The present study addressed the association between the inter-hospital transfer of COVID-19 patients and the period of hospitalization or mortality. In accordance with most previous findings, transfer was not significantly linked to mortality. In contrast, the period of hospitalization was robustly associated with transfer, even after considering multiple factors that could have had an influence. These factors comprised clinical characteristics and complications, as well as the fact whether patients living alone at home. Patients without ICU stay showed an average prolongation of the period of hospitalization by more than 12 days, while in-hospital death was associated with a reduction by about 5 days. As reasons for transfer, capacity reasons were slightly more frequent than medical reasons. In patients with ICU stay, the period of hospitalization was again longer in transferred patients by about 12 days, but the result was dependent on the fact whether patients survived, since in-hospital death was associated with an average reduction by about 16 days. In these patients, capacity reasons for transfer were more than three times as frequent as medical reasons, besides ECMO. In all patients, the reasons for transfer were correlated with the change in the hospitals’ level of care after transfer. Capacity reasons were predominantly associated with lowering, and medical reasons with raising of the level. Taken together, these findings indicate that at least in the region and at the time analyzed, COVID-19 patients transferred between hospitals showed a markedly longer period of hospitalization than non-transferred patients did. To which extent the prolongation was due to the transfer per se or due to specific characteristics of the preferentially transferred patients and corresponding physicians’ decisions, cannot be conclusively inferred from our data, despite the consideration of individual risk profiles. Despite this, one should keep in mind the fact that inter-hospital transfers were statistically associated with an increased overall utilization of the health care system.
In-hospital mortality did not significantly depend on the fact whether a transfer had occurred or not, provided that the groups of ICU and non-ICU patients were analyzed separately. The separate analysis was necessary, as ICU patients were relatively more often transferred and at the same time showed a much higher mortality than non-ICU patients. The finding regarding mortality of transferred ICU patients is in line with the great majority of the previous studies on mortality in ICU patients performed in different countries at different periods of time [5, 7, 8, 11–13], suggesting the validity and comparability of our data. When focusing on the mortality of non-ICU patients, our results appear novel but compatible with the literature, as in a previous study [5] no higher mortality of transferred patients was observed, when pooling ICU and non-ICU patients after adjustment for several risk factors. This finding is of interest, as transfer per se can pose a risk especially in ICU patients [21, 22]. Similarly, in the non-ICU patients studied by us, we observed only a tendency, which completely disappeared in a multivariable analysis considering risk factors such as age (see Table 1), bacterial infection and acute kidney injury that differed between transferred and non-transferred non-ICU patients (see Table 2). In-hospital death was associated with much greater average reduction of the period of hospitalization in ICU compared to non-ICU patients, which not only underlined the importance of performing separate analyses for these patients but also suggested that ICU patients who did not survive, died relatively early in their hospital stay.
Particularly at times of high burden from COVID-19, inter-hospital transfer of patients was frequent, whereby many of the transfers were due to the need for ICU treatment. However, transfers to general wards also occurred, especially in hotspots as found in many places including the region of the present study in Southern Germany. It may be argued that transfers are trivially associated with prolongation of hospitalization due to the time required for discharge and subsequent readmission; this, however, would account for no more than 2 days. Our study suggests that clinical and social factors could also be involved, besides the additional efforts and time for disinfection, allocation of rooms and similar needs particularly in COVID-19 patients [22]. Indeed, a recently published study [6] demonstrated a longer time of transportation in COVID-19 compared to other patient, implying a longer total period of hospitalization, if transfer is included, as in our study.
For better understanding of the transfer of COVID-19 patients, we categorized the reasons into medical and capacity reasons, besides ECMO. Our study took advantage from the fact that data could be collected in a network of four acute care hospitals including a dedicated COVID-19 hospital, all of which were experienced in the management of COVID-19 and had detailed and comparable documentation. In studying the impact of transfer, we thus benefited from the availability of a multitude of data describing patients’ characteristics and circumstances of transfer, in addition to the fact that the study comprised a time and location with high numbers of COVID-19 patients and stress on the health-care system.
Regarding the reasons for transfer, a previous study already yielded clues in a comprehensive analysis during the first wave of COVID-19, using the level of care of the hospitals, type of transport vehicle and incidence numbers [6]. Our results demonstrate that the change of level of care associated with transfer was closely linked to its reason suggesting that the change in level is a useful proxy for the reason. The previous study [6] revealed that in times of highest incidence, transfers towards higher levels dominated, whereas 4 weeks later there was a peak for transfers to lower levels. Using our result, this would imply that initially medical reasons dominated and later capacity reasons played a major role. This interpretation seems plausible from clinical experience, particularly in view of the dynamic situation encountered in the first wave. We did not perform a time series analysis of our data because the case numbers would have been too low to allow for statistically reliable conclusions, but a tentative analysis showed a similar picture as in the previous study [6]. It should also be acknowledged that our study refers to a period when the therapeutic and transfer management of COVID-19 patients was already well established, in contrast to the first wave.
The differences in the periods of hospitalization cannot be understood without consideration of the decision processes of the treating physicians. According to clinical practice, transfer of ICU patients preferentially occurred in patients who were hemodynamically stable but expected to need prolonged care, with a likely beneficial outcome. Naturally, ICU patients with either fulminant course or expected rapid recovery were not the primary candidates for transfer. This was reflected in the high proportion of capacity reasons as well as the finding that in the multiple regression analysis death had an extreme shortening effect on the period of hospitalization.
In non-ICU patients, the decision process was more complex and less dictated by medical reasons; instead, capacity reasons dominated. There was an interference with medical reasons in the sense that patients, in whom a prolonged recovery occurred or who were in late stage COVID-19, were probably more often transferred. As suggested by our finding that in non-ICU patients living at home alone (16% of this group) was associated with a period of hospitalization being longer by about 5 days indicated that the management of COVID-19 patients extended beyond acute hospital care. The result was robust in the sense that it was obtained after adjustment for several factors associated with disease severity and the period of hospitalization. Thus, other explanations may be needed. One of them could be that stress situations of the health-care system imply shortage of resources, with the consequence that more time is needed to organize a sufficient degree of care after discharge from the hospital in patients living alone compared to patients living in a shared household or nursing facility. Quarantine requirements may also have interfered with the period of hospitalization, for example if discharge was delayed in order to reduce the risk for transmission of infection in shared households and nursing environments. If this should have been true, the prolongation associated with living alone at home would have been even greater. These considerations suggest that shortages outside the hospital environment may have had repercussions on the time spent in the hospital. The problem may have been further exaggerated by shortages of staff in hospitals and nursing institutions.
When comparing the prevalence of comorbidities between transferred and non-transferred patients, ICU patients did not show significant differences between the two groups, except for tendencies for COPD and malignant diseases being less frequent in transferred patients. There was also no difference in the number of comorbidities. This suggests that in ICU patients these risk factors played a minor role for transfer, in accordance with the observation that most transfers in these patients occurred due to capacity reasons. In contrast, medical reasons appeared to play a relatively greater role in non-ICU patients, corresponding to the fact that the number of comorbidities was greater in transferred compared to non-transferred patients. The number of comorbidities that significantly differed between transferred and non-transferred patients was larger than for ICU patients, but this might be partially due to differences in statistical power. Despite this, it seems plausible that certain comorbidities might have favored a transfer, in particular those that might have required prolonged care. Correspondingly, in non-ICU patients, bacterial infection was associated with a prolongation of the period of hospitalization by about 4 days. A further factor linked to extended care could be the presence of cerebrovascular diseases and dementia. This assumption is again supported by our data (see Supplemental Table S1), as in line with the increased age, the observations suggested an impaired clinical state in transferred non-ICU patients; this was indicated by the impaired renal function and lower hemoglobin level, as well as the higher need for corticosteroids and antibiotics. As the differences in the prevalence of cardiovascular diseases were minor, the higher levels of troponin and NT-proBNP in transferred non-ICU patients might have been linked to the higher age. In ICU patients, the most obvious finding was the reduction of the Horovitz-Score in the transferred patients, potentially indicating a greater severity of the respiratory impairment.
Limitations
Our study focused on the period of hospitalization as major outcome and did not analyze the impact of transfer on long-term outcomes. In addition, the determination of the reasons for transfer had to rely on the information given in the files and medical records. If the reason was not explicitly stated, a medical reason was assumed. Thus, we may have underestimated the percentage of transfers due to capacity reasons. Moreover, we had only limited information to compute the APACHE-II-Score, which might have affected the finding that transferred and non-transferred ICU patients did not differ in the simplified score. Although we collected a broad variety of data, we also might have missed clinical factors potentially explaining the differences in the period of hospitalization. The clinical characteristics considered primarily referred to admission but we also included parameters of clinical course, especially complications and treatment modalities. Moreover, the relevant factors might have comprised psychological and social factors such as living alone at home, which we indeed identified as relevant. Our result is confined to the second half of 2021 and a limited region of Southern Germany and thus might not be applicable to other regions and times. On the other hand, it was obtained in a standardized and well-experienced hospital environment. Considering, that Germany is known to have a high number of ICU resources compared to many other countries and that despite this fact issues for ICU patients emerged, it could be of interest to compare our findings on the period of hospitalization with those of other countries.