The majority of patients included in the trial were men (approx. 89%), aged 62 years, and in the NYHA II class (approx. 68%; 20% in NYHA III and 12% in NYHA I).

In the context of the present analysis, it is important that the mean age at enrolment is less than 65–69 years, which is the interval for which the median survival data were presented above. In this sense, the assumption of 3.9 years life expectancy seems conservative, i.e., the actual average life expectancy may exceed 3.9 years.

Methods of assessing the effects

A CUA was performed, as the TELEREH-HF trial showed benefits of telerehabilitation for tertiary outcomes, suggesting an improvement in QoL. Two outcomes reported in the study were used in the present analysis: the NYHA class and SF-36 QoL assessment. One method was not chosen as a base case due to its complementary nature. The NYHA class can only take 4 values, which may adversely affect its ability to express the patient’s health through the utility accurately. The SF-36 is a generically-used instrument; however, it is more complex, which may be considered a disadvantage as converting the health condition factor to the utility would require the use of more parameters.

Each outcome was converted into the utility. For the NYHA class, a literature review was carried out (see Supplementary material) to evaluate the utility difference between the NYHA classes I and II, I and III, or II and III. Three studies that reported (respectively) large numbers of patients with the LVEF <40% were chosen for the utility parametrization [11–13]. In two of the studies the results were given in the form of utility differences between the NYHA classes (e.g. II and I), not the levels; however, this does not affect the outcome of the CUA as it is only based on utility differences [11, 12]. Nevertheless, to compare the results between studies, the utility level for the NYHA class I was adopted as the value of 0.823 [13]. It is also worth noting that among the studies found in the review, in the present analysis we eventually used those with the largest number of patients. Additionally, the results are very consistent across these studies.

The final data used to assess the utility of the NYHA classes are presented in Table 2. The average baseline utility derived from the NYHA class structure is 13% × 0.823 + 71% × 0.746 + 16% × 0.657 = 0.7419, and is 0.7538 after 9 weeks.

To convert the SF-36 into the utility, the following approach was used. Although state utility sets for the Polish population are available for the health states defined in the EuroQol-5-Dimensions-5-Level (EQ-5D-5L) questionnaire [7], there are currently no algorithms to assign this utility to the SF-36 questionnaire. Therefore, we used an algorithm developed for the United Kingdom [14].

By comparing the health utility calculated at baseline and after 9 weeks, we can estimate the change in the utility, which is then related to life expectancy, to calculate the therapeutic benefit in QALYs. The analysis assumed that the effect obtained during telerehabilitation lasts until the end of the patient’s life. This does not mean that a constant QoL is assumed from telerehabilitation to the end of life, but that the difference in life utility between patients using telerehabilitation and those receiving standard care remains unchanged over the lifespan. Meanwhile, based on the TELEREH-HF trial, it was conservatively assumed that telerehabilitation does not translate into longer life expectancy [5]. Finally, in the base case analysis, the further life expectancy adopted was 3.9 years. Considering the mean patient’s age of 63 years, and the median reported life expectancy for those aged 65–69 years of 3.6 years, this assumption also seems to be conservative.

Methods of assessing the costs

The costs of the initial (i.e., diagnosis and the first week of hospitalization) and the final assessment were not included, as these were identical for telerehabilitation and standard care. For telerehabilitation, the costs covered 8 weeks (i.e., 5 days/week, excluding weekends, when no telerehabilitation sessions were conducted), with a daily cost of 74 Polish zloty (PLN) (according to Regulation No. 133/2019/DSOZ issued by the President of the NHF). Thus, the total cost of telerehabilitation was estimated at 2960 PLN per patient.

In those undergoing standard care, most did not generate extra costs. However, rehabilitation was also used in 12% (51/425) of patients, 24% (12/51) of whom were rehabilitated in the hospital, and the rest (76% [39/51]) in an outpatient setting. Inpatient rehabilitation (7 days/week for 5 weeks) was estimated at 218.40 PLN/day (according to Regulation No. 133/2019/DSOZ issued by the President of the NHF), resulting in a cost of 7644 PLN per patient. Outpatient rehabilitation (24 days) was estimated as 74 PLN/day (according to Regulation No. 133/2019/DSOZ issued by the President of the NHF), resulting in a cost of 1776 PLN per patient. Finally, the mean cost of standard care was calculated as: 12% × (12/51 × 7644 + 39/51 × 1776) = 378.80 PLN.

Therefore, the cost difference between telerehabilitation and standard care is 2581.20 PLN. The costs are incurred in the first weeks after enrolment and are not discounted.

RESULTS

Base case analysis

Using the NYHA class shift approach, we obtained an average utility difference of 0.01187 between telerehabilitation and standard care. For the SF-36-based approach, the average change in utility in the telerehabilitation group during the ninth week was 0.01 (95% confidence interval [CI], 0.0016 to 0.0184) compared to 0.0028 (95% CI, –0.0056 to 0.0112) in the standard care group. This difference in the mean effect (i.e., 0.00726) was taken as the annual measure of telerehabilitation benefits. The cumulative effect difference over 3.9 years, at the discount rate of 3.5% per year, gives a total of:

• • 3.71 × 0.01187 = 0.044 QALY for the NYHA-based approach, and
• • 3.71 × 0.00726 = 0.0269 QALY for the SF-36-based approach.

The difference in the costs is 2581.20 PLN. Therefore, the ratio of the additional cost to the additional effect expressed in QALY, i.e., incremental cost-utility ratio (ICUR), is:

• • 2581.20/0.044 = 58 663.42 PLN/QALY for the NYHA-based approach, and
• • 2581.20/0.0269 = 95 955.39 PLN/QALY for the SF-36-based approach.

Incremental cost-utility ratio is interpreted as an average cost of getting an additional unit of the therapeutic effect when using the analyzed technology in place of the comparator. In Poland, the threshold differentiating cost-effective technologies from cost-ineffective technologies has been adopted by law and equals three times the annual gross domestic product per person, i.e., currently 155 514 PLN/QALY [15]. Therefore, in both approaches of calculating the effect in the form of utility, we found that telerehabilitation is cost-effective, i.e., the additional costs are justified by the outcomes gained.

Sensitivity analysis

Influence of lifespan and effect persistence

In the base case, a further life expectancy equal to 3.9 years, and the persistence of the effect difference between groups of patients throughout life was adopted. Supplementary material, Table 3 presents the impact of these assumptions on the results. Various variants concerning the number of years of the effect’s persistence have been presented (the life expectancy itself is irrelevant in the model, but it exceeds the duration of the effect). The effects obtained during these years were discounted the same way as in the base case.

The longer the duration of the effect, the lower the ICUR value, which proves the higher cost-effectiveness of telerehabilitation. Importantly, when parameterizing the effect based on the NYHA, even when the effect endures for 2 years, it can already be concluded that it is cost-effective. With the SF-36 approach, a 3-year duration of the effect is needed.

Considering both approaches collectively, a 2-year period of the telerehabilitation effect is sufficient for cost-effectiveness.

Lack of discounting

No discounting in the base case resulted in:

• • 3.9 × 0.01187 = 0.0463 QALY for the NYHA-based approach, and
• • 3.9 × 0.00726 = 0.0283 QALY for the SF-36-based approach.

In this case, the ICURs are:

• • 2581.20/0.0463 = 55 757.89 PLN/QALY for the NYHA-based approach, and
• • 2581.20/0.0269 = 91 63.38 PLN/QALY for the SF-36-based approach.

Probabilistic analysis

To determine the effect of stochastic uncertainty in the data on the results obtained, a probabilistic sensitivity analysis was performed using the Monte Carlo method. In terms of the NYHA-based approach, the impact of the target structure uncertainty of the NYHA classes was investigated. Ten thousand bootstrap samples were generated, and their ICUR values calculated, resulting from the observation of the population structure appropriate for a given sample after 9 weeks. As shown in Figure 1, for almost all Monte Carlo iterations, the ICUR value is below the applicable profitability threshold, which indicates a very high resistance of the obtained conclusions to the uncertainty associated with the use of random tryouts. Importantly, studies on the relationship between the NYHA class and utility cover a very large number of patients and are characterized by high convergence of the presented results. Therefore, this element of the model is not an important source of additional uncertainty.

Regarding the SF-36-based approach, the effect of uncertainty related to the value of the utility difference was examined for both subgroups of the TELEREH-HF study. The parametric bootstrap method was used. The mean and standard deviation of change in utility between baseline and week 9 were calculated for each group, yielding 0.00276 and 0.0846 for the control group, and 0.01 and 0.08375 for the telerehabilitation group. Based on the size of both groups (393 and 385 patients, respectively), a normal distribution was assumed to estimate the uncertainty of the mean: the standard error of the mean was approximately 0.0043 in both groups. The Monte Carlo simulation generated 10 000 utility difference values from this distribution. As shown in Figure 2, the uncertainty of the estimates using this approach is much greater; still, for approximately 68%, the ICUR value does not reach the profitability threshold in Poland, indicating the cost-effectiveness of telerehabilitation.

DISCUSSION

This study analyzed the economic viability of telerehabilitation in heart failure. When estimating health-related utility gain, two approaches were used: one based on the improvement in the NYHA class, and another based on QoL measured with SF-36. For both approaches, telerehabilitation was found cost-effective, i.e., the cost of one QALY gained is below the statutory threshold in Poland.

The specific results obtained for both approaches are different: such differences are expected in the case of instruments based on the subjective assessment of QoL. Additionally, for the NYHA and SF-36, different algorithms were used to convert their values into utility, as they are different instruments. Nonetheless, despite the quantitative differences, both approaches gave consistent findings.

Unfortunately, in the case of both the NYHA and SF-36, no Polish data was available that could be converted into utility values to specifically reflect societal preferences in Poland. In this context, it is important that the difference between the obtained ICUR coefficients and the statutory threshold is large and gives a large margin of freedom for the model parameters. For example, if we take 2.5 years instead of the assumed effect duration of 3.9 years, telerehabilitation remains a cost-effective option.

The greater uncertainty of the estimates based on the SF-36 is interesting. Perhaps this is due to the greater number of aspects covered by this instrument, leading to greater ‘noise’ in a given sample. Another reason may be that in the probabilistic analysis based on SF-36, uncertainty was considered for both the telerehabilitation and control groups, both in terms of the initial and target utility levels. For the NYHA-based approach, only the structure of the telerehabilitation group was considered uncertain, and only after 9 weeks of care. In conclusion, the results of both the base case analysis and the sensitivity analysis indicate the cost-effectiveness of telerehabilitation.

The compared groups were not homogeneous. In the telerehabilitation group, some patients did not follow the doctor’s instructions. In the group of patients who did, the results expressed by SF-36 were slightly better (utility increases by 0.0107 vs 0.01), indicating a slightly larger difference in the change in utility between the telerehabilitation and standard care groups (by approx. 10%). Additionally, different results were reported for the participating centers, which could be interpreted in a future post-hoc analysis with corresponding limitations. Moreover, some patients in the control group received rehabilitation, and likely achieved better results; thus, the comparison is conservative for telerehabilitation. It should be noted the analysis also covered the costs of extra rehabilitation in the control group, so the overall impact on ICUR should not be significant.

The analysis assumed a constant difference in utility gain related to telerehabilitation over standard care, which might be perceived as a less conservative approach. To explore that assumption, we estimated the necessary duration of effects, both in the NYHA and SF-36, to maintain the cost-effectiveness of telerehabilitation. The estimated 2-year duration of the effects represents roughly half of anticipated mean survival.

Telerehabilitation has been shown to be cost-effective in Belgium [16]. In the randomized study of total 126 patients, a 6-month telerehabilitation resulted in QoL improvement, which translated into incremental QALY gain of 0.22. This stronger effect may be due to longer duration of telerehabilitation itself, longer follow-up period, greater sensitivity of the EQ-5D instrument, or the sensitivity of the utility attribution algorithm to various health conditions to telerehabilitation outcomes. Lower costs were additionally observed in the telerehabilitation group, mainly due to lower costs of rehospitalization. In the TELEREH-HF study, no such differences were observed.

Lower costs in the telerehabilitation group were also observed in an Australian study [17], although no improvement in the QoL was found, which may reflect its significantly shorter follow-up than in the Belgium study [16]. Telerehabilitation was reported cost-effective due to the savings generated. Similar results were reported for New Zealand [18].

In Denmark, an analysis of the effectiveness of telerehabilitation compared to traditional rehabilitation also showed that the additional costs generated were small (approx. 1700 EUR) [19]. However, the differences in QALY over the one-year horizon, although in favor of telerehabilitation, were not significant. This study used the SF-36 to measure QoL, which may suggest a low sensitivity of this instrument in the context of the considered intervention.

The economic modeling reported savings generated by telecare and an additional effect of approximately 0.03 QALY per year, although authors stressed that the financial benefits may not apply to patients with low baseline risk [20].

The benefits of telerehabilitation not included in the present study should also be noted. Better health outcomes can lead to lower patient resource use and costs. Telerehabilitation may improve access to healthcare, as care can be provided regardless of whether a patient lives close to the rehabilitation center. Thus, telerehabilitation may increase the equity between patients. Telerehabilitation may also increase the possibilities of using the existing base by releasing resources in outpatient clinics and hospitals, which may motivate service providers to promote telerehabilitation care.

CONCLUSIONS

The telerehabilitation compared to standard care was found to be cost-effective, i.e., additional costs are justified by the clinical effects gained. The additional cost of gaining one year of healthy life is in the range of 58 000––96 000 PLN/QALY, depending on the adopted approach, clearly below the profitability threshold in Poland. The therapeutic benefits are driven by the improvement of the patients’ clinical condition, the NYHA class, or their QoL as measured by the SF-36 questionnaire.

Article information

Acknowledgements: The authors thank all medical and technical staff of the TELEREH-HF team.

Funding: The study was supported by the National Centre for Research and Development, Warsaw, Poland (grant number STRATEGMED1/233547/13/NCBR/2015).

Conflict of interest: The authors were supported by the National Center for Research and Development, Warsaw, Poland.

Open access: This article is available in open access under Creative Common Attribution-Non-Commercial-No Derivatives 4.0 International (CC BY-NC-ND 4.0) license, allowing to download articles and share them with others as long as they credit the authors and the publisher, but without permission to change them in any way or use them commercially. For commercial use, please contact the journal office at kardiologiapolska@ptkardio.pl.

How to cite: Niewada M, Tabor B, Piotrowicz E, et al. Cost-effectiveness of telerehabilitation in patients with heart failure in Poland: an analysis based on the results of Telerehabilitation in the Heart Failure Patients (TELEREH-HF) randomized clinical trial. Kardiol Pol. 2021; 79(5): 510–516, doi: 10.33963/KP.15885.

REFERENCES