SMART-ESAS: Smartphone Monitoring and Assessment in Real Time of Edmonton Symptom Assessment System Scores for Patients With Cancer

PURPOSE Serial patient-reported outcome (PRO) measurements in clinical practice are associated with a better quality of life and survival. Recording electronic PROs using smartphones is an efficient way to implement this. We aimed to assess the feasibility of the electronically filled Edmonton Symptom Assessment System (e-ESAS) scale in the lower-middle–income country (LMIC) setting. METHODS Baseline clinical features and conventional paper-based ESAS (p-ESAS) were collected in newly diagnosed patients with solid organ tumors. Text message link was sent to these patients for filling e-ESAS. ESAS was categorized into physical, psychological, and total symptom domains. Scores were divided into none to mild (0-3) and moderate to severe (4-10). Intraclass correlation coefficients (ICCs) were used to determine the correlation between p-ESAS and e-ESAS. Multivariable logistic regression was used to identify independent factors affecting symptom burden. RESULTS Of 1,160 participants who filled out p-ESAS, 595 completed both e-ESAS and p-ESAS questionnaires and were included in the final analysis. Moderate to severe physical, psychological, and total symptom scores were seen in 39.8%, 40%, and 39% of participants. Tiredness and anxiety were the most common physical and psychological symptoms, respectively. ICCs between the p-ESAS and e-ESAS varied between 0.75 and 0.9. Total symptom scores were independently predicted by metastatic disease (odds ratio [OR], 1.83; 95% CI, 1.26 to 2.67; P = .001) and a higher level of education (OR, 0.42; 95% CI, 0.25 to 0.72; P = .001). CONCLUSION Paper-based and electronically filled ESASs have good intraobserver reliability across individual symptoms and domain scores in a representative cohort at a tertiary care institute in the LMIC. This may help us incorporate e-ESAS in routine clinical care in the real-world setting with financial, infrastructural, and manpower limitations.


INTRODUCTION
Patients with cancer suffer from significant physical, psychological, and emotional symptoms that are attributed to both the disease and its treatment.][3][4] Patient-reported outcomes (PROs) are measurements of any aspect of a patient's health status that comes directly from the patient. 5They are a true reflection of the patient's physical, psychological, emotional, and social well-being, bypassing the interpretation of these responses by a health care provider, thus eliminating bias and interobserver variability. 6,7e Edmonton Symptom Assessment System (ESAS) 8 is one such scale designed for reporting, with intensity, the common symptoms of patients with cancer.It was originally developed in 1991 and has been revised and validated over the years for clinical and research purposes. 9It includes nine symptom domains, including six physical symptoms-pain, tiredness, nausea, drowsiness, appetite, and shortness of breath, and two psychological symptoms-depression and anxiety, along with one overall well-being.It has been validated in over 20 languages, including Hindi, [10][11][12] and is widely used in oncologic settings, especially in advanced malignancies. 13Os assist in documenting the symptom burden objectively, and their use in routine practice has been associated with better symptom control 4 and quality of life, 14 fewer hospitalizations, and better overall survival (OS). 6ESAS scores at baseline also have a prognostic value in metastatic cancers. 13ntegration of these scales into clinical practice improves symptom control and helps tailor treatment for each patient. 4ince its inception, ESAS measurement has involved paperbased collection using visual analog scales and symptom assessment graphs. 8However, routine implementation of these scales necessitates an efficient and user-friendly way to collect data.Even with sufficient data to advocate the implementation of routine PRO monitoring in clinical practice, it might not be feasible in the outpatient setting in centers with very high patient load, especially in resource-limited settings. 15ectronic PRO (ePRO) reporting using web-or mobilebased surveys is a convenient method to collect symptom burden scores. 16Online self-reporting has proven to be an effective long-term strategy for monitoring symptoms and making treatment decisions as well. 17They have also been associated with significantly better completion rates, 18 functional status, 19 and overall survival 20,21 compared with conventional surveillance using paper-based scales.Globally, more than 60% of users have shifted to using smartphones, including the lower-middle-income countries (LMICs) in the Asia-Pacific. 22According to the Global System for Mobile Communications Association's 2022 report, there are more than 1.5 billion smartphone users, with almost 90% having access to third generation (3G) or fourth generation (4G) network. 22,23Smartphone-based PROs can be a cost-effective way to monitor patients' symptoms, especially in resource-limited countries, where smartphone usage has increased significantly over the past few years.To our knowledge, there have been no data on the use of ePROs in LMICs, and their implementation and utility in such settings are largely unknown.We aimed to assess the feasibility and utility of the electronically filled ESAS scale (e-ESAS) in our setting.The objectives of the study were to (1) compare the consistency of PROs between the conventional paper-based ESAS questionnaire (p-ESAS) and e-ESAS filled on a smartphone and ( 2) analyze the clinical features associated with higher intensity of symptoms at baseline in patients with solid organ tumors.

Study Design and Population
This was a prospective study conducted at a tertiary care center in India.All patients older than 18 years with newly diagnosed cancer registered at our cancer center, Dr B.R.A. IRCH, All India Institute of Medical Sciences, New Delhi, between July 2021 and July 2022 were screened.Our center caters to the Northern states of India, with a predominantly Hindi-speaking population.We register and treat approximately 10,000 new cancer cases annually.Among these, patients with a confirmed diagnosis of solid organ cancer, who were able to comprehend and answer the questionnaire as assessed by the study investigators on the basis of an initial conversation regarding educational status, were included.Patients with hematolymphoid malignancies or those who received previous treatment were excluded from the study.The study was approved by the Institute Ethics Committee before its conduct.An informed consent was obtained from all participants before inclusion in the study.

Clinical Variables and ESAS Questionnaire
Baseline demographics and clinical features, including age, sex, residence-rural or urban (as reported by the patients themselves), level of education, distance traveled to reach the hospital, underlying comorbid conditions, addictions, primary diagnosis, and stage of the disease, were recorded.Level of education was categorized into illiterate-someone who cannot read and write 24 ; the ones who were literate were categorized into those who have completed primary, middle, or high school or those who have a graduation or postgraduation degree.Patients were asked to fill the ESAS in two formats-paper and electronic.All participants filled out the ESAS questionnaire on paper (p-ESAS) on the day of registration at our center.The e-ESAS questionnaire was then delivered as a text message with a unique link to those patients who possessed a smartphone (Appendix Fig A1).It could be answered either in English or in Hindi in both physical and digital formats on the basis of the patient's language preference. 11Each patient was instructed to rate each symptom from zero to 10, with zero being no symptom at all and 10 representing the worst possible severity.If the patients were unable to complete it independently, however, they were capable of providing responses with assistance, then the caregiver or health care provider involved in the patient's care was allowed to help.][9][10] This was based on the receiver operating characteristic curve, which defined a cutoff of four for combined moderate to severe symptoms. 13,26,27They were also categorized into three groups-physical (first six domains), psychological (next two domains), and total symptom scores. 28,29ubmitted p-ESAS questionnaires, which were <50% complete, were not included in the analysis.Incomplete questionnaires with more than 50% responses, subscores of the physical and psychological groups, and the total symptom scores were calculated using the average of the answered items. 13,27These were rounded off to the nearest integer and then divided into mild versus moderate to severe as above.For those that were completely filled, physical, psychological, and total scores were calculated by adding the individual domain scores and calculating the average scores.For e-ESAS, the  software was developed such that submission was not allowed until all the questions had been answered; hence, complete responses were obtained for all patients.

Statistics
Descriptive statistics were used to present demographics and baseline clinical details.Categorical data were described as percentages, and continuous variables were described as median with IQR.Comparison between groups was performed using the chi-square test for categorical variables and paired t-tests or the Wilcoxon signed-rank test for continuous data on the basis of the normality of the data distributed.The significant P value used was specified to be <.05 a priori.
Reliability assessment was performed using intraclass coefficients (ICCs, average measure, using the two-way mixed-effects model) between e-ESAS and p-ESAS to determine the correlation of the p-ESAS and e-ESAS measurements in each domain.Pooled mean and standard deviation were calculated, and ICC (or the tendency to be similar) was calculated as the ratio of the variance of interest to the total variance. 30By definition, ICC <0.5 is suggestive of poor reliability, 0.5-0.75 is moderate, 0.75-0.9 is good, and more than 0.9 is indicative of excellent reliability. 31Among the e-ESAS submissions, univariable logistic regression was applied to identify potential factors associated with moderate to severe symptoms.

Patient Characteristics
During the study period between July 2021 and July 2022, a total of 1,160 eligible patients were screened for participation in the study and p-ESAS responses were collected.Of these patients, 634 patients completed the e-ESAS questionnaire sent on their phones.Five hundred twenty-six patients did not complete the e-ESAS questionnaire.This was because they did not possess a smartphone, or know how to use a smartphone even if they possessed one, or due to some technical glitch in receiving the text message despite it being sent multiple times.Few participants did not wish to fill the questionnaire one more time on their phones (after filling the physical one on paper).
The dropout rates were not associated with any baseline factor, including rural residence, level of education, stage of cancer at baseline, or site of cancer.Thirty-nine patients had filled <50% of the p-ESAS forms, which were excluded from the analysis.Fifteen responses were incomplete, but contained more responses to more than 50% questions; therefore, these data were averaged and included in the analysis.A total of 595 patients were included in the final analysis (Fig 1).
The median age at diagnosis was 53 years (IQR, 44-63 years), and 307 (51.6%) patients were male.Twenty-five percent of patients were illiterate (n 5 147), 23% resided in a rural area, and the median distance traveled by them to reach the hospital was 80 km (IQR, 25-300 km).The most common primary cancer was breast cancer (29.7%), followed in order of decreasing frequency by lung (29.2%), urothelial (13.3%), and GI (6.4%) cancers.More than 50% of patients presented with metastatic disease at baseline (Table 1).

ESAS Responses and Correlation Between e-ESAS and p-ESAS
Individual category and domain-wise mean symptom scores are shown in Figures 2A-2C.Severity-and site-wise symptom scores are shown in Table 2. Moderate to severe physical symptoms were reported maximally by patients with lung cancer (49%, P < .001),followed by female genital tract (48.4%) and urothelial malignancies (43%) in decreasing frequency.Physical symptom intensity reported was similar for GI (37% moderate to severe), hepatobiliary (33.3%), and other malignancies.The most commonly reported physical symptom was tiredness, with 88% of patients experiencing the same.It was significantly more severe in those with metastatic disease.Moderate to severe psychological symptoms were numerically reported more in female genital tract cancers.However, there was no statistically significant difference in any other sites (P 5 .51).The most common psychological symptom reported was anxiety, seen in around 70% of all the patients, more commonly seen in those with stage IV disease.
The total symptom scores were reported as moderate to severe in more than 50% patients with female genital tract cancers, followed by lung cancer.This was significantly higher than other sites (P 5 .025),which reported similar numbers of patients with moderate to severe symptoms.Site-wise individual symptom scores are shown in Appendix Table A1.For all patients, irrespective of the primary site of cancer, moderate to severe physical, psychological, and total symptom scores were seen in 39.8%, 40%, and 39% of the patients.The most common comments on other problems reported were constipation (n 5 17 of 76, 22.4%) and gastritis (n 5 15 of 76, 19.7%).
ICCs between the physically and online filled ESAS questionnaire (p-ESAS v e-ESAS) varied between 0.75 and 0.9 (Table 3).Bland-Altman plots to determine the difference between the two categories' means for physical, psychological, and total symptom scores are shown in Figure 3. Graphically, most of the points fell between 61.96 SD.

Factors Associated With Symptom Burden
On univariable analysis of the patients, rural residence; lung, urothelial, and hepatobiliary primary sites; education less than graduation level; and metastatic disease at baseline were associated with higher odds of moderate to severe total symptom scores.4).

DISCUSSION
Our study aimed to evaluate the feasibility and utility of using a mobile-based secure system for monitoring ePROs.Despite their vast clinical utility, 20,21,32 the assessment of PROs is not prevalent, given the logistic strain it places on health care delivery systems.In this regard, electronic reporting is a pragmatic solution with additional benefits as well.Randomized controlled trials have proven the superiority of ePROs in the developed nations 19,21 ; however, to our knowledge, there are no supporting real-world data on PROs collected electronically from LMICs like India, where high cancer burden and resource-limited setting make ePROs a necessity.ESAS questionnaire, developed over 30 years back, has been widely used and validated in oncologic and palliative settings. 8,12,33We found a good reliability (ICC varying between 0.74 and 0.87, P < .001) between e-ESAS and p-ESAS across individual symptoms and domain scores obtained in our population.
We also found that the electronically filled ESAS scores were consistently higher than the physical ones.This might be because the e-ESAS scores were filled after the physical questionnaire, and by that time, the participants would have received medications to palliate symptoms, which might have improved the symptoms compared with the first contact.
Electronic systems allow easy compilation and storage of patient symptom scores, easier standardization, and longitudinal follow-up. 16Therefore, electronic PROs (e-ESAS) may be incorporated in routine clinical practice, with the additional benefit of avoiding a dedicated clinic visit for PRO measurements, which is especially important for high-burden centers.
Implementation of eESAS, however, might be encountered with some difficulties.Nearly half of the participants in our study did not complete the online form, despite 95% of them possessing smartphones.Regression analysis did not reveal any association with the level of education, urban or rural residence, or metastatic/localized disease.The speculated reasons include an inability to comprehend the questions or the lack of patient motivation, which may be due to a higher symptom burden, a sense of hopelessness, or poor social support and not possessing a smartphone in a minority of patients.These shortcomings, however, may be overcome with the help of health care providers, 34 who can periodically encourage patients and train them to ensure high completion rates.This might also improve with a better understanding of the e-ESAS tool with future research.
Among the participants who filled the e-ESAS scale, moderate to severe physical and overall symptoms were  reported maximally by patients with lung and female genital tract cancers.Similar results have been seen in multiple studies validating paper-based ESAS, 13,27,35 as are seen with our e-ESAS.We also found that metastatic disease at presentation predicted a higher symptom burden, as shown in previous studies as well. 27,36Higher education was independently predictive of lower symptom burden.This is in contrast to previous data; 27 however, our results may be biased by the fact that more educated patients filled the e-ESAS compared with the illiterate ones.
The strengths of our study include its prospective nature, the representativeness of the enrolled cohort, and the documentation of real-time data collection.To the best of our knowledge, it is the first study that has tried to validate the use of ePROs in the LMIC setting.It was, however, limited by the number of smartphone-based responses-only 634 patients of 1,160 responded to the e-ESAS link sent via text message.This might question the generalizability of the study.In addition, there was no formal assessment of the patients' cognitive abilities before recording their responses, only a subjective idea of which participant would be able to comprehend the questions.In addition, it was a onetime assessment of the PROs, predominantly for validation of the electronic measurements, and there was no longitudinal follow-up.Another limitation of the study is the presence of baseline metastasis in more than half of the total number of patients, which is slightly higher than the Indian epidemiologic data. 37This can be explained by the presence of a referral bias in the outpatient department of a tertiary care center as ours.The ESAS scale itself also has some inherent limitations, 33 such as the difficulty in differentiating between drowsiness and tiredness; reversal of scale for lack of appetite where zero might be coded as the worst and 10 as the best, as opposed to the actual scale; and general understanding of categories like best well-being.
A majority of patients with solid organ cancers present with up-front metastatic disease, 37 where the intent of treatment is palliative, and primary goals of care include symptom control and improvement in the quality of life.Ironically, for these patients, periodic visits to the out-patient department for PRO monitoring are more difficult since they have higher symptom scores to begin with.Apart from this, serial symptom measurements (in palliative and curative settings) have shown to improve quality of life and OS.PRO monitoring may not be feasible because of financial, infrastructural, and manpower limitations in our settings.e-ESAS is one such pragmatic tool to help us overcome these barriers.With encouraging results from our study, we may extend the use of e-ESAS in routine clinical care to help us tailor therapy for each patient and improve outcomes beyond the traditional biochemical or radiologic indices.However, evaluation of the cause of high dropout rates for e-ESAS via interviews and thematic analysis represents the next step in expanding the reach of this pragmatic assessment method.

FIG 1 .
FIG 1. Workflow of the study.e-ESAS, electronically filled Edmonton Symptom Assessment System.
FIG 2. (A) Box plot showing physical, psychological, and overall symptom scores in e-ESAS and p-ESAS (mean 6 SD labeled at the top of each box plot).(B) Bar graph showing means of individual psychological and best well-being domains in p-ESAS and e-ESAS (mean 6 SD labeled at the top of each bar).(C) Bar graph showing means of individual physical domains in p-ESAS and e-ESAS (mean 6 SD labeled at the top of each bar) e-ESAS, electronically filled Edmonton Symptom Assessment System; p-ESAS, paper-based Edmonton Symptom Assessment System; SD, standard deviation.

TABLE 2 .
Smartphone-Based ESAS Severity Scores and for Different Primary Sites Abbreviation: ESAS, Edmonton Symptom Assessment System.JCO Global Oncology ascopubs.org/journal/go| 5 Electronic Patient-Reported Outcomes in Patients With Cancer

TABLE 4 .
Multivariable Analysis for Association of Physical, Psychological, and Overall Symptom Scores With Baseline Characteristics