The ageing of the global population is increasing the demand for medical services [1]. Although acute admission is the standard solution for treating acute serious illness, elderly patients have been recognised as being highly susceptible to functional decline, iatrogenic events, delirium, and hospital-acquired infections while hospitalised [2, 3]. Furthermore, adverse events occur during the hospital-to-home transition due to poor communication between the different sectors in the health care system [4].

There are substitutes to acute admissions, often referred to as hospital-at-home (HaH) service, defined as “a service that provides active treatment, by health care professionals, in the patient’s home of a condition that otherwise would require acute hospital in-patient care, always for a limited time period” [5]. In Denmark and other places the patient is sometimes offered a transfer from their home to the local nursing home for a short time period [6, 7].

These solutions have been evaluated in a range of randomised clinical trials [7, 8]. The included patients tend to have frequently occurring, relatively uncomplicated conditions, with well-defined treatments that could be delivered safely at home [7, 9]. Nursing care is always provided and physician inputs are available [7], either from a general practitioner (GP) [10, 11] or a hospital- specialist [6, 12]. The Patients might be admitted to HaH services either by the hospital specialist or the GP [5].

These solutions have been evaluated in a range of randomised clinical trials (RCT) comparing HaH with hospital admission [13]. Most of the studies favour the alternatives to hospital admission and demonstrate a reduction in subsequent admissions, faster recovery, and economic advantages [11, 14–17].

However, no studies have evaluated whether the patients in a HaH model should be cared for primarily by a hospital intern specialist or by the patient’s own GP. In a model, based on the hospital specialist, the patient will be initially transported to a hospital, examined by physicians, who are specialists in the conditions from which the patient is suffering with quick and easy access to advanced diagnostic procedures such as laboratory tests and diagnostic imaging, and the physician can easily confer with a variety of other specialists [18]. The hospital specialist will be responsible for the treatment and either visit the patient or otherwise be in contact with the patient and local community nurses during the next days. The hospital specialist may not be aware of the patient’s psychosocial conditions or familiar with the local community resources and health staff. In contrast, in a GP based model, the patient’s own GP might have the advantage of familiarity with the patient’s life situation and can follow the patient closely during the acute course of treatment, but may have less access to advanced diagnostic facilities or knowledge and experience at the specialist level.

Acknowledging these advantages and disadvantages, it is not obvious if the patient’s own GP or the hospital specialist should be responsible for the patients in a HaH setting. No randomised clinical trials have investigated this aspect so far. We thus did a pragmatic, randomised controlled trial (Acute Combined CarE for Seniors in Southern Jutland, (ACCESS)) to evaluate whether the patient’s own GP is more effective than a hospital specialist at reducing hospital admissions without affecting the recovery or death rates in elderly patients with acute medical conditions cared for in a HaH setting where the local community provides the nursing resources.

Between November 2013 and February 2015, 139 patients were referred from 72 (43% of all) GPs for the study and 131 patients were found eligible and included. We randomly assigned 67 patients to the community arm and 64 patients to the hospital arm (Fig. 1). Eight hospital specialists and 44 GPs delivered the intervention..

The primary outcome was that almost twice as many patients were admitted in the hospital specialist arm, 29 (45%) versus 16 (24%) of the patients in the GP arm within the first 7 days (Effect size 2.7, 95% CI 1.3–5.8; p = 0.01). The absolute risk difference was 21%; thus, nearly five patients needed the GP based service to avoid one case of admissions within the first 7 days. This significant difference was also found at 14 days and 30 days, but no difference was seen after 90 days (Table 2). The hospital specialist decided to admit 39% of all the patients within the first day after inclusion, while the GP only admitted 15% within the same period (p: 0.003).

No significant differences between the groups were found in death within the next 3 months or in municipal care level, changes in function tests, or perceived quality of life 1 week after the inclusion (Table 2).

We have shown that among a group of old and fragile patients with acute medical complaints partipating in a HaH service a hospital specialist admitted almost twice as many patients to hospital as a GP (45% versus 24%) within the first 7 days after the onset of the disease. This difference was also seen 30 days after the onset of the acute condition. We did not detect any differences in 7 days recovery rate or perceived quality of life between the groups or in numbers of death within 30 or 90 days.

The strength of our study is that it was carried out as a pragmatic study in the setting where the patients are usually identified and handled, involving the community nurses, GPs, and the hospital. There are however, some important limitations. A major limitation is that we did not manage to reach the expected number of inclusions. While our main result, difference in numbers of admission on day 7, remained significant despite a lower number of patients than expected, we were unable to find significant differences after day 30. We included a range of potentially adverse outcomes, like differences in death rates and recovery rates and were not able to show any significant differences in these outcomes, which might be due to the limited number of included patients.

Another limitation is that it was not possible to conceal the randomisation allocation for the patients or the health care professionals. Furthermore, since 44 GPs were involved, the individual GP gained only little experience in the HaH concept, and further experience with HaH would likely result in even higher differences between the GP and hospital specialist model. Finally, not all patients were interested or mentally able to participate in the physical and mental function tests on day 0 and 7, which weaken these results.

This study is the first to evaluate a GP versus hospital specialist based HaH model, and we are unable to compare our results directly with other findings in a similar design. Of the many HaH publications, few have examined the same group of elderly with a broad spectrum of acute diseases as we studied [8]: Two British studies revealed admission rates of 21% and 35% in a GP-based HaH in the patient’s own home [11, 29], quite similar to our findings. Two other studies of hospital specialist based HaH services from New Zealand and Australia found that 21% and 19%, respectively, were admitted to hospital [8, 12, 30], which was lower than our findings. These four studies suggest that admission rates are lower in a hospital setting than in the GP setting, which was also our expectation. However, the hospital-specialist HaH studies were based on a comprehensive hospital-outreach service in the patient’s home, including follow-up visits from hospital doctors, which was not an option in our case.

Our results raise some questions which deserve to be discussed in more details. Firstly, were the GP based HaH care truly preventing or rather delaying admissions of these patients? The prevention of hospitalisation seems to last for some weeks only. The design of the study does not allow us to judge whether later admissions were related to the acute illness which caused the inclusion the study or were related to a new situation. On the other hand, no later excessive admissions were seen in the GP group within the next months. Since the patients were not allowed to be included more than once within the observation period, we do not know if repeated HaH solutions managed by the GPs would have avoided further admissions. It might be argued that in this late stage of life even a single avoided or delayed admission would be appreciated by the patient if alternatives to admissions are possible.

Secondly, was the high admission rate of patients in the hospital specialist arm merely a result of differences between a hospital specialist and a GP strategy? If the patient had already reached the outpatient department of the hospital, it might be safer for a specialist to resort to a “keep and watch” strategy, where the patient was admitted to hospital rather than returned to home. The high number of admissions on the inclusion day and short-stay admissions in the hospital arm may indicate this. Although the hospital specialists had been thoroughly introduced to the HaH alternative to admission most of them had little or no practical experience with management of patients in a HaH setting. Increasing experience with HaH concepts might in future reduce the hospital specialist’s preference for admission. In contrast, the patient’s own GPs had activated the HaH team and initiated a “carefully watch at home” strategy, where the patient was closely observed at home andthe majority of the admissions in the GP arm occurred one to 3 days after the inclusion. However, since the GP’s practical experience with HaH was also very limited, and only increased a little in this project due to the low number of patients per GP, additional experience in future might reduce the admission numbers even more in the GP based HaH models.

Our study has a number of clinical implications. In contrast to previous results and our expectations, the GP based model was more successful than the hospital specialist model in avoiding hospital admissions when caring for the elderly patients in the HaH setting. We believe that the GPs’ advantage of familiarity with the patient and close collaboration with the municipal care system outweigh the hospital specialists’ specific knowledge of the disease and access to advanced diagnostic procedures.

For the fragile, old patients with high morbidity and mortality, it is an exhausting experience to be transported to the hospital outpatient ED, assessed by specialists and passing through diagnostic procedures including blood tests and often diagnostic imaging procedures. As our results indicate this does not increase the recovery rate or reduce death rate compared to the simpler GP controlled HaH alternative, but merely increases the number of hospital admissions. As long as the community care is able to provide immediate HaH services this patient group is better handled by the community and own GP. However, before the community and GPs assume responsibility for some of the elderly patients in HaH models, many problems must first be solved. HaH models are complex interventions and require close collaboration between the GPs, the municipal care system, and the hospitals and involves resource allocation, economy, cultural and legal issues [31]. In Denmark, for instance, resources and funds need to be transferred from the secondary health care, the hospitals, to the primary health care sector and GPs, if HaH models replaces hospital admissions. Furthermore, the patients and relatives need to be reassured that the HaH model is not an inferior health care solution, compared to a hospital admission. It also requires that the community nurses and GPs are properly trained in HaH care, and that legal aspects concerning the responsibility for the patients in HaH care is clear.

We believe that our results are also valuable outside Denmark in countries with similar health care systems. Future work should assess whether our findings are replicable in other community and own GP-based HaH models together with the economic consequences for the total health care system.

In conclusion, we found that a GP based HaH model including the patient’s own GP was superior to a hospital specialist based HaH model in avoiding hospital admissions among elderly patients with acute medical diseases. There were no measurable differences in terms of mental or physical recovery rates or deaths between the two models.