Evaluation of the National Health Services (NHS) Direct Pilot Telehealth Programme
Total Page：16
File Type：pdf, Size：1020Kb
1 Evaluation of the National Health Services (NHS) Direct Pilot Telehealth programme: 2 cost-effectiveness analysis 3 Abstract 4 Objective to evaluate the cost-effectiveness of a pilot telehealth programme applied to a wide 5 population of patients with chronic obstructive pulmonary disease (COPD). 6 Design: Vital signs data was transmitted from the home of the patient on a daily basis using a 7 patient monitoring system for review by community nurse to assist decisions on management. 8 Setting: Community services for patients diagnosed with COPD 9 Participants: Two Primary Care Trusts (PCTs) enrolled 321 patients diagnosed with COPD 10 into the telehealth programme. 227 patients having a complete baseline record of at least 88 11 days of continuous remote monitoring and meeting all inclusion criteria were included in the 12 statistical analysis. 13 Intervention: Remote monitoring 14 Methods: Resource and cost data associated with patient events (in-patient hospitalisation, 15 accident and emergency (A&E) and home visits) 12 months before, immediately before and 16 during monitoring, equipment, start-up and administration were collected and compared to 17 determine cost-effectiveness of the programme. 18 Main outcome measures: Cost-effectiveness of programme, impact on resource usage, and 19 patterns of change in resource usage. 20 Results: Cost-effectiveness was determined for the two PCTs and the two periods before 21 monitoring to provide four separate estimates. Cost-effectiveness had high variance both 22 between the PCTs and between the comparison periods ranging from a saving of £140800 23 ($176,000) to an increase of £9600 ($12,000). The average saving was £1023 ($1280) per 24 patient per year. The largest impact was on length of stay with a fall in the average length of 25 in-patient care in PCT1 from 11.5 days in the period 12 months before monitoring to 6.5 days 26 during monitoring, and similarly in PCT2 from 7.5 days to 5.2 days. 27 Conclusion: There was a wide discrepancy in the results from the two PCTs. This places 28 doubt on outcomes and may indicate also why the literature on cost-effectiveness remains 29 inconclusive. The wide variance on savings and the uncertainty of monitoring cost does not 30 allow a definitive conclusion on the cost effectiveness as an outcome of this study. It might 31 well be that the average saving was £1023 ($1280) per patient per year but the variance is too 32 great to allow this to be statistically significant. Each locality based clinical service provides 33 a service to achieve the same clinical goal, but they do so in significantly different ways. The 34 introduction of remote monitoring has a profound effect on team learning and clinical 35 practice and thus distorts the cost effectiveness evaluation of the use of the technology. Cost- 36 effectiveness studies will continue to struggle to provide a definitive answer because outcome 37 measurements are too dependent on factors other than the technology. 38 Keywords: telehealth, telemedicine, cost-effectiveness, remote patient monitoring
39 1 Introduction
40 Seventy per cent of the total health and social care budget is spent on treatment and care of 41 patients with long term conditions (LTC) [1]. Telehealth services are suggested as a solution 42 to control the spiralling cost and improve clinical outcomes. This involves installing 43 equipment in patients’ homes to measure daily physiological parameters. These are assessed 44 by clinicians (nurse, doctor or a medically trained technician) to determine the potential for 45 deterioration in health status and assess their well-being. 46 Measurements for health monitoring can include blood glucose, blood pressure, weight, 47 saturated blood oxygen (SpO2) and electrocardiogram (ECG). Social care monitoring can 48 include activity monitoring, bed and chair usage, falls alarms, and use of utilities. Telehealth 49 equipment can be configured in many ways, and the level and complexity of monitoring 50 technology can vary from a simple phone call to report data through to the use of 51 videoconferencing [2]. 52 The term telehealth is used interchangeably in this document with telemonitoring and remote 53 monitoring. We will refer exclusively to telehealth. Many advantages of telehealth are 54 reported in the literature: improved clinical outcomes, reduced number of unplanned hospital 55 and A&E admissions [3], efficient handling of caseloads, and facilitation of self-management 56 of condition by patients. A meta-analysis confirms these findings [2, 3]. 57 Exploiting technology to good effect proves complex because service re-design and clinical 58 engagement is found to be both expensive and challenging. High quality studies on cost are 59 few and inconclusive. In her review of studies on cost-effectiveness of telehealth, Mistry 60 (2012) reports that many studies lacked adequate methodological rigor and details of design 61 [4]. Mistry argues that this lowered the overall quality and further that most studies were pilot 62 in nature and presented findings on small sample size (100 patients or less) [4]. Polisena et al. 63 (2009) also conclude that although telehealth might hold the potential to reduce cost, the 64 quality of studies assessing and reporting telehealth cost were poor [5]. Whitten et al. (2002) in 65 a review of cost-effectiveness of telemedicine intervention similarly conclude that there was a 66 lack of evidence demonstrating cost-effectiveness [6]. A recent review on telehealth continues to 67 report that the evidence for cost effectiveness of telehealth is “weak and contradictory” [7]. 68 The primary concern is that outcomes are based on small pilot studies that run over short periods 69 of time (typically 3-6 months) [7], and with little consideration of the impact of services 70 introduced by the pilot. Outcomes are expected to be different when at scale over a prolonged 71 period and integrated into normal healthcare services. To establish whether this is true, the Whole 72 system Demonstrator project in the UK undertook a randomized trial with 6000 patients. The 73 conclusion is that a telehealth service in combination with standard care is not cost effective 74 when compared with standard usual care. However, only savings are published without any 75 comparison with the costs of the telehealth services. The authors recommend that future 76 research on cost-effectiveness of telehealth should be “targeted towards specific patient 77 population and sub population” and should explore the “association between area level 78 factors, patient characteristics (demographics, needs levels for each index condition), and the 79 variation in their service use and cost.” [8]. 80 Further studies [9] also conclude that additional home telemonitoring care is not cost effective 81 compared to usual care. This study [9] also reports variation in cost for the groups; there was 82 significant difference in costs pre and during the project; and many patients had no admission to 83 hospital during the period of the study. 84 In contrast, this paper reports on the cost-effectiveness of a telehealth programme when 85 applied to a wide population of patients with chronic obstructive pulmonary disease (COPD). 86 It takes into account the characteristics and structure of service delivery.
87 2 Methods
88 2.1 Programme Description 89 National Health Service Direct (NHSD) launched a pilot telehealth programme for patients 90 with COPD in two Primary Care Trusts (PCTs) (November 2010 in PCT1 and March 2011 in 91 PCT2) in order to evaluate cost-effectiveness of a service. The remote monitoring was 92 managed centrally through NHSD. 93 (PCTs were the administrative bodies in England responsible for commissioning primary, 94 community and secondary health services from providers. Until 31 May 2011 they also 95 provided community health services directly. Collectively PCTs were responsible for 96 spending around 80 per cent of the total NHS budget. Primary care trusts were abolished on 97 31 March 2013 as part of the Health and Social Care Act 2012, with their work taken over by 98 Clinical Commissioning Groups.) 99 The NHSD telehealth programme included: planning and administration of the programme; 100 developing operating policy and procedures and technical requirements; device 101 commissioning and management of the installation process; developing clinical process 102 workflow for the call centre; and reporting and management of data elements for evaluation. 103 The PCTs were responsible for: the operational and service elements of the programme at 104 their respective sites, and participation in planning and reporting of results for programme 105 evaluation. 106 Evaluation for the NHSD telehealth programme was designed by the joint research team of 107 Brunel University and Chorleywood Health Centre, with input from NHSD and the PCTs. 108 The evaluation was designed so that the majority of data from the patient was gathered 109 through the Health Buddy in order to: simplify implementation (no paper), reduce impact on 110 PCT staff workload, and make the project as close to normal service as possible. The PCTs 111 were selected on the basis of having well-established community teams to manage patients 112 with COPD. In this way we would be able to determine the impact of introducing the 113 technology and not the introduction of service. Ethical approval was gained from NHS 114 Research Ethics Committee. 115 The BOSCH® Health Buddy remote monitoring system was used for the programme. 116 Peripheral devices included temperature, weight, oxygen level, and blood pressure. It also 117 provided structured questionnaires on symptoms for respiratory distress, cough and sputum, 118 their severity and user acceptability for both patient and nurse professionals. Health education 119 messages were transmitted to the patient.
120 2.2 Population Identification 121 The evaluation was commissioned by NHSD to determine cost-effectiveness and other 122 outcomes of a telehealth programme. Two PCTs participated and a total of 321 patients were 123 enrolled into the monitoring programme. Of these, 227 were included for statistical analysis 124 using the criteria of having a complete baseline record of at least 88 days of continuous 125 remote monitoring. 126 The decision to refer a patient to the telemonitoring service was made by the PCT, and was 127 based on the inclusion criteria as in section Inclusion criteria.
128 2.2.1 Inclusion criteria for receiving the monitoring 129 Patients having a primary diagnosis of COPD and at least one of the following: 130 • Two or more admissions or emergency attendances in the 12 months 131 • Difficulty following a medication regime 132 • Difficulty following a treatment regime 133 Patients had to satisfy the following criteria: 134 • Have a telephone 135 • Reasonable cognitive ability that enables them to report observations 136 • Be able to understand English (to use the Health Buddy) 137 • Reasonable dexterity (to use the Health Buddy and peripherals) 138 • Not have a severe visual impairment
139 2.2.2 Inclusion in the statistical data set 140 • Completion of all baseline measures 141 • Three months (88 days) of continuous remote monitoring 142 2.3 Patient Management 143 Each PCT was responsible for referring a patient to the telemonitoring service and how the 144 patient was managed within the service. Each patient was provided with the Health Buddy 145 unit and a finger pulse oximeter and instructed to take one SpO2 measurement each day. In 146 the same measurement session, the Health Buddy would ask questions relating to the 147 symptoms of COPD, such as the type of cough and sputum production. Devices to monitor 148 other vital signs were provided according to the needs of the patient. Data was reviewed daily 149 by the nurse and was used to assist decisions on the management of the patient.
150 2.4 Resource and Cost Utilisation 151 Resource utilisation data was obtained from multiple sources and include A&E visits, 152 ambulance usage, and hospitalisation. Cost reports were submitted by NHSD, PCT1 and 153 PCT2. 154 For comparative purposes, each patient acted as their own control. To avoid the effects of 155 seasonal variation and expected deterioration in health, two control periods were compared. 156 For each patient we selected: an equal period of monitoring taken to start 12 months prior to 157 monitoring and so cover the same time of the year (seasonal); and an equal period taken to 158 end immediately before the start of the remote monitoring (deterioration). The levels of 159 service usage for each patient were found for both these periods and the monitoring period. 160 For the comprehensive cost analysis, pertinent information was defined as useful by the 161 project partners. Two templates were made for the project; the first recorded data about 162 patient usage of health services other than home monitoring, and the second recorded data 163 about the telemonitoring service. These were completed by the PCTs. Analysis of resource 164 usage was separated into two recognisable groups defined by clinical event outcomes, all- 165 cause and COPD specific clinical events. This was based on coding whenever possible. 166 Identifying trends and changes attributable to remote monitoring was to be aided by 167 comparing the service user data of high cost users (as defined by a-priori admission to 168 hospital) before and during monitoring. Two data collections were made, one midway 169 through the project to verify methods and data quality and at the end for final analysis. 170 Resource utilisation and costs included those to establish the programme (training, planning 171 meetings) and for on-going running (administration).
172 2.4.1 Health resource utilisation 173 Analysis of health resource utilisation was performed by using each patient as their own 174 control and comparing metrics before and during monitoring. In order to compensate for 175 seasonal effects and because COPD is a degenerative disease, data was taken for an epoch of 176 equal duration 12 months before the monitoring period and immediately before monitoring.
177 2.4.2 Healthcare system cost 178 The telehealth programme targeted a population of patients with known high cost and 179 likelihood of future high cost. The eligibility criteria were designed to select those having 180 high resource utilisation and consequent cost. The purpose of the investigation was to 181 compare the savings in resource with the cost of the programme. 182 Cost data included the initial cost for start-up (attending planning meetings, staff training), 183 on-going management of the programme, and increased personnel time to examine data and 184 visit patients. We assumed a monthly service charge of £100 ($125) per patient was made by 185 NHS Direct to provide the equipment, manage the data and provide the service. The cost also 186 included installation and decommissioning. 187 2.4.3 Healthcare system cost-effectiveness 188 For this study we chose to undertake define cost-effectiveness as a comparison between the 189 costs to provide the remote monitoring service and the direct savings that in health cost that 190 could be attributed to to the introduction of the service. Service costs included equipment, 191 remote monitor service costs and additional staff resources; direct savings were attributed to 192 reduction in utilization of health service resource including hospitalization avoided, reduced 193 in-hospital stay, and reduced contacts with health services. 194 Our data did not include quality of life, and so we were unable to undertake cost-benefit 195 analysis.
196 3 Results
197 3.1 Baseline data 198 Based on the inclusion criteria, 107 patients from PCT1 and 120 patients from PCT2 were 199 included in the statistical data set. The average length of monitoring in PCT1 was 203 days 200 and in PCT2 was 215 days, giving a total of 21,738 and 25,764 monitoring days for each PCT 201 respectively. The mean age of patients on entry to monitoring was 70.9 (s.d. 8.9) with 112 202 male, and 112 female participants, and was the same for each PCT. Data on gender was not 203 provided for 3 patients. 204 Health service utilisation data used in this study was provided by the PCTs. Resource usage 205 for hospitalisation and A&E visits was separated by cause of admission (COPD, all other 206 causes). 207 The study was designed to evaluate service, rather than a controlled trial, and patients were 208 enrolled and withdrawn based on clinical judgement. This resulted in wide variation length of 209 participation. We chose the minimum period for inclusion in the data set in order to be of 210 sufficient duration that we might capture clinical events and was comparable to the period of 211 monitoring in other home monitoring services, such as used for hospital discharge. This 212 period is typically 90 days, but inspection of the data showed that a choice of 88 days of 213 continuous remote monitoring would include several further patients.
214 3.2 Initial and Start-up Costs 215 Each PCT provided detailed costs associated with establishing the programme. This included 216 a breakdown of the impact on all staff involved, together with all aspects such as training, 217 attending meetings, planning meetings, additional administration and travel. These costs are 218 reported in Error: Reference source not found. 219 There were differences in approaches and costs between the PCTs. PCT 1 allocated all COPD 220 patients to a member of its long term conditions team, whereas PCT 2 allocated the less 221 severe patients to its community matrons and the more severe patients to the respiratory team. 222 PCT 1 also allocated significantly greater time to training than PCT 2. It also appears that 223 they appointed dedicated administration support. As a consequence, PCT 1 reported 224 significantly higher start-up costs. 225 226 Table 1 Resource Utilisation and Cost Setup Running Internal Salary External Meeting Plannin Total Cost+30 Meeting Patient Total Salary Training s g Admin Total Salary % Travel Travel Cost (£/hour) (Hours) (Hours) (Hours) (Hours) Hours (£) (£) (£) (£) (£) PCT 1 Nurse staffing Long term Conditio ns Team 29.05 1506.4 8A 1.00 31.50 6.00 0.00 38.50 1118.43 1453.96 6.56 45.92 4 25.73 10737.1 11448. 7 100.00 101.50 4.00 115.50 321.00 8259.33 3 13.76 697.16 05 6 18.78 31.00 0.00 0.00 0.00 31.00 582.18 756.83 0 0 756.83 5 15.59 18.00 0.00 0.00 0.00 18.00 280.62 364.81 0 0 364.81 12.14 33355.2 33355. 4 1.00 0.00 0.00 2112.50 2113.50 25657.89 6 0 0 26 Total for PCT 1 Respirat ory 47431. Team 46667.99 39
PCT 2 Nurse staffing Respirat ory Team 1890.4 8C 40.66 0.00 30.50 5.00 0.00 35.50 1443.43 1876.46 0 14 6 8B 34.39 0.00 16.25 2.00 0.00 18.25 627.62 815.91 0 0 815.91 28.85 4727.8 8A 2.00 73.25 44.15 4.50 123.90 3574.52 4646.88 0 81 8 7 23.39 34.75 7.00 50.00 120.30 212.05 4959.85 6447.81 0 45 6492.8 1 19.59 3809.4 6 32.75 5.50 11.00 91.50 140.75 2757.29 3584.48 0 225 8 4 13.42 0.00 0.00 0.00 1.00 1.00 13.42 17.45 0 27 44.45 11.54 2617.8 3 7.00 4.00 0.00 163.50 174.50 2013.73 2617.85 0 0 5 Total for PCT 2 Respirat ory 20398. Team 20006.84 84 Commu nity Matron 8C 40.66 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0 0.00 8B 34.39 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0 0.00 28.85 6236.4 8A 4.00 26.00 15.00 97.50 142.50 4111.13 5344.47 0 892 7 7 23.39 1.00 0.00 0.00 0.00 1.00 23.39 30.41 0 0 30.41 6 19.59 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0 0.00 4 13.42 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0 0 0.00 11.54 1464.1 3 4.00 4.00 0.00 89.00 97.00 1119.38 1455.19 0 9 9 Total for PCT 2 Commu nity 7731.0 Matron 6830.07 7
Total for all PCT 2 staff 26836.91 28129.91
Total for all staff 73504.90 75561.30 227
228 3.3 Comparison of Outcomes 229 We determined that inpatient stays were the single dominant cost for health resource 230 utilisation. Changes seen in A&E and outpatient usage were small and thus savings were 231 small in comparison. Cost analysis was only performed for inpatient stays.
232 3.4 Health Resource Utilisation - Inpatient Stays 233 234 Table 2 Overall health resource utilisation 12 Months Before Immediately Before During Change 12 Months Before Change Immediately Before COPD Other COPD Other COPD Other COPD % Other % COPD % Other % A&E 3 80 8 112 7 99 4 133 19 24 -1 -13 -13 -12 Outpatients 144 78 149 86 122 83 -22 -15 5 6 -27 -18 -3 -3 Inpatient all Stays 64 43 100 52 71 39 7 11 -4 -9 -29 -29 -13 -25 Bed days 603 209 581 329 398 195 -205 -34 -14 -7 -183 -31 -134 -41 Cost 385920 133760 371840 210560 254720 124800 -131200 -34 -8960 -7 -117120 -31 -85760 -41
Average stay 9.4 4.9 5.8 6.3 5.6 5 -3.8 -40 0.1 2 -0.2 -3 -1.3 -21 235 236 Total resource utilisation data (Table 2) indicates a decrease in total inpatient bed days during 237 the monitoring period when compared with both the period 12 months before (34%) and 238 immediately before monitoring (31%). This results in savings of £131,200 ($164,000) and 239 £117,120 ($146700) respectively. There is a difference in the cause of the savings made. The 240 savings made during monitoring when compared with data from 12 months before are due 241 largely to reduced average length of stay (-40%) despite a comparative increase in the 242 number of stays (11%). Comparison of hospital usage data during monitoring and 243 immediately before monitoring shows a reduction in the number of admissions (-29%) with a 244 negligible reduction in length of stay (-3%). 245 Table 3 Health resource utilisation PCT1 12 Months Before Immediately Before During Change 12 Months Before Change Immediately Before COPD Other COPD Other COPD Other COPD % Other % COPD % Other % A&E 3 54 8 63 7 51 4 133 -3 -6 -1 -13 -12 -19 Outpatients 144 65 149 68 122 66 -22 -15 1 2 -27 -18 -2 -3 Inpatient all Stays 31 18 38 25 21 17 -10 -32 -1 -6 -17 -45 -8 -32 Bed days 356 107 258 206 136 83 -220 -62 -24 -22 -122 -47 -123 -60 Cost 227840 68480 165120 131840 87040 53120 -140800 -62 -15360 -22 -78080 -47 -78720 -60
Average stay 11.5 5.9 6.8 8.2 6.5 4.9 -5 -43 -1 -17 -0.3 -4 -3.3 -40 246
247 The PCT1 data (Table 3) shows a significant decrease in total inpatient bed days in the 248 monitoring period when compared with both the period 12 months before (62%) and 249 immediately before monitoring (47%). This resulted in savings of £140,800 ($176,000) and 250 £78,080 ($97,800) respectively. The savings made during monitoring when compared with 251 data from 12 months before are due largely to reduced average length of stay (-43%) along 252 with a reduction in the number of stays (-32%). Comparison of hospital usage data during 253 monitoring and immediately before monitoring shows a reduction in the number of 254 admissions (-45%) and a very small reduction in the length of stay (-4%). The data for ‘all 255 other’ clinical causes for admissions showed a reduction. We also observed reductions for in- 256 patient bed days of 22% and 60% respectively. 257 Table 4 Health resource utilisation PCT2 12 Months Before Immediately Before During Change 12 Months Before Change Immediately Before COPD Other COPD Other COPD Other COPD % Other % COPD % Other % A&E 26 49 48 22 85 -1 -2 Outpatients 13 18 17 4 31 -1 -6 Inpatient all Stays 33 25 62 27 50 22 17 52 -3 -12 -12 -19 -5 -19 Bed days 247 102 323 123 262 112 15 6 10 10 -61 -19 -11 -9 Cost 158080 65280 206720 78720 167680 71680 9600 6 6400 10 -39040 -19 -7040 -9
Average stay 7.5 4.1 5.2 4.6 5.2 5.1 -2.3 -31 1 24 0 0 0.5 11 258 259 The PCT2 data (Table 4) shows that during the monitoring period there was an increase in the 260 number of in-patient bed days when compared with the period 12 months before monitoring 261 (6%) but a decrease for the period immediately before monitoring (-19%). This resulted in an 262 increase in cost of £9,600 ($12,000) and savings of £39,040 ($48,900) respectively. The 263 increased spend seen when the monitoring period data is compared with period 12 months 264 before monitoring may be explained by the 50% relative increase in the number of 265 admissions although its’ impact was perhaps eased by a reduction in length of stay (31%). 266 For “all other causes”, there was a reduction in patient bed days during the monitoring period 267 when compared to the period 12 months before (9%) but an increase (10%) when compared 268 to the period immediately before monitoring.
269 3.5 Other Outcomes 270 The cause of A&E visits in PCT2 could not be found in the data and it was not possible to 271 count visits caused by either COPD or ‘all other causes’. The number of visits due to COPD 272 in the PCT1 data was too small to provide statistically significant changes. Outpatient 273 appointments had a consistent reduction of about 15% for PCT1 but the change was 274 inconsistent for PCT2. Numbers were too small to have significant impact on the cost when 275 compared to inpatient stays.
276 3.6 Costs of the Service 277 Each of the PCTs was asked to provide details of direct costs that could be attributed to 278 setting up and running the service. This included staff time spent on planning meetings, travel 279 to planning meetings, administration, and education sessions. PCT1 estimated costs to be 280 £47,000 ($58,900) and PCT2 estimated costs as £28,000 ($35,000). A separate estimate of 281 on-going costs of providing the service (administration) is shown in table 1.
282 NHS Direct assumed a cost of £100 ($125) per patient per month to provide the service (or 283 £3.29 ($4.12) per day). PCT1 ran the service for 16.9 months for 107 patients, and accrued a 284 cost of £180,000 ($225,600) to provide the monitoring service for the duration of the project. 285 PCT2 ran the service for 17.9 months for 120 patients, and accrued a cost of £215,000 286 ($269400) to provide the monitoring service for the duration of the project. 287 4 Discussion
288 4.1 Overall Outcomes 289 The project aim was to determine the cost effectiveness of using a telehealth service delivery 290 model to manage patients with severe COPD. Our project includes two separate pilot sites 291 and by taking the prior monitoring period from two separate periods into account, we have 292 created four independent data sets for comparison. 293 The most striking observation is the significant difference in outcome from the data sets, with 294 high variance both between the PCTs and between the comparison periods. This result would 295 support the observation of poor correlation that is reported in meta-analysis of cost 296 effectiveness studies. Within our data we see outcome ranging from a saving of £140800 297 ($176,500) to an increase of £9600 ($12,000). The cause of this variation may be related to 298 the process utilised within the PCT and additional research on the impact of other variables 299 that influence patient outcomes is needed.
300 4.2 Cost-effectiveness 301 PCT1 made a saving of £140,800 ($176,500) at a cost of £180,000 ($225,600). PCT2 made 302 a saving of £39,040 ($48,900) at a cost of £215,000 ($269,500). The saving for each patient 303 for a year in PCT1 was £2,366 ($2965) (12 months before) and £1,312 ($1645) (immediately 304 before). In PCT2 a loss occurred of £136 ($170) (12 months before) and a saving of £552 305 ($690) (immediately before). 306 There are a number of factors that would affect cost-effectiveness. No allowance was made 307 for the re-use of equipment and the actual number of monitoring devices used was not 308 reported. For example, if equipment could be hired for only each day it was used for 309 monitoring the monitoring cost for PCT1 would then have been £71,518 ($89,600) rather 310 than £180,000 ($225,600) and for PCT2 £84,760 ($106,200) rather than £215,000 311 ($269,400)). 312 The reported costs to establish and run the monitoring service were additional to the usual 313 cost of providing a COPD service in the community and were particular to the project. The 314 actual cost of providing a telehealth programme that was integrated fully into usual service 315 provision would be less than that that was reported. 150 of the patients, approximately, were 316 never admitted to hospital and therefore would not have resulted in savings and would skew 317 the findings of costs and savings. Overall implementation and administrative strategies as 318 well as deployment factors may have influenced the cost differences noted between the 2 319 PCTs.
320 4.3 Impact on Hospital Admissions 321 Despite the possibility that the admissions policies of each acute trust had changed over the 322 twelve months before monitoring, the telehealth programme appeared to have an effect on the 323 number of admissions and the clinical outcomes. It can be argued that the latter is shown by 324 the finding of a fall in the average length of in-patient care: in PCT1 from 11.5 days in the 325 period 12 months before monitoring to 6.5 days during monitoring, and similarly in PCT2 326 from 7.5 days to 5.2 days. There is no evidence to explain these changes. It had not been 327 possible to conduct a study of the patient care pathways that might have shown the impact of 328 home monitoring on either patient or professional carer education or an earlier recognition of 329 exacerbations leading to more prompt hospital admission. If the latter was the case, it can be 330 assumed that in-patient management would be more effective allowing earlier hospital 331 discharge. In addition, the professional skills and clinical effectiveness of the community 332 team may have developed over the year before monitoring and that there was more effective 333 patient management irrespective of monitoring. 334 In order to understand how the length of stay was being affected by the monitoring we 335 considered the profile of the length of stay for each of the periods of interest; 12 months 336 before, immediately before, and during monitoring. Figure 1 shows that many of the short 337 stays of 3-5 days were reduced to 2 days, and the number of stays of 12-15 days was reduced. 338 We assume that prompt admission may result in a much reduced stay and having the 339 technology in the home gave confidence for early discharge. Figure 1 also shows that many 340 patients were not admitted to hospital during the periods of interest (0 days). 341
342 Figure 1: Length of hospital stay for the periods 12 months prior, immediately prior and 343 during monitoring. 344 345 Figure 2 depicts the change in the total length of stay for each patient, and shows the 346 difference between the total length of stays immediately prior to monitoring and the total 347 length of stays during monitoring. The difference for each patient is shown as a histogram of 348 the change (note that the change for 0 days is suppressed). The greatest peak occurs for a 349 decrease of 2 days, and many of the lengthy stays (normally arising from complications) are 350 eliminated. We may also propose that by having monitoring available, patients are being 351 released early. 352
353 Figure 2: Change in length of stay for each patient compared to length of stay in the period 354 immediately before monitoring
355 4.4 Limitations of this Study 356 The study has a number of recognisable limitations that must affect the results of the study. 357 Although the average length of monitoring was 200 days, there was a wide variation in 358 monitoring times with a minimum monitoring period of 88 days. This will have impacted on 359 capturing health resource usage. Illness will wax and wane and resolves when disease is 360 cured or causes death. COPD is a chronic long term condition that is progressive and is 361 affected by seasonal change and the well-being of patients can be changed markedly by other 362 co-morbidities and social factors. The study design attempts to compensate for this by 363 comparing data from the monitoring period with both the period 12 months before and the 364 period immediately before monitoring. There was anecdotal evidence from the research team 365 of professional learning and team development in both PCTs. As the project progressed, it is 366 believed that the nurses developed better management of their patients. This alone may have 367 had an impact on their recognition and treatment of a sudden deterioration in a patient’s 368 condition and the need for admission and leading to greater clinical effectiveness. For 369 example, PCT1 showed a saving of £60,000 ($75,000) without any monitoring. In this study 370 the telemedicine service was added as an ‘additional’ layer of service making it difficult to 371 discern which element of the combined service had the greatest impact. Others have reported 372 large gains in terms of cost savings and improved clinical outcomes with telemedicine in less 373 well developed community health services. 374 Analysis of health resource was restricted to the data from the patient administration system 375 (PAS). A fuller analysis with data from the patient’s clinical record could have revealed 376 greater detail about the causes of admission, reasons for extended stay and clinical outcome. 377 The project was to have evaluated cost benefit by applying SF-12v2 to all patients at baseline 378 and at period intervals. However a system error resulted in the questionnaire being given to 379 only 14 patients. Analysis of this small set of data showed no significant change in health. 380 We made no attempt to evaluate the project from a user perspective, and so we were unable 381 to determine how patient management strategy had changed as a result of monitoring being 382 introduced.
383 4.5 Implications Derived from this Study 384 The study provides a number insights and lessons to be learned. Future evaluations of cost- 385 effectiveness, and clinical effectiveness for that matter, of remote patient monitoring have to 386 be different because such evaluations seen in this study and in others are highly variable, and 387 thus unreliable. Despite the fact that each locality based clinical service provides a service to 388 achieve the same clinical goal, they do so in significantly different ways, with varying levels 389 of investment and skill. The introduction of remote monitoring itself has a profound effect on 390 team learning and clinical practice and thus distorts the cost effectiveness evaluation of the 391 use of the technology. 392 There was a wide discrepancy in the results from the two PCTs. This places doubt on 393 outcomes and may indicate also why the literature on cost-effectiveness remains 394 inconclusive. Others [9] have similarly reported that statistical parameters for the respective 395 cohorts are not stationary. 396 Moreover, to gain a true assessment of the cost-effectiveness remote patient monitoring, the 397 effect of the service must be considered in how the experiment is established, and how 398 outcomes are interpreted, in order that the effect of the technology alone is assessed. 399 The analysis of length of stay shows a marked decrease that could be attributed to the early 400 discharge of patients to a home where monitoring is available. This suggests a need for 401 systems to be deployed rapidly with the patient on discharge. 402 The wide variation in nce on savings and the uncertainty of monitoring cost does not allow a 403 definitive conclusion on the cost effectiveness as an outcome of this study. It might well be 404 that the average saving was £1023 ($1280) per patient per year but the variationnce is too 405 great to allow this to be statistically significant. 406 The authors conclude that cost-effectiveness studies will continue to struggle to provide a 407 definitive answer because outcome measurements are too dependent on factors other than the 408 technology. This question could be answered by developing standardised measures of clinical 409 and organisational activity that can be considered in addition to basic costs and thereby 410 minimise the effect of cultural, educational and financial differences between health care 411 services on the outcomes of studies.
412 Acknowledgements
413 This research was part funded by NHS Direct.
414 References
415 1. Department of Health. Improving the health and well-being of people with long term 416 conditions, 2010. 417 2. Clarke M, Shah S, and Sharma, U. Systematic review of studies on telemonitoring of 418 patients with congestive heart failure: a meta-analysis. J Telemed Telecare 2012; 17:7–14 419 3. Cleland J, Louis A, MD, Rigby A, Janssens U, Balk A. Noninvasive Home Telemonitoring 420 for Patients With Heart Failure at High Risk of Recurrent Admission and Death The Trans- 421 European Network–Home-Care Management System (TEN-HMS) Study. Journal of the 422 American College of Cardiology. 2005;45:1654–64. 423 4. Mistry H. Systematic review of studies of the cost-effectiveness of telemedicine and 424 telecare. Changes in the economic evidence over twenty years. J Telemed Telecare 425 2012; 18 doi: 10.1258/jtt.2011.1105055. 426 5. Polisena J, Coyle D, Coyle K, McGill S. Home telehealth for chronic disease management: 427 A systematic review and an analysis of economic evaluations. Int J Technol Assess 428 2009;25:339–349. 429 6. Whitten SP, Mair SF, Haycox A, May RC, Williams LT, Hellmich S. Systematic review of 430 cost effectiveness studies of telemedicine interventions. Br Med J 2002; 324:1434-1437. 431 7. Wootton, R. Twenty years of telemedicine in chronic disease management - an evidence 432 synthesis. J Telemed Telecare 2012;18 doi:10.1258/jtt.2012.120219. 433 8. Henderson C, Knapp M, Fernández JL, Beecham J, Hirani SP, Cartwright M, Rixon L, 434 Beynon M, Rogers A, Bower P, Doll H, Fitzpatrick R, Steventon A, Bardsley M, Hendy J, 435 Newman SP: Cost effectiveness of telehealth for patients with long term conditions (Whole 436 Systems Demonstrator telehealth questionnaire study): nested economic evaluation in a 437 pragmatic, cluster randomised controlled trial. Br Med J 2013;346:f1035 doi: 438 10.1136/bmj.f1035. 439 9. Upatising B, Wood DL, Kremers WK, Christ SL, Yih Y, Hanson GJ, Takahashi PY. Cost 440 Comparison Between Home Telemonitoring and Usual Care of Older Adults: A Randomized 441 Trial (Tele-ERA). Telemedicine and e-Health 2015;21:3-8. 442