In our previous APHP direct-admission study, DA to an AGU was associated with greater effectiveness (lower hospital LOS, as well as lower likelihood of post-acute care transfer, including follow-up and rehabilitation care) than admission to an AGU after an ED visit (17). No significant association was found with the risk of ED return visit (17). In this economic evaluation, we aimed to assess the cost-effectiveness of DA to an AGU versus admission after an ED visit for the elderly to avoid a return ED admission. At baseline, we found a negative ICER (€-2788 per ED return visit averted), which means that DA was more effective in avoiding an ED return visit and less costly than admission after an ED visit. An acceptability curve showed that DA can be considered a cost-effective intervention at a threshold of €-2409 per ED return visit averted. It also demonstrated that if the payer is not willing to pay additional euros per ED return visit avoided, DA is cost-effective in 63% of cases, i.e., 63% of the 5000 ICERs are situated in the south-east quadrant. Thus, our results are strongly in favor of DA implementation.
To our knowledge, this study is the first cost-effectiveness analysis of DA to an AGU for elderly patients, compared with admission after an ED visit. Some observational studies have already shown that admissions to AGUs (compared with non-geriatric units) are associated with better outcomes and lower costs (23, 24). Another study, conducted on nearly 1 million ED visits resulting in over 187 acute care hospitalizations in California, found that periods of ED overcrowding were associated with 1% increased costs per admission (25). However, none of these studies reported ICERs, which are nonetheless essential to inform stakeholders' decision-making. In a context of limited resources, decision makers must consider the allocation of resources. If €100 is allocated to a new health program, for example to gain an additional unit of effectiveness due to the implementation of such a program (here, an ED return visit averted thanks to the implementation of DA to an AGU), it implies that the same €100 cannot be allocated to a competing health program (in the same or alternative field of health) (26). This is considered to be the opportunity cost (21). Because of our analysis of the uncertainty surrounding the cost-effectiveness ratios, it should be borne in mind that in 37% of cases, the payer will have to be willing to pay additional euros if they choose to favor DA to an AGU over admission after an ED visit. It is difficult to define what is an acceptable incremental cost-effectiveness ratio. The threshold for willingness to pay may vary depending on the context in which decisions are made, and this may be different between countries due to different health policies, organization, and financing of health care. We therefore used analytical tools such as acceptability curves, a guarantee that cost-effectiveness studies were of good quality, which can inform decision makers about the likelihood that a new health program may be cost-effective, based on a variety of the Willingness to Pay schedule. If we extend the reasoning, as Bourel et al. did in a cost-effectiveness analysis in a completely different field of care, should the payer decide to invest €100,000 in the DA of elderly patients to the AGU rather than continuing to hospitalize this cohort via the emergency room, there is a 68% chance of averting 100,000/1000 = 100 ED return visits (26). The results of such economic calculations favorable to the implementation of DA of elderly people to the AGU are reinforced by the fact that this group of patients is less likely to be discharged in follow-up and rehabilitation care than those admitted after an ED (17). Indeed, the daily hospitalization cost in follow-up and rehabilitation care is high, and the LOS is often long, on average 35 days in 2019 (27), before the patient returns to the institution or home.
While the results of the economic analysis are important to consider when choosing one intervention over another, there are other important considerations, such as the feasibility of DA intervention, especially in hospitals with problems related to access block and ED overcrowding (17). Increasing the total number of AGU beds, as well as follow-up and rehabilitation care beds, might be important levers (28–32). In a large study involving 17,111 patients experiencing acute hospital discharge delays in Canada (30), patients waiting for nursing home admission accounted for 41.5% of such bed days while only accounting for 8.8% of acute hospital discharge delay patients. This means that a small number of patients with non-medical days waiting for nursing home admission contribute to a substantial proportion of total non-medical days in acute hospitals. Some authors described the end of acute hospitalization as “push” rather than “pull” systems, patients being pushed to the next stage by pressure of patients behind them rather than pulled to the next stage (32). Higher availability of follow-up and rehabilitation care beds might help the transition to a “pull” system. Increasing the number of both AGU and follow-up and rehabilitation care beds would lead to an obvious increase in a hospital’s functioning costs. However, according to the results of our study, these investments could be offset by the costs of ED return visits averted and related re-hospitalizations. Feasibility of DA is also related to better management of patient flow over the entire geriatric pathway. General practitioners should play an important gatekeeping role for DA, but this is conditional on their availability. In Norway, which has a gatekeeper-based healthcare system, Blinkenberg et al. found that only 65% of the emergency-admitted patients came through the primary healthcare gatekeeping system (general practitioners and out-of-hours doctors) (33). DAs were more common in central areas (45%), where only 18% of referrals were from a GP. Among hospital inpatients admitted for unscheduled care in the UK, patients able to get a general practice appointment on their last attempt were more likely to have been admitted via a GP than after an ED visit (34). Better coordination between outpatient and inpatient care results in a reduction in avoidable costs (35).
This study has some limitations. The first, already mentioned in the APHP direct-admission study (17), relates to the comparison of effectiveness between the two intervention groups: the choice of DA vs. ED was not randomly assigned, and potential confounding by indication could bias our analyses. IPW weighting based on a propensity score was used to balance baseline characteristics between groups, although unmeasured confounding can never be ruled out in observational studies. Second, we were unable to value hospitalizations in follow-up care and rehabilitation, as we used the APHP Health Data Warehouse, in which patient data were not linked to that regarding follow-up and rehabilitation care in public and private hospitals, most often outside the APHP. The cost implications of this lower hospitalization in the DA group have been discussed above. Finally, whilst we could have considered the societal perspective, the method most often used as it is sufficiently broad to take into account all those affected by the treatments studied, it would have been necessary to estimate travel costs, personal expenses, productivity costs/sick days to qualify for such an analysis (36). The database we used was not designed for such an analysis and our payer perspective analysis follows Peter J. Neumann’s recommendation, according to which "more attention needs to be paid to the question of what cost data decision makers themselves find most useful” (36).