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.

How to self-examine for tender and swollen joints: co-producing a training video for people with rheumatoid arthritis.

Objective: This paper describes the co-production of a training video to support people with RA to self-examine for tender and swollen joints. Methods: The patient and public involvement and engagement (PPIE) group supporting a remote monitoring study elected to develop a video to train people with RA how to self-examine for tender and swollen joints, because nothing appropriate was publicly available to fulfil their needs. A core team of PPIE group members and clinicians developed the video, with input from conception to dissemination from the PPIE group. The video was posted, open access, on a YouTube website in February 2021, alongside supporting materials. The number of monthly hits was tracked and a survey developed to ascertain feedback. Results: The video received 1000 hits in the first week, and >40 000 at 10 months. The top three countries viewing the video were India, the USA and the UK, with a range of ages and gender profile broadly corresponding to those of RA patients. Forty-eight survey responses were received (26 patients and 22 clinicians). Patients reported an improvement in their ability to self-examine after watching this video. Eighty-six per cent of patients and 71% of clinicians would recommend the video. It has been used and disseminated by a number of national organizations within the UK. Conclusion: This co-produced, open-access training video for people with RA, originally intended to support a research study into remote monitoring, has been well received, reflecting an international interest in self-examination.

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.

Using patient-generated health data in clinical practice: How timing influences its function in rheumatology outpatient consultations.

OBJECTIVE: Utilizing patient-generated health data (PGHD) in clinical consultations and informing clinical and shared decision-making processes has the potential to improve clinical practice but has proven challenging to implement. Looking at consultations between people with rheumatoid arthritis (RA) and rheumatologists, this study examines when and how daily PGHD was discussed in outpatient consultations. METHODS: We conducted a secondary qualitative analysis of 17 audio-recorded research outpatient consultations using thematic and interactional approaches. RESULTS: Clinicians decided when to look at the PGHD and what symptoms to prioritise during the consultation. When PGHD was introduced early in consultations, it was usually used to invite patients to collaborate (elicit new information). When introduced later, PGHD was used to corroborate patient accounts and to convince the patient about proposed actions and treatments. Clinicians occasionally disregarded PGHD if it did not fit into their clinical assessment. CONCLUSION: The time that PGHD is introduced may influence how PGHD is used in consultations. Further research is needed to understand how best to empower patients to discuss PGHD. PRACTICE IMPLICATIONS: Educating patients and clinicians about the importance of timing and strategies when using PGHD in consultations may help promote shared decision-making.

Do associations between education and obesity vary depending on the measure of obesity used? A systematic literature review and meta-analysis.

BACKGROUND: Consistent evidence suggests a relationship between lower educational attainment and total obesity defined using body mass index (BMI); however, a comparison of the relationships between educational attainment and total obesity (BMI ≥30 kg/m2) and central obesity (waist circumference (WC) > 102 cm for men and WC > 88 cm for women) has yet to be carried out. This systematic literature review (SLR) and meta-analyses aimed to understand whether i) the associations between education and obesity are different depending on the measures of obesity used (BMI and WC), and ii) to explore whether these relationships differ by gender and region. METHODS: Medline, Embase and Web of Science were searched to identify studies investigating the associations between education and total and central obesity among adults in the general population of countries in the Organisation for Economic Co-operation and Development (OECD). Meta-analyses and meta-regression were performed in a subset of comparable studies (n=36 studies; 724,992 participants). RESULTS: 86 eligible studies (78 cross-sectional and eight longitudinal) were identified. Among women, most studies reported an association between a lower education and total and central obesity. Among men, there was a weaker association between lower education and central than total obesity (OR central vs total obesity in men 0.79 (95% CI 0.60, 1.03)). The association between lower education and obesity was stronger in women compared with men (OR women vs men 1.66 (95% CI 1.32, 2.08)). The relationship between lower education and obesity was less strong in women from Northern than Southern Europe (OR Northern vs Southern Europe in women 0.37 (95% CI 0.27, 0.51)), but not among men. CONCLUSIONS: Associations between education and obesity differ depending on whether total or central obesity is used among men, but not in women. These associations are stronger among women than men, particularly in Southern European countries.

Remote symptom monitoring integrated into electronic health records: A systematic review.

OBJECTIVE: People with long-term conditions require serial clinical assessments. Digital patient-reported symptoms collected between visits can inform these, especially if integrated into electronic health records (EHRs) and clinical workflows. This systematic review identified and summarized EHR-integrated systems to remotely collect patient-reported symptoms and examined their anticipated and realized benefits in long-term conditions. MATERIALS AND METHODS: We searched Medline, Web of Science, and Embase. Inclusion criteria were symptom reporting systems in adults with long-term conditions; data integrated into the EHR; data collection outside of clinic; data used in clinical care. We synthesized data thematically. Benefits were assessed against a list of outcome indicators. We critically appraised studies using the Mixed Methods Appraisal Tool. RESULTS: We included 12 studies representing 10 systems. Seven were in oncology. Systems were technically and functionally heterogeneous, with the majority being fully integrated (data viewable in the EHR). Half of the systems enabled regular symptom tracking between visits. We identified 3 symptom report-guided clinical workflows: Consultation-only (data used during consultation, n = 5), alert-based (real-time alerts for providers, n = 4) and patient-initiated visits (n = 1). Few author-described anticipated benefits, primarily to improve communication and resultant health outcomes, were realized based on the study results, and were only supported by evidence from early-stage qualitative studies. Studies were primarily feasibility and pilot studies of acceptable quality. DISCUSSION AND CONCLUSIONS: EHR-integrated remote symptom monitoring is possible, but there are few published efforts to inform development of these systems. Currently there is limited evidence that this improves care and outcomes, warranting future robust, quantitative studies of efficacy and effectiveness.

Three Quality Improvement Initiatives and Performance of Rheumatoid Arthritis Disease Activity Measures in Electronic Health Records: Results From an Interrupted Time Series Study.

OBJECTIVE: Applying treat-to-target strategies in the care of patients with rheumatoid arthritis (RA) is critical for improving outcomes, yet electronic health records (EHRs) have few features to facilitate this goal. We undertook this study to evaluate the effect of 3 health information technology (health-IT) initiatives on the performance of RA disease activity measures and outcomes in an academic rheumatology clinic. METHODS: We implemented the 3 following initiatives designed to facilitate performance of the Clinical Disease Activity Index (CDAI): an EHR flowsheet to input scores, peer performance reports, and an EHR SmartForm including a CDAI calculator. We performed an interrupted time-series trial to assess effects on the proportion of RA visits with a documented CDAI. Mean CDAI scores before and after the last initiative were compared using t-tests. Additionally, we measured physician satisfaction with the initiatives. RESULTS: We included data from 995 patients with 8,040 encounters between 2012 and 2017. Over this period, electronic capture of CDAI scores increased from 0% to 64%. Performance remained stable after peer reporting and the SmartForm were introduced. We observed no meaningful changes in disease activity levels. However, physician satisfaction increased after SmartForm implementation. CONCLUSION: Modifications to the EHR, provider culture, and clinical workflows effectively improved capture of RA disease activity scores and physician satisfaction, but parallel gains in disease activity levels were missing. This study illustrates how a series of health-IT initiatives can evolve to enable sustained changes in practice. However, capture of RA outcomes alone may not be sufficient to improve levels of disease activity without a comprehensive treat-to-target program.

Using Health Information Technology to Support Use of Patient-Reported Outcomes in Rheumatology.

Technology can help health care providers understand their patients' experience of illness in a way that was previously impossible. Experience in using health information technology (IT) to capture this information through PROs within rheumatology suggests that careful attention to human centered design, including detailed workflow planning, consideration of patient and physician burden, integration into the health IT ecosystem, and delivering information to the right person at the right time are all important. Technology applications must be tested in diverse health systems and populations to ensure they are simple to interpret, useful for clinical decision making and effective in impacting outcomes.

Effects of Language, Insurance, and Race/Ethnicity on Measurement Properties of the PROMIS Physical Function Short Form 10a in Rheumatoid Arthritis.

OBJECTIVE: Most studies that have evaluated patient-reported outcomes, such as those utilizing the Patient-Reported Outcomes Measurement Information System (PROMIS) Physical Function Short Form 10a (PF10a) in rheumatoid arthritis (RA), have been performed in white and English-speaking populations. The aim of our study was to assess the measurement properties of the PF10a in a racially/ethnically diverse population with RA and to determine the effect of non-English language proficiency, insurance status, and race/ethnicity on the validity and responsiveness of the PF10a. METHODS: Data were abstracted from electronic health records for all RA patients seen in a university-based rheumatology clinic between 2013 and 2017. We evaluated the use of the PF10a, floor and ceiling effects, and construct validity across categories of language preference, insurance, and race/ethnicity. We used standardized response means and linear mixed-effects models to evaluate the responsiveness of the PF10a to longitudinal changes in the Clinical Disease Activity Index (CDAI) across population subgroups. RESULTS: We included 595 patients in a cross-sectional analysis of validity and 341 patients in longitudinal responsiveness analyses of the PF10a. The PF10a had acceptable floor and ceiling effects and was successfully implemented. We observed good construct validity and responsiveness to changes in CDAI among white subjects, English speakers, and privately insured patients. However, constructs evaluated by the PF10a were less correlated with clinical measures among Chinese speakers and Hispanic subjects, and less sensitive to clinical improvements among Medicaid patients and Spanish speakers. CONCLUSION: While the PF10a has good measurement properties and is both practical and acceptable for implementation in routine clinical practice, we also found important differences across racial/ethnic groups and those with limited English proficiency that warrant further investigation.

"Am I OK?" using human centered design to empower rheumatoid arthritis patients through patient reported outcomes.

OBJECTIVE: Use of patient reported outcomes (PROs) in the routine care of rheumatoid arthritis (RA) has been shown to improve health outcomes, However, integration of PROs into the clinical visit is inconsistent. We aimed to develop a "dashboard" for RA patients to display relevant PRO measures for discussion during a routine RA clinical visit. METHODS: Patients (N = 45) and providers (N = 12) were recruited from rheumatology clinics at a university center and a safety net hospital. Using a human-centered design process involving patients, clinicians, designers, and health-IT experts, we performed interviews, clinic observations, and focus groups, which subsequently guided an iterative phase of prototype testing. RESULTS: RA patients and their providers shared the goals of assessing wellbeing and developing a personalized treatment plan. We found conflicting views of which data were most important for guiding decision-making and for answering the patient's overarching question of "Am I OK?" CONCLUSION: The final dashboard simplified the display of PRO data and correlated it longitudinally to the patient's medication regimen. It also included laboratory values relevant for RA care. PRACTICE IMPLICATIONS: By presenting data graphically, the dashboard may provide a platform for patients and providers to communicate around PROs and shared goals.

Patient-reported outcome measures for use in clinical trials of SLE: a review.

Inclusion of patient-reported outcomes is important in SLE clinical trials as they allow capture of the benefits of a proposed intervention in areas deemed pertinent by patients. We aimed to compare the measurement properties of health-related quality of life (HRQoL) measures used in adults with SLE and to evaluate their responsiveness to interventions in randomised controlled trials (RCTs). A systematic review was undertaken using full original papers in English identified from three databases: MEDLINE, EMBASE and PubMed. Studies describing the validation of HRQoL measures in English-speaking adult patients with SLE and SLE drug RCTs that used an HRQoL measure were retrieved. Twenty-five validation papers and 26 RCTs were included in the indepth review evaluating the measurement properties of 4 generic (Medical Outcomes Study Short-Form 36 (SF36), Patient Reported Outcomes Measurement Information System (PROMIS) item-bank, EuroQol-5D, and Functional Assessment of Chronic Illness Therapy-Fatigue) and 3 disease-specific (Lupus Quality of Life (LupusQoL), Lupus Patient Reported Outcomes, Lupus Impact Tracker (LIT)) instruments. All measures had good convergent and discriminant validity. PROMIS provided the strongest evidence for known-group validity and reliability among generic instruments; however, data on its responsiveness have not been published. Across measures, standardised response means were generally indicative of poor-moderate sensitivity to longitudinal change. In RCTs, clinically important improvements were reported in SF36 scores from baseline; however, between-arm differences were frequently non-significant and non-important. SF36, PROMIS, LupusQoL and LIT had the strongest evidence for acceptable measurement properties, but few measures aside from the SF36 have been incorporated into clinical trials. This review highlights the importance of incorporating a broader range of SLE-specific HRQoL measures in RCTs and warrants further research that focuses on longitudinal responsiveness of newer instruments.