Inequalities and the NHS App (quantitative study 2)

Introduction

Parts of this chapter have been reproduced and/or adapted with permission from Kc et al.57 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.

As we highlighted in the introduction to this monograph, the impacts of patient portal adoption and use on inequalities are not well understood.16,17 Inequality patterns not only relate to patient portal adoption, which is generally lower in those who are elderly,18,19 ethnic minorities18,20 and from a lower socioeconomic group.19 They are also reflected on sustained engagement trends and the use of different portal functionalities.20 User preferences for specific patient portal features, and their ability to engage with certain functions, is influenced by factors such as age, digital literacy, health conditions and personal healthcare needs.21–23 For example, a study looking at patient portal adoption by older adults revealed a notable interest among this group to use functions enabling access to their health information and appointment booking. However, there were also reported issues surrounding usability of the technology itself and a need for increased user support.58 In a separate study, younger adults and females exhibited greater use of patient portal features overall. However, older patients utilised functions associated with education more frequently and individuals experiencing higher poverty used features related to billing and insurance more22 Such variations in patient engagement with different patient portal features may indicate an inherent user preference related to patient-specific needs. However, they also may indicate barriers to sustained use that extend beyond initial enrolment.

Previous research on patient portal use in NHS primary care services identifies similar disparities and they show greater use of certain functions such as medication requests and appointment booking, especially among people with more long-term health conditions.59 In a more recent study, males, non-White ethnic groups, and older adults (over 85 years) showed reduced use of certain NHS online health services such as online appointment booking and prescription ordering, whereas, those with long-term conditions and from a higher socioeconomic status were generally more likely to use these functions.60 Our earlier study on the NHS App use also found similar patterns of inequalities in its adoption, with higher registration rates among larger-sized, less deprived and less ethnically diverse GP practices and among younger population groups.40 While an overall increase in the adoption of the various app features was seen over time, particularly in response to COVID events40 relationship between the patterns of inequalities in uptake and the use of different NHS App features remains unexplored.

These issues need careful consideration as digital innovations are known to follow an inverse care law, where the availability and use of health care is inversely related to the population who need it most61,62 Furthermore, as national platforms such as the NHS App continue to evolve and play more prominent role in healthcare delivery,63 prioritising sustained patient engagement and participation beyond initial uptake is essential to achieve meaningful success. Therefore, identifying patterns of uptake of the various app features and understanding how different population groups are engaging with those functions may be important first steps in advancing equitable outcomes. This will not only allow identification of groups who most benefit from using digital channels but will also help to reallocate resources to support those who remain at risk of exclusion. This study aims to address these knowledge gaps by analysing the patterns of uptake of various NHS App functions and changes in use across different population groups.

Methods

Data sources

An ecological study using data at the GP practice level exploring differences in the usage of NHS App features by the GP-registered population in England was conducted. The outcome variables were weekly registrations, logins, prescriptions ordered, medical record views and appointments booked. NHS App usage data were available for the period between March 2020 and June 2022, provided by the NHS Digital team as anonymised daily logs from the NHS App dashboard.41

Data on the sociodemographic characteristics of the GP-registered population and their healthcare needs at the GP level were used as covariates. Data on the age and sex of all GP-registered populations were obtained from the NHS Digital website, which is a reliable source for the collection and processing of national data across the health and social care sector in the UK.42 We excluded data for those under 15 years as the NHS App is not offered to younger age groups,64 and the resultant total for all GP registered males and females was combined to calculate the GP practice size. Population aged 15–34 years was used as the age identifier for analyses, and the percentage of males was used as the identifier for sex.

Data on long-term healthcare needs were extracted from the GPPS, which is a survey of approximately 2.4 million adult GP-registered patients in the UK.44 Population-weighted positive response to the question, ‘Do you have any long-term physical or mental health conditions, disabilities or illnesses?’ obtained from the GPPS database was included.

Information on ethnic composition and the IMD quintile were obtained from Fingertips public health profile, which is an online repository of a range of health and social care indicators mapped as public health profiles at local and national levels.65 The identifier for ethnicity included the per cent of population who reported their ethnicity as ‘White’.

Data obtained from these public health data sources were then ranked into quartiles (four groups). Quartile 1 included practices with the lowest population percentage for the given variable, and quartile 4 the highest. The only exception to this was the IMD split which followed the ONS's method of ranking into quintiles (five groups)65 where quintile 1 included practices with the lowest IMD score (i.e. least deprived practices), and quintile 5 included practices with the highest (i.e. most deprived practices).

For analysis, covariate data were linked to the NHS App data using the GP practice code, which serves as a unique identifier code for each GP practice in England. Practices with incomplete data on practice size or those with missing area codes were removed to avoid errors during analysis (n = 126). Practices with < 200 registered patients were also excluded (n = 20) to remove practices that service atypical populations or those with unreliable practice-level data.

Analyses

We had complete weekly usage data for a period between 23 March 2020 and 27 June 2022 for a total of 6386 GP practices in England. Outcomes were measured as weekly rates per 1000 registered GP practice population and as cumulative totals at the end of the study period. Negative binomial regression models using IRR were used to explore subgroup differences in the patterns of NHS App use by patient sociodemographic characteristics, their long-term healthcare needs and GP practice size. The regression models compared variations in app usage for the different covariates in comparison to the reference group Q1 (i.e. for IMD = least deprived practices and for all other covariates = practices with the lowest population percentage for the given variable). IRR values were then used to calculate the percentage change to report the relative difference between the different covariate quintiles. All statistical analyses were performed using Stata 17.0 (StataCorp LP, College Station, TX, USA).

Results

We present cumulative totals for all NHS App features in Table 6, and figures of percentage changes from fully adjusted models with full IRRs are presented in Appendix 3, Table 8.

Registrations

Across the whole study period, there were 24,168,835 NHS App registrations (Table 6). Results showing subgroup differences in NHS App registrations using the fully adjusted models are presented in Figure 5 and these show strong associations between IMD and app registration, with lower rates of registration in all quintiles than Q1 (i.e. least deprived) (test for trend p < 0.001). For example, registration rates were 34.0% lower in Q5 than Q1 (p < 0.001). Registrations were also lower in practices with higher proportions of male patients (e.g. 7.1% lower in Q4 than Q1 p < 0.001) and in practices with greater proportions of patients with long-term health needs (e.g. 3.6% lower in Q4 compared with Q1, p < 0.001) (test for trend p < 0.001).

FIGURE 5 Weekly NHS App registration rates per 1000 GP registered population for the different covariate quantiles. Where IMD = Index of Multiple Deprivation. IMD Q1 (reference group) = least deprived practice and IMD Q5 = most deprived practices. For all other covariates, Q1 = practices with the lowest population percentage for the given variable and Q4 = practices with the highest population percentage for the given variable. Reproduced with permission from Kc et al.57 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure above includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Registrations were higher in practices with higher proportions of White patients (e.g. 1.9%, lower in Q4 compared with Q1 p < 0.001) and those with more patients from the youngest age group (e.g. 3.1%, lower in Q4 compared with Q1 p < 0.001). Registration rates were also higher in larger-sized practices (e.g. 3.2% lower in Q4 than Q1 p < 0.001) (test for trend p < 0.001).

Logins

Across the whole study period, there were 447,976,852 NHS App logins (see Table 6). Results showing subgroup differences in NHS App registrations using the fully adjusted models are presented in Figure 6 and they show strong associations between IMD and app logins, with lower login rates in all quintiles compared to Q1 (i.e. least deprived) (test for trend p < 0.001). For example, login rates were 34.9% lower in Q5 than Q1 (p < 0.001). Logins were also lower in practices with higher proportions of male patients (e.g. 10.4% lower in Q4 than Q1, p < 0.001) and in practices with greater proportions of patients with long-term health needs (e.g. 3.4% lower in Q3 than Q1 and 1.7% lower in Q4 than Q1, p = 0.001) (test for trend p < 0.001).

FIGURE 6 Weekly NHS App login rates per 1000 GP registered population for the different covariate quantiles. Where IMD = Index of Multiple Deprivation. IMD Q1 (reference group) = least deprived practice and IMD Q5 = most deprived practices. For all other covariates, Q1 = practices with the lowest population percentage for the given variable and Q4 = practices with the highest population percentage for the given variable. Reproduced with permission from Kc et al.57 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure above includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

App logins were generally higher in practices with a higher proportion of White patients (e.g. 16.4% higher in Q2 than Q1 and 9.1% higher in Q4 than Q1, p < 0.001) and in those with more patients from the youngest age group aged 15–34 years (e.g. 5.6% higher in Q4 than Q1, p < 0.001). NHS App login rates were also higher in the larger-sized practices (e.g. 11.7% higher in Q4 than Q1, p < 0.001) (test for trend p < 0.001).

Appointments booked

Across the whole study period, there were 1,757,975 GP appointments booked using the NHS App. Results showing subgroup differences in the rates of appointment booked using the fully adjusted models are presented in Figure 7 and they show strong associations between IMD and appointments booked, with lower rates of appointment booking rate in all quintiles than Q1 (i.e. least deprived) (test for trend p < 0.001). For example, rates of appointments booked were 45.7% lower in Q2 than Q1 and 39.7% lower in Q5 compared to Q1 (p < 0.001). Appointments booking rates were also lower in practices with higher proportions of male patients (e.g. 36.4% lower in Q4 than Q1, p < 0.001) and in those with more patients with long-term health needs (e.g. 20.0% lower in Q4 than Q1, p = 0.001) (test for trend p < 0.001).

FIGURE 7 Weekly rates of appointments booked using the NHS App per 1000 GP registered population for the different covariate quantiles. Where IMD = Index of Multiple Deprivation. IMD Q1 (reference group) = least deprived practice and IMD Q5 = most deprived practices. For all other covariates, Q1 = practices with the lowest population percentage for the given variable and Q4 = practices with the highest population percentage for the given variable. Reproduced with permission from Kc et al.57 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure above includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

The rates of appointments booked were higher in practices with the highest proportion of White patients (e.g. 14.1% higher in Q4 than Q1, p < 0.001) and those with the most patients from the youngest age group aged 15–34 years (i.e. 46.5% higher in Q4 than Q1, p < 0.001). The rates of appointments booked were also higher among the larger-sized practices (e.g. 73.4% in Q4 than Q1, p < 0.001) (test for trend p < 0.001).

Medical record views

Across the whole study period, there were 117,962,559 medical record views through the NHS App. Results showing subgroup differences in the rates of medical records viewed using the fully adjusted models are presented in Figure 8 and they show strong associations between IMD and medical record views, with higher rates in all quintiles than Q1 (i.e. least deprived) (test for trend p < 0.001). For example, rates of medical record views were 32.3% lower in Q5 than Q1 (p < 0.001). They were also lower in practices with higher proportions of male patients (e.g. 12.0% lower in Q4 than Q1, p < 0.001) (test for trend p < 0.001).

FIGURE 8 Weekly rates of medical record views using the NHS App per 1000 GP registered population for the different covariate quantiles. Where IMD = Index of Multiple Deprivation. IMD Q1 (reference group) = least deprived practice and IMD Q5 = most deprived practices. For all other covariates, Q1 = practices with the lowest population percentage for the given variable and Q4 = practices with the highest population percentage for the given variable. Reproduced with permission from Kc et al.57 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure above includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

Rates of medical record views were higher in practices with a higher proportion of White patients (e.g. 34.0% higher in Q2 than Q1 and 28.7% higher in Q4 than Q1, p < 0.001) and in those with more patients from the youngest age group aged 15–34 years (e.g. 6.7% higher in Q2 than Q1, p < 0.001 and 1.7% higher in Q4 than Q1, p = 0.003). The rates of medical records views were also higher in larger-sized practices (e.g. 23.9% higher in Q4 than Q1, p < 0.001) and in practices with more people with long-term healthcare needs (e.g. 6.0% in Q4 than Q1, p < 0.001) (test for trend p < 0.001).

Prescriptions ordered

Across the whole study period, there were 21,324,472 prescriptions ordered using the NHS App. Results showing subgroup differences in the rates of prescriptions ordered using the fully adjusted models are presented in Figure 9 and they show strong associations between IMD and rates of prescriptions ordered, with lower rates in all quintiles compared to Q1 (i.e. least deprived) (test for trend p < 0.001). For example, rates of prescriptions ordered were 9.9% lower in Q5 than Q1 (p < 0.001). Prescription order rates were also lower in practices with higher proportions of male patients (e.g. 14. 5% lower in Q4 than Q1, p < 0.001) and in practices with a higher proportion of people aged 15–34 years old (e.g. 14.8% lower in Q4 than Q1, p < 0.001).

FIGURE 9 Weekly rates of prescriptions ordered using the NHS App per 1000 GP registered population for the different covariate quantiles. Where IMD = Index of Multiple Deprivation. IMD Q1 (reference group) = least deprived practice and IMD Q5 = most deprived practices. For all other covariates, Q1 = practices with the lowest population percentage for the given variable and Q4 = practices with the highest population percentage for the given variable. Reproduced with permission from Kc et al.57 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The figure above includes minor additions and formatting changes to the original text. (Figure not reproduced here; see the original article.)

The rates of prescriptions ordered were significantly higher in practices with a higher proportion of White patients (e.g. 119.4% higher in Q3 than Q1 and 130.40% higher in Q4 than Q1, p < 0.001) and in larger practice sizes (e.g. 20.7% higher in Q4 than Q1, p < 0.001). The rates of prescriptions ordered were also higher in practices with more people with long-term healthcare needs (e.g. 18.3% higher in Q4 than Q1, p < 0.001) (test for trend p < 0.001).

Discussion

Principal results

This study exploring NHS App use showed high patient engagement with variations in the use of different features overall and across the different covariate categories. Subgroup analyses found a clear deprivation gradient with lower rates of use of the various NHS App features among practices in the most deprived areas and in those with more registered males. In contrast, the use of the different features was higher among the largest-sized practices and in those with more White patients, with a 130% difference in the rates of prescriptions ordered in practices with the highest proportion of White patients than Q1. These patterns varied in practices with more younger patients, which had higher appointment booking rates, but lower rates of prescription ordered. Similarly, in practices with the most people with long-term healthcare needs, there were significantly higher rates of medical record views and prescriptions ordered but lower appointment booking rates, registrations, and logins.

Some results stand out as particularly striking, for example, with ordering prescriptions online – one of the most used features – there is a strong ethnicity gradient, with practices with a higher proportion of White patients using this function more than twice as much as practices with a lower proportion of White patients, suggesting particular opportunities for changes in design and practice to reduce inequality.

Strengths and limitations

This is the first study to explore variations in the use of different NHS App features in relation to different sociodemographic groups. It provides a crucial understanding of digital health engagement in a real-world setting. While these initial findings suggest an unequal pattern of patient engagement with the NHS App features in relation to patient age, sex, ethnicity, deprivation and patient healthcare needs, there are limitations in using population-level data to understand individual-level differences. Although patients can select their data-sharing preferences through the NHS App,64 individual-level data are not yet available for research, which restricts a more direct evaluation of these usage patterns.

Furthermore, this study used data over 27 months since the launch of the NHS App, providing insight into its usage trends. However, challenges in capturing and controlling the effects of the influences of wider confounding factors such as regulatory changes, competing availabilities of similar commercial digital health platforms and changes in strategic decisions impacting provider and public response to the app48 have implications for our research findings. In addition, this study includes a number of sociodemographic variables and their proxies identified in consultation with the wider research team and the PPIE group. However, the full range of indicators used (e.g. additional age ranges and ethnic groups) and broader socioeconomic variables have been excluded due to analytical considerations and data acquisition challenges. While some variables are encompassed within existing indicators (e.g. income and education for IMD value calculation) and their proxies offer reference points for the different sociodemographic categories, these factors may operate in silos to influence how people engage with digital resources,66 and they merit further exploration in future studies.

Also, exclusions were applied to remove practices with incomplete data and small sample sizes to remove practices serving unconventional populations or those with unreliable practice-level data for the outcomes. This along with the ecological nature of the analysis may affect the generalisability of results and present missed opportunities to capture the demography these practices serve and their distinct challenges. Nevertheless, our study provides a comprehensive overview of public uptake of the NHS App using a nationally representative sample of the GP-registered population in England.

Comparison with prior work

The results of the current study further the findings presented in Chapter 3. They indicate a continued pattern of inequality in patient engagement with the NHS App. A recent publication from NHS Digital highlights that the NHS App currently has over 30 million registrations, with over 65 million patient record views and 22 million repeat prescription orders through the app in the year 2022 alone.4 While our data support these findings by highlighting high usage across all app features, evidence of a varied pattern of use of the different app functions among different population groups is congruent with wider research in the field.59,60 Most notably, reductions in use of the different functions linked to increases in deprivation and reduction in ethnic diversity across practices reinforce concerns related to the digital inverse care law62 and underscore persistent equity challenges that may warrant targeted support and interventions. A previous study exploring disparities in the uptake of NHS primary care services including the NHS App highlighted the scale of digital health inequality related to non-activation of the app, which corresponded to approximately 4.27 million patients living in the most deprived quintiles in England.67 This figure could be more pronounced when considering the inequalities in patient interaction with the NHS App features as highlighted by our findings, indicating that activation may not necessarily translate to sustained utilisation. This emphasises the multifaceted nature of digital health disparities and prompts a critical examination of the factors contributing to unequal access, as well as continued user interaction.

Although the diffusion of innovation theory tells us that some of these differences may level out as the technology is more widely used by the general population,68 certain patient portal features may continue to have differential adoption based on individual capacity, need and access.22 Discrepancies in use of the functions, as highlighted by our results, such as younger adults showing increased appointment bookings but decreased prescriptions, underscore these complexities confounded by various factors. Younger adults generally have better capacity for engagement with digital health technology, yet they generally use healthcare utilisation less overall due to their low healthcare needs.69 Therefore, the reduction in prescription orders through the NHS App among this group may reflect a good health status overall and low care needs. However, the evidence of increased uptake of certain transactional services such as appointment booking in this user group may also be indicative of their technological literacy and motivation to utilise these services. While we could not measure these associations directly for the NHS App, it warrants further exploration in future studies.

In congruence with previous studies, our results also indicate pronounced variations in different app feature usage among individuals with long-term health conditions.59,60 While appointment booking rates decreased, medical record views and prescription orders increased, signalling heightened patient engagement with features catering to their ongoing healthcare needs, such as prescription refills and health record monitoring.70 The decline in appointment booking could be associated with the configuration of this feature, which is contingent upon the specific provisions implemented by individual practices. It could also be linked to variations in the self-care ability of individuals with long-term healthcare needs, as they might often depend on family, social networks, and caregivers for support, potentially benefiting from delegated digital access for care co-ordination.16,71 Although the NHS App offers access to linked profiles for family members and carers, the availability of these proxy access entitlements relies on individual GP practice provisions72 Therefore, further analysis of app functions, variations in the activation of features dependent on GP practice provisions and patients’ ability for self-care can offer further insight into broader system issues influencing NHS App use.

However, disparities in access and use of digital tools may not inherently lead to inequalities in healthcare outcomes. If those with better capacity, interest and opportunity to engage with technology utilise digital services more, this may reduce pressures on the healthcare system and offer cost-saving opportunities to improve care for those who continue to benefit from in-person support.73 This may be particularly relevant to a patient-facing tool such as the NHS App, which is scaled at a national level, offering a range of functions for diverse population groups. Therefore, identifying early users and non-users of the NHS App and supporting engagement with features most relevant to their specific needs may be crucial to creating equal outcomes for all.74 There is also a need for ongoing assessment of the uptake trends of the NHS App contingent on continued policy efforts and facilitation strategies to understand its ongoing relevance, user satisfaction, and equitable access, contributing to the overall success of such digital health initiatives.

Our results, showing variation in use of the app overall and specific functions by various patient characteristics, suggest opportunities for design, policy and practice change. This might include targeted marketing of the app and its functions towards specific groups – such as advertising the prescription function to wider ethnic groups, or ensuring user research and testing with diverse populations, to make sure future design decisions do not reinforce existing variation.

Conclusions

We found ongoing patient engagement with the various NHS App functions. However, inequalities in the rates of use of the different app features indicate an ongoing digital divide. NHS App uptake and use followed a strong deprivation gradient and influences of patient age, ethnicity and healthcare needs, indicating the need for dedicated efforts to meaningfully engage different population groups. Further research is also needed to explore how these differences affect disparities in health and care outcomes.

Our study contributes to the growing body of evidence on digital health engagement and lays the groundwork for future research and policy-making. It underscores the importance of continuous monitoring, evaluation, and adaptation of digital health initiatives to ensure they meet the evolving needs of diverse populations and contribute to equitable health outcomes for all. The NHS App has the potential to be a valuable tool for patient care in the NHS in England, but it is important to address the digital divide and ensure that all individuals can use the app effectively irrespective of their demographic, ethnic or socioeconomic characteristics.