In our data set, which includes deaths until 1 June 2020, the median age at death after a confirmed SARS-CoV-2 infection was 85 years. The proportion of deaths (lethality) was low (2.9 %) in Norway, but increased significantly with age. In our data, the lethality among persons aged 90 years and over was 52 %. High lethality among the oldest patients has been shown in several countries, proving the importance of protecting the most vulnerable segments of the population (14). Not all those infected with SARS-CoV-2 develop symptoms, and some have symptoms that are so mild as to evade testing. The test criteria and capacity may also vary over time and across regions and countries. The number of people who are tested and have the disease confirmed is decisive for determining its lethality (14). In countries where many persons with a mild or no disease have been tested, this will lower the lethality rate. As of 7 June 2020, Iceland, where widespread testing has been undertaken, reported a lethality of 0.6 %, Finland 4.6 %, Denmark 4.9 % and Sweden, which has been severely affected by the pandemic, reported a lethality of 10.4 % (15, 16).
We found that two of every three deaths after a confirmed SARS-CoV-2 infection occurred outside of hospitals. Figures from the MSIS registry show that the majority of these deaths occurred in other healthcare institutions, most likely in nursing homes, and only a minority died outside an institution. Since positive tests for SARS-CoV-2 were retrieved from the MSIS registry and linked to deaths recorded in the National Population Register, the location of the test was assumed to be the same as the place where the patient died. In most cases this will be correct, but we cannot exclude the possibility of erroneous classifications. Reporting to the MSIS registry includes information on whether the patient was at home at the time when the test was taken or admitted to a hospital, nursing home or other healthcare institution, but does not indicate the type of healthcare institution in question, nor whether persons have been long-term nursing home residents, have been admitted for short-term periods from their homes or have been transferred from a hospital for a temporary stay in a nursing home. In our study, 15 patients died during the first week after discharge from a hospitalisation episode with a COVID-19 diagnosis, but we have no information on their clinical histories or functional status. The Norwegian Directorate of Health recommends that nursing home residents who fall ill with COVID-19 be treated in the nursing home without being hospitalised, unless there are especially strong grounds to indicate that a hospitalisation will significantly prolong their lives and enhance their quality of life (17). The nursing home population is especially vulnerable to a serious trajectory and death from COVID-19 due to their advanced age, underlying chronic diseases and general frailty, and a high risk of rapid spread of the virus among nursing home residents has also been shown (18, 19). Old and frail patients often have atypical symptoms, which increase the risk that the disease goes unnoticed or is discovered at a late stage. High-quality infection control, good monitoring of symptoms, a low threshold to testing for SARS-CoV-2 and isolation of infected patients are therefore especially important to prevent outbreaks in nursing homes.
In deaths that occurred outside hospitals, cardiovascular diseases and dementia were the most common underlying diseases. We have previously investigated the prevalence of underlying diseases in persons hospitalised for COVID-19 (11). In that study, the proportion of patients in the hospital population with a diagnosis of dementia was identical to the proportion of persons with a diagnosis of dementia in the general population. It is therefore interesting to note that no diagnoses of dementia could be identified in any of the patients who died while in hospital. This could indicate that the recommendations from the Norwegian Directorate of Health regarding treatment of nursing home residents with severe COVID-19 outside hospital are being complied with (17). In a study of COVID-19 outbreaks in Bergen, very few nursing home residents were hospitalised. The authors discussed whether the reason could be that the nursing homes had made thorough efforts to establish the treatment level for each resident prior to the COVID-19 epidemic (19). In line with Danish (5) and Italian studies (20), we found a larger proportion of patients with diabetes among those who died while in hospital than among those who died elsewhere. During the first weeks of the pandemic there was special concern about persons with underlying conditions or undergoing treatment that affects the immune system. In our study, only five persons with such conditions were registered as deceased after a confirmed SARS-CoV-2 infection, and there were no deaths among persons receiving active cancer therapy or users of biological drugs prescribed and paid for by the hospital (h-prescription).
The strength of this study is that it includes all persons resident in Norway and all deaths among people with a confirmed SARS-CoV-2 infection. The registry linkages provide us with data for prospectively registered somatic diagnoses from both the specialist and the primary health service. Even though the quality of some diagnoses varies between the health registries, their advantage is that these diagnoses have been registered on an ongoing basis, thus avoiding the biases that may arise when the risk factors are reported by the deceased person's doctor.
A weakness of this study is that we cannot know whether COVID-19 was the direct cause of death. Theoretically, the patient could have recovered from the infection and later died from another cause. However, for several reasons we assume it is likely that the vast majority of the persons included in the statistical base in fact died from SARS-CoV-2, because only a short time had passed between confirmation of the infection and death, all those who had a confirmed infection and died while in hospital had COVID-19 as their main or secondary diagnosis, and the number of deaths from COVID-19 reported to the MSIS registry was numerically quite equal to our sample; 239 and 244 persons respectively. A review of patient records to verify COVID-19 as the cause of death would have strengthened the validity of the study.
In September 2020, the Norwegian Institute of Public Health published preliminary data from the Causes of Death Registry on COVID-19 as a cause of death and the correlation with chronic diseases (21). These analyses encompass the period March–May 2020, and the Causes of Death Registry reports that COVID-19 was registered as the underlying cause of death in 215 (91 %) of the 236 cases that had been confirmed by a laboratory and reported to the registry. Chronic diseases had been registered in the death certificates for 89 % of all deaths associated with COVID-19, and by registering chronic diseases in this manner, the Causes of Death Registry has applied a wider definition of chronic disease than we have done in our analysis.
The disease profile differs to some extent from what we found, with less reported diabetes and more dementia in the data from the Causes of Death Registry. This underscores the need for using multiple sources in parallel when studying such complex issues.
Aggregated data based on registry linkages can be a good source of information on risk groups during a pandemic, but are not sufficient for more refined analyses. A major research project is therefore now being established, in which individual-level data on underlying disease and risk factors are also being linked to causes of death from the Causes of Death Registry, drug use from the Prescription Registry and data on socioeconomic conditions and country of origin from Statistics Norway.