Smartphone GPS signatures of patients undergoing spine surgery correlate with mobility and current gold standard outcome measures

Alessandro Boaro Computational Neuroscience Outcomes Center, Department of Neurosurgery, Brigham and Women’s Hospital, Harvard Medical School;
Institute of Neurosurgery, Department of Neurosciences, Biomedicine and Movement Sciences, University of Verona, Italy; and

Search for other papers by Alessandro Boaro in
jns
Google Scholar
PubMed
Close
 MD
,
Jeffrey Leung Computational Neuroscience Outcomes Center, Department of Neurosurgery, Brigham and Women’s Hospital, Harvard Medical School;

Search for other papers by Jeffrey Leung in
jns
Google Scholar
PubMed
Close
 BA
,
Harrison T. Reeder Department of Biostatistics, Harvard T.H. Chan School of Public Health;

Search for other papers by Harrison T. Reeder in
jns
Google Scholar
PubMed
Close
 AM
,
Francesca Siddi Computational Neuroscience Outcomes Center, Department of Neurosurgery, Brigham and Women’s Hospital, Harvard Medical School;

Search for other papers by Francesca Siddi in
jns
Google Scholar
PubMed
Close
 MD
,
Elisabetta Mezzalira Computational Neuroscience Outcomes Center, Department of Neurosurgery, Brigham and Women’s Hospital, Harvard Medical School;

Search for other papers by Elisabetta Mezzalira in
jns
Google Scholar
PubMed
Close
 RN, MSc
,
Gang Liu Department of Biostatistics, Harvard T.H. Chan School of Public Health;

Search for other papers by Gang Liu in
jns
Google Scholar
PubMed
Close
 MSc
,
Rania A. Mekary Computational Neuroscience Outcomes Center, Department of Neurosurgery, Brigham and Women’s Hospital, Harvard Medical School;
School of Pharmacy, MCPHS University, Boston, Massachusetts;

Search for other papers by Rania A. Mekary in
jns
Google Scholar
PubMed
Close
 PhD, MSc
,
Yi Lu Department of Neurosurgery, Brigham and Women’s Hospital, Harvard University, Boston, Massachusetts

Search for other papers by Yi Lu in
jns
Google Scholar
PubMed
Close
 MD, PhD
,
Michael W. Groff Department of Neurosurgery, Brigham and Women’s Hospital, Harvard University, Boston, Massachusetts

Search for other papers by Michael W. Groff in
jns
Google Scholar
PubMed
Close
 MD
,
Jukka-Pekka Onnela Department of Biostatistics, Harvard T.H. Chan School of Public Health;

Search for other papers by Jukka-Pekka Onnela in
jns
Google Scholar
PubMed
Close
 DSc
, and
Timothy R. Smith Computational Neuroscience Outcomes Center, Department of Neurosurgery, Brigham and Women’s Hospital, Harvard Medical School;
Department of Neurosurgery, Brigham and Women’s Hospital, Harvard University, Boston, Massachusetts

Search for other papers by Timothy R. Smith in
jns
Google Scholar
PubMed
Close
 MD, PhD, MPH
Free access

OBJECTIVE

Patient-reported outcome measures (PROMs) are currently the gold standard to evaluate patient physical performance and ability to recover after spine surgery. However, PROMs have significant limitations due to the qualitative and subjective nature of the information reported as well as the impossibility of using this method in a continuous manner. The smartphone global positioning system (GPS) can be used to provide continuous, quantitative, and objective information on patient mobility. The aim of this study was to use daily mobility features derived from the smartphone GPS to characterize the perioperative period of patients undergoing spine surgery and to compare these objective measurements to PROMs, the current gold standard.

METHODS

Eight daily mobility features were derived from smartphone GPS data in a population of 39 patients undergoing spine surgery for a period of 2 months starting 3weeks before surgery. In parallel, three different PROMs for pain (visual analog scale [VAS]), disability (Oswestry Disability Index [ODI]) and functional status (Patient-Reported Outcomes Measurement Information System [PROMIS]) were serially measured. Segmented linear regression analysis was used to assess trends before and after surgery. The Student paired t-test was used to compare pre- and postoperative PROM scores. Pearson’s correlation was calculated between the daily average of each GPS-based mobility feature and the daily average of each PROM score during the recovery period.

RESULTS

Smartphone GPS features provided data documenting a reduction in mobility during the immediate postoperative period, followed by a progressive and steady increase with a return to baseline mobility values 1 month after surgery. PROMs measuring pain, physical performance, and disability were significantly different 1 month after surgery compared to the 2 immediate preoperative weeks. The GPS-based features presented moderate to strong linear correlation with pain VAS and PROMIS physical score during the recovery period (Pearson r > 0.7), whereas the ODI and PROMIS mental scores presented a weak correlation (Pearson r approximately 0.4).

CONCLUSIONS

Smartphone-derived GPS features were shown to accurately characterize perioperative mobility trends in patients undergoing surgery for spine-related diseases. Features related to time (rather than distance) were better at describing patient physical and performance status. Smartphone GPS has the potential to be used for the development of accurate, noninvasive and personalized tools for patient mobility monitoring after surgery.

ABBREVIATIONS

GPS = global positioning system; ODI = Oswestry Disability Index; PROM = patient-reported outcome measure; PROMIS = Patient-Reported Outcomes Measurement Information System; VAS = visual analog scale.

OBJECTIVE

Patient-reported outcome measures (PROMs) are currently the gold standard to evaluate patient physical performance and ability to recover after spine surgery. However, PROMs have significant limitations due to the qualitative and subjective nature of the information reported as well as the impossibility of using this method in a continuous manner. The smartphone global positioning system (GPS) can be used to provide continuous, quantitative, and objective information on patient mobility. The aim of this study was to use daily mobility features derived from the smartphone GPS to characterize the perioperative period of patients undergoing spine surgery and to compare these objective measurements to PROMs, the current gold standard.

METHODS

Eight daily mobility features were derived from smartphone GPS data in a population of 39 patients undergoing spine surgery for a period of 2 months starting 3weeks before surgery. In parallel, three different PROMs for pain (visual analog scale [VAS]), disability (Oswestry Disability Index [ODI]) and functional status (Patient-Reported Outcomes Measurement Information System [PROMIS]) were serially measured. Segmented linear regression analysis was used to assess trends before and after surgery. The Student paired t-test was used to compare pre- and postoperative PROM scores. Pearson’s correlation was calculated between the daily average of each GPS-based mobility feature and the daily average of each PROM score during the recovery period.

RESULTS

Smartphone GPS features provided data documenting a reduction in mobility during the immediate postoperative period, followed by a progressive and steady increase with a return to baseline mobility values 1 month after surgery. PROMs measuring pain, physical performance, and disability were significantly different 1 month after surgery compared to the 2 immediate preoperative weeks. The GPS-based features presented moderate to strong linear correlation with pain VAS and PROMIS physical score during the recovery period (Pearson r > 0.7), whereas the ODI and PROMIS mental scores presented a weak correlation (Pearson r approximately 0.4).

CONCLUSIONS

Smartphone-derived GPS features were shown to accurately characterize perioperative mobility trends in patients undergoing surgery for spine-related diseases. Features related to time (rather than distance) were better at describing patient physical and performance status. Smartphone GPS has the potential to be used for the development of accurate, noninvasive and personalized tools for patient mobility monitoring after surgery.

In Brief

The objective of this study was to investigate the feasibility of using personal smartphones to accurately characterize patients' mobility after spine surgery. Smartphone-derived GPS features were shown to accurately characterize perioperative mobility trends and significantly correlate to current patient-reported outcome measures. The results of this study unveiled an opportunity for the use of personal devices as objective, accurate, and noninvasive tools for the follow-up of patients undergoing spine surgery.

Spine-related diseases are a major cause of disability worldwide. Low-back pain, radicular pain, neurogenic claudication, mobility and sensory deficits are some of the most common symptoms leading patients to approach neurosurgery clinics.1,2 The surgical indication for a patient presenting with a spine issue depends on the correct interpretation of the patient history, physical examination, and imaging findings and has the final goal of improving the patient’s clinical condition and quality of life.3,4

The best way to assess the results of a surgical operation to the spine is to evaluate the ability to return to a normal level of activity in a context of resolution of symptoms.5,6 The gold standard to evaluate patient functional status and post-procedure recovery consists of a small number of visits and a set of questionnaires composed of qualitative or semiquantitative questions regarding the patient’s pain and the ability to perform self-care, physical activity, and work. These questionnaires are limited in the amount of information they can generate and in their ability to provide continuous and objective patient monitoring.7–9

There is a need for more objective, patient centered and less invasive tools to assess patients’ ability to return to activity after surgery. Recently, there has been increasing interest in the use of smartphone and wearable devices to collect this type of information.10–13 The data generated by these devices can be used to study social, behavioral, and cognitive phenotypes in naturalistic settings. This approach, now known as digital phenotyping, has been defined as moment-by-moment quantification of the individual-level human phenotype in situ using data from smartphones and other personal digital devices.14,15 Digital phenotyping relies on data streams derived from personal digital devices, i.e., devices that individuals already own and use, thus bypassing the need of having the patient wear an additional device for data acquisition. Smartphone data can be divided into active and passive: the former includes surveys and voice recordings, the latter includes sensor data (accelerometer, gyroscope, etc.) and usage data (communication logs, screen logs, etc.).16 This approach has been recently used to successfully characterize the behavior and functional status of different types of patients, unveiling the potential to develop clinically valid tools for patient follow-up.17–20 By coupling passively acquired data streams with gold standard patient-reported outcomes measures (PROMs), there is an opportunity to validate smartphones as noninvasive, scalable tools for patient follow-up and quality of life assessment.

The aim of this study was to investigate the feasibility of personal smartphones to accurately characterize patients’ mobility after spine surgery. We hypothesized that daily metrics derived from smartphone global positioning system (GPS) data could be used to effectively monitor patients’ postoperative return to baseline activity. Our results suggest that this methodology can be used to noninvasively follow patients over time. In the near future, this approach could be used to evaluate trajectories of recovery for patients and it could provide additional information for patient counseling and presurgical decision-making.

Methods

Patient Selection

Patient enrollment was conducted at two US academic hospitals in Boston, Brigham and Women’s Hospital and Faulkner Hospital. Patients presenting with brain or spine disease of neurosurgical interest and who provided consent, downloaded and activated the Beiwe smartphone application, which represents the front end of the research platform created by a member of our research team.16 Patients were not enrolled if they did not possess a smartphone or if they were younger than 18 years. By protocol, data were acquired and patients prospectively followed for a period of 6 months. According to the design of this study, only patients with evidence of spine disease (degenerative, tumoral, or traumatic) who underwent a surgical operation and had available smartphone GPS data and PROMs results for 60 days starting 3 weeks before the surgery were included. This study was approved by the Ethics Committee and reviewed under the protocol number 2016P000095.

Data Collection

Three types of data were collected in this study: smartphone GPS data; PROMs for pain and physical and mental health, collected both on the smartphone and during patient follow-up visits; and basic demographic and clinical information.

All smartphone data were immediately encrypted on the smartphone. The data were then uploaded to the Beiwe back-end system based on the Amazon Web Services (AWS) cloud. Once the data were received by the system back-end, they were re-encrypted by the system and deleted from the front-end smartphone application. The data were stored in industry-standard secure storage, in compliance with HIPAA (Health Insurance Portability and Accountability Act) regulations. Because the GPS sensor consumes a significant amount of power, it has to be sampled alternatingly between an on cycle and an off cycle, corresponding to time intervals when the sensors actively collected data and were dormant, respectively. In our study, the GPS was configured to follow a 1-minute on cycle and a 10-minute off cycle. Our previous work has shown that missing GPS data must be imputed prior to constructing daily summary statistics of the data. We used our existing resampling method for imputation, which has been demonstrated to result in a 10-fold reduction in the error averaged across all mobility features compared to simple linear interpolation of data.21 The daily summaries computed (following imputation) included the following: overall distance covered (in kilometers), radius of gyration (average radius traveled over a period of 1 day), diameter (largest distance between any two locations visited in a day), maximal distance from home (in kilometers), time spent at home (in hours), number of locations visited (defined as a location where the phone has stationed for at least 5 minutes), proportion of time spent in significant locations (percentage of day) and time of day spent moving (in hours).

PROMs included the following scores: visual analogic scale (VAS) for pain, Oswestry Disability Index (ODI), and Patient Reported Measures Information System 10 (PROMIS-10) physical and mental scores.

The following demographic and clinical variables were collected for each patient: age, sex, type of spinal pathology, spine level involved, presenting symptoms, and type of surgical operation.

Data Preprocessing

The collected PROMs data were stored in folders named with a unique identification code for every patient. Inside each subfolder were comma-separated value (.csv) files containing individual responses to the surveys. The files were named according to the date and time of submission.

To consolidate the files, a Python script was written for each subfolder to concatenate all the responses into one data frame, resulting in three data frames per patient. Occasionally, a patient would submit a response to the pain scale survey twice in one day. To keep consistency to a one-per-day format, the latter response was removed.

Statistical Analysis

A set of tables and plots was generated to explore the GPS-based variables and PROMs trends over time for the entire observation period. Each variable was summarized at the cohort level using daily mean and standard deviation values.

For each GPS feature, a segmented regression analysis, adjusted for age and strength deficit, was conducted to evaluate the change in data trend over time between the following three periods: the immediate preoperative phase (day −20 to day −3), the perioperative phase (day −2 to day +3), and the recovery phase (day +4 to day +40).

Pearson’s correlation coefficient (r) was calculated between each pair of GPS features and PROMs collected during the recovery period, defined as the period from postoperative day 3 to postoperative day 40. In case 1 or more days presented missing values, a linear interpolation function was used for imputation. The strength of the correlation was categorized as follows: 0.2–0.39 as weak, 0.40–0.59 as moderate, 0.6–0.79 as strong, and 0.8–1 as very strong.22

Student paired t-tests were conducted to evaluate the change in PROM scores between the end of the recovery period (postoperative weeks 4 and 5) and baseline (2 immediate preoperative weeks) values. The alpha level for each statistical test conducted was set at 0.05 in the context of a two-tailed test.

Results

Population Characteristics

A total of 238 patients with brain or spine disease of neurosurgical interest downloaded Beiwe between June 2016 to July 2018. Of these, 56 patients underwent surgery for spine disease and were enrolled before their intervention. Seventeen patients were excluded (15 had no PROMs and 2 deleted the application after a few days), and a final group of 39 patients (69.6%) who fulfilled the inclusion criteria with regard to smartphone and PROMs data availability were included in this study (details in Table 1).

TABLE 1.

Data acquisition details for patients included in the study and summary of reasons for exclusion

Values
Total no. of pts included39
 Smartphone operating system, no (%)
  iOS31 (79)
  Android8 (21)
 Passive data acquisition time, days
  Overall period61
  Total days for all pts1875
  Average days per pt 48.1
 PROMs data, no. of surveys
  Pain survey875
  ODI211
  PROMIS226
Total no. of pts excluded17
 Reason for exclusion
  No PROMs data15
  Dropped out2

iOS = iPhone operating system; pt = patient.

The average age of the cohort was 52.8 ± 13.8 (range 25–80) years, and the male to female ratio was 0.77:1 (17 males, 22 females). The most frequent diagnosis was herniated disc with 19 cases (48.7%), followed by spondylosis, central canal stenosis or foraminal stenosis, failed back surgical syndrome, spondylolisthesis, synovial cyst, scoliosis, or vertebral fracture. These pathologies were distributed along the spinal column with the following proportions: 64.1% lumbar level, 30.8% cervical level, 2.6% thoracic level, and 2.6% sacral level. The most common clinical presentation was radiculopathy, reported by 33 patients (84.6%), followed by low-back pain, neurogenic claudication, neck pain, and myelopathy. A posterior surgical approach was chosen in the majority of cases (69.2%) and the anterior approach was reserved just for patients with cervical disease. In 66.6% of cases, patients received spinal instrumentation. Detailed demographic and clinical characteristics of the patient cohort are reported in Table 2.

TABLE 2.

Demographic, clinical, and surgical features of patients included in the study

Values
Age, yrs 52.8 (25–80)
Sex
 Male17 (43.5)
 Female22 (56.5)
Spine disease
 Herniated disc19 (48.7)
 Spinal stenosis 8 (20.5)
 Failed back surgical syndrome 3 (7.7)
 Spondylolisthesis3 (7.7)
 Synovial cyst2 (5.1)
 Scoliosis2 (5.1)
 Vertebral fracture2 (5.1)
Spine level
 Cervical12 (30.8)
 Thoracic1 (2.6)
 Lumbar 25 (64.1)
 Sacral1 (2.6)
Clinical presentation
 Radiculopathy33 (84.6)
 Low-back pain7 (17.9)
 Neurogenic claudication4 (10.2)
 Neck pain1 (2.6)
 Myelopathy2 (2.6)
Op approach
 Anterior10 (25.6)
 Posterior27 (69,2)
 Combined 2 (5.1)
Instrumentation
 Yes26 (66.6)
 No13 (43.4)

Values are presented as mean (range) or number of patients (%).

GPS-Based Mobility Trends

The average number of locations visited per day ranged between 1.9 and 4.8, with an average of 3.8 ± 0.3 during the preoperative period. On the first postoperative week the mean value dropped to 2.2 ± 0.2, and over the following weeks a steady increase was observed, reaching a level comparable to baseline during postoperative week 5 (3.5 ± 0.3) (Table 3 and Fig. 1). The proportion of the day spent at these locations (outside home) varied between 19% and 80%, with a preoperative average value of 66% ± 4%. The time spent at these locations dropped during the first postoperative week (29.5% ± 6%) and progressively returned toward baseline values during the following weeks (postoperative week 5: 55% ± 6%).

TABLE 3.

Passive and active data weekly summaries in preoperative period (2 weeks) and postoperative period (5 weeks)

Preop period (2 wks)POW1POW2POW3POW4 POW5
Passive data
 Number of locations visited3.8 ± 0.32.2 ± 0.22.7 ± 0.23.3 ± 0.43.5 ± 0.33.5 ± 0.3
 Time spent in locations (% of day)66 ± 429.5 ± 637 ± 649 ± 1053 ± 555 ± 6
 Time spent moving (hrs)2.8 ± 0.32 ± 0.32.5 ± 0.22.8 ± 0.32.8 ± 0.22.9 ± 0.3
 Time spent paused (hrs)21.2 ± 0.322 ± 0.321.4 ± 0.321.2 ± 0.321.2 ± 0.221 ± 0.3
 Time spent at home (hrs)15.4 ± 0.615.8 ± 216.4 ± 115.7 ± 116.3 ± 0.815.9 ± 0.8
 Distance traveled (km)92 ± 6421.3 ± 6.633.9 ± 1087 ± 6943 ± 10264.5 ± 560.5
 Maximal distance from home (km)287.7 ± 103.350.9 ± 38.960.2 ± 31.8160.3 ± 70.775.1 ± 2897.2 ± 78.3
 Diameter (km)55.4 ± 61.810.9 ± 5.412.6 ± 4.457.2 ± 67.116.3 ± 5.466.3 ± 116
 Radius (km)18.4 ± 27.93.5 ± 2.72.8 ± 1.923 ± 33.74.4 ± 1.511.6 ± 12.2
Active data
 Pain7.3 ± 0.36.7 ± 0.94.4 ± 0.63.4 ± 0.63 ± 0.43.1 ± 0.3
 ODI score47.2 ± 6.551.6 ± 23.144.2 ± 18.644.2 ± 17.436.7 ± 5.827.8 ± 15.1
 PROMIS physical score38.3 ± 2.934.8 ± 740.2 ± 4.942.3 ± 3.143.8 ± 2.846.3 ± 3.2
 PROMIS mental score44.9 ± 4.747.3 ± 10.945.1 ± 4.746.5 ± 2.846.3 ± 348.4 ± 3.7

POW = postoperative week.

All values are expressed at the cohort level as mean ± SD.

FIG. 1.
FIG. 1.

Variation over the perioperative period of time–related GPS features. A: Variation over time of time spent paused (pause). B: Variation over time of time spent in significant locations (time locs). C: Variation over time of time spent moving (move). D: Variation over time of number of significant locations visited (locs). Vertical line indicates day of surgery; dots indicate observations. Pause and move are expressed in hours. Time locs are expressed as percentage of the day. Figure is available in color online only.

The average amount of time spent at home ranged between 9.4 and 17.8 hours. Patients spent on average 15.4 ± 0.7 hours at home every day during the 2 weeks preceding the surgery. After the operation, a general increase in the values for the amount of time spent at home was observed, exceeding an average of 16 hours in postoperative weeks 2 and 4 (Table 3 and Fig. 1).

Time spent moving, defined as the time in which the smartphone is not stationary, ranged between 1.5 and 3.5 hours per day. In the preoperative period, the average time was 2.8 ± 0.3 hours. The time spent moving presented a nadir during the first postoperative week (2.0 ± 0.3), followed by an increase in which time spent moving reached and passed the baseline values by postoperative week number 5 (2.9 ± 0.3). Predictably, the observed trends in time spent moving showed an inverse relationship with time spent not moving (Table 3 and Fig. 1).

The average distance covered on a daily basis ranged between 12.8 and 1548 km. The average distance recorded during the preoperative period was 92 ± 64 km. The lowest average value was recorded during the first postoperative week (21.3 ± 7). The following weeks presented a general value increase, which was particularly evident on postoperative weeks 3 and 5 (Table 3). Similarly, the radius of gyration, diameter, and maximal distance from home (respective preoperative values 18.4 ± 28, 55.4 ± 62, and 287.7 ± 103 km) presented reduced values on the first postoperative weeks and then progressively increased (Table 3).

With regard to time spent paused and moving, number of locations visited, and time spent at these locations, the segmented linear regression analysis was able to detect and characterize the change in direction of the data trends between the preoperative, perioperative, and recovery periods observed in the exploratory analysis (all p values of the respective slopes were < 0.05; Table 4). No statistically significant change in the direction of the data trend was detected with regard to GPS features measuring distances (distance traveled, maximal distance from home, diameter, radius of gyration). Interestingly, time spent at home was the only time-related variable for which a statistically significant change in the trend was not detected (Table 4).

TABLE 4.

Segmental regression analysis results for passive data features during the preoperative, perioperative, and postoperative period

Slope Estimate: Preop/Baseline, Days −20 to −3Direction & Slope Estimate Change (p value)
Periop, Days −2 to +3Postop, Days +4 to +40
Time spent at home0.07 −0.16 (0.57)0.07 (0.80)
Locations visited−0.01−0.31 (<0.01)0.36 (<0.001)
Time spent paused−0.030.18 (0.01)−0.18 (0.01)
Distance traveled−17.30.11 (0.99)28.4 (0.65)
Time spent in locations0.0003−0.08 (<0.0001)0.09 (<0.0001)
Time spent moving0.03−0.18 (0.01)0.18 (0.01)
Maximal distance from home−4.955.13 (0.72)1.98 (0.89)
Diameter−1.45−1.71 (0.91)4.92 (0.74)
Radius1.33−1.62 (0.74)0.46 (0.92)

All segmented linear regression models were adjusted for age and strength deficit. Boldface type indicates statistical significance.

With regard to time spent paused and moving, number of locations visited, and time spent at these locations, the segmented linear regression analysis was able to detect and characterize the change in direction of the data trends between the preoperative, perioperative, and recovery periods observed in the exploratory analysis (all p values of the respective slopes were < 0.05; Table 4). No statistically significant change in the direction of the data trend was detected with regard to GPS features measuring distances (distance traveled, maximal distance from home, diameter, radius of gyration). Interestingly, time spent at home was the only time-related variable for which a statistically significant change in the trend was not detected (Table 4).

Patient Reported Outcome Measures

The daily average pain reported in the 2 weeks preceding the surgery was 7.3 ± 0.3, followed by a reduction in pain level particularly evident in the first 3 postoperative weeks (postoperative week 1: 6.7 ± 0.9, postoperative week 2: 4.4 ± 0.6, postoperative week 3: 3.4 ± 0.6) and then stabilized around a value of 3 (Table 3 and Fig. 2).

FIG. 2.
FIG. 2.

Variation over the perioperative period of PROMs. A: Variation over time of pain level (Pain level). B: Variation over time of ODI score (ODI). C: Variation over time of PROMIS physical score (PROMIS P). D: Variation over time of PROMIS mental score (PROMIS M). Vertical line indicates day of surgery; dots indicate observations. Figure is available in color online only.

The average reported ODI score in the preoperative period was 47.2% ± 6.5%, describing a condition where pain was the main problem and daily activities were affected (40%–60% ODI score). The first postoperative week presented an increase in the percentage of disability to 51.6% ± 23.1%, followed by a progressive decrease, which in the fourth week dropped under the 40% threshold (ODI score postoperative week 4 36.7% ± 5.8%, postoperative week 5 27.8% ± 15.1%). The values in weeks 4 and 5 highlighted an effective change in the category of disability reported by the patients (from 40%–60% to 20%–40% ODI score), indicating that at this time daily activities were not grossly affected even in the presence of some pain (Table 3 and Fig. 2).

The average reported PROMIS physical score in the preoperative period was 38.3 ± 2.9. The first postoperative week presented a slight reduction in the average value to a score of 34.8 ± 7. These values represented the presence of physical difficulties almost entailing the need for a cane to walk. The following weeks presented a progressive increase in the score reported, which stabilized around a value of 45 at 1 month post-surgery (postoperative week 4: 43.8 ± 2.8, postoperative week 5: 46.3 ± 3.2) representing a clinical improvement where patients were now able to be physically independent, with no need of support (Table 3 and Fig. 2).

The difference between pain, ODI, and PROMIS physical scores at 1 month post-surgery compared to the preoperative baseline proved to be statistically significant (p < 0.001).

With regard to the PROMIS mental score, the average values reported remained between 44 and 49 points in the whole observation period with no detected significant difference between postoperative and preoperative periods (p = 0.44).

Correlation Between GPS Mobility Data and PROMs During the Recovery Period

Pain was found to have the highest correlation with passive mobility measures during the recovery period. Strong correlations were found with time spent moving (r = 0.77) or paused (r = −0.77) as well as with time spent in significant locations (r = −0.75) and number of locations visited (r = −0.68) (Fig. 3); time spent at home presented a moderate positive correlation (r = 0.48). All the reported correlations were statistically significant (p < 0.05). Interestingly, measures of spatial movement (overall distance, radius of gyration, diameter), presented a weak and nonsignificant correlation with pain, with the exception of maximal distance from home, which presented a mild negative correlation (r = −0.39).

The PROMIS physical score presented significant correlation to mobility data as well. Strong correlations were found with time spent moving (r = 0.70) or paused (r = −0.70) and with time spent in a significant location (r = 0.65) (Fig. 4), followed by a moderate correlation to number of locations visited (r = 0.52) and time spent at home (r = −0.45). All reported correlations were statistically significant (p < 0.05). Similarly to pain, all measures of spatial movement, except for maximal distance from home, presented weak and nonsignificant correlations with the PROMIS physical score.

FIG. 3.
FIG. 3.

Pain and GPS feature variation over recovery period and correlation plots. (A) Left: Variation over time of time spent paused (pause, full circles) and pain (P, empty circles). Right: Scatter plot showing the linear relationship between time spent paused and pain (r = 0.77). (B) Left: Variation over time of time spent moving (move, full circles) and pain (P, empty circles). Right: Scatter plot showing the linear relationship between time spent moving and pain (r = −0.77).FIG. 3. (C) Left: Variation over time of time spent in significant locations (time locs, full circles) and pain (P, empty circles). Right: Scatter plot showing the linear relationship between time spent in significant locations and pain (r = −0.75). (D) Left: Variation over time of number of significant locations visited (locs, full circles) and pain (P, empty circles). Right: Scatter plot showing the linear relationship between number of significant locations visited and pain (r = −0.68). Pause and move are expressed in hours. Time locs are expressed as percentage of the day. Figure is available in color online only.

FIG. 4.
FIG. 4.

PROMIS physical score and GPS features variation over time and correlation plots. (A) Left: Variation over time of time spent paused (pause, full circles) and PROMIS physical score (PP, empty circles). Right: Scatter plot showing the linear relationship between time spent paused and PROMIS physical score (r = −0.7). (B) Left: Variation over time of time spent moving (move, full circles) and PROMIS physical score (PP, empty circles). Right: Scatter plot showing the linear relationship between time spent moving and PROMIS physical score (r = 0.7). (C) Left: Variation over time of time spent in significant locations (time locs, full circles) and PROMIS physical score (PP, empty circles). Right: Scatter plot showing the linear relationship between time spent in significant locations and PROMIS physical score (r = 0.65). Pause and move are expressed in hours. Time locs are expressed as percentage of the day. Figure is available in color online only.

The ODI score was found to be moderately correlated to time spent moving (r = 0.55) and paused (r = −0.55) and with time spent in significant locations (r = −0.43); correlation to number of locations visited was present but mild (r = −0.32). In the case of ODI, the daily distance traveled was found to be significantly associated, even if with mild strength (r = −0.36).

Finally, with regard to mental status expressed as PROMIS mental score, mild correlations were found with time spent in significant locations (r = 0.37), maximal distance from home (r = 0.36), number of locations visited (r = 0.32), and time spent at home (r = −0.31).

Discussion

The results of this study demonstrated the ability of smartphone-based GPS mobility features to accurately characterize perioperative mobility trends in patients undergoing surgery for spine-related diseases and the presence of significant correlation with gold standard PROMs.

We found that different GPS metrics behaved differently in the perioperative period. Under conditions in which a patient just had a surgical operation, we would expect an immediate reduction in their mobility and increase in the time spent resting, followed by a progressive inversion of these trends. This was exactly what we observed in the form of an initial increase of the time spent without moving and a reduction in the number of locations visited as well as in the time spent in such locations, followed by a return to baseline values by the time of postoperative week 4 or 5. The time spent at home presented an overall slight increase over the first postoperative month, highlighting the need for home rest. A more variable picture was depicted for distance-related variables such as distance traveled, maximal distance from home, diameter, and radius of gyration, due to the presence of massive outliers.

Previous studies have used digital phenotyping to characterize mobility patterns in different patient populations and have studied the correlation between passively acquired mobility data and patient outcome data. Cote et al. observed a significant correlation between self-reported pain and GPS-based measures in a population of patients with spine disease, where an increased reported pain was associated with reduced mobility.18 Panda et al. used smartphone accelerometer data from a cohort of cancer patients who underwent surgery to develop a summary statistic called daily exertional activity; these authors observed that patients experiencing postoperative events presented a significantly lower level of activity in the recovery period.17,23

To fully evaluate patients who are recovering from spine surgery, apart from pain and occurrence of adverse events, additional aspects regarding the ability to be independent and return to work, physical performance, and mental status need to be considered. It has been observed that different outcome measures such as VAS, ODI, and PROMIS scores do not correlate very well with one another, suggesting that they are measuring different aspects of patients’ lived experience.8,9,24

In this study, we found strong correlations between mobility measures, not only with pain, but also with physical and performance status, measured through the PROMIS physical score, which included the ability to perform self-care, walk, perform daily activities, and the presence of fatigue. Conversely, only a moderate or mild correlation was found with ODI score and the mental section of the PROMIS score. An explanation for these findings is that the ODI score and PROMIS mental score were developed to assess, partially or exclusively, the mental, emotional, and social health of patients, whereas spine surgery tended to present its toll more in the form of physical limitation, which is better captured by an assessment tool specific for physical activity, such as the PROMIS physical score.

As described above, the GPS features used in this study characterized mobility from different perspectives, specifically in terms of time and space. This distinction needs special consideration because while the current standard for the evaluation of patient recovery after spine surgery is to investigate patient performance based on space (i.e., “how much are you able to walk?"),5,6,25 our study shows that time-related variables present a better correlation to patient-reported experience and might therefore provide a more adequate understanding of how patients are recovering and how effective the surgery has been (i.e., “how much time did you spend on your feet and outside home?"). It was, in fact, surprising that the most obvious GPS signatures, such as distance covered as well as radius of gyration and diameter, were also the ones with the weakest correlation indices, and ones for which the segmented analysis could not detect a significant change in the data trend between pre-, peri-, and postoperative periods. We believe this is due to the fact that GPS distances are highly affected by the use of different means of transportation, as clearly demonstrated by the observation of several hundred kilometers traveled per day—which doesn’t necessarily correlate to the actual physical condition of a patient.

Of particular interest was then the comparison of the GPS mobility measures to PROMs at 1 month after surgery. While objective GPS-based measures presented values comparable to preoperative baseline levels, PROMs of physical status and disability scores detected a clinically relevant and statistically significant improvement of the preoperative condition (from category 40%–60% to 20%–40% in ODI score with a change of more than 10% and a change of more than 7 points in PROMIS physical score).26 These findings demonstrate the presence of a detectable difference between an objective measure of mobility and a subjective evaluation of physical activity and performance status: patients objectively reached preoperative baseline mobility values (which are expected to further improve) but feel they are doing more and significantly better compared to the period before surgery. The observation of this difference powerfully demonstrates the ability of digital phenotyping to objectively assess and characterize patients’ return to activity and physical performance, free from influence and biases caused by the patient’s conscious recollection.7,27 It is important to consider that the baseline values we utilized refer to the patients’ immediate preoperative condition rather than a perfectly normal functional level. In this regard, we do not consider reaching the preoperative values as the best outcome possible, which we in fact expect the patients to further improve upon with time. Similarly, while an initial postoperative reduction of mobility followed by progressive improvement can be observed in all patients who undergo spine surgery, we do expect patients with different diseases and surgical interventions to have significantly different recovery experiences.

This study presents several limitations that need to be taken into consideration. First, patients who did not own a smartphone were a priori excluded, potentially inserting a selection bias against specific groups, for example, older patients. This highlights an important limit in the generalization of the results derived by personal digital devices, which tend to be used by younger sections of the population. Similarly, non–English-speaking patients were excluded. In terms of spine diseases and type of surgical intervention, our analysis did not take into consideration disease- or treatment-specific features, rather providing a general characterization of patient mobility immediately after surgery and during the recovery period. It is also important to remember that GPS data and the related variables extracted refer to the smartphone and not the patient him/herself, and we considered the location of the smartphone as a proxy for the location of the patient.

In spite of these limitations, smartphone-based GPS data proved to be able to characterize perioperative mobility trends that were consistent with patient reported experience in the perioperative period.

Future steps need to be taken to expand the study of digital phenotyping in spine patients; it will be necessary to include more patients to allow the development of disease- and treatment-specific mobility signatures that would permit a closer, quantifiable, and more reliable follow-up for patients undergoing spine surgery as well as provide useful information to be used to inform patients before surgery on what to expect from the recovery period. Furthermore, additional smartphone data sources such as accelerometer and gyroscope data should be combined with GPS signatures to provide an even more accurate characterization of the actual functional status and recovery trajectory,28 eventually also allowing the use of space-related variables, which GPS alone does not permit.

Conclusions

Our study showed a strong correlation between specific smartphone GPS features and current pain and physical performance outcome measures in patients undergoing spine surgery. The results of this study unveiled an opportunity for the use of personal devices and the development of objective, accurate, and noninvasive tools for the follow up of patients undergoing spine surgery. The development of such tools for the objective quantification of patient physical performance and their use, along with the information provided by the patients themselves, would allow a paradigmatic shift from patient assessment based only on “how they feel,” to the possibility of observing, and therefore considering, how the patients are actually doing.

Acknowledgments

Sources of support: H.T.R. was supported by an NIH BD2K training grant (T32LM012411) and fellowship F31HD102159.

Disclosures

Dr. Groff reports receiving royalties from Spine Art and NuVasive. Dr. Onnela is a co-founder of Phebe Health, a newly established company that operates in digital phenotyping.

Author Contributions

Conception and design: Boaro, Onnela. Acquisition of data: Boaro, Leung, Reeder, Siddi, Liu. Analysis and interpretation of data: Boaro, Leung, Reeder, Liu, Mekary. Drafting the article: Boaro, Leung. Critically revising the article: Leung, Siddi, Mezzalira, Liu, Mekary, Lu, Groff, Onnela, Smith. Reviewed submitted version of manuscript: Boaro, Siddi, Mezzalira, Lu, Groff, Onnela, Smith. Statistical analysis: Boaro, Reeder, Mekary. Study supervision: Onnela, Smith.

Supplemental Information

Previous Presentations

Portions of this work have been presented in poster form at the Computational Neuroscience Outcomes Center Symposium 2019, October 9, 2019, in Boston, MA.

References

  • 1

    Lurie J, Tomkins-Lane C. Management of lumbar spinal stenosis. BMJ. 2016;352:h6234.

  • 2

    Kato S, Fehlings M. Degenerative cervical myelopathy. Curr Rev Musculoskelet Med. 2016;9(3):263271.

  • 3

    Malmivaara A, Slätis P, Heliövaara M, et al. Surgical or nonoperative treatment for lumbar spinal stenosis? A randomized controlled trial. Spine (Phila Pa 1976).2007;32(1):18.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Ghobrial GM, Harrop JS. Surgery vs conservative care for cervical spondylotic myelopathy: nonoperative operative management. Neurosurg. 2015;62(CN Suppl 1):6265.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5

    Ogura Y, Kobayashi Y, Kitagawa T, et al. Outcome measures reflecting patient satisfaction following decompression surgery for lumbar spinal stenosis: comparison of major outcome measures. Clin Neurol Neurosurg. 2020;191:105710.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Zuckerman SL, Devin CJ. Outcomes and value in elective cervical spine surgery: an introductory and practical narrative review. J Spine Surg. 2020;6(1):89105.

  • 7

    Rabah NM, Levin JM, Winkelman RD, et al. The association between physicians’ communication and patient-reported outcomes in spine surgery. Spine (Phila Pa 1976).2020;45(15):10731080.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Levin JM, Winkelman RD, Smith GA, et al. The association between the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey and real-world clinical outcomes in lumbar spine surgery. Spine J. 2017;17(11):15861593.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Stokes OM, Cole AA, Breakwell LM, et al. Do we have the right PROMs for measuring outcomes in lumbar spinal surgery?. Eur Spine J. 2017;26(3):816824.

  • 10

    Gordon BA, Bruce L, Benson AC. Physical activity intensity can be accurately monitored by smartphone global positioning system ‘app’. Eur J Sport Sci. 2016;16(5):624631.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Krummel TM. The rise of wearable technology in health care. JAMA Netw Open. 2019;2(2):e187672.

  • 12

    Madhushri P, Dzhagaryan AA, Jovanov E, Milenkovic A. A smartphone application suite for assessing mobility. Annu Int Conf IEEE Eng Med Biol Soc. 2016:31173120.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13

    Wang Q, Markopoulos P, Yu B, et al. Interactive wearable systems for upper body rehabilitation: a systematic review. J Neuroeng Rehabil. 2017;14(1):20.

  • 14

    Onnela JP. Opportunities and challenges in the collection and analysis of digital phenotyping data. Neuropsychopharmacology. 2021;46(1):4554.

  • 15

    Onnela JP, Rauch SL. Harnessing smartphone-based digital phenotyping to enhance behavioral and mental health. Neuropsychopharmacology. 2016;41(7):16911696.

  • 16

    Torous J, Kiang MV, Lorme J, Onnela JP. New tools for new research in psychiatry: a scalable and customizable platform to empower data driven smartphone research. JMIR Ment Health. 2016;3(2):e16.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 17

    Panda N, Solsky I, Huang EJ, et al. Using smartphones to capture novel recovery metrics after cancer surgery. JAMA Surg. 2020;155(2):123129.

  • 18

    Cote DJ, Barnett I, Onnela JP, Smith TR. Digital phenotyping in patients with spine disease: a novel approach to quantifying mobility and quality of life. World Neurosurg. 2019;126:e241e249.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 19

    Barnett I, Torous J, Staples P, et al. Relapse prediction in schizophrenia through digital phenotyping: a pilot study. Neuropsychopharmacology. 2018;43(8):16601666.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 20

    Aledavood T, Torous J, Triana Hoyos AM, et al. Smartphone-based tracking of sleep in depression, anxiety, and psychotic disorders. Curr Psychiatry Rep. 2019;21(7):49.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Barnett I, Onnela JP. Inferring mobility measures from GPS traces with missing data. Biostatistics. 2020;21(2):e98e112.

  • 22

    Chan YH. Biostatistics 104: correlational analysis. Singapore Med J. 2003;44(12):614619.

  • 23

    Panda N, Solsky I, Hawrusik B, et al. Smartphone Global Positioning System (GPS) data enhances recovery assessment after breast cancer surgery. Ann Surg Oncol. 2021;28(2):985994.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 24

    DeVine J, Norvell DC, Ecker E, et al. Evaluating the correlation and responsiveness of patient-reported pain with function and quality-of-life outcomes after spine surgery. Spine (Phila Pa 1976).2011;36(21)(suppl):S69S74.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 25

    Hung M, Saltzman CL, Kendall R, et al. What Are the MCIDs for PROMIS, NDI, and ODI instruments among patients with spinal conditions?. Clin Orthop Relat Res. 2018;476(10):20272036.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 26

    Steinhaus ME, Iyer S, Lovecchio F, et al. Minimal clinically important difference and substantial clinical benefit using PROMIS CAT in cervical spine surgery. Clin Spine Surg. 2019;32(9):392397.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 27

    Schwartz CE, Zhang J, Rapkin BD, Finkelstein JA. Reconsidering the minimally important difference: evidence of instability over time and across groups. Spine J. 2019;19(4):726734.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 28

    Nishiguchi S, Yamada M, Nagai K, et al. Reliability and validity of gait analysis by android-based smartphone. Telemed J E Health. 2012;18(4):292296.

  • Collapse
  • Expand
Images and illustration from Akinduro et al. (pp 834–843). Copyright Tito Vivas-Buitrago. Published with permission.
  • FIG. 1.

    Variation over the perioperative period of time–related GPS features. A: Variation over time of time spent paused (pause). B: Variation over time of time spent in significant locations (time locs). C: Variation over time of time spent moving (move). D: Variation over time of number of significant locations visited (locs). Vertical line indicates day of surgery; dots indicate observations. Pause and move are expressed in hours. Time locs are expressed as percentage of the day. Figure is available in color online only.

  • FIG. 2.

    Variation over the perioperative period of PROMs. A: Variation over time of pain level (Pain level). B: Variation over time of ODI score (ODI). C: Variation over time of PROMIS physical score (PROMIS P). D: Variation over time of PROMIS mental score (PROMIS M). Vertical line indicates day of surgery; dots indicate observations. Figure is available in color online only.

  • FIG. 3.

    Pain and GPS feature variation over recovery period and correlation plots. (A) Left: Variation over time of time spent paused (pause, full circles) and pain (P, empty circles). Right: Scatter plot showing the linear relationship between time spent paused and pain (r = 0.77). (B) Left: Variation over time of time spent moving (move, full circles) and pain (P, empty circles). Right: Scatter plot showing the linear relationship between time spent moving and pain (r = −0.77).FIG. 3. (C) Left: Variation over time of time spent in significant locations (time locs, full circles) and pain (P, empty circles). Right: Scatter plot showing the linear relationship between time spent in significant locations and pain (r = −0.75). (D) Left: Variation over time of number of significant locations visited (locs, full circles) and pain (P, empty circles). Right: Scatter plot showing the linear relationship between number of significant locations visited and pain (r = −0.68). Pause and move are expressed in hours. Time locs are expressed as percentage of the day. Figure is available in color online only.

  • FIG. 4.

    PROMIS physical score and GPS features variation over time and correlation plots. (A) Left: Variation over time of time spent paused (pause, full circles) and PROMIS physical score (PP, empty circles). Right: Scatter plot showing the linear relationship between time spent paused and PROMIS physical score (r = −0.7). (B) Left: Variation over time of time spent moving (move, full circles) and PROMIS physical score (PP, empty circles). Right: Scatter plot showing the linear relationship between time spent moving and PROMIS physical score (r = 0.7). (C) Left: Variation over time of time spent in significant locations (time locs, full circles) and PROMIS physical score (PP, empty circles). Right: Scatter plot showing the linear relationship between time spent in significant locations and PROMIS physical score (r = 0.65). Pause and move are expressed in hours. Time locs are expressed as percentage of the day. Figure is available in color online only.

  • 1

    Lurie J, Tomkins-Lane C. Management of lumbar spinal stenosis. BMJ. 2016;352:h6234.

  • 2

    Kato S, Fehlings M. Degenerative cervical myelopathy. Curr Rev Musculoskelet Med. 2016;9(3):263271.

  • 3

    Malmivaara A, Slätis P, Heliövaara M, et al. Surgical or nonoperative treatment for lumbar spinal stenosis? A randomized controlled trial. Spine (Phila Pa 1976).2007;32(1):18.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 4

    Ghobrial GM, Harrop JS. Surgery vs conservative care for cervical spondylotic myelopathy: nonoperative operative management. Neurosurg. 2015;62(CN Suppl 1):6265.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 5

    Ogura Y, Kobayashi Y, Kitagawa T, et al. Outcome measures reflecting patient satisfaction following decompression surgery for lumbar spinal stenosis: comparison of major outcome measures. Clin Neurol Neurosurg. 2020;191:105710.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 6

    Zuckerman SL, Devin CJ. Outcomes and value in elective cervical spine surgery: an introductory and practical narrative review. J Spine Surg. 2020;6(1):89105.

  • 7

    Rabah NM, Levin JM, Winkelman RD, et al. The association between physicians’ communication and patient-reported outcomes in spine surgery. Spine (Phila Pa 1976).2020;45(15):10731080.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 8

    Levin JM, Winkelman RD, Smith GA, et al. The association between the Hospital Consumer Assessment of Healthcare Providers and Systems (HCAHPS) survey and real-world clinical outcomes in lumbar spine surgery. Spine J. 2017;17(11):15861593.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 9

    Stokes OM, Cole AA, Breakwell LM, et al. Do we have the right PROMs for measuring outcomes in lumbar spinal surgery?. Eur Spine J. 2017;26(3):816824.

  • 10

    Gordon BA, Bruce L, Benson AC. Physical activity intensity can be accurately monitored by smartphone global positioning system ‘app’. Eur J Sport Sci. 2016;16(5):624631.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 11

    Krummel TM. The rise of wearable technology in health care. JAMA Netw Open. 2019;2(2):e187672.

  • 12

    Madhushri P, Dzhagaryan AA, Jovanov E, Milenkovic A. A smartphone application suite for assessing mobility. Annu Int Conf IEEE Eng Med Biol Soc. 2016:31173120.

    • PubMed
    • Search Google Scholar
    • Export Citation
  • 13

    Wang Q, Markopoulos P, Yu B, et al. Interactive wearable systems for upper body rehabilitation: a systematic review. J Neuroeng Rehabil. 2017;14(1):20.

  • 14

    Onnela JP. Opportunities and challenges in the collection and analysis of digital phenotyping data. Neuropsychopharmacology. 2021;46(1):4554.

  • 15

    Onnela JP, Rauch SL. Harnessing smartphone-based digital phenotyping to enhance behavioral and mental health. Neuropsychopharmacology. 2016;41(7):16911696.

  • 16

    Torous J, Kiang MV, Lorme J, Onnela JP. New tools for new research in psychiatry: a scalable and customizable platform to empower data driven smartphone research. JMIR Ment Health. 2016;3(2):e16.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 17

    Panda N, Solsky I, Huang EJ, et al. Using smartphones to capture novel recovery metrics after cancer surgery. JAMA Surg. 2020;155(2):123129.

  • 18

    Cote DJ, Barnett I, Onnela JP, Smith TR. Digital phenotyping in patients with spine disease: a novel approach to quantifying mobility and quality of life. World Neurosurg. 2019;126:e241e249.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 19

    Barnett I, Torous J, Staples P, et al. Relapse prediction in schizophrenia through digital phenotyping: a pilot study. Neuropsychopharmacology. 2018;43(8):16601666.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 20

    Aledavood T, Torous J, Triana Hoyos AM, et al. Smartphone-based tracking of sleep in depression, anxiety, and psychotic disorders. Curr Psychiatry Rep. 2019;21(7):49.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 21

    Barnett I, Onnela JP. Inferring mobility measures from GPS traces with missing data. Biostatistics. 2020;21(2):e98e112.

  • 22

    Chan YH. Biostatistics 104: correlational analysis. Singapore Med J. 2003;44(12):614619.

  • 23

    Panda N, Solsky I, Hawrusik B, et al. Smartphone Global Positioning System (GPS) data enhances recovery assessment after breast cancer surgery. Ann Surg Oncol. 2021;28(2):985994.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 24

    DeVine J, Norvell DC, Ecker E, et al. Evaluating the correlation and responsiveness of patient-reported pain with function and quality-of-life outcomes after spine surgery. Spine (Phila Pa 1976).2011;36(21)(suppl):S69S74.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 25

    Hung M, Saltzman CL, Kendall R, et al. What Are the MCIDs for PROMIS, NDI, and ODI instruments among patients with spinal conditions?. Clin Orthop Relat Res. 2018;476(10):20272036.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 26

    Steinhaus ME, Iyer S, Lovecchio F, et al. Minimal clinically important difference and substantial clinical benefit using PROMIS CAT in cervical spine surgery. Clin Spine Surg. 2019;32(9):392397.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 27

    Schwartz CE, Zhang J, Rapkin BD, Finkelstein JA. Reconsidering the minimally important difference: evidence of instability over time and across groups. Spine J. 2019;19(4):726734.

    • Crossref
    • PubMed
    • Search Google Scholar
    • Export Citation
  • 28

    Nishiguchi S, Yamada M, Nagai K, et al. Reliability and validity of gait analysis by android-based smartphone. Telemed J E Health. 2012;18(4):292296.

Metrics

All Time Past Year Past 30 Days
Abstract Views 1464 0 0
Full Text Views 2217 886 31
PDF Downloads 636 262 29
EPUB Downloads 0 0 0