Health progression for Covid-19 survivors hospitalized in geriatric clinics in Sweden

Objective To analyse if the health progression of geriatric Covid-19 survivors three months after an acute Covid-19 infection was worse than in other geriatric patients. Specifically, we wanted to see if we could see distinct health profiles in the flow of re-admitted Covid-19 patients compared to re-admitted non-Covid-19 controls. Design Matched cohort study. Setting and participants Electronic medical records of geriatric patients hospitalised in geriatric clinics in Stockholm, Sweden, between March 2020 and January 2022. Patients readmitted three months after initial admission were selected for the analysis and Covid-19 survivors (n = 895) were compared to age-sex-Charlson comorbidity index (CCI)-matched non-Covid-19 controls (n = 2685). Methods We assessed using binary logistic and Cox regression if a previous Covid-19 infection could be a risk factor for worse health progression indicated by the CCI, hospital frailty risk score (HFRS), mortality and specific comorbidities. Results The patients were mostly older than 75 years and, already at baseline, had typically multiple comorbidities. The Covid-19 patients with readmission had mostly had their acute-phase infection in the 1st or 2nd pandemic waves before the vaccinations. The Covid-19 patients did not have worse health after three months compared to the matched controls according to the CCI (odds ratio, OR[95% confidence interval, CI] = 1.12[0.94–1.34]), HFRS (OR[95%CI] = 1.05[0.87–1.26]), 6-months (hazard ratio, HR[95%CI] = 1.04[0.70–1.52]) and 1-year-mortality risk (HR[95%CI] = 0.89[0.71–1.10]), adjusted for age, sex and health at baseline (the CCI and HFRS). Conclusions and implications The overall health progression of re-hospitalized geriatric Covid-19 survivors did not differ dramatically from other re-hospitalized geriatric patients with similar age, sex and health at baseline. Our results emphasize that Covid-19 was especially detrimental for geriatric patients in the acute-phase, but not in the later phase. Further studies including post-vaccination samples are needed.


Introduction
Emerging evidence shows a wide range of systemic and organ-specific health problems weeks or months after an acute phase of Covid-19 (SARS-CoV-2) infection [1][2][3]. These health issues can be persistent and/or new conditions. After the acute phase, the quality of life is lower, and signs, symptoms, and diseases including fatigue, malaise, neurological, mental, cardiac, vascular, gastrointestinal, renal, musculo-skeletal, and dermatological disorders are more prevalent [1][2][3]. As older individuals have a higher risk for severe Covid-19 outcomes and the risk factors in acute Covid-19 differ by age [4][5][6][7], age group-specific analyses on health progression after Covid-19 infection are needed. This has been addressed by Cohen et al. (2022), and they reported the excess burden in multiple clinical conditions four months after initial Covid-19 infection in adults aged >65 years [8]. Another study with Covid-19 survivors aged >65 years by Liu et al (2022) reported 12% decrease in cognitive functions one year after Covid-19 hospitalization. Yet, the research evidence of the health progression after Covid-19 infection in the aged populations is limited and analyses in the most vulnerable and frail individuals with a high death rate from acute Covid-19 [9,10], are needed.
In this study, we sought to analyse whether a previous acute Covid- 19 infection is a risk factor for worse health progression three months after initial infection in hospitalized geriatric patients surviving at least three months. The health indicators were obtained from the electronic medical records including the admission with acute Covid-19 diagnosis and readmission after three months. In addition, we assessed whether the risk for new diseases was higher in Covid-19 patients compared to the controls. We hypothesised health progression to be worse and the risk for pathologies to be increased in the Covid-19 survivors.

Methods
The analysis was performed in older patients with and without Covid-19 infection readmitted to geriatric hospitals (age at baseline at least 50 years). In the main analyses, we assessed the risk for worse health progression in the Covid-19 survivors compared to the matched non-Covid-19 controls using CCI�2 (higher than the median (1)) (i), HFRS�5 (intermediate or high risk of frailty)(ii), 6-months and 1-year-mortality (iii), and occurrence of new diseases (iv) as the outcomes.

Clinical cohort
All sample information used in the analysis were extracted from the electronic medical records and Death Register. International classification of diseases, 10th revision (ICD-10) system was used for the diagnoses. Selection and the timeline of the analytical sample for the main analysis are visualized in Fig 1. Data consisted of medical records of 33,452 patients admitted to nine geriatric hospitals in Stockholm, Sweden between March 1, 2020, and January 17, 2022. Of these, patients were excluded if: 1) data were missing (discharge date, diagnoses as ICD-codes or age, n = 398), 2) they were considered outliers, with age<50 years (n = 5), or 3) they had ICD-codes U089, U099, U109, ZV100 or Z861A in their 1 st admission records referring to post-infectious condition or history of Covid-19 disease (n = 214) (Fig 1) [11].
The 1 st admission was considered as the baseline in which 33,049 patients with and without acute Covid-19 diagnosis (U07) were divided into Covid-19 (n = 5,117, 16%) and non-Covid-19 (n = 27,718, 84%) patients, respectively (Fig 1). In our main analysis, only readmissions three months after the 1 st admission were considered, which is the time after acute and postacute Covid-19 disease [12,13]. Of the Covid-19 survivors with readmission after three months, 14 were excluded because they had an ICD-code U07 in their readmission records, and of the non-Covid-19 survivors, 34 patients were excluded because of ICD-codes U089, U099, U109, ZV100 or Z861A in their records referring to post-infectious condition or history of Covid-19 disease (Fig 1).
For the comparative main analysis, a case-control setting was assigned in patients with readmission data. For each Covid-19 survivor with readmission (n = 895, U07 only in the 1st admission), three age-sex-CCI-matched non-Covid-19 controls were selected out of the 5,241 non-Covid-19 patients, thus, resulting in a control group with 2,685 patients (Fig 1). Matching is described in paragraph, Statistical analysis.

Covid-19 diagnosis
Acute Covid-19 disease was determined either by a positive laboratory result performed using transcriptase polymerase chain reaction (RT-PCR), or a clinical diagnosis for those with a negative RT-PCR but with typical symptoms and computed tomography scan findings (ICD-10-codes U07.1 or U07.2 [11]).

Charlson comorbidity index, Hospital frailty risk score and mortality
In the analysis, a comorbidity measure (Charlson Comorbidity Index [CCI]) and a frailty measure (hospital frailty risk score [HFRS]) were based on ICD-codes. The CCI was computed using ICD-10-codes and weights as previously described [14]. The HFRS was defined from 109 ICD-10-codes that are overrepresented in the frail individuals [15]. The CCI and HFRS were calculated at baseline (in the 1 st admission) and after three months (in the 1 st readmission after three months).
Death dates between March 1, 2020, and January 17, 2022, were available for all 33,452 patients in the electronical medical records. We monitored if death occurred within 6 months or within one year after the 1 st admission to a hospital, and time to death for 6-months-and 1-year mortality was calculated from the 1 st admission date. Time between the 1 st admission date and end of data collection (17 January, 2022)

Diseases
In the analysis of individual diseases, two approaches were used: a) all ICD-10-codes in hierarchical groups and b) selected diagnoses with specific ICD-codes. In the hierarchical approach (a), ICD-codes in the medical records in an admission were assigned into categories shown in  Table 1 and  Table 1). These selected diseases (b) were chosen for analysis as they are common conditions in the geriatric patients (i.e. a reasonable number of cases for statistical analysis) and/or have been previously suggested or shown to be one of the late effects of Covid-19 in younger and/or older adults [2,8,16,17]. Occurrence of individual diseases (yes/no) for each patient were explored at baseline (in the 1 st admission) and after three months (in any readmission after three months until 17 January, 2022).

Statistical analysis
Control matching. The matching of the non-Covid-19 control patients with the 895 Covid-19 survivors was performed based on the 1 st admission data. Three controls were selected for  Heart failure  I099, I110, I130, I132, I255, I420, I425, I426,  I427, I428, I429, I43, I50   Stroke  G45, G46, H341, I60, I61, I63, I64 Dementia � or confusion F00, F01, F02, F03, F051, G30, G31, A810, R41, F059, F051, F050 Two approaches were used: all ICD-10-codes in hierarchical groups (a) and selected diagnoses with specific ICDcodes (b). � ) Anti-dementia medication (ATC-codes N06DA or N06DX01) was also considered https://doi.org/10.1371/journal.pone.0283344.t001 each Covid-19 survivor and a control patient was required to have a readmission after three months, and to have the same sex, age quartile, CCI category (0,1 or 2+), and the 1 st admission inthe 1 st pandemic wave or after (cut point August 31, 2020). If more than three matches were found, of these, three controls were selected randomly. Main analyses. Binary logistic regression was used to assess odds ratios (ORs) and corresponding CIs for the relationship between baseline Covid-19 diagnosis and the outcomes i, ii and iv. In the mortality analysis (iii), Cox regression was used to assess the hazard ratios (HRs) and corresponding CIs for the effect of the baseline Covid-19 diagnosis on the outcome.
All analyses (i-iv) were performed using the cross-validation in which 50 iterations were performed, each time randomly selecting 90% of the data. For each outcome, the cross-validated estimate (OR[i, ii, iv], HR[iii] and their corresponding CIs) was the mean of all estimates in the iterations. All models (i-iv) were adjusted for age and sex, as well as CCI (i), HFRS (ii) and CCI + HFRS (iv) at baseline. The analysis on individual diseases (iv) was performed using a negative-outcome-specific approach in which only patients without the disease in question at baseline were included. The health outcomes (i-iii) and individual diseases (iv) were analysed in separate models, one by one. All analyses were performed using R software (version 4.0.5).
In the mortality analysis, R package survival (version 3.2-10) was used.
Additional analyses. Additionally, as a descriptive analysis, we visualized the high level of comorbidities in these geriatric patients using network presentation where node size corresponds to the prevalence of patients with the disease and the line that connects each node corresponds to the prevalence of co-occurrence of each disease pair. As the matched control group was three times larger than the Covid-19 patient group, node sizes and line widths were adjusted to be visually comparable between the groups by dividing the numbers of patients with a disease and co-occurrences of the diseases in the control group by the number 3. Networks were produced using R package igraph v1.2.9. Statistical significance (p<0.05) for the difference between baseline and readmission after 3 months within a patient group was assessed using Mann Whitney-U test for continuous and Fisher's Exact Test for categorical variables.

Patient characteristics
Characteristics of the 3,580 patients used in the main analysis (895 Covid-19 3-months survivors and 2,685 age-sex-CCI-matched non-Covid-19 controls) are shown in Table 2, and S1 and S2 Tables. Of these patients, five out of six were more than 75 years old (lower quartile 78, median 84, and upper quartile 89 years) and 56% were women. Median Fig 1) are shown in S1 and S2 Tables. Of the 5,103 Covid-19 patients, 1,093 (21%) died within 3 months, 895 (18%) had readmission after three months, and 3,115 (61%) survived and had no readmission after three months (Fig 1). Of the 27,684 non-Covid-19 patients, 2,967 (11%) died within three months, 5,241 (19%) had a readmission, and 19,476 (70%) survived and had no readmission after three months (Fig 1). The proportions of patients with CCI�2 at baseline within these patient groups are shown in Fig 2.

Health status three months after acute Covid-19 infection
Patients in the main analysis had a statistically significant increase in the CCI, HFRS and number of diseases at readmission (

Individual diseases
An overview of comorbidities in the geriatric patients at baseline and readmission after three months (S3 and S4 Tables) is shown in Fig 3. We found previous acute Covid- 19

Discussion
In this study, we explored in 3,580 geriatric hospitalised geriatric patients, mostly older than 75 years, the risk of previous acute Covid-19 infection for poorer health progression and  Table. In the network presentation, node size corresponds to prevalence of a disease and the line that connects each node corresponds to prevalence of co-occurrence of each disease pair. As the control group was three times larger than the Covid-19 patient group, node sizes and line widths were adjusted to be visually comparable between the groups by dividing the numbers of patients with the disease and co-occurrences in the control group by the number, 3.
https://doi.org/10.1371/journal.pone.0283344.g003 specific diseases three months after the initial infection. As opposite to our hypothesis, the health progression of the Covid-19 survivors in this cohort did not differ from the matched control patients significantly. That is, we did not observe an association between previous acute Covid-19 infection and greater health decline indicated by the different health indicators (the CCI and HFRS) at the readmission, controlled for age, sex and the health at the baseline. Furthermore, previous acute Covid-19 disease was not a risk factor for a higher 6-months or 1-year-mortality among the 3-months survivors, adjusted for age, sex and health at the baseline. We based our initial hypothesis of Covid-19 being a clear risk factor for poorer health progression in these older hospitalised patients on the fact that homeostasis and resilience of the body lowers with advancing age [18,19] and the fatality of acute Covid-19 disease has been the highest in the oldest ages [10,20]. In addition, previous studies have shown a wide range of Covid-related long-term effects on health [1][2][3]. Age-related health risk factors including preexisting comorbidities and frailty are suggested as likely contributors to more severe late effects of the Covid-19 infection after surviving from the acute phase of the disease [21]. However, extensive research evidence of these late effects in the geriatric hospitalised patients or nursing home residents has been lacking [22]. We also analysed previous acute Covid-19 infection as a risk factor for new individual diseases, adjusted for age and sex, and found out that the risk was significantly higher for two disease categories: 'Diseases of the eye and adnexa' and 'Diseases of the ear and mastoid process'. Among the hierarchical disease categories, their incidence rates were the lowest in this geriatric population, less than six percent. For the other diseases than these two, the odds ratio was most often greater than one but not statistically significant. Hearing loss, visual disturbance and conjunctivitis are one of the late or persistent effects of Covid-19 infection (also referred as 'post-Covid-19 syndrome' or postacute sequelae of COVID-19, PASC) [2,23,24], and viral infections, in general, have manifestations in the eyes and ears [25,26]. While in PASC, the late effects are often lingering symptoms for weeks or months after the initial infection, our disease-specific analysis focused on new conditions occurring after the acute Covid-19 infection. Mechanistically, Covid-19 virus can infect the inner ear cells directly [27], and possibly the conjunctival epithelium as well [28]. However, whether the location of initial infection has a role in the late pathophysiological consequences of Covid-19 infection remains to be disentangled, and in overall, the pathways leading to Covid-19-related late manifestations in the eyes and ears are still poorly characterised.
Previous studies focusing on the late effects of Covid-19 in individuals older than 65 years are limited. In the study by Cohen et al. (2022) with 90,000 older Covid-19 survivors, the incidence rates of multiple conditions in different organ systems were higher in the Covid-19 survivors compared to the '2020 controls' or to 'historical control group from 2019'. This clinical sequelae resembled in most parts those of viral lower respiratory tract illness. A major difference between the study by us and Cohen et al was that we did not exclude any individuals based on the health status while in the study by Cohen, exclusion was made if a person had some severe and/or disabling chronic conditions or was living in an institution. In addition, we used medical records from hospitalisations and general death registry data, while Cohen et al. used also other patient registries from outpatient hospital services.
We were expecting to see a clear difference in the health progression of the Covid-19 survivors compared to matched controls but this was not the case in the re-hospitalised geriatric patients. Instead, our data demonstrated the ageing-related health decline as well as a substantially high 3-months mortality rate in Covid-19 patients. Already at baseline, the patients with many diseases as indicated by the CCI and as visualised in the network diagrams in this report. Of the 33,000 patients initially hospitalised in these geriatric hospitals, one out of five died within three months if they had had acute Covid-19 infection and if not, one out of ten. Of the survivors, 20% had a readmission after three months, and according to the increased number of diseases, CCI and HFRS, the health of both Covid-19 and control patients was worse in the readmission than in the 1 st hospitalisation.

Strengths and limitations
To the best of our knowledge, our study offers completely new information in the context of hospitalised geriatric patients. Primarily, we wanted to avoid exclusion of any patients and used comprehensive health indicators and disease diagnoses and mortality that were well available for analysis. Thus, data set in this study is representative of the whole population as it includes practically all patients hospitalized in these clinics in Stockholm, Sweden within the study period, between March 2020 and January 2022. Other advantages in the analysis were that we used as controls, non-Covid-19 patients hospitalized in the same clinics who were matched for age, sex and the health at the baseline. We were also able to use clinical data from several hospitals.
However, when interpreting our findings, some issues should be considered. First, the results cannot be generalised to all other older populations, such as community-dwelling older individuals. Health progression was evaluated only in geriatric patients with readmission to the nine geriatric hospitals and the CCI and HFRS after three months were not available for analysis for those patients who survived three months and were not readmitted to these clinics. Second, as we were able to use electronical medical records only, further research in the elderly is needed with questionnaire data targeted for Covid-19-related later symptoms. Third, in this analysis, most of the Covid-19 patients with readmission after three months had their 1 st admission during the 1 st two pandemic waves in which e.g. the prognosis of the geriatric Covid-19 patients was worse than in the later waves [29]. Therefore, Covid-19-related late effects need to be explored in geriatric patients with more later follow-up data. Fourth, many adverse health conditions in the oldest individuals may have increased during the pandemic also without Covid-19 infection because of restrictions and pressured health care services [30]. The recommendations in Sweden and other countries for individuals older than 70 years included limiting e.g. social contacts and being at crowded places. By decreasing daily activities and social interactions, cognitive and physical well-being are compromised and the risk for many conditions such as falls, dementia, and psychiatric symptoms increase [31,32]. Historical data helps to explore whether the health profiles of geriatric patients in overall have changed during the pandemic. However, data before the pandemic were not available to our analysis. Lastly, due to sample size issue, we analysed individual diseases only in wide disease categories.

Conclusions and implications
The health progression of Covid-19 3-months-survivors readmitted for geriatric care, did not differ dramatically from the other readmitted patients with similar age, sex and baseline health. Our results emphasize that Covid-19 was especially detrimental for geriatric patients in the acute-phase, but at least in this cohort not in the later phase. Future research is needed to assess the health progression after Covid-19 in the later pandemic waves.
Supporting information S1 Table. Characteristics of all geriatric patients in the 1 st admission (baseline). Timeline and the grouping of the patients is described in Fig 1. (DOCX) S2 Table. Descriptive of the health indicators in all geriatric patients. The CCI and HFRS are described in the 1 st admission (baseline) and a readmission after three months. Vital status was observed six months and one year after the 1 st admission. Timeline and the grouping of the patients are described in