Results of Literature Searches

The search and selection of systematic reviews are summarized in the literature flow diagram (Figure 2). Database searches of published literature resulted in 1,311 potentially relevant articles. We identified an additional 183 potentially relevant articles through the grey literature search and searches of Web sites for telehealth organizations and government agencies. After dual review of abstracts and titles, 617 articles and grey literature reports were selected for full-text dual review. Of these, we determined that 58 systematic reviews met the inclusion criteria and we included these in the literature map. A list of included reviews is in Appendix C. We did not identify additional eligible reviews or relevant reports via the search of government or organization Web sites. The two reasons we excluded reviews that were about telehealth were: 1) because they included mixed interventions, some of which did not meet our definition of telehealth, and they did not report results separately for the interventions and outcomes we included or 2) they did not meet our criteria for a systematic review (i.e., they did not state questions, search citation databases, and assess the quality of identified studies; see Methods above). The reviews excluded for these two reasons are listed in Appendix D.

Figure 2

Literature flow diagram: search results to included studies. ^a Cochrane Database of Systematic Reviews. ^b Grey literature search included the New York Academy of Medicine Grey Literature Collection, Web sites for the American Telemedicine Association, (more...)

Description of Included Systematic Reviews

The 58 included reviews met the strict criteria we established to distinguish systematic reviews that provided content organized, analyzed, and presented in a way that could support contemporary policy and practice decisions about telehealth.^38-95 Information abstracted from each included systematic review is provided in Appendixes E and F. Figure 3 presents the publication year of the included reviews. Almost 80 percent (46 of 58) were published since 2011, indicating a high level of interest in this topic and enough studies to support numerous reviews.

Figure 3

Included systematic reviews by year of publication.

Table 1 includes descriptive information on basic characteristics of the included reviews, such as the final year of the searches performed in the review, which we used as a means of checking how current the information was that formed the basis for the reviews. Table 1 also reports the setting (i.e., where the patients were located). In the majority of reviews, the patients were in their home (60%), while a small number of reviews addressed telehealth when the patient was hospitalized (5%), and some included a mixture of settings including home, hospital and clinic (35%). In terms of the types of outcomes reported, 55 percent reported clinical outcomes, 12 percent reported cost outcomes, and 33 percent reported both clinical and cost outcomes.

Table 1

Characteristics of included systematic reviews.

Table 1 also includes three characteristics of the reviews that can be used to consider the rigor and utility of the reviews. The current standards for systematic reviews require more than simply listing and describing individual studies – they require that the evidence for a topic be synthesized across studies and that the body of evidence is evaluated. This evaluation is often referred to as strength of evidence (SOE) assessment.^34-36 We defined SOE as an evaluation of a group of studies that incorporated prespecified criteria. Quality assessment of individual studies is always considered in determining strength of evidence but other criteria are often added. For example, the SOE used in AHRQ reviews frequently includes the following criteria in addition to the quality of studies: consistency (i.e., whether the results are consistent across studies), directness (i.e., whether the studies provide direct or indirect evidence given the questions asked in the review) and precision (i.e., how precise the effect estimates are). SOE may also consider other factors such as evidence of publication bias. Less than half (41%) of the included reviews reported some form of SOE assessment. We also reported on whether the reviews included a meta-analysis. While quantitative synthesis is not appropriate in all cases, it was used in 26 (45%) reviews. The final characteristic of the included reviews reported in Table 1 is whether the review did or did not present a conclusion about the effectiveness of telehealth. Just over one-fifth (22%) reported that the evidence was inconsistent or contradictory (we used the term unclear to include both) and did not draw a conclusion after reviewing the literature.

The modality or type of technology used for telehealth is also included in Table 1 and represented in Figure 4. Half of the reviews (50%) included mix modalities, meaning either they included primary studies that used multiple technologies or they included primary studies of several different technologies.^{41-45,47,50,51,54-56,58,65,70,71,73,76,79-82,85-87,90,91,94,95} Other reviews limited their inclusion to studies of asynchronous communication, employing various technologies, including special monitors or internet-based applications on standard computers(29%) to facilitate communication.^{38,46,49,59,60,64,66,67,69,72,77,83,84,88,89,92,93} Another common modality was videoconferencing, which was the focus of 17 percent of included reviews.^{39,40,48,57,61,62,68,74,75,78} A smaller number of reviews included only studies that used mobile phones for telehealth (4%).^53,63

Figure 4

Distribution of telehealth modality across included systematic reviews.

Because technology has changed over time, we also looked at the distribution of the modalities covered by the included reviews by publication date (Figure 5). The number of reviews published that reviewed several technologies (mixed) peaked in 2012 at six but was also high in 2014 and 2015. One or two reviews of studies of video have been published every year from 2008 to 2014. More recently, since 2010, reviews have been published that evaluate the use of asynchronous communication and mobile phones.

Figure 5

Systematic reviews by telehealth modalities and publication year.

We classified the included systematic reviews according to the clinical focus and telehealth function. As explained in Methods, these were developed based on our review of the included studies. “Clinical focus” is the clinical condition, indication, or situation telehealth was used to address. Function is the role telehealth played in health care. These categories are defined below and the number of reviews in each category is provided in Table 2.

Table 2

Characteristics of systematic review evidence by clinical focus and telehealth function.

The categories identified and used to describe the clinical focus of each systematic review are:

• Cardiovascular disease: These reviews included studies of the use of telehealth for the management of heart failure, acute care and followup for myocardial infarction, management of patients with implantable defibrillators, and primary and secondary prevention of coronary disease.
• Diabetes: Reviews in this group included management of type 1, type 2, and gestational diabetes and a target range of activities from regulating glucose levels to promoting physical activity.
• Respiratory disease: This category included reviews of telehealth interventions for managing chronic obstructive pulmonary disease (COPD), cystic fibrosis, asthma, and lung transplantation.
• Mixed chronic conditions: These reviews considered that the uses of telehealth are similar across chronic conditions and included studies conditions such as asthma, hypertension, diabetes, COPD, and kidney failure in their reviews.
• Physical rehabilitation: These reviews included telehealth uses for rehabilitation for stroke, traumatic brain injury, or multiple reasons in children or adults.
• Behavioral health: Telehealth was evaluated in these reviews for mental health treatment for unspecified conditions, treatment of depression and anxiety, and addiction treatment, including substance abuse, smoking cessation, alcohol abuse, and pathological gambling.
• ICU or Surgery: This is a comparatively diverse category, including use of telehealth to allow physicians to advise on ICU patient management or surgery remotely and to facilitate communication between NICUs and parents of preterm babies.
• Dermatological conditions: This category included a review of studies that focused on treating several skin conditions.
• Preterm birth: This refers to telehealth interventions designed to monitor maternal and fetal health and prevent preterm birth.
• Mixed: When a review included uses for a wide range of conditions it was assigned to this category.
• Burn care: This refers to telehealth interventions designed to address the clinical needs of patients recovering from burns.

In creating the function categories we looked at the activities telehealth was used for and, when appropriate, what activities telehealth supplemented or replaced. We grouped the reviews in to the following categories:

• Remote patient monitoring: This category included interventions that are often called by other names such as home telehealth, or home telemonitoring, but it is broadly defined to also include remote monitoring and management of patients in other settings such as patients in critical care units or patients during transport to the hospital. Home telehealth may require special devices or may use computer applications and networks while in hospital monitoring may include video as well as transmission of data from monitoring devices. The key characteristic is that it involves the collection of data about a patient, usually physiological data such as blood glucose, weight, and blood pressure over time, and this data is transmitted to a health care provider or care team which reviews the data and adjusts care (often medications) based on this data.
• Communication and counseling: This category included the use of technology to facilitate the exchange of information between a patient and health care provider as well as the provision of advice. This could be synchronous, as is the case with videoconferencing and chat or asynchronous such as via Web sites or email. These interventions are often designed to increase access and can be used to replace or supplement face-to-face interactions with health care professionals.
• Psychotherapy: This differs from general or limited counseling and includes the use of technology to provide a course of treatment for a mental health condition.
• Consultation: This category was applied to interventions designed to facilitate involvement of another provider, often a specialist, across time and/or distance. While the patient may or may not be involved in the consultation, the consultation was required to be about a specific patient in order to differentiate this from training (which would not meet our definition of telehealth).
• Telementoring: This category was similar to consultation but refers specifically to the use of technology to allow a remote provider to view and advise on a procedure being conducted in another location in real time.
• Telerehabilitation: This included any type of rehabilitation services delivered via technology so the patient can be in a different location or can be engaged in rehabilitation activities at different times.
• Multiple Functions: In some cases, the intervention included more than one function and in these cases we classified it as mixed.

The second largest group in the clinical focus classification is “mixed chronic conditions.” These nine systematic reviews all defined their inclusion criteria such that the reviews either combined studies of several individual conditions, included primary studies with patients with more than one condition, or both. For example, the review by de Jong that evaluated internet communication between health providers and patients with chronic conditions included studies in which all patients had the same condition (e.g., diabetes), studies in which patients had related conditions (e.g., chronic neurological conditions), and studies in which included patients had different conditions (e.g., chronically ill women with a variety of clinical conditions).⁴⁶ The similarity in all these reviews was that their scopes were limited to chronic conditions.

Mixed conditions was the label given to reviews that included a wide range of conditions, all of which may not have typically been considered chronic. In many cases these reviews focused on a particular technology or health care function and included studies from varied patient populations. For example, a review of electronic patient portals included studies with populations undergoing in vitro fertilization, diabetes, and congestive heart failure, and patients without specific conditions.³⁸ Another review of electronic symptom reporting included studies of patients with several conditions including cancer and diabetes.⁵⁶

Telehealth function included a similar category: multiple functions. Ten reviews were coded this way when the included studies stated telehealth was used for more than one function. For example, several reviews had a focus on a specific technology: video conferencing^74,78 or the internet,⁴⁹ and the technology was used to communicate with, monitor, and treat patients.

Data from Table 2 as well as the conclusions of the included systematic reviews were used to generate the bubble plots presented later in this report.

Figures 6 and 7 graphically present the distribution of included systematic reviews across the three major characteristics of the reviews (clinical focus, telehealth function, and telehealth modality).

Figure 6

Distribution of clinical focus across included systematic reviews. ICU = intensive care unit

Figure 7

Distribution of telehealth function across included systematic reviews.

Figure 6 depicts the distribution across clinical focus areas for the included reviews. Taking into account the number of reviews, primary studies, and patients, the most common clinical focus areas studied were cardiovascular disease (12 reviews). ^{41,42,45,47,67,73,81,82,84,87,89,95} The next largest group was mixed chronic conditions (9 reviews),^{46,48-50,54,60,68,74,76} followed by diabetes (8),^{43,53,63,69,80,88,91,92} behavioral health (7),^{39,51,65,70,72,90,93} and mixed conditions (6).^{38,56,59,64,78,86} Focus areas with five or fewer included systematic reviews were physical rehabilitation (5),^{54,57,62,71,85} respiratory disease (5),^{44,55,58,66,83} ICU or surgery support (3),^40,61,75 burn care (1),⁷⁹ dermatology conditions (1),⁹⁴ and preterm birth (1).⁷⁷ Over one-quarter of included systematic reviews (26%) focused on mixed chronic or mixed but not exclusively chronic conditions.

Figure 7 depicts the distribution of the function the telehealth interventions perform in health care delivery. The included reviews examined telehealth used to provide treatment, monitor patients' signs and symptoms, or facilitate communication between provider and patient. These functions could replace or supplement in person service delivery. Telehealth was most frequently used for remote patient monitoring (17 reviews)^{41,42,45,54,55,58,66,68,73,76,77,81,83,87-89,91} and communication and counseling (14 reviews).^{38,43,46,56,59,60,63,64,67,71,72,75,92,95} Ten reviews combined research on multiple functions,^{44,49,50,53,69,74,78,80,84,86} seven summarized studies in which telehealth was used for deliver psychotherapy,^{39,48,51,65,70,90,93} and five reviews focused on telerehabilition.^{54,57,62,82,85} Four reviews examined studies in which telehealth was used to provide consultations about patient care^47,61,79,94 and one review focused on telementoring.⁴⁰

Indicators of Rigor of the Reviews

In Table 1 we reported selected characteristics of the included reviews that can be interpreted as indicators of the rigor or utility of the review. Specifically assessing the results across studies using either a “strength of evidence approach” or quantitative synthesis (i.e., a meta-analysis) are of interest, as reviews that incorporate these approaches may be more in accordance with contemporary standards for high-quality systematic reviews. While meta-analyses can be done poorly it may produce results and conclusions that are more definitive and easier to interpret.

In Tables 3 and 4 we report the percentage of included systematic reviews that used these approaches (strength of evidence and meta-analysis) as well as the number of studies in the reviews that were RCTs, according to clinical focus and telehealth function. While it is possible for RCTs to be of poor quality, randomized studies are generally considered to be higher in the hierarchy of evidence than observational studies and a preponderance of RCTs is often an indication, albeit imperfect, both of interest in the topic and the quality of the evidence.

Table 3

Data synthesis methods used in systematic reviews by clinical focus.

Table 4

Data synthesis methods used in systematic reviews by telehealth function.

Table 3 reports these systematic review characteristics by clinical focus. From this table it is possible to see that some clinical areas, such as burn care and ICU/surgery support, had been the focus of at least one systematic review, but that these reviews contained no or few RCTs and have not included meta-analyses. However, of the three reviews about ICU/surgery support, two included a strength of evidence assessment. Reviews of other topic areas also had a large proportion of RCTs among the included studies such as cardiovascular disease (85%) and diabetes (82%), and about half of the systematic reviews for these clinical focus areas reported strength of evidence (58% and 50%, respectively).

Table 4 includes the same information by telehealth functions. Reviews of telehealth for consultation and telementoring had fewer RCTs. Reviews of communication and counseling studies contained 88 percent RCTs.

We also evaluated the included reviews (n=58) for relationships between the conclusion (i.e., whether the telehealth provided benefit) and several independent variables; use of quantitative analysis (meta-analysis vs. not); use of strength of evidence (or not reported); and type of outcome (clinical, cost or utilization, or combined). Conclusions (dependent variables) were defined with two approaches: 1) benefit vs. no benefit and 2) reported positive or negative conclusion vs. no clear conclusion. All included studies were code for these independent and dependent variables. Chi square tests were used to explore differences between the observed counts and the expected counts using SPSS^® (IBM SPSS^® Statistics for Windows, Version 23). We found no statistically significant relationships (p>0.05) across all the chi-square analyses; however, the cell sizes for some comparisons were less than 10, suggesting that this quantitative approach was not appropriate for the analysis of this number of reviews and variables. To better understand the relationships between clinical condition, telehealth function, and effectiveness, we used qualitative approaches and the graphical presentations included in the next section.

Evidence Map Core

The bubble and intersection plots in the sections below constitute the core of our evidence map and help to clarify the type of evidence that exists on telehealth and how useful it is for policymaking and clinical decisionmaking. We constructed the bubble plots for clinical focus and for telehealth function. After examining the results, we also created an intersection plot in order to examine how clinical focus and function overlap. Combined, we believe these provided the insights that most closely matched our stated objectives and questions and provided the best way to summarize and assess the state of the evidence about telehealth. In this section we presented a more detailed analysis after an overview of the three plots.

Clinical Focus

Figure 8 is the bubble plot by clinical focus. In this plot each bubble is a clinical focus area. The y-axis is the number of patients in studies in the systematic reviews, so the higher up the bubble is on the grid, the more patients were studied. The lists of studies were deduplicated, so that each patient is counted only once within a bubble. The size of the bubble is the number of studies included in the reviews, again with each study counted only once when determining the size of the bubble. The color of the bubble represents the percent of the reviews that included strength of evidence assessment. The horizontal placement along the x-axis is determined by weighting the overall conclusion of each review (coded as 0=no benefit, 1=unclear, 2=potential benefit, and 3=positive benefit) by the number of studies in the review. As stated above in the Methods section, this weighted estimate of reported effect was created by multiplying the overall conclusion code by the number of studies in the review and then averaging across the reviews for the clinical area or function. Bubbles more to the right indicate more positive findings while bubbles to the left represent findings that include more unclear conclusions or more reviews reporting no benefit. While the weighting does not create a value with absolute meaning, it allows comparisons of the consistency and direction of the conclusions across clinical areas.

Figure 8

Telehealth literature map of systematic reviews by clinical focus. Bubble size reflects the unduplicated number of individual studies included in the systematic reviews about that clinical focus. The number label on each bubble is the number of systematic (more...)

As noted above (Table 2), the research volume, as measured by both the number of studies and the number of patients, is largest for mixed and mixed chronic conditions, followed by cardiovascular disease and behavioral health. This finding is represented on the plot by the fact that the bubbles are large and higher up on the y-axis. The mixed and mixed chronic condition bubbles are also farther to the right, indicating the conclusions of the reviews were that telehealth consistently provides benefit. The bubble representing diabetes shows that it is the single condition with fewer studies (the bubble is smaller) than mixed chronic conditions but about the same number as cardiovascular disease. However, diabetes studies included with fewer patients (the bubble is lower) than the cardiovascular disease studies but the findings were more positive findings (the bubble is farther to the right).

Telehealth Function

In addition to the evidence map by clinical focus, we also looked at the evidence by telehealth function. Figure 9 represents the same included systematic reviews as shown in Figure 8 except the reviews are summarized by the function telehealth played instead of clinical focus.

Figure 9

Telehealth literature map of systematic reviews by function of telehealth. Bubble size reflects the unduplicated number of individual studies included in the systematic reviews about that clinical focus. The number label on each bubble is the number (more...)

Each bubble is a function of telehealth. The other variables are the same as in Figure 8. The y-axis is the number of patients in a deduplicated list of studies in the systematic reviews for that function; the size of the bubble is the number of unique studies included in the reviews about that function; and the color of the bubble is the percentage of reviews that include strength of evidence assessment. The horizontal placement along the x-axis is determined by weighting the overall conclusion of each review by the number of studies in the review (bubbles more to the right indicate more positive findings while bubbles to the left represent finding that that are unclear or found no benefit).

In this bubble chart, communication and counseling is the function bubble highest and farthest to right, indicating the most reports of positive benefits of telehealth when used for these purpose and that the studies in these reviews contained the highest number of patients among the function categories. Remote patient monitoring is lower than communication, as these studies included fewer patients, but it is higher than other functions. Remote patient monitoring is also toward the right, indicating that most reviews about remote patient monitoring conclude that telehealth provides benefits in quality of care or in utilization.

Reviewing the bubble plot provides a means of both comparing the characteristics of available evidence across topics and identifying areas where systematic reviews are not available to support decisions. The next steps in our analyses and mapping were designed to explore where clinical focus and function overlap.

The Intersection of Clinical Areas and Telehealth Intervention Function

Much can be learned and several conclusions drawn from examining the literature on telehealth by clinical area/population of patients or by the health care function telehealth serves. However, examining the intersection of clinical areas and functions provides additional, more finely grained and potentially more useful insight for determining which telehealth intervention(s) could or should be used in specific patient populations. Figure 10 displays how the evidence clusters by telehealth clinical area and function. Each of the systematic reviews that provide evidence for the intersection of the clinical area and function are represented in the corresponding cell by a circle that is shaded to represent the overall conclusion of the review. The number of studies in each review is included to the right of the circles. This intersection plot demonstrates how the research evidence about telehealth clusters into a few clinical area/telehealth function pairs, the extent to which the conclusions are consistent within and across these pairs, and the volume of research for each pair.

Figure 10

Evidence from systematic reviews: the intersection of clinical focus and telehealth function.

For example, the pair with the most reviews is cardiovascular disease and remote patient monitoring. There are seven circles in this cell representing seven systematic reviews. The shading indicates that the conclusions included five reviews finding telehealth provided benefit, one citing potential benefit, and one reporting no benefit. The reviews range in size from 4 to 30 studies.

Identifying subgroups of reviews allows more details to be examined in order to better understand patterns and diversity within the pairs. Empty cells include some potentially important topics for which we did not identify any reviews, and therefore could be topics for future reviews if primary literature is available. Some empty cells are intersections that are likely not applicable (e.g., psychotherapy and physical rehabilitation).

The diabetes and communication pair provides an example of the diversity of reviews even within a cell, the range of information available, and the challenges researchers conducting reviews and users of the reviews face. Three reviews concluded that telehealth resulted in benefit or potential benefit. One review summarized studies of social networking services in diabetes care and concluded that their use was feasible and effective.⁹² Another review focused on how mobile phones were used in several ways to provide support and encouragement for patient self-management activities such as monitoring glucose, exercise, and maintaining diets, and found strong evidence of improvement in glycemic control in all patients, but the strongest for type 2 diabetic patients.⁶³ A third review reported potential benefits based on included studies that evaluated the use of a range of technologies to promote physical activity as part of type 2 diabetes management and concluded that telehealth is effective but that additional interventions were needed to sustain adherence, noting that the high dropout rate also raised concerns about potential bias in the results.⁴³

Reviews of telehealth for diabetes that included multiple functions varied in that the conclusions were less strong, with two concluding the evidence was unclear and one reporting potential benefits from telehealth. One study in this clinical focus-function pair included studies of different technologies (e.g., electronic messaging, Web sites, and video conferencing) used to support glycemic control in patients with type 2 diabetes, but concluded that the evidence in their review was “unconvincing” due to concerns about publication bias and a small effects.⁸⁰ Another review related to diabetes was in the multiple function group as it included studies of mobile phone use both for communication as well as remote patient monitoring. This review concluded that telehealth shows promise in this area but the evidence lacked rigorous study designs. Specifically, studies had insufficient sample sizes and short interventions and followup periods.⁵³ A third review of Web-based interventions for type 2 diabetes concluded that the Web could be used for behavioral interventions and to support self-management, however, the favorable results were enhanced if these were supplemented by other interventions such as case managers or mobile phone support and followup.⁶⁹

Examining this plot shows where synthesized bodies of evidence (i.e., systematic reviews) about telehealth are available and allows more in-depth examination of details such as those included above about telehealth for diabetes and communication. Considering the empty cells or those with few or limited reviews allows consideration of the importance of these areas and whether they are gaps that should be addressed in future reviews and/or primary research. In the next sections of the results we identify gaps and delve more into selected topics. Then in the Discussion section we use this information to create categories related to the sufficiency and need for research in selected areas.

Gaps and Priority Topics

Evidence Gaps

In order to identify which clinical and functional focus areas were not covered in the included systematic reviews, we assembled lists of telehealth practice domains generated by selected organizations and in reports on uses of telehealth, reviewed the notes from our KI interviews, and drew on our team's experience and expertise. Examples are included in Table 5. Certain domains on these lists may not be relevant if they do meet the definition of telehealth used for this report. For example, remote health care data management and some ancillary telemedicine services may not involve or augment an interaction between a provider and patient or interactions among providers about a specific patient, and would not be included here. It is also important to note that there is no definitive or authoritative list of domains, that these lists do not exactly correlate with our clinical focus areas and our definition of telehealth, and that the domains across these lists may overlap. This could be problematic if mutually exclusive categories were needed. However, for our purpose, which is to identify areas where systematic reviews that could support decisions are not available, these are useful. An initial review of these lists led us to identify certain areas that were not represented in our included reviews. For example, one such area is urgent/primary care.

Table 5

Examples of telehealth practice domains from four sources.

In order to determine if systematic reviews were underway on additional topics, we searched PROSPERO, the international prospective register of systematic reviews maintained by the University of York, Centre for Reviews and Dissemination.⁹⁸ We searched from August 1, 2013 through February 2016 for any ongoing reviews with the following words in any field: “telehealth” OR “telecare” OR “telemedicine” OR “eHealth” OR “mHealth.” We reviewed the titles and identified 82 registered reviews that were listed as ongoing and appeared to be relevant. A list of the topics covered and the number of reviews on each topic is included in Appendix H. The most frequent specific topics of these reviews in process that were not well-represented in our included completed reviews are weight loss, cancer, and maternal/child health. Other topics such as diabetes (6 reviews in progress) and mixed chronic conditions (3 reviews in progress) are represented in our included reviews, however, the PROSPERO entries means additional evidence syntheses will be available in the near future.

We also looked at the reviews we excluded (see Appendix D for a list and Appendix I for a table with the clinical focus area and the number of reviews). While these reviews did not meet our inclusion criteria, knowing there are substantial numbers of these reviews and which clinical areas they cover is useful because it could indicate that a more formal, rigorous, or differently structured systematic review could be conducted using the primary studies in these reviews. An assessment of the clinical focus areas covered in these excluded reviews revealed that they included additional areas not well-covered in our included reviews such as cancer, chronic pain, autism, and pregnancy (our map is limited to one review on uterine monitoring to prevent preterm birth). At the same time, this list also includes many of the areas that were covered in our included reviews, suggesting it is possible additional research exists that could be added to the body of evidence for these areas. It may also suggest that the utility of the evidence in these reviews could be increased if the included studies were summarized and analyzed in a different way. Specifically, reviews could 1) include quality assessments of the studies they include and examine whether their conclusions would differ if only high-quality studies were included and 2) include subgroup analysis by clinical focus or telehealth function or modality for reviews that have a wide range of telehealth interventions.

We identified one example of this type of supplemental analysis in our literature search. The original review⁹⁹ was excluded by us because structured telephone calls were included interventions in the review along with videophone and telemonitoring and the results were not summarized by type of intervention. In the subgroup analysis completed and published later, Conway and colleagues⁸¹ reanalyzed the data by type of technology used in remote monitoring for heart failure. This allowed us to include the review for the telehealth interventions, and exclude the studies of telephone calls. It is likely that this analysis would also be more informative for practice and policy decisionmaking.

After reviewing the topics covered in the included systematic reviews, reviews in progress, and the reviews that were excluded, we identified three areas in which telehealth had been proposed as appropriate or studied that were not well-represented: 1) triage for urgent care/primary care, 2) maternal health, and 3) pediatric cancer and chronic pediatric health conditions, and one area, 4) dermatology, where telehealth has been widely used, but the focus of the research included in reviews had been on diagnostic accuracy/agreement, not patient outcomes. We conducted a targeted search for primary studies on each of these topics (from 2006) and reviewed the abstracts for clinical focus, telehealth function, and modality. (Search strategies appear in Appendix A.)

Triage for urgent care has historically been provided most often in the form of advice from a nurse by phone; however, recent studies suggest telehealth is playing an increasing role, particularly related to heart health. Telehealth interventions are being used help decide the level of care needed in different situations. These have been used to address the following indications: chronic heart failure, arrhythmias causing dizziness/presyncope, flu, and a variety of primary care indications. Telehealth interventions are being used for the following functions: communication, monitoring, and diagnosis of heart arrhythmia conditions. Several modalities are described in the literature: mobile phone images, patient portals, single-lead electrocardiogram, mobile applications, and continuous mobile cardiac outpatient telemetry. We reviewed 353 abstracts of primary studies but identified only five potentially relevant studies, two of which were about cardiac care and may overlap with our included reviews on cardiovascular disease.

Much of this literature on triage was about telephone-only advice or nurse lines, which is not telehealth as defined for our evidence map. Another subtopic identified that did not fit our definition is use of telehealth by first responders forwarding data to the emergency room about myocardial function or other cardiovascular problems so the emergency department can be prepared when the patient arrives.

For maternal health, we reviewed 129 abstracts and identified 33 articles that evaluated several telehealth functions (remote fetal monitoring, antepartum cardiotocography monitoring, triage, consultation, counseling and health promotion, communication, screening, and diagnosis) in managing the following: gestational diabetes, perinatal depression, high-risk pregnancy, fetal and pediatric cardiology, pre-eclampsia, pregnancy termination, and fetal alcohol spectrum disorder. The studies used several modalities, including robotic ultrasound, videoconferencing, patient portals, text messaging/SMS, customized Web sites, mobile applications, and electronic logbooks.

We reviewed 61 abstracts from our search on telehealth for pediatric cancer and other chronic pediatric conditions. We found 12 potentially relevant studies—articles that performed the following telehealth functions: counseling in the form of support to families, remote case management, monitoring, psychotherapy, and consultation. These studies addressed the following indications: five were about pediatric cancer and the others covered several conditions including asthma, tic disorders, and other complex illnesses.

For dermatology, we identified and included one systematic review⁹⁴ and reviewed references from one narrative review¹⁰⁰ which included studies of clinical outcomes in addition to diagnostic concordance. As this suggests that research in the field may be expanding, we searched for studies of teledermatology that included clinical outcomes. We identified 315 abstracts on telehealth and dermatology of which only 15 included indexing terms for clinical outcomes. Our review of both the subset and the larger set of results failed to identify a discrete group of primary studies of teledermatology with clinical outcomes. The results included the studies in the reviews mentioned above, as well as abstracts of descriptive articles, articles not in English, feasibility studies, studies of diagnostic concordance, and studies with outcomes that were mixed or not clearly stated in the abstract. While further analysis of the literature would be needed to definitively confirm this, it appeared there were still few studies of teledermatology that include clinical outcomes.

Priority Topics

In creating the plots and tables, we sorted the included reviews into 11 different clinical categories and 7 different functions. While it is not unusual for bubble plots in literature maps to have 30 to 50 categories,^5,8 we also wanted to look across categories and summarize the results related to selected key policy and practice questions. In this section we describe in more detail the findings related to two subsets of the reviews that cut across categories and overlap, but that represent important approaches to considering telehealth.

Chronic Disease/Older Patients

A frequently cited target population for telehealth is patients with chronic disease, most of whom are older. The logic is straightforward, if over simplified here: patients with chronic disease are likely to require frequent visits for monitoring and management as well as support to self-manage their conditions. However, they may have barriers to access, or office visits may not be the best type of support and these challenges can be ameliorated by telehealth. Furthermore, by avoiding acute episodes (e.g., hospitalization for COPD) or adverse effects (e.g., amputations in diabetics), telehealth can reduce costs while increasing function and quality of life.

Thirty-one of the systematic reviews we identified for this report examined telehealth in either multiple chronic conditions or specific chronic conditions. This included the 9 reviews we classified as mixed chronic conditions,^{46,48-50,52,60,68,74,76} all 8 of the reviews that focused specifically on diabetes,^{43,53,63,69,80,88,91,92} 10^{41,42,45,73,81,82,84,87,89,95} of the 12 reviews on cardiovascular disease (those not considered chronic included 1 about acute myocardial infarction⁴⁷ and 1 about primary prevention),⁶⁷ and 4 of 5 about respiratory illness (1 included telehealth use in transplant),⁵⁵ including 3 about COPD^58,66,83 and 1 about cystic fibrosis.⁴⁴

Only two of these reviews, one about diabetes⁵³ and one about cystic fibrosis,⁴⁴ included studies with children as well as adults, and one other review included pregnant women with pre-existing diabetes.⁸⁸ While most of the reviews did not specify elderly, the patients in these studies were adults with chronic conditions and most were older. The majority of the systematic reviews (16 of 31) included telehealth interventions used for remote patient monitoring : six reviews focused on telehealth used to provide counseling or facilitate other communication and seven reviews included multiple functions, while one review examined the use of telehealth for psychological therapies and one considered rehabilitation.

As a group, the conclusions of the systematic reviews of research on the effectiveness of telehealth for chronic conditions were generally positive. Of the 31 studies, 13 (42%) reported benefits in primary or most outcomes, 11 reported potential benefits, 4 found no benefit, and 3 stated that the impact was unclear. Details on the findings from the thirteen reviews reporting benefits are included in Table 6 below. These reviews have characteristics associated with rigorous systematic review methods: 8 of 13 included some approach to assessing the strength of evidence across studies and 10 included a quantitative meta-analysis.

Table 6

Selected results: Systematic reviews of telehealth for chronic conditions.

Remote Patient Monitoring

Remote patient monitoring is a frequently studied telehealth function. Seventeen of the included reviews assessed telehealth use for monitoring and managing illnesses^{41,42,45,52,55,58,66,68,73,76,77,81,83,87-89,91} and five assessed multiple functions that included remote patient monitoring.^{49,50,78,84,86} Remote monitoring is of particular interest in considering telehealth because it makes new or significantly different forms of information and treatment available that can supplement and extend office-based care rather than replace face-to-face interactions. Specifically, many remote monitoring applications of telehealth allow patients to provide more data to providers, in a timelier way than could be obtained in outpatient visits, or allow patients to be monitored in their homes rather than in hospitals. With this information, providers can then tailor their recommendations and treatment. In this usage, telehealth changes not just the mode of care delivery (from face-to-face and in real time to something distant and/or asynchronous); rather it transforms the form of care.

In some studies, remote monitoring uses specialized devices to record and transmit data, but some types of remote monitoring may be done using more standard devices with specialized applications (e.g., mobile phones and computers with internet connections). Most, but not all of the reviews we identified used remote patient monitoring in the context of single or multiple common chronic conditions (e.g., diabetes, COPD, and congestive heart failure). The exceptions were a review of uterine monitoring of pregnant women to prevent preterm births⁷⁷ and monitoring that allowed parents to monitor babies in neonatal ICU and communicate with providers.⁷⁵

Of the 22 reviews that synthesized studies of monitoring, 10 concluded telehealth lead to positive benefits, 6 concluded benefits were possible, 2 were inconclusive, and 4 reported no benefit from telehealth. The 10 that reported benefits overlapped with those discussed in relation to chronic disease and are in Table 6 and the 6 that concluded potential benefit for remote patient monitoring also addressed chronic conditions.

Three of the four that reported no benefits explored very different applications of remote monitoring: a review of the use of home uterine monitors to prevent preterm birth that found no impact on maternal and perinatal outcomes,⁷⁷ a review of the addition of real time video as part of home care,⁶⁸ and a review that identified only one study of the use of monitoring for parents with babies in neonatal intensive care unit that did not find a significant difference in the primary outcome (i.e., length of hospital stay).⁷⁵

While the overlap of telehealth for chronic conditions and monitoring is not unexpected, it reinforces the potential of telehealth as a positive, transformative force in the care of chronic illnesses; one that may require more attention, development, or more adaptation for other uses.

Impact on Costs and Utilization

Fewer of the systematic reviews included in the map focused on costs or economic impact of telehealth exclusively. Out of 58 included reviews, 32 contained some cost/utilization outcomes and of these 7^{41,57,59,61,68,73,75} focused on these outcomes exclusively. In the remaining 25 reviews, clinical outcomes and cost or utilization were included. ^{38,42,44-46,49,50,52,54-56,58,65,66,77-81,83,84,86,87,89,94}

In general, the results reported for clinical outcomes were more extensive and the syntheses more sophisticated than those reported for costs or utilization. For example, there were fewer meta-analyses of utilization than of clinical outcomes, and there were very few true cost-effectiveness or cost-benefit analyses identified in the included reviews. It is also important to note that costs are perhaps more affected by the organization of health services, coverage policies, and health policy in general than clinical outcomes. For this reason, it is necessary to underscore that the studies included in currently available reviews were not conducted under new models of care. In addition, not all of these studies were conducted in the United States and costs may be very different in different health care systems (e.g., 2 studies of telehealth for gastrointestinal care were conducted in Sweden, and one study that focused on travel costs was conducted in Newfoundland).

Figure 11 is a variation of Figure 10. For this version we re-reviewed the 32 systematic reviews that included any information on cost, cost-effectiveness, or health services utilization. Each review is represented by a bubble placed in the grid consisting of rows for telehealth functions and columns for clinical focus areas. The shading of the bubble represents the overall conclusion: whether for these outcomes the research suggests that telehealth provides a benefit, a potential benefit, is inconclusive, or provides no benefit. For this figure, benefit is defined as cost savings or reduction in health services utilization. For reviews that included both clinical outcomes and cost/utilization, only the cost/utilizations outcomes are included here.

Figure 11

Evidence on cost and utilization from systematic reviews: the intersection between clinical focus and telehealth function.

Telehealth functions and clinical areas represented in the literature overall (including clinical as well as cost outcomes) are not all represented when we limit our focus to costs and utilization. None of the reviews contained cost or utilization outcomes for telementoring, and behavioral health is less well represented (1 review included costs/utilization out of 6 reviews included). In general, the cost/utilization results were less positive, in that they reported less benefit. Four reviews (13%) concluded telehealth provides benefit in terms of reduced costs or utilization, 11 (34%) potential benefit, 10 (31%) were inconclusive, and 7 (22%) found no benefit or increases in cost or utilization.

Table 7 contains a row for each included systematic review that included reports on cost or utilization outcomes and provides the key findings cited in the reviews. Most of these finding are not the result of complex, sophisticated, or even comprehensive analyses. A few meta-analyses on utilization such as hospital admissions and emergency department visits were available. Furthermore, very few studies considered the overall cost-impact or cost-effectiveness of an intervention; rather they documented individual costs or resource use measures taken in isolation. Comprehensive cost-analyses are needed to understand the full implications of telehealth in various situations. Several of the authors of the reviews underscored that cost information was incomplete or inconsistently reported.

Table 7

Telehealth cost and utilization: Findings from systematic reviews.

We re-examined our search of PROSPERO, a database of systematic review protocols for reviews in progress. From this we identified four pending reviews. Two planned to review the evidence for specific uses; costs of telehealth in home care and cost effectiveness of teleconsultations for patients in rural areas. The other two reviews were more general, examining the impact of mHealth and cost-effectiveness of health information technologies; this will only be useful for telehealth policy decisions if the reviews create subsets of health information technologies or include costs in their assessment of impact.

This initial review of information complied to create an evidence map suggests that information on costs is limited and costs and utilization may be an appropriate topic for additional research. While there may be primary studies that could be synthesized in a new review, there will be applicability challenges because current pending policy decisions are likely concerned with newer, integrated models of care, and existing research is likely to be based predominately on experiences in fee-for-service and nonintegrated-care organizations. More primary research is needed about how telehealth impacts costs and utilization of health services, although restrictions on funding cost-effective research may be a barrier to research that could address these current policy and practice questions.

Telehealth and New Models of Payment and Service Delivery

A key policy consideration is how telehealth might figure into new service delivery and payment models. Initiatives such as value-based purchasing and Accountable Care Organizations have been designed to create incentives for care that is high quality, accessible, and lower cost.¹⁰¹ Indeed, the US Centers for Medicare and Medicaid Services has a stated goal of increasing Medicare and Medicaid reimbursement to value-based models that share risk with provider organizations and give incentives for more coordinated and efficient care.¹⁰² The incentives are to improve or meet targets for performance on multiple measures. For example in the Medicare hospital value-based purchasing program, payments are adjusted based how a hospital scores on several measures in four domains (clinical processes, patient experiences, outcomes, and efficiency).¹⁰³

Many of the evaluations of telehealth we identified considered a variety of outcomes, including clinical outcomes and health services utilization (e.g., hospitalizations and urgent visits) and costs. While most studies did include more than a single outcome, none purposefully examined the impact of telehealth on all the domains or the groups of measures used in these new models. This is understandable, as the widespread use of these models of reimbursement is relatively new to health care in the United States, and studies of telehealth have not yet been able to assess the contribution of telehealth to value-based models.

While it may be possible to make some inferences with regards to value-based care across studies, doing so would require re-examining the literature and organizing a review around groups of measures similar to those used in these new models. However, this approach is unlikely to identify studies where the suite of measures is used within an organization, making it difficult to determine how telehealth and performance measures interact across these domains. Understanding this would require additional primary research that evaluates telehealth on all the relevant domains, though reviews of existing evidence could be used to inform the development of a demonstration or evaluation by suggesting which combinations of telehealth technologies and functions combined with specific patient populations should be the focus of larger studies.

Carrying out telehealth research under models of value-based care presents an important opportunity for future work, as any intervention or innovation that delivers care in more coordinated and efficient ways could be of great benefit to organizations entering into shared-risk models. For example, the processes and outcomes related to managing chronic disease could potentially be enhanced by some of the beneficial aspects of telehealth identified in the systematic reviews described above. Research would need to go beyond clinical factors and focus on delivering benefits not only from an individual-patient perspective, but from a population health management perspective as well.