Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Feb 1.
Published in final edited form as: Ann Surg Oncol. 2020 Aug 18;28(2):985–994. doi: 10.1245/s10434-020-09004-5

Smartphone Global Positioning System (GPS) Data Enhances Recovery Assessment after Breast Cancer Surgery

Nikhil Panda 1,2,*, Ian Solsky 1,3,*, Becky Hawrusik 1, Greyson Liu 4, Harrison Reeder 4, Stuart Lipsitz 1, Eesha V Desai 1, Kurt W Lowery 1, Kate Miller 1, Michele A Gadd 2, Carrie C Lubitz 2, Barbara L Smith 2, Michelle Specht 2, Jukka-Pekka Onnela 1,4,**, Alex B Haynes 1,5,**
PMCID: PMC7855656  NIHMSID: NIHMS1621617  PMID: 32812109

Abstract

Purpose

We sought to determine whether smartphone GPS data uncovered differences in recovery after breast-conserving surgery (BCS) and mastectomy, and how these data aligned with self-reported quality of life (QoL).

Methods

In a prospective pilot study, adult smartphone-owners undergoing breast surgery downloaded an application that continuously collected smartphone GPS data for 1 week preoperatively and 6 months postoperatively. QoL was assessed with the Short-Form-36 (SF36) via smartphone delivery preoperatively and 4 and 12 weeks postoperatively. Endpoints were trends in daily GPS-derived distance traveled and home time, as well as SF36 Physical (PCS) and Mental Component Scores (MCS) comparing BCS and mastectomy patients.

Results

Thirty-one patients were included. Sixteen BCS and fifteen mastectomy patients were followed for a mean of 201 (SD 161) and 174 (107) days, respectively. There were no baseline differences in demographics, PCS/MCS, home time, or distance traveled. Through 12 weeks postoperatively, mastectomy patients spent more time at home [e.g., week 4: 16.7 h 95%CI (14.3,19.6) versus 11.0 h (9.4,12.9), p < 0.001] and traveled shorter distances [e.g., week 4: 52.5 km 95%CI (36.1,76.0) versus 107.7 km (75.8–152.9), p = 0.009] compared with BCS patients. There were no significant QoL differences throughout the study as measured by the MCS [e.g., week 4 difference: 7.83 95%CI (−9.02,24.7), p = 0.362] or PCS [e.g., week 4 difference: 8.14 (−6.67,22.9), p = 0.281]. GPS and QoL trends were uncorrelated (< ± 0.26, p > 0.05).

Conclusions

Differences in BCS and mastectomy recovery were successfully captured using smartphone GPS data. These data may describe currently unmeasured aspects of physical and mental recovery, which could supplement traditional and QoL outcomes to inform shared decision-making.

INTRODUCTION

Many patients with resectable breast cancer are offered treatment with breast-conserving therapy [breast conserving surgery (BCS) with adjuvant radiation therapy] or mastectomy, two strategies that result in equivalent survival, but may have notable differences in postoperative physical, emotional, and social quality of life (QoL)1–7. There is a paucity of empirical data on the impact of breast surgery on these aspects of recovery. Surgeons, therefore, prepare patients for recovery using traditional outcome measures (e.g., long-term oncologic outcomes or morbidity rates), which may not completely capture aspects of postoperative physical and psychosocial health. Clinicians, therefore, recommend patients make treatment decisions based on personal goals and values, which encompass recovery outcomes that currently remain difficult to measure. As such, patients frequently enter treatment with unmet information needs and misaligned expectations, which may negatively influence QoL9.

There is a growing interest in harnessing the increased prevalence and technologic sophistication of smartphones to help address these nuanced aspects of postoperative recovery10. Smartphone-generated behavioral data may be used to inform various aspects of physical and psychosocial health. Our research team previously introduced smartphone-based digital phenotyping and how these data may be used to describe currently unmeasured aspects of postoperative recovery11–13. For example, smartphone accelerometer data can quantify differential postoperative physical activity among patients experiencing postoperative events (e.g., morbidity, readmission, reoperation)14,15. Previous studies in non-surgical patients have suggested that GPS data may be used to measure physical and psychosocial health outcomes. In terms of surgical recovery, the amount of distance traveled each day may reflect changes in gross mobility, pain, and fatigue. Similarly, the amount of time spent at home may be associated with loneliness, community engagement, and amount of social interactions16–20. However, it remains unknown whether these data are clinically meaningful among patients recovering from surgery and, importantly, how they align with or supplement QoL as assessed by patient-reported outcome measures (PROMs)21.

The objective of this pilot study was to determine whether high frequency sampling of smartphone GPS data could describe differences in recovery among patients undergoing BCS versus mastectomy. Furthermore, we aimed to understand the alignment between perioperative trends in passively collected smartphone GPS data and patient-reported physical and mental QoL. We hypothesized that smartphone GPS data could be used to derive metrics that would capture suspected differences in recovery among patients after BCS versus mastectomy, and that these measurements would quantify novel aspects of postoperative recovery. By achieving these aims, these data may improve surgeons’ and patients’ understanding of recovery after breast cancer surgery.

Methods

Inclusion and Exclusion Criteria

Patients were approached if scheduled to undergo surgery for suspected or confirmed breast cancer between July 2017 and December 2018 at a single, academic, tertiary-care cancer center in the United States. Patients with metastatic disease were excluded. English-speaking adult (age ≥ 18 years) iOS- or Android-based smartphone owners who were scheduled for BCS or mastectomy were eligible for inclusion. Patients were provided with printed study materials during the initial outpatient visit with their surgeons and were then contacted by members of the research team (NP/IS) to further discuss the study (e.g., potential benefits for postoperative recovery, minimal risk, steps to maximize patient privacy). Verbal consent was obtained from all patients, each of whom downloaded a smartphone application, Beiwe. Beiwe is an open source research platform created by our research team for high-throughput smartphone-based digital phenotyping intended for use in biomedical research13. The application continuously and passively samples data from various smartphone sensors (e.g., GPS, accelerometer, gyroscope) and logs (e.g., communication logs, screen activity logs) in their raw form, and also allows for active data collection from patients using audio diary entries and surveys. This prospective cohort pilot study was approved by the Partners Human Research Committee.

Collection of Clinical and GPS Data

Medical records of enrolled patients were reviewed for demographic, baseline characteristics, cancer- and disease-specific variables, operative details, and postoperative clinical outcomes. All data were stored on Research Electronic Data Capture (REDCap)22. Throughout the 25-week study period (1 week preoperatively and 6 months postoperatively), smartphone GPS sensor data were collected passively and prospectively through the Beiwe application. These data were processed to determine daily “home time” and “distance traveled” for each enrolled patient. Home time was defined as the number of hours each day when patients were within a 200-m radius of their home23. The second metric, distance traveled, was defined as the daily total distance traveled by a patient irrespective of mode of transportation (e.g., walking, cycling, driving). A detailed description of the sampling scheme and procedure for computing these summary statistics are provided in Appendix 1.

Collection of QoL Data

QoL was assessed through smartphone delivery of the Short-Form-36 (SF36, version 1) preoperatively and at 4 and 12 weeks postoperatively. The SF36 is a 36-item survey used to quantify perceived generic health24–26. Survey responses provide QoL scores in eight health domains: (1) physical functioning, (2) physical role limitations, (3) bodily pain, (4) general health, (5) vitality, (6) social functioning, (7) mental health, and (8) emotional role limitations. Weighted sums of domain scores derived from validated algorithms can be used to generate physical (PCS) and mental component (MCS) scores.27,28.

All passively collected GPS data and actively self-reported QoL data were encrypted and securely stored in compliance with the Health Insurance Portability and Accountability Act (HIPAA). Our research team has previously described the processes of data upload, asymmetric encryption, and storage and how they have been optimized for maximum patient safety15.

Exposure and Outcome Measures

GPS-derived metrics and QoL data were compared among patients scheduled to undergo BCS (resection of lesion and margin of normal tissue often with adjuvant radiation therapy) versus mastectomy (resection of the entire affected breast) with or without surgical lymph node assessment as the initial surgical treatment for confirmed or suspected breast cancer. The primary endpoints were postoperative trends in GPS-derived metrics (home time, distance traveled) and self-reported QoL as measured by SF36 PCS and MCS scores in the first 12 weeks of recovery. In addition, data from the entire cohort of both BCS and mastectomy patients were analyzed to determine correlations between passively collected GPS and actively reported QoL data. This was performed in order to understand whether observed differences in GPS-derived metrics may reflect differences in self-reported QoL.

Statistical Analysis

Baseline data were summarized using means with standard deviation or medians with interquartile ranges for continuous measures and frequencies with percentage for categorical variables. To examine differences in baseline characteristics between patients who underwent BCS versus mastectomy, exact p-values for small samples were calculated for means (two-sample exact t-tests), medians (Wilcoxon tests) and proportions (Fisher’s exact test)29. Log-normal and gamma distributions were fitted to the rightward-skewed GPS data; the gamma distribution was ultimately chosen using the Akaike Information Criterion30. A generalized linear mixed model using the gamma distribution was fitted with a log-link for the outcomes with patient random effects to account for repeated measures on a patient. To determine the trends in GPS-derived metrics over time, separate unadjusted restricted cubic spline regression models with three equally spaced knots were applied to the BCS and mastectomy groups. From these models, the predicted means at three time points (2, 4, and 12 weeks postoperatively) were compared.

For the QoL analysis, SF36 surveys that were completed within 1 week of the assigned dates (preoperatively, 4, and 12 weeks postoperatively) were included. For the SF36 PCS and MCS component scores, a repeated measures mixed model was fitted with a linear link and the following covariates: a treatment group indicator, categorical time variables, and an interaction term between group and time. From the mixed model, differences in these scores at baseline, 4, and 12 weeks postoperatively were compared among BCS and mastectomy groups using a robust Wald test that does not require the outcome data to be normally distributed. Correlations between 4- to 12-week differences in predicted GPS-derived metrics and 4- to 12-week differences in the PCS and MCS component scores were calculated, pooling over groups. Data from patients who underwent BCS and subsequent completion mastectomy during the study period were analyzed according to intention-to-treat principles. All statistical analyses were performed using SAS software (version 9.4). Statistical significance was defined as a two-sided p-value < 0.05 after multiple testing correction using the Benjamini-Hochberg step-up Bonferroni method31.

RESULTS

Patient Selection

Seventy-two patients scheduled for elective breast surgery for suspected or confirmed malignancy were approached for participation. Of the 62 eligible patients, 33 patients consented to participate in the study (53% enrollment rate). Thirty-one patients (16 BCS, 15 mastectomy) were ultimately included in final analyses (Fig. 1). Sixteen BCS patients were followed for a mean of 201 days [standard deviation (SD) 161] and fifteen mastectomy patients were followed for a mean of 174 days (SD 107).

Fig. 1.

Fig. 1.

Open in a new tab

Selection of patients for inclusion. Shown are reasons for exclusion and barriers to obtaining informed consent for participation

Surgical and Clinical Data

There were no differences in demographics among patients who underwent BCS versus mastectomy (Table 1). Patients who underwent mastectomy had greater blood loss [mean 68.2 ml (SD 61.5) versus 10.1 (SD 7.9), p < 0.001] and longer operations [mean 146.7 min (SD 55.7) versus 50.7 (SD 18.2), p < 0.001] compared with those who underwent BCS, given the majority were bilateral [11/15 (73%)] and underwent direct-to-implant reconstruction [9/15 (60%)]. Of the remaining six patients who underwent mastectomy, three underwent immediate placement of tissue expander followed by scheduled interval implant exchange and three chose mastectomy without reconstruction. Patients who underwent mastectomy were also more likely to have lymph nodes sampled through sentinel lymph node biopsy or lymphadenectomy [15 patients (100%) versus 11 (69%), p = 0.008]. Perioperatively, patients who underwent mastectomy had longer hospital length of stay [mean 1.9 days (SD 1.0) versus 0, p < 0.001] and more frequently received visiting nursing services [13 patients (87%) versus 0, p < 0.001] compared with those who underwent BCS. Following mastectomy, all patients were discharged with a surgical drain; the mean duration of drains was 14.2 (SD 3.1) days.

Table 1.

Baseline characteristics and clinical outcomes of BCS and mastectomy patients included in final analyses. Weeks of available data refer to number of weeks where individuals had at least 5/7 days of GPS sensor data captured. Abbreviations: standard deviation (SD); interquartile range (IQR); body mass index (BMI); American Society of Anesthesiology (ASA); minutes (min); milliliters (ml).

BCS (n=16) Mastectomy (n=15)
Demographics
Age (mean ±SD) 52.0 (11.2) 53.4 (10.9)
Female sex (n, %) 16 (100) 15 (93)
Race/ethnicity (n, %)
 Non-Hispanic White 13 (82) 13 (86)
 Non-Hispanic Black 1 (6) 0 (0)
 Hispanic 1 (6) 1 (7)
 Asian 0 (0) 1 (7)
 Unknown 1 (6) 0 (0)
BMI (kg/m2, mean ±SD) 26.1 (4.3) 26.3 (6.6)
Prior treatment (n, %)
 Chemotherapy 2 (13) 1 (7)
 Radiation 0 (0) 1 (7)
Operative details
ASA classification (mean ±SD) 1.7 (0.6) 1.9 (0.4)
Operative timea (min, mean, ±SD) 50.7 (18.2) 146.7 (55.7)
Blood lossa (ml, mean, ±SD) 10.1 (7.9) 68.2 (61.5)
Lymph node samplinga (n, %) 11 (69) 15 (100)
Immediate reconstruction (n, %) n/a 9 (60)
Final surgical pathology (n, %)
 Invasive carcinoma 11 (69) 11 (73)
 Carcinoma in situ 2 (13) 3 (20)
 Other 3 (18) 1 (7)
Recovery Information
Hospital length of staya (mean ±SD) 0 (0) 1.9 (1.0)
Discharge with visiting nursea (n, %) 0 (0%) 13 (86)
Perioperative morbidity (n, %) 0 (0) 2 (13)
Emergency room visit (n, %) 0 (0) 1 (7)
Return to operating room (n, %) 4 (25) 4 (27)
Adjuvant treatment (n, %)
 Chemotherapya 2 (13) 8 (53)
 Radiation 10 (63) 5 (33)
Number of follow-up visits (mean ±SD)
 With surgeon 2 (1.7) 2 (2.4)
 With multidisciplinary team 9 (6.8) 10.7 (4.9)
Smartphones and data volume
Smartphone operating system (n, %)
 iPhone 16 (100) 14 (93)
 Android 0 (0) 1 (7)
Available data (weeks, mean ±SD)
 Passively collected GPS sensor data 18.4 (9.6) 19.9 (9)
 Actively self-reported SF36 survey dataa 1.7 (0.9) 2.7 (1.3)
Open in a new tab
a

Denotes statistically significant differences (two-sided p-value < 0.05)

During postoperative follow-up, four patients (25%) who received BCS required reoperation based on pathologic margin status: two underwent re-excision and two chose to undergo completion mastectomy. Reoperations occurred at a mean of 4.3 (SD 1.8) weeks after the initial operation. Following mastectomy, four patients (27%) required reoperation: one based on pathologic margins and axillary hematoma, and three for scheduled tissue expander to implant exchange. Reoperations occurred at a mean of 11.6 (SD 5.2) weeks postoperatively. Among patients who underwent BCS, 10 patients (63%) completed breast conserving therapy and received adjuvant radiation [start of treatment: mean 8.4 (SD 4.8) weeks postoperatively] and two patients (13%) received adjuvant chemotherapy [start of treatment: mean 5.2 (SD 4.8) weeks postoperatively]. Following mastectomy, five patients (33%) received adjuvant radiation [start of treatment: mean 18.5 (SD 8.2) weeks postoperatively] and eight patients (53%) received adjuvant chemotherapy [start of treatment: mean 7.7 (SD 4.8) weeks postoperatively].

GPS-Derived Recovery Metrics

There were no significant preoperative differences between patients scheduled to undergo BCS versus mastectomy in home time [mean home time 13.9 h 95% CI (9.01, 21.1) versus 17.4 (9.86, 23.4), p = 0.806] and distance traveled [mean distance 69.5 km 95% CI (37.7, 127.3) versus 68.9 (29.9, 157.5), p = 0.821]. Estimated trends in postoperative home time and distance traveled are shown in Figs. 2 and 3, respectively. Compared with those recovering from BCS, patients recovering from mastectomy spent a greater number of hours at home 2 and 4 weeks after surgery [e.g., week 4 home time: 16.7 hours 95% CI (14.3,19.6) versus 11.0 (9.4,12.9), p < 0.001]. Similarly, patients traveled shorter distances during their recovery after mastectomy compared with those patients recovering from BCS [week 4 distance: 52.5 kilometers 95% CI (36.1,76.0) versus 107.7 (75.8–152.9)]. There were no differences in either GPS-derived recovery metrics between patients recovering from BCS versus mastectomy 12 weeks after surgery.

Fig. 2.

Fig. 2

Open in a new tab

Trends in postoperative daily home time in first 12 weeks after surgery as derived from smartphone GPS data. Hollow circles for each trend represent daily mean home time (hours) among patients recovering from BCS (n = 16) or mastectomy (n = 15)

Fig. 3.

Fig. 3

Open in a new tab

Trends in postoperative daily distance traveled in first 12 weeks after surgery as derived from smartphone GPS data. Hollow circles for each trend represent daily mean distance traveled (km) among patients recovering from BCS (n = 16) or mastectomy (n = 15). Note, 15 observations above 200 km distance traveled censored with adjustment of scale

Correlation Between GPS-Derived Data and Self-Reported QoL

Overall response rates to the QoL assessments were 77% (preoperatively), 68% (week 4 postoperatively), and 45% (week 12 postoperatively, Table 2). Figure 4 shows the predicted estimates of self-reported physical (PCS) and mental (MCS) QoL. There were no differences within or between BCS and mastectomy cohorts’ PCS and MCS scores at baseline or during recovery at 4 and 12 weeks. Correlation testing of 4- to 12-week differences in GPS-derived and QoL data in the combined cohort of BCS and mastectomy patients demonstrated that trends in home time and distance traveled were not well-correlated with changes in PCS and MCS component scores ( < ± 0.26, p > 0.05 for all, Table 3).

Table 2:

Response rates to quality of life assessments among BCS and mastectomy patients included in final analysis

Quality of life assessment BCS (n=16) Mastectomy (n=15) Overall (n=31)
Preoperative (n, %) 10 (63) 14 (93) 24 (77)
Postoperative, week 4 (n, %) 10 (63) 11 (73) 21 (68)
Postoperative, week 12 (n, %) 5 (31) 9 (60) 14 (45)
Open in a new tab

Fig. 4.

Fig. 4

Open in a new tab

Differences in self-reported QoL after BCS and mastectomy. Differences in PCS (left) and MCS (right) component scores between BCS and mastectomy (n = 15) patients are shown preoperatively (graphically at week 0) and at 4 and 12 weeks postoperatively

Table 3.

Correlation testing of 4- to 12-week trends in GPS-derived metrics and self-reported QoL as measured by SF36 PCS and MCS scores. Correlation coefficients () with 95% confidence intervals are provided

Parameter Home time Distance traveled
Home time 1 −0.44 [−0.88, 0.00]
p=0.050
Distance traveled −0.44 [−0.88, 0.00]
p=0.050		 1
PCS −0.02 [−0.72, 0.67]
p=0.944		 0.26 [−0.22, 0.75]
p=0.281
MCS −0.08 [−0.81, 0.63]
p=0.814		 −0.16 [−0.63, 0.30]
p=0.482
Open in a new tab

DISCUSSION

The results of this study demonstrate how two smartphone GPS-derived metrics, home time and distance traveled, described differences in aspects of recovery among patients who underwent BCS versus mastectomy for breast cancer. These differences in recovery, which may represent aspects of physical and psychosocial health, existed despite no differences in self-reported QoL. These data suggest that passively collected smartphone sensor data may supplement the recovery information that is captured by PROMs and traditional surgical outcome measures to create a more in-depth picture of recovery from these operations.

Previous studies have suggested that GPS-derived metrics may be used to improve patient-centered care by serving as surrogates for both physical and psychosocial health outcomes16–18. A greater amount of time spent at home may represent fatigue or a lack of motor functioning19. Home time may also describe a patient’s social support during recovery, as other studies have demonstrated the association between increased home time and higher self-reported loneliness20. Petersen et al. used a passive in-home digital platform of infrared sensors combined with PROMs and found that more hours at home each day was associated with worse physical ability and emotional state32. Like home time, GPS-derived distance metrics have similarly been utilized to describe an individual’s scope of mobility and community engagement, suggesting that lower distances traveled may signal potential challenges related not only to physical, but also emotional and cognitive health33. These considerations are especially important during recovery after cancer surgery, as they may influence loss to follow-up and ability to comply with adjuvant care. For example, the observed differences in home time and distance traveled between patients recovering from BCS versus mastectomy may reflect the requisite mobility to participate in prescribed adjuvant therapy. The results of this study add to these findings by applying GPS data to describe differential recovery patterns after distinct surgical interventions. Additionally, unlike non-smartphone based methods of GPS data collection, which require significant resources and supportive infrastructure, we describe a methodology that harnesses the near ubiquity of smartphone ownership worldwide20.

The trends in the passively collected smartphone GPS sensor data and actively reported QoL outcomes were shown to be uncorrelated in the current study. Previous research in correlating QoL with physical activity metrics measured by wearable digital activity trackers (e.g., step counts) has demonstrated moderate correlation with certain generic health domains34. However, this study suggests that differential aspects of postoperative recovery seen in passively collected GPS data were not captured in self-reported QoL using the SF36. There are several potential explanations for contextualizing these findings. One of the most compelling is that GPS-derived metrics and the SF36 PCS and MCS component scores may quantify different aspects of physical and psychosocial recovery. The lack of statistically significant differences in PCS and MCS component scores between patients recovering from BCS versus mastectomy may indicate, in broad terms, similar overall postoperative QoL outcomes. It is possible that adjuvant therapy following BCS represents similar treatment intensity as mastectomy, which is then reflected in the self-reported QoL.

However, the differences in home time and distance traveled also make apparent that the recovery experiences of BCS and mastectomy patients are not the same. If these nuanced findings are validated in larger studies, they could have significant implications for shared decision-making. For patients to make informed decisions, they must understand the impact of different treatments in patient-centered terms. For certain patients, home time and distance traveled likely have more meaning and are more easily interpreted compared with morbidity rates, or even PROMs. If so, these metrics can be used to bring patients’ goals and values to the forefront of discussions about treatment options, and provide surgeons with a concrete way to engage patients in shared decision-making. This, in turn, may also improve patient satisfaction and adherence to therapy while reducing undesired care35.

Other possible explanations for the differences seen in patient-reported and passively collected GPS data relate to the research tools and methods of this study. The SF36 is a generic QoL instrument and inclusion of a breast-, surgery-, or cancer-specific instrument may have improved sensitivity to measure changes in QoL36–38. QoL was assessed 3 times during the study period: preoperatively and at 4 and 12 weeks postoperatively. The optimal timing and frequency of QoL assessment through PROMs remains unknown, although it is possible that more frequent survey administration may have produced different results21,39. However, similar to the addition of disease- and treatment-specific instruments, the burden of more surveys on each patient must be considered. Lastly, the lack of correlation between passively collected GPS data and self-reported QoL may be a consequence of the small study sample. Both statistically significant and minimal clinically important differences may emerge in larger analyses40. Each of these considerations underscore that as new forms of patient-generated health data are being introduced into clinical practice, these data alone cannot be assumed to replace self-reported QoL. Rather, these data provide insights into nuanced aspects of recovery that may not be completely captured by currently available generic health instruments.

The results of this study must be interpreted in the context of the study design and its limitations. While there was a significant amount of passively collected smartphone sensor data for each patient, this was a relatively small study and the findings were not adjusted for any differences in demographics or postoperative events (e.g., type of reconstruction following mastectomy). The comparative analysis of PCS and MCS QoL scores may have approached statistical significance in a larger cohort; however, there were potentially clinically meaningful differences observed in the passively collected GPS sensor data despite the relatively small study base. Similarly, we introduced this pilot study within a single institution with a relatively homogeneous study base (e.g., 82–86% non-Hispanic White patients). We continue to enroll patients in smartphone-based digital phenotyping research, with efforts to target recruitment toward a more generalizable population. Given existing data showing similar ownership and usage of smartphones across social determinants of health, this approach to measuring nuanced aspects of postoperative recovery and QoL should be scalable with equity41.

In addition, we did not collect data on patients who opted out of the study, which may help inform feasibility and acceptability of a clinical intervention that uses passively collected sensor data or a smartphone platform. As hospitals begin to rapidly adopt and leverage digital tools for health delivery, patient perceptions must be considered. In addition, it is not possible to determine reasons for travel or whether patients were traveling with others. Future studies will incorporate reasons for travel (e.g., return to work, adjuvant treatment, etc.) from not only patients, but also their caregivers. Lastly, enrollment of patients occurred at the time of surgical consultation, resulting in fewer data points collected during the preoperative settings, thereby preventing a rich understanding of each patient’s baseline home time and distance traveled. Enrolling patients at an earlier date prior to surgery will allow for a better assessment of preoperative functioning.

CONCLUSION

The findings of this study have immediate implications for surgeons, patients, and innovators in an era where digital-health research continues to grow in popularity. These results demonstrate how smartphone-based patient-generated data can improve the way clinicians measure the impact of different surgical treatments for breast cancer. If successfully implemented into clinical practice, these data can help surgeons to improve counseling and better engage in shared decision-making by discussing aspects of recovery that remain currently unmeasured by traditional outcome measures and PROMs.

SYNOPSIS.

Differences in recovery after breast conserving surgery and mastectomy were successfully captured using smartphone GPS data. These data may describe currently unmeasured aspects of physical and mental recovery, which could supplement traditional and quality-of-life outcomes to inform shared decision-making.

ACKNOLWEDGEMENTS

This study was supported by a National Institutes of Health Research Training in Alimentary Tract Surgery grant (grant T32 DK007754-18 [NP]), an Ariadne Labs Spark Grant funded by the Paul G. Allen Family foundation (NP, IS, JPO, ABH), and the National Institutes of Health/National Institute of Mental Health [grant 1DP2MH103909 (JPO)].

Conflict of Interests/Disclosures: NP reports a contract agreement with Aptima, a human-centered engineering and performance assessment contractor of the DARPA/Department of Defense. JPO receives sole compensation as a member of Harvard University. His research at Harvard T.H. Chan School of Public Health is supported by research awards from the NIH, Otsuka Pharmaceutical, Boehringer Ingelheim, and Apple. He received an unrestricted gift from Mindstrong Health in 2018. He is cofounder and board member of a recently established commercial entity that operates in digital phenotyping.

Appendix 1. GPS Data Sampling and Procedure Used to Process Raw GPS Data for Computing Daily Home Time and Distance Traveled

GPS data were sampled from smartphones by alternating between on-cycles (GPS data sampled) and off-cycles (no GPS data sampled). In this study the duration of the on-cycle was set to 2 min and the duration of the off-cycle to 10 min. This simple sampling scheme enables collection of enough data to infer mobility patterns while minimizing smartphone battery drainage23. Importantly, this approach also enables quantification of the extent of missingness in data that is not due to design, but due to behavioral or other factors, such as patients turning off their phones. Patients were encouraged to use and interact with their phones as they normally would for the duration of the study.

GPS data were processed to compute daily “home time” (in hours) and “distance traveled” (in kilometers) metrics23. Raw GPS data were collected at specific intervals according to a data collection schedule specified by the study design. Each sample recorded subjects’ latitude and longitude that were converted into a mobility trace defined by a sequence of flights and pauses. Flights were defined to be segments of linear movement and pauses were defined to be periods of time when a person did not move. Curved movement was approximated by multiple sequential flights. Also, if a missing interval was flanked by two pauses at the same location (situated within 50 m of one another), the missing interval was assumed to be a longer pause at the same location. Because the data stream has long periods of structured missingness, we next used a resampling method to estimate a complete connected path of flights and pauses. Specifically, each missing period was filled with random draws from the subjects’ empirical distributions of observed flights and pauses to create complete paths that reflected individuals’ observed mobility patterns. The resulting trajectories were summarized into fifteen daily summary statistics or metrics. Given the aims of this study, we selected two specific metrics that potentially describe aspects of postoperative recovery. The “distance traveled” metric is computed as the sum of the lengths (in kilometers) of the flights in each day (irrespective of mode of transportation). To compute “home time,” we first fixed the location of each subject’s home by identifying the set of significant locations the subject visited. In order to determine the set of all significant locations for a person, we ran a ‘k-means’ clustering algorithm on the set of all pause locations with a minimum duration of 10 minutes to identify a small number of locations where the subject spent the most time. The significant location with the largest total amount of time during the night hours between 9PM and 6AM over the course of the study period was assumed to be the location of the person’s home. The “home time” metric was the amount of time (in hours) per day spent within a 200-m radius of home.

Footnotes

Previous Presentation: Results from this study were presented as an oral podium presentation at the 2020 Academic Surgical Congress.

Publisher's Disclaimer: This Author Accepted Manuscript is a PDF file of an unedited peer-reviewed manuscript that has been accepted for publication but has not been copyedited or corrected. The official version of record that is published in the journal is kept up to date and so may therefore differ from this version.

REFERENCES

  • 1.Veronesi U, Cascinelli N, Mariani L, et al. Twenty-Year Follow-up of a Randomized Study Comparing Breast-Conserving Surgery with Radical Mastectomy for Early Breast Cancer. N Engl J Med. 2002;347(16):1227–1232. doi: 10.1056/NEJMoa020989 [DOI] [PubMed] [Google Scholar]
  • 2.Blichert-Toft M, Nielsen M, Düring M, et al. Long-term results of breast conserving surgery vs. mastectomy for early stage invasive breast cancer: 20-year follow-up of the Danish randomized DBCG-82TM protocol. Acta Oncol (Madr). 2008;47(4):672–681. doi: 10.1080/02841860801971439 [DOI] [PubMed] [Google Scholar]
  • 3.Litière S, Werutsky G, Fentiman IS, et al. Breast conserving therapy versus mastectomy for stage I-II breast cancer: 20 year follow-up of the EORTC 10801 phase 3 randomised trial. Lancet Oncol. 2012;13(4):412–419. doi: 10.1016/S1470-2045(12)70042-6 [DOI] [PubMed] [Google Scholar]
  • 4.Michael YL, Kawachi I, Berkman LF, Holmes MD, Colditz GA. The persistent impact of breast carcinoma on functional health status. Cancer. 2000;89(11):2176–2186. doi: [DOI] [PubMed] [Google Scholar]
  • 5.Lee ES, Lee MK, Kim SH, et al. Health-related quality of life in survivors with breast cancer 1 year after diagnosis compared with the general population: a prospective cohort study. Ann Surg. 2011;253(1):101–108. doi: 10.1097/sla.0b013e3181f662ce [DOI] [PubMed] [Google Scholar]
  • 6.Abrahams HJG, Gielissen MFM, Schmits IC, Verhagen CAHHVM, Rovers MM, Knoop H. Risk factors, prevalence, and course of severe fatigue after breast cancer treatment: a meta-analysis involving 12 327 breast cancer survivors. Ann Oncol. 2016;27(6):965–974. doi: 10.1093/annonc/mdw099 [DOI] [PubMed] [Google Scholar]
  • 7.Montazeri A Health-related quality of life in breast cancer patients: a bibliographic review of the literature from 1974 to 2007. J Exp Clin Cancer Res. 2008;27(1):32. doi: 10.1186/1756-9966-27-32 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Lee CN, Chang Y, Adimorah N, et al. Decision Making about Surgery for Early-Stage Breast Cancer. J Am Coll Surg. 2012;214(1):1–10. doi: 10.1016/j.jamcollsurg.2011.09.017 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Oskay-Ozcelik G, Lehmacher W, Konsgen D, et al. Breast cancer patients’ expectations in respect of the physician-patient relationship and treatment management results of a survey of 617 patients. Ann Oncol. 2006;18(3):479–484. doi: 10.1093/annonc/mdl456 [DOI] [PubMed] [Google Scholar]
  • 10.Iribarren SJ, Schnall R. Health Informatics: An Interprofessional Approach. In:; 2017:255.
  • 11.Onnela J-P, Rauch SL. Harnessing Smartphone-Based Digital Phenotyping to Enhance Behavioral and Mental Health. Neuropsychopharmacology. 2016;41(7):1691–1696. doi: 10.1038/npp.2016.7 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Panda N, Solsky I, Haynes AB. Redefining shared decision-making in the digital era. Eur J Surg Oncol. July 2019. doi: 10.1016/j.ejso.2019.07.025 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Torous J, Kiang MV, Lorme J, Onnela J-P. New Tools for New Research in Psychiatry: A Scalable and Customizable Platform to Empower Data Driven Smartphone Research. JMIR Ment Heal. 2016;3(2):e16. doi: 10.2196/mental.5165 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Panda N, Solsky I, Huang EJ, et al. Passively Collected Smartphone Sensor Data to Detect Postoperative Events after Cancer Surgery: A Prospective, Multicenter, Proof-of-Principle Study. J Am Coll Surg. 2019;229(4):S159–S160. doi: 10.1016/j.jamcollsurg.2019.08.352 [DOI] [Google Scholar]
  • 15.Panda N, Solsky I, Huang EJ, et al. Using Smartphones to Capture Novel Recovery Metrics After Cancer Surgery. JAMA Surg. October 2019:1–7. doi: 10.1001/jamasurg.2019.4702 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Armstrong KA, Coyte PC, Brown M, Beber B, Semple JL. Effect of Home Monitoring via Mobile App on the Number of In-Person Visits Following Ambulatory Surgery. JAMA Surg. 2017;152(7):622. doi: 10.1001/jamasurg.2017.0111 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Hyder JA, Hirschberg RE, Nguyen LL. Home Discharge as a Performance Metric for Surgery. JAMA Surg. 2015;150(2):96. doi: 10.1001/jamasurg.2014.1725. [DOI] [PubMed] [Google Scholar]
  • 18.Myles PS, Shulman MA, Heritier S, et al. Validation of days at home as an outcome measure after surgery: a prospective cohort study in Australia. BMJ Open. 2017;7(8):e015828. doi: 10.1136/bmjopen-2017-015828 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Yurkiewicz IR, Simon P, Liedtke M, Dahl G, Dunn T. Effect of Fitbit and iPad Wearable Technology in Health-Related Quality of Life in Adolescent and Young Adult Cancer Patients. J Adolesc Young Adult Oncol. 2018;7(5):579–583. doi: 10.1089/jayao.2018.0022 [DOI] [PubMed] [Google Scholar]
  • 20.Petersen J, Austin D, Kaye JA, Pavel M, Hayes TL. Unobtrusive in-home detection of time spent out-of-home with applications to loneliness and physical activity. IEEE J Biomed Heal informatics. 2014;18(5):1590. doi: 10.1109/JBHI.2013.2294276 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Panda N, Haynes AB. Prioritizing the Patient Perspective in Oncologic Surgery. Ann Surg Oncol. 2019:6–7. doi: 10.1245/s10434-019-07753-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.REDCap: Citing REDCap.
  • 23.Barnett I, Onnela J-P. Inferring mobility measures from GPS traces with missing data. Biostatistics. October 2018. doi: 10.1093/biostatistics/kxy059 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Panda N, Rattner DW, Morse CR. Third-time (“redo-redo”) anti-reflux surgery: patient-reported outcomes after a thoracoabdominal approach. Surg Endosc. August 2019. doi: 10.1007/s00464-019-07059-4 [DOI] [PubMed] [Google Scholar]
  • 25.RAND. 36-Item Short Form Survey (SF-36). https://www.rand.org/health-care/surveys_tools/mos/36-item-short-form.html Accessed October 14, 2019.
  • 26.Contopoulos-Ioannidis DG, Karvouni A, Kouri I, Ioannidis JPA. Reporting and interpretation of SF-36 outcomes in randomised trials: systematic review. BMJ. 2009;338:a3006. doi: 10.1136/bmj.a3006 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Taft C, Karlsson J, Sullivan M. Do SF-36 summary component scores accurately summarize subscale scores? Qual Life Res. 2001;10(5):395–404. doi: 10.1023/A:1012552211996 [DOI] [PubMed] [Google Scholar]
  • 28.Laucis NC, Hays RD, Bhattacharyya T. Scoring the SF-36 in orthopaedics: A brief guide. J Bone Jt Surg - Am Vol. 2014;97(19):1628–1634. doi: 10.2106/JBJS.O.00030 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Mehta CR, Patel NR, Tsiatis AA. Exact Significance Testing to Establish Treatment Equivalence with Ordered Categorical Data. Biometrics. 1984;40(3):819. doi: 10.2307/2530927 [DOI] [PubMed] [Google Scholar]
  • 30.Akaike H A new look at the statistical model identification. IEEE Trans Automat Contr. 1974;19(6):716–723. doi: 10.1109/TAC.1974.1100705 [DOI] [Google Scholar]
  • 31.Multiple comparisons - Handbook of Biological Statistics. http://www.biostathandbook.com/multiplecomparisons.html Accessed February 20, 2020.
  • 32.Petersen J, Austin D, Mattek N, Kaye J. Time Out-of-Home and Cognitive, Physical, and Emotional Wellbeing of Older Adults: A Longitudinal Mixed Effects Model. PLoS One. 2015;10(10):e0139643. doi: 10.1371/journal.pone.0139643 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Wettstein M, Wahl H-W, Shoval N, et al. Out-of-Home Behavior and Cognitive Impairment in Older Adults. J Appl Gerontol. 2015;34(1):3–25. doi: 10.1177/0733464812459373 [DOI] [PubMed] [Google Scholar]
  • 34.Bade BC, Brooks MC, Nietert SB, et al. Assessing the Correlation Between Physical Activity and Quality of Life in Advanced Lung Cancer. Integr Cancer Ther. 2018;17(1):73–79. doi: 10.1177/1534735416684016 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.de Mik SML, Stubenrouch FE, Balm R, Ubbink DT. Systematic review of shared decision-making in surgery. Br J Surg. 2018;105(13):1721–1730. doi: 10.1002/bjs.11009 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Brady MJ, Cella DF, Mo F, et al. Reliability and validity of the Functional Assessment of Cancer Therapy-Breast quality-of-life instrument. J Clin Oncol. 1997;15(3):974–986. doi: 10.1200/JCO.1997.15.3.974 [DOI] [PubMed] [Google Scholar]
  • 37.FACT-B - Functional Assessment of Cancer Therapy - Breast Cancer. https://eprovide.mapi-trust.org/instruments/functional-assessment-of-cancer-therapy-breast-cancer Accessed October 15, 2019.
  • 38.Pusic AL, Klassen AF, Scott AM, Klok JA, Cordeiro PG, Cano SJ. Development of a New Patient-Reported Outcome Measure for Breast Surgery: The BREAST-Q. Plast Reconstr Surg. 2009;124(2):345–353. doi: 10.1097/PRS.0b013e3181aee807 [DOI] [PubMed] [Google Scholar]
  • 39.Goodwin PJ, Black JT, Bordeleau LJ, Ganz PA. Health-Related Quality-of-Life Measurement in Randomized Clinical Trials in Breast Cancer—Taking Stock. JNCI J Natl Cancer Inst. 2003;95(4):263–281. doi: 10.1093/jnci/95.4.263 [DOI] [PubMed] [Google Scholar]
  • 40.Antonescu I, Scott S, Tran TT, Mayo NE, Feldman LS. Measuring postoperative recovery: What are clinically meaningful differences? Surgery. 2014;156(2):319–327. doi: 10.1016/j.surg.2014.03.005 [DOI] [PubMed] [Google Scholar]
  • 41.Smith A. U.S. Smartphone Use in 2015 | Pew Research Center.

RESOURCES