LAUSR

The article by Beaney and colleagues just published in the EMJ evaluates the outcomes from a national programme of home oxygen monitoring instituted by NHS England during the first waves of the COVID19 pandemic. This may sound familiar; we ran a paper by this same author group on this subject in our August 2022 issue. At first glance, you might think we accidentally reran the same paper. Both studies ask what appears to be the same research question: was there benefit? However, the studies are indeed quite different in what effect they are examining. The study just published looks at whether patients who enrolled in the monitoring programme had improved outcomes over those not enrolled. The previously published study assessed whether this national programme resulted in fewer hospitalisations and deaths in the country. In other words, what was the effect of the programme in reducing adverse outcomes from COVID in England? These are very different questions, and therefore we agreed with the authors on the need to present the findings as two separate papers. The bottom line is that, depending on how you look at it, the programme was and was not beneficial. In essence, the current paper reads like a standard observational study, looking at the outcomes of patients who were and were not enrolled in the programme. As an observational study where patients were not randomised, it is important that the authors adjust for the underlying characteristics of the patients who were and were not enrolled; factors such as age, comorbidities, sex and severity of illness. If after that, the group enrolled in the programme did better (fewer healthcare visits, hospitalisations or deaths), one can conclude that being in the programme was ‘associated’ with better outcome, but whether or not it caused better outcomes is something that cannot be said in an observational study. Those undergoing O2 monitoring may have benefited for other reasons or were less sick and while the investigators attempted to identify and measure such confounders, observational studies are always at risk of unmeasured confounding. The previously published populationbased study, also an observational study, touches on perhaps a thornier issue, which is whether the programme’s impact in the country was worth the resources expended on it. While undoubtedly every life saved is important, rolling out a nationwide programme during a pandemic requires significant resource use in terms of personnel, equipment and time (eg, time taken away from another patient while you explain the home O2 monitoring scheme, or salary for a programme administrator who might be able to instead run a prevention programme). If only a minority of people who might benefit enrol, is it worth the expense and time to set up this programme, or are there better ways to try to avoid hospitalisations and save lives.? This evaluation is similar to the analysis we use when we estimate a number needed to treat (NNT) for a medication; how many people need to receive this (expensive) medication to benefit one. In the case of the O2 monitoring programme, the question might be how many people we need to offer this programme to, and enrol, for us to save an extra life. There is no magic number here (and it very much depends on cost, and other tradeoffs) but clearly if every second person you offer the programme to enrols, and perhaps every sixth person has a better outcome, that sounds like a reasonable investment. But if its 50 people that need to be offered the programme for one to enrol (or if a busy ED or primary care physician fails to offer it), and of these, only a few benefit, you can imagine the NNT will become quite large and worth a rethink. Spoiler alert: in the treatment/nontreatment study published in this issue, deaths were 52% lower in those enrolled (95% CI 7% to 75%) compared with those not enrolled, while the odds of any ED attendance were higher at 37% (95% CI 16% to 63%) in the enrolled group, as were the odds for admission at 59% (95% CI 32% to 91%). However, we know from the populationbased paper that the O2 monitoring programme only enrolled 2.5% of the eligible cohort. In the previously published populationbased study, the authors found no association between mortality rates from COVID during the period that the monitoring programme was in effect, and there was a small increase in healthcare utilisation. It is not surprising that, with such a small proportion of eligible patients enrolled, an overall impact was not seen. But what the study published in this issue suggests is that for those who were actually enrolled, there was a benefit. Given the small numbers of patients enrolled in the O2 monitoring programme, one might rightly worry about selection bias in who actually was treated. That bias could have gone either way: if patients who had better prognoses at the outset were more likely to enrol in the programme, then the programme could have had no effect on their outcomes, and the difference in outcome we see is only because the enrolled patients were less sick to begin with. To a large extent, the matching strategy used by the authors suggests they compared patients in the montoring program who were equally sick (by objective parameters) to those not enrolled. And the authors argue that, if anything, emergency physicians would be more likely to try to recruit patients on the borderline to be in the programme rather than those who look well. So it does seem unlikely that the benefits for the enrolled patients are due to selection bias. However, this does lead to another point, which is that the O2 monitoring programme may only be beneficial in certain individuals (those on the borderline) and thus any future iterations would want to use this and other data to determine who should be enrolled. While it may be that the newer COVID variants are less likely to affect our lungs, and therefore, O2 monitoring may no longer be an appropriate method to decrease hospitalisations safely, these papers together raise policy questions that will persist in the next phases of the COVID19 and future pandemics, not just in England but worldwide. They mirror the way we evaluate screening and prevention programmes. While the programme may benefit the individuals who take part, if it cannot be deployed on a widescale basis, perhaps we should try another path. Therefore, it is worth the effort to acknowledge what happened Emergency Medicine, University of California San Francisco, San Francisco, California, USA