To estimate the effectiveness of Covid-19 vaccines in the elderly population in Finland, we conducted a nationwide, register-based cohort study starting on December 27, 2020, and ending on February 19, 2022. The study population was defined as all individuals aged 70 years and over at the beginning of the study and registered in the Population Information System as resident in Finland since January 1, 2020. The unique personal identity code assigned to all permanent residents in Finland allowed linking individual-level data from different sources.
The primary outcome was Covid-19-related hospital admission timely associated with a laboratory-confirmed (by polymerase chain reaction or antigen detection assay) SARS-CoV-2 infection. Using the Care Register for Health Care, which records data on all patients discharged from inpatient care in Finland, we defined Covid-19-related hospitalization as any inpatient encounter with a primary diagnosis of Covid-19 (International Classification of Diseases, 10th revision: U07.1, U07.2), acute respiratory tract infections (J00–J22, J46) or severe complications of lower respiratory tract infections (J80–84, J85.1, J86). A hospitalization was considered timely associated with an infection recorded in the National Infectious Diseases Register if the positive specimen was collected up to 14 days before or seven days after the hospital admission.
The secondary outcome was Covid-19-related ICU admission. ICU admissions were identified from the Finnish Intensive Care Consortium's Quality Register for Intensive Care, which records data on all patients treated in an ICU in Finland. We considered any admission as Covid-19 related if it was marked by the treating physician as due to Covid-19 and if the patient was laboratory-confirmed SARS-CoV-2 positive during the stay.
The exposure was Covid-19 vaccination recorded in the National Vaccination Register, which covers the whole population irrespective of whether they are served by public or private primary health care providers. We distinguished between the three vaccine brands Comirnaty, Spikevax and Vaxzevria and the number of administered doses. The time since vaccination was taken into account by categorizing the time since the last dose using the following cut points: days 21 and 84 after the first dose, days 14, 91 and 181 after the second dose, and days 14 and 61 after the third dose. Thus, a vaccinee’s exposure state changed over time. Being unvaccinated was the reference state.
We considered age, sex, region of residence, residence in a long-term care facility, influenza vaccination in 2019–2020, number of nights hospitalized between 2015 and 2019 and presence of predisposing comorbidities as confounders. The first three confounders were taken from the Population Information System. Information on whether a subject was in long-term care at the beginning of the study or vaccinated against influenza in the last pre-pandemic season were collected from the Care Register for Social Care and the National Vaccination Register, respectively. We used the data in the Care Register for Health Care from 2015 onwards to count the number of nights hospitalized between 2015 and 2019 and to assess the presence of comorbidities that predispose to severe Covid-19 according to a recent study of predictors of Covid-19 hospitalization (26). We completed the collection of data on predisposing comorbidities with primary health care records and prescription data as described in a previous study (3).
Each study subject was considered at risk of the primary and secondary outcomes from the beginning of the study until the first occurrence of any of the following events: outcome of interest, death, day 14 after any laboratory-confirmed SARS-CoV-2 infection, vaccination with an unidentified vaccine, a heterologous second vaccination, vaccination with the third dose prior to the start of the booster campaign (approximated by September 17, 2021), or end of study. All those events other than the outcome of interest led to censoring before the end of the study.
Using Cox regression with time in the study as the underlying time scale, we compared the hazard of the two outcomes in vaccinated study subjects with the corresponding hazard in the unvaccinated. The effect measure of interest was VE, quantified as one minus the hazard ratio adjusted for the seven confounders categorized as outlined in Table 1. For each VE estimate, we computed either the 95% Wald confidence interval (CI) or, if there were no cases in one of the two groups, the p-value of the likelihood-ratio test. We stratified the analysis by age group and presence of comorbidities. Because of the emergence of Omicron (Supplementary Fig. 1), we also estimated VE by calendar time, conducting a separate regression analysis for each quarter.
Table 1
Distribution of person-years in the study of vaccine effectiveness against Covid-19 hospitalization.
| Not vaccinated, person-years (%) | Vaccinated, person-years (%) |
Age in years |
70–79 | 171 816 (26) | 480 532 (74) |
80–89 | 57 024 (20) | 223 267 (80) |
90–115 | 12 790 (23) | 43 687 (77) |
Sex |
Male | 104 371 (25) | 316 048 (75) |
Female | 137 258 (24) | 431 438 (76) |
Region of residence |
Helsinki-Uusimaa | 55 738 (23) | 182 496 (77) |
Åland | 1130 (21) | 4343 (79) |
Northern and Eastern Finland | 60 654 (25) | 185 294 (75) |
Southern Finland | 58 517 (25) | 177 691 (75) |
Western Finland | 65 590 (25) | 197 663 (75) |
In long-term care |
No | 233 001 (25) | 707 795 (75) |
Yes | 8629 (18) | 39 691 (82) |
Influenza vaccination in 2019–2020 |
No | 130 699 (30) | 302 083 (70) |
Yes | 110 930 (20) | 445 403 (80) |
Nights hospitalized between 2015 and 2019 |
0 | 131 127 (25) | 386 023 (75) |
1–5 | 49 542 (23) | 163 688 (77) |
6–20 | 35 150 (23) | 114 631 (77) |
21+ | 25 811 (24) | 83 144 (76) |
Presence of comorbiditiesa |
No predisposing comorbidities | 106 823 (25) | 312 782 (75) |
Moderately predisposing comorbidities only | 52 882 (24) | 170 635 (76) |
At least one highly predisposing comorbidity | 81 925 (24) | 264 069 (76) |
Presence of particular comorbiditiesa |
Diabetes mellitus type 2 | 51 465 (24) | 165 036 (76) |
Heart failure or ischemic heart disease | 42 614 (23) | 140 974 (77) |
Malignant neoplasms | 27 975 (23) | 93 192 (77) |
Asthma | 18 906 (23) | 64 024 (77) |
Chronic pulmonary disease | 7412 (26) | 21 426 (74) |
Chronic renal disease | 5410 (24) | 17 236 (76) |
Hematologic malignancy | 3866 (24) | 12 460 (76) |
Diseases of myoneural junction and muscle | 1569 (26) | 4535 (74) |
Organ or stem cell transplant | 556 (24) | 1798 (76) |
Cerebral palsy or other paralytic syndromes | 501 (24) | 1606 (76) |
Severe disorders of the immune system | 372 (24) | 1163 (76) |
a See study of predictors of Covid-19 hospitalization (26). |
This tabulation uses a binary vaccination status without distinguishing between vaccine brands, number of doses and time since vaccination. |
To rule out residual confounding, we quantified the association between Covid-19 vaccination and a potential negative control outcome: inpatient encounters due to injury, poisoning and certain other consequences of external causes (International Classification of Diseases, 10th revision: S00–T98) recorded in the Care Register for Health Care. Each study subject was considered at risk of this outcome from the beginning of the study until the first occurrence of any of the following events: injury, death, vaccination with an unidentified vaccine or a heterologous two-dose series, vaccination with the third dose prior to September 17, 2021, or end of study. We compared the hazard of the control outcome in vaccinated study subjects with the corresponding hazard in the unvaccinated using Cox regression and expected to find no difference between the groups.
The significance level was set to 5%. All analyses were performed in R 4.1.2 (R Foundation for Statistical Computing, Vienna, Austria). Results concerning the first dose, less common (less than 1200 person-years) vaccine series and the first two weeks after vaccination are presented in the supplementary data only.