THE EVIDENCE FOR THE ECONOMIC VALUE OF eHEALTH IN THE UNITED STATES TODAY: A SYSTEMATIC REVIEW

Leslie Wilson PhD, Ashley Kim, David Szeto

Department of Clinical Pharmacy, University of California, San Francisco, USA


Abstract

The United States healthcare system continues to face increasing costs, with year-over-year projected cost increases that now exceed the rate of increase of the gross domestic product. The rapid expansion and integration of eHealth within the United States healthcare system is driven primarily by a perceived ability to increase access in a cost-efficient manner. However, there is little economic research that addresses the large diversity of eHealth products being integrated into this healthcare landscape. The goal of this study is to evaluate the published economic evidence for eHealth in the United States, analyse how well it supports the growth of the current eHealth environment, and suggest what evidence is needed. This systematic literature review, conducted through the PubMed and The Cochrane Library databases, found that few studies addressed today’s eHealth environment. The current landscape is broader and less tailored to the traditional telemedicine initiatives represented by existing studies. We suggest more rigour in the design and scope of economic studies and that current eHealth technologies be identified for analysis. These studies must be comprehensive from the healthcare system’s perspective and conducted for relevant initiatives and patient groups to allow for evidence-based discussions on the cost-effectiveness of eHealth.

Keywords: eHealth; telemedicine; cost-effectiveness; health economics; health outcomes

J Int Soc Telemed eHealth 2016;4:e21

Introduction

New mobile and online interactive services are transforming the landscape of healthcare with the goals of reducing costs and improving access for patients, doctors, and healthcare organizations. Software and technologies built upon existing and new communication infrastructures, such as email and video conferencing, continue to allow healthcare stakeholders to interact in increasingly novel ways. For example, patients can take a picture of a new skin rash post-treatment, upload it, and send it to their doctors at any time of the day. But also being developed are 24/7 virtual worlds such as the VA Virtual Medical Center which is completely virtual and in beta testing.  However, there is little hard evidence whether any of these approaches are actually bending the healthcare cost curve. Currently, the healthcare sector continues to produce unsustainable trends in cost. The National Health Expenditures increased by 5.3 percent in 2014 to $3.0 trillion, which accounted for 17.5% of the Gross Domestic Product (GDP) in the United States. More alarmingly, however, is the projection that expenditure will grow at 5.8 percent per year, 1.3 percentage points faster than the GDP  placing the health share of GDP at 20.1 percent by 2025.1

We aim to review the economic evidence that supports the use of  eHealth in the United States. We use the term eHealth to refer to a broad set of health services and information delivery technologies leveraged to improve health care.2 eHealth as used here is meant to encompasses all levels of integration, ranging from single isolated technologies to complete virtual systems of healthcare. To our knowledge, no one has yet reviewed the recent economic evidence. Therefore, we conducted a systematic review to evaluate the availability of economic evidence for all types of eHealth in the U.S. to better understand the support and validation being given for the growth and direction of this field.

Methods

We searched PubMed and The Cochrane Library databases using MeSH terms from January 1, 2010 to July 3, 2016 across all years and languages. The comprehensive search strategy and systematic review are adapted from the PRISMA Statement and can be found in Appendix A.3 This process identified 20 final studies: five randomised controlled trials (RCTs), 14 different types of cohort studies, and one case series study, as seen in Figure 1.

Figure 1
Figure 1. Flow diagram for study identification adapted from PRISMA.

For each study, we described the type of economic evaluation, quality, risk of bias, area of eHealth evaluated, and relevance to the current landscape. The gold standard economic study is one that measures both costs and outcomes, such as a cost-effectiveness analysis (CEA) or cost-benefit study, from a societal perspective.4 Cost-consequence analyses (a type of CEA) measure costs and vary the outcomes to provide evidence of when the optimal balance between cost and outcome is reached. Inferior economic studies only examine costs or incomplete costs, without providing evidence of the effects on outcomes. All types of economic evaluations and perspectives were represented in our review, from simple cost discussions to comprehensive cost-effectiveness Markov models for a health system. The quality of the studies was reviewed using the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist by the ISPOR Task Force Report as a guide.5 This review can be found in Appendix B. The risk of bias and validity of each study was evaluated and assessed independently by two reviewers, using the Cochrane Risk of Bias Tool, and results can be found in Appendix C.6 The assessment of an “unclear” risk of bias was largely due to lack of transparency in modelling studies. We found that the risk of bias did vary across studies, but was relatively low overall.

Results and Discussion

We discuss our findings  by grouping them into two common goals of eHealth care delivery that they represent. The first is patient access and convenience and the second is prevention and continuity of care. Many of the economic studies focus only on single item costs, such as time off work or transportation, rather than total savings important to the healthcare system itself.

Patient Access and Convenience: Rural and Specialty Care

A key obstacle in healthcare is the lack of accessibility to care at the optimal time and price, particularly in rural areas. Our review covered 12 economic studies that focus on improving patient access and convenience: three single item cost assessments,7-9 three full cost assessments,10-12 and six cost-effectiveness studies.13-18

Single item cost assessments

Urquhart et al. (2011) used a retrospective cohort design to compare in-person to remote video telemedicine for post-operative visits post-parathyroidectomy, demonstrating that a remote video telemedicine consultation with a specialist could save an average of $357 per patient in transportation costs.7 A similar retrospective cohort study by Canon et al. (2014) found that surgical paediatric patients from rural Arkansas saved $88 per trip in travel costs when their post-operative visit was performed in a teleclinic with an on-site nurse and off-site physician.8 Their incision was examined by the remote physician using a video camera or by digital images sent by the on-site nurse. A third study aimed to address the steep start-up costs of incorporating telehealth networks by examining the incorporation of an iPhone 4 into a real-time telestroke network.9 This study compared time spent for physicians using an iPhone to perform a remote examination or conduct a scored bedtime risk assessment with a remote telehealth assessment. Results were congruent between the remote and bedside physicians. While methods were taken to prove inter-rater reliability, generalisability was limited by a small sample size (n=20). A common limitation between these three studies was the lack of comprehensive cost data and outcomes, and therefore inadequate evidence for economic conclusions.

Full Cost Assessments

A cohort study by Kim et al. (2014) compared tele-assessment of long-term home parenteral nutrition with mobile tablets to view the home infusion process for bowel disease.10 Researchers measured a comprehensive set of costs for providing mobile distance clinic appointments. Total initial appointment setup costs were $916.64 per patient and included equipment ($590), supplies ($82), delivery ($54), personnel ($190 per hour), and follow-up visits ($362 per month). While initial and maintenance costs were expensive, the intervention showed transportation cost savings for the patient. However, savings were not compared to existing costs of care and patient outcomes were only evaluated via a statement of patient satisfaction.

Another retrospective cost assessment by Morland et al. (2013) of an RCT compared two methods for treating post-traumatic stress disorder for veterans living in rural islands in Hawaii.11 They compared usual care where psychiatrists were flown to the islands for consultations, to teleconferencing for anger management therapy. Comparable outcomes were demonstrated using clinical anger scales. Both intervention costs, such as intervention equipment and maintenance costs, and sunk costs, such as Internet bandwidth fees, were included. The intervention costs were calculated to account for average use over time as a per-minute cost. This study highlighted a cost savings of $713 per teleconference consultation and led clinics to end their usual care practice.

A third pre-post study by Franzini et al. (2010) is a more robust cost analysis of a tele-ICU programme comparing the constant availability of intensivists to the usual care that involved less intensivist care, from the hospital’s perspective. They calculated the change in costs per case after adoption of the new intensivist telehealth intervention. The telehealth costs increased 16% for floor daily average costs and 48% for overall ICU costs per case.12 Daily costs by revenue centres, hourly provider fees, fees for hardware, software, installation, and operation of all tele-ICU equipment, and monthly per-bed fees were all considered. They also measured the change in hospital mortality across the intervention and found a significant 11.4% decrease, however, stopped short of calculating cost-effectiveness.

These cost assessment studies are specific to either a rural location or specialty care, and although they do not demonstrate cost-effectiveness or value, they do contribute to eHealth savings.

Cost-effectiveness Studies

Two studies testing efficiency in rural and specialty care conducted a full cost-effectiveness study. Hitt et al. (2013) studied the ability of telecolposcopy to improve accessibility to cervical cancer screening for women in rural Arkansas.13 Medicaid treatment coverage was available for colposcopic evaluation, but there were no specialists to perform the colposcopy and its follow-up. They compared usual care without a specialist to the intervention, which paired local examiners with remote experts in a hub-and-spoke arrangement via interactive video services. Based on an analysis of salaries and hours utilised for these two approaches, the authors found telecolposcopy yielded comparable positive predictive values and sensitivities. Expected survival was modelled and resulted in a cost-effective outcome of $11,830 per quality-adjusted life-year (QALY).13

A complete cost-effectiveness study by Pyne et al. (2010) based on a Markov model using a large RCT compared an interactive televideo mental health visit in seven rural VA outpatient clinics with usual primary care, which was not specifically defined.14 Only the primary care physician was on-site, and the rest of the televideo team, consisting of a telepsychiatrist, depression care manager, clinical pharmacist, and supervising psychiatrists, was off-site. Use of televideo enhanced off-site team care by improving adherence to antidepressant medications and demonstrating depression response and remission. Costs included outpatient, pharmacy, and intervention expenses, and the effectiveness outcomes were depression-free days. The time horizon was only 12 months and assumed equal survival in that period. QALYs were based on the health survey for veterans with a standard gamble conversion formula. The cost-effectiveness of the intervention was $72,636 to $144,990 per QALY depending on intervention costs, which were primarily driven by care team salaries and travel costs. Sensitivity analyses were performed, and acceptability curves demonstrated a willingness to pay threshold of $100,000 per QALY, demonstrating that the intervention was cost-effective 65% of the time.

Telestroke networks have become more prevalent recently due to the increasing need to provide timely diagnosis and care for patients in an environment with inadequate resources and expertise, as shown in two cost-effective studies in our review. First, a retrospective cost-effectiveness analysis (CEA) using a lifetime Markov model reported, from a societal perspective, a 1-to-7 hub-and-spoke telestroke network in rural areas to be dominantly cost-effective compared to care with no network. Data for model inputs were from Georgia Health Sciences University and Mayo Clinic telestroke networks.15 Costs included telemedicine setup and maintenance, initial and recurrent stroke treatment, rehabilitation, long-term care, and caregiver costs. Effectiveness was measured in QALYs and utility was measured by EuroQol. This study by Demaerschalk et al. (2013) found an incremental effectiveness of 0.002 QALYs in the one-year time horizon and 0.02 QALYs in the lifetime horizon and concluded that the up-front investment in telehealth for stroke was justified for their health system.15 A related study by Switzer et al. (2012) examined the same hub-and-spoke telestroke network data as Demaerschalk et al. (2013), but lacked calculation of a true cost-effectiveness ratio.16 The study devised a model that compared the costs and effectiveness separately over a five-year time frame, from three perspectives: a network, a hub hospital, and a spoke hospital. The model described a savings of $358,435. Each spoke showed a $109,080 savings while each hub showed a cost of $44,804 annually, leading them to suggest cost sharing arrangements so that all telestroke network players can enjoy savings.16

Another decision analytic model was built to assess the cost-effectiveness of two-way, audiovisual hub-and-spoke telestroke care from a stroke expert compared with usual stroke care in a remote emergency department.17 Nelson et al. (2011) took a societal perspective in the lifetime horizon cost-effectiveness model. Cost inputs were sourced from the literature and included equipment, expert personnel time, hospital, and caregiver costs. Additional model inputs were from the University of Utah and Stroke Telemedicine for Arizona Rural Residents telehealth networks. They found that emergency department telestroke resulted in an incremental cost-effectiveness ratio (ICER) of $2,449 per QALY. Their analysis demonstrated that for a $50,000 willingness-to-pay threshold, there was a 99.8% certainty of cost-effectiveness over a lifetime. Similarly, Yoo et al. (2011) also designed a decision analysis model and examined the cost-effectiveness of telemedicine equipment in the ICU, from a healthcare system perspective.18 They considered costs at the systems level, with cost inputs including per-patient per-hospital stay cost while, and after, being in the ICU. Most model inputs were based on previously published U.S. literature on patients and personnel. The base case CEA estimated telemedicine in the ICU to extend 0.011 QALYs per patient with incremental cost of $516, when compared to ICU without telemedicine. However, the probabilistic CEA estimated an ICER of $50,265 per QALY, with a 95% confidence interval of 1,000 ICER estimates ranging widely from -$229,016 to $375,870. The relatively wide CI may suggest a wide range of estimate cost-effectiveness with programmes having variable impacts on costs and outcomes, but it also projects a cost savings of 37% per 1,000 iterations. This analysis model therefore indicates that telemedicine in the ICU is cost-effective in most cases using a broad range of assumptions.

Many economic evaluations focus on identifying the efficiencies of using telehealth in areas with minimal specialist care. Approximately 46.2 million people lived in nonmetropolitan counties in 2014, which constitutes 15% of the entire population of the U.S.19 Although small in scope, our reviewed studies begin to suggest cost savings in these areas and hint at the potential of eHealth to expand much further, but further economic evaluations need to be done.

Prevention and Continuity of Care in Disease Management

The remaining reviewed studies focus on the potential to improve disease prevention and longitudinal care management. Our review found six cost-consequence analyses20-25 and two cost-effectiveness studiesin this area.26,27

Cost and Cost-consequence Studies

A common microvascular complication of diabetes is diabetic retinopathy. Due to its asymptomatic nature, early detection is highly recommended; however, adherence to screening is low. Brady et al. (2014) presented a model that examined whether tele-ophthalmology screenings in adult diabetic patients for routine care at an urban internal medicine practice are cost-effective and cost-saving.20 These screenings were done via fundus camera photographs and sent to outside expert readers. After expert recommendation, the original physician decided whether to refer patients for subspecialty care. Using a decision tree and Medicare fee schedule-based costs, they compared photo-based screening versus no photo-based screening for ability to diagnose accurately. They demonstrated cost savings of $48 per patient for the tele-ophthalmology screening, but did not calculate cost-effectiveness. Similar to diabetes management, tele-ophthalmology is associated with large start-up costs; however, improving adherence to screenings or early detection can offset these costs.

Another study provided a breakdown of the specific direct costs in a telemedicine-based diabetic retinopathy screening when compared to standard ophthalmologic evaluations.21 Similar to the study by Brady et al. (2014),20 retinal images were taken, stored, and sent to a separate facility to be examined by specialists. The intervention arm considered costs for healthcare providers, training, equipment, maintenance, and transportation. In addition, the cost of any digital retinal imaging evaluation that indicated a severity warranting a follow-up evaluation was averaged across the eHealth cohort and included. Costs of an estimated $45 per screening were less than the total Medicaid reimbursement of $77.80 per exam. 

A final RCT in patients with diabetes mellitus by Palmas et al. (2010) conducted a cost-consequence analysis of a telemedicine case management compared to usual care.22 This study used Medicare claims data for underserved urban and rural New York primary care practices. The telemedicine diabetes management intervention consisted of trained nurse care managers assisting patients with a home glucose meter, blood pressure readings, lifestyle changes, and follow-up. However, costs of implementation were high and reductions in Medicare costs were not demonstrated, with mean annual payments of $9,040 and $9,669 for the usual care and telemedicine group, respectively.

The Butler et al. (2012) cohort study was a retrospective pilot cost analysis of an asynchronous telepsychiatry model for primary care.23 Telepsychiatry can be delivered either synchronously or asynchronously. This study compared the cost of synchronous and asynchronous management models with in-person psychiatric care. Synchronous telepsychiatry had the highest fixed and marginal costs when compared to asynchronous telepsychiatry and in-person psychiatric care. When comparing the two models, the number of patient encounters at which the total calculated model costs were equal was 249. After 249 patient encounters, each additional patient encounter represented savings for the asynchronous telepsychiatry model. This study, which was performed from a health system perspective, counted implementation and set-up costs, but did not count classified components, such as electronic medical records, and considered them to be sunk costs.

A budget impact analysis of a collaborative eHealth team intervention for depression care was reported by Fortney et al. (2011).24 In this analysis, 395 patients in a VA community-based outpatient clinic were randomised to telemedicine-based collaborative care or usual care. Utilisation and cost data were collected from an automation centre and included provider time, medical supplies, and administrative overhead. Electronic transfer of information and other off-site information transfer costs were all considered sunk costs. The authors hypothesised that patients with depression use more health services and are therefore costlier. They suggested that a collaborative team-based approach may temporarily increase costs but may improve outcomes and reduce long-term costs. There were no significant outcome differences demonstrated between the two groups for the 12-month duration evaluated.

Soran et al. (2010) analysed costs of telehealth in the management of heart failure in a retrospective cost analysis of a RCT.25They compared heart failure management using a home monitoring device and interactive programme with an enhanced patient education and follow-up programme in Medicare patients. The alternative care management consisted of a digital home scale managed by a PCP along with patient and physician education programmes. The telemedicine group’s mean 6-month Medicare costs ($17,837) were more than that of the standard of care group ($13,886), demonstrating that the latter was less costly. There was no significant difference in six-month mortality. One weakness was the short time horizon, which may not allow for a fair comparison of mortality. Therefore, an accurate assessment of cost-effectiveness cannot be made based on the cost comparisons alone.

Cost-effectiveness Studies

One study by Fishman et al. (2013) looked at the cost-effectiveness of care in treating hypertension, which increases national healthcare costs by $73 billion a year, from a health plan perspective.26 The authors compared a web-based collaborative care model using electronic communications, home blood pressure monitoring, and web-based pharmacy care, to usual care, which included only asynchronous home blood pressure monitoring. There was a significant improvement in blood pressure control and an increase in message threads between the patient and pharmacist. Unlike in previous telestroke studies, the value of this intervention came from the consistent monitoring of the patients’ hypertension treatment. Micro-costing was used to determine the costs of all resources, and Group Health estimates were used to estimate healthcare utilisation costs. Literature sources were used to determine survival from blood pressure control, and utility was used to adjust life years downward based on each blood pressure health state. The authors demonstrated an ICER of $1,850 and $2,220 per QALY for men and women, respectively.

A retrospective electronic medical record review by Kirkizlar et al. (2013) looked at the cost-effectiveness of a teleretinal screening based on specific parameters of patient pool size, age, and race compared with pre-teleretinal care.27 They used a discounted lifetime Markov model with inputs and costs from a diabetic VA patient population before and after a newly introduced teleretinal screening programme. The screening was shown to be cost-effective for most treatment scenarios ($20,000 to $92,000 per QALY). The screening was cost-effective for patient pool sizes of 3,500 or greater and aged 50 to 80 years, and cost-saving for those younger than 50. They also demonstrate that their CEA results were insensitive to telemedicine-related costs.

Conclusion

Our review uncovered that most studies focused on savings in two categories: rural areas and areas lacking clinical specialists. Most studies were simple cost assessments and few were "gold standard" CEA studies.4 Promisingly, the results in all presented CEA studies, suggested that eHealth was cost-effective. The available  CEA studies focused on three specific interventions, namely telestroke, tele-ophthalmology, and telepsychiatry, with considerable good evidence of the tele-approach being cost-effective. Other cost studies focused on management of chronic diseases, where the addition of web-based monitoring did not consistently demonstrate clear evidence of a cost advantage and did not address full cost-effectiveness.

Our reviewed studies showed a wide variation in type of economic evaluation and also how costs were analysed and considered. Most importantly, we found that there were too few complete economic evaluations, such as cost-effectiveness and cost-benefit studies which address both costs and effects. The many cost studies, while providing valuable examples of how to assess costs, leave uncertain, how the eHealth savings affect the outcomes of care. Both are essential to best evaluate an eHealth intervention. 

Other differences were in how those costs were analysed and considered.  Some studies focused on implementation of the eHealth intervention and included all costs, including infrastructure, while others focused only on the cost and effects after the eHealth system was implemented, especially when the infrastructure was already in place.  

It is also important to consider the completeness of each economic analysis in light of the perspective of that analysis. Studies with a societal perspective will include all costs, while those from a health system perspective will include only those costs relevant to that system. Additionally, the economic studies also differed by taking either or short-term or a long-term view point which could dramatically affect their results. As Fortney et al (2011) suggested, eHealth may increase short-term costs and decrease long-term costs due to either better health outcomes or the distribution of implementation costs.24

Finally, as highlighted by Li et al (2012), it may be difficult to draw relevant conclusions from previous cost analyses and apply them to the current eHealth environment because of the different technologies represented in the literature.21  Most noticeably, studies on the cost-effectiveness of any system-level eHealth initiatives were absent, as were  any economic assessments of the many new technologies in eHealth that are rapidly being introduced into the current healthcare marketplace.  This gap may be indicative of the paradigm shift towards eHealth that is already changing the way healthcare is practiced in the U.S. and  illustrating how quickly adoption of current initiatives appear without strong economic evidence to support it. The eHealth community needs accurate economic validation as it continues to grow. We suggest movement towards more cost-effectiveness studies that begin to comprehensively address all levels of eHealth and virtual integration, especially at the system level.

Corresponding author:
Leslie Wilson
Department of Clinical Pharmacy
University of California, San Francisco
3333 California Street
Laurel Heights
San Francisco, CA 94118

Leslie.Wilson@ucsf.edu

Conflict of interest: The authors declare no conflicts of interest.

References

  1. Centers for Medicare & Medicaid Services. (2016). NHE Fact Sheet. Available at: https://www.cms.gov/research-statistics-data-and-systems/statistics-trends-and-reports/nationalhealthexpenddata/nhe-fact-sheet.html accessed 22 September 2016.
  2. Eysenbach G. What is e-health? J Med Internet Res 2001;3(2):e20. doi: 10.2196/jmr.3.2.e20.
  3. Moher D, Liberati A, Tetzlaff J, Altman DG. The PRISMA Group. Preferred Reporting Items for Systematic Reviews and Meta-Analyses: The PRISMA Statement. PLoS Med 2009;6(7) e1000097. doi:10.1371/journal.pmed.1000097e1000097. doi:10.1371/journal.pmed.1000097.
  4. Gold MR, Siegel JE, Russell LB, Weinstein MC. Cost-effectiveness in Health and Medicine. Oxford, England: Oxford University Press; 1996.
  5. Husereau D, Drummond M, Petrou S, et al. Consolidated Health Economic Evaluation Reporting Standards (CHEERS)—explanation and elaboration: a report of the ISPOR health economic evaluation publication guidelines good reporting practices task force. Value Health 2013;16(2):231-250. doi: 10.1016/j.jval.2013.02.002.
  6. Higgins JPT, Altman DG, Sterne JAC. Cochrane Handbook for Systematic Reviews of Interventions. The Cochrane Collaborative®. Version 5.1.0, 2011.
  7. Urquhart AC, Antoniotti NM, Berg RL. Telemedicine—an efficient and cost-effective approach in parathyroid surgery. Laryngoscope 2011;121(7):1422-1425. doi: 10.1002/lary.21812.
  8. Canon S, Shera A, Patel A, et al. A pilot study of telemedicine for post-operative urological care in children. J Telemed Telecare 2014;20(8):427-430. doi: 10.1177/1357633X14555610.
  9. Anderson ER, Smith B, Ido M, Frankel M. Remote assessment of stroke using the iPhone 4. J Stroke Cerebrovasc Dis 2013;22(4):340-344. doi: 10.1016/j.jstrokecerebrovasdis.2011.09.013.
  10. Kim H, Spaulding R, Werkowitch M, et al. Costs of multidisciplinary parenteral nutrition care provided at a distance via mobile tablets. JPEN 2014;38(2 Suppl):50S-57S. doi: 10.1177/0148607114550692.
  11. Morland LA, Raab M, Mackintosh MA, et al. Telemedicine: a cost-reducing means of delivering psychotherapy to rural combat veterans with PTSD. Telemed J E Health 2013;19(10):754-759. doi: 10.1089/tmj.2012.0298.
  12. Franzini L, Sail KR, Thomas EJ, Wueste L. Costs and cost-effectiveness of a telemedicine intensive care unit program in 6 intensive care units in a large health care system. J Crit Care 2011;26(3):329.e1-6. doi: 10.1016/j.jcrc.2010.12.004.
  13. Hitt WC, Low G, Bird TM, Ott R. Telemedical cervical cancer screening to bridge medicaid service care gap for rural women. Telemed J E Health 2013;19(5):403-408. doi: 10.1089/tmj.2012.0148.
  14. Pyne JM, Fortney JC, Tripathi SP, Maciejewski ML, Edlund MJ, Williams DK. Cost-effectiveness analysis of a rural telemedicine collaborative care intervention for depression. Arch Gen Psychiatry 2010;67(8):812-821. doi: 10.1001/archgenpsychiatry.2010.82.
  15. Demaerschalk BM, Switzer JA, Xie J, Fan L, Villa KF, Wu EQ. Cost utility of hub-and-spoke telestroke networks from societal perspective. Am J Manag Care 2013;19(2):976-985.
  16. Switzer JA, Demaerschalk BM, Xie J, Fan L, Villa KF, Wu EQ. Cost-effectiveness of hub-and-spoke telestroke networks for the management of acute ischemic stroke from the hospitals’ perspectives. Circ Cardiovasc Qual Outcomes 2013;6(1):18-26. doi: 10.1161/CIRCOUTCOMES.112.967125.
  17. Nelson RE, Saltzman GM, Skalabrin EJ, Demaerschalk BM, Majersik JJ. The cost-effectiveness of telestroke in the treatment of acute ischemic stroke. Neurology 2011;77(17):1590-1598. doi: 10.1212/WNL.0b013e318234332d.
  18. Yoo BK, Kim M, Sasaki T, Melnikow J, Marcin JP. Economic evaluation of telemedicine for patients in ICUs. Crit Care Med 2016;44(2):265-274. doi: 10.1097/CCM.0000000000001426.
  19. United States Department of Agriculture Economic Research Service. (2016). Population & Migration: Overview. Available at: http://www.ers.usda.gov/topics/rural-economy-population/population-migration.aspx accessed 1 August 2016.
  20. Brady CJ, Villanti AC, Gupta OP, Graham MG, Sergott RC. Tele-ophthalmology screening for proliferative diabetic retinopathy in urban primary care offices: an economic analysis. Ophthalmic Surg Lasers Imaging Retina 2014;45(6):556-561. doi: 10.3928/23258160-20141118-11.
  21. Li Z, Wu C, Olayiwola JN, Hilaire DS, Huang JJ. Telemedicine-based digital retinal imaging vs standard ophthalmologic evaluation for the assessment of diabetic retinopathy. Conn Med 2012;76(2):85-90.
  22. Palmas W, Shea S, Starren J, et al. Medicare payments, healthcare service use, and telemedicine implementation costs in a randomized trial comparing telemedicine case management with usual care in medically underserved participants with diabetes mellitus (IDEATel). J Am Med Inform Assoc 2010;17(2):196-202. doi: 10.1136/jamia.2009.002592.
  23. Butler TN, Yellowlees P. Cost analysis of store-and-forward telepsychiatry as a consultation model for primary care. Telemed J E Health 2012;18(1):74-77. doi: 10.1089/tmj.2011.0086
  24. Fortney JC, Maciejewski ML, Tripathi SP, Deen TL, Pyne JM. A budget impact analysis of telemedicine-based collaborative care for depression. Med Care 2011;49(9):872-880. doi: 10.1097/MLR.0b013e31821d2b35.
  25. Soran OZ, Feldman AM, Piña IL, et al. Cost of medical services in older patients with heart failure: those receiving enhanced monitoring using a computer-based telephonic monitoring system compared with those in usual care: the heart failure home care trial. J Card Fail 2010;16(11):859-866. doi: 10.1016/j.cardfail.2010.05.028.
  26. Fishman PA, Cook AJ, Anderson ML, et al. Improving BP control through electronic communications: an economic evaluation. Am J Manag Care 2013;19(9):709-716.
  27. Kirkizlar E, Serban N, Sisson JA, Swann JL, Barnes CS, Williams MD. Evaluation of telemedicine for screening of diabetic retinopathy in the Veterans Health Administration. Ophthalmology 2013;120(12):2604-2610. doi: 10.1016/j.ophtha.2013.06.029.

APPENDIX A.

Systematic review search strategy Search Algorithm

Terms were: (telemedicine OR "Mobile Health" OR "Health, Mobile" OR mHealth OR mHealths OR Telehealth OR eHealth) AND (“Cost-Benefit Analysis” OR "Analyses, Cost-Benefit" OR "Analysis, Cost-Benefit" OR "Cost-Benefit Analyses" OR "Cost Benefit Analysis" OR "Analyses, Cost Benefit" OR "Analysis, Cost Benefit" OR "Cost Benefit Analyses" OR "Cost Effectiveness" OR "Effectiveness, Cost" OR "Cost-Benefit Data" OR "Cost Benefit Data" OR "Data, Cost-Benefit" OR "Cost-Utility Analysis" OR "Analyses, Cost-Utility" OR "Analysis, Cost-Utility" OR "Cost Utility Analysis" OR "Cost-Utility Analyses" OR "Economic Evaluation" OR "Economic Evaluations" OR "Evaluation, Economic" OR "Evaluations, Economic" OR "Marginal Analysis" OR "Analyses, Marginal" OR "Analysis, Marginal" OR "Marginal Analyses" OR "Cost Benefit" OR "Costs and Benefits" OR "Benefits and Costs" OR "Cost-Effectiveness Analysis" OR "Analysis, Cost-Effectiveness" OR "Cost Effectiveness Analysis"). In addition, the non-MeSH term, “virtual healthcare,” was also searched.

Search Results
The search returned 238 results and contained literature from a wide variety of study designs, including evaluation studies, randomized controlled trials, reviews, and meta-analyses. An additional seven studies were identified through screening references. The search was limited to studies from January 1, 2010 to July 3, 2016 in order to solely capture the most recent evidence likely to be applicable to the current virtual healthcare environment. These were further filtered by only English-language articles, relevance to humans, and the existence of an abstract. After removing duplicates, 243 abstracts were read to determine whether they met our inclusion and exclusion criteria. The inclusion criteria included having a virtual healthcare component and having an economic outcome or cost analysis, as defined by our search algorithm above. The exclusion criteria included studies that were telephone-only, web-based information dissemination-only, and non-U.S.-based population studies. Papers with both a qualifying and non-qualifying component were included in our review. For example, if an intervention included telephone-use with another means of virtual healthcare, the paper was also included.

This process identified 43 abstracts that were read for full-text assessment. Two of the most common reasons studies were further excluded were due to location (i.e. study population outside of the U.S.) and irrelevance to virtual healthcare costs (e.g. intervention was solely web-info based or yielded no economic results). There were 20 studies of which five were based on RCTs, 14 were based on different types of cohort studies, and one was a case series study.

Appendix B

A review of studies adapting the Consolidated Health Economic Evaluation Reporting Standards (CHEERS) checklist by the ISPOR Task Force Report.

Author; Journal (Year)

Population,
Study Type, &
Perspective

Intervention

Control

Variables

Economic Results

Strengths

Weaknesses

Urquhart et al.; The Laryngoscope (2011)7

Population: Retrospective review of parathyroidectomy post-operative visits (intervention n=39, total n=149)
Study Type: Retrospective cohort study with cost savings analysis
Perspective: Patient

Telehealth appointment with specially-trained nurse and room camera connected to physician at distant site

Patients visit the surgical site for their postoperative visit. No further details provided.

Cost:
Transportation was the only cost variable

Outcome:
There were no post-operative surgical complications noted with any of the visits

- Average round-trip travel translated into $357 (119 miles) saved per patient
- Immeasurable benefits to both the patient and the healthcare system mentioned (e.g. reduced risk to patient, improved surgeon productivity, internal cost reductions, greater patient satisfaction)

- Simple study clearly highlighting cost savings in specific situation

- Only one outcome variable mentioned (no post-operative complications). Difficult to interpret whether cost-effectiveness, is demonstrated.
- Study only looked at transportation cost
- Patient perspective limits applicability of results

Canon et al.; Journal of Telemedicine and Telecare (2014)8

Population: Paediatric patients from northwest Arkansas who required post-operative urology follow-up (n=61). 10 patients chose tele-medicine and 51 patients chose to be evaluated at the physical clinic.

Study Type: Retrospective cohort study with cost analysis


Perspective: Patient

The physician at the Arkansas Children's Hospital used the camera in the room to obtain close-up views of the surgical area. When the clarity of the video image was poor, the nurse at the remote clinic sent a digital image via encrypted email.

Description not given but assumed to be similar to intervention arm with the exception that a physician is physically present and not remote

Cost:
- Distance to the remote site and ACH clinic
- Travel time to the remote site and ACH clinic

 

Outcome:
None, other than mention that previous studies found patients to be highly satisfied with telemedicine especially in rural areas

Participants saved approximately $88 and 2.6 hours of their time per trip

Straightforward logic for cost savings clearly demonstrates validity of result

- Low sample size for intervention arm (n=10). Which arm patients enrolled in was voluntary which may bias results.
- Results difficult to interpret without understanding comparative clinical outcome

Anderson et al.; Journal of Stroke and Cerebrovas-cular Diseases (2013)9

Population: Acute ischemic stroke patients (n=20)
Study Type: Cohort study with cost discussion
Perspective: Rural hospital

NIH Stroke Scale (NIHSS) assessment done by a remote provider via iPhone 4

NIHSS assessment done by a provider at the patient's bedside

Cost:
Telemedicine implementation
Outcome:
- Average NIHSS scores - Mean remote assessment time 

- Startup costs for currently available systems for assessment of acute stroke cost upwards of several thousands of dollars as opposed to the several hundred dollar cost of an iPhone 4

- Using a standardized scoring metrix: National Institutes of Health Stroke Scale (NIHSS)
- Compares clinical outcomes

- Small sample size
- Focus is not a cost-analysis, but just a general cost discussion, which is lacking and provides no quantitative cost information

Kim et al.; Journal of Parenteral and Enteral Nutrition (2014)10

Population: Patients requiring long-term home parenteral nutrition (HPN) infusion care for nonmalignant bowel disease (n=45)
Study Type:

Case series study with cost accounting
Perspective:

Patient

Mobile distance clinic appointments using video teleconferencing software, mobile tablets with an unlimited data plan, encryption equipment program for HIPPA security, encrypted email accounts and firewalled websites

Standard in-person home parenteral nutrition care appointments

Cost:
- Supply of home HPN materials
- Equipment, connection, and delivery expenses
- Initial setup costs (mobile tablet devices, 4G data plans, personnel time of multidisciplinary team, intervention materials)
Outcome:
- Quality of care and level of satisfaction with virtual examinations

Costs for mobile clinic appointment:
- Total setup cost of initial appointment: $916.64/patient
- Total equipment (devices, encryption equipment): $590.62/patient
- Average follow-up: $361.63/month
- Supply costs were $82.18 for the Internet connection, device maintenance, and intervention materials
- Average delivery fees were $53.84
- Personnel costs (2 interventionists and 1 telehealth system coordinator) $190/hr

- Showed clear breakdown of all costs related to the intervention

- Lacks discussion of clinical benefit
- Outcome variables are unclear
- Patient perspective limits applicability of results to stakeholders

Morland et al.; Telemedicine and e-Health (2013)11

Population: Male veterans with PTSD and anger problems recruited across 3 VA clinical sites and 3 Veterans Centers in Hawaii (n=74)
Study Type: Retrospective cost analysis of a RCT
Perspective:

VA healthcare system

Psychotherapy care delivered via clinical videotelecon-ferencing (CVT)

Traditional in-person psychotherapy care

Cost:
Travel , Personnel per session, Therapy per session, Partici-pant per session, Equipment
- Total cost per participant
Outcome:
- Average Novasco Anger Scale, Trait Anger, and Anger Expression Index scores

- Unadjusted mean cost of CVT significantly (p=0.00) lower by $713, relative to the mean cost of in-person delivery
- Significant (p=0.00) cost reduction with CVT was relatively stable across the 3 clinical outcomes (NAS-T, T-ANG, AEI), from $703 for the T-ANG, $708 for the AEI, and $710 for the NAS-T

- Clinical outcomes were controlled by testing sensitivity of cost findings

- Lacks indirect cost factors at the patient (ie. quality of life, travel costs, lost wages, time, family burden) or societal (ie. productivity, lost taxes, early mortality, out-of-network service provider costs, insurance) levels
- Travel costs included flying, which may not be generalisable

Franzini et al.; Journal of Critical Care (2010)12

Population: Two independent groups of patients during a pre-tele-ICU period (n=1,913) and post-tele-ICU period (n=2,067) in 6 ICUs at 5 hospitals in the Gulf Coast region
Study Type: Observational cohort study with cost consequence analysis
Perspective: Healthcare system

Tele-ICU system in the administrative offices of the health care system. Equip-ped with audio-visual monit-oring providing real-time vitals and audiovisual connections to patients' rooms. Staffed by 2 intensivists, 4 nurses and 2 admin tech-nicians. Tele-ICU physicians conducted rounds based on subjective assessments of illness severity. Progress notes faxed daily from monitored units to the tele-ICU, orders from the tele-ICU were entered into the computer workstation and printed in the monitored units.

Intensivist-led multidisciplinary teams conduct rounds among all patients at least once a day. Patients never had non-intensivists as attending physicians. Some ICUs had patients cared for by several types of physicians. One unit had intensivist coverage for approximately 50% of the patients, and another had coverage for 10% or less of the patients. Remaining patients had no intensivist coverage.

Cost:
- Hospital costs
- ICU costs and floor costs
- Costs per case and costs per patient
Outcome:
- ICU mortality only

Average cost changes from pre-tele-ICU to post-tele-ICU:
- Daily ICU: $2,851 to $3,653 or 28% increase
- Daily floor: $1,451 to $1,687 or 16% increase
- Overall ICU per case: $13,029 to $19,324 or 48% increase
- Overall floor per case: $8,938 to $11,994 or 34% increase
- Hospital cost per patient: $20,231 to $25,846 or 28% increase (patients with SAPS II ≤ 50 saw statistically significant 35% increase while patients with SAPS II > 50 did not see a statistically significant increase)
Patients with SAPS II > 50 saw costs per patient increase $2,985 (not statistically significant), however hospital mortality decreased significantly by 11.4% suggesting that tele-ICU intervention is cost-effective in this sub-group of stroke patients

- Comprehensive consideration of costs and economics
- Large study with diverse mix of ICU sites and high quality data

- Use of health care system perspective in the cost-effectiveness analysis
- Inability to address impact of the tele-ICU on hospital volume and revenue due to lack of data
- Lack of randomization which may result in ICUs with better physician acceptance and integration of information systems more favourable clinical outcomes

Hitt el al.; Telemedicine and e-Health (2013)13

Population: Arkansas underserved/ Medicaid rural female patients with an average age of 26.2 years (n=1,812)
Study Type: Cohort study with cost consequence analysis
Perspective: Hospital

Colposcopy services via interactive telemedicine at four separate sites. Weekly 3-hour clinic. A nurse/nurse practitioner onsite per-formed the exams and collected biopsy specimens under the real-time, super-vision of a faculty member at the hub site. Total personnel is 1 MD and 4 NP. and 4 assistants.

Traditional in-person colposcopy incorporating 4 MD examiners and 4 assistants

Cost:
- Cost per exam
- Hourly rates of all practictioners and assistants involved
Outcome:
- Identification and designation of any precancerous lesions (high-grade squamous intraepithelial lesions are rated on severity) or cancer (squamous cell carcinoma or adenocarcinoma) in biopsy specimens

- Traditional model produces hourly cost of $416, or a cost per exam of $52
- Telecolposcopy model produces hourly cost of $321, or a cost of $4 per exam
- For a conventional pap test every 3 years up to the age of 75 years, estimated QALY was $11,830/QALY saved in year 2000 dollars

- Large sample size allows for sensitivity and positive predictive value calculations

- Very specific patient population (geographically isolated, underserved rural areas, vulnerable) which may not be generalizable
- Marginal cost of $11,830/QALY does not include costs of technology, telecommunications, or overhead
- Regarding the biopsy results, there exists a discrepancy between the sensitivity for finding high-grade lesions and the sensitivity/PPV of the impression

Pyne et al.; Archives of General Psychiatry (2010)14

Population:

Primary care patients who screened positive for depression and had a physician who felt comfortable treating the patients for depression were recruited from 7 VA Community-Based Outpatient Clinics. Patients with serious mental illness were excluded. (n=395)
Study Type:

RCT with cost-effectiveness analysis
Perspective: Veterans Health Administration

Stepped-care model for depression treatment by an off-site depression care team to make treatment recommendations via electronic medical records. Team included a nurse depression care manager, clinical pharmacist, and psychiatrist, collectively made up the remote portion of the Telemedicine Enhanced Antidepressant Management (TEAM) team.

Usual care still included provider and patient education via interactive video, mail, and website. Psychiatrist was on-site. A key difference between control and intervention was the TEAM intervention.

Cost:
- Intervention costs (including patient education pamphlets, care provider edu-cation, develop-ment of partic-ipant and care provider sections of the website, interactive video equipment, DCM intervention train-ing, and time spent by intervention personnel), health care expenditures
Outcome:
Primary outcome, depression-free days
Clinical outcomes could be used with clinical success.
- intervention group had significantly better treatment adherence at 6 and 12 months greater odds of
- significantly more likely to demonstrate depression treatment response (odds ratio=1.9; P=.02) at 6-month follow-up and by 12 months, the intervention group had significantly greater odds of depression remission (odds ratio=2.4; P=.02)
- Mental health status measured by 12-item Short Form for Veterans (SF-12V) improved more in the intervention group than in the usual care group at 6 months (P=.07) and at 12 months (P=.009)
- Health-related quality of life as measured by the Quality of Well-being (QWB) scale improved significantly more in the intervention group at 6 months (P=.003), but not at 12 months (P=.70), compared with the control group

- Unadjusted mean 12-month health care utilization expenditures by category were all greater for the intervention group
- Clinical outcomes significantly improved in the intervention group but the intervention did not significantly improve 6- or 12-month QALYs
- In base case analysis (existing sample, not boot-strapped), incremental intervention effects on SF-12 QALYs (β=0.018; SE=0.009; P=.04) and expenditures (β=$1,528; SE=$298; P=.04) were significant
- Mean incremental cost-effectiveness ratio (ICER) using SF-12 QALYs and expenditures from the bootstrapped-with-replacement sample was $85,634/QALY (median, $85,932.QALY; interquartile range, $48,911/QALY-$122,952/QALY)
- Adding depression-related inpatient expenditures, the incremental intervention effect on expenditures was significant (β=$1,510; SE=$326; P<.001) and the mean ICER using SF-12 QALYs and expenditures from the boot-strapped-with-replacement sample was $132,175/QALY(median, $83,174/QALY; interquartile range, $36,722/QALY-$119,869/QALY)
- Adding all inpatient expenditures to the base case analysis expenditures, the incremental intervention effect on expenditures was significant (β=$1,355; SE=$404; P=.001) and the mean ICER was $111,999/QALY (median, $71,028/QALY; interquartile range, $32,057/QALY-$103,085/QALY)
- Adding patient expenditures to the base case expenditures, the incremental intervention effect on expenditures was significant (β=$1,304; SE=$371; P<.001) and the mean ICER was $72,636/QALY (median, $74,390/QALY)

- Detailed study design and protocol
- Strong analysis of clinical outcomes to provide strong cost-effectiveness argument

- Study intervention is not specific to telemedicine, but rather, a depression management team that partially operates remotely

Demaerschalk et al.; American Journal of Managed Care (2013)15

Population: Hypothetical cohort of acute ischemic stroke patients with mean age of 68 years (base case n=1,112)
Study Type: Markov model for hypothetical cohort study with cost effectiveness analysis
Perspective: Societal

Tele-stroke network (modelled using data inputs from Georgia Health Sciences University and the Mayo Clinic telestroke networks)

No network setting for stroke care

Cost:
Telestroke setup and maintenance, initial hospital-ization, post acute stroke care  (rehabilitation and nursing home), and caregiver
Outcome:
Utility values as measured by the EuroQol, were obtained from literature.

- Telestroke network resulted in incremental cost savings of $1,436 per patient over a lifetime horizon
- Incremental effectiveness in QALYs was 0.002 per patient in the 1-year time horizon, but increased to 0.02 per patient in the lifetime scenario

- Quality data source with inclusion of an array of cost inputs
- Societal perspective taken, allowed for by comprehensive independent telestroke network data as inputs

- Lack of transparency with telestroke network data and cost decision model logic
- Only disability considered for cost-effectiveness analysis

Switzer et al.; Circulation: Cardiovascular Quality and Outcomes (2012)16

Population: Hypothetical cohort of acute ischemic stroke patients (base case n=1,112)
Study Type: Decision analytic model for a retrospective cohort study with cost consequence analysis
Perspective: Management of a telestroke network from the perspectives of a network, a hub hospital, and a spoke hospital

Tele-stroke network (modelled using data inputs from Georgia Health Sciences University and the Mayo Clinic telestroke networks)

No network setting for stroke care

Costs:
Telestroke setup and maintenance, acute ischemic stroke treatment, reimbursements
Outcome:
- Incremental home discharges, inpatient rehab-ilitation/nursing home discharges, and in-hospital deaths
- Incremental acute ischemic stroke patients treated with intravenous thrombolysis and endovascular stroke therapy and admitted to each hospital

- Network perspective: average cost savings of $358,435 per year with a telestroke network versus without one during the first 5 years. Cost savings increase over time, from $234,836 at the end of 1 year to $393,712 at the end of 5 years.
- Hub hospital bear positive costs of $405,121 per year
- Spoke hospitals save $109,080 per year
- Each hospital can achieve equal cost savings of $44,805 per year during a 5-year time horizon

- Quality data source

- Lack of transparency with telestroke network data and cost decision model logic
- Study claims to be a cost-effectiveness study but only presents cost differences and effectiveness separately

Nelson et al.; Neurology (2011)17

Population: Patients with ischemic stroke
Study Type: Decision analytic model with cost effectiveness analysis
Perspective: Societal

2-way, audiovisual technology that links stroke specialists to remote emergency department physicians and their stroke patients

Remote emergency departments without telestroke consultations or stroke experts

Costs:
Telestroke equipment, staffing, and training.

Patients transfer.
Outcome:
Total cost, outcomes included tissue plasminogen activator (tPA), , and (QALYs)

Telestroke results in an ICER of $108,363/QALY in the 90-day horizon and $2,449/QALY in the lifetime horizon. For the 90-day and lifetime horizons, 27.5% and 99.7% of 10,000 Monte Carlo simulations yielded ICERS<$50,000/QALY

- Assumptions and variabilities were accounted for, by using various sensitivity analyses at different time horizons, and a second-order Monte Carlo simulation

- Model was based only on initial strokes
- Costs were converted to 2008 US dollars

Yoo et al.; Critical Care Medicine (2016)18

Population: Adult ICU patients
Study Type: Decision analytic model with cost effectiveness analysis
Perspective: Healthcare system

Audiovisual telemedicine with data communications in the ICU, link- ing hospital ICUs to intensivists and other critical care profess-ionals at remote locations"

Traditional ICU care without telemedicine

Costs:
Per-patient per-hospital-stay, ICU, and floor cost after ICU
Outcome:
ICU mortality, floor-mortality, mortality rate after hospital discharge, cumulative utility, QALY after discharge

- Tele-ICU estimated ICER of $45,320/QALY compared to ICU without telemedicine
- Tele-ICU extended 0.011 QALYs/patient with incremental cost of $516 compared to ICU without telemedicine
- 95% CI of 1,000 ICER estimates ranged from -$229,016 to $375,870

- Model inputs use U.S. literature data, giving model good applicability
- Consideration of aggregated costs ICU can provide more easily interpreted results

- Takes on system perspective and considers only costs at the system level
- Results with very large 95% CI are difficult to interpret

Brady et al.; Ophthalmic Surgery, Lasers and Imaging Retina (2014)20

Population: Routine medical care for adult patients with diabetes at an urban practice (n=99)
Study Type: Cohort study with cost consequence analysis
Perspective: Hospital

Non-mydriatic fundus photography that screens for proliferative diabetic retinopathy (PDR) with remote grading

In-office screening for PDR

Cost:
- Tele-ophthal-mology screening cost per patient
- Treatment costs
- Screening costs
Outcome:
- Prevalence of PDR
- Prior adherence to PDR screening

- Monte Carlo simulation indicated screening saved a median of $48 per patient versus $30 in the base case

- One-way and probabilistic sensitivity analyses were performed to evalute the robustness of study findings

- Used local Medicare reimbursement rates which reduces generalizability of the study due to varying reimbursements

Li et al.; Connecticut Medicine (2012)21

Population: Diabetic patients from large multi-site Federally Qualified Health Center (n=611)
Study Type: Cohort study with cost consequence analysis
Perspective: Healthcare system

Store and forward non-mydriatic retinal screening examination by trained tech-nicians. Images and medical history sent to secure Web server. Consultation report transmitted to the referring primary care providers electronically.

Standard opthalmologic examination after  patients referred by a primary care provider to an ophthalmologist. Appointment is made, and the patient is evaluated by the ophthalmologist in the office several weeks to months later.

Cost:
Initial costs: Equipment, training overhead
Annual costs: Human resources, device mainten-ance, overhead
Other: Transportation, cost for conventional fundus ophthalmoscopy, and imaging examination
Outcome:
Digital retinal imaging highly sensitive in detecting diabetic retinopathy. No other outcome variables are mentioned.

- Costs for telemedicine-based retinal screening were $17.60 for capital (equipment + training), $1.50 for annual maintenance, $18.30 for staff  labour ($15.00 for ophthalmologist, $3.80 for medical assistant), and $2.50 for transportation fees. With the cost of (12.3% of patients)  requiring subsequent examinations total was $49.95.
- Costs for the conventional fundus ophthalmoloscopy $8.70 on average for roundtrip transportation, $65.30 for bilateral eye examination, and $3.80 for nurse labour costs. Total cost of the conventional intervention was $77.80.

- Considered depreciation, maintenance costs, and net present value for equipment costs

- Publication seemingly has several errors (e.g. missing citations, incorrectly labeled subeadings)
- Population is unique and data sources specific (to Medicaid) therefore may be difficult to extrapolate results to other populations
- Lack of outcome variables result in difficulty interpreting economic results

Palmas et al.; Journal of the American Medical Informatics Association (2010)22

Population: Participants (aged 55 and older) were ethnically diverse with diabetes mellitus, and resided in federally designated medically underserved areas of New York State (n=1665)
Study Type: Cost consequence analysis
Perspective: Medicare

Home telemedicine unit (containing a web camera, home glucose meter and blood pressure cuff, access to patients' own clinical data and educational page) with nurse case management

Patients received clinical care from their primary care provider, without other guidance or direction from study personnel

Cost:
- Commercial vendors costs

 Equipment, service, travel
- Clinical teams, salaries
- Annual Medicare payments
Outcome:
Changes in Pre-specified clinical endpoints (HgbA1c, LDL cholesterol, and blood pressure levels)

- Mean annual payments estimated as $9,040 and $9,669 for the usual care and telemedicine groups, respectively
- Telemedicine arm estimated project intervention costs of $622 per participant per month of intervention delivered. More than 57% of expenditures incurred by commercial vendors.

- Robust telemedicine intervention; to assess the robustness of the intention-to-treat (ITT) analysis, expenditures were compared between the two treatment groups at 3 different time periods during the follow-up process
- An additional analysis was performed to account for claims data for both censored and non-censored participants

- Analysis was only on Medicare claims, which may be skewed
- Limitation is that cost analysis lacks cost-effectiveness; as such, stating that the implementation costs are high "compared to other diabetes case management interventions evaluated in the literature" is arbitrary and lacks societal perspective

Butler et al.; Telemedicine and e-Health (2012)23

Population:

Patients in Tulare County, CA identified by primary care providers as having non-urgent psychiatric problems (n=125)
Study Type: Retrospective cost-analysis of a cohort study
Perspective: Healthcare system

Web based store and forward review of video-taped patient interviews at rural clinic, with mental state assessments, electronic data, medical records and treatment plan. Consu-ltation opinion electronically transmitted to primary care physician with the option of follow-up phone or email consu-ltations with the psychiatrist.

In-person consultations are regularly performed at University of California Davis

Cost:
- Infrastructure costs (laptops, video equipment)
- Development cost for proprietary Web-based consultation program
- Hourly labour costs by providers
Outcome:
- Clinical outcomes were not considered, state the authors, due to the retrospective nature of the data collection

- Asynchronous telepsychiatry had a fixed cost of $7,000 ($4,000 consisted of development costs for the Web-based consultation software, $2,000 consisted of equipment costs, and $1,000 consisted of provider training costs)
- Fixed costs for in-person psychiatry were considered sunk costs
- Asynchronous telepsychiatry produced a marginal cost of $68.18 per person, compared to in-person psychiatry's marginal cost of $96.36

- Simple cost analysis clearly showing potential for cost savings

- Without clinical outcomes, it is difficult to interpret cost-effectiveness of the asynchronous telepsychiatry intervention

Fortney et al.; Medical Care (2011)24

Population:

Primary care patients positive for depression with a physician who felt comfortable treating for depression were recruited from 7 VA Community-Based Outpatient Clinics (n=395)
Study Type:

 RCT with budget impact analysis
Perspective: Healthcare system

Stepped-care model for depression treatment by off-site depression care team of a nurse, depression care manager, clinical pharmacist, and psychiatrist to make recommendations via EMRs..

Usual care included provider and patient education via interactive video, mail, and website. Psychiatrist on-site.

Cost:
- Direct costs reflect costs associated with provider time and medical supplies
- Indirect costs refer to resources that enable patient encounters, such as administrative overhead and clinic space
Outcome:
- Total number of primary care encounters (depression- or non-depression- related)

- Average cost of TEAM intervention: $794/patient
- Intervention group had significantly higher depression-related mental health costs (ME=$107.55, P=0.03), higher total outpatient cost (ME=$599.28, P=0.012) and higher specialty physical health costs (ME=$490.60, P=0.003)

- Robust data set allowing for comprehensive subset analysis (e.g. service line cost data) for more informed hospital decision making

- Study intervention is not specific to telemedicine, but rather, a depression management team that partially operates remotely

Soran et al.; Journal of Cardiac Failure (2010)25

Population: Medicare patients (aged 65 or older) with the diagnosis of heart failure, specifically systolic heart failure, secondary to systolic dysfunction and had symptoms despite having received prior therapy (n=315)
Study Type: Retrospective cost consequence analysis of a RCT
Perspective: Medicare

After patient and clinician education, patients used a commercial computer-based telephonic heart failure monitoring system that detects early signs and symptoms of heart failure using telecommunication equipment. Patients were managed by a primary care physician

Standard heart failure care (SC) with enhanced patient education and follow-up in Medicare-eligible patients

Cost:
- Avg 6m costs for Medicare services: hospice, durable medical equipment, home health, outpatient, physician, and inpatient care in
both hospitals and nursing homes
Physician costs
Intervention and imputed drug costs
Outcome:
- Changes in heart failure symptoms or weight
- Cardiovascular death or rehos-pitalization for heart failure
- Length of hospital stay

At 6 months, the mean Medicare costs were estimated to be $17,837 and $13,886 for the HFMS and the SC groups, respectively. Also, the overall medical costs of Medicare patients were significantly higher for patients who were randomized to the HFMS arm, than they were for the patients randomized to the SC arm.

- Recruitment of patients through established cooperative networks in the respective cities, thereby conferring more generalizable and applicable results

- Geographical differences arising from the conduction of trial at 3 different sites (Miami, Cleveland, Pittsburgh) leads to differing Medicare reimbursement costs

Fishman et al.; American Journal of Managed Care (2013)26

Population: Enrolled adults (aged 25-75) diagnosed with hypertension (HTN) and taking HTN meds in 10 primary care medical centres within Group Health's Washington group practice (n=778)
Study Type:

RCT with cost-effectiveness analysis
Perspective:

Health plan (Group Health Cooperative)

Home BP Monitoring plus Pharmacist Care (e-BP) consists of the following:
- Home BP monitor and training on its use
- Direct care supervision from a clinical pharmacist trained in hypertension evidence-based care
- Patient-centred techniques for addressing behavioural issues

Home BP Monitoring (BPM) incorporates Usual Care (UC), which consists of information materials geared towards providing patients with the necessary resources to control hypertension, with a home BP monitor and proper training on its use

Cost:
- Identification of enrolees and development of self-management materials
- Patient training
- Protocol devel-opment and training for pharmacists
- Pharmacist services
- Home BP monitor
- Overhead costs
Outcome:
Change in diastolic/systolic BP; percentage of patients with controlled BP at 12 months

The ICERs for e-BP relative to BPM are:
- $16.65 for each percent increase in the percent of patients with BP control
- $65.29 and $114.82 for each decrease in systolic and diastolic BP, respectively
- $1850 and $2220 per year for life-years saved for men and women, respectively
- No statistically significant improvement in BP control or for the change in diastolic BP. Only significant improvement in systolic BP seen, with an ICER of $29.63 per mmHg systolic BP.

- More tailored health-plan perspective

- Lacked costs incurred by patients; unclear control variable due to two ICERs, which do not directly compare each of the 3 arms to one another

Kirkizlar et al.; Ophthalmology (2013)27

Population: Type 1 and type 2 diabetic patients in the Veterans Health Administration (n=900)
Study Type: Retrospective cohort study (medical chart review) with cost-effectiveness analysis
Perspective: Veterans Health Administration

Digital images are taken by a technician and sent electronically to a central location for reading by a retinal specialist or certified reader

Eye care professional performing a conventional retinal examination in a clinical setting

Cost:
-Testing, treatment, and corresponding complication costs
Outcome:
Diabetic retinopathy disease state (mild non-proliferative, moderate non-proliferative, severe non-proliferative, and proliferative), glucose level, diastolic and systolic blood pressure levels

Assumed cost-effective cost per QALY threshold of $50,000. Teleretinal screening cost-effective for pool sizes of ≥3500. For patient pool size of 9000, confidence interval for the average cost per QALY never exceeds the threshold.
- Patients under 50 are cost-saving, between 50 and 80 are cost-effective and over 80 are not cost-effective. All groups of populations differentiated by race are suggested to be cost-effective except for Native Hawaiian or other Pacific Islander group, which data suggests are cost savings.

- Study uses quality data input for discounted Markov decision process model
- Study designed to be sufficiently powered and allow for robust statistical analysis

- Takes on system perspective, specifically Veterans data, and therefore of limited applicability for society and other systems

Appendix C

An analysis of study risk using the Cochrane Risk Bias Tool.

AppendixB

Wilson L et al., J Int Soc Telemed eHealth 2016;4:e21

.