A panel of experts discussed the evidence base of telehealth, including what is really known, the barriers to creating evidence, the challenges of study designs, and how to translate the evidence base into policy changes and changes in standards of care. The following sections reflect the individual speaker’s comments and reflections.
Elizabeth Krupinski, Ph.D.
University of Arizona
A huge body of literature and evidence for telemedicine already exists, and not just in the telemedicine-specific journals. Figure 8-1 shows a classic pyramid of the hierarchy of evidence (based in part on their importance and relevance) and the types of studies that one can conduct. At the top are systematic reviews and meta-analyses. Under that are best evidence, evidence guidelines, and evidence summaries, which can only appear as the body of evidence evolves because scientifically sound studies are needed to create these kinds of standards. Below this, there are essentially two categories for study designs: experimental and observational (see Box 8-1). In experimental studies, the primary investigator assigns and chooses the test intervention (e.g., telemedicine versus no telemedicine), there is always a control or a comparison group, and the subjects are allocated randomly. With observational studies, the primary investigator studies the people and the exposures in situ. A control or comparison group may or may not be a
FIGURE 8-1 Hierarchy of evidence.
NOTE: The following acronyms are used for spacing purposes: SR is systematic review, MA is meta-analysis, BE is best evidence, EG is evidence guidelines, and ES is evidence summaries.
SOURCE: Reprinted with permission, Elizabeth Krupinski (2012).
part of the study design, which is why some people consider observational studies to be a little less reliable than experimental studies, but that is not necessarily the case. RCTs and observational studies (with a control or comparison group) both have advantages and disadvantages, and in the long run, both are valid and have been used in many telemedicine studies. Observational studies without a control or comparison group are generally considered less rigorous.
Each of these types of studies serves a different purpose. Although the RCT is the gold standard and there should be more of these, other viable alternatives are being used in traditional medicine as well, so there is no reason why these other study designs should not be used and have the same impact in telemedicine as occurs in the rest of medical care.
Experimental and Observational Studies
|Experimental Studies||Observational Studies|
|•||Randomized controlled trials||With a control or comparison group:|
|•||Lab trials||•||Analytic study|
|•||Case control study|
|Without a control or comparison group:|
Randomized Controlled Trials
There is much disagreement about whether RCTs are needed for telemedicine. The strengths of RCTs include the randomization and the fact that they are typically prospective studies. One of the main goals of RCTs is to eliminate, or at least minimize, the different types of biases. Generally, RCTs seek to falsify a hypothesis (rather than confirm). Also, RCTs allow for meta-analysis because of the quantitative data that are collected. Weaknesses include that RCTs are expensive and time consuming, and randomization can be impractical. There may be ethical issues related to withholding a potentially beneficial treatment. For telemedicine, a double-blind study is nearly impossible, which may or may not be important.
Cohort study strengths include that the same characteristic is measured in two different groups. Typically, the groups differ in only one parameter— for telemedicine this would be the use of the technology versus traditional care. Eligibility and outcome assessments are usually standardized. The weaknesses are that these studies are observational, and they are often not randomized. For example, patients may be placed into a group based on a characteristic such as the presence of Internet connectivity in their homes, which may introduce some bias.
Cross-sectional studies are rife in the medical literature, yet they are often criticized in telemedicine as being not rigorous enough. Cross-sectional studies are based on interviews, surveys, or study. The data are collected at a single point in time from a representative sample. This can be difficult for telemedicine because the impact of telemedicine will mostly occur in the long run, but a lot of valuable information can be collected here as well. Sometimes the data collection relies on history or recall, which may introduce bias. Typically, these studies establish associations rather than causality, but are very useful for developing future research in the field.
Case Studies and Case Series
Case studies and case series are a little less rigorous, but still informational. A case study is a description of a single case, such as how something impacted an individual in a unique manner. These studies typically describe rare events, early trends, or unusual manifestations or responses. Case studies are especially useful because they are the stories that are used when trying to educate and influence policy makers. These studies elucidate mechanisms and look at impact, but usually do not have hypotheses or statistical analyses. Typically they are retrospective, but are more scientific than mere anecdotes.
A case series is a little more powerful than a case study. A case series typically represents about 10 to 30 subjects, which is not statistically powerful, yet does provide a body of evidence. Case series include more detailed descriptions, offer a well-described treatment or intervention, have specific exclusion and inclusion criteria, and can be prospective or retrospective. The main weakness is that there is no comparison group.
RCTs and Chronic Disease Management
In 2012, Wootton published a literature review of RCTs for chronic disease management of five conditions: asthma, chronic obstructive pulmonary disease, diabetes, heart failure, and hypertension (Wootton, 2012). This included all RCTs conducted between 1990 and 2011 that had a control group and used one or more telemedicine interventions (of any type). This review yielded more than 1,300 publications. Wootton then looked at the number of subjects, the patient types (e.g., disease and severity), the intervention type (i.e., what type of telemedicine it was), how long the studies went on, and what the outcomes were. He also rated the overall value of the intervention based on whether primary or secondary outcomes in the intervention group were significantly better, worse, or equivalent when
compared with the control group. Wootton found a total of 141 RCTs in which 148 telemedicine interventions had been tested with nearly 37,000 patients. He determined that 108 of the trials were favorable toward the telemedicine intervention and 38 trials showed no statistical difference between the telemedicine intervention and traditional care. So this meta-analysis showed that in 99 percent of the studies, telemedicine interventions were as good as or better than traditional approaches to care.
What were limitations of the study? Some argue there is a publication bias in telehealth because usually only studies with positive outcomes are published. However, this “bias” exists in traditional medicine as well. No significant differences were noted among the diseases for the telemedicine effectiveness, which might arouse suspicion. Finally, the median duration of the studies was only 6 months, which may not be long enough to observe significant impact.
Review of Reviews to Determine Assessment Methodologies
Also in 2012, Ekeland and colleagues published a review of systematic reviews of telemedicine to determine assessment methodologies (Ekeland et al., 2012). They narrowed nearly 1,600 studies to the 50 that described their methodologies. As a result, they made four recommendations for the development of a stronger evidence base for telemedicine:
- larger, more rigorous design-control studies that assess the impact of telemedicine;
- better standardization of populations, interventions, and outcome measures to reduce heterogeneity and to facilitate meta-analyses;
- combination of quantitative and qualitative methods; and
- more naturalistic methods and settings.
Krupinski asserted that the existing body of evidence for telemedicine is robust enough to permit the creation of evidence guidelines and standards, which is a clear indication of the maturity of the field. The ATA has established and is developing a number of guidelines. Other bodies have developed guidelines for telemedicine, including the American College of Radiology, American Dermatology Association, American Medical Association, and many others. There are also many international guidelines, such as the European Code of Practice for Telehealth.
In conclusion, Krupinski stated that the telemedicine community needs to work with funding agencies (e.g., the National Institutes of Health [NIH]) to develop Requests for Proposals for telemedicine research in
line with Ekeland’s recommendations. In addition, meta-analysis projects should be supported; these are difficult to conduct and time consuming, but they provide a more comprehensive perspective. They can be specific to disease or intervention. Finally, standards and guideline development should continue to be supported. This brings together all the evidence that exists, leading to even better research (because people do research based on the standards and guidelines) and, eventually, payment.
Lee H. Schwamm, M.D.
Massachusetts General Hospital;
Harvard Medical School
Acute stroke is a high-impact, but low-frequency, event. It is the fourth leading cause of death and the leading cause of disability. There are 800,000 strokes each year across the United States, and about one practicing neurologist for every 20,000 Americans, or 50 to 60 strokes per neurologist per year. However, not all of these neurologists are experts in stroke care. Telestroke is a way to reallocate the expertise and make it available broadly.
Intravenous tissue plasminogen activator (tPA), a clot-dissolving medication, is known to be beneficial in stroke, but it requires immediate expertise to be available and there is disparity in its use. When it works, it dramatically increases the likelihood that patients will improve and return to a near-normal life. Treatment within the first 90 minutes of an acute stroke leads to a roughly 18-fold better chance of being helped than harmed, and benefit continues for 4.5 hours after stroke onset. Stroke patients need to get to the right location to receive care as quickly as possible. In fact, American Heart Association guidelines support bypassing hospitals that do not have the resources to treat acute stroke when transporting a patient in whom acute ischemic stroke is suspected.
Telestroke is an early and sustainable application of telehealth for stroke care, particularly acute stroke care. The evolution of telestroke first required proof that acute stroke evaluation could be done safely and effectively via telehealth. It also required collaboration of all the key stakeholders to achieve consensus on the need to regionalize care (i.e., specification of criteria and designation of centers to provide acute stroke care), to have ongoing education to make use of those centers, to certify the sites, and to have public reporting of outcomes. Finally, sustainability of telestroke depends on increasing rates of adoption, promoting diffusion and innovation of the technology, and improving reimbursement of services.
Telestroke is a disruptive technology that is changing the delivery of
health care. Clayton M. Christensen, a professor at the Harvard Business School, has been recognized for his conceptualization of the force of disruptive innovation in the world of business and health care. He said:
The challenge that we face—making health care affordable and conveniently accessible to most people—is not unique to health care. […] The transformational force that has brought affordability and accessibility to other industries is disruptive innovation. […] Most disruptions have three enablers: a simplifying technology, a business model innovation, and a disruptive value network. (Christensen, 2011)
Within that context, the goal of telestroke is to expand access to acute stroke care to rural hospitals and smaller hospitals, many of which are neurologically underserved even if they do not reside in areas that meet the definitions of rural or health professional shortage area.1
Disruptive Innovation: Simplifying Technologies
The simplifying technologies for telestroke were
- the use of brain computed tomography (CT) imaging, which allowed for examination of the brain and detection of stroke in vivo;
- an injectable clot dissolver (i.e., tPA) which could be given anywhere, not just at a high-technology center;
- DICOM standards that allowed for the reliable transfer of brain images;
- video standards that enabled lower cost and high reliability videoconferencing; and
- innovations in chip design and video compression that made videoconferencing an everyday event.
Disruptive Innovation: Business Model Innovation
The first enabler for telestroke was the distillation of acute stroke care into a standard process. Also, hospitals without stroke expertise needed to figure out how to provide that service. However, stroke experts are in short supply, and are needed 24 hours per day, 7 days per week, which is expensive. Easy access via telehealth to regional stroke centers means lower
1 A health professional shortage area includes any of the following that the Secretary determines has a shortage of health professional(s): (1) An urban or rural area (which need not conform to the geographic boundaries of a political subdivision and which is a rational area for the delivery of health services); (2) a population group; or (3) a public or nonprofit private medical facility (Code of Federal Regulations, Public Health Service, Department of Health and Human Services, title 42, chapter 1, part 5 ).
costs and access to higher quality care than was previously available locally. Thus this redefined access to acute stroke expertise became an outsourced commodity that hospitals could now purchase.
In the telestroke model, the patient is with a health care provider at the bedside. Images are encoded digitally, sent to a distant location (e.g., stroke center at a hub hospital) on a dedicated server, and then another physician can be connected remotely. The remote provider can zoom in and interact with and visually inspect the patient for key signs. Research has shown that the quality and interrater reliability of the use of the NIH Stroke Scale, the standard measure of stroke disability, by two different observers at the bedside is equivalent to the interrater reliability of its use by one remote observer and one observer at the bedside. Subsequent studies have shown and confirmed that the rates of agreement can increase even more with more clearly defined protocols and guidelines for how to score these tests.
Acute stroke expertise is needed, in part, because reading a CT scan is difficult. An analysis of the first 2 years of the Massachusetts General Hospital telestroke pilot demonstrated not only that the technology increased the likelihood of treatment in a statistically significant manner, but did not significantly add to the time of treatment. It also demonstrated that neurologists could read a CT scan for acute stroke without the need of an additional radiologist. The American Heart Association recommends the use of high-quality videoconferencing for assessment and recommendations and teleradiology for review of brain CT scans when onsite experts are not available.
Telestroke is a cost-effective, sustainable way to allow smaller hospitals with limited neurology coverage to provide equitable access to high-quality stroke care. The use of telestroke also standardizes care across a network or a state, supports the development of stroke centers in the community, and leads to the quicker evaluation and treatment of more patients with tPA. Additionally, remote supervision of tPA before transfer to a regional center has been shown to be feasible and safe. The outcomes of patients treated through the telestroke network have been shown to be equal to those of patients treated within the parent facility.
Disruptive Innovation: Disruptive Value Network
Increases in public reporting and requirements for meaningful use create incentives for hospitals to gain expertise in stroke care. Participation in a telestroke network enables hospitals to access highly experienced stroke experts from a comprehensive center at a reasonable cost. These
spoke hospitals can thereby retain or grow their stroke care portfolio. Additional value comes through education exchange, enhanced patient and provider satisfaction in the local emergency department, and improvements in overall stroke care. Hub hospitals can extend their expertise into the community, which aligns with their academic mission and goals. They grow their stroke business and attract patients for innovative treatments and clinical trials. At the hub level, it can increase provider compensation and satisfaction by broadening their audience of patients and providing opportunities for both academic growth and driving innovation in care. Overall, participation in the network deepens relationships within the community, keeps patients within their own communities, and drives demand for other telehealth delivery models.
Survey of Telestroke Network Participants
A survey of programs providing telestroke network services showed that hub hospitals support spoke hospitals that are both in and out of their corporate networks (Silva et al., 2012). Formal written agreements or contracts exist in nearly all networks. Ninety-five percent of all sites were using high-quality, two-way, real-time interactive video, but only 68 percent were reviewing brain imaging as part of their consultation process. Many incorporated telephone-only (49 percent) or S&F consultations (11 percent) when appropriate. All the networks included ED consultation as a main purpose of the program, and 84 percent said they use the network for the triage of patients for admission versus transfer. Benefiting the community, improving clinical outcomes, and enhancing clinical process effectiveness or provider knowledge were the three most common internal factors for initiating the program. Reducing cost was among the least important factors. That is, hospitals are not viewing this as a cost-saving approach, but instead as a way to improve quality. Less than half of hub hospitals had a dedicated telemedicine software package; many were using their own electronic medical record, dictating into the spoke hospital’s record, or using paper only.
Respondents noted that barriers to growing the telestroke program include the lack of infrastructure funds, inability to obtain physician licensure, and lack of physician reimbursement. Lack of physician buy-in is also an issue. At this point it is not about the technology, or more and better technology, but about eradicating the conventional barriers.
Potential Next Steps
Schwamm said several things can be done to promote the growth and diffusion of telestroke programs, all of which could also be applied to other diseases, including
- providing federal grants to help ensure access to care anywhere that patients are neurologically underserved;
- simplifying the main administrative processes related to licensure, credentialing, and billing, which are frequently different in each state;
- requiring that federal and third-party reimbursements for telestroke services are at rates equivalent to in-person acute stroke care;
- removing requirements for physical proximity to the patient from current critical care billing code regulations;
- encouraging the use of telestroke within a state-based, system of care model or integrated delivery networks rather than a geographically disconnected, one-off transactional model; and
- requiring participation of telestroke sites in national quality improvement programs, including measurement and reporting their outcomes.
In terms of further research or inquiry:
- determine the most effective models of telestroke care delivery;
- explore whether the models can be applied effectively to other disease conditions;
- determine if disparities in access to digital broadband services mirror disparities in health care access; and
- provide funding to measure the actual costs of telestroke versus conventional stroke care (i.e., clinical effectiveness research).
Finally, Schwamm said, the IOM should convene a committee to gather evidence on the current barriers, make recommendations for solutions, and create a clearinghouse of information for states, patients, and providers.
Moderator: Kamal Jethwani, M.D., M.P.H.
Partners Healthcare Center for Connected Health;
Harvard Medical School
An open discussion followed the panelists’ presentations. Audience members were able to give comments and ask questions of the panelists. The following sections summarize the discussion session.
Evidence and Policy
One participant agreed that much evidence already exists, but challenged the field to produce the evidence needed to educate and influence policy makers. Krupinski responded that part of the problem is the rapidity with which technology changes. She said that some sophisticated systems (e.g., smart homes) are impractical and not cost-efficient, but because the technology changes so rapidly, people may eventually find these devices for a very low cost and install them themselves. She also noted that because of the rapidity of technological change, it is nearly impossible to do an effective study and then get the results published, even within a year.
One participant commented that policy makers have a keen interest in evidence of cost effectiveness. Schwamm agreed, stating that within his ACO, they spend a lot of time thinking about ways to make stroke care more cost-effective.
Technology in the Patient-Centered Medical Home
One participant asked how telehealth might help fulfill the promises of the PCMH, contending that rural communities do not have the necessary resources or facilities. He further suggested looking at how the entire spectrum of telehealth, including telephones and fax machines, can contribute to improvements without having to bring patients to the larger medical centers and without having to increase the workforce in the rural community. Krupinski opined that technology in the PCMH will be “off the shelf”—meaning that patients will buy it on their own. She noted that the iPad is an FDA-approved device for reading CTs, ultrasounds, and magnetic resonance imaging (MRI). She also predicted developments in software algorithms to automatically analyze data. Krupinski disagreed with the participant in part, stating that telehealth will be used for PCMHs in rural areas because the cost is dropping dramatically. Schwamm added that hospitals were willing to purchase telestroke services because policy was actually moving ahead of the business model, viewing the availability of stroke care as a health care rights issue. He added that the movement toward global payments liberates resources to allow hospitals to invest in the infrastructure to improve access and create some downstream return on that investment to defray the costs. Schwamm added that technologies like videoconferencing are becoming higher quality and more ubiquitous, and people are more open to its use because of experiences with technologies like Skype.
One participant argued that there is still a role for studies of provider readiness, citing data on provider buy-in as a barrier to adoption, especially in the acute care setting where telemedicine can substantially disrupt workflow. Krupinski disagreed somewhat, stating that those studies can be useful on a local level to understand an individual organization or help to implement a new program, but that they do not need to be published widely.
One participant asked whether existing literature adequately reflects the right control groups. He noted that several speakers had called for telemedicine to be considered a tool for quality improvement in clinical care. He argued that because it is expensive, it should be compared to other less expensive quality-improvement tools (e.g., traditional education and outreach), rather than to nothing. He argued that comparing telemedicine to no intervention may introduce bias to make it appear that telemedicine is better than it might be if compared to a less expensive alternative. Krupinski stated that sometimes the intervention or treatment itself can define the control group. For example, she said, a clear comparison is someone coming into the door of a stroke center versus someone coming to the door of a non-stroke center. She added that the participant’s question directly related to the recommendation for more naturalistic settings, noting that many studies are now mixing interventions—that is, it is not “nothing” versus the intervention but instead, different degrees of telemedicine (e.g., telephone alone versus telephone with video versus telemedicine with supplemental nurse visitation).
Schwamm added that telehealth is an intersection between a technology and human behavior, and so there is an important role for involving social scientists in the study designs. He agreed that methodology questions are to a certain extent shaped by the actual intervention. He noted that the use of telehealth is a systematic change to the way care is provided, and so in his opinion, the more effective studies are those that use a cluster randomized design at the hospital level wherein a cohort of hospitals is identified to participate, and they are randomly allocated to treatment or control groups so that all subjects at the site are allocated to the same intervention, and the comparison is at the hospital level. This prevents cross contamination of the intervention into the control.
One participant noted that telehealth is a broad and encompassing term, so precision is needed in language. Schwamm agreed, adding that precision of language is also needed when talking about affordability and costs. He noted that costs are incurred by patients, by third-party payers, and by the provider institutions, and so clarity is needed about who saves money. He said some interventions may save the patient money that is not accounted for in health care dollars (e.g., patient transportation, time lost from work), which is not the dollars that most legislators or payors are focused on saving. He added that telehealth can reduce usage, such as by using each health care provider to his or her highest capacity (e.g., a nurse practitioner providing care with telehealth support from a physician, a physician providing care with support from a more senior physician). He summarized that precision is needed about the true costs and where real savings occur. Krupinski agreed, stating that definition of terms in a cost analysis is a form of standardization. She stated that differing interpretations of cost makes comparison of cost analysis studies difficult. Krupinski further added that cost of care cannot be separated from quality of care.
Incentives of Health Care Reform
One participant asked if the Health Information Technology for Economic and Clinical Health (HITECH) Act of 20092 and the ACA might motivate the adoption of telehealth. Schwamm stated that the contingencies, costs, and efficiencies associated with the HITECH Act created incentives for his own organization to adopt a single, unified medical record on the platform of an integrated delivery system. He noted, however, that discussions of the medical record have lacked consideration for the integration of telehealth as one of the key priorities in selecting a vendor. He argued that as organizations start to embrace and adopt electronic medical records, vendors will need to ensure the links will be there to incorporate telehealth in all modalities as another domain of the medical record.
One participant asked whether there might be a public push for telehealth technologies and how that might be leveraged. Schwamm agreed that the public is very engaged in access to health care information, and frustrated by the lack of access to providers, especially through modalities
2 The HITECH Act was enacted as part of the American Recovery and Reinvestment Act, P.L. 5, 111th Cong., 1st sess. (February 17, 2009).
such as smartphones and the Internet. He suggested working with disease-based organizations that advocate for patients (e.g., PatientsLikeMe) to determine how to provide patients with trusted sources of information. Krupinski agreed, noting that the ATA works closely with many patient advocacy groups, especially when they speak to policy makers about the advantages of telehealth.