Analysis: What’s the impact of telehealth on hospital use?
Adam Steventon from the Nuffield Trust analyses the results of the much-debated Whole Systems Demonstrator WSD trial.
Of the five investigative strands studied in the Whole Systems Demonstrator (WSD) trial, the first (led by a team at the Nuffield Trust) examined the impact on hospital admissions, associated costs and mortality. Results from the formal cost effectiveness study and quality of life measures are due soon.
The rigour and size of the trial (with over 3000 participants) means that it constitutes a sizeable step forward. Over a billion records of administrative data were extracted and linked to make the analysis possible. As a result we were able to look not just at the position at the start and end of key trial – but also what happened throughout the year. Algorithms were used to standardise for risk and ensure the comparisons between telehealth and control groups were not biased.
In the trial telehealth was tested for people with diabetes, chronic obstructive pulmonary disease or heart failure. Patients used technology in their own home to record items like blood glucose level and weight, and transmitted their readings to health professionals working remotely. Advocates claim that this approach helps patients manage their conditions and enables faster response from professionals when needed. As a result, telehealth may improve the quality of care while reducing the use of expensive forms of care, such as admissions to hospital.
The findings suggest that the telehealth group had 0.14 fewer emergency admissions per person per year compared to the control group. However this figure must be interpreted with caution. Admissions appeared to increase for control patients shortly after joining the trial, perhaps because these patients felt anxious after learning that they would not receive telehealth. These caveats, which are explored in more detail in the article in the British Medical Journal, suggest the impact of telehealth might be different when deployed outside of a trial setting.
Our best estimate was that telehealth reduced the cost of hospital care by around £188 per person per year. These findings also require caution, as they were not statistically significant, so could have been the result of chance. We made no conclusions about the impact of telehealth on costs – there was no evidence from the trial of a reduction of costs, but this does not mean that telehealth has no impact.
We did observe fewer deaths among the telehealth group than controls, over the 12 months of the trial (4.6 per cent compared with 8.3 per cent). This has been seen as a strong motivator to deploy telehealth but, as the BMJ's editorial pointed out, the reasons for this reduction are not yet understood. Caution is therefore required. Other randomised trials of telehealth have found increases in mortality, so telehealth may not have the same effect if deployed elsewhere.
While these findings are valuable, decisions about telehealth should be based on the complete evaluation. The other four strands have not yet published and the impacts on other patient outcomes such as quality of life are clearly important. We should not assume that fewer emergency hospital admissions mean better quality of life. Further work will report on the impact of telehealth on other services such as primary care, as well as the cost of operating the telehealth intervention – not taken account in findings published so far. However as of writing there is not yet any firm evidence that telehealth will deliver any savings to the NHS, even if the cost of the intervention is reduced to zero.
Perhaps most importantly there are critical questions on how telehealth may have different impacts on different population subgroups. The low levels of emergency admissions observed in this trial (0.68 per person per year for controls, 0.54 for telehealth patients) are a reflection of the relatively broad eligibility criteria used. The trial was open to all patients with one of the three chronic conditions, so many of the participants had relatively low or moderate levels of risk. What we do not yet know is whether for example some higher risk patients may show greater impact. If that is the case the key will be in targeting the technology to the right types of patient.
Adam Steventon is a senior research analyst for the Nuffield Trust