Background

According to projections, the number of individuals in Europe aged above 80 is expected to double by 2050 [31]. The consequent rise in multimorbidity, frailty, and chronic diseases is placing increasing pressure on intensive (ICU) and intermediate care units (IMC) [1, 6]. Understanding challenges linked to this demographic shift is crucial for contemporary societies and healthcare systems [20]. Despite these trends, there is a significant lack of evidence-based knowledge regarding the treatment of older adults, who are often excluded from randomized controlled trials [9]. This knowledge gap raises concerns, particularly given the pivotal role of early frailty recognition—a marker for biological age—in shaping medical care for older patients. Such recognition not only facilitates a more accurate assessment of mortality risk but also positively influences patient outcomes [28]. One precise frailty assessment tool is the Multidimensional Prognostic Index (MPI), based on a comprehensive geriatric assessment (CGA) [36]. Despite its commendable clinimetric properties and successful prognostications across various outcomes, including mortality and rehospitalizations [16, 25], it has rarely been used in critical care units. Another well-established tool for predicting mortality risk, especially in intensive care medicine, is the Sequential Organ Failure Assessment (SOFA) [33]. This study aimed to compare the predictive power of the MPI [24] and of the SOFA for older patients in critical care units not only in terms of mortality risk, but also with respect to quality of life and nursing needs.

Patients and methods

Over a 6-month daily recruitment period, patients (≥ 65 years) admitted to ICU or IMC were enrolled within 48 h. Exclusion criteria were a lack of German language skills, machine-invasive ventilation, immediate life-threatening conditions, or a stay > 48 h in the respective ward before screening. Main endpoint of this study was mortality at 90-day follow-up (FU).

Clinical evaluation

The key assessments were a CGA-based calculation of the MPI as well as the SOFA. Moreover, 17 geriatric syndromes (GS) and 10 geriatric resources (GR) were collected. The GS include incontinence, instability, cognitive alteration, depression or irritability, inanition, sensory impairment, chronic pain, insomnia, irritable bowel syndrome, impoverishment, isolation, immobility, polypharmacy, iatrogenic disease, incoherence/delirium, fluid/electrolyte imbalance, and swallowing disorders. GR comprise favorable intellectual, physical, social, and economic resources, good living conditions, and motivational, emotional, mnemonic, and competence-related resources [19]. Subsequent to recruitment and obtaining consent, study participants underwent a FU visit by phone 30 and 90 days after initial evaluation. Information on survival, admission to long-term care facilities, nursing needs (German: Pflegegrad), GP/outpatient visits, rehospitalizations, drug prescription, falls, and quality of life by means of the European Quality of Life‑5 Dimensions (EQ-5D-5L) scale [10] was collected at discharge and FU. The EQ-5D-5L consists of five dimensions: mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. Each dimension is evaluated on a 5-point scale, ranging from 1 to 5, with 1 indicating no issues and 5 indicating severe impairment. The scores of these individual dimensions can be transformed into a unified index score, ranging from −0.661 to 1, using population-based preference weights. Higher index values, thereby, correspond to greater levels of quality of life [32].

MPI

The MPI comprises eight subscales, including Activities of Daily Living (ADL) [13], Instrumental Activities of Daily Living (IADL) [15], Cumulative Illness Rating Scale—Comorbidity Index (CIRS) [17], Mini Nutritional Assessment Short Form (MNA-SF) [30], Short Portable Mental Status Questionnaire (SPMSQ) [21], Exton Smith Scale (ESS) [3], number of medications, and social housing situation [24]. The resulting continuous MPI value ranges from 0 to 1, with higher values indicating a higher risk. This allows a classification into three risk groups: MPI‑1 (0.0–0.33), MPI‑2 (0.34–0.66), and MPI‑3 (0.67–1.0) [24]. Due to a limited number of patients in the low-risk MPI‑1 group (N = 6), a combined low- and medium-risk group was formed for this analysis.

SOFA

The SOFA score evaluates six organ systems (respiratory, cardiovascular, hepatic, coagulation, renal, and neurological) with a separate score from 0 to 4. For instance, liver function is determined by bilirubin levels, with scores ranging from 0 to 4. The cumulative score is then calculated by adding up the individual dimensions and ranges from 0–24. Higher values reflect a higher mortality risk and organ dysfunction [33]. SOFA scores were computed using pre-existing participant data in this study.

Registration, participant consent, and ethics

This is a prospective observational study, conducted by the Ageing Clinical Research team at the University Hospital Cologne, collaborating with the Department I for Oncology, Hematology, and Internal Intensive Medicine. Ethical approval was granted by the Cologne Ethics Committee (EK19-1049_1, 18 June 2019), and the study was registered in the German Clinical Trials Registry (DRKS00016951). Informed consent was obtained from all patients or their authorized representatives.

Statistical analysis

Statistical analyses were performed using SPSS Statistics (Version 26.0; IBM, Armonk, NY, USA). Descriptive statistics included absolute numbers and relative frequencies for categorical variables and medians with interquartile ranges (IQR) for continuous variables. Given the nonnormal distribution of all variables (Shapiro–Wilk tests), nonparametric methods were employed to compare continuous variables between groups. Rates were compared using Χ2 or Fisher’s exact test. Logistic regression analyses reported as average marginal effects (ME) and adjusted for age and gender were conducted to evaluate the predictive power of MPI and SOFA score on mortality and nursing needs. MEs indicate how the probability of an event changes on average when an independent variable is increased by one unit. They are particularly useful because they allow a more intuitive interpretation of the results of a logistic regression by quantifying the change in probability rather than just looking at the log-odds. Additionally, a receiver operating characteristic (ROC) analysis was employed to compare both scores for mortality risk. Pearson correlation analyses were performed to examine the correlation between MPI and SOFA score with quality of life at both FUs. Ordinary least squares (OLS) regressions were used to check for robustness and control for age and gender.

Results

Study population and demographics

Figure 1 depicts the stepwise inclusion and exclusion of patients, yielding a final sample size of 96. Detailed demographics and clinical characteristics, including MPI and SOFA subgroups, are presented in Table 1. The median age was 74 (44% women). Of all patients, 81.3% were admitted to IMC and 18.7% to ICU. After 3 months, 46.9% had died. At admission, the median MPI was 0.69 (IQR = 0.56–0.75), and the median SOFA score was 4 (IQR = 2–7). The median hospital stay was 25 days, with a daily medication count of 9 (IQR = 7–12). Common admission reasons included infections (30.2%), malignancies (17.7%), and electrolyte disorders (12.5%).

Fig. 1
figure 1

Flow chart of the study

Table 1 Study population

MPI

The MPI showed a relationship with mortality risk after the 90-day FU for older patients admitted to IMC or ICU. The mortality rate in the high-risk group tended to be higher than in the low- and medium-risk group (p = 0.087). When looking at the continuous score, deceased patients had a median MPI of 0.69 (IQR = 0.56–0.78), while survivors scored 0.63 (IQR = 0.50–0.69). A logistic regression showed significant MEs after adjusting for age and gender (p = 0.039). Subassessments revealed poorer ADL (p = 0.005) and IADL (p = 0.022) scores in deceased patients. SPMSQ (p = 0.004) and ESS (p = 0.005) also showed significantly worse results for deceased patients. In a separate evaluation of the MPI for ICU and IMC, only the result for the ICU was statistically significant (ICU p = 0.004 vs IMC p = 0.257).

SOFA

The predictive power of the SOFA score for patients’ outcome was assessed by comparing median scores between deceased and surviving patients at the 90-day FU. The median SOFA score for deceased patients (5, IQR = 3–7) was significantly higher than for survivors (4, IQR = 2–5) (logistic regression, p = 0.014, Mann–Whitney U‑test, p = 0.040). In the SOFA subdomains, deceased patients displayed a significantly lower mean arterial pressure (p = 0.027) and a higher number of taken vasopressors (p = 0.055). As with the MPI, the separate evaluation of the SOFA score for the ICU and IMC only yielded significant results for the ICU (ICU p < 0.01 vs. IMC p = 0.238).

Comparison of the MPI and SOFA score

To compare predictive performance, a ROC analysis, illustrated in Fig. 2, was conducted for MPI and SOFA score in predicting mortality after 90 days. The area under the curve (AUC) was 0.618 (95% confidence interval [CI] 0.502–0.733) for the MPI and 0.621 for the SOFA score (95% CI 0.507–0.735). The sensitivity was 61.0% for the MPI (cut-off = median MPI (0.688)) and 68.9% for the SOFA score (cut-off = median SOFA score (4)). The specificity was 52.9% for MPI and 45.1% for SOFA score.

Fig. 2
figure 2

Prediction of mortality by Multidimensional Prognostic Index (MPI) and the Sequential Organ Failure Assessment (SOFA) score. AUC Area under the curve

Follow-up evaluation

A higher MPI at admission showed a significant negative correlation with quality of life (p < 0.001, r = −0.631 at discharge, p = 0.005, r = −0.377 at 30-day FU; p = 0.004, r = −0.409 at 90-day FU) and was associated with higher nursing needs (Mann–Whitney U‑test, p = 0.002 at 30-day FU; p = 0.011 at 90-day FU) at FU. Figure 3 summarizes the predictive power of MPI and SOFA score for different patient outcomes after the 90-day FU. The figure reports coefficients from age- and gender-adjusted logistic/OLS regressions of different patient outcomes on MPI and SOFA score. As can be seen, both scores significantly predicted mortality rates, with higher scores being associated with a higher mortality risk (MPI: p = 0.039, SOFA score: p = 0.014). By contrast, only the MPI showed a significant relationship when it comes to quality of life (MPI: p = 0.006, SOFA score: 0.172) and nursing needs (MPI: p = 0.001, SOFA score: p = 0.248). A higher MPI was correlated with lower quality of life and a higher likelihood of the patient having nursing needs. Even though the effects went in the same direction for the SOFA score, they were not significant. In addition to MPI and SOFA score, people who survived also differed from those who died with respect to geriatric resources and syndromes. Survivors had overall more GR (p = 0.046)—physical (p = 0.005), intellectual (p = 0.018), and general competencies (p = 0.005)—than patients who were deceased at FU. When it comes to GS, survivors showed less instability (p = 0.039), cognitive impairment (p = 0.027), and delirium/incoherence (p = 0.046) than patients deceased at FU.

Fig. 3
figure 3

Predictive power of the Multidimensional Prognostic Index (MPI, a) and the Sequential Organ Failure Assessment (SOFA) score (b) for different patient outcomes (mortality, quality of life, and nursing needs). Each estimate is the result of a separate regression. In the case of mortality and nursing needs, estimations are based on logistic regressions, and in the case of quality of life, estimates are based on ordinary least squares (OLS) regressions. All regressions are adjusted for patient age and gender. Points on the horizontal axes represent the size of the marginal effects, and whiskers indicate 95% confidence intervals. ***p < 0.001, **p < 0.01, *p < 0.05

Discussion

Due to the demographic change and the rising age of patients globally, the first AWMF (Arbeitsgemeinschaft der wissenschaftlichen medizinischen Fachgesellschaften) guideline was just developed to support clinical decision-making in hospitalized older patients [2]. This study aimed to assess the efficacy of the CGA-based MPI in comparison to the SOFA score for predicting mortality risk, nursing needs, and quality of life among older patients in critical care units. The investigation reveals that both instruments are comparable in terms of prognostic power for mortality. In particular the MPI showed highest accuracy in predicting not only mortality risk but also quality of life and nursing needs.

The FU assessments after 30 and 90 days showed that the MPI, but not the SOFA score was correlated with a lower quality of life and a higher probability of having nursing needs, indicative of greater prognostic complexity and healthcare needs [5, 12, 27]. This suggests that the physiological assessment captured by the SOFA score may not fully encapsulate the multidimensional aspects of health status and nursing needs among older patients in critical care settings, as captured by the MPI. In line with our findings, a study conducted in the emergency department already showed that there is a correlation between poor quality of life and a higher MPI [29].

Previous research has consistently established the MPI’s significant and sensitive association with various outcomes, including mortality, hospital stays, nursing needs, (re)hospitalization, and geriatric syndromes [8, 19, 22,23,24]. While affirming the existing evidence in the geriatric field, this study extends the application of the MPI to ICU and IMC. The research emphasizes that the major impact on poor outcomes is rooted in social, functional, cognitive, and nutritional aspects rather than chronological age alone [14]. Long-term health is a multifaceted concept, encompassing not only the absence of disease but also mental, social, and overall well-being [35]. Recognizing the importance of frailty and individual biological age, the MPI emerges as a suitable tool for objectively assessing frailty as a marker of biological age in the clinical setting [11, 28]. Additionally, the MPI contributes essential prognostic information and aids in clinical decision-making by classifying patients into risk groups [26, 34, 36]. The results of the study indicate that also in critical care units, a higher MPI is associated with higher mortality rates after 3 months, addressing a critical gap in intensive care decision-making.

Traditionally, the MPI has been predominantly employed for geriatric patients in internal or geriatric wards, while the SOFA is commonly utilized in intensive care units [7, 18, 24]. The study’s results demonstrate that both the MPI and SOFA score exhibit comparable performance in predicting mortality after 3 months (MPI p = 0.039 vs. SOFA score p = 0.014). The ROC analyses indicate similar sensitivity (61.0% vs. 68.9%), specificity (52.9% vs. 45.1%), and nearly identical AUC values (0.618 vs. 0.621).

While both tests effectively predict mortality, the study may initially suggest that the SOFA score is more suitable for critical care scenarios due to its shorter assessment time (approximately 10 min) compared to the MPI (approximately 30 min). However, a 2022 study introduced a short version of the MPI, called BRIEF-MPI, which was shown to be as accurate as the original MPI with an average application time of 5 min [4]. Given our study’s findings of comparable mortality risk prediction after 90 days between the MPI and SOFA score and the MPI’s superior performance in assessing quality of life and nursing needs, the MPI may be a better tool for older patients in critical care than the SOFA score.

The MPI provides valuable insights into the social and functional characteristics, quality of life, and nursing needs of older intensive care patients. The results suggest that the SOFA score is sufficient when the main interest is in understanding mortality. If, by contrast, the aim is to understand other aspects of health status and quality of life, the MPI may be more suitable. Furthermore, the comparable performance of both scores in the overall cohort is mirrored in the subgroups of ICU and IMC, although larger data collection is needed for conclusive evidence. The fact that the original MPI provides estimates of mortality risk up to 1 year [24] and, thus, exceeds the time frame of SOFA implies further potential superiority for longer FU periods. This emphasizes the importance of future research with longer FU periods to draw more differentiated conclusions.

Acknowledging limitations, the study is constrained by a single-center retrospective design, a relatively modest sample size, and observed attrition between FU assessments. While retention rates align with other studies involving multimorbid patients, findings may require replication across diverse centers with varying environmental factors. The utilization of distinct evaluators introduces potential interexaminer variations, compounded by the inability to cross-validate their assessments. Additionally, the pooling of observations from ICU and IMC, with an emphasis on IMC, introduces a potential source of bias, warranting consideration in the interpretation of results.

Summary

  • Person-centered, multidimensional prognostic evaluation and organ-based prognosis are equally valid instruments for clinical decision-making in older, critically ill patients.

  • The Multidimensional Prognostic Index (MPI) is superior to the Sequential Organ Failure Assessment (SOFA) score when it comes to predicting quality of life and nursing needs (Pflegegrad) up to 3 months after ICU discharge.

  • Findings underscore the importance of broader validation for both MPI and SOFA score, revealing a research gap in treatments for older patients in critical care settings.

  • Further research is needed to determine whether MPI excels in assessing long-term prognosis for older critically ill patients, addressing a current knowledge gap.