, we have observed that apathy can directly affect many of the measures within the FIM (e.g., apathetic patients are scored lower on language measures because they do not initiate speech, even though their language skills are intact). Actimeter Quantity of movement was assessed using a wrist-watch style actimeter (GENEActiv, Activinsights Ltd, Kimbolton, UK) that uses a tri-axial accelerometer. It was set to record acceleration at 40 Hz with all data recorded onto the device in units of milli-g (1000 milli-g = 1 g = 9.81 m/s2). Subjects wore the actimeter on the wrist of an upper extremity with intact strength (normal Fugl-Meyer score of 66). Data were collected for at least 36 hours (two nights and the intervening day), though only analyzed from the day and the second night. We documented the schedule of therapy sessions and meals for all subjects during the day. Actimeter data processing
Author Manuscript
Acceleration values were extracted from the device and analyzed further in custom Matlab scripts (Matlab 2012b, Mathworks, USA) unless otherwise noted. In order to minimize the effect of gravity on acceleration values, we removed slowly changing acceleration values (greater than approximately two seconds) using a detrending algorithm with a moving window (“locdetrend” from chronux.org; {16}). We then converted the three axes into one by calculating the total euclidean distance of all axes from the origin. Finally, we binarized all acceleration values empirically into movement (>0.3 milli-g) or not movement. This
J Neuropsychiatry Clin Neurosci. Author manuscript; available in PMC 2017 July 01.
Goldfine et al.
Page 4
Author Manuscript
binarization removed the influence of movement amplitude from the final analyses as we were interested in amount of time moving rather than its amplitude. Results are all displayed as “total movement time” meaning the amount of time where the acceleration value was considered to be movement versus not movement. Statistical analyses Statistical analyses were also performed in Matlab using built in and Statistics Toolbox codes. Scatter plots were produced with the code “scatter.m” after confirming that there were no overlapping data points. As some data were not Gaussian distributed, we report medians and interquartile ranges, and compare groups using the Wilcoxon Rank Sum Test. Boxplots were created using Matlab’s “boxplot.m” with default parameters. Multiple linear regression analysis was performed using Matlab’s “linearmodel.fit” with default parameters.
Author Manuscript
Results Subjects studied 57 subjects consented and completed data acquisition; we collected data from an additional 14 subjects, but in 11 subjects, these data were not used because the subjects removed the actimeters prior to study completion, and device error occurred in 3 subjects. Of the 11 subjects who removed the device, 4 of them had apathy. Of the 57 subjects studied, 33 had sufficient communication ability to perform the MADRS interview for depression symptoms.
Author Manuscript
Subjects had a wide range of apathy scores (Figure 1) with 21 (36%) having apathy (defined as AI-C ≥ 4) and the remaining 36 (44%) without apathy. Note that this does not reflect the demographics of the population at this hospital as reported in {3}, but rather reflects our attempt to obtain data on a similar number of subjects with versus without apathy. Apathetic and non-apathetic subjects were well matched on age and gender, but those with apathy were generally more disabled from their stroke and had less recovery during their stay in rehabilitation (Table 1). Correlation of apathy severity versus quantity of movement Our primary aim was to investigate whether there was a correlation between apathy severity and amount of movement during the hours of 9AM–5PM (Figure 2). During this time subjects typically participated in physical therapy, occupational therapy, speech therapy, meals and free time. We found that apathy severity inversely correlated with total movement per hour (Pearson r = −0.49, p
J Neuropsychiatry Clin Neurosci. Author manuscript; available in PMC 2017 July 01. Published in final edited form as:
J Neuropsychiatry Clin Neurosci. 2016 ; 28(3): 199–204. doi:10.1176/appi.neuropsych.15090235.
Quantifying post-stroke apathy with actimeters Andrew M. Goldfine, M.D., M.Sc., Dept. of Neurology, Stony Brook Medicine Behdad Dehbandi, M.Sc., Burke Medical Research Institute, White Plains, NY and Weill Cornell Medical College of Cornell University, New York, NY
Author Manuscript
Juliana M. Kennedy, B.S., Burke Medical Research Institute, White Plains, NY Briana Sabot, Burke Medical Research Institute, White Plains, NY Cory Semper, M.E., and Burke Medical Research Institute, White Plains, NY David Putrino, P.T., Ph.D. Burke Medical Research Institute, White Plains, NY and Weill Cornell Medical College of Cornell University, New York, NY
Author Manuscript
Abstract The authors tested the hypothesis that wrist-worn actimeters can quantify the severity of poststroke apathy. The authors studied 57 patients admitted to an acute rehabilitation unit for ischemic or hemorrhagic stroke. After accounting for motor deficit of the affected arm and age, each increment of Apathy Inventory correlated with 5.6 fewer minutes moving per hour. The overall statistical model had an R2 of only 0.341, suggesting unexplained factors for total movement time. Wrist-worn actimeters may serve as an objective, quantifiable measure of post-stroke apathy in patients with an intact upper extremity, but cannot alone be used to diagnose apathy.
Keywords Stroke; rehabilitation; geriatric neuropsychiatry
Author Manuscript
Introduction Apathy occurs in approximately 35% of patients after stroke and is associated with more disability and lower quality of life {1–4}. Apathy is defined as a reduction in goal-directed behavior and flattening of emotions in comparison with the patient’s prior degree of
Address for Reprints: Andrew M. Goldfine, MD, Health Sciences Center T12-020, Department of Neurology, Stony Brook Medicine, Stony Brook, New York 11794-8121, [email protected]. Location of work: Burke Rehabilitation Hospital, 785 Mamaroneck Ave, White Plains, NY
Goldfine et al.
Page 2
Author Manuscript
function, not due to an impairment of arousal or medication side effect {5}. It is typically defined using scales, based on patient report or clinician observation, all of which involve subjective interpretation {6}. As in most behavioral disorders, there is currently no objective, quantifiable definition of apathy. This serves as a limitation for the conduct of observational and clinical trial research, as subjective measures limit inter-rater reliability and the interpretation of outcomes (e.g., quantifying change over time).
Author Manuscript
One approach to quantify human behavior is the use of actimeters. These are small, typically wireless, accelerometers or gyroscopes that can be secured to subjects’ bodies to record movement, typically for many days at a time. They have been used to quantify total movement, inter-limb patterns, and body position with the potential for secondary analyses including sleep duration, circadian rhythms, and steps taken per day (reviewed in {7}; recent examples in stroke recovery are {8; 9}). Actimeters are a promising tool for clinical research because they are affordable, user-friendly, non-invasive, and produce outputs that are ecologically valid and linearly quantifiable, unlike many standard clinical research tools. Actimeters have already been used in studies investigating apathy in order to determine if there is an association between apathy and the amount of time that subjects move {10; 11}. These studies were performed in patients with a variety of brain injuries, as well as in Alzheimer disease, and found that apathetic patients generally move less during the day. These findings may not be extrapolated to stroke survivors, as post-stroke motor dysfunction directly affects movement ability. Thus, metrics associated with quantity of movement may reflect stroke severity rather than apathy severity.
Author Manuscript
Here, we asked whether amount of movement measured by a wrist-watch style actimeter on the unaffected wrist correlates with severity of apathy in patients with stroke. We studied patients undergoing acute rehabilitation as this setting provides a controlled environment with similar daily demands placed on all patients. This study aimed to determine the potential utility of actimeters as a measure of apathy severity, and therefore their potential utility in future clinical trials. We hypothesized that subjects with apathy would move less throughout the day. We also performed exploratory analyses on the correlation of amount of movement with recovery during rehabilitation.
Methods Subjects
Author Manuscript
From July 2014 to May 2015 we studied patients admitted to an acute rehabilitation hospital for deficits due to an ischemic or hemorrhagic stroke. To ensure that the actimeter values were not affected by disorders of movement, we did not study patients with bilateral upper extremity weakness (one arm had to have a Fugl-Meyer score of 66 out of 66), or movement disorders such as Parkinson disease. We also excluded patients where we could not diagnose apathy severity including those on sedating or anti-psychotic medications during the day, hypoarousal such as from an infection or sleep disturbance, or active psychiatric disease other than depression. The study was approved by the Burke Rehabilitation Hospital Committee for Human Rights in Research, and all subjects or their legally authorized representatives gave signed informed consent.
J Neuropsychiatry Clin Neurosci. Author manuscript; available in PMC 2017 July 01.
Goldfine et al.
Page 3
Behavioral outcome measures
Author Manuscript
Apathy was quantified using the Apathy Inventory Clinician Version (AI-C) {12}, scored weekly by the subjects’ treating speech therapist as part of standard of care. The AI-C asks the clinician to report on: (1) emotional blunting; (2) loss of initiative (i.e., goal-directed behavior); and (3) loss of interest in activities and other people. Each question is scored 0 = no problem to 4 = major problem, with a total score ranging 0–12. Scores of 4 and above are considered clinically significant apathy. Scores were based on behaviors observed over the previous week. We report the AI-C score closest to the time of the actimetry recording. To quantify depressive symptoms, we interviewed patients with the Montgomery-Asberg Depression Rating Scale (MADRS; 0–60 with higher scores representing worse depression) {13}. Note that this was only performed on the subset of patients with intact communication ability (mild or no aphasia as judged by treating speech pathologist).
Author Manuscript
Motor dysfunction was assessed with the upper extremity Fugl-Meyer motor scale by subjects’ treating occupational therapists {14; 15}. This scale assesses reflexes, strength, and dexterity with scores ranging 0 (flaccid paralysis) to 66 (full strength). Disability was assessed at admission and discharge by treating therapists using the FIM™ (UB Foundation Activities, Inc.). Scores range 18–126 with lower scores representing more disability.
Author Manuscript
For the multiple regression analysis, the Fugl-Meyer upper limb assessment was used as a measure of stroke severity. The Fugl-Meyer was chosen in place of the FIM because FuglMeyer is an impairment measure, which means it is less affected by other factors. For example, we have observed that apathy can directly affect many of the measures within the FIM (e.g., apathetic patients are scored lower on language measures because they do not initiate speech, even though their language skills are intact). Actimeter Quantity of movement was assessed using a wrist-watch style actimeter (GENEActiv, Activinsights Ltd, Kimbolton, UK) that uses a tri-axial accelerometer. It was set to record acceleration at 40 Hz with all data recorded onto the device in units of milli-g (1000 milli-g = 1 g = 9.81 m/s2). Subjects wore the actimeter on the wrist of an upper extremity with intact strength (normal Fugl-Meyer score of 66). Data were collected for at least 36 hours (two nights and the intervening day), though only analyzed from the day and the second night. We documented the schedule of therapy sessions and meals for all subjects during the day. Actimeter data processing
Author Manuscript
Acceleration values were extracted from the device and analyzed further in custom Matlab scripts (Matlab 2012b, Mathworks, USA) unless otherwise noted. In order to minimize the effect of gravity on acceleration values, we removed slowly changing acceleration values (greater than approximately two seconds) using a detrending algorithm with a moving window (“locdetrend” from chronux.org; {16}). We then converted the three axes into one by calculating the total euclidean distance of all axes from the origin. Finally, we binarized all acceleration values empirically into movement (>0.3 milli-g) or not movement. This
J Neuropsychiatry Clin Neurosci. Author manuscript; available in PMC 2017 July 01.
Goldfine et al.
Page 4
Author Manuscript
binarization removed the influence of movement amplitude from the final analyses as we were interested in amount of time moving rather than its amplitude. Results are all displayed as “total movement time” meaning the amount of time where the acceleration value was considered to be movement versus not movement. Statistical analyses Statistical analyses were also performed in Matlab using built in and Statistics Toolbox codes. Scatter plots were produced with the code “scatter.m” after confirming that there were no overlapping data points. As some data were not Gaussian distributed, we report medians and interquartile ranges, and compare groups using the Wilcoxon Rank Sum Test. Boxplots were created using Matlab’s “boxplot.m” with default parameters. Multiple linear regression analysis was performed using Matlab’s “linearmodel.fit” with default parameters.
Author Manuscript
Results Subjects studied 57 subjects consented and completed data acquisition; we collected data from an additional 14 subjects, but in 11 subjects, these data were not used because the subjects removed the actimeters prior to study completion, and device error occurred in 3 subjects. Of the 11 subjects who removed the device, 4 of them had apathy. Of the 57 subjects studied, 33 had sufficient communication ability to perform the MADRS interview for depression symptoms.
Author Manuscript
Subjects had a wide range of apathy scores (Figure 1) with 21 (36%) having apathy (defined as AI-C ≥ 4) and the remaining 36 (44%) without apathy. Note that this does not reflect the demographics of the population at this hospital as reported in {3}, but rather reflects our attempt to obtain data on a similar number of subjects with versus without apathy. Apathetic and non-apathetic subjects were well matched on age and gender, but those with apathy were generally more disabled from their stroke and had less recovery during their stay in rehabilitation (Table 1). Correlation of apathy severity versus quantity of movement Our primary aim was to investigate whether there was a correlation between apathy severity and amount of movement during the hours of 9AM–5PM (Figure 2). During this time subjects typically participated in physical therapy, occupational therapy, speech therapy, meals and free time. We found that apathy severity inversely correlated with total movement per hour (Pearson r = −0.49, p
