Advertisement

Association of Body Mass Index With All-Cause Mortality in Acutely Hospitalized Older Patients

Open AccessPublished:August 10, 2021DOI:https://doi.org/10.1016/j.jamda.2021.07.015

      Abstract

      Objectives

      The aim was to examine the relationship between body mass index (BMI) and mortality in older hospitalized patients taking activities of daily living (ADLs) into account.

      Design

      Retrospective cohort study.

      Setting and Participants

      Nationwide population-based study of all patients aged ≥65 years admitted to Danish geriatric medical departments during 2005 to 2014 and included in the National Danish Geriatric Database.

      Methods

      Patients were followed until death, emigration, or study termination (December 31, 2015). Primary outcome was all-cause mortality. BMI and ADLs were routinely assessed on admission and linked at an individual level to the Danish national health registers. Kaplan-Meier analysis was used to estimate crude survival according to each BMI subcategory and Cox regression to examine the association with mortality adjusting for age, comorbidity, polypharmacy, ADLs, marital status, prior hospitalizations, and admission year.

      Results

      In total, 74,589 patients (63% women) were included aged [mean (SD)] 82.5 (7.5) years with BMI [mean (SD)] of 23.9 (5.1) kg/m2. During follow-up 51,188 died. Follow-up time was 191,972 person-years. Unadjusted and adjusted hazard ratio (HR) for overall, 30-day, and 1-year mortality decreased significantly with increasing BMI. In women, the highest adjusted HR (95% confidence interval) for overall mortality was seen for underweight patients (BMI <16) 1.83 (1.72–1.95) and the lowest for obesity grade II patients (BMI = 35.0–39.9) 0.66 (0.60–0.73) when using normal weight (BMI = 18.5–24.9) as reference. In men, the HR for BMI <16 and BMI = 35.0–39.9 were 1.98 (1.76–2.23) and 0.56 (0.49–0.65), respectively.

      Conclusions and Implications

      In hospitalized older patients, association between mortality and BMI did not show a U-shaped or J-shaped curve after adjustment of multiple confounders, including ADLs. Instead, mortality was highest in patients with low BMI and decreased with increasing BMI before leveling off in the obese range. Our study highlights the need for a debate and reassessment of what should be the ideal BMI in this vulnerable patient group.

      Keywords

      The prevalence of obesity is high among adults,
      • Flegal K.M.
      • Kit B.K.
      • Orpana H.
      • Graubard B.I.
      Association of all-cause mortality with overweight and obesity using standard body mass index categories: A systematic review and meta-analysis.
      including those in the older age group.
      Centers for Disease Control and Prevention
      Prevalence of obesity among adults and youth: United States, 2015–2016.
      ,
      • Villareal D.T.
      • Apovian C.M.
      • Kushner R.F.
      • Klein S.
      Obesity in older adults: Technical review and position statement of the American Society for Nutrition and NAASO, The Obesity Society.
      Although decreasing with age, still 15% are considered obese at age 70 years and older.
      • Villareal D.T.
      • Apovian C.M.
      • Kushner R.F.
      • Klein S.
      Obesity in older adults: Technical review and position statement of the American Society for Nutrition and NAASO, The Obesity Society.
      Previous studies on the association between body mass index (BMI) and mortality show conflicting results depending on the studied population. In the general population, high BMI is associated with disability, additional costs, and both disease-specific mortality like cardiovascular and cancer, and all-cause mortality.
      • Bhaskaran K.
      • Dos-Santos-Silva I.
      • Leon D.A.
      • et al.
      Association of BMI with overall and cause-specific mortality: A population-based cohort study of 3.6 million adults in the UK.
      • Berrington de Gonzalez A.
      • Hartge P.
      • Cerhan J.R.
      • et al.
      Body-mass index and mortality among 1.46 million white adults.
      • Whitlock G.
      • Lewington S.
      • Sherliker P.
      • et al.
      Body-mass index and cause-specific mortality in 900 000 adults: Collaborative analyses of 57 prospective studies.
      Most studies report a U-shaped association when assessing the general population.
      • Flegal K.M.
      • Kit B.K.
      • Orpana H.
      • Graubard B.I.
      Association of all-cause mortality with overweight and obesity using standard body mass index categories: A systematic review and meta-analysis.
      ,
      • Berrington de Gonzalez A.
      • Hartge P.
      • Cerhan J.R.
      • et al.
      Body-mass index and mortality among 1.46 million white adults.
      • Whitlock G.
      • Lewington S.
      • Sherliker P.
      • et al.
      Body-mass index and cause-specific mortality in 900 000 adults: Collaborative analyses of 57 prospective studies.
      • Calle E.E.
      • Thun M.J.
      • Petrelli J.M.
      • et al.
      Body-mass index and mortality in a prospective cohort of U.S. adults.
      • Di Angelantonio E.
      • Bhupathiraju Sh N.
      • et al.
      Global BMI Mortality Collaboration
      Body-mass index and all-cause mortality: Individual-participant-data meta-analysis of 239 prospective studies in four continents.
      Other studies addressing specific diseases such as type 2 diabetes
      • Zaccardi F.
      • Dhalwani N.N.
      • Papamargaritis D.
      • et al.
      Nonlinear association of BMI with all-cause and cardiovascular mortality in type 2 diabetes mellitus: A systematic review and meta-analysis of 414,587 participants in prospective studies.
      show a J-shaped or even an inverse association with increasing BMI after disease onset of heart failure,
      • Curtis J.P.
      • Selter J.G.
      • Wang Y.
      • et al.
      The obesity paradox: Body mass index and outcomes in patients with heart failure.
      acute myocardial infarction,
      • Bucholz E.M.
      • Rathore S.S.
      • Reid K.J.
      • et al.
      Body mass index and mortality in acute myocardial infarction patients.
      pneumonia,
      • Nie W.
      • Zhang Y.
      • Jee S.H.
      • et al.
      Obesity survival paradox in pneumonia: A meta-analysis.
      and hip fracture.
      • Modig K.
      • Erdefelt A.
      • Mellner C.
      • et al.
      "Obesity paradox" holds true for patients with hip fracture: A registry-based cohort study.
      Among oldest-old individuals, high BMI has been reported to be associated with lower mortality,
      • Winter J.E.
      • MacInnis R.J.
      • Nowson C.A.
      The influence of age the BMI and all-cause mortality association: A meta-analysis.
      • Veronese N.
      • Cereda E.
      • Solmi M.
      • et al.
      Inverse relationship between body mass index and mortality in older nursing home residents: A meta-analysis of 19,538 elderly subjects.
      • Yamazaki K.
      • Suzuki E.
      • Yorifuji T.
      • et al.
      Is there an obesity paradox in the Japanese elderly population? A community-based cohort study of 13 280 men and women.
      • Thinggaard M.
      • Jacobsen R.
      • Jeune B.
      • et al.
      Is the relationship between BMI and mortality increasingly U-shaped with advancing age? A 10-year follow-up of persons aged 70–95 years.
      • Lang P.O.
      • Mahmoudi R.
      • Novella J.L.
      • et al.
      Is obesity a marker of robustness in vulnerable hospitalized aged populations? Prospective, multicenter cohort study of 1 306 acutely ill patients.
      • Landi F.
      • Onder G.
      • Gambassi G.
      • et al.
      Body mass index and mortality among hospitalized patients.
      maybe because it requires good health to maintain a high BMI at advanced ages.
      • Lv Y.B.
      • Yuan J.Q.
      • Mao C.
      • et al.
      Association of body mass index with disability in activities of daily living among Chinese adults 80 years of age or older.
      Although many risk factors are associated with increased mortality, activities of daily living (ADLs) are the strongest predictors of survival, and hence the best indicators of general health among oldest-old individuals.
      • Ryg J.
      • Engberg H.
      • Mariadas P.
      • et al.
      Barthel Index at hospital admission is associated with mortality in geriatric patients: a Danish nationwide population-based cohort study.
      ,
      • Thinggaard M.
      • McGue M.
      • Jeune B.
      • et al.
      Survival prognosis in very old adults.
      However, studies documenting the positive association between BMI and survival did not take general health into account
      • Winter J.E.
      • MacInnis R.J.
      • Nowson C.A.
      The influence of age the BMI and all-cause mortality association: A meta-analysis.
      • Veronese N.
      • Cereda E.
      • Solmi M.
      • et al.
      Inverse relationship between body mass index and mortality in older nursing home residents: A meta-analysis of 19,538 elderly subjects.
      • Yamazaki K.
      • Suzuki E.
      • Yorifuji T.
      • et al.
      Is there an obesity paradox in the Japanese elderly population? A community-based cohort study of 13 280 men and women.
      • Thinggaard M.
      • Jacobsen R.
      • Jeune B.
      • et al.
      Is the relationship between BMI and mortality increasingly U-shaped with advancing age? A 10-year follow-up of persons aged 70–95 years.
      or only included a few patients.
      • Lang P.O.
      • Mahmoudi R.
      • Novella J.L.
      • et al.
      Is obesity a marker of robustness in vulnerable hospitalized aged populations? Prospective, multicenter cohort study of 1 306 acutely ill patients.
      Furthermore, underweight and malnutrition also play a major role in the health and well-being of frail older adults and are therefore essential to address as well.
      • Volkert D.
      • Beck A.M.
      • Cederholm T.
      • et al.
      ESPEN guideline on clinical nutrition and hydration in geriatrics.
      Weight, height, and ADLs are routinely assessed when an older acutely ill patient is hospitalized and admitted to a geriatric department in Denmark. Therefore, the aim of the present study was to explore the association between BMI and survival in a nationwide population of hospitalized older patients after taking ADLs into account.

      Methods

      Design and Setting

      This is a nationwide population-based longitudinal cohort study. Each Danish citizen has a unique personal registry number, which enables accurate linkage of information at the individual level between the many population-based national registers.
      • Frank L.
      Epidemiology. When an entire country is a cohort.
      We used data from 4 different Danish national registers (Supplementary Table 1): the Danish National Database of Geriatrics,
      • Kannegaard P.N.
      • Vinding K.L.
      • Hare-Bruun H.
      National database of geriatrics.
      the Danish National Patient Register (all somatic inpatient contacts since 1977; psychiatric inpatient, emergency department, and outpatient specialty clinic contacts since 1995),
      • Schmidt M.
      • Schmidt S.A.
      • Sandegaard J.L.
      • et al.
      The Danish National Patient Registry: A review of content, data quality, and research potential.
      the Danish Civil Registration System (deaths, migration, and marital status with 100% completeness),
      • Schmidt M.
      • Pedersen L.
      • Sorensen H.T.
      The Danish Civil Registration System as a tool in epidemiology.
      and the Danish National Database of Reimbursed Prescriptions (all individually redeemed reimbursed prescriptions).
      • Johannesdottir S.A.
      • Horvath-Puho E.
      • Ehrenstein V.
      • et al.
      Existing data sources for clinical epidemiology: The Danish National Database of Reimbursed Prescriptions.
      Data were used to assess the association between BMI and overall, 30-day, and 1-year mortality.
      The method by which the cohort was established has previously been published and described in detail elsewhere.
      • Ryg J.
      • Engberg H.
      • Mariadas P.
      • et al.
      Barthel Index at hospital admission is associated with mortality in geriatric patients: a Danish nationwide population-based cohort study.
      In short, the study population was identified using data from the Danish National Database of Geriatrics,
      • Kannegaard P.N.
      • Vinding K.L.
      • Hare-Bruun H.
      National database of geriatrics.
      which is a clinical quality database established in 2005 designed to include all patients admitted to geriatric medical departments in Denmark. Patients were admitted directly from the general practitioner, through the emergency department, or by transfer from other hospital departments. The database has a patient completeness of 90%, reaching the standard requirement for Danish national databases.
      • Kannegaard P.N.
      • Vinding K.L.
      • Hare-Bruun H.
      National database of geriatrics.
      We included all patients aged ≥65 years with a first registration (index date) in the database from January 1, 2005 to December 31, 2014. Patients were followed from index date until time of death (outcome), emigration, or end of study (December 31, 2015). Exact date of death from all-cause mortality was retrieved from the Danish Civil Registration System.
      • Schmidt M.
      • Pedersen L.
      • Sorensen H.T.
      The Danish Civil Registration System as a tool in epidemiology.
      At the time of admission to the geriatric department, level of ADL dependency was routinely assessed by a geriatric nurse or nursing assistant using the Barthel-Index-100.
      • Mahoney F.I.
      • Barthel D.W.
      Functional evaluation: The Barthel Index.
      ,
      • Shah S.
      • Vanclay F.
      • Cooper B.
      Improving the sensitivity of the Barthel Index for stroke rehabilitation.
      This instrument is a sum score across 10 domains of ADLs, each scored on a weighted numerical scale with lowest score indicating total dependency and highest score indicating complete independence. Time for completion is approximately 5 minutes, and the index has high reliability.
      • Shah S.
      • Vanclay F.
      • Cooper B.
      Improving the sensitivity of the Barthel Index for stroke rehabilitation.
      ,
      • Katz P.
      Measures of adult general functional status: The Barthel Index, Katz Index of Activities of Daily Living, Health Assessment Questionnaire (HAQ), MACTAR Patient Preference Disability Questionnaire, and Modified Health Assessment Questionnaire (MHAQ).
      Furthermore, weight and height allowed calculation of BMI as the weight in kilograms divided by the height in meters squared (kg/m2). The BMI was divided into standard categories according to the World Health Organization: BMI <16 (severe underweight), BMI = 16.0–18.4 (underweight), BMI = 18.5–24.9 (normal weight), BMI = 25.0–29.9 (overweight), BMI = 30.0–34.9 (obesity grade 1), BMI = 35.0–39.9 (obesity grade 2), or BMI ≥40.0 (obesity grade 3).
      Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000;894:i–xii, 1–253.
      ,
      Physical status: the use and interpretation of anthropometry. Report of a WHO Expert Committee. World Health Organ Tech Rep Ser 1995;854:1–452.
      Data on comorbidity were collected based on International Classification of Diseases, 10th Revision, hospital discharge diagnoses extracted from the Danish National Patient Register,
      • Schmidt M.
      • Schmidt S.A.
      • Sandegaard J.L.
      • et al.
      The Danish National Patient Registry: A review of content, data quality, and research potential.
      allowing for calculation of Charlson comorbidity index (CCI). Because patients in our study were included from January 1, 2005, we calculated CCI 10 years before the index date to allow the same time scale for comorbidity calculation in all patients. Number of medications was defined as the number of different medications purchased up to 120 days before the index date using information from the Danish National Database of Reimbursed Prescriptions.
      • Johannesdottir S.A.
      • Horvath-Puho E.
      • Ehrenstein V.
      • et al.
      Existing data sources for clinical epidemiology: The Danish National Database of Reimbursed Prescriptions.
      Finally, the national registers were also used to assess number of prior hospitalizations and marital status. In general, confounders were chosen due to their known impact on burden of disease (prior hospitalizations) and mortality in older adults (ADLs,
      • Ryg J.
      • Engberg H.
      • Mariadas P.
      • et al.
      Barthel Index at hospital admission is associated with mortality in geriatric patients: a Danish nationwide population-based cohort study.
      polypharmacy,
      • Gnjidic D.
      • Hilmer S.N.
      • Blyth F.M.
      • et al.
      Polypharmacy cutoff and outcomes: Five or more medicines were used to identify community-dwelling older men at risk of different adverse outcomes.
      CCI,
      • Charlson M.E.
      • Pompei P.
      • Ales K.L.
      • MacKenzie C.R.
      A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation.
      and marital status
      • Manzoli L.
      • Villari P.
      • Pirone G.M.
      • Boccia A.
      Marital status and mortality in the elderly: A systematic review and meta-analysis.
      ). The present study is a post hoc analysis from a cohort study assessing the association between ADL and mortality.
      • Ryg J.
      • Engberg H.
      • Mariadas P.
      • et al.
      Barthel Index at hospital admission is associated with mortality in geriatric patients: a Danish nationwide population-based cohort study.

      Statistics

      Descriptive data are reported as means with corresponding standard deviation (SD) (normally distributed data) or as medians with corresponding interquartile range (IQR) (25%–75% percentile) (skewed data). Tests of differences in the categorical variables were performed using the χ2 test. Differences between groups in the numeric variables were tested using the Wilcoxon rank sum test (median differences) or the Student t-test (mean differences), as appropriate. Kaplan-Meier survival curves were used to estimate crude survival according to each of the pre-specified BMI subcategories. Spearman correlation (ρ) was used to test the association between BMI and Barthel Index, age, CCI, or number of prescribed medications. Cox regression was used to carry out univariable and multivariable analysis adjusting for age, marital status, Barthel Index, CCI, number of prescribed medications, previous hospital admission, and year of index admission. All variables were treated as categorical in the models. The statistical significance of the categorical variables included in the multivariable Cox regression model was tested using Wald statistics. The proportional hazard assumption was inspected graphically for the BMI variable using a Log-Log plot and was found to be satisfactory. In the fully adjusted model, analyses were conducted as complete case analysis excluding patients with missing data on one or more of the included variables. Imputation methods were not used.
      • Bennett D.A.
      How can I deal with missing data in my study?.
      An additional descriptive nonresponse analysis was carried out for missing data on BMI or Barthel Index to examine whether patients with missing versus not missing data differed on outcome variable. Furthermore, a robustness analysis was performed in which all patients with missing data on BMI were added to either category BMI <18.5 or BMI ≥30 to address potential implications. Analyses were stratified by sex at birth according to the health registers. All analyses were performed using the statistical software STATA (StataCorp, College Station, TX). A P value of .05 indicated statistical significance.

      Ethics

      Informed consent from patients was not needed according to Danish legislation on medical ethics, due to the design using register-based data only. The study was approved by the Danish data protection agency (J.nr. 16/23359). Data are reported according to STROBE (STrengthening the Reporting of OBservational studies in Epidemiology) guidelines.
      • von Elm E.
      • Altman D.G.
      • Egger M.
      • et al.
      Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies.

      Results

      The final study population consisted of 74,589 patients (63% women) (Table 1). The age [mean (SD)] and BMI [mean (SD)] of the total cohort were 82.5 (7.5) years and 23.9 (5.1) kg/m2, respectively, and the median follow-up time was 2 years. More men were obese (BMI ≥30) (9.1%) than underweight (BMI <18.5) (5.9%), whereas more women were underweight (11.6%) than obese (8.8%) (Table 1).
      Table 1Baseline Characteristics of the Study Population
      MeasuresTotal Cohort (n = 74,589)Women 63% (n = 46,815)Men 37% (n = 27,774)
      Age (y), mean (SD)82.5 (7.5)83.4 (7.3)80.9 (7.5)
      Barthel Index, median (IQR)
      ADLs were assessed using Barthel-Index-100.
      54 (29–77)55 (30–77)52 (26–77)
      CCI, median (IQR)
      The CCI was calculated based on hospital discharge diagnoses during 10 years before baseline.
      2 (1–3)2 (1–3)2 (1–4)
      Number of drugs, median (IQR)
      Redeemed prescriptions within 120 days before index date.
      6 (4–9)6 (4–9)6 (4–9)
      Prior hospitalization (1 year), median (IQR)
      Based on hospital admissions during 1 year before baseline. Normally distributed data are presented with mean (SD).
      0 (0–1)0 (0–1)1 (0–2)
      BMI (kg/m2), mean (SD)23.9 (5.1)23.6 (5.3)24.5 (4.7)
      <16, n (%)1774 (2.4)1449 (3.1)325 (1.2)
      16.0–18.4, n (%)5313 (7.1)3997 (8.5)1316 (4.7)
      18.5–24.9, n (%)30,111 (40.4)18,982 (40.5)11,129 (40.1)
      25.0–29.9, n (%)15,060 (20.2)8659 (18.5)6401 (23.0)
      30.0–34.9, n (%)4955 (6.6)2938 (6.3)2017 (7.3)
      35.0–39.9, n (%)1235 (1.7)835 (1.8)400 (1.4)
      ≥40, n (%)447 (0.6)324 (0.7)123 (0.4)
      Missing, n (%)15,694 (21.0)9631 (20.6)6063 (21.8)
      ADLs were assessed using Barthel-Index-100.
      The CCI was calculated based on hospital discharge diagnoses during 10 years before baseline.
      Redeemed prescriptions within 120 days before index date.
      § Based on hospital admissions during 1 year before baseline. Normally distributed data are presented with mean (SD).

      Crude Survival

      A total of 51,188 deaths occurred, with no patients lost to follow-up. Age [mean (SD)] at time of death was 87.2 (7.1) years in women and 84.2 (7.2) years in men. Among patients still hospitalized 3108 had died within 30 days. Total follow-up corresponded to 191,972 person-years. Survival (years) [median (95% confidence interval [CI])] was higher in women [3.1 (3.0–3.1)] compared with men [2.0 (2.0–2.1)]. Crude survival proportions decreased with decreasing BMI category for the total study population and both sexes, with men in the lowest BMI categories having the shortest median survival (Figure 1). No correlation was found between BMI and Barthel Index (ρ = 0.05), age (ρ = −0.18), CCI (ρ = 0.08), or numbers of prescribed medications (ρ = 0.13). Supplementary Table 2 displays baseline characteristics of study population with complete data used for the multivariable analysis (n = 56,564). Of these, 38,966 patients died during follow-up, with 3574 dying within 30 days and 16,488 within 1 year.
      Figure thumbnail gr1
      Fig. 1Kaplan-Meier crude survival curves for the total cohort (A) and stratified by sex [women (B), men (C)]. Illustrated for each of the 7 BMI subcategories (<16.0, 16.0–18.4, 18.5–24.9, 25.0–29.9, 30.0–34.5, 35.0–39.9, ≥40.0).

      Association Between BMI Category and Mortality

      Unadjusted hazard ratio (HR) (95% CI) for mortality decreased significantly with increasing BMI category (reference: normal weight (BMI = 18.5–24.9) HR = 1) in the total cohort and in women (n = 37,181) and men (n = 21,709) separately (Table 2). In the multivariable model, adjusting for Barthel Index and other covariables, the relationship persisted with the strongest association seen for 30-day mortality (Table 2). Obesity (BMI ≥30) was associated with an adjusted HR (95% CI) for overall mortality of 0.71 (0.64–0.78) in women and 0.64 (0.55–0.75) in men. The significant association was present in all 3 obesity grades for overall, 30-day, and 1-year mortality (Table 2). Furthermore, patients with BMI ≥30 also had significantly lower mortality when compared with patients with BMI = 25.0–29.9 (data not shown).
      Table 2Overall, 30-Day, and 1-Year Mortality Risk (HR and 95% CI) According to BMI in Total Cohort and by Sex Using the Subcategory “BMI = 18.5–24.9 (Normal Weight)” as the Reference Category
      TotalWomenMen
      HR (95% CI)HR (95% CI)HR (95% CI)
      UnivariableMultivariable
      Adjusted for age, marital status, Barthel Index, CCI, number of different medications purchased in the 120 days before index date, number of hospital admissions during 1 year before baseline, and period of index admission.
      UnivariableMultivariable
      Adjusted for age, marital status, Barthel Index, CCI, number of different medications purchased in the 120 days before index date, number of hospital admissions during 1 year before baseline, and period of index admission.
      UnivariableMultivariable
      Adjusted for age, marital status, Barthel Index, CCI, number of different medications purchased in the 120 days before index date, number of hospital admissions during 1 year before baseline, and period of index admission.
      Overall mortality
       BMI (kg/m2)
      <16.01.67 (1.59–1.76)1.71 (1.62–1.81)1.82 (1.71–1.93)1.83 (1.72–1.95)1.81 (1.61–2.03)1.98 (1.76–2.23)
      16.0–18.41.37 (1.32–1.41)1.37 (1.33–1.42)1.40 (1.35–1.46)1.41 (1.35–1.46)1.60 (1.50–1.70)1.61 (1.51–1.72)
      18.5–24.91 (reference)1 (reference)1 (reference)1 (reference)1 (reference)1 (reference)
      25.0–29.90.77 (0.75–0.79)0.79 (0.77–0.81)0.77 (0.75–0.80)0.78 (0.75–0.81)0.73 (0.70–0.75)0.76 (0.73–0.79)
      30.0–34.90.67 (0.65–0.70)0.71 (0.68–0.74)0.71 (0.67–0.74)0.74 (0.70–0.78)0.60 (0.56–0.64)0.65 (0.61–0.70)
      35.0–39.90.60 (0.55–0.65)0.61 (0.57–0.67)0.66 (0.60–0.72)0.66 (0.60–0.73)0.52 (0.45–0.60)0.56 (0.49–0.65)
      ≥40.00.63 (0.56–0.72)0.68 (0.59–0.77)0.66 (0.56–0.77)0.72 (0.61–0.84)0.64 (0.51–0.81)0.71 (0.56–0.90)
      30-day mortality
       BMI (kg/m2)
      <16.02.15 (1.88–2.45)2.11 (1.84–2.42)2.35 (2.00–2.75)2.15 (1.82–2.54)2.55 (2.00–3.25)2.73 (2.13–3.50)
      16.0–18.41.49 (1.36–1.64)1.46 (1.32–1.61)1.50 (1.33–1.70)1.45 (1.27–1.65)1.79 (1.54–2.08)1.70 (1.45–1.98)
      18.5–24.91 (reference)1 (reference)1 (reference)1 (reference)1 (reference)1 (reference)
      25.0–29.90.65 (0.60–0.71)0.67 (0.62–0.74)0.67 (0.59–0.76)0.69 (0.60–0.78)0.60 (0.53–0.68)0.64 (0.57–0.72)
      30.0–34.90.52 (0.45–0.61)0.55 (0.47–0.64)0.57 (0.46–0.71)0.58 (0.47–0.72)0.46 (0.37–0.58)0.52 (0.41–0.65)
      35.0–39.90.42 (0.30–0.59)0.36 (0.26–0.51)0.57 (0.38–0.83)0.49 (0.33–0.74)0.27 (0.14–0.50)0.22 (0.11–0.43)
      ≥40.00.52 (0.32–0.85)0.44 (0.27–0.74)0.73 (0.42–1.26)0.60 (0.34–1.06)0.26 (0.08–0.81)0.25 (0.08–0.79)
      1-y mortality
       BMI (kg/m2)
      <16.01.88 (1.75–2.01)1.90 (1.77–2.04)2.09 (1.92–2.26)2.03 (1.86–2.20)2.07 (1.79–2.38)2.21 (1.91–2.56)
      16.0–18.41.45 (1.39–1.52)1.44 (1.37–1.51)1.48 (1.39–1.57)1.44 (1.36–1.53)1.75 (1.62–1.89)1.72 (1.59–1.87)
      18.5–24.91 (reference)1 (reference)1 (reference)1 (reference)1 (reference)1 (reference)
      25.0–29.90.72 (0.69–0.75)0.73 (0.70–0.76)0.73 (0.69–0.77)0.74 (0.70–0.78)0.66 (0.63–0.70)0.69 (0.65–0.73)
      30.0–34.90.61 (0.57–0.65)0.64 (0.60–0.68)0.69 (0.63–0.75)0.70 (0.64–0.77)0.51 (0.46–0.57)0.57 (0.51–0.63)
      35.0–39.90.49 (0.42–0.56)0.46 (0.40–0.53)0.55 (0.46–0.66)0.50 (0.42–0.60)0.42 (0.33–0.53)0.43 (0.34–0.54)
      ≥40.00.61 (0.50–0.75)0.60 (0.49–0.74)0.66 (0.51–0.85)0.63 (0.48–0.82)0.60 (0.42–0.85)0.63 (0.44–0.90)
      Women: n = 37,181 (univariable), n = 35,814 (multivariable); Men: n = 21,709 (univariable), n = 20,750 (multivariable).
      Adjusted for age, marital status, Barthel Index, CCI, number of different medications purchased in the 120 days before index date, number of hospital admissions during 1 year before baseline, and period of index admission.

      Association Between BMI as a Continuous Variable and Mortality

      Graphical presentation of BMI as a continuous variable plotted against risk of overall, 30-day, and 1-year mortality in the fully adjusted model (using BMI = 25 as reference) is shown in Figure 2. Low BMI was associated with the highest HR in both short-term and long-term mortality, whereas increasing BMI was associated with decreasing mortality. The most optimal BMI (nadir) for overall mortality was observed for BMI = 37.8 in the total cohort, BMI = 35.2 in men, and BMI = 37.0 in women. For 30-day mortality, nadir was BMI = 37.3 in the total cohort, BMI = 45.0 in men, and BMI = 36.5 in women. And for 1-year mortality, nadir was BMI = 38.5 in the total cohort, BMI = 35.7 in men, and BMI = 45.0 in women. Dividing the cohort into 5-year age subcategories (65–69, 70–74, 75–79, 80–84, 85–89, 90–94, ≥95) did not change the association between BMI and mortality (data not shown).
      Figure thumbnail gr2
      Fig. 2BMI as continuous variable plotted against risk of mortality using BMI = 25 as reference. Broken lines indicate 95% CIs. Illustrated for each sex separately for overall mortality (A), 30-day mortality (B), and 1-year mortality (C).
      The fully adjusted model with BMI as a continuous variable revealed a significantly decreased overall mortality risk [HR (95% CI)] of 6% (6%–7%) in the total cohort for every 1 kg/m2 increase in BMI up to BMI = 24.9 and 2% (2%–3%) for BMI range 25.0–34.9 (Table 3). The corresponding figures for 30-day mortality were 9% (7%–10%) and 5% (2%–7%) and for 1-year mortality 8% (7%–8%) and 3% (2%–4%). Similar associations were seen when looking at data stratified for sex (Table 3). No change was seen for BMI ≥35 in any outcome (Table 3).
      Table 3Overall, 30-Day, and 1-Year Mortality Risk (HRs and 95% CIs) for Every 1 kg/m2 Increase in BMI in the Fully Adjusted Model
      Adjusted for age, marital status, Barthel Index, CCI, number of different medications purchased in the 120 days before index date, number of hospital admissions during 1 year before baseline, and period of index admission.
      With BMI as a Continuous Variable
      TotalWomenMen
      HR (95% CI)HR (95% CI)HR (95% CI)
      Overall mortality
       BMI (kg/m2)
      <25.00.94 (0.93–0.94)0.93 (0.92–0.93)0.92 (0.91–0.92)
      25.0–34.90.98 (0.97–0.98)0.99 (0.98–1.00)0.96 (0.95–0.98)
      ≥35.01.01 (0.99–1.03)1.01 (0.99–1.03)1.03 (1.00–1.06)
      30-day mortality
       BMI (kg/m2)
      <25.00.91 (0.90–0.93)0.91 (0.89–0.93)0.89 (0.87–0.91)
      25.0–34.90.95 (0.93–0.98)0.97 (0.93–1.01)0.95 (0.91–0.99)
      ≥35.01.03 (0.97–1.09)1.03 (0.97–1.10)1.02 (0.88–1.17)
      1-y mortality
       BMI (kg/m2)
      <25.00.92 (0.92–0.93)0.92 (0.91–0.93)0.90 (0.89–0.91)
      25.0–34.90.97 (0.96–0.98)0.99 (0.97–1.01)0.96 (0.94–0.97)
      ≥35.01.02 (0.99–1.04)1.02 (0.98–1.05)1.04 (0.99–1.08)
      Women: n = 35,814; Men: n = 20,750.
      Adjusted for age, marital status, Barthel Index, CCI, number of different medications purchased in the 120 days before index date, number of hospital admissions during 1 year before baseline, and period of index admission.

      Robustness Analysis

      In a robustness analysis, BMI was divided into the following 4 subcategories: <18.5, 18.5–24.9, 25.0–29.9, and ≥30. Adding all patients with missing data on BMI to either subcategory <18.5 or ≥30 did not alter the significant association between BMI and overall mortality when using BMI 18.5–24.9 as reference (data not shown).

      Discussion

      In this nationwide population-based longitudinal cohort study, we found BMI independently associated with mortality in hospitalized geriatric patients even when adjusting for relevant confounders including ADL. The mortality decreased steeply with increasing BMI while leveling off at obesity levels. This finding is in contrast to what would be expected in the general population where a U-shaped association is a common finding.
      ADLs are known to be very strong predictors of survival in older patients,
      • Ryg J.
      • Engberg H.
      • Mariadas P.
      • et al.
      Barthel Index at hospital admission is associated with mortality in geriatric patients: a Danish nationwide population-based cohort study.
      ,
      • Thinggaard M.
      • McGue M.
      • Jeune B.
      • et al.
      Survival prognosis in very old adults.
      but only a few studies have taken this into account when assessing association between BMI and mortality. One study assessing 1306 older hospitalized patients (mean age 85 years), reported decreased 1- and 2-year mortality with higher BMI when adjusting for ADL and comorbidity, but not on short-term mortality, possibly due to lack of power.
      • Lang P.O.
      • Mahmoudi R.
      • Novella J.L.
      • et al.
      Is obesity a marker of robustness in vulnerable hospitalized aged populations? Prospective, multicenter cohort study of 1 306 acutely ill patients.
      An older study reporting data on 18,316 hospitalized patients [mean age 71 years (women) and 70 years (men)] found a significant J-shaped curve association between BMI and mortality in the adjusted model including ADL and morbidity.
      • Landi F.
      • Onder G.
      • Gambassi G.
      • et al.
      Body mass index and mortality among hospitalized patients.
      In the present study, combining data from reliable Danish registers on an individual level with objective measurements allowed adjustment of relevant confounders including ADL. The fully adjusted model only changed the association marginally on both short- and long-term follow-up (Tables 2 and 3) indicating that BMI is an independent prognostic indicator on top of ADL in frail oldest-old individuals.
      Optimal BMI according to mortality risk has been increasingly and heavily debated in the past decades. In contrast to previous studies, we found no U- or J-shaped curve,
      • Flegal K.M.
      • Kit B.K.
      • Orpana H.
      • Graubard B.I.
      Association of all-cause mortality with overweight and obesity using standard body mass index categories: A systematic review and meta-analysis.
      ,
      • Berrington de Gonzalez A.
      • Hartge P.
      • Cerhan J.R.
      • et al.
      Body-mass index and mortality among 1.46 million white adults.
      • Whitlock G.
      • Lewington S.
      • Sherliker P.
      • et al.
      Body-mass index and cause-specific mortality in 900 000 adults: Collaborative analyses of 57 prospective studies.
      • Calle E.E.
      • Thun M.J.
      • Petrelli J.M.
      • et al.
      Body-mass index and mortality in a prospective cohort of U.S. adults.
      • Di Angelantonio E.
      • Bhupathiraju Sh N.
      • et al.
      Global BMI Mortality Collaboration
      Body-mass index and all-cause mortality: Individual-participant-data meta-analysis of 239 prospective studies in four continents.
      but a continuously inverse relationship before leveling off when a BMI of 35.0–39.9 was reached. This optimal BMI is higher than previously reported. A meta-analysis of 19,538 old nursing home residents reported same relationship as seen in the present study, but all residents with overweight were pooled in one category (BMI ≥30).
      • Veronese N.
      • Cereda E.
      • Solmi M.
      • et al.
      Inverse relationship between body mass index and mortality in older nursing home residents: A meta-analysis of 19,538 elderly subjects.
      Another meta-analysis on 64,076 community-dwelling older adults ≥65 years reported a J-shaped curve with an optimal BMI of 27.5–30.0 (extracted from their graphical presentation).
      • Winter J.E.
      • MacInnis R.J.
      • Nowson C.A.
      The influence of age the BMI and all-cause mortality association: A meta-analysis.
      Several prior studies report an optimal BMI of 20–25 in U-shaped curves,
      • Berrington de Gonzalez A.
      • Hartge P.
      • Cerhan J.R.
      • et al.
      Body-mass index and mortality among 1.46 million white adults.
      • Whitlock G.
      • Lewington S.
      • Sherliker P.
      • et al.
      Body-mass index and cause-specific mortality in 900 000 adults: Collaborative analyses of 57 prospective studies.
      • Calle E.E.
      • Thun M.J.
      • Petrelli J.M.
      • et al.
      Body-mass index and mortality in a prospective cohort of U.S. adults.
      • Di Angelantonio E.
      • Bhupathiraju Sh N.
      • et al.
      Global BMI Mortality Collaboration
      Body-mass index and all-cause mortality: Individual-participant-data meta-analysis of 239 prospective studies in four continents.
      whereas a meta-analysis on 2.88 million individuals found an optimal nadir between 25 and 30 in adults ≥65 years.
      • Flegal K.M.
      • Kit B.K.
      • Orpana H.
      • Graubard B.I.
      Association of all-cause mortality with overweight and obesity using standard body mass index categories: A systematic review and meta-analysis.
      Some argue against the higher optimal BMI found in the meta-analysis since it used normal weight as reference in the analysis and the span from 18.5–24.9 may be too wide.
      • Flegal K.M.
      • Kit B.K.
      • Orpana H.
      • Graubard B.I.
      Association of all-cause mortality with overweight and obesity using standard body mass index categories: A systematic review and meta-analysis.
      However, this cannot explain our findings because we plotted BMI against mortality HR using BMI 25 as reference following a prior study showing increased mortality already at BMI levels of 25.0 to 27.5.
      • Di Angelantonio E.
      • Bhupathiraju Sh N.
      • et al.
      Global BMI Mortality Collaboration
      Body-mass index and all-cause mortality: Individual-participant-data meta-analysis of 239 prospective studies in four continents.
      Others have shown that using self-reported weight and height are less predictive compared to studies using measured BMI as in our study.
      • Flegal K.M.
      • Kit B.K.
      • Orpana H.
      • Graubard B.I.
      Association of all-cause mortality with overweight and obesity using standard body mass index categories: A systematic review and meta-analysis.
      Combining this with no patients lost to follow-up in our nationwide study allows wide generalizability of our results in geriatric patients.
      According to the World Health Organization, the goal for individual adults is to maintain BMI in the range of 18.5–24.9 to stay healthy irrespective of age because comorbidity and mortality increase with increasing degrees of overweight.
      World Health Organization. Global Health Observatory
      Overweight: Issues and trends.
      Our study pinpoints the paradox that while BMI increases the risk of several diseases having a BMI in the obese range seems to increase the chance of survival in older acutely ill patients. This does not advocate for weight increase in general. Instead, it highlights that in contrast to younger age groups, higher BMI is beneficial in older people, which adds to a more recent article showing that high BMI is associated with a lower risk of developing ADL disability in oldest-old individuals.
      • Lv Y.B.
      • Yuan J.Q.
      • Mao C.
      • et al.
      Association of body mass index with disability in activities of daily living among Chinese adults 80 years of age or older.
      Other studies have not reported the same high optimal BMI as we find. Although studies addressing mortality after disease onset (heart failure,
      • Curtis J.P.
      • Selter J.G.
      • Wang Y.
      • et al.
      The obesity paradox: Body mass index and outcomes in patients with heart failure.
      acute myocardial infarction,
      • Bucholz E.M.
      • Rathore S.S.
      • Reid K.J.
      • et al.
      Body mass index and mortality in acute myocardial infarction patients.
      pneumonia,
      • Nie W.
      • Zhang Y.
      • Jee S.H.
      • et al.
      Obesity survival paradox in pneumonia: A meta-analysis.
      and hip fracture
      • Modig K.
      • Erdefelt A.
      • Mellner C.
      • et al.
      "Obesity paradox" holds true for patients with hip fracture: A registry-based cohort study.
      ) have observed optimal BMI in the same range as we found, studies looking at mortality in the background population show different results. This may partly be because geriatric patients are admitted to hospital with many different acute illnesses combined with a common phenotype of multimorbidity, polypharmacy, and functional decline. Therefore, our data should be interpreted with caution and may not be generalizable outside a geriatric cohort.
      The optimal BMI differs between groups of patients-at-risk. In geriatric patients, weight reduction is not the main focus.
      • Volkert D.
      • Beck A.M.
      • Cederholm T.
      • et al.
      ESPEN guideline on clinical nutrition and hydration in geriatrics.
      Our study is yet another argument for having a strong focus on nutrition in older people to avoid weight loss, and for this BMI adds useful information to the clinician when discussing personalized care plans. Furthermore, our study shows that acutely ill older people do not fit into official health guidelines of an optimal BMI for the general population. Our data suggest that the presence of an energy reserve is beneficial during acute illness in the oldest patients. Because the population of older adults is increasing.
      • Christensen K.
      • Doblhammer G.
      • Rau R.
      • Vaupel J.W.
      Ageing populations: The challenges ahead.
      this demographic change highlights the need for official recommendations on optimal BMI in this vulnerable patient group.

      Limitations

      This study has important limitations. First, some of the increased risk seen in patients with low BMI may be partly facilitated by the effect of smoking as shown by others.
      • Berrington de Gonzalez A.
      • Hartge P.
      • Cerhan J.R.
      • et al.
      Body-mass index and mortality among 1.46 million white adults.
      ,
      • Calle E.E.
      • Thun M.J.
      • Petrelli J.M.
      • et al.
      Body-mass index and mortality in a prospective cohort of U.S. adults.
      Unfortunately, we had no data on smoking to address this. Second, no data on body shape or distribution of fat and muscle exist in the database. We can therefore neither address speculations of sarcopenia nor sarco-obesity. Third, the acutely admitted older patients with overweight might be those with genetic or other advantages allowing them to achieve higher ages despite overweight. According to Statistics Denmark,
      News from Statistics Denmark No 72, February 17, 2011 [in Danish: Nyt fra Danmarks Statestik].
      age at time of death in Denmark was 80.2 years in women and 77.1 years in men during the study period, compared with 87.2 years and 84.2 years in our study population. Our data on lower mortality seen in patients with high BMI might partly be explained by this survival bias. Fourth, we had 21% missing data on BMI. However, our analyses adding all patients with missing data on BMI to either subcategory <18.5 or ≥30.0 lowered the effect size of association between BMI and mortality, but not the significance association. Fifth, preexisting disease is a potential confounder. Disease status affects both exposure and outcome because disease often causes weight loss, but also increases mortality risk. Therefore, low BMI may be a result of underlying disease. We did adjust for known diseases in terms of CCI, but may have missed any underlying undiagnosed disease. If such an undiagnosed condition would have an impact on weight loss of major importance, it should be possible to detect differences in short-term mortality outcome. However, the association between mortality and BMI was the same when looking at short-term mortality (30 days), mid-term mortality (1 year), or long-term mortality (overall survival allowing up to 11 years of follow-up). Sixth, nonstandardization of weight measurement may have led to risk of misclassification of BMI subcategories in some patients. Finally, not all patients aged ≥65 years are hospitalized in a geriatric department. Therefore, our data should be interpreted with caution and may not be generalizable to all patients within this age group.

      Conclusions and Implications

      Association between BMI and mortality in hospitalized older patients did not show a U-shaped or a J-shaped curve when adjusting for relevant confounders including ADLs. Instead, mortality decreased continuously with increasing BMI before leveling off in the obese range. Our study highlights the need for a debate and reassessment of what should be the ideal BMI in this vulnerable patient group.

      Supplementary Data

      Supplementary Table 1Description of Data Sources
      Data SourcesDescription
      The Danish National Database of Geriatrics
      • Kannegaard P.N.
      • Vinding K.L.
      • Hare-Bruun H.
      National database of geriatrics.
      A clinical quality database established in 2005. Designed to include all patients admitted to 1 of the 24 geriatric departments in Denmark. Patients are admitted directly from the general practitioner, through the emergency department, or by transfer from other hospital departments. Contains information on a number of variables collected at the time of hospital admission (ie, height, weight, and assessment of Barthel Index).
      The Danish National Patient Register
      • Schmidt M.
      • Schmidt S.A.
      • Sandegaard J.L.
      • et al.
      The Danish National Patient Registry: A review of content, data quality, and research potential.
      Established in 1977. Contains individual-level information on all hospital admissions, discharge diagnoses, and dates of admission and discharge.
      The Danish Civil Registration System
      • Schmidt M.
      • Pedersen L.
      • Sorensen H.T.
      The Danish Civil Registration System as a tool in epidemiology.
      Each Danish citizen and residents on immigration have since 1968 been assigned a unique 10-digit civil personal registry number. This enables accurate linkage of information at the individual level between the many population-based national registers. Contains data on deaths, migration, and marital status.
      The Danish National Database of Reimbursed Prescriptions
      • Johannesdottir S.A.
      • Horvath-Puho E.
      • Ehrenstein V.
      • et al.
      Existing data sources for clinical epidemiology: The Danish National Database of Reimbursed Prescriptions.
      Contains information on individually redeemed reimbursed prescriptions from all pharmacies in Denmark since 2004.
      Supplementary Table 2Baseline Characteristics of Study Population With Complete Data Used for the Multivariable Analysis
      MeasuresTotal Cohort (n = 56,564)Women 63% (n = 35,814)Men 37% (n = 20,750)
      Age (y), mean (SD)82.4 (7.4)83.3 (7.3)80.8 (7.4)
      Barthel Index, median (IQR)
      ADLs were assessed using Barthel-Index-100.
      56 (31–78)57 (33–78)54 (29–78)
      CCI, median (IQR)
      The CCI was calculated based on hospital discharge diagnoses during 10 years before baseline.
      2 (1–3)2 (1–3)2 (1–4)
      Number of drugs, median (IQR)
      Redeemed prescriptions within 120 days before index date.
      6 (4–9)6 (4–9)6 (4–9)
      Prior hospitalization (1 y), median (IQR)
      Based on hospital admissions during 1 year before baseline. Normal distributed data are presented with mean (SD).
      0 (0–1)0 (0–1)1 (0–2)
      BMI (kg/m2), mean (SD)24.0 (5.1)23.6 (5.3)24.5 (4.6)
      <16.0, n (%)1669 (3.0)1364 (3.8)305 (1.5)
      16.0–18.4, n (%)5061 (8.9)3825 (10.7)1236 (6.0)
      18.5–24.9, n (%)28,872 (51.0)18,257 (51.0)10,615 (51.2)
      25.0–29.9, n (%)14,538 (25.7)8386 (23.4)6152 (29.6)
      30.0–34.9, n (%)4800 (8.5)2865 (8.0)1935 (9.3)
      35.0–39.9, n (%)1191 (2.1)802 (2.2)389 (1.9)
      ≥40.0, n (%)433 (0.8)315 (0.9)118 (0.6)
      ADLs were assessed using Barthel-Index-100.
      The CCI was calculated based on hospital discharge diagnoses during 10 years before baseline.
      Redeemed prescriptions within 120 days before index date.
      § Based on hospital admissions during 1 year before baseline. Normal distributed data are presented with mean (SD).

      References

        • Flegal K.M.
        • Kit B.K.
        • Orpana H.
        • Graubard B.I.
        Association of all-cause mortality with overweight and obesity using standard body mass index categories: A systematic review and meta-analysis.
        JAMA. 2013; 309: 71-82
        • Centers for Disease Control and Prevention
        Prevalence of obesity among adults and youth: United States, 2015–2016.
        (Available at:)
        • Villareal D.T.
        • Apovian C.M.
        • Kushner R.F.
        • Klein S.
        Obesity in older adults: Technical review and position statement of the American Society for Nutrition and NAASO, The Obesity Society.
        Am J Clin Nutr. 2005; 82: 923-934
        • Bhaskaran K.
        • Dos-Santos-Silva I.
        • Leon D.A.
        • et al.
        Association of BMI with overall and cause-specific mortality: A population-based cohort study of 3.6 million adults in the UK.
        Lancet Diabetes Endocrinol. 2018; 6: 944-953
        • Berrington de Gonzalez A.
        • Hartge P.
        • Cerhan J.R.
        • et al.
        Body-mass index and mortality among 1.46 million white adults.
        N Engl J Med. 2010; 363: 2211-2219
        • Whitlock G.
        • Lewington S.
        • Sherliker P.
        • et al.
        Body-mass index and cause-specific mortality in 900 000 adults: Collaborative analyses of 57 prospective studies.
        Lancet. 2009; 373: 1083-1096
        • Calle E.E.
        • Thun M.J.
        • Petrelli J.M.
        • et al.
        Body-mass index and mortality in a prospective cohort of U.S. adults.
        N Engl J Med. 1999; 341: 1097-1105
        • Di Angelantonio E.
        • Bhupathiraju Sh N.
        • et al.
        • Global BMI Mortality Collaboration
        Body-mass index and all-cause mortality: Individual-participant-data meta-analysis of 239 prospective studies in four continents.
        Lancet. 2016; 388: 776-786
        • Zaccardi F.
        • Dhalwani N.N.
        • Papamargaritis D.
        • et al.
        Nonlinear association of BMI with all-cause and cardiovascular mortality in type 2 diabetes mellitus: A systematic review and meta-analysis of 414,587 participants in prospective studies.
        Diabetologia. 2017; 60: 240-248
        • Curtis J.P.
        • Selter J.G.
        • Wang Y.
        • et al.
        The obesity paradox: Body mass index and outcomes in patients with heart failure.
        Arch Intern Med. 2005; 165: 55-61
        • Bucholz E.M.
        • Rathore S.S.
        • Reid K.J.
        • et al.
        Body mass index and mortality in acute myocardial infarction patients.
        Am J Med. 2012; 125: 796-803
        • Nie W.
        • Zhang Y.
        • Jee S.H.
        • et al.
        Obesity survival paradox in pneumonia: A meta-analysis.
        BMC Med. 2014; 12: 61
        • Modig K.
        • Erdefelt A.
        • Mellner C.
        • et al.
        "Obesity paradox" holds true for patients with hip fracture: A registry-based cohort study.
        J Bone Joint Surg Am. 2019; 101: 888-895
        • Winter J.E.
        • MacInnis R.J.
        • Nowson C.A.
        The influence of age the BMI and all-cause mortality association: A meta-analysis.
        J Nutr Health Aging. 2017; 21: 1254-1258
        • Veronese N.
        • Cereda E.
        • Solmi M.
        • et al.
        Inverse relationship between body mass index and mortality in older nursing home residents: A meta-analysis of 19,538 elderly subjects.
        Obes Rev. 2015; 16: 1001-1015
        • Yamazaki K.
        • Suzuki E.
        • Yorifuji T.
        • et al.
        Is there an obesity paradox in the Japanese elderly population? A community-based cohort study of 13 280 men and women.
        Geriatr Gerontol Int. 2017; 17: 1257-1264
        • Thinggaard M.
        • Jacobsen R.
        • Jeune B.
        • et al.
        Is the relationship between BMI and mortality increasingly U-shaped with advancing age? A 10-year follow-up of persons aged 70–95 years.
        J Gerontol A Biol Sci Med Sci. 2010; 65: 526-531
        • Lang P.O.
        • Mahmoudi R.
        • Novella J.L.
        • et al.
        Is obesity a marker of robustness in vulnerable hospitalized aged populations? Prospective, multicenter cohort study of 1 306 acutely ill patients.
        J Nutr Health Aging. 2014; 18: 66-74
        • Landi F.
        • Onder G.
        • Gambassi G.
        • et al.
        Body mass index and mortality among hospitalized patients.
        Arch Intern Med. 2000; 160: 2641-2644
        • Lv Y.B.
        • Yuan J.Q.
        • Mao C.
        • et al.
        Association of body mass index with disability in activities of daily living among Chinese adults 80 years of age or older.
        JAMA Netw Open. 2018; 1: e181915
        • Ryg J.
        • Engberg H.
        • Mariadas P.
        • et al.
        Barthel Index at hospital admission is associated with mortality in geriatric patients: a Danish nationwide population-based cohort study.
        Clin Epidemiol. 2018; 10: 1789-1800
        • Thinggaard M.
        • McGue M.
        • Jeune B.
        • et al.
        Survival prognosis in very old adults.
        J Am Geriatr Soc. 2016; 64: 81-88
        • Volkert D.
        • Beck A.M.
        • Cederholm T.
        • et al.
        ESPEN guideline on clinical nutrition and hydration in geriatrics.
        Clin Nutr. 2019; 38: 10-47
        • Frank L.
        Epidemiology. When an entire country is a cohort.
        Science. 2000; 287: 2398-2399
        • Kannegaard P.N.
        • Vinding K.L.
        • Hare-Bruun H.
        National database of geriatrics.
        Clin Epidemiol. 2016; 8: 731-735
        • Schmidt M.
        • Schmidt S.A.
        • Sandegaard J.L.
        • et al.
        The Danish National Patient Registry: A review of content, data quality, and research potential.
        Clin Epidemiol. 2015; 7: 449-490
        • Schmidt M.
        • Pedersen L.
        • Sorensen H.T.
        The Danish Civil Registration System as a tool in epidemiology.
        Eur J Epidemiol. 2014; 29: 541-549
        • Johannesdottir S.A.
        • Horvath-Puho E.
        • Ehrenstein V.
        • et al.
        Existing data sources for clinical epidemiology: The Danish National Database of Reimbursed Prescriptions.
        Clin Epidemiol. 2012; 4: 303-313
        • Mahoney F.I.
        • Barthel D.W.
        Functional evaluation: The Barthel Index.
        Md State Med J. 1965; 14: 61-65
        • Shah S.
        • Vanclay F.
        • Cooper B.
        Improving the sensitivity of the Barthel Index for stroke rehabilitation.
        J Clin Epidemiol. 1989; 42: 703-709
        • Katz P.
        Measures of adult general functional status: The Barthel Index, Katz Index of Activities of Daily Living, Health Assessment Questionnaire (HAQ), MACTAR Patient Preference Disability Questionnaire, and Modified Health Assessment Questionnaire (MHAQ).
        Arthritis Care Res. 2003; 49: S15-S27
      1. Obesity: preventing and managing the global epidemic. Report of a WHO consultation. World Health Organ Tech Rep Ser 2000;894:i–xii, 1–253.
        (Available at:)
      2. Physical status: the use and interpretation of anthropometry. Report of a WHO Expert Committee. World Health Organ Tech Rep Ser 1995;854:1–452.
        (Available at:)
        • Gnjidic D.
        • Hilmer S.N.
        • Blyth F.M.
        • et al.
        Polypharmacy cutoff and outcomes: Five or more medicines were used to identify community-dwelling older men at risk of different adverse outcomes.
        J Clin Epidemiol. 2012; 65: 989-995
        • Charlson M.E.
        • Pompei P.
        • Ales K.L.
        • MacKenzie C.R.
        A new method of classifying prognostic comorbidity in longitudinal studies: Development and validation.
        J Chronic Dis. 1987; 40: 373-383
        • Manzoli L.
        • Villari P.
        • Pirone G.M.
        • Boccia A.
        Marital status and mortality in the elderly: A systematic review and meta-analysis.
        Soc Sci Med. 2007; 64: 77-94
        • Bennett D.A.
        How can I deal with missing data in my study?.
        Aust N Z J Public Health. 2001; 25: 464-469
        • von Elm E.
        • Altman D.G.
        • Egger M.
        • et al.
        Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: Guidelines for reporting observational studies.
        BMJ. 2007; 335: 806-808
        • World Health Organization. Global Health Observatory
        Overweight: Issues and trends.
        • Christensen K.
        • Doblhammer G.
        • Rau R.
        • Vaupel J.W.
        Ageing populations: The challenges ahead.
        Lancet. 2009; 374: 1196-1208
      3. News from Statistics Denmark No 72, February 17, 2011 [in Danish: Nyt fra Danmarks Statestik].
        (Available at:)