The economic effects of home telehealth can be discussed using direct or indirect measures of costs and cost savings. Direct measures include figures such as cost per patient, cost per telehealth unit, and annual costs. Examples of indirect measures are probability of patient hospitalization, number of emergency room visits, and other types of health service use. Though a formal economic analysis would typically assign these services a fixed dollar value, other study designs often report changes in service use without attempting to translate these changes into costs or cost savings.
The majority of economic analyses are of home telehealth programs in the United States. Their findings are of limited applicability in the context of the Canadian health care system. Though significant cost savings were seen in one analysis of a program that significantly reduced bed days of care, short-term implementation costs can be high enough to offset any savings from reduced use of health care services. A recent study reported that a clinical decision support system for patients and physicians was clinically advantageous, but not cost-effective. However, the authors note that widespread adoption of the system would mean possible economies of scale. As this study was retrieved in a final scan of 2011-2012 literature and was not subject to the same level of analysis as the other studies included in this review, we will not attempt to extrapolate beyond the conclusions provided by its authors. See O’Reilly et al. (2011) for more details.
Another study –the lone study to offer a full cost-effectiveness analysis – yielded some intriguing findings: although the program was only cost-effective for 23% of the user group as a whole, it had much higher cost-effectiveness among patients who were disabled or had a history of myocardial infarction. Marital status was also predictive of greater cost-effectiveness. Though more research into the cost-effectiveness of home telehealth for various populations is called for, it is possible that targeting select patient sub-groups could improve the average cost-effectiveness of home telehealth interventions.
The evidence on the effects of home telehealth on primary care use is inconclusive, but there appears to be a tendency towards increased use by those in home telehealth programs. It should be recognized that an increase in need-based primary visits can be a positive thing if it prevents emergency department visits and hospitalizations. At the moment, there is not enough evidence to say whether this is what is occurring.
Little information is available on the effects of home telehealth on emergency services use. There is some evidence that it has the potential to do so, but so few studies have examined the question that it is not clear whether reductions are widely achievable. Findings from one study suggest that HbA1c levels are a better predictor of emergency service use than use of home telehealth, suggesting that an intervention’s effectiveness in improving glycemic control may be the salient factor in is likelihood of reducing emergency service use. As home telehealth does generally appear to be effective in reducing HbA1c, one might expect further research to show reduced use of emergency services as well. However, long-term studies are needed.
Hospitalizations frequently decrease among users of home telehealth services. However, it has not been convincingly demonstrated that this improvement is attributable to the intervention.
At present, there is no reason to believe that home telehealth reduces specialist appointments.
Emergency Services Use
Summary: One multi-part study reported on emergency services use. There is some evidence that home telehealth has the potential to reduce use of emergency services, but so little research has been done that it is not clear whether reductions are widely achievable. While studies of the VA-CCHT program have yielded some thought-provoking findings, there are notable weaknesses in several of the study designs. The effects of diabetes telehealth applications on emergency service use require further investigation.
Study Details: The effects of the Veterans Affairs’ Care Coordination/Home Telehealth (VA-CCHT) home telehealth program for diabetes were reported in Chumbler et al. (2005a, b, c).
Chumbler et al. (2005a) was a single-group retrospective study of VA-CCHT. The probability that a patient would have 1 or more visits to the emergency department decreased significantly following enrollment in the program (11% reduction; p=.04). The authors note that this may be explained by regression to the mean.
Chumbler et al. (2005b) was a 2-group retrospective study in which 2 variants of the VA-CCHT program were compared. One group of patients was required to mail instant photographs of wounds to care coordinators on a weekly basis. These photographs were evaluated to determine whether there was any need for follow-up. The remaining patients received daily monitoring via 1 of the following: a hand-held messaging device, a 2-way audio/video connected telemonitor, or a videoconferencing system.
The weekly monitored group had a 15% decrease in diabetes-related emergency department visits at post-test when compared to the pre-test (Chumbler et al., 2005b). However, there was no significant difference in all-cause emergency department visits between the groups over the 12 months of the study. Differences between the groups at baseline make the significance of this outcome difficult to ascertain. All of the patients in the weekly monitored group had severe wounds that needed to be monitored, while only some of the patients in the daily monitored group had wounds that required careful attention.
Chumbler et al. (2005c) reported on a study with a retrospective, concurrent matched cohort design. They found a statistically significant decrease in the likelihood of 1 or more emergency department visits for both the intervention and control groups. There was also a statistically significant difference between the groups, favourable to the intervention group (effect size 0.29 ; p<.05). The results of a sub-group analysis at 1 of the sites are noteworthy: when controlling for HbA1c, the telehealth effect on emergency department visits to be insignificant.
Summary: (Re-) hospitalization outcomes were reported in 5 studies. Hospitalizations frequently decrease among users of home telehealth services. However, it has not been convincingly demonstrated that this improvement is attributable to the intervention.
Study Details: The majority of the evidence on (re-) hospitalizations was found in studies of Veterans Affairs’ programs that used home telehealth in conjunction with care coordination. Publications reporting hospitalization rates for interventions that took place within this context included Barnett et al. (2006), Chumbler et al. (2005a, b, c), Dang et al. (2007), and Jia et al. (2009). Ralston et al. (2009), in which patients were offered online educational material and access to their electronic medical records, also reported on system use.
Studies of the VA-CCHT programs indicate that home telehealth has a positive effect on all-cause hospitalization (Barnett et al., 2006; Chumbler et al., 2005a, b, c; Dang et al., 2007; Jia et al., 2009). Interventions delivered via hand-held home messaging devices, 2-way audio/video connected telemonitors, and videoconferencing systems were all found to decrease all-cause hospitalization. Note, however, that concurrent comparison groups and prospective measurement of exposure and outcomes were not used in any of these studies. For further details, see Table C.8.3.3: System Outcomes – Hospitalizations- All-Cause Hospitalizations (below).
Diabetes-related hospitalizations also appeared to decrease among participants in the VA-CCHT program, although generally not significantly (Barnett et al., 2006; Chumbler et al., 2005b, c; Jia et al., 2009). Whether this effect was attributable to the intervention is uncertain. Chumbler et al. (2005b) does not support conclusions of a significant intervention effect; the lack of baseline comparability of the 2 study groups, and the distinct objectives of the interventions they received, prohibit meaningful comparison. Jia et al. (2009) were the only researchers to report a significant decrease in diabetes-related hospitalizations in the intervention group, as well as a significant difference between the intervention and control group outcomes. This study is noteworthy because it was 4 years in duration, and therefore able to capture the long-term impact of the intervention on health care service use. See Table C.8.3.3: System Outcomes – Hospitalizations – Diabetes-Related Hospitalizations for more details (above).
Bed Days of Care
Bed days of care were tracked in Chumbler et al. (2005a, b, c), Dang et al. (2007), and Ralston et al. (2009). Findings are conflicting. Statistically significant reductions were reported in Chumbler et al. (2005a) and Dang et al. (2007). Chumbler et al. (2005b) found a significant reduction in bed days of care for patients receiving daily monitoring, but a significant increase for those receiving weekly monitoring. Chumbler et al. (2005c) found insignificant increases in intervention and control groups. Ralston et al. (2009) found no significant differences between intervention and control groups in inpatient days; however, the study was not sufficiently powered to detect such a difference, nor designed to produce 1. Additional details can be found in Table C.8.3.3: System Outcomes – Hospitalizations – Bed Days of Care (above). As mentioned above, limitations in study design mean that these results must be interpreted with caution.
Primary Care Use
Summary: Primary care use was reported in 3 studies, including a multi-part study of the VA-CCHT programs. The evidence on the effects of home telehealth on primary care use is inconclusive, but there appears to be a tendency towards increased use by those in home telehealth programs. It should be recognized that an increase in need-based primary visits can be a positive thing if it prevents emergency department visits and hospitalizations. At the moment, there is not enough evidence to say whether this is what is occurring.
Study Details: As with other outcomes related to system use, the bulk of the evidence on primary care use is traceable to studies of VA-CCHT programs (Barnett et al., 2006; Chumbler et al., 2005a, b, c). Ralston et al. (2009), though reporting on the number of visits made to primary care providers over the course of the intervention, prefaced findings with the statement that the study was not designed to reduce or powered to detect the effects of the intervention on use of the health care system.
Barnett et al. (2006) reported that the intervention group had an increase in care coordinator-initiated primary care clinic visits over the 24-month study, whereas the control group experienced a reduction in this measure. An 8.7 percentage point increase in the number of visits was attributed to the intervention (p=.04). Results from a sub-analysis of 1 study site that controlled for HbA1c contradicted this finding, reporting that the intervention group had a significant reduction in primary care clinic visits over the 2 years (59.0% to 21.0%; p<.001). The decrease within the control group was not as great and failed to reach significance (38.0% to 22.6%; p= .06). A statistically significant reduction of 21.0 percentage points was attributed to the intervention (p =.03).
The 3 analyses of the VA-CCHT program (Chumbler et al., 2005a, b, c), each 12 months in length, reported intervention impact on unscheduled and need-based primary care visits. In Chumbler et al. (2005a), a single-group study, there was a significant increase in the number of patients with 1 or more unscheduled primary care visits (46.82% to 63.50%; p<.0001). This translated into a significant increase in a patient’s probability of having 1 or more unscheduled primary care clinic visits (35% increase; p=.0004). Chumbler et al. (2005b) found ‘the change in the number of unscheduled primary care visits was 55% lower at 12 months’ in the daily monitoring group than in the weekly monitoring group (p. 5). This difference was significant (p<.01). However, the clinical characteristics of the groups were significantly different at baseline. In Chumbler et al. (2005c), the intervention group experienced an insignificant increase in need-based primary care visits, while the control group had a significant decrease (40.7% to 28.7%, p=.01). The telehealth effect was significant (0.18, p<.05).
Ralston et al. (2009) reported on the impact on primary care provider visits of an intervention that lasted 9 to 15 months. They found no significant difference within or between the intervention and control groups at pre and post-test observations. As noted above, the study was not designed to reduce, nor powered to detect an intervention impact on, health resource use or cost.
Specialist Care Use
Summary: Specialist care use is reported in 4 studies (Barnett et al., 2006; Boaz et al., 2009; Chumbler et al., 2005a, b, c; Ralston et al., 2009). At present, there is no reason to believe that home telehealth reduces specialist appointments more than usual care.
Study Details: Changes in use of specialist care use are reported in Boaz et al. (2009), Ralston et al. (2009), and a number of analyses of VA-CCHT programs (Barnett et al., 2006; Chumbler et al., 2005a, b, c). Chumbler et al. (2005b) reported a significant increase in ophthalmology visits in a group of patients using weekly monitoring (19% increase; p<.01). However, the difference between this group and the daily monitoring group was not significant. No other significant differences within or between groups were reported. See Table 8.8.3: System Outcomes – Specialist Appointments, below, for details.
Telehealth Costs and Cost Comparisons
Summary: Telehealth costs and cost comparisons were reported in 6 studies, none of which took place in Canada (Barnett et al., 2007; Boaz et al., 2009; Cho et al., 2009; Dang et al., 2007; Forjuoh et al., 2007, 2008; Shea et al., 2006). Significant cost savings were seen in one analysis of a program that significantly reduced bed days of care. Generally, the short-term implementation costs of home telehealth were high enough to offset savings from reduced use of health care services in some cases. There is some evidence that targeting select patient sub-groups could improve the average cost-effectiveness of home telehealth interventions.
Study Details: Few comparative analyses of telehealth costs have been undertaken. An exception is Boaz et al. (2009), in which intervention group patients made no unscheduled visits to the diabetes clinic during the 6 month intervention, while the control group made an average of 2.5 visits per patient. Each extra visit cost $50. Savings from reduced system use were therefore calculated at $125 per patient. The telemedicine system itself cost $150 per patient.
In Barnett et al. (2007), a VA-CCHT program was subjected to a cost-effectiveness analysis. Pre-post differences in total costs of care were divided by pre-post differences in quality-adjusted life years per patient. When calculations were based on the entire sample, a mean incremental cost-effectiveness ratio of $60,941 resulted. The intervention was cost-effective for 23% of the sample when each quality adjusted life year was valued at $20,000; for 27-34% of the sample in a valuation range of $50,000 to $100,000; and for 42% of the sample when the threshold was set at $200,000. Patients who were married were twice as likely to have cost-effective management at the various valuations, and patients with a history of myocardial infarction, hemiplegia, or paraplegia were nearly 5 times as likely to have cost-effective management. Patients with congestive heart failure were less likely to have cost-effective management.
Dang et al. (2007) also enrolled patients from a Veterans Affairs medical centre. The telehealth intervention studied had an average cost per patient of $2,385 for the first year. This included salaries, equipment, and other items associated with the program. Cost savings from the reduced bed days of care achieved by the intervention were estimated at $292,000 per year for 41 patients.
The intervention used in Cho et al. (2009) involved a specialized mobile device, or ‘diabetes phone’, that enabled transmission of blood glucose data. The expense of this phone was described as prohibitive and was identified as a barrier to clinical implementation. The authors anticipated that emerging technologies might provide lower-cost alternative to the device, thereby making widespread adoption more feasible.
The home telehealth equipment featured in the multi-part IDEATel studies (Shea et al., 2006) had an average unit cost of $3,425, plus $110 for the glucometer and cable. This price does not include health human resource costs.
Forjuoh et al. (2007, 2008) estimated that the cost of their Central Texas intervention was roughly $650 per patient. This included $200 for the personal digital assistant (PDA) used in the study, $25 for software, $65 for manual development, $220 for training and monitoring, $80 for HbA1C testing, and $60 in monetary incentives.
Recent Developments: A scan of material from 2011-2012, a time period not covered by our initial searches, found 1 additional article that addressed telehealth costs and cost savings (O’Reilly et al., 2011). In this study, use of a clinical decision support system for patients and physicians was associated with clinical improvements, but was not considered cost effective. However, the authors note that widespread adoption of the system would mean possible economies of scale. As this study was not subject to the same level of analysis as the other studies included in this review, we will not attempt further analysis. See O’Reilly et al. (2011) for a full breakdown of costs.