The impact of data from remote measurement technology on the clinical practice of healthcare professionals in depression, epilepsy and multiple sclerosis: survey

Data were analysed using SPSS v.26 and Stata SE 16.1. We used descriptive statistics to understand the rates of responses from different demographic groups. Chi-squared tests were used to compare the rate of reported app usage (question 8) between different job roles and age groups, as well as to compare reported use of apps and devices by respondents’ patients. Responses to items in question 10, on the impact of patient RMT data on five aspects of clinical work, were dichotomised as 1 for responses of ‘definitely’ or ‘sometimes’, and 0 for responses of ‘unsure’, ‘hardly’, and ‘never’. ‘Unsure’ was grouped with the negative polarity items since respondents may have responded in this way if they did not have experience of using these technologies. Dichotomising the data rather than treating it as a 5-point scale had the benefit of resolving this ambiguity. For question 10 on the impact of RMT data on respondents’ clinical practice, univariate, then multivariate logistic regression was used to explore the effect of demographic variables and app usage on views of the impact of RMT data. A p-value of 0.05 was considered significant. For each categorical variable, the reference category chosen was the category with the largest number of responses. Missing responses were accounted for using multiple imputation, also conducted in STATA. 83 imputations were calculated for each missing entry, as the highest fraction of missing information (FMI) found in the analysis was 0.83 [27]. A sensitivity analysis was used to examine the influence of missingness on parameter estimates by comparing the consistency of parameters estimates between models with imputed and observed datasets.

Survey responses were received from 1009 HCPs. The greatest proportion of these were completed as a result of efforts to recruit via the UK National Institute for Health Research (NIHR) Clinical Research Network: 21 National Health Service (NHS) healthcare trusts and Clinical Commissioning Groups (CCGs) across the UK advertised the survey to healthcare staff working with patients with epilepsy, MS or depression in their area. These trusts and CCGs reported that 650 staff had notified them that they had completed the survey, however it is likely a greater proportion of the total survey responses were obtained via this recruitment method, as it was not compulsory for respondents to notify their trust of their involvement. We did not include a survey item to ask where respondents had heard about the survey to preserve anonymity, so it is not possible to further attribute respondents to different recruitment methods.

Despite the survey being available in 5 European languages in addition to English, and efforts made to disseminate the survey to clinicians working in other countries (via social media, LinkedIn groups, contacts of the consortium, and via the Innovative Medicines Initiative steering group), the number of respondents reporting to be working in countries outside the UK was small (n = 30, 3% of total). These were analysed together with the UK responses as there were too few from each individual country to viably compare them statistically against the UK responses.

Of the 1009 completed surveys, 1006 were retained for analysis. Reasons for exclusion were insincere completion of the survey (n = 1) and improper completion of the consent form (n = 2). Figure 1 shows a STROBE flow diagram.

Table 1 shows a breakdown of the included survey responses by demographics. Age group categories followed a normal distribution, with the largest number of respondents in the middle category (ages 41–50, n = 293, 29% of respondents) and progressively fewer in older and younger age groups. The most frequently selected item under specialism was ‘mental health’ (n = 587, 38%), with the fewest selecting ‘social care’ (n = 32, 2%). Although relatively few respondents reported a specialism of ‘multiple sclerosis’ (n = 55, 4%) or ‘epilepsy’ (n = 73, 5%), a larger number reported working in neurology (n = 112, 7%), which is a specialism covering both these conditions. Among job roles, ‘nursing’ was the category with the largest proportion of responses. Doctors would have been the second most numerous category, however doctors were split into two categories of ‘GPs’ (n = 118, 12%) and ‘Doctor (excluding GP)’, (n = 138, 14%). ‘Clinical psychology professionals’ were therefore the second most numerous category, with 157 responses (16%). There was good representation from all clinical settings, with fewest (5%) from ‘specialist tertiary care’ settings, which might be expected due to their specialised nature. Most responses (97%) were from HCPs working in the UK. The second most reported country was Portugal, with 21 survey completions (2%).

559 respondents (56%) indicated that they use digital health smartphone apps for at least one purpose in their clinical role. Figure 2 shows the percentages of respondents who reported using smartphone apps of each of the five types listed in this question. Guidelines apps were the most used (25%), while calculation apps were least used (11%). There was a spread of responses across all app types, with no app type receiving fewer than 10% of total responses.

Figure 3 shows the percentage of respondents in each job category who reported using smartphone apps in their daily clinical practice. A chi-squared test showed that there were significant differences between job roles for this question (χ² = 106.348, p < 0.01). Pharmacists and medical students were most likely to report using any apps, although there were fewer survey completions overall from these job roles. Further analysis revealed that of the different types of app listed, pharmacists were most likely to report using prescribing/dosing apps (87% of pharmacist respondents), and medical students were most likely to use guidelines apps (80% of student respondents). 40% of GP respondents and 38% of doctors excluding GPs reported using guidelines apps. Clinical psychology professionals and allied health professionals were least likely to report using smartphone apps of any kind.

Question 8a invited respondents to provide examples of other types of app that they used in their daily clinical practice. Table 2 provides a random subsample of the responses, with information on the publisher, category, cost and target condition of these, where relevant. Responses covered apps published by the NHS, charities, commercial organisations, and public bodies. Apps mentioned included a large number targeting mental health, with fewer focussed on neurological conditions, which could be expected given the larger proportion of responses from mental health specialists in the sample. Many of these apps are either patient-facing or have a focus on wellbeing and it is possible that respondents understood this question to mean apps they recommend to their patients, rather than apps they use themselves to facilitate their job role.

Across all age categories, between 50 and 60 percent of respondents reported using apps. The highest percentage was reported by those in the over-60 category. However, a chi-squared test on these age groupings was non-significant (χ2 = 1.848, p = 0.76).

78% of respondents indicated that their patients used RMT (one or more of smartphone apps, wearables, or other type of device) to improve or increase awareness of their health. Figure 4 shows the relevant percentages for different purposes. Activity monitoring was selected by the largest proportion of respondents (62%). Sleep monitoring was also selected by more than half of respondents (53%). Of the given categories, monitoring for a specific condition was least selected (31%).

Figure 5 shows the split of these responses across different device types. Further to the above, this shows that wearable sensing devices are most commonly reported to be used for passive monitoring of activity and sleep, while smartphone apps were used more often than wearable devices for monitoring mood, setting personal health goals and weight management.

A sub-group analysis by job role showed that there were significant differences between job roles for this question (χ² = 154.952, p < 0.01). This is shown in Fig. 6. A greater proportion of GPs than other types of health professional had experienced their patients using devices to monitor activity (88% of GP respondents), monitor sleep (66%) and manage weight (66%).

Sub-questions 10.1–10.5 asked respondents about the impact of the data from patients’ RMT devices on different aspects of respondents’ clinical work. Figure 7 shows these results. Across all sub-questions, the mean percentage of respondents indicating that the data from RMT ‘definitely’ or ‘sometimes’ had an impact on their work was 62%. The mean percentage of those indicating that it ‘hardly’ or ‘never’ had an impact was 27%. A greater proportion of respondents selected ‘definitely’ or ‘sometimes’ for sub-question 10.4 (Does RMT data impact patients’ awareness of their own health?) than for any of the other sub-questions.

Results from univariate and multivariate logistic regression analyses with multiple imputation and using questions 3–8 as independent variables and sub-questions 10.1–10.5 as dependent variables are provided in Tables 3, 4, 5, 6 and 7. An analysis for multicollinearity revealed no variance inflation factor (VIF) greater than 3.5, suggesting that multicollinearity did not affect the results. Results from multiple imputation and observed datasets were consistent in term of the parameter estimate and its precision.

Across all five sub-questions in question 10, the independent variable on HCP use of smartphone apps (question 8), gave a significant result in the multivariate logistic regression models, indicating that those respondents who reported using apps were more likely to respond with ‘definitely’ or ‘sometimes’ to the items discussing the impact of data from RMT on their work.

In relation to clinical setting, the reference category across all regression analyses in sub-questions 10.1 to 10.5 was ‘primary care’. The univariate analyses showed that in all 5 sub-questions, respondents from at least one category of setting gave responses which were significantly different to those in primary care. The odds ratios in these cases were below one, indicating that respondents working in multiple settings outside of primary care were less likely than those in primary care to respond positively (choose ‘definitely’ or ‘sometimes’ rather than ‘unsure’, ‘hardly’ or ‘never’) concerning the impact of RMT on their work. However, in the multivariate regression analyses, only 4 out of 5 sub-questions (10.1, 10.2, 10.4 and 10.5) showed significant differences between primary care and other settings, when other variables were controlled for.

With regard to job role, the reference category was ‘doctors (excluding GPs)’. In 3 out of the 5 question items, the general practitioner job role was significantly different to the reference category (doctors excluding GPs) in the univariate regression analysis, however this was not sustained in the multivariate models for these sub-questions where other variables were controlled for.

In relation to age, only sub-question 10.5, concerning the impact of RMT data on decision-making processes in consultations, showed any difference between age groups. Here, according to both univariate and multivariate models, respondents in the 18–30 age category were less likely than those in the reference category (ages 41–50) to indicate that patients’ RMT data had an impact on their decision-making processes during consultation. There was no significant difference between any specialism and the reference category (mental health) in the regression models for any of the question items.

While HCPs reported using apps, some of the app names they reported appear to be patient-facing, and it is not clear from our data whether they are using these apps themselves or are recommending these to their patients. We cannot tell from our study the extent to which apps and wearables are prescribed or recommended to patients by healthcare professionals. In the UK, some smartphone apps are now highlighted in the NHS apps library [33], and the National Institute for Health and Care Excellence (NICE) has issued guidance that HCPs consider digital mobile health interventions, including smartphone apps, for behaviour change [34]. However, the present study has not been able to ascertain how many HCPs actively recommend these to their patients and this is an interesting question for future work. In other European countries including Germany and Sweden, governments incentivise HCPs to recommend digital solutions to their patients [35]. Further work could also usefully compare use of treatment pathways involving apps and wearables in countries with different incentivisation policies to understand the benefits of these approaches. There are also questions around the regulation of apps and the effect of regulation on their adoption that could be usefully explored.

Recently, NHS guidance has been aimed at primary and secondary care services reducing face-to-face appointments during the COVID-19 pandemic [23, 36] and there is some suggestion that healthcare professionals may have become more favourably disposed to the use of digital technologies to support their clinical practice during this period [37]. However, this evidence relates largely to the use of video conferencing software and telephone consultations rather than the use of wearables and apps to generate data on behavioural and physiological aspects of health, although these are beginning to be evaluated in practice. Given that clinicians would be required to review the data from such RMT on a digital device (e.g. computer) even if patients were present at a face-to-face appointment, it is not clear whether the recent reduction in face-to-face appointments would cause a change of attitudes towards the use of this data in practice. Further work would be required to understand precisely how views on these specific areas may have changed or not because of the reorganisation of care for non-COVID-19 patients during the pandemic.

A key strength of this study is the large sample size, with over 1000 respondents. Another is the exploration of the impact that RMT data has on patient care. The application of regression techniques has enabled us to control for demographic factors when exploring the variation in these views. Despite a coordinated effort to facilitate responses from clinicians working outside of the UK (translation into multiple European languages, advertisement via social media, network connections), the proportion of those working in other countries among respondents to the survey was small. However, we were able to reach a representative sample of respondents from different age groups, job roles and clinical settings, which has enabled sub-group analyses of the views and experiences presented. There is some risk that the self-selecting sample for this study were those most interested in the use of technology and this may mean that views expressed are more positive than is the case across all healthcare professionals. A small percentage of respondents (6%) were from management or healthcare science roles where their contact with patients, and therefore their understanding of factors relevant to the application of RMT, may be more limited. As an exploratory study, the data were analysed without multiplicity adjustment and results were mainly interpreted for exploratory and not confirmatory purposes [38].