To the Editor: We note with interest the studies published in the Journal by Richards et al and Ong et al.1,2 The authors conclude that treating pneumonia and pulmonary emboli in an ambulatory setting is safe for selected patients. However, this represents a large change in the conditions traditionally treated on this basis, from conditions that are associated with a very low mortality (such as cellulitis) to a subgroup of patients with potentially serious infections that are identified as being of low risk. We feel that safety is of prime importance in hospital-in-the-home programs because of limited or delayed access to acute medical care, and that both studies were underpowered to define this endpoint.

Both studies incorrectly quote previous work that suggests that the groups they have identified have mortality rates of up to 5% (for pulmonary emboli) and up to 9.2% (for mild to moderate pneumonia). Published data suggest that the mortality of mild pneumonia (with CURB-65 scores ≤ 2) is in the range 1.7%–3%,3,4 and that mortality from treated sub-massive pulmonary emboli is in the range 1.0%–1.3% within the first week.5 These rates, although seemingly small, are still much higher than that associated with the treatment of soft tissue infections on ambulatory care programs. Recurrent pulmonary embolus, in particular, may be sudden and unexpected. Although admission to hospital may not necessarily prevent these deaths, the additional trauma of a death at home, particularly soon after transfer to ambulatory care, may carry a higher significance in the minds of patients, their families and the public than a death in hospital.

We acknowledge that benefits for patients in ambulatory treatment programs need to be balanced against potential adverse outcomes. However, if these conditions are to be treated where access to medical attention may be delayed, it is imperative that informed consent be obtained from patients (including an awareness of the possibility of death), a mechanism be available for patients to summon urgent attention at any time, and patients and health care providers be aware that readmission to the hospital may be necessary in the event of clinical deterioration.

In reply: Thank you for the opportunity to reply to the letter from Cheng et al. The mortality figures we cited are correct.1 The cited article by Lim et al supports our statement that “Patients with a CURB-65 score of 0–2 have a low mortality (0.7%–9.2%)” (Table 4 shows mortality for CURB-65 score 0 is 0.7% and for score 2 is 9.2%).2 The 3% figure in the abstract is a summary measure obscuring the difference across the CURB 0–2 range — important information for anyone considering community management of community-acquired pneumonia where, we agree, safety is paramount. The rate of 1.7% cited by Cheng et al is for a modified CURB-65 score, which adds a further point, and thus is for the equivalent of CURB-65 scores of < 1.3

Cheng et al correctly observe our study was not powered to detect mortality differences. As explained in our discussion, mortality was not a primary outcome measure. With low mortality, large numbers are required to detect a statistically significant difference — the base rate of 3% in the validation study would require 10 602 patients in a randomised controlled trial to detect a 33% relative (1% absolute) increase in mortality. The study did provide for informed consent (including the possibility of readmission) and the ability to summon urgent attention.

Careful patient selection, routine twice-daily nurse and daily doctor visits, along with a highly trained nurse available by telephone 24 hours a day who can dispatch a doctor or nurse immediately, provides a structure that should match hospital care. Careful patient monitoring will detect failure to respond as expected.

It is important to treat in hospital those who will benefit, but not feasible to admit all with potential mortality risk (nor is there evidence of benefit). Hospitalisation also has risks. With this tool for predicting accurately who will suffer worse outcomes, it could be argued there has to be good evidence that better outcomes will result from continuing inpatient treatment of mild to moderate community-acquired pneumonia.

These wider issues are worthy of debate. There is an assumption by some professionals and consumers that hospital-sanctioned death is more acceptable, that everything possible has been done, and that community-based death implies unsatisfactory management. As a counterpoint to this, there is a clear patient preference for treatment in the home where possible. Avian influenza may, of course, drastically redefine our expectations about locus of care and of death.

In reply: We acknowledge the concerns of Cheng and colleagues regarding the safety of patients with pulmonary embolism (PE) treated in an ambulatory care setting. Caution is important as this is a relatively new area of treatment in ambulatory care compared with the management of deep venous thrombosis. The main objective of our study was to describe our experience in the management of PE in ambulatory care; it was not a randomised controlled study to conclusively define safety as such.

As stated in our paper, there have been reports of the management of PE in the ambulatory care setting.1,2 We now know that more than 90% of patients with sub-massive PE will have a good response to treatment. The challenge is to accurately define this group. The mortality rate we quoted of less than 5% was derived from a review article on prognosis of patients with PE.3 This article quoted three studies on sub-massive PE, one of which was referenced by Cheng and colleagues in their letter.4 We note also that the specific study that was referenced4 included patients with cyanosis and shock; these patients would have been excluded by our selection criteria.

We do not advocate management of all patients with sub-massive PE in the ambulatory care setting. It is also important to be conservative initially in the selection of these patients. There have been various studies examining prognostic indicators for PE, which we have referenced in our paper. There is evidence now that, for patients with specific prognostic indicators, the risk of death and adverse outcomes is significant and such patients should always be admitted.

The practice of managing patients with sub-massive PE should only occur in ambulatory care units which are appropriately resourced, have strict admission criteria and well defined protocols and specialist medical input, consistent with the recommendation of the British Thoracic Society.5 In the meantime, further studies are required before this becomes standard practice in ambulatory care or hospital-in-the-home units.