Outcomes of hospitalized patients with COVID-19 according to level of frailty

Design

We performed an observational, descriptive and retrospective

Population and sample

A total of 3,581 clinical episodes were extracted from a database of 3,112 patients from La Paz University Hospital (LPUH) during the period between 15 July 2020 and 31 July 2020, meaning a patient could have been admitted to the hospital several times. The database included sociodemographic data, clinical status, laboratory findings, and clinical management of the patients admitted with a respiratory infection by SARS-CoV-2 (determined by polymerase chain reaction) since the beginning of the current pandemic. A sample of 254 patients was randomly chosen from the total population, and with this sample size, in terms of the most frequent complications, a test power of at least 0.8744 was obtained.

Frailty assessment

Frailty was assessed by the CFS, which is a tool initially developed by the Canadian Health Study of Aging based on the Frailty Index (FI) to be used in patients older than 65 years. It was created with the objective of developing a simple instrument to stratify patients according to their degree of vulnerability and to relate it to patients’ needs and prognoses. However, contrary to the FI, that includes comorbidity within its score, which has been independently associated with worse clinical outcomes in COVID-19 patients, the CFS does not take into account comorbidities for its assessment, allowing to focus on the investigation of the accurate relationship between frailty and clinical outcomes (Kow & Hasan, 2020).

The CFS is a practical, efficient tool that assigns a score between 1 and 7 based on activity, functionality and disability (Rockwood et al., 2005). It classifies patients as robust (score 1–3), pre-frail (score 4), and frail (score >5) (Dent et al., 2019). This stratification allowed to identify not only frail individuals, who usually have higher risk of serious health complications and worse prognosis; but pre-frail subjects as well, who have reported an increased risk of negative events, such as an elevated mortality from all causes (Changfeng, Shouyan & Yu, 2019).

The CFS scale was not taken directly from the medical record but involved analyzing family medical history, nursing history, and in some cases interviewing the patient directly. Six doctors spent 15 days collecting this information and classifying patients according to the CFS scale. Parallelly, a group of experts trained in the use of frailty scales classified the same patients according to their level of frailty obtaining a Kappa 0,89 (95% CI [0.82–0.94])

Variables

Statistical analysis

Quantitative variables were described using robust statistics, such as median and interquartile interval, whereas for qualitative variables a frequency distribution was used. For the comparison of quantitative not normally distributed variables between frailty groups, the Kruskal–Wallis non-parametric H test was used, based on the Shapiro–Wilk test. Finally, for the comparison of qualitative variables, the chi-squared test was used.

The survival estimate was made using the Kaplan–Meier method comparing the survival curve between groups with the log-rank test. The multivariate analysis was carried out by means of Cox regression, with the forward conditional method, introducing as independent variables the CFS scale and the variables that obtained statistical significance in the bivariate analysis or that could have a clinically plausible implication. The results of the multivariate model were presented as a Hazard ratio (95% CI).

Statistical analysis was performed with STATA v16.0 and a pvalue of 5% was considered statistically significant.

Ethical considerations

The study was performed in accordance with the principles of the Declaration of Helsinki (2008 update; available on the World Medical Association website-http://www.wma.net/e/policy/b3.htm) and in accordance with the standards of good clinical practice as described in the ICH Harmonized Tripartite Guidelines for Good Clinical Practice (2001), and the guidelines for Good Epidemiological Practice (http://www.ieaweb.org/GEP07.htm). The study was approved by the Clinical Research Ethics Committee of LPUH, Madrid, with LPUH code: PI-4155. It was not necessary a formulary of informed consent due to the anonymized database was extracted.

This study was conducted in accordance with European and Spanish regulations for the protection of personal data (Organic Law 3/2008).

Patients characteristics

Complete information for all variables was collected from 254 patients. According to the CFS instrument 69.29% were classified as robust (CFS = 1–3), 13.39% as pre-frail (CFS = 4) and 17.32% as frail (CFS > 5). Table 1 shows demographic and housing characteristics, COVID-19 contagion risks, level of dependency and previous comorbidities of the patients studied according to their level of frailty.

The individuals classified as frail were mostly women (56.82%), with an older age (median (IQR): 81.5 (72–87.75)) compared with the robust (median (IQR): 69.50 (55.00–79.00)) and pre-fail (median (IQR): 76.00 (66.50–82.50)) subjects. Frail patients had a higher prevalence of suspected nosocomial transmission (27.27%), compared with robust (1.72%) or prefrail (3.03%) individuals.

In Table 1 it can also be observed that most comorbidities had a similar distribution among the three groups defined above, except for heart disease—for which the robust group had a very low prevalence compared with the pre-frail and frail individuals (18.75% vs. 41.18% and 40.91%, respectively; p = 0.001)—and chronic bronchitis, for which the prevalence grew linearly among the three groups (p = 0.007). With the same pattern as heart disease, chronic renal disease was less prevalent among robust patients (7.95%) compared to pre-frail and frail groups (20.95% and 20.45%, respectively; p = 0.017). Differences were also observed regarding the prevalence of previous neurological disorders, obesity and dementia.

At arrival to the emergency room the clinical presentation of COVID-19 was similar among groups; however, as shown in Table 2, the highest incidence of altered behavior and level of consciousness occurred in frail patients (11.93%, p =< 0.001)

COVID-19 outcomes

Table 3 shows all the in-hospital complications related to COVID-19 according to level of frailty. The most common complications among patients with COVID-19 were acute renal failure (46 patients), delirium (38 patients) and acute respiratory distress syndrome (ARDS) (51 patients).

Despite the length of hospital stay was similar between groups (p = 0.296), ICU admissions (p = 0.014) were less prevalent among frail individuals. Moreover, mortality rates increased linearly among the three groups (p = 0.012).

A statistically significant difference was found regarding delirium, whose incidence was 8.52% in robust patients, 11.76% in pre-frail subjects and 43.18% in frail individuals. Delirium was also associated with a higher mortality rate, being presented in 27.30% of frail patients versus in 4.55% without it; p < 0.001.

No differences were found between groups regarding the rest of the COVID-19 outcomes shown in Table 3. However, it should be noticed that heart failure appeared as a complication of the disease in 5.88% and 6.82% of pre-frail and frail individuals, respectively, compared with an incidence of 1.14% in the robust group; nonetheless, this difference did not reach statistical significance (p = 0.059).

Figure 1 shows the Kaplan–Meier survival estimation curves for each group of patients, according to the CFS frailty scale. In this graph a huge difference during the first fortnight can be detected, especially for patients identified as frail. The differences observed were statistically significant (p value < 0.001).

Finally, a multivariate model was performed using a Cox regression model from the variables shown in the Table 4. The results demonstrated that frailty was associated with mortality (HR: 1.39, 95% CI [1.07–1.81]). However, other factors such as chronic heart disease, chronic kidney disease, and high blood pressure were also associated with death.

Sociodemographic characteristics of interest

In our study, we found that 17.32% of the patients were classified as frail, however, if the subjects with pre-frailty were considered, the total prevalence would increase to 30.71%, which is still a much lower result compared with another study that have reported proportions of 39.6–42.0% (Hewitt et al., 2020). Choi, Ahn & Kim (2015) demonstrated that the most affected frail patients were women and residents of nursing homes, which could be related to the advanced age and the dependency situation that feature these patients. In our study, old age, female sex, dependency, living in a nursing home, and having comorbidities were the high-risk profiles of frail individuals. These data are relevant because frailty is a condition that increases the risk of hospitalization, unwanted incidents and mortality (Martínez-Reig et al., 2016).

The devastating effects of the pandemic on nursing homes are well known causing a disproportionate number of infections and deaths worldwide (Rutten et al., 2020). The typical response to prevent the transmission of the virus in nursing homes has been the aggressive isolation of patients, sometimes ignoring its possible collateral effects in patients with dementia, delirium, or other mental health problems, for whom human relationships are essential (Tan & Seetharaman, 2020). It has been reported that in similar previous epidemics (SARS-CoV in 2003 and MERS-CoV in 2012) approximately 35% of the survivors showed psychiatric disorders during their early recovery, including depression, stress and anxiety, in addition to other minor symptoms (Mak et al., 2009).

The preexisting medical conditions that showed significant associations with frailty in our study were kidney disease, dementia, cardiovascular and respiratory diseases; therefore, the patients with COVID-19 could fit the typical profile of frail individuals: women, elderly, many of them institutionalized in nursing homes, and with comorbidities (Romero Cabrera, 2011). Comorbid conditions are also a notable element as a prognostic factor for the disease; some studies have especially highlighted advanced age, frailty, and the prior presence of chronic diseases (comorbidity) as important risk factors in COVID-19 (Landi et al., 2020). Hence, the control of comorbidities could be a key measure to improve COVID-19 outcomes, especially considering that a large proportion of deaths occurred during influenza pandemics were due to a decompensation of previous medical conditions and their subsequent complications, which is actually being reported during the current pandemic (Garnier-Crussard et al., 2020).

Clinical symptoms and complications

Although CFS does not consider multimorbidity within its classification, it is known that they are related to worse outcomes. Our results showed a similar distribution of multimorbidity (without prioritizing any disease over another) among the three groups with some exceptions: neurological disease (20.45%), dementia (34.09%), heart disease (40.91%), chronic bronchitis (22.73%) and obesity (13.64%) that were more prevalent among frail patients.

Some of the chronic pathologies that were associated with poor prognosis were chronic heart disease, chronic kidney disease and hypertension, which are frequent conditions among elderly and frail people that increase their vulnerability to suffering worse outcomes.

The relationship between frailty/multimorbidity, in view of the results, since multimorbidity and frailty may be bidirectionally related, should lead us to a deep reflection about discuss the role of a multidimensional approach integrating multimorbidity and frailty in the management of hospitalized patients (Boeckxstaens et al., 2015). As Price et al. (2020) argue in their research protocol, it is important to know the association between multimorbidity and frailty with health outcomes, as this could enable better allocation of specialized resources, and especially holistic and shared decision making with patients.

It is well known that respiratory viruses, such as coronaviruses, can produce neuroinvasion and neurological symptoms, in addition to the severe hypoxia that respiratory symptoms can generate, especially in the elderly (Carod Artal, 2020). According to our results, neurological complications from COVID-19 were uncertain, with nonspecific and systemic symptoms (headaches, dizziness, etc.), highlighting the most common and more specific disorders of smell and taste. However, the most remarkable results observed in frail people were alterations of behavior and delirium (43.18%). Mao et al. established that alterations in the level of consciousness can be severe in almost 15% of cases, with only 2.4% considered mild (Mao et al., 2020). However, our results were identical to Mcloughlin et al. (2020) (42%) and higher than that reported by Zazzara et al. (2020) (38%). Hence, an episode of delirium in elderly and frail patients could be the first symptom of COVID-19, as suggested by another author (O’Hanlon & Inouye, 2020). In addition, since frailty implies physical and mental deterioration, these data highlight the importance of including sociopsychological support in our clinical management in order to reduce fatal outcomes among frail patients.

Frailty has been considered a predictor of mortality (Changfeng, Shouyan & Yu, 2019; Dent et al., 2019). In our study, in addition to confirming the relationship between frailty and mortality in patients with COVID-19 (HR: 1.39, 95% CI [1.07–1.81]), as previously hypothesized, delirium behaved as a predictor of death in frail people. Despite that another study (Mao et al., 2020) did not find an association between delirium and mortality, this complication is being studied and some studies have already offered clinical guidance for treating delirium in patients with COVID-19 in ICUs (Kotfis et al., 2020).

We also found, with statistical significance, a lower tendency for COVID-19 pneumonia to be mild in frail people. However, without statistical significance, a greater tendency for pneumonia to be severe in the groups of frail and pre-frail individuals was observed (in more than 60% of the patients of both groups pneumonia was severe, while in the robust group it occurred in 47.73% of the cases). These findings are concerning, given SARS-CoV-2 appears to be a potentially similar virus to SARS-CoV and MERS-CoV-6; in which, elderly patients had a greater tendency to develop serious respiratory diseases, greater complications, more disability and higher mortality (Garnier-Crussard et al., 2020). However, these clinical overviews are similar to the findings of our study, which demonstrated an increased risk of developing severe pneumonia and a high risk of mortality for the elderly, frail, and those with multimorbidity (Weiss & Murdoch, 2020).

A previous study (Miles et al., 2020) has suggested that frailty might not be a good prognostic factor for death; however, our estimations showed a remarkable reduction in survival rates in frail patients which was not observed in prefrail and robust individuals. Other remarkable findings were the significant difference observed regarding the need for ICU admission among frail subjects, and its association with mortality. It was concerning to find that the subgroup of frail people had a mortality rate close to 60%, but only 11.36% were admitted to the ICU. Lower rates were obtained by Hewitt et al. (2020) who reported a mortality rate close to 50% in frail people from the UK, with 7% of terminal patients. However, it must be taken into consideration that during the initial period of the pandemic the burden of hospital strains triggered protocols of limited ICU admissions that discouraged patients who were not considered for resuscitation, i.e., patients with a poor prognosis; therefore, frail individuals could have also been discouraged from ICUs. In addition, due to frail subjects were older than pre-frail and robust individuals, it suggests that age could have been a determining factor in the decision-making of COVID-19 treatment and health resource allocation. Nevertheless, a recent study has concluded that even severely frail elderly patients with a poor prognosis can be benefited from hospitalization for COVID-19, and their survival could increase if enough resources were available for them (De Smet et al., 2020); therefore, the risk of worse outcomes and less ICU admission among the elderly could be better explored to avoid ageism in a setting of limited health resources.

This serious situation is not exclusive to Spain. When COVID-19 hospital admissions exceeded the availability of beds, hospitals in China and Italy (among other countries in Europe) also had to face the difficult task of classifying patients in order to allocate resources (Boreskie, Boreskie & Melady, 2020). Age was an important criterion in the triage of limited resources, as we observed (Boreskie, Boreskie & Melady, 2020); for example, in Italy, generating some rejection by equating old age with frailty (Rosenbaum, 2020). It is remarkable that this situation has affected countries with a very advanced life expectancy, such as Italy or Spain. In the UK, the criterion recommended by the NICE COVID-19 Quick Guide was the assessment of income frailty as a criterion to assist in decision-making in the allocation of resources, with less attention to biological age (NICE Guideline, 2020).

Assessing the level of frailty as a regulatory condition for the allocation of health resources is not new. Some economic studies have previously addressed this problem in response to the high demand for care of frail, elderly, and multimorbidity patients (Cesari et al., 2016). Therefore, it should not be surprising that when health services are strained - such as during the peaks in the COVID-19 pandemic- some patients are ruled out for ICU admission, as seen in our study and in other countries.

Expanding our knowledge about frailty in COVID-19 is essential. The relationship between frailty and mortality, ICU admission, and hospital stay appears clear. However, as a recently published in a systematic review, research should now be directed toward interventions and treatments, given the guidelines for care of frail patients with COVID-19 are not based on strong evidence. As previously addressed, the implications of COVID-19 for the physical and mental health of patients and health professionals should lead us to reflect on the need for developing specific care protocols, based on solid evidence, to facilitate optimal decision-making in vulnerable and frail individuals, especially in terms of the detection of early symptoms for the activation of healthcare processes (Ramírez-Ortiz et al., 2020)

Limitations

Given that our work was a descriptive observational study carried out in a random sample selected from a larger cohort of COVID-19 indivduals, we cannot demonstrate causality in the relationship between the variables studied. In addition, due to the CFS was determined by the revision of medical records there could be slight variability in the assessment when it is performed in a medical setting.