Classifying Self-Reported Rheumatoid Arthritis Flares Using Daily Patient-Generated Data From a Smartphone App: Exploratory Analysis Applying Machine Learning Approaches.

BACKGROUND: The ability to predict rheumatoid arthritis (RA) flares between clinic visits based on real-time, longitudinal patient-generated data could potentially allow for timely interventions to avoid disease worsening. OBJECTIVE: This exploratory study aims to investigate the feasibility of using machine learning methods to classify self-reported RA flares based on a small data set of daily symptom data collected on a smartphone app. METHODS: Daily symptoms and weekly flares reported on the Remote Monitoring of Rheumatoid Arthritis (REMORA) smartphone app from 20 patients with RA over 3 months were used. Predictors were several summary features of the daily symptom scores (eg, pain and fatigue) collected in the week leading up to the flare question. We fitted 3 binary classifiers: logistic regression with and without elastic net regularization, a random forest, and naive Bayes. Performance was evaluated according to the area under the curve (AUC) of the receiver operating characteristic curve. For the best-performing model, we considered sensitivity and specificity for different thresholds in order to illustrate different ways in which the predictive model could behave in a clinical setting. RESULTS: The data comprised an average of 60.6 daily reports and 10.5 weekly reports per participant. Participants reported a median of 2 (IQR 0.75-4.25) flares each over a median follow-up time of 81 (IQR 79-82) days. AUCs were broadly similar between models, but logistic regression with elastic net regularization had the highest AUC of 0.82. At a cutoff requiring specificity to be 0.80, the corresponding sensitivity to detect flares was 0.60 for this model. The positive predictive value (PPV) in this population was 53%, and the negative predictive value (NPV) was 85%. Given the prevalence of flares, the best PPV achieved meant only around 2 of every 3 positive predictions were correct (PPV 0.65). By prioritizing a higher NPV, the model correctly predicted over 9 in every 10 non-flare weeks, but the accuracy of predicted flares fell to only 1 in 2 being correct (NPV and PPV of 0.92 and 0.51, respectively). CONCLUSIONS: Predicting self-reported flares based on daily symptom scorings in the preceding week using machine learning methods was feasible. The observed predictive accuracy might improve as we obtain more data, and these exploratory results need to be validated in an external cohort. In the future, analysis of frequently collected patient-generated data may allow us to predict flares before they unfold, opening opportunities for just-in-time adaptative interventions. Depending on the nature and implication of an intervention, different cutoff values for an intervention decision need to be considered, as well as the level of predictive certainty required.

Feasibility and acceptability to use a smartphone-based manikin for daily longitudinal self-reporting of chronic pain.

Background: As management of chronic pain continues to be suboptimal, there is a need for tools that support frequent, longitudinal pain self-reporting to improve our understanding of pain. This study aimed to assess the feasibility and acceptability of daily pain self-reporting using a smartphone-based pain manikin. Methods: For this prospective feasibility study, we recruited adults with lived experience of painful musculoskeletal condition. They were asked to complete daily pain self-reports via an app for 30 days. We assessed feasibility by calculating pain report completion levels, and investigated differences in completion levels between subgroups. We assessed acceptability via an end-of-study questionnaire, which we analysed descriptively. Results: Of the 104 participants, the majority were female (n = 87; 84%), aged 45-64 (n = 59; 57%), and of white ethnic background (n = 89; 86%). The mean completion levels was 21 (± 7.7) pain self-reports. People who were not working (odds ratio (OR) = 1.84; 95% confidence interval (CI), 1.52-2.23) were more likely, and people living in less deprived areas (OR = 0.77; 95% CI, 0.62-0.97) and of non-white ethnicity (OR = 0.45; 95% CI, 0.36-0.57) were less likely to complete pain self-reports than their employed, more deprived and white counterparts, respectively. Of the 96 participants completing the end-of-study questionnaire, almost all participants agreed that it was easy to complete a pain drawing (n = 89; 93%). Conclusion: It is feasible and acceptable to self-report pain using a smartphone-based manikin over a month. For its wider adoption for pain self-reporting, the feasibility and acceptability should be further explored among people with diverse socio-economic and ethnic backgrounds.

Examining the variability of multiple daily symptoms over time among individuals with multiple long-term conditions (MLTC-M/multimorbidity): An exploratory analysis of a longitudinal smartwatch feasibility study.

Introduction: People living with multiple long-term conditions (MLTC-M) (multimorbidity) experience a range of inter-related symptoms. These symptoms can be tracked longitudinally using consumer technology, such as smartphones and wearable devices, and then summarised to provide useful clinical insight. Aim: We aimed to perform an exploratory analysis to summarise the extent and trajectory of multiple symptom ratings tracked via a smartwatch, and to investigate the relationship between these symptom ratings and demographic factors in people living with MLTC-M in a feasibility study. Methods: 'Watch Your Steps' was a prospective observational feasibility study, administering multiple questions per day over a 90 day period. Adults with more than one clinician-diagnosed long-term condition rated seven core symptoms each day, plus up to eight additional symptoms personalised to their LTCs per day. Symptom ratings were summarised over the study period at the individual and group level. Symptom ratings were also plotted to describe day-to-day symptom trajectories for individuals. Results: Fifty two participants submitted symptom ratings. Half were male and the majority had LTCs affecting three or more disease areas (N = 33, 64%). The symptom rated as most problematic was fatigue. Patients with increased comorbidity or female sex seemed to be associated with worse experiences of fatigue. Fatigue ratings were strongly correlated with pain and level of dysfunction. Conclusion: In this study we have shown that it is possible to collect and descriptively analyse self reported symptom data in people living with MLTC-M, collected multiple times per day on a smartwatch, to gain insights that might support future clinical care and research.

Does body mass index mediate the relationship between socioeconomic position and incident osteoarthritis?

OBJECTIVES: To investigate associations of socioeconomic position (SEP) and obesity with incident osteoarthritis (OA), and to examine whether body mass index (BMI) mediates the association between SEP and incident OA. METHODS: Data came from the English Longitudinal Study of Ageing, a population-based cohort study of adults aged ≥50 years. The sample population included 9,281 people. Cox regression analyses were performed to investigate the associations between SEP (measured by education, occupation, income, wealth and deprivation) and obesity (BMI ≥30 kg/m2) at baseline and self-reported incident OA. The mediating effect of BMI on the relationship between SEP and incident OA were estimated using Structural Equation Models. RESULTS: After a mean follow-up time of 7.8 years, 2369 participants developed OA. Number of person-years included in the analysis was 65,456. Lower SEP was associated with higher rates of OA (for example, hazard ratio (HR) lowest vs highest education category 1.52 (95% confidence interval (CI) 1.30, 1.79)). Obesity compared with non-obesity was associated with increased rates of incident OA (HR 1.37 (95% CI 1.23, 1.52)). BMI mediated the relationship between a lower SEP and OA (ß = 0.005, p < 0.001) and the direct effect was not significant (ß = 0.004, p = 0.212). CONCLUSIONS: Strategies to reduce social inequalities and obesity prevalence may help to reduce OA risk.

Using patient-reported data from a smartphone app to capture and characterize real-time patient-reported flares in rheumatoid arthritis.

Objective: We aimed to explore the frequency of self-reported flares and their association with preceding symptoms collected through a smartphone app by people with RA. Methods: We used data from the Remote Monitoring of RA study, in which patients tracked their daily symptoms and weekly flares on an app. We summarized the number of self-reported flare weeks. For each week preceding a flare question, we calculated three summary features for daily symptoms: mean, variability and slope. Mixed effects logistic regression models quantified associations between flare weeks and symptom summary features. Pain was used as an example symptom for multivariate modelling. Results: Twenty patients tracked their symptoms for a median of 81 days (interquartile range 80, 82). Fifteen of 20 participants reported at least one flare week, adding up to 54 flare weeks out of 198 participant weeks in total. Univariate mixed effects models showed that higher mean and steeper upward slopes in symptom scores in the week preceding the flare increased the likelihood of flare occurrence, but the association with variability was less strong. Multivariate modelling showed that for pain, mean scores and variability were associated with higher odds of flare, with odds ratios 1.83 (95% CI, 1.15, 2.97) and 3.12 (95% CI, 1.07, 9.13), respectively. Conclusion: Our study suggests that patient-reported flares are common and are associated with higher daily RA symptom scores in the preceding week. Enabling patients to collect daily symptom data on their smartphones might, ultimately, facilitate prediction and more timely management of imminent flares.

Engagement with consumer smartwatches for tracking symptoms of individuals living with multiple long-term conditions (multimorbidity): A longitudinal observational study.

INTRODUCTION: People living with multiple long-term conditions (multimorbidity) (MLTC-M) experience an accumulating combination of different symptoms. It has been suggested that these symptoms can be tracked longitudinally using consumer technology, such as smartphones and wearable devices. AIM: The aim of this study was to investigate longitudinal user engagement with a smartwatch application, collecting survey questions and active tasks over 90 days, in people living with MLTC-M. METHODS: 'Watch Your Steps' was a prospective observational study, administering multiple questions and active tasks over 90 days. Adults with more than one clinician-diagnosed long-term conditions were loaned Fossil® Sport smartwatches, pre-loaded with the study app. Around 20 questions were prompted per day.Daily completion rates were calculated to describe engagement patterns over time, and to explore how these varied by patient characteristics and question type. RESULTS: Fifty three people with MLTC-M took part in the study. Around half were male ( = 26; 49%) and the majority had a white ethnic background (n = 45; 85%). About a third of participants engaged with the smartwatch app nearly every day. The overall completion rate of symptom questions was 45% inter-quartile range (IQR 23-67%) across all study participants. Older patients and those with greater MLTC-M were more engaged, although engagement was not significantly different between genders. CONCLUSION: It was feasible for people living with MLTC-M to report multiple symptoms per day over 3 months. User engagement appeared as good as other mobile health studies that recruited people with single health conditions, despite the higher daily data entry burden.

Recruitment and Ongoing Engagement in a UK Smartphone Study Examining the Association Between Weather and Pain: Cohort Study.

BACKGROUND: The huge increase in smartphone use heralds an enormous opportunity for epidemiology research, but there is limited evidence regarding long-term engagement and attrition in mobile health (mHealth) studies. OBJECTIVE: The objective of this study was to examine how representative the Cloudy with a Chance of Pain study population is of wider chronic-pain populations and to explore patterns of engagement among participants during the first 6 months of the study. METHODS: Participants in the United Kingdom who had chronic pain (≥3 months) and enrolled between January 20, 2016 and January 29, 2016 were eligible if they were aged ≥17 years and used the study app to report any of 10 pain-related symptoms during the study period. Participant characteristics were compared with data from the Health Survey for England (HSE) 2011. Distinct clusters of engagement over time were determined using first-order hidden Markov models, and participant characteristics were compared between the clusters. RESULTS: Compared with the data from the HSE, our sample comprised a higher proportion of women (80.51%, 5129/6370 vs 55.61%, 4782/8599) and fewer persons at the extremes of age (16-34 and 75+). Four clusters of engagement were identified: high (13.60%, 865/6370), moderate (21.76%, 1384/6370), low (39.35%, 2503/6370), and tourists (25.44%, 1618/6370), between which median days of data entry ranged from 1 (interquartile range; IQR: 1-1; tourist) to 149 (124-163; high). Those in the high-engagement cluster were typically older, whereas those in the tourist cluster were mostly male. Few other differences distinguished the clusters. CONCLUSIONS: Cloudy with a Chance of Pain demonstrates a rapid and successful recruitment of a large, representative, and engaged sample of people with chronic pain and provides strong evidence to suggest that smartphones could provide a viable alternative to traditional data collection methods.