Study population and setting
This study detailed a matched, retrospective cohort study using national electronic health records and death registration data in England, including 47,780 hospitalized COVID-19 cases (10% in ICU, mean age: 64.5, 55% male) matched 1:1 to hospitalized non-COVID-19 controls on age group, sex, ethnicity, region of residence, index of multiple deprivation quintile, smoking status, BMI group, and chronic disease history (hypertension, major adverse cardiovascular event, respiratory disease, chronic kidney disease, chronic liver disease, diabetes, and cancer). Laboratory-confirmed or clinically-diagnosed COVID-19 cases who were discharged alive from their first hospital episode between January 1 and August 31, 2020 were identified from the Hospital Episode Statistics Admitted Patient Care records in England. Controls were identified from the General Practice Extraction Service Data for Pandemic Planning and Research (GDPPR) dataset, as individuals with at least one GDPPR record from January 1 2019 to September 30, 2020, who did not meet the COVID-19 case definition, and who were alive on the COVID-19 case’s index date. The index date was defined as the date of discharge following first hospitalization with COVID-19, and individuals were followed from that date until September 30, 2020 or date of death. Outcomes of interest included death, readmission, or any diagnoses of severe organ impairment (respiratory, cardiovascular, metabolic, kidney, and liver diseases) following hospital discharge.
Summary of Main Findings
Follow-up times were relatively similar between cases and matched-controls with mean follow-up of 140 days (SD: 50 days) and 153 days (SD: 33 days), respectively. Among the 47,780 hospitalized COVID-19 patients, 29.4% (766 per 1000 person-years (PY)) were readmitted and 12.3% (320 per 1000 PY) died following discharge, which were 3.5 and 7.7 times higher than rates in matched controls, respectively. Among COVID-19 cases, post-discharge new-onset of respiratory disease (21.5%, 538.9 per 1000 PY), major adverse cardiovascular events (MACE: 2.6%, 65.9 per 1000 PY), diabetes (1.1%, 28.7 per 1000 PY), chronic kidney disease (CKD: 0.6%, 14.6 per 1000 PY), and chronic liver disease (CLD: 0.2%, 4 per 1000 PY) occurred at higher rates than controls, with rate ratios 27.3 (respiratory disease), 3.0 (MACE), 1.5 (diabetes), 1.9 (CKD), and 2.8 (CLD) times higher, respectively, than matched controls. The increased rates of long-term outcomes (death, readmission, respiratory disease, diabetes, MACE, CKD, CLD) in COVID-19 patients compared to matched controls were even more pronounced among non-white individuals and individuals <70 years, but relatively consistent by sex.
This was a very large study using national electronic health records and death registration data across England. The study used a matched cohort design, including a comparison group of non-COVID controls, who were matched on several factors that could be related to COVID-19 and post-discharge outcomes of interest. Outcomes of interest were presented as prevalence and rates, the latter of which accounts for differences in follow-up times between COVID-19 patients and matched controls.
Hospitalized COVID-19 patients, who were discharged alive, but who had missing age or sex or those who could not be matched to a control were excluded from the analysis (9.2% of all eligible patients), which may have resulted in selection bias if the excluded participants were different than those included in terms of their risk of outcomes. With increased attention on long-term outcomes of COVID-19, post-discharge follow-up of severe COVID-19 cases may be systematically different than that of the matched controls, leading to a lower threshold for readmission or increased rates of post-discharge diagnoses, compared to the matched controls (detection bias, resulting in a bias away from the null or potential exaggeration of the association). The population included only individuals who were hospitalized for COVID-19 (or a control condition, which are not well described), so the results are unlikely generalizable to the population of COVID-19 cases who do not require hospitalization. This study focused solely on re-admission, mortality, and organ dysfunction, following hospitalization, rather than symptom longevity, which is of additional interest in COVID-19 research. The control group is somewhat poorly defined, limiting complete assessment of residual confounding and bias.
This is the largest known study of post-discharge follow-up among COVID-19 patients, using a matched control population for comparison.
This review was posted on: 4 February 2021