The COVID-19 pandemic has had a profound impact on global health systems, with Italy — one of the hardest-hit countries — experiencing a significant strain on its healthcare infrastructure and economy. The empirical literature has produced extensive evidence to establish and characterize the spread of the disease and its consequences for infected individuals [e.g. [1], [2]]. However, the medical consequences of the pandemic are not limited to COVID-19 patients because the outbreak forced policymakers and hospital managers to implement drastic measures and divert resources away from the treatment of other conditions [3], especially in the first phases of the pandemic, when the available resources and specific medical knowledge on how to treat the disease were limited. On the one hand, the insufficient capacity of hospitals and the necessary prioritization of infected individuals in very serious conditions have led to an impressive number of COVID-19 deaths occurring without hospitalization [4]. On the other hand, the prioritization of COVID-19 patients over patients with other conditions and the efforts made to avoid in-hospital contagion have led to profound disruptions in hospital practices, which have resulted in a dramatic reduction in the level of care supplied to all other patients.

While previous works have highlighted and documented some of these aspects, these studies typically focus on limited areas or specific settings, and we are not aware of any study that has previously attempted to quantify the full extent of these phenomena. A more comprehensive understanding of the consequences of the outbreak for all patients, including those suffering from nonrespiratory conditions, is crucial for at least three reasons. First, given the magnitude of these disruptions, any quantification of the health costs of the pandemic that focuses only on infected individuals would severely underestimate the aggregate social costs of the outbreak. Second, it would be impossible to critically appraise the emergency strategies adopted to cope with the pandemic by focusing only on their clinical impact on infected individuals and overlooking their consequences for other patients. Third, we expect the consequences of missed hospitalizations and reduced care to persist for years, if not for decades, and the echo of the pandemic to become one of the main challenges for healthcare providers.

In this study, we applied regression analysis and machine learning techniques to complete-coverage administrative data for more than 50 million hospitalization records from Italy from 2012 to 2021, representing all hospital discharge data for acute care; the aim was to not only provide a precise estimate of out-of-hospital deaths but also quantify the effects of the COVID-19 pandemic on patients affected by other conditions and thus provide a more comprehensive account of the disruptions in healthcare provision induced by the pandemic.

Our results show that 53.7% of COVID-19 deaths occurred outside of hospitals throughout the whole period. We interpret this result as evidence of the scarcity of available resources and hospitals’ inability to cope with the unprecedented increase in demand during the pandemic. We also estimate a 22% reduction in the number of hospitalizations for nonrespiratory patients and a 0.5-day reduction (corresponding to 7.5% of the average) in average hospital length of stay (LOS), conditional on patients’ conditions and characteristics. We interpret these changes as attributable to a combination of the fear of in-hospital contagion on the part of both patients and healthcare providers [5], [6], [7], [8], [9] and a significant diversion of healthcare resources away from the treatment of nonrespiratory conditions. Consistent with these interpretations, both the incidence of out-of-hospital deaths and the decrease in nonrespiratory admissions are stronger in areas more severely affected by the pandemic, as proxied by population-level excess mortality.

The literature on excess mortality associated with the COVID-19 outbreak is vast and provides estimates of the aggregate death toll of the pandemic for several countries [2], [10], [11]. By comparing these estimates with available statistics on officially recognized COVID-19 deaths, some authors have concluded that several COVID-19 deaths occurred without diagnosis or hospitalization in a variety of settings, including for instance the US [12], Germany [13], England and Wales [14], and Lombardy [15]. Focusing on the metropolitan area of Milan, Jackson et al. [4] found the phenomenon to be particularly relevant in the early phases of the pandemic and to disproportionally affect older men, care home residents, and people with comorbidities. Our contribution to the literature on COVID-19 excess mortality is twofold. First, we are not aware of any study that has estimated in-hospital excess deaths and compared it to population-level excess deaths to estimate the share of COVID-19 deaths that occurred outside of hospitals. Second, while population-level estimates of COVID-19 excess mortality have been used extensively to test for statistical associations with demographic [16], [17], environmental [18], and institutional factors [19] and to evaluate the efficacy of policy interventions [20], [21], we are not aware of any study that has previously associated excess mortality with out-of-hospital death or with the reduced utilization of healthcare services.

The empirical literature has also documented disruptions to the provision of healthcare to non-COVID-19 patients in different forms [22], [23], [24], [25], including reduced access to emergency care [26], [27], [28], a reduction in elective surgeries [29], [30], [31], and reduced hospitalizations for other conditions [9], [32], [33]. Although several studies have documented these phenomena, we are aware of only one that has done so with national-level full-coverage administrative data, with a focus on elective surgeries in the Netherlands [29]; to the best of our knowledge, no previous study has quantified the aggregate loss of hospitalizations for acute care at the national level. We contribute to this stream of literature by providing the first national-level comprehensive estimate of the aggregate loss in hospitalizations for acute care patients induced by the pandemic and by providing novel evidence of a reduced LOS for nonrespiratory patients, conditional on their conditions and characteristics, which is a previously undocumented consequence of the diversion of resources away from non-COVID-19 patients. Finally, we contribute to this literature by providing further evidence that the reduced utilization of healthcare services is primarily driven by patients with less serious conditions. This finding is also consistent with previous works on emergency department (ED) admissions in more confined settings, such as Golinelli et al.[34], and suggests that healthcare providers prioritize patients with greater needs. Our evidence thus suggests that the results of previous studies documenting increases in mortality for hospitalized patients [35] may be better rationalized by compositional changes in hospitalizations in favour of patients with more serious conditions rather than a direct effect of these disruptions on patient care.