Telemonitoring at scale for hypertension in primary care: An implementation study

Background While evidence from randomised controlled trials shows that telemonitoring for hypertension is associated with improved blood pressure (BP) control, healthcare systems have been slow to implement it, partly because of inadequate integration with existing clinical practices and electronic records. Neither is it clear if trial findings will be replicated in routine clinical practice at scale. We aimed to explore the feasibility and impact of implementing an integrated telemonitoring system for hypertension into routine primary care. Methods and findings This was a quasi-experimental implementation study with embedded qualitative process evaluation set in primary care in Lothian, Scotland. We described the overall uptake of telemonitoring and uptake in a subgroup of representative practices, used routinely acquired data for a records-based controlled before-and-after study, and collected qualitative data from staff and patient interviews and practice observation. The main outcome measures were intervention uptake, change in BP, change in clinician appointment use, and participants’ views on features that facilitated or impeded uptake of the intervention. Seventy-five primary care practices enrolled 3,200 patients with established hypertension. In an evaluation subgroup of 8 practices (905 patients of whom 427 [47%] were female and with median age of 64 years [IQR 56–70, range 22–89] and median Scottish Index of Multiple Deprivation 2012 decile of 8 [IQR 6–10]), mean systolic BP fell by 6.55 mm Hg (SD 15.17), and mean diastolic BP by 4.23 mm Hg (SD 8.68). Compared with the previous year, participating patients made 19% fewer face-to-face appointments, compared with 11% fewer in patients with hypertension who were not telemonitoring. Total consultation time for participants fell by 15.4 minutes (SD 68.4), compared with 5.5 minutes (SD 84.4) in non-telemonitored patients. The convenience of remote collection of BP readings and integration of these readings into routine clinical care was crucial to the success of the implementation. Limitations include the fact that practices and patient participants were self-selected, and younger and more affluent than non-participating patients, and the possibility that regression to the mean may have contributed to the reduction in BP. Routinely acquired data are limited in terms of completeness and accuracy. Conclusions Telemonitoring for hypertension can be implemented into routine primary care at scale with little impact on clinician workload and results in reductions in BP similar to those in large UK trials. Integrating the telemonitoring readings into routine data handling was crucial to the success of this initiative.

TITLE: Telemonitoring at scale for hypertension in primary care: an implementation study SUMMARY Background: While evidence from randomised controlled trials shows that telemonitoring for hypertension is associated with improved blood pressure (BP) control, health care systems have been slow to implement it, partly because of inadequate integration with existing clinical practices and electronic records. Neither is it clear if trial findings will be replicated in routine clinical practice at scale. We aimed to explore the feasibility and impact of implementing an integrated telemonitoring system for hypertension into routine primary care.
Methods and findings: This was a quasi-experimental implementation study with embedded qualitative process evaluation set in primary care in Lothian, Scotland. We described the overall uptake of telemonitoring and in a sub-group of representative practices, used routinely acquired data for a records based controlled before and after study; plus qualitative data from staff and patient interviews and practice observation. The main outcome measures were intervention uptake, change in BP, change in clinician appointment use, and participants' views on features that facilitated or impeded uptake of the intervention.
Seventy-two primary care practices enrolled 3,200 patients with established hypertension. In an evaluation subgroup of eight practices (905 patients of whom 427 (47%) were female with median age of 64 , and median SIMD 2012 decile of 8 (IQR 6 -10), mean systolic BP fell by 6.55mmHg (SD 15.17) and mean diastolic BP by 4.23mmHg (SD 8.68). Compared with the previous year, participating patients made 19% fewer face-to-face appointments compared with 11% fewer in patients with hypertension who were not telemonitoring. Total consultation time for participants fell by 15.4 mins (SD 68.4) compared with 5.5 mins (SD 84.4) in non-telemonitored patients. The convenience of remote collection of BP readings and integrating these readings into routine clinical care was crucial to the success of the implementation. Limitations include the fact that practices and patient participants were self-selected, younger and more affluent than nonparticipating patients and the possibility that regression to the mean may have contributed to the reduction in BP. Routinely acquired data is limited in terms of completeness and accuracy.
Conclusions: Telemonitoring for hypertension can be implemented into routine primary care at scale with little impact on clinician workload and resulting in similar reductions in BP to large UK trials. Integrating the telemonitoring readings into routine data handling routines was crucial to the success of this initiative.
Keywords: Telemonitoring; hypertension; implementation research Why was this study done?
 Although uncontrolled hypertension is the biggest remediable cause of stroke and myocardial infarction, and anti-hypertensive medications are effective, many patients have uncontrolled BP.
 Despite strong evidence that telemonitoring encourages medication use and is effective at lowering BP, clinicians have been slow to adopt it, in part due to poor integration with routine clinical processes. What did the researchers do and find?  Based on our previous research with clinicians on the desirable attributes of a telemonitoring system, NHS Lothian developed an integrated system which provided regular summaries of patient home-monitored BP readings to their GP practice which were delivered alongside routine laboratory results.
 We observed the roll-out of this system, interviewing patients and clinicians about their attitudes to the innovations to determine what worked and did not in terms increasing uptake.
 In a group of eight practices we collected routinely acquired data on blood pressure, clinician appointments and other resource use and compared these to the previous year.
Resource use was compared with patients in these practices with high BP who did not use telemonitoring.
 We found that the intervention was popular in many but not all practices. Patients who used the system took up fewer appointments in the year of the intervention compared with the previous year than those who did not. BP fell in the intervention group in line with findings of UK randomised controlled trials.
What do these findings mean?
 The findings suggest that introducing telemonitoring to routine practice at scale is feasible.
 Although not definitive, they provide some reassurance that the intervention did not increase practice workload and that in routine practice the improvements in BP control were similar to those in controlled trials.
 However, the people who took part were not entirely typical of practice populations as a whole, being younger and slightly more affluent.

INTRODUCTION 1 2
Hypertension is common among people over the age of fifty and an 3 important risk factor for cardiovascular disease [1]. Although effective 4 management greatly reduces the risk of cardiovascular events, blood 5 pressure (BP) remains uncontrolled in many people [2]. This is in part 6 due to poor adherence to medication [3], but also reluctance on the 7 part of clinicians to intensify therapy [4] and patients to accept 8 intensified therapy [5]. 9 In the United Kingdom (UK), hypertension is managed in primary care 10 mainly by practice nurses assisted by health care assistants with 11 specific training in BP monitoring, supported by general practitioners 12 (GPs). In Scotland (population 5.45  Why then, given consistent findings from randomised controlled trials 36 (RCTs), has telemonitoring not been more widely adopted? 37 Implementing new models of care at scale is challenging [18,19], 38 particularly in the context of clinical teams already working at full 39 stretch. Telehealth trials have tended to recruit relatively few, often 40 highly selected individuals who have been followed-up for relatively 41 short periods leaving unanswered questions about the day-to-day 42 practicality of managing large numbers of patients [20]. In addition, 43 the adoption, effectiveness and impact on resources of delivering BP 44 telemonitoring as a routine approach to care is unknown and may 45 differ from the experience in trials. 46 Following our earlier randomised controlled trial of telemonitoring for 47 The telemonitoring intervention implemented is described in detail in the online 86 supplement (Appendix S1 and figure S1). Participating people with hypertension were 87 provided with an electronic, oscillometric sphygmomanometer and shown how to submit 88 BP readings via a low-cost third-party text-based telemonitoring system (Florence) [22] 89 using their own mobile phone. These readings were stored in a central server and made 90 available to practices via an internet link. However, previous research had shown that a 91 major barrier to adoption was the need for clinicians to log-on to third party web-sites 92 during busy clinics [20] and so a novel element [23] was developed which automatically 93 extracted patient-generated data from the third-party website. This displayed a mean BP 94 and summarised BP values in graphical and tabular format, (Appendix S1, figure S2) and 95 dispatched these summaries, at intervals chosen by the clinician, through the routinely 96 used Primary Care data management system, Docman [24]. This obviated the need for 97 third-party log-ons and, critically, systematically presented the telemonitoring results in 98 manageable numbers on a daily basis with relatively infrequent, but data-rich reports 99 integrated into the electronic medical record and seen alongside routine laboratory results 100 and hospital communications. Between reports, patients were informed by automated text 101 responses if submitted readings were low, normal, high or very high and advised to follow  Table S1 of the 119 supplementary appendix lists how these barriers were addressed. In essence this 120 comprised of: motivating clinicians and service planners by demonstrating how an 121 evidence-based BP telemonitoring service could improve outcomes and potentially save 122 time; enhancing capability by providing initial training and (importantly) retraining and 123 ongoing support as required; and providing opportunity by supporting recruitment drives 124 within individual practices. Local practitioner 'champions' were recruited to be early 125 adopters in the expectation that they would demonstrate and promote the intervention to 126 colleagues [29]. Practices were encouraged to adapt the system to suit their routines and 127 priorities with flexibility in terms of the initial patient groups targeted (some focusing on 128 uncontrolled BP). Clinicians were continuously involved in the development of the implementation strategy, trialling different recruitment methods and in managing 130 telemonitoring data. Learning from practices was shared in regular newsletters. 131

132
Outcomes of interest 133 134 Our primary implementation outcome was the number of participating practices and the 135 overall number of patients recruited. In addition, in a group of eight practices, chosen to 136 represent a range of socio-economic status, list size, city centre/small town, and 137 fast/slow patient recruitment, we measured patient engagement and adherence, 138 resource and medication use, BP change over time, and gauged clinician and patient 139 acceptability of the implementation. 140 These practices agreed for data on BP, anti-hypertensive medication prescriptions, 141 practice workload (as recorded in the appointment system) for all patients with 142 hypertension including those not being managed by the Scale-Up BP system to be 143 extracted from their electronic health record (EHR), linked to data from the 144 telemonitoring system, de-identified and transferred to the local NHS safe haven (a 145 secure analysis server which does not permit any data to be taken away). 146 This allowed comparison of resource use between participating and non-participating 147 patients. Primary Care stored data were collected and transferred by a data extraction 148 service (Albasoft) through software already resident on the practice computers. Time to 149 action of very high (>160mmHg) and very low (<90 mmHg) average systolic BP readings 150 was obtained from manual searching of Docman reports and practice records. Change in 151 BP over time was derived from telemonitored data from Scale-Up BP patients only. 152 As the purpose of the study was primarily to determine uptake of the intervention a 153 formal power calculation was not deemed appropriate. 154

Assessment of attrition 155
Patients were considered to have discontinued monitoring if they did not record any 157 readings for seven months during the observation period. This period was chosen 158 because we did not want to exclude patients with good BP control some of whom were 159 asked to submit readings only every six months. We collected data on (i) the total number of appointments, (ii) total number of face-to-face 172 appointments, (iii) total consultation time, and (iv) total consultation time in the surgery. 173 The practice appointment system was used to identify all interactions including face-to-174 face and planned telephone consultations, home visits or administration activities (for 175 example, prescribing a requested repeat prescription). Face-to-face consultations could be 176 reliably identified as they had different arrival and start times. We determined time spent 177 in activities recorded in the appointment system in total and in face-to-face activities for all 178 people with hypertension in the eight practices and could compare those taking part in the 179 intervention with those who were not. 180 There was no automated extraction method that could determine time spent in 181 consultations which were not recorded in the appointment system such as ad-hoc phone 182 calls or record checking. 183 Analysis of resource use outcomes was conducted by computing totals/averages across all 184 patients and comparing before-and-after, and separately performing a patient-level analysis 185 by computing sum totals/averages within patients. We only included patients with a full 186 year of data before and after the start of telemonitoring. Regarding the consultation time 187 analysis, the first analysis weighted patients with a greater number of appointments more 188 heavily than in the second patient-level analysis, which is appropriate when simply trying The telemonitored patients' appointments data were compared against comparator 195 patients (all patients with a diagnosis of hypertension, who did not use the 196 telemonitoring system), from the same eight evaluation practices. For the 197 comparator group we could not choose the start of telemonitoring as the "anchor 198 point" to define the before and after groups, we therefore used a randomly chosen 199 date for the anchor point within a date range consistent with the telemonitoring 200 group (between October 2015 and June 2016). Outlier consultation durations (>30 201 minutes: three times the length of a typical primary consultation in the UK) were 202 excluded from all analyses involving consultation time as these are very likely to be 203 erroneous and so it was not appropriate to retain them in the analyses. These long 204 consultation times are almost certainly because clinicians fail to close the electronic 205 record, for example before taking breaks or other tasks. It would be very unlikely that 206 a consultation related to BP management would be so long; including these would 207 lead to an overestimate of resource use. 208 Linear mixed effects models were then fitted to the resource outcomes to compare 209 telemonitored patients with the comparator group, adjusting for female gender, SIMD ≥ 210 5 (less deprived), second surgery measured systolic BP (after Sept 2015), age (in years), 211 and GP practice as a random effect. As for the attrition analysis, natural B-spline 212 functions were used to model the effects of age and systolic BP on outcome. 213 Change in BP over six-months and proportion controlled 214 215 The difference between the second BP reading and the last BP reading occurring 6-12 216 months after the second reading was calculated for each patient (Second -Last). Because 217 of the known disparity between home and office BP readings we used the second 218 telemonitored reading as the first was sometimes taken in the GP practice as a 219 demonstration of the system to the patient. Only patients with a full year's follow-up were 220 included. Summary statistics (mean, median, standard deviation, 221 lower quartile, upper quartile, minimum, maximum) were calculated based on these BP 222 differences, overall and stratified by age, gender, SIMD and starting systolic BP. 223 To assess the proportion of patients whose BP was uncontrolled as defined by National 224 Institute of Health and Clinical Excellence Guidelines [31] (home systolic BP>135mmg) we 225 categorised the BP data according to level of control of BP, and compared the proportions 226 with uncontrolled BP at baseline with 6-12 months later. Patients were only included in this 227 analysis if they had complete data at baseline and 6-12 months. 228

Assessment of medication prescribed 229 230
For each patient, we calculated the total number of anti-hypertensive medication 231 prescriptions in the year before the patient first started telemonitoring and in the year 232 after. To avoid bias, we only included patients with a full year of data before and after the 233 start of telemonitoring. Summary statistics were then computed overall and stratified by age, gender, SIMD and initial systolic BP. SAS software version 235 9.4 (SAS Institute Inc., Cary, NC, USA) was used to analyse the data. 236 237 Mean defined daily doses (DDD) were calculated for the 6 months before and after date of 238 first monitoring for all patients with at least 6 months follow-up, based on WHO DDD index 239 values [32] and compared using a paired t-test (analysis undertaken in R version 3.5.3). 240 Qualitative exploration of patient and professional perceptions 241 242 In the eight practices, we undertook semi-structured face-to-face interviews with patients, 243 GPs, practice nurses and health care assistants, to explore perceptions of the 244 telemonitoring intervention in terms of ease of use, organisational adoption and 245 adaptation, barriers, facilitators and potential improvements to the implementation of 246 Scale-Up BP. See Appendix S2 for details of methodology. 247

249
Progress of the implementation. were older (median age71, IQR 62-79, range 20-90), with a slightly higher percentage of 274 females (54%) and slightly less affluent on average (median SIMD of 7). First systolic 275 blood pressure reading was similar between groups (see Table 1). The median number of 276 patients using the telemonitoring system per practice was 80 (IQR 44 to 153), compared 277 to 450 (IQR 332 to 599) controls. Apart from being slightly more affluent, generally the 278 practices were representative of practices in Lothian (see Table 2 Of the 655 patients who had been registered for more than seven months, 49 (7%) stopped texting readings. Gender, age or SIMD did not significantly predict discontinuing 293 (see appendix Table S2). However, patients with higher systolic blood pressure were 294 more likely on average to discontinue (odds ratio 1.03, 95% CI 1.01 to 1.05). Figures S7  295 and S8 show how the predicted probability of drop-out varies with systolic blood 296 pressure and age when these variables are fitted using flexible spline functions and are 297 not constrained to be linear in the statistical model. 298 Change in resource use over time recorded in the appointment system 299 300 There were 1260 appointments in total (671 face-to-face) in the year before the start of 301 telemonitoring compared to 1158 appointments in total (569 face-to-face) in the year 302 after. This corresponded to an observed 8% reduction in total appointments, and 15% 303 reduction in the number of face-to-face appointments. 304 There was an observed increase in appointments just prior to the start of telemonitoring 305 probably due to patients attending for training to measure BP and text results which might 306 potentially exaggerate potential reductions in consultation numbers (see supplementary 307 figure S5). In further analysis, in comparing changes in resource use with a comparator 308 group of non-participating patients we therefore excluded consultations within two weeks 309 of the anchor point which for telemonitoring patients may have involved appointments to 310 set up the system. We also excluded 7% of consultations for which the appointment times 311 were not recorded and an additional 3% of consultations which were recorded as being over 312 30 minutes in duration. 313 Both comparator group and telemonitoring group showed falls in numbers of all 314 appointments and face-to-face appointments. Falls in the number of appointments were 315 notably more pronounced in females, relatively more deprived (SIMD<5), and people 316 whose BP was controlled in the telemonitored group than in the comparator group (see 317 supplementary table S3). Table 3 shows the number of consultations and consultation times per patient and 319 comparator group before and after the intervention. A total of 118 telemonitoring patients 320 were included who recorded at least one consultation before and/or after the 321 intervention). As above, we excluded consultations within two weeks of the anchor point. 322 These 118 patients had a median age of 64 (IQR 54 to 69), 46 (39%) female, median SIMD 323 of 9 (IQR 7 to 10, relatively less deprived), and mean first surgery systolic BP value of 139.4 324 (SD 16.0). The patient characteristics of the comparator group were as shown in Table 1. 325 Therefore, the difference in demographics between telemonitoring and control were more 326 pronounced than for the overall cohort, except average initial systolic BP values which were 327 very similar between groups. 328 Mixed effects models were fitted to the outcomes to compare telemonitored patients with 329 the comparator group, adjusting for SIMD category (<5 or 5+), age, gender, or initial systolic 330 BP (see supplementary table S4). Time spent in all appointments in the year was 331 significantly reduced in the telemonitoring group (adjusted mean difference 16.1 minutes, 332 95% CI 0.1 to 32.1 minutes, p=0.048). Although the reduction in time spent in face-to-face 333 appointments did not achieve statistical significance, the observed reduction was clinically 334 relevant (12.7 minutes (95% CI -0.5 to 25.9 minutes, p=0.059). Note that confidence 335 intervals were wide due to the relatively small sample size in the telemonitoring group. 336 There was no significant difference in total number of appointments (face-to-face or 337 overall) in the adjusted analysis. 338 Change in BP over time 339 340 During the study period, there were 399 patients in the eight evaluation practices who had 341 two telemonitored blood pressures at least six months apart, and with at least one year of 342 follow-up. This group had similar first surgery systolic blood pressure reading compared to 343 other hypertension patients in the practices who had not been telemonitored (who also 344 had at least two readings six months apart). However, there was a lower percentage of females in the telemonitoring group 346 compared to the comparator group, patients were younger on average, and with a higher 347 median SIMD decile (median 9 versus 7) indicating a more affluent population (see table   348 4).

349
The median difference between the second telemonitored systolic BP and the last was 6 350 mmHg (IQR -3 to 15) and for diastolic BP 4 mmHg (IQR -1 to 10). Falls in BP were greater 351 for those whose BP was initially uncontrolled (people whose systolic BP was >135mmHg 352 had a median fall in BP of 13mmHg (IQR 6 to 23) whereas people whose BP was <135 had 353 no change in the median BP (IQR -7 to 7). This reduction in systolic BP was consistent across 354 age, gender and deprivation (see supplementary table S5) Clinician responsiveness to high BP readings.

373
We searched the Docman reports in the eight practices for instances where BP control was 374 poor and would normally result in management changes. We found 44 instances of average 375 systolic BP >160mmHg and three instances of an average systolic <90mmHg in Docman 376 reports. 377 The median number of days between first systolic BP >160mmHg or <90mmHg appearing 378 in the medical record (transcribed from the Docman report) to clinical review and action 379 (or not) was 13 (range 0-91 days). Actions recorded in the Primary Care records occurring 380 in response to these reports are summarised in supplementary table S6.

Change in number of anti-hypertensive medications prescribed 382 383
For 622 patients taking part in Scale-Up BP for one year we calculated average defined 384 daily doses (DDDs) of BP lowering medications in the 45 days before they started 385 telemonitoring and, to allow for discovery of a raised BP and time to respond to it, for a 386 similar period 4-6 months after they started telemonitoring. DDDs rose from 2.08 to 2.35 387 between those times, a rise of 12% (See appendix table S7). Additionally, there was a 388 small increase in the total number of prescriptions issued for anti-hypertensive 389 medications to patients over the whole year (1.06 SD 4.77 additional prescriptions). 390 Prescription numbers increased more in people whose baseline BP was > 135mmHg and 391 who were from more deprived areas. Similar changes were found in patients with raised 392 BP not taking part in Scale-Up BP (see supplementary tables S8a and b). 393

Perceptions of the implementation 394 395
Details of the process evaluation including observation of the implementation process 396 and interviews with clinicians and patients can be found in Appendix S3 and 397 supplementary tables S9 and S10. In summary clinicians found that getting regular reports 398 integrated with their usual data-handling practices was particularly helpful. Continued 399 support from the implementation team, the involvement of local champions and patient 400 enthusiasm for the service were all instrumental in building confidence in the process. 401 Initially starting patients on the system was seen as time consuming but this improved 402 with time and was perceived as balanced by subsequent time saving. Practice teams 403 adapted aspects of the intervention, particularly data handling, to their own routines. 404

406
Summary of findings 407 408 This study shows that a telemonitoring system for BP monitoring using software to 409 integrate it with normal Primary Care work patterns can be implemented at scale. BP control improved in line with that found in RCTs of telemonitoring, probably mediated by 411 an intensification in therapy [16,[33][34][35][36]. This new model of care was associated with an 412 observed reduction in the number of face-to-face appointments and consulting time. The 413 well-recognised barriers to implementation of new technologies (lack of confidence in 414 technology, workload fears, lack of time to learn and introduce new things, and scepticism 415 that the implementation will improve patient care or efficiency) were overcome through 416 engaging frontline clinicians in the development of the system, particularly local 417 champions, and strong continuous support from a facilitator team. While it was relatively 418 straightforward to persuade practices of the likely benefits of the intervention and to 419 undertake training, they varied in translating that training into action. Some practices 420 required several training sessions before they started regular recruitment while some did 421 not get started usually citing lack of time. Others, however, recruited large numbers and 422 their success persuaded others to follow them. Patients liked the system and relatively 423 few discontinued, however, it is concerning that people with less well controlled BP were 424 over-represented in those that did and this requires further investigation. 425 426   427 This intervention was implemented in typical primary care practices by clinical staff for 428 patients for routine care purposes rather than a research setting. Integration with practice 429 data management routines and the ability to adapt the intervention encouraged uptake 430 by practices and apart from initial training and remote technical support needed little 431 additional support. The use of routine data sources means that any evaluation can be 432 continued in future to determine the longer-term effects of the intervention. 433

Strengths and limitations
The Scale-Up BP patients in the eight evaluation practices were self-selected or chosen 434 by clinical staff, were slightly less deprived and younger than others in their practices with 435 hypertension and may have been more able in terms of self-management or more 436 motivated. Nonetheless, older and more deprived participating patients benefitted 437 equally or more from the intervention than others. A component of the large reduction 438 in blood pressure over time for patients who were initially uncontrolled (15mmHg) could 439 be considered as regression to the mean, but this was a largely unselected group in terms 440 of BP control and the reduction in overall mean is unlikely therefore to be a totally 441 random effect. However, the purpose of our study was not to prove effectiveness (as this 442 has already been established) [13]  There have been no direct trials exploring the impact of telemonitoring on cardiovascular 477 outcomes, but based on previous of anti-hypertensive agents, BP reductions of the 478 magnitude achieved in this study, probably through intensification of anti-hypertensive 479 therapy, if sustained, would be expected to lead to a greater than 15% reduction in risk of 480 stroke and a greater than 10% reduction in risk of coronary heart disease [39]. 481 Implications for policy, practice and future research 482 483 Telemonitoring of hypertension has been shown to be cost-effective [14,15]. 484 Sphygmomanometers have become less expensive in recent years and in Lothian will in 485 future be provided by the Health Board. Low cost telemonitoring systems have become 486 available potentially lowering costs further. We believe that the integrated system we 487 tested will improve efficiency. The system has recently been adopted by the Scottish 488 Government with plans to roll it out across the country. Further work is underway to 489 determine the optimum strategies to enrol patients in the system and to get them to 490 adhere to it. Further integration with the Primary Care medical record and making the reports available to secondary care is planned, as is the development of an app based 492 solution and improved asynchronous communication with patients. An economic 493 evaluation will be the subject of a future paper. 494

496
Scale-Up BP has demonstrated that improvements in BP control similar to those found in 497 RCTs of telemonitoring in hypertension can be achieved when implemented in routine 498 practice and at scale and that this is accomplished with no increase in workload. The 499 strategy of integrating the telemonitoring system with current data handling routines was 500 critical in the adoption and is a model for managing other long-term conditions. Based on 501 the findings from this implementation study, we conclude that Scale-up BP is ready for 502 routine use across NHS Scotland and possibly also other parts of the UK. 503