Association between Patient-Reported Outcomes and Physical Activity Measured on the Apple Watch in Patients with Hematological Malignancies

Carrie A Thompson, Paul Novotny, Jeff A. Sloan, Alicia Bartz and Kathleen Yost



Patient-reported outcomes (PRO's) are important measures in patients with malignancies, but may be burdensome to complete. Wearable activity monitors provide objective, continuous activity data which may correlate with PRO's. The aims of this study were: 1) develop electronic technology to measure PRO's; 2) examine associations between PRO's and wearable technology data; and 3) develop emoji PRO scales.


Patients were recruited from Mayo Clinic hematology and oncology practices. Eligibility included: adult patients with a diagnosis of lymphoma, multiple myeloma, brain cancer, pancreatic cancer, breast cancer, and ovarian cancer within the past 5 years, life expectancy of >6 months, and own an iPhone version 5.0 or greater. All patients were provided with an Apple Watch and downloaded the study app onto the iPhone and watch at enrollment. Baseline PRO data was collected on the iPhone, including the PROMIS physical function, fatigue, sleep, social/role function short forms; single item linear analog self-assessment (LASA) of QOL, fatigue, and physical function; and emoji scales. Two emoji scales were developed for this study: a mood scale and a scale for physical, emotional, and overall QOL. Participants were instructed to wear the watch at least 8 hours/day. Activity levels were analyzed using the square root of the average daily values in order to minimize the effects of outliers. Univariate associations between PROs and activity levels were assessed using Spearman correlations. Multivariate associations were assessed using stepwise linear regression models. This analysis is reporting the baseline activity and PRO results of the lymphoma and myeloma patients recruited to date in this ongoing study.


From 2/2017-8/2017, 89 patients with lymphoma and 26 with multiple myeloma were recruited. 52% were male and median age was 55 years (range 21-79). The median time since diagnosis was 8.4 months (range 0-60); 74% were on active therapy, 16% were in observation post-treatment, and 10% had not been treated. Patients had an ECOG PS of 0 (47%), 1 (32%), 2 (12%), 3 (7%), and 4 (3%). The majority (64%) of patients were overweight or obese. All patients were "extremely" confident filling out forms, 99% had used a computer in the past 12 months, and 26% had ever used a smart watch.

During week 1 of the study, patients wore the watch for an average of 9.3 hours per day, did 3760 mean steps per day (SD 3417), exercised 8.3 minutes/day (SD 11.8), were sedentary 224.9 minutes/day (SD 154.2), and burned 115.8 kcal/day (SD 111.5). Mean heart rate was 82 beats per minute (SD 8.2).

There were significant correlations between PRO's and activity data; the strongest correlation was between steps per day and PROMIS physical function (Spearman correlation 0.33, p=0.0005). The association between steps per day and PROMIS physical function remained significant after adjusting for age and performance score (p=0.0076, figure 1). There were no variables significantly associated with active energy, flights, workout minutes, cycling distance, or heart rate.

Emoji responses were significantly associated with PROs; the Spearman correlations between the emoji and LASAs were -0.84 for fatigue, 0.68 for physical well-being, 0.72 for emotional well-being, and 0.77 for overall QOL (all p<0.0001).

Out of the 15 patients that have completed the study, all of them reported they would recommend filling out forms using the smart phones and watches, collecting data using emoji scales, and were very willing to complete future surveys using the watch compared to 87% by iPad, 33% by computer, 20% by paper, 20% by live telephone interview, and 0% by automated telephone interview.


Apple Watch activity data is significantly associated with PRO's in patients with lymphoma and myeloma, and emoji scales are a promising tool to collect PRO's. Collecting data utilizing wearable technology and smart phones is feasible and preferred by patients. Further studies will analyze the ability of activity data to predict longitudinal changes in PRO's.

Disclosures No relevant conflicts of interest to declare.

  • * Asterisk with author names denotes non-ASH members.