N. Panda1,2, I. Solsky1, J. P. Onnela3, A. B. Haynes1,2 1Ariadne Labs,Boston, MA, USA 2Massachusetts General Hospital,Surgical Oncology,Boston, MA, USA 3Harvard School Of Public Health,Biostatistics,Boston, MA, USA
Introduction:
Digital phenotyping is the moment-by-moment quantification of physical, social, behavioral, cognitive, and emotional functioning as reflected in smartphone data collected passively from smartphone sensors and actively through surveys. This approach to measuring individual functioning has the potential to improve shared decision-making, patient engagement, and improve outcomes that matter most to patients. The feasibility of active data gathering via surveys as acquired using digital phenotyping requires further investigation.
Methods:
Patients enrolled in a prospective study who underwent surgery for abdominal cancer downloaded the Beiwe smartphone app, developed by our research team for digital phenotyping. Beiwe, in addition to collecting passive data (i.e. GPS, accelerometer, call frequency) from smartphone sensors, pushed perioperative surveys to patients. Pre-operative surveys included the validated Short Form-36 (SF36) and Decision Conflict Scale (DCS) as well as a baseline survey on treatment expectations developed by our research team (ExpB). Post-operative surveys included a single item on the alignment of treatment expectations created by our team (ExpL, pushed at week 1, 4, 12, and 24 after surgery), SF36 (pushed at week 4, 12, and 24 after surgery), and the validated Decision Regret Scale (DRS, pushed at 12 and 24 weeks after surgery). Daily microsurveys (5-question survey generated randomly from SF36) were also administered perioperatively. Enrollment time and survey completion rates were analyzed. Responses gathered via intermittent SF36 versus daily microsurvey pushes were also compared.
Results:
Twenty patients with primary abdominal malignancies (10 sarcoma, 5 liver, 5 colon) who underwent surgery were followed for a median of 117.5 days [IQR 65.5, 158.0]. Pre-operative survey completion rates were 0.90 (±0.31), 0.90 (±0.31), and 0.95 (±0.22) for the SF36, ExpB, and DCS, respectively. Microsurvey completion rate throughout the study was 0.63 (±0.29). Post-operatively, the completion rates of the ExpL (pushed at 1 week), ExpL+SF36 (pushed at 4 weeks), and DRS+SF36+ExpL (pushed at 12 and 24 weeks) were 0.53 (±0.29), 0.71 (±0.32), 0.44 (±0.46), and 0.27 (±0.36), respectively. The mean number of items completed by patients through the intermittent complete SF36 administration was lower (75.9 (±38.5)) compared to daily microsurveys (347.8 (± 276.6), p=0.002).
Conclusion:
Active gathering of patient-reported quality of life outcomes is feasible several months into post-operative recovery in a digital phenotyping study. The daily administration of SF36 microsurveys resulted in a significantly greater number of total items completed throughout the study period in comparison to intermittent distribution of the SF36 in full. This suggests that microsurveys may allow for greater data collection; however, the validity of this approach must be studied further.