The value of patient-reported outcomes (PROs) on increasing physician awareness of patients’ functioning and well-being can be useful, but it can have drawbacks depending on the methods used to assess the PROs, according to several presentations at the “Health Services Research, Clinical Informatics, and Quality of Care” Oral Abstract Session, held June 1.
Dr. Fabrice Denis
An overwhelming majority of lung cancer relapses are symptomatic—somewhere between 75% and 90%, he said, but patients may have symptoms weeks before a follow-up visit with their treating physician. Moovcare is an algorithm based on the chaos theory model that reports on 12 symptoms every week, with notifications sent to nurses and/or oncologists.
“In a nonrandomized study, we assessed the survival of 98 patients after their treatment for a stage III/IV lung cancer with our algorithm,” Dr. Denis said. “Those first results suggested a 1-year survival that was 27% greater in patients with our web-mediated follow-up than in patients with a standard follow-up.”
The confirmatory phase III, randomized, multicenter study enrolled 133 patients with lung cancer (either small cell [SCLC] or non–small cell) who had internet access (12 were deemed ineligible after randomization); of the remaining 121 patients, median age was 65, 41% were receiving maintenance or tyrosine kinase inhibitors, and 17% had SCLC. At 9 months, OS improvement was observed and patients in the standard follow-up arm were eligible to cross over to the web-based arm.
After 2 years of follow-up, 70 deaths occurred and the median OS was 22.5 months in the web-based follow-up arm (60 patients) and 14.9 months in the standard follow-up arm (61 patients), without adjustment for cross-over from the control arm (HR 0.594, 95% CI [0.368, 0.959]; p = 0.03). Censoring cross over resulted in a hazard ratio of 0.496 (95% CI [0.305, 0.806]; p = 0.004).
Other studies have evaluated intensive follow-up, Dr. Denis said, “but our randomized study is the first assessing the survival rates as a primary outcome.” Future plans include initiating large (more than 1,000 patients), multicenter, international studies in other cancers; a similar program will be available soon for cancer screening among patients who smoke.
Emojis and the Apple Watch
Dr. Carrie A. Thompson
Recruited adult patients had a range of different cancers with a life expectancy of at least 6 months. All patients had an iPhone version 5.0 or higher and received an Apple Watch. Dr. Thompson and colleagues created the study app by using Apple’s ResearchKit, which is an open-source framework for building medical research, that the patients downloaded to their iPhones and watches.
“All patients were asked to wear the Apple Watch for a minimum of 8 hours per day for 12 weeks, from which we collected physical activity data,” Dr. Thompson said. “Patients were randomly assigned into three groups for mode of survey collection. These groups were stratified by cancer state—active treatment, survivorship, and observation alone—as these groups were likely to have different PRO and activity levels.”
Group 1 answered weekly surveys on paper only, Groups 2 and 3 answered weekly surveys on their iPhones, with Group 2 receiving daily emoji questions on their iPhone and Group 3 receiving daily emoji questions on the Apple Watch.
The weekly surveys included Patient-Reported Outcomes Measurement Information short forms: global health scale (physical and mental health); physical function, fatigue, sleep disturbance, social/role function; and anxiety. Additionally, patients were asked four single-item linear analog self-assessment questions: physical well-being, emotional well-being, fatigue, and quality of life.
The median age of the 294 patients was 53 (range, 20-79) and the median time since cancer diagnosis was 14.4 months. There were 99 patients in Group 1 (paper only), 98 patients in Group 2, and 97 patients in Group 3. The majority of patients were receiving active therapy, predominantly white, and predominantly female.
“In general, this was a computer-savvy group, but only 27% had used a smartwatch prior to the study,” Dr. Thompson said. Patients wore the watches an average of 9.8 hours a day, meaning capturing the activity data on the device was feasible.
Response rates over the 12-week study varied: Group 1 had an average response rate of 76%, Group 2 had an average of 77%, and Group 3, who completed their weekly surveys on the iPhone app but the emoji component on the watch, had a 60% response rate. In all groups, the response rates decreased over time. In Group 3, the response rate decreased to about 40% by week 12; the other two groups had a response rate of more than 60% at week 12.
Activity levels were analyzed using square root of average daily values in order to minimize the effect of outliers. The associations between PROs and activity levels were analyzed with Spearman correlations (SC) for univariate analysis, stepwise linear regression models for multivariate associations, and mixed models for longitudinal associations; in all, the researchers collected more than 21.4 million discrete data points.
“We found having more steps per day was associated with less fatigue, better physical function, better global physical well-being, better social function, and less sleep disturbance,” Dr. Thompson said. “Having more minutes of exercise per day was associated with better global mental well-being and less sleep disturbance.”
The researchers developed two emoji scales (one ordinal, one mood), each of which met the criteria for a valid ordinal scale. There was a high association between the emoji scale and fatigue (SC -0.80; p < 0.0001), between the emoji scale and physical function (SC 0.70; p < 0.0001), on the emotional scale (SC 0.68; p < 0.0001), and overall QOL (SC 0.75; p < 0.0001). Patients who selected the “happy face” emoji in the mood analysis had better overall ratings than other groups, whereas the “thinking face” group had the lowest QOL, physical health, and mental health, and the highest anxiety scores.
Dr. Thompson’s group will continue following patients every 3 months for 24 months for events, including relapse, retreatment, hospitalization, and death to determine associations between activity level and PROs with clinical outcomes.
“In addition, we are performing further analysis to understand the discrepancies in response rates between groups,” she said.
Lay Health Worker–Led Symptom Assessment
Dr. Manali I. Patel
Patients were enrolled if they were newly diagnosed with stage III or IV cancer, required medical oncology, planned to receive all of their care at the oncology practice where the assessment was implemented, and were enrolled in CareMore Medicare Advantage. The primary outcome was feasibility (retaining 75% of patients), with secondary outcomes of health care use, patient satisfaction, and self-reported health. There were a total of 186 patients in the intervention group and 102 patients in the control group. Baseline characteristics were similar between the two groups, but the intervention group had a higher risk adjustment factor (3.25 ± 1.78 vs. 2.80 ± 1.43, respectively).
Patients who received the proactive symptom assessment had significantly lower mean number of emergency department visits per quarter (0.60 vs. 0.92; p = 0.03) and lower mean number of inpatient admissions per quarter (0.72 vs. 1.02; p = 0.03).
“There was an almost $10,000 difference in total costs of care between the intervention group and the control group,” she said.
Self-reported overall health and mental/emotional health also improved at the 5-month follow-up compared to the 1-week post-oncology visit.
‘Important,’ But Unanswered Questions Remain
Discussant Martin J. Taphoorn, MD, of Haaglanden Medical Center and Leiden University Medical Center, Netherlands, called all three papers “important,” but he said they all left some unanswered questions. For instance, although web-mediated follow-up for PROs allows for easy assessment and earlier detection and treatment of symptoms, it’s not applicable for all patients. “There may be cross-cultural differences in the use of a web-based system,” he said.
Using mobile technologies or apps within the technology “may reduce the response burden and increase compliance,” he said. “But it is important to see in the longer-term follow up that there was still a difference in OS with or without cross-over. This could be cost-effective because it may reduce the number of CT scans.”
Although the Apple Watch activity data was associated with PROs and emojis are showing promise for PRO assessment, Dr. Taphoorn was concerned about the moderate response rate on the watch and with the lack of comparative data between the three groups.
“Emojis are easily interpretable, which could be beneficial for those with cognitive difficulties, illiteracy, or for children,” Dr. Taphoorn said.
Using a lay health worker to assess symptoms can improve overall health and patient satisfaction while substantially reducing health care use and costs, but the short-term results and lack of control group should not be overlooked.
–Michelle Dalton, ELS