One-Year Mortality Prediction in Elderly Patients With Suspected Pneumonia: The Usefulness of a Holistic Approach

Pneumonia has an impact on long-term mortality in elderly patients. The risk factors associated with poor long-term outcomes are understated. The purpose of this study was to identify the predictors of 1-year mortality in older patients having a suspicion of pneumonia, using usual pneumonia severity scores and geriatric evaluation’s scores focused on comorbidities, nutritional status and functionality. Consecutive patients over 65 years old and hospitalized with a suspicion of pneumonia were enrolled in a monocentric cohort from May 2015 to April 2016. Three scores were used to assess patients’ comorbidities (Cumulative Illness Rating Scale-Geriatric, CIRS-G), malnutrition (Mini Nutritional Assessment, MNA), functionality (Functional Independence Measure, FIM) respectively. Severity of pneumonia was assessed by using the Confusion, Urea, Respiratory Rate, Blood Pressure, and 65-years old score (CURB65), the Pneumonia Severity Index (PSI) and Sequential Organ Failure Assessment score (SOFA). With the exception of CIRS-G, all the scores were obtained prospectively within 48 hours after admission. The main outcome was 1-year mortality. Dates of death were obtained by consulting the cantonal register of deaths. Each score was analysed in univariate and multivariate models and logistic regressions were used to identify contributors to 1-year mortality.


Background
Pneumonia represents one of the greatest causes of hospitalization and mortality from infection in patients aged 65 or older and a major challenge for physicians 1 . Multiple morbidities and age-related modi cations can interfere with the outcome of pneumonia. Effective prognostication at patients' admission is therefore required. The two most widely used and validated tools for determining pneumonia 1-month mortality are Pneumonia Severity Index (PSI) 2 and Confusion, Urea, Respiratory Rate, Blood Pressure, and 65-years old score (CURB-65) 3 . It has also been shown that the Sequential Organ Failure Assessment score and its quick version (SOFA and qSOFA) [4][5] are good predictors of pneumonia prognosis 6 . However, these scores may have limitations in elderly patients because they don't estimate the biological reserve and systemic functionality, which contribute to short-and long-term prognosis 7 .
Hence, in this population it is important to evaluate comorbidities and to perform a comprehensive geriatric assessment, including the evaluation of their functional, cognitive and nutritional status.
Our research aimed to determine predictors of long-term mortality related to pneumonia in older patients, including commonly used pneumonia and sepsis severity scores and other scores evaluating patients' comorbidities (Cumulative Illness Rating Scale-Geriatric, CIRS-G 8 and Charlson Comorbidity Index, CCI 9 ), malnutrition (Mini Nutritional Assessment, MNA 10 ), functionality (Functional Independence Measure, FIM 11 ) and cognitive status (Mini-Mental State Examination, MMSE 12 ). We used a prospective observational cohort including patients over 65 years with a suspicion of pneumonia. We hypothesized that the latter predicted better 1-year mortality than pneumonia and sepsis severity scores.

STUDY DESIGN AND PARTICIPANTS
This study took place at a Department of Internal Medicine, Rehabilitation and Geriatrics in Switzerland, a 1,800-bed tertiary care health institution serving a population of about 500,000 inhabitants.
Two hundred consecutive hospitalized patients aged 65 years or older suspected of pneumonia were enrolled in a prospective cohort study between the 1st of May 2015 and the 30th of April 2016, which aimed to determine whether low-dose Computed Tomography (LDCT) had the capacity to enhance the probability of diagnosing pneumonia in elderly patients, and is described elsewhere 13 .
Brie y, the clinical suspicion of community or hospital-acquired pneumonia (de ned as an infection developing in 2 or more days after hospital admission) was based, in accordance with the Infectious Disease Society of America/American Thoracic Society (IDSA/ATS) guidelines, on the presence of at least 1 respiratory symptom (new or worsening cough, purulent sputum, pleuritic chest pain, new or worsening dyspnoea, respiratory rate > 20 breaths/min, auscultatory ndings or oxygen saturation < 90 % on room air) and at least 1 clinical or serological nding compatible with pneumonia (body temperature > 38°C or < 35°C, CRP > 10 mg/l, white blood cells > 10 G/l with > 85 % neutrophils or band forms).
Patients who had been treated for a pulmonary infection during the previous 6 months, who had already undergone a CT scan during that speci c episode or needed a contrast-enhanced CT, who had to be hospitalized in an Intensive Care Unit (ICU), who had received antibiotics for more than 48 hours before inclusion, or patients considered as unable to give their consent were excluded.

DATA COLLECTION
Data were retrieved from the electronic patient record system. For each included patient, collected data were categorized according to the following domains: demographics, previous treatments, laboratory ndings, and comorbidities. The following scores were obtained prospectively within 48 hours after admission: CURB-65, PSI, SOFA, qSOFA, CCI, MNA, MMSE, and FIM. CIRS-G was retrieved retrospectively by two medical doctors and a research nurse. Dates of death were obtained by consulting the institutional database and the cantonal register of deaths. All scores are brie y described in the Appendix.

OUTCOMES
In this research, we aimed to study the performance of comorbidities (CIRS-G, CCI), malnutrition (MNA), cognitive status (MMSE) and functionality (FIM) assessment tools which were evaluated individually and in a multivariate analysis to predict 1-year mortality in elderly patients hospitalized for suspicion of pneumonia. These were then compared with pneumonia severity scores (CURB-65, PSI), and sepsis severity scores (SOFA and qSOFA).

STATISTICAL ANALYSIS
Numbers and percentages are reported for categorical variables; medians and interquartile ranges (IQRs) are reported for continuous variables with non-normal distributions, whereas means and standard deviations (SDs) are reported for those continuous variables with a normal distribution.
Categorical variables were compared using the χ 2 Test or Fisher Exact Test, while continuous variables were compared using the t-Test or nonparametric Mann-Whitney U Test.
Logistic regression analyses were used to examine associations between long-term mortality and risk factors. In the rst step, each risk factor was tested individually. In the second step, excluding prognostic scores, risk factors showing an association in the univariate model (p < .1) were added to multivariate and adjusted models. To compare the accuracy of comorbidity, malnutrition, cognitive status and functionality assessment tools (CIRS-G, CCI, MNA, MMSE and FIM) to predict mortality, the CURB-65 severity score was added into the logistic regression model. Prognostic values of scores demonstrating an individual capacity to predict 1-year mortality were compared using criteria of sensitivity, speci city, positive and negative likelihood ratios (LR), and diagnostic odds ratios (DOR). The most performing cut-offs were determined using the Youden Index The study was carried out in accordance to the Declaration of Helsinki II Principles (W.M.Association, 2001) and was approved by the local Ethics Committee (CCER 14-250)

Discussion
To our knowledge, the present study is the rst to compare the prognostic value of both pneumonia and sepsis severity scores and scores evaluating comorbidities, malnutrition, cognitive status and functionality in predicting 1-year mortality in an elderly population hospitalized with a suspicion of pneumonia.
In our study, CURB-65 and SOFA were the only pneumonia and sepsis severity scores to prognosticate 1year mortality. Among tools performed for the comprehensive geriatric assessment at patients' admission, CIRS-G, MNA, and FIM evaluating comorbidity, malnutrition and functionality, were strong predictors of 1-year mortality.
Many studies reported a series of risk factors associated with long-term mortality in patients suffering from pneumonia 7 . Indeed, pulmonary infections may have signi cant impacts on various organ systems, such as respiratory, cardiovascular, and neurological ones, leading to the potential worsening of preexisting comorbidities and subsequent higher fatality rates 7 . Therefore, a better understanding of longterm mortality prediction, measured at 30% in our study, seems urgent.
Amongst the risk factors commonly associated with poor long-term outcomes, we investigated the role of comorbidities, malnutrition, functionality and dementia. Concerning the latter, its contribution to elderly patients' prognosis has been studied by Uranga et al. In their research, dementia was found to be the best predictor of one-year mortality in patients hospitalized with community-acquired pneumonia 14 . Similar results were found in a meta-analysis by Foley et al who showed that the odds of pneumonia-associated mortality increased more than 2-fold in patients with cognitive disorders 15 . We found that cognitive disorders were associated with poorer 1-year outcomes but only in univariate analysis.
Another aspect often playing a role in elderly patients' mortality is malnutrition. The MNA was developed as a nutritional screening tool. Using this tool, we were able to identify a very strong correlation between malnutrition and poor outcomes at one year, indicating that assessment of the nutritional status at admission may help in reducing elderly patients' mortality. Few other studies detected similar results.
Yoon et al 16 , studying an elderly population with aspiration pneumonia, identi ed lower BMI and hypoalbuminemia as independent prognostic factors for 5-year mortality. Yeo et al 17 recently highlighted that malnutrition was strongly linked with higher 2-year mortality in people suffering from pneumonia, particularly in the elderly, making essential a routine nutritional assessment at admission. Among elderly patients who have recovered from pneumonia, those who are malnourished have an increased risk of developing impaired muscle and respiratory function, which may lead to more severe long-term outcomes 18 .
Regarding comorbidities, we took into consideration the CCI and the CIRS-G. One of our main ndings was a strong correlation between the CIRS-G and mortality at one year. Similar mortality results, although not focused on a speci c disease, were found in the recent literature. A systematic review on the performance of different morbidity scores to predict mortality in inpatients hospitalized for any medical condition showed that CIRS-G, as per 1 point increase, was signi cantly associated with post-discharge mortality 19 22 . In alignment with the limitation of the Charlson Comorbidity Index, which is considered to be potentially misleading in rating elderly patients' multimorbidity and not su ciently able to predict longterm prognosis in geriatric populations 19-20 , we did not nd any relevant correlation between this score and patients' 1-year mortality.
Amongst several scales estimating patients' dependence and functionality [23][24][25] , we decided to use the FIM, which in a recent study among critically ill elderly patients admitted to an intermediate care unit proved a correlation between low ratings on the scale and higher 1-year mortality rates 26 . Another research showed that frailty, de ned as unintentional weight loss, self-reported exhaustion, weakness slow walking speed, and low physical activity, was strongly associated with the severity of pneumonia and a higher 1-year mortality in older patients, suggesting that frailty should be detected early to improve their management 27 . In our ndings, lower FIM ratings were individually associated with a poorer longterm prognosis but failed in nding signi cant 1-year mortality correlations in multivariate models.
We investigated the ability of the two most widely used prognostic tools (CURB-65 and PSI), validated as 30-day pneumonia mortality predictors, to determine long-term prognosis amongst a population of elderly people. We did not nd any association between the PSI and the prognosis of our patients. Interestingly, CURB-65 was an e cient long-term mortality predictor in our cohort. Similar results were found by Wesemann et al 28 in a predominantly older population.
It has recently been shown that SOFA and qSOFA are useful scores for evaluating pneumonia mortality in geriatric populations [29][30][31] . In our study, we found a weak but signi cant correlation between SOFA scores at admission and 1-year mortality rates, probably due to a persistent impact of organ failure caused by pneumonia on frail people's homeostasis. No signi cant association was found between the qSOFA and long-term prognosis.
The present study's main strength was the consecutive inclusion of elderly patients hospitalized and treated for a suspicion of pneumonia. It has several limitations, however. As a single-centre study carried out with a relatively small number of patients, it should not be generalized to other hospitals. Since we focused our attention on elderly individuals with suspected pneumonia, our results should not be exported to other clinical contexts. Additionally, CIRS-G was retrieved retrospectively from medical records, exposing our analysis to potentials biases. Finally, we did not record the 'Do not resuscitate' orders in our cohort, which might affect patients' outcomes and be a source of bias.

Conclusions
The CIRS-G and MNA were found to be promising prognostic tools for assessing long-term mortality in a cohort of elderly people suffering from suspected pneumonia. Apart from CURB-65 and SOFA, no other pneumonia severity nor sepsis scores could e ciently predict 1-year mortality. Our ndings suggest that a more holistic approach, including nutritional and comorbidity assessments, should be performed when treating older patients with a suspicion of pneumonia and discussing about their prognosis and goals of care. Nevertheless, further studies are needed to evaluate the concrete impact of managing comorbidities and malnutrition on long-term prognosis.