Bedside cognitive screening to detect dementia and predict poor outcomes in hospitals

Abstract

INTRODUCTION

Cognitive impairment in hospitalized older adults without delirium is often undiagnosed. We evaluated the diagnostic accuracy and prognostic value of two brief bedside screening tools—the 10‐point Cognitive Screener (10‐CS) and Short Portable Mental Status Questionnaire (SPMSQ).

METHODS

Multicenter prospective cohort of 2003 patients ≥65 years of age without delirium, admitted to 43 hospitals in five countries of the Creating a Hospital Assessment Network in Geriatrics (CHANGE) Study. Screening was completed within 48 h of admission. Dementia was defined as informant‐based Clinical Dementia Rating ≥1.

RESULTS

Dementia prevalence was 22%, with 35% undiagnosed. Both tools showed good accuracy for detecting dementia, with the 10‐CS slightly more accurate than SPMSQ (area under the curve [AUC] = 0.87 vs 0.85; p = 0.02). After adjustment, cognitive impairment detected by either tool was associated with increased risk of incident delirium, hospital‐associated disability in self‐care activities, and 90‐day mortality, but not prolonged stay.

DISCUSSION

Brief bedside screening identified dementia and predicted adverse outcomes, supporting its integration into routine hospital care.

Highlights

• The 10‐pt Cognitive Screener (10‐CS) and Short Portable Mental Status Questionnaire (SPMSQ) are brief bedside tools requiring no equipment for hospital use.

• They had good accuracy for detecting dementia in older inpatients without delirium.

• The 10‐CS showed slightly higher accuracy than the SPMSQ for detecting dementia.

• Cognitive impairment by either tool predicted delirium, disability, and mortality.

• Brief cognitive screening may improve recognition of dementia and guide acute care.

1. BACKGROUND

As populations age, dementia has become a major global public health concern, currently affecting an estimated 57 million people, with the fastest growth projected in lower‐ and middle‐income countries. ¹ This demographic shift is reshaping hospital care, as more older adults with dementia are admitted for acute medical conditions. ² , ³ Dementia refers to an acquired, progressive syndrome of cognitive and functional decline that interferes with independence in everyday activities. ¹ It is present in ≈40% of hospitalized older adults, and this proportion is likely higher when cognitive impairment is considered—a broader construct that includes dementia and mild cognitive impairment, characterized by detectable cognitive decline exceeding normal aging without interference with independence in daily activities. ² , ³ Dementia and cognitive impairment complicate clinical management and increase the risk of adverse outcomes, in part by increasing vulnerability to delirium. ⁴ , ⁵

Delirium is an acute, fluctuating syndrome of inattention and altered awareness that develops over hours to days, often accompanied by additional disturbances (e.g., disorientation or memory impairment), and precipitated by an underlying medical condition or drug effect. ⁶ Timely recognition of preexisting cognitive impairment may help clinicians anticipate complications, tailor treatment strategies, and support safer transitions. ⁷ Despite its relevance, cognitive impairment frequently goes undetected in hospitals. ⁸ , ⁹ Assessment is challenged by operational, environmental, and clinical barriers. ⁷ Clinicians face time constraints and often lack formal training in cognitive evaluation. Environmental factors (e.g., noise, lack of privacy, and frequent interruptions) interfere with a patient's ability to follow instructions. ¹⁰ In addition, fatigue, acute illness, pain, and centrally active medications may impair cognition and blur the distinction between preexisting cognitive impairment and delirium. ⁷ , ¹⁰

Although delirium is a well‐known predictor of poor outcomes, it reflects only part of a patient's cognitive vulnerability. ¹¹ Because cognitive assessment is unreliable in the presence of delirium, a practical approach is to screen for delirium first. When present, patients can be considered cognitively vulnerable and reassessed for dementia after recovery; when absent, cognitive screening can be undertaken during hospitalization. ⁴ , ¹¹ However, commonly used tools such as the Mini‐Mental State Examination (MMSE) and the Montreal Cognitive Assessment (MoCA) require motor coordination, sustained attention, and visual processing skills, which are often impaired in acutely ill patients. ¹² In addition, hospital circumstances, such as intravenous access in the dominant arm or the absence of corrective lenses, may further hinder test performance. Other comprehensive tools are too lengthy to be feasible in acute care settings. ¹³

Brief cognitive screeners are pragmatic tools designed for rapid identification of cognitive impairment. ¹⁴ , ¹⁵ Unlike comprehensive neuropsychological assessments or instruments such as the Clinical Dementia Rating (CDR) scale, these screeners can be completed quickly, require limited training, and do not require an informant, but they do not provide a definitive dementia diagnosis or staging. ⁴ , ¹⁶ Their primary indication is to identify patients who warrant more comprehensive evaluation. Validated tools such as the 10‐point Cognitive Screener (10‐CS) and the Short Portable Mental Status Questionnaire (SPMSQ) are feasible at the bedside, impose minimal motor or visual demands, and do not rely on ancillary materials, which is advantageous in acute care. ¹⁴ , ¹⁵ Because hospitalization‐related factors (e.g., pain, medications, and the illness prompting admission) may affect performance, brief screeners are best viewed as identifying cognitive difficulties that should be reassessed after stabilization and, when indicated, followed by further evaluation. ⁷ , ⁸ , ⁹ In this context, they support pragmatic standardization of inpatient cognitive screening when delirium is absent. ⁴ However, their diagnostic accuracy for identifying patients at risk for underlying dementia during hospitalization remains underexplored, as does their prognostic value in hospitalized older populations from low‐ and middle‐income countries. ¹⁷

This study aimed to: (1) evaluate the diagnostic accuracy of the 10‐CS and SPMSQ for detecting dementia in a large cohort of hospitalized older adults without delirium across 43 hospitals, using the validated informant‐based Clinical Dementia Rating (IB‐CDR) as the reference standard ⁸ and (2) examine the association between these screening scores and adverse outcomes, including in‐hospital incident delirium, hospital‐associated disability, prolonged stay, and 90‐day mortality.

2. METHODS

2.1. Study design, setting, and population

This study was conducted within the prospective multicenter CHANGE (Creating a Hospital Assessment Network in Geriatrics) cohort, a large international initiative involving 43 hospitals across five countries: Brazil (n = 38), Angola (n = 1), Chile (n = 1), Colombia (n = 2), and Portugal (n = 1). ⁵ For the present analyses, diagnostic accuracy was evaluated using cross‐sectional baseline data, and associations of baseline cognitive impairment with adverse outcomes, including 90‐day mortality, were examined using the prospective follow‐up component of the cohort. Most participating sites were recruited through TeleGero, a tele‐education network for geriatric care that has been active for over two decades. This network has nationwide coverage in Brazil and includes an international center in Porto, Portugal. ⁵ Additional hospitals not affiliated with TeleGero were enrolled after expressing interest in participating in the CHANGE Study following outreach by the coordinating center, thereby increasing the diversity of geographic regions, health care funding models, and hospital care structures represented. All participating sites were located in urban areas; 86% were in central urban regions and 14% in peripheral urban regions. A complete list of participating hospitals is provided in Table S1. The overarching objective of the CHANGE Study was to promote the use of rapid geriatric assessment tools adapted to the needs of acutely hospitalized older adults, particularly in lower‐ and middle‐income countries. ⁵ , ⁸

Eligible participants were adults 65 years of age or older, consecutively admitted to internal medicine or geriatric wards between June 2022 and December 2023, and assessed within 48 h of admission (n = 4464). We excluded patients with terminal illness under exclusive palliative care (n = 232), hospital stays shorter than 48 h (n = 892), individuals without consent (n = 518), language barrier (n = 1), readmissions of previously enrolled participants (n = 177), and incomplete data, including missing baseline measures (n = 47), most commonly missing delirium assessment, and loss to follow‐up (n = 41). We additionally excluded 553 patients with delirium at admission, as identified by the Confusion Assessment Method (CAM), ¹⁸ because delirium can compromise the validity of bedside cognitive testing and preclude detection of any superimposed preexisting or longstanding cognitive difficulties. ⁴ , ⁶ The final sample comprised 2003 older inpatients (Figure S1). This sample size was not calculated a priori but was defined pragmatically as all consecutively enrolled CHANGE Study participants without delirium at admission during the recruitment period.

The study protocol was approved by the research ethics committee at each participating site. Written informed consent was obtained from all participants or a representative in cases of impaired decision‐making capacity. The study is registered in the Brazilian Registry of Clinical Trials (ReBEC; RBR‐6p5bvvg) and adheres to the 2015 Standards for Reporting Diagnostic Accuracy Studies (STARD 2015) guidelines (see Appendix in the online supplementary material). ¹⁹

RESEARCH IN CONTEXT

1. Systematic review: We searched PubMed, Scopus, and Web of Science for studies on cognitive screening in hospitals. Dementia is common but often undetected in acute care. Brief cognitive tools perform well in outpatients, but their hospital use is limited by clinical and operational barriers such as delirium and time constraints. Evidence on their performance in detecting dementia and predicting adverse outcomes in hospitals remains scarce.

2. Interpretation: Our findings demonstrate that the 10‐point Cognitive Screener (10‐CS) and Short Portable Mental Status Questionnaire (SPMSQ) accurately identify dementia in hospitalized older adults without delirium. Cognitive impairment detected by either tool predicted incident delirium, hospital‐associated disability in self‐care activities, and 90‐day mortality, supporting the integration of brief cognitive screening into routine hospital care.

3. Future directions: Implementation studies should assess whether integrating brief cognitive screening into hospital workflows improves outcomes for older inpatients, a priority as health systems worldwide face the challenges of population aging.

2.2. Data collection

Trained local investigators conducted a standardized baseline assessment within 48 h of hospital admission using structured interviews with the patient whenever they could reliably participate, complemented by hospital chart review. Decision‐making capacity was assessed clinically; when capacity was impaired, interviews and consent were obtained from a proxy (legally authorized representative) identified via the medical record or the attending team. Assessments were performed within 48 h of admission to capture early baseline status and minimize bias from subsequent in‐hospital complications, treatments, or clinical deterioration that could affect cognitive performance. ⁶ This comprehensive assessment included sociodemographic characteristics, medical history, frailty, functional status before hospitalization, severity of acute illness, and cognitive screening. Sociodemographic data included age (in years), sex at birth (male or female), self‐identified race/ethnicity (categorized as White, Black, or Other), years of formal education, and marital status (dichotomized as partnered [married or in a relationship] vs unpartnered [single, divorced, or widowed]). The primary reason for hospitalization was extracted from medical records and grouped into major system‐based categories. Comorbidities were measured using the Charlson Comorbidity Index, a validated summary measure of overall disease burden. ²⁰ We also recorded prior diagnoses of dementia or mild cognitive impairment (MCI) based on patient and/or proxy reports of a prior clinician diagnosis and review of the medical record (problem list and prior documentation); these sources served as complementary information to capture prior clinical recognition. Baseline frailty was assessed using the Fatigue, Resistance, Ambulation, Illnesses, and Loss of weight (FRAIL) Scale, a five‐item screening tool that evaluates fatigue, resistance, ambulation, illness count, and unintentional weight loss over the 4 weeks preceding admission. ²¹ FRAIL scores range from 0 to 5 and classify individuals as robust (0), pre‐frail (1–2), or frail (3–5). Functional status was measured using the Katz Index of Activities of Daily Living (ADL), which evaluates independence in performing six self‐care activities—bathing, dressing, toileting, transferring, eating, and continence—based on the patient's condition 4 weeks before admission. ²² Acute illness severity at admission was quantified using the National Early Warning Score 2 (NEWS‐2), calculated from routinely collected physiological parameters (respiratory rate, oxygen saturation, temperature, systolic blood pressure, heart rate, and level of consciousness). ²³

All variables were collected using standardized electronic case report forms hosted in Research Electronic Data Capture (REDCap), which featured real‐time data validation checks and completeness monitoring. ²⁴ A centralized coordination team supervised the process to ensure protocol fidelity and data quality across participating sites.

2.2.1. Cognitive assessment

Cognitive status was assessed using informant‐based and patient‐administered tools within 48 h of hospital admission. Trained investigators (primarily geriatric‐trained physicians and nurses) conducted structured interviews with a knowledgeable informant—defined as someone who had contact with the patient at least once a week over the 6 months preceding admission—to administer the IB‐CDR. ⁸ When applicable, the informant was the same person as the proxy. Interviews were conducted in person whenever possible; when the informant was unavailable on‐site, the IB‐CDR was administered by telephone. ⁸ This validated tool assesses cognitive and functional abilities across six domains (memory, orientation, judgment and problem‐solving, community affairs, home and hobbies, and personal care) based on the patient's baseline status ≈3 months before hospitalization. Each domain is scored on a scale from 0 (no impairment) to 3 (severe impairment), and a global score is calculated following standardized rules. ²⁵ , ²⁶ The IB‐CDR global score was subsequently assigned by a blinded assessor not involved in the baseline assessment using standard CDR scoring rules. ²⁵ It was used to classify cognitive status (0 = no impairment; 0.5 = questionable dementia; 1 = mild dementia; 2 = moderate dementia; 3 = severe dementia), with scores ≥1 indicating dementia. The sum of domain scores was also used to compute the IB‐CDR Sum of Boxes, providing a continuous measure of severity. ²⁷ The diagnostic criteria used for the IB‐CDR have been established previously and validated in this population. ⁸ It demonstrated excellent diagnostic accuracy for detecting dementia in hospitalized older adults, with a global score ≥1 showing an area under the receiver‐operating characteristic (ROC) curve (AUC) of 0.92 (95% confidence interval [CI]: 0.85–0.98) against a gold‐standard clinical diagnosis, supporting its use as a reference standard for hospital‐based dementia detection and staging. ⁸

As part of the comprehensive baseline assessment, investigators administered two brief cognitive screening tools at the bedside to all patients, first the 10‐CS and then the SPMSQ. ¹⁴ , ²⁸ Both instruments were administered by local investigators after brief standardized training to ensure consistent administration across sites. The 10‐CS is a verbally administered instrument that assesses temporal orientation (day, month, year), verbal fluency (animal naming in 1 min), and three‐word recall. It takes ≈2 min to complete, requires no written materials or motor tasks, and yields a total score from 0 to 10, with higher scores indicating better cognitive performance (0–5: probable cognitive impairment; 6–7: possible cognitive impairment; 8–10: normal cognition). Patients with very low literacy receive score adjustments (illiterate: +2 points; 1–3 years of education: +1 point), with the final score capped at 10. In the validation study, the 10‐CS showed excellent diagnostic accuracy for dementia, outperforming the MMSE. ¹⁴ A cutoff of ≤5 has also been linked to a higher risk of adverse outcomes in hospitalized older adults. ⁴ , ¹⁶

The SPMSQ is also a verbally administered, 10‐item instrument designed to easily assess cognitive function in older adults, and it typically takes about 5 min to administer. ¹⁵ , ²⁹ It includes questions assessing orientation to time and place, recall of personal information (e.g., birthdate, mother's name), general knowledge (e.g., the names of current and past presidents), and concentration (e.g., serial subtraction). Each response is scored as either correct or incorrect, resulting in a total error score ranging from 0 to 10, with lower scores indicating better cognitive performance. Cognitive status is categorized as normal cognition (0–2 errors), mild impairment (3–4 errors), moderate impairment (5–7 errors), or severe impairment (8–10 errors). ¹⁵ Because educational attainment can influence performance, one additional error is allowed in interpretation for individuals with ≤5 years of education. The SPMSQ has shown excellent diagnostic accuracy for dementia, and it correlated highly with the MMSE. ³⁰

Investigators administering the cognitive assessment had access to clinical information recorded during the baseline evaluation and available in the hospital medical records. We did not systematically collect adverse events related to the administration of the IB‐CDR, 10‐CS, or SPMSQ.

2.2.2. Outcomes

The study outcomes were incident delirium during hospitalization, hospital‐associated disability in ADL, prolonged hospital stay, and all‐cause mortality within 90 days of hospital admission. In‐hospital outcomes were registered during the final 24 h of hospitalization. Incident delirium, defined as a new onset after hospital admission, was identified by staff physicians trained in the Confusion Assessment Method (or CAM). ¹⁸ Hospital‐associated ADL disability was defined as new dependence in at least one basic ADL (bathing, dressing, toileting, transferring, eating, and continence) at discharge compared with the baseline level 4 weeks before admission, excluding patients who were entirely dependent at baseline. ²² Prolonged hospital stay was defined as a length of stay in the upper tertile of the sample distribution. ³¹ Vital status was recorded at discharge. Patients discharged alive were followed up at 30 and 90 days through structured telephone interviews conducted by trained investigators blinded to baseline assessments, and mortality status was additionally confirmed through linkage with official national death registries.

2.3. Statistical analysis

We conducted descriptive analyses to summarize participant characteristics by dementia status, as defined by the IB‐CDR. Continuous variables were reported as means with standard deviations (SD) when approximately normally distributed or medians with interquartile ranges (IQRs) when non‐normally distributed. Normality was assessed visually using histograms. Group comparisons used two‐sample t‐tests for approximately normal variables and the Mann–Whitney U test for skewed variables. Categorical variables were described as frequencies and percentages and compared using chi‐square tests.

We examined the correlations between 10‐CS and SPMSQ scores and the IB‐CDR Sum of Boxes using scatter plots with a smoothed regression line and Spearman's rank correlation coefficients; values between 0.36 and 0.67 were interpreted as moderate and values ≥0.68 as strong correlations. ³²

We evaluated the diagnostic performance of the two brief cognitive screening tools—the 10‐CS and SPMSQ—against the IB‐CDR as the reference standard. We calculated the AUCs with 95% CIs derived from bias‐corrected and accelerated (BCa) bootstrap methods with 1000 iterations. AUC values of 0.70–0.79 were considered fair, 0.80–0.89 good, and ≥0.90 excellent discrimination. ³³ Comparisons between AUCs were conducted using nonparametric methods. ³⁴ Sensitivity, specificity, likelihood ratios, and predictive values were calculated for multiple cutoff points, with optimal thresholds selected based on the highest Youden index (sensitivity + specificity − 1). We compared agreement between the 10‐CS and SPMSQ at their optimal cutoffs using Cohen's kappa, and illustrated the overlap in a Venn diagram. We also described participant characteristics according to 10‐CS and SPMSQ categories.

We computed Kaplan–Meier curves for 90‐day mortality by cognitive screening category and compared them using log‐rank tests. We fitted Weibull proportional hazards models to examine the association of preexisting cognitive impairment, defined by the 10‐CS and SPMSQ, with 90‐day mortality. For in‐hospital adverse outcomes, we used mixed‐effects logistic regression models. All models included random intercepts to account for clustering, specified at the state/province level and the level of each participating hospital (study center). We used a sequential adjustment strategy to quantify how effect estimates changed after adjustment for sociodemographic factors and then additional clinical and hospital‐level covariates. ³⁵ Three sequential models were specified: (1) Model 1 (base model) included random intercepts only; (2) Model 2 adjusted for sociodemographic variables (age, sex, race/ethnicity, education, and marital status); and (3) Model 3 included all covariates from Model 2, plus the Charlson Comorbidity Index (excluding dementia), baseline frailty status, illness severity (NEWS‐2), and hospital type (public vs private). In addition, we applied the same models, treating 10‐CS and SPMSQ scores as continuous measures (0–10), to further assess associations with baseline cognitive status. We also assessed whether adding the 10‐CS or SPMSQ separately improved outcome discrimination by comparing the AUCs of Model 2 with and without each tool. ³⁴

To test the robustness of our findings, we conducted several sensitivity and complementary analyses. First, we explored potential cultural variation by stratifying the sample into Brazil versus other countries. Second, we assessed whether the diagnostic performance of the brief cognitive tools varied by education level, given prior reports of reduced SPMSQ accuracy in individuals with fewer than 6 years of schooling. ²⁸ Third, we examined an alternative definition of preexisting cognitive impairment based on documented or reported diagnoses of dementia or MCI, acknowledging that brief tools may not outperform prior clinical information. Finally, to address the potential influence of in‐hospital death on outcome assessment, we repeated analyses of in‐hospital adverse outcomes in patients who survived to discharge. For hospital‐associated ADL disability, additional analyses were limited to participants who were fully independent before admission to ensure comparable baseline function. ¹² , ³⁶

We conducted complete‐case analyses using Stata version 17.0 (StataCorp LLC, College Station, TX), and statistical significance was set at p < 0.05 (two‐sided).

3. RESULTS

3.1. Cohort characteristics

Among the 2003 patients included, the mean (SD) age was 78.1 (8.6) years; 56.0% were women, 52.1% identified as White, and the median (IQR) education was 5 (3–10) years (Table 1). The most common reasons for hospitalization were infectious diseases (36.2%), evaluation and management of constitutional symptoms such as unintentional weight loss (18.1%), acute cardiovascular conditions (15.6%), and acute kidney injury (9.5%).

TABLE 1.

Participant characteristics by dementia status (n = 2003).

		No dementia	Dementia	p‐value
	Total	(IB‐CDR <1)	(IB‐CDR ≥1)
Variables	(n = 2003)	(n = 1559)	(n = 444)
Brief cognitive screening tools
10‐CS (0–10), median (IQR)	6 (4, 8)	7 (5, 8)	2 (1, 4)	<0.001
10‐CS ≤5, n (%)	853 (42.6)	462 (29.6)	391 (88.1)	<0.001
SPMSQ (0–10), median (IQR)	2 (1, 5)	2 (1, 3)	7 (4, 9)	<0.001
SPMSQ ≥5, n (%)	548 (27.4)	238 (15.3)	310 (69.8)	<0.001
Sociodemographic factors
Age, years, mean (SD)	78.1 (8.6)	76.8 (8.1)	82.7 (8.4)	<0.001
Women, n (%)	1121 (56.0)	822 (52.7)	299 (67.3)	<0.001
Race/ethnicity, n (%)				0.007
White	1043 (52.1)	785 (50.4)	258 (58.1)
Black	911 (45.5)	738 (47.3)	173 (39.0)
Other	49 (2.4)	36 (2.3)	13 (2.9)
Education, years, median (IQR)	5 (3, 10)	5 (3, 9)	4 (3, 10)	0.38
Married or in a partnership, n (%)	911 (45.5)	764 (49.0)	147 (33.1)	<0.001
Clinical measures
Charlson comorbidity index, median (IQR)	2 (1, 5)	2 (1, 5)	2 (1, 4)	0.05
Prior dementia or MCI diagnosis, n (%)	370 (18.5)	81 (5.2)	289 (65.1)	<0.001
Frail status (FRAIL Scale), n (%)				<0.001
Robust (0)	266 (13.3)	253 (16.2)	13 (2.9)
Pre‐frail (1–2)	789 (39.4)	642 (41.2)	147 (33.1)
Frail (3–5)	948 (47.3)	664 (42.6)	284 (64.0)
Katz ADL Index (0–6), median (IQR)	6 (5, 6)	6 (5, 6)	2 (1, 5)	<0.001
Acute illness severity (NEWS‐2), n (%)				<0.001
Low to mild risk (0–4)	1304 (65.1)	1102 (70.7)	202 (45.5)
Moderate (urgent) risk (5–6)	384 (19.2)	263 (16.9)	121 (27.3)
High (emergency) risk (≥7)	315 (15.7)	194 (12.4)	121 (27.3)
Health care system
Public hospital, n (%)	1524 (76.1)	1236 (79.3)	288 (64.9)	<0.001

Note: p‐values represent differences between patients with and without dementia, as defined by the IB‐CDR, and were calculated using the chi‐square test for categorical variables and the two‐sample t‐test or Mann–Whitney U test for continuous variables, as appropriate.

Abbreviations: IB‐CDR, informant‐based Clinical Dementia Rating; 10‐CS, 10‐point Cognitive Screener (lower score = worse performance), with scores ≤5 indicating probable cognitive impairment; SPMSQ, Short Portable Mental Status Questionnaire (higher score = worse performance), with scores ≥5 indicating moderate cognitive impairment; IQR, interquartile range; SD, standard deviation; MCI, mild cognitive impairment; NEWS‐2, National Early Warning Score 2; ADL, activities of daily living, with Katz ADL Index based on the patient's health status 4 weeks prior to admission (higher scores indicating greater independence).

According to the IB‐CDR global score, 709 participants (35.4%) had no cognitive impairment (IB‐CDR = 0), 850 (42.4%) had questionable dementia (IB‐CDR = 0.5), and 444 (22.2%) met criteria for dementia (IB‐CDR ≥ 1), including 220 (11.0%) with mild dementia (IB‐CDR = 1), 94 (4.7%) with moderate dementia (IB‐CDR = 2), and 130 (6.5%) with severe dementia (IB‐CDR = 3). Among patients with dementia, 34.9% had no prior diagnosis of dementia or MCI. Patients with dementia had lower scores on the 10‐CS and higher scores on the SPMSQ compared to those without dementia. The prevalence of cognitive impairment, defined by 10‐CS ≤ 5, was 88.1% in the dementia group versus 29.6% in the non‐dementia group (p < 0.001); for SPMSQ ≥ 5, prevalence was 69.8% versus 15.3% (p < 0.001). In addition, compared with patients without dementia, those with dementia were older, more often women, more likely to be unpartnered, frail, and functionally dependent before admission, and had higher acute illness severity at presentation (Table 1).

3.2. Diagnostic accuracy of brief cognitive screening tools

The correlation between the 10‐CS and the IB‐CDR Sum of Boxes was Spearman's rho = ‐0.59 (95% CI: ‐0.62 to ‐0.56). For the SPMSQ, the correlation was rho = 0.57 (95% CI: 0.54–0.60) (Figure 1).

FIGURE 1.

Scatter plot showing the correlation between the brief cognitive screening tools and Clinical Dementia Rating (CDR) Sum of Boxes scores (n = 2003). 10‐CS = 10‐point Cognitive Screener (higher score = better performance); SPMSQ = Short Portable Mental Status Questionnaire (higher score = worse performance). Correlations: 10‐CS, Spearman's rho = ‐0.59 (95% confidence interval [CI]: ‐0.62 to ‐0.56); SPMSQ, rho = 0.57 (95% CI: 0.54–0.60).

The AUC for dementia (IB‐CDR ≥1) was 0.87 (95% CI: 0.85–0.89) for the 10‐CS and 0.85 (95% CI: 0.83–0.87) for the SPMSQ, with a p = 0.02 for the difference in AUC between the two tools (Figure 2). AUC values were similar between Brazilian and non‐Brazilian participants. Among participants with ≥6 years of schooling, AUCs were 0.89 (95% CI: 0.86–0.92) for the 10‐CS and 0.88 (95% CI: 0.85–0.91) for the SPMSQ. Among those with <6 years (58% of the sample), AUCs were 0.86 (95% CI: 0.83–0.88) and 0.84 (95% CI: 0.82–0.87), respectively (Table S2). The AUC for prior clinical documentation of dementia or MCI was 0.80 (95% CI: 0.78–0.82), which was significantly lower than that of the 10‐CS (p < 0.001) and the SPMSQ (p < 0.001).

FIGURE 2.

Receiver‐operating characteristic (ROC) curves of brief cognitive screening tools for detecting dementia (n = 2003). 10‐CS = 10‐point Cognitive Screener; SPMSQ = Short Portable Mental Status Questionnaire; AUC = area under the receiver‐operating characteristic (ROC) curve; 95% CI = 95% confidence intervals (calculated using bootstrap with 1000 iterations). Dementia was defined as a score ≥1 on the informant‐based Clinical Dementia Rating (IB‐CDR) scale. AUC comparison (10‐CS vs SPMSQ), p = 0.02.

At a cutoff of ≤5, the 10‐CS had a sensitivity of 88.4% and specificity of 70.5%. For the SPMSQ, a cutoff of ≥5 yielded a sensitivity of 69.6% and a specificity of 84.9% (Table 2). At the optimal cutoff values, the 10‐CS and SPMSQ demonstrated high negative predictive values (95.4% and 90.8%, respectively) but lower positive predictive values (49.8% and 56.6%, respectively), as shown in Table 2.

TABLE 2.

Sensitivity, specificity, likelihood ratios, and predictive values for dementia detection at the cutoff points of each brief cognitive screening tool (n = 2003).

Cutoff	Sensitivity (95% CI)	Specificity (95% CI)	LR+	LR–	PPV	NPV
10‐CS
=0	24.3 (20.4–28.6)	99.0 (98.4–99.5)	25.28	0.76	87.8	82.1
≤1	37.2 (32.7–41.8)	97.1 (96.1–97.8)	12.59	0.65	78.2	84.4
≤2	54.3 (49.5–59.0)	93.7 (92.4–94.9)	8.63	0.49	71.1	87.8
≤3	67.3 (62.8–71.7)	88.3 (86.6–89.9)	5.77	0.37	62.2	90.5
≤4	77.6 (73.8–81.7)	80.7 (78.9–82.9)	4.09	0.27	52.8	91.8
≤5	88.4 (84.7–90.9)	70.5 (68.0–72.6)	2.98	0.15	49.8	95.4
≤6	93.5 (90.8–95.6)	55.7 (53.2–58.2)	2.11	0.12	37.6	96.8
≤7	96.0 (93.7–97.6)	38.7 (36.3–41.1)	1.56	0.10	30.8	97.1
≤8	98.7 (97.1–99.5)	22.6 (20.6–24.8)	1.28	0.06	26.6	98.3
≤9	99.6 (98.4–99.9)	9.7 (8.3–11.3)	1.10	0.05	23.9	98.7
≤10	100.0 (99.2–100.0)	0.0 (0.0–0.2)	1.00	–	22.2	77.8
SPMSQ
≥0	100.0 (99.2–100.0)	0.0 (0.0–0.2)	1.00	–	22.1	77.9
≥1	98.0 (96.2–99.1)	21.8 (19.8–23.9)	1.25	0.09	26.2	97.4
≥2	93.9 (91.2–95.9)	43.8 (41.4–46.4)	1.67	0.14	32.2	96.2
≥3	86.8 (83.3–89.8)	62.8 (60.3–65.2)	2.33	0.21	39.8	94.4
≥4	78.2 (74.0–82.0)	75.8 (73.5–77.9)	3.22	0.29	47.8	92.4
≥5	69.6 (65.0–73.8)	84.9 (83.0–86.6)	4.59	0.36	56.6	90.8
≥6	60.9 (56.2–65.5)	91.2 (89.6–92.5)	6.90	0.43	66.2	89.2
≥7	51.1 (46.4–55.9)	94.7 (93.5–95.8)	9.67	0.52	73.3	87.2
≥8	43.9 (39.2–48.6)	97.1 (96.1–97.9)	15.12	0.58	81.1	85.9
≥9	35.5 (31.0–40.1)	98.1 (97.3–98.7)	18.33	0.66	83.9	84.3
=10	22.1 (18.3–26.2)	99.1 (98.5–99.5)	24.42	0.79	87.4	81.8

Bolded values indicate the best trade‐off between sensitivity and specificity, based on the Youden index (sensitivity + specificity − 100).

Abbreviation: 10‐CS, 10‐point Cognitive Screener (higher score = better cognitive performance); SPMSQ, Short Portable Mental Status Questionnaire (higher score = worse cognitive performance); 95% CI, 95% confidence interval; LR, likelihood ratio; PPV, positive predictive value; NPV, negative predictive value.

Of the 2003 participants, 853 screened positive using a 10‐CS score ≤5 and 548 using an SPMSQ score ≥5, with 461 classified as impaired by both tools, 392 by the 10‐CS only and 87 by the SPMSQ only (Figure 3). Overall agreement between the tools was 76.1%, with a Cohen's kappa of 0.49 (95% CI: 0.45–0.53; p < 0.001).

FIGURE 3.

Overlap between the 10‐point Cognitive Screener (10‐CS) and the Short Portable Mental Status Questionnaire (SPMSQ) in detecting dementia Venn diagram showing overlap between 10‐CS ≤5 (probable cognitive impairment) and SPMSQ ≥5 (moderate cognitive impairment); reference: informant‐based Clinical Dementia Rating ≥1. Agreement: 76.1%; Cohen's κ = 0.49 (95% confidence interval [CI]: 0.45–0.53; p < 0.001).

Patients classified as cognitively impaired in both the 10‐CS (≤5) and the SPMSQ (≥5) were more likely to be older, women, have fewer years of education, not live with a partner, be frail, be functionally dependent before admission, and present with higher acute illness severity scores (Table S3). Among participants who screened positive, a prior dementia or MCI diagnosis was documented in 306/853 (35.9%) using the 10‐CS and in 247 of 548 (45.1%) using the SPMSQ (Table S3).

3.3. Association of cognitive impairment with adverse outcomes

Over a median (IQR) hospital stay of 8 (5–16) days, 403 patients (20.1%) developed incident delirium, and 1103 (55.1%) experienced hospital‐associated ADL disability. Incidence rates stratified by cognitive impairment status are shown in Table 3. Prolonged hospital stay, defined as ≥14 days (upper tertile), did not differ significantly between patients with and without cognitive impairment. During the 90‐day follow‐up, 375 patients (18.7%) died; 195 (52.0%) of them died during hospitalization. Kaplan–Meier curves for 90‐day mortality differed between participants with and without cognitive impairment as classified by either the 10‐CS and the SPMSQ, with log‐rank p <  0.001 for both tools (Figure 4). After adjustment for sociodemographic, clinical, and hospital‐level factors (Model 3), a 10‐CS score ≤5 and an SPMSQ score ≥5 were each associated with a higher risk of 90‐day mortality (Table 3). In addition, they had nearly a threefold higher odds of developing incident delirium during hospitalization. Cognitive impairment by both tools was associated with hospital‐associated ADL disability; neither was associated with prolonged hospital stay (Table 3). The use of cognitive screening scores as continuous measures showed similar results (Table S4). When cognitive impairment was defined based on prior diagnosis of dementia or MCI, it was associated with incident delirium during hospitalization but not with other adverse outcomes (Table S5).

TABLE 3.

Association between preexisting cognitive impairment based on brief screening tools and adverse health outcomes (n = 2003).

		Odds ratio or hazard ratio (95% confidence interval)
	N (%)	Model 1: Base model	Model 2: base model + sociodemographic factors	Model 3: fully adjusted model
In‐hospital incident delirium
10‐CS categories
>5	127 (11.0)	(Reference)	(Reference)	(Reference)
≤5	276 (32.4)	3.85 (3.01–4.93)	3.61 (2.81–4.64)	2.91 (2.24–3.78)
SPMSQ categories
<5	203 (14.0)	(Reference)	(Reference)	(Reference)
≥5	198 (36.6)	3.56 (2.77–4.57)	3.60 (2.75–4.72)	2.92 (2.20‐3.87)
Hospital‐associated ADL disability
10‐CS categories
>5	568 (49.6)	(Reference)	(Reference)	(Reference)
≤5	535 (71.2)	2.48 (2.01–3.05)	2.31 (1.87–2.85)	1.86 (1.49–2.33)
SPMSQ categories
<5	773 (53.6)	(Reference)	(Reference)	(Reference)
≥5	322 (73.0)	2.13 (1.65–2.73)	1.79 (1.37–2.33)	1.43 (1.09–1.90)
Prolonged hospital stay
10‐CS categories
>5	336 (29.2)	(Reference)	(Reference)	(Reference)
≤5	277 (32.5)	1.14 (0.93–1.40)	1.19 (0.97–1.47)	1.02 (0.82–1.27)
SPMSQ categories
<5	435 (30.0)	(Reference)	(Reference)	(Reference)
≥5	175 (32.3)	1.12 (0.89–1.41)	1.18 (0.92–1.52)	1.03 (0.80‐1.33)
90‐day mortality
10‐CS categories
>5	178 (18.8)	(Reference)	(Reference)	(Reference)
≤5	197 (27.6)	1.72 (1.40–2.11)	1.66 (1.34–2.05)	1.27 (1.02–1.58)
SPMSQ categories
<5	243 (19.9)	(Reference)	(Reference)	(Reference)
≥5	129 (29.2)	1.76 (1.41–2.21)	1.74 (1.36–2.22)	1.40 (1.09–1.80)

Abbreviations: 10‐CS, 10‐point Cognitive Screener (0‐10; lower score = worse performance), with scores ≤5 indicating probable cognitive impairment; SPMSQ, Short Portable Mental Status Questionnaire (0‐10; higher score = worse performance), with scores ≥5 indicating moderate cognitive impairment; ADL, activities of daily living (eating, transferring, toileting, dressing, bathing, continence); CAM, Confusion Assessment Method.

Estimates were computed from mixed‐effects logistic regression for in‐hospital adverse outcomes (reported as odds ratios) and Weibull proportional hazards models for 90‐day mortality (reported as hazard ratios). In‐hospital outcomes were assessed in the final 24 h of hospitalization. Delirium in the hospital was identified by CAM‐trained physicians; ADL disability was defined as new dependence versus 4 weeks pre‐admission, excluding 107 (5.3%) fully dependent patients. Prolonged stay was defined as ≥14 days (upper tertile).

Model 1: Base model with random intercepts to account for clustering, specified at the state/province level and the level of each participating hospital (study center).

Model 2: Base model + sociodemographic factors (age, sex, race/ethnicity, education, marital status).

Model 3: Model 2 + Charlson Comorbidity Index, baseline frailty status (FRAIL Scale), acute illness severity measured by the National Early Warning Score 2 (NEWS‐2), and hospital type (public vs private).

FIGURE 4.

Kaplan–Meier curves for 90‐day mortality by categories of brief cognitive screening tools (n = 2003): (A) 10‐point Cognitive Screener (10‐CS); (B) Short Portable Mental Status Questionnaire (SPMSQ). Cognitive categories were defined using validated cutoffs: 10‐CS ≤5 (probable cognitive impairment) and SPMSQ ≥5 (moderate cognitive impairment). Log‐rank p < 0.001 for both tools.

When added to a base model of sociodemographic and clinical factors, the 10‐CS increased the AUC for incident delirium from 0.75 to 0.79 (p < 0.001); results for the SPMSQ were similar (Table 4). Incremental AUC values for hospital‐associated ADL disability, prolonged hospital stay, and 90‐day mortality are also shown in Table 4.

TABLE 4.

Impact of adding brief cognitive screening tools on the discrimination of prediction models for adverse outcomes.

Outcomes	AUC (95% confidence interval)	p‐value a
In‐hospital incident delirium
Model 1	0.751 (0.724–0.777)	(Reference)
Model 1 + 10‐CS	0.794 (0.769–0.819)	<0.001
Model 1 + SPMSQ	0.786 (0.760–0.811)	<0.001
Hospital‐associated ADL disability
Model 1	0.728 (0.706–0.751)	(Reference)
Model 1 + 10‐CS	0.741 (0.718–0.763)	0.006
Model 1 + SPMSQ	0.739 (0.716–0.761)	0.01
Prolonged hospital stay
Model 1	0.708 (0.684–0.732)	(Reference)
Model 1 + 10‐CS	0.710 (0.685–0.734)	0.31
Model 1 + SPMSQ	0.710 (0.686–0.734)	0.23
90‐day mortality
Model 1	0.725 (0.697–0.751)	(Reference)
Model 1 + 10‐CS	0.745 (0.718–0.771)	0.001
Model 1 + SPMSQ	0.751 (0.724–0.777)	<0.001

Abbreviation: ROC, receiver‐operating characteristic; 10‐CS, 10‐point Cognitive Screener (0‐10); SPMSQ, Short Portable Mental Status Questionnaire (0‐10).

Model 1: Multivariable predictive model using mixed‐effects logistic regression with random intercepts to account for clustering, specified at the state/province level and at the level of each participating hospital (study center), adjusted for sociodemographic variables (age, sex, race/ethnicity, education, marital status), Charlson Comorbidity Index, baseline frailty (FRAIL Scale), acute illness severity (NEWS‐2), and hospital type (public vs private).

^a p‐values refer to comparisons between areas under the ROC curves of the models with and without each cognitive screening tool.

Sensitivity analyses restricted to patients who survived hospitalization did not alter the associations between brief cognitive screening categories and in‐hospital adverse outcomes (Table S6). When considering only patients who were fully independent in ADLs before admission, patients with cognitive impairment (10‐CS ≤ 5) had a nearly threefold higher odds of hospital‐associated ADL disability compared to those without impairment (adjusted odds ratio [OR] = 3.04, 95% CI: 2.24–4.13). The corresponding OR for the SPMSQ (≥5) was 2.24 (95% CI: 1.49–3.35).

4. DISCUSSION

In this large, international cohort study conducted across 43 hospitals in five countries, we found that the 10‐CS and SPMSQ demonstrated good diagnostic accuracy in screening for dementia risk among older inpatients without delirium. The 10‐CS showed slightly higher accuracy than the SPMSQ in this admixed geriatric population, drawn predominantly from a middle‐income country, with a median of 5 years of education. Of note, more than one‐third of dementia cases identified using the IB‐CDR had no prior diagnosis of dementia or MCI, reflecting a significant gap in diagnostic recognition. The 10‐CS and SPMSQ predicted 90‐day mortality, even after adjustment for sociodemographic characteristics, comorbidity burden, frailty, and illness severity. Screening scores on both tools were also associated with in‐hospital adverse outcomes, including incident delirium and hospital‐associated ADL disability. These findings provide compelling evidence that brief cognitive screening tools—requiring minimal time and no specialized equipment—can identify underlying cognitive vulnerability and support prognostic stratification in older adults hospitalized without delirium.

Brief cognitive screening tools, such as the MMSE, MoCA, and Mini‐Cog, showed good to excellent diagnostic performance for dementia in outpatient settings (AUCs 0.85–0.92). ¹⁷ , ³⁷ However, their use in hospitals is limited by acute illness, pain, time constraints, and a scarcity of trained personnel. Informant‐based tools may offer an alternative, but they are impractical when reliable informants are not present. ⁷ , ⁸ , ⁹ , ¹⁰ Most hospital‐based studies have relied on small samples in high‐income settings and rarely account for how education or cross‐country context may influence test performance. ⁹ , ¹⁰ Our multicenter study addresses these gaps by enrolling a diverse hospital population, primarily composed of older patients with fewer years of formal education in lower‐ and middle‐income countries. The 10‐CS and SPMSQ achieved good diagnostic accuracy for identifying patients likely to have underlying dementia, using the IB‐CDR as reference, outperforming previously reported or documented diagnoses of dementia or MCI. The lower diagnostic accuracy of prior documentation likely reflects poor linkage between outpatient and hospital records and limited patient or family awareness or willingness to report cognitive concerns. ⁷ , ³⁸ In our cohort, the optimal operating thresholds based on the Youden index (10‐CS ≤5; SPMSQ ≥5 errors) were concordant with the pre‐specified “probable impairment” and “moderate impairment” categories, respectively. ¹⁴ , ¹⁵ At these thresholds, both tools demonstrated good diagnostic performance relative to the IB‐CDR reference. Their performance remained consistent across countries and educational strata, with the 10‐CS showing slightly higher accuracy (AUC 0.87 vs 0.85). Unlike the MMSE and Mini‐Cog, these tools do not rely on motor or visual components, making them easy to administer at the bedside and scalable for dementia risk identification in resource‐limited hospital environments. ⁴ , ¹² , ¹⁶

Even in the absence of a formal diagnosis, cognitive impairment increases the risk of poor outcomes, particularly when accompanied by delirium. ³⁹ , ⁴⁰ Delirium frequently coexists with dementia and serves as a powerful clinical marker of cognitive vulnerability in acutely ill older adults. ¹¹ , ⁴¹ , ⁴² However, delirium alone does not capture the full spectrum of baseline cognitive impairment. ⁴ Many hospitalized patients with undiagnosed dementia or cognitive impairment are admitted without delirium and may still be at elevated risk. Understanding whether preexisting cognitive impairment independently predicts adverse outcomes, particularly in the absence of delirium, is essential for improving risk stratification. ⁴ In our multicenter study involving a large, admixed cohort of hospitalized older adults without delirium, we demonstrated that preexisting cognitive impairment conveyed prognostic information beyond sociodemographic and clinical measures. Patients identified as cognitively impaired by the 10‐CS or SPMSQ had significantly higher risks of developing in‐hospital delirium, hospital‐associated ADL disability, and 90‐day mortality.

Some cognitive screening scores have been associated with adverse outcomes in hospitalized older adults; however, most existing evidence is limited to single‐center studies in high‐income countries. ¹⁰ , ³⁹ Preexisting cognitive impairment detected by the MMSE or Mini‐Cog at hospital admission has been linked to an increased risk of incident delirium. ⁴³ , ⁴⁴ In our multicenter study, cognitive impairment identified by the 10‐CS and SPMSQ was associated with a nearly threefold higher odds of developing delirium during hospitalization. Previous research has also shown that lower MMSE scores are associated with longer hospital stays and higher mortality rates. ⁴⁵ In contrast, the Mini‐Cog did not retain a significant association after adjustment for comorbidities. ⁴⁶ In our fully adjusted models—including sociodemographic characteristics, comorbidity burden, frailty, and acute illness severity—the 10‐CS and SPMSQ scores were associated with 90‐day mortality and hospital‐associated ADL disability, an important but often under‐recognized patient‐centered outcome that signifies loss of independence and is strongly linked to increased caregiver burden, greater need for post‐acute services, and institutionalization. ⁴⁷ Prior studies, including single‐center analyses in emergency departments, have reported associations between lower 10‐CS scores and longer hospitalizations. ⁴ , ¹⁶ We found no association with prolonged length of stay, possibly reflecting variability in discharge practices, staffing, and access to post‐acute care across the 43 hospitals. ⁴⁸ Across multiple studies, hospital length of stay is driven primarily by admitting diagnosis, comorbidity burden, and acute illness severity. ⁴⁹ , ⁵⁰

Early detection of cognitive impairment during hospitalization offers a valuable opportunity to recognize a problem that often goes unnoticed and to optimize care, even in resource‐constrained settings. ⁸ The 10‐CS is especially suitable in settings with limited time, personnel, and training, as it takes ≈2 min to administer, requires no writing or collateral history, and performs well among patients with low educational attainment. ¹⁶ With high sensitivity (88.4%) and moderate specificity (70.5%) at the optimal cutoff (≤5), it can help clinicians flag patients at a higher likelihood of underlying dementia who warrant further assessment. ⁵¹ Its high negative predictive value (95.4%) further supports its use to exclude dementia in screen‐negative patients. The 10‐CS may be preferred as a first‐line bedside screener in time‐constrained or lower‐education settings. The SPMSQ, which takes about 5 min to administer, may be a reasonable alternative when time allows, particularly among patients with higher educational attainment. Because concordance was only moderate, the tools should not be considered interchangeable. In practice, such cognitive screeners should be administered early during hospitalization after delirium has been excluded or resolved. Integrating brief, free, and easy‐to‐administer tools into hospital workflows may facilitate identification of patients who would benefit from geriatric assessment, reduce potentially inappropriate medication use, support anticipatory care planning, and improve communication and discharge processes. ⁴ , ¹⁶ The identification of cognitive impairment may also promote caregiver engagement and facilitate post‐discharge follow‐up. ¹² Positive screens should be followed by a comprehensive diagnostic assessment, ideally after discharge. ¹⁷

This study presents several important strengths. We conducted a large, multicenter investigation across 43 hospitals in Brazil and four additional countries, encompassing a diverse geriatric population that is often underrepresented in hospital‐based dementia research. Our analyses evaluated the diagnostic and prognostic properties of two brief cognitive screening tools that can be feasibly applied at the bedside, even in busy acute care settings. Finally, we found that reliance on routine clinical documentation—a common approach in hospital‐based dementia studies—yielded inferior diagnostic and prognostic performance compared to the brief cognitive tools. ⁸

This study also has several limitations. Although the IB‐CDR is a reference tool for assessing cognitive impairment and its severity, it is not a clinical gold standard for diagnosing dementia. ⁵² We were also unable to validate the brief cognitive tools against a comprehensive cognitive assessment performed after discharge in a clinically stable context, which would have been ideal but challenging in this study context. It is also possible that administering the 10‐CS before the SPMSQ may have introduced order effects, such as practice or sustained attention, which we did not evaluate. A fixed administration order may have affected performance on the second instrument through fatigue, reduced engagement, or interference from prior items, thereby attenuating SPMSQ performance. In addition, trained investigators administered all cognitive measures and reviewed medical records; therefore, observer bias cannot be excluded. Although we adjusted for key sociodemographic and clinical covariates and accounted for clustering using random intercepts at the region and hospital levels, residual confounding remains possible due to unmeasured factors, including dementia subtype and severity, sensory impairment, language and literacy, medical diagnoses, medication exposures, and site‐level differences in clinical assessment practices and resources. Moreover, we cannot extrapolate our findings to older patients treated in rural settings or to those admitted with delirium. Because acute fluctuations may obscure preexisting deficits, direct cognitive testing is not feasible in the presence of delirium. ⁴ For this reason, we excluded individuals admitted with delirium. Still, delirium itself warrants follow‐up cognitive evaluation after discharge and is independently associated with poor outcomes. ¹¹

In conclusion, brief bedside cognitive screening using the 10‐CS and SPMSQ showed good accuracy for identifying patients likely to have underlying dementia and predicting adverse outcomes in a large, admixed cohort of hospitalized older adults. Their brevity, simplicity, and consistent performance across settings support their integration into routine hospital workflows, including in resource‐limited environments. Early identification of cognitive vulnerability may prompt comprehensive geriatric assessment, support anticipatory care planning, facilitate safer care transitions, and drive further diagnostic evaluation. Future work should incorporate patient‐ and caregiver‐reported outcomes (e.g., quality of life and satisfaction with care) and determine whether integrating such tools into hospital workflows improves outcomes for older inpatients. Doing so is an urgent priority as health care systems worldwide confront the challenges of population aging.

AUTHOR CONTRIBUTIONS

Márlon Juliano Romero Aliberti conceived and designed the study, acquired data, performed analyses, interpreted results, and drafted the manuscript. Thiago Junqueira Avelino‐Silva contributed to study conception and design, data acquisition, interpretation, and drafting of the manuscript. Kenneth E. Covinsky, Laiss Bertola, Regina Miksian Magaldi, Esther S. Oh, Christopher R. Carpenter, Maria Aparecida Camargos Bicalho, and Mônica Sanches Yassuda contributed to study conception and design, data interpretation, and manuscript preparation. Bruna Macêdo de Carvalho, Maria Eduarda Pires Lins e Silva Lima, Pérola Quintans de Almeida, Victor José Dornelas Melo, Tatiana de Carvalho Espíndola Pinheiro, Raiza Tavares Lira, and Ênio Simas Macedo acquired data, contributed to interpretation, and assisted with manuscript preparation. Claudia Kimie Suemoto conceived and designed the study, supervised the project, contributed to data interpretation, and revised the manuscript for important intellectual content. Funding acquisition was undertaken by Márlon Juliano Romero Aliberti through a research scholarship grant, with supervision from Claudia Kimie Suemoto and support from Kenneth E. Covinsky. All authors reviewed and approved the final version of the manuscript.

CONFLICT OF INTEREST STATEMENT

The authors declare no conflicts of interest. Author disclosures are available in the supporting information. Author disclosures are available in the Supporting information.

CONSENT STATEMENT

Informed consent was obtained from all participants prior to enrollment in the CHANGE Study across 43 hospitals in five countries.

Supporting information

Supporting Information

Supporting Information

Supporting Information