The association of smartphone-based activity space measures with cognitive functioning and pain sickle cell disease

Individuals living with sickle cell disease (SCD) are at risk for significant health complications including chronic pain, neuropsychological dysfunction, depression and poor mobility (Vichinsky et al, 2010; Zempsky et al, 2013; Jonassaint et al, 2016). Monitoring the day-to-day health of patients often requires burdensome daily symptoms diaries or expensive, time-consuming clinical tests. There are limited objective methods for capturing health outcomes of patients outside the clinic.

Location data obtained by global positioning systems (GPSs) are increasingly used to assess physical mobility and activity spaces (Loveday et al, 2015). Activity space was found to increase with improvements in functioning post-surgery (Barzilay et al, 2011) and is smaller in older adults with depressive symptoms (Kerr et al, 2012) and cognitive deficits (Wettstein et al, 2015). However, no reported studies have assessed the utility of GPS activity space measures in younger populations with a chronic condition, such as SCD. GPS could provide an objective measure of health and well-being for SCD patients, as well as a method for early identification of patients at risk for poor outcomes.

We recruited adult patients with SCD, aged 18–65 years, enrolled in a longitudinal study examining brain functioning in SCD. Participants completed a neuropsychological battery, had clinical measures, and an application tracking their location was installed on their smartphone. It was emphasized that their identifying data would be kept confidential and their location linked only to their study identification. All participants provided informed consent.

The application was compatible with both iOS and Android devices and recorded the location of the phone every two minutes based on GPS and Wi-Fi positioning capabilities. The data were processed and analysed using geographic information system software. Data were filtered for location accuracy. Participants with less than 2 weeks of valid data were excluded from analysis.

A standard deviation ellipse (SDE) of each participant was calculated in km² to represent the directional distribution of a series of location points. At one standard deviation, SDE will contain approximately 68% of all geographic points for an individual. The location samples and SDE of an example participant are visualized in Figure 1A.

Fig 1.

Data on global positioning system activity space and health outcomes among adult patients with sickle cell disease. A The entire dataset of a representative study participant. Each colour represents a different day. Standard deviation ellipse (SDE) can serve as an indicator for “activity space” and includes at least 68% of the samples by definition. Size of the SDE shown is 238·70 km² (map has been altered to de-identify the exact home location). B Scatter plot and fitted values for activity space score (log km²) and Center for Epidemiologic Studies Depression Scale (CES-D) depression scores. C Box plot showing distribution of activity space scores (km²) by patients reporting 0/10 pain versus those reporting 1/10 pain or greater. D Scatter plot and fitted values for activity space score (log km²) and haemoglobin levels (g/l). [Colour figure can be viewed at wileyonlinelibrary.com]

Clinical measures

A 5 m gait speed test screened for potential mobility issues. The digit symbol substitution test (DSST) measured processing speed (Wechsler, 2008). Higher scores indicate faster processing speed. Depressive symptoms were assessed by the 20-item Center for Epidemiologic Studies Depression Scale (CES-D), reported as the number of symptoms endorsed (Radloff, 1977). Participants verbally reported pain intensity on a 0 (least) to 10 (worst) scale. Haemoglobin (Hb) value was recorded from the most recent value entered in the medical records.

Statistical analysis

We calculated two SDEs for each participant: one for all days recorded and one for a representative 7-day period. Only data with spatially accurate and temporally intact information were analysed. SDE data were log-transformed to account for non-normal distribution (log km²). Pairwise Pearson correlations tested associations between log activity space and DSST scores, CES-D scores, pain scores and Hb. Due to the exploratory nature of the analyses, we considered a significance level of P < 0·1 as statistically meaningful and did not include covariates.

The application was installed on phones of 42 participants. A total of 37 participants (88%) had adequate spatial information from which SDEs could be calculated. Twenty-six participants (62%) had 13 or more consecutive days of valid data (range 13–227 days); 60% were female; average age 40 years (range 33–42). There were no differences between those with and without analysable data. Gait speed times ranged from limited to fully functional community ambulators (0·65–1·5 m/s).

The average activity space SDE was 52·91 km² (range 0.63 – 336.62). DSST was marginally correlated with total activity space (r = 0·31; P = 0·14; Fig 1B) and depressive symptoms were negatively correlated with week-long activity space (r = −0·40; P = 0·10; Fig 1C). Pain scores were not correlated with activity space; however, patients reporting 0/10 pain showed slightly greater activity than patients reporting ≥ 1/10 pain (Fig 1D). Haemoglobin showed a positive correlation with total activity space (r = 0·38; P = 0·07). There was no correlation between activity space and gait speed.

Few studies have examined the connection between SCD and mobility, and none have assessed objective measures of day-to-day activity. To our knowledge, this was the first study to test the feasibility and utility of GPS activity data as a means for health monitoring in SCD.

Our results validate the utility of activity space for assessing outcomes in SCD. Consistent with other GPS studies (Kerr et al, 2012; Wettstein et al, 2015), these data suggest that increased mobility is associated with more positive mood and cognition in SCD. Further, we found patients with lower Hb had smaller activity spaces. In another study of adults with SCD, six-minute walk distance was associated with Hb level and indicators of cardiovascular health (Marinho et al, 2016).

Although these preliminary findings are promising, feasibility of this study approach is limited by the challenges of using patient phones to capture GPS data, including phone incompatibility, insufficient storage space, inferior GPS hardware and short battery life. Despite these limitations, almost 90% of the phones produced reliable data and only 4 phones produced no data. In future studies using this technology, providing feedback and awarding status badges or reward tokens for successful days completed may increase compliance.

The current data are preliminary, but suggest the use of GPS activity space to identify patients at early risk for poor outcomes is promising. By adding additional streams of data, such as accelerometer data, number of text messages sent, phone talking time, and a diary application that records patient pain and mood symptoms, we may be able to construct statistical models that very accurately assess and predict clinical outcomes. Using these technologies will be important in determining which patients are most in need of intervention and how patients respond to changes in treatment without patients having to record, track and recognize these symptoms themselves. This study is an important first step to understanding how GPS can be used to monitor the health of patients living with SCD or other chronic conditions.