Authors: Je Min Suh, Laurence Weinberg, Nattaya Raykateeraroj, Andrew Hardidge, Andrew Ng, Tony Huang, Yao Hui Lim, Sophia Grobler, Jad Jurji, Jae Hyun Lee, David Pilcher, Dong-Kyu Lee
Categories: Research, Nonagenarians, Anaemia, ICU, Knee arthroplasty, Mortality, Complications
Source: Perioperative Medicine
Authors: Je Min Suh, Laurence Weinberg, Nattaya Raykateeraroj, Andrew Hardidge, Andrew Ng, Tony Huang, Yao Hui Lim, Sophia Grobler, Jad Jurji, Jae Hyun Lee, David Pilcher, Dong-Kyu Lee
Anaemia is one of the most common complications following major surgery, arising from perioperative blood loss, hemodilution, and impaired erythropoiesis. Although preoperative anaemia is a recognised risk factor for adverse outcomes, the prognostic importance of early postoperative anaemia remains underexplored, particularly in very elderly patients. Given the rising prevalence of osteoarthritis and total knee arthroplasty (TKA) in the ≥ 90-year-old population, understanding the implications of anaemia in this high-risk cohort is increasingly relevant.
We conducted a retrospective binational cohort study using the Australian and New Zealand Intensive Care Society (ANZICS) Adult Patient Database. All patients ≥ 90 years admitted to the ICU following TKA between 2010 and 2024 with haemoglobin (Hb) measured within 24 h were included. Anaemia was defined according to World Health Organization criteria and stratified as non-anaemia, mild, or moderate/severe. The primary outcome was all-cause mortality, examined in short-term (< 365 days) and long-term (≥ 365 days) intervals. Secondary outcomes included ICU and hospital length of stay (LOS), illness severity scores, and postoperative biochemical and physiological parameters. Time-to-event outcomes were assessed with Cox regression, and LOS outcomes with log-transformed linear models.
Among 493 patients, 22.3% were non-anaemic, 30.0% had mild anaemia, and 47.7% had moderate/severe anaemia. Severe anaemia was associated with higher illness severity (median APACHE III 57 vs 54; p = 0.013), prolonged ICU stay (18% longer; p = 0.008), and longer hospitalisation (p = 0.037). Anaemia was not associated with short-term mortality. However, beyond 365 days, both mild (HR 2.08; 95% CI 1.05–4.13; p = 0.036) and severe anaemia (HR 1.98; 95% CI 1.05–3.73; p = 0.035) independently predicted higher mortality.
In nonagenarians undergoing TKA, anaemia at ICU admission was highly prevalent and associated with increased illness severity, resource utilisation, and elevated long-term mortality. These findings highlight postoperative Hb as an important prognostic marker in the oldest old and support strategies for early detection, optimisation, and structured follow-up to improve survival and recovery in this vulnerable population.
The online version contains supplementary material available at 10.1186/s13741-026-00655-8.
Anaemia is one of the most common complications following major surgery, most often resulting from perioperative blood loss, hemodilution, and impaired erythropoiesis. The World Health Organization (WHO) defines anaemia as haemoglobin (Hb) < 130 g/L in men or < 120 g/L in women, and severe anaemia as < 110 g/L regardless of sex (World Health Organization 2011). Although preoperative anaemia is well recognised as an independent predictor of adverse surgical outcomes, (Musallam et al. 2011; Fowler et al. 2015; Klein et al. 2016), the prognostic importance of early postoperative anaemia remains underexplored. Emerging evidence suggests that low postoperative Hb is associated with elevated all-cause mortality, risk of myocardial infarction, functional decline, and impaired recovery (Muñoz et al. 2017; Baron et al. 2014; Meybohm et al. 2017; Wu et al. 2007).
The clinical importance of this issue is magnified by demographic shifts. By 2080, more than 2.2 billion people worldwide will be ≥ 65 years of age, and by 2030, an estimated 31 million people will be ≥ 90 years of age (United Nations Department of Economic and Social Affairs, Population Division 2022; Christensen et al. 2009). Osteoarthritis, which is responsible for 7.1% of the global musculoskeletal disease burden (Safiri et al. 2020), is highly prevalent in this cohort. Total knee arthroplasty (TKA), the definitive surgical treatment for osteoarthritis, is among the most frequently performed procedures in high-income countries (more than 700,000 annually in the United States) (Sloan et al. 2018; Australian Orthopaedic Association National Joint Replacement Registry (AOANJRR) 2022).
Despite rising case volumes, little is known regarding the prognostic effects of postoperative anaemia in the oldest populations, particularly nonagenarians undergoing TKA. This group has diminished physiologic reserve, multiple comorbidities, and high susceptibility to perioperative complications (Seib et al. 2018; Makary et al. 2010). Understanding the relationship between postoperative anaemia and outcomes in this population is essential to guide transfusion thresholds and inform discussions regarding surgical risk. In this multicentre, binational retrospective cohort study, we evaluated the association between early postoperative anaemia and prognosis in nonagenarians after TKA. Our findings address a critical evidence gap and might inform perioperative strategies to increase survival, recovery, and quality of life in this uniquely vulnerable population.
Ethical approval for this study was obtained from the Alfred Hospital Human Research Ethics Committee (project No. 253/24). Given the study’s retrospective design and use of de-identified data, the requirement for individual informed consent was waived. Data analysis was initiated only after ethics approval was granted.
This retrospective cohort study was conducted in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines (Elm et al. 2007). De-identified patient-level data were obtained from the Australian and New Zealand Intensive Care Society (ANZICS) Adult Patient Database, a binational repository that captures detailed clinical and outcome data from more than 98% of intensive care units (ICUs) in Australia and 68% of ICUs in New Zealand (Secombe et al. 2023). Data collection and quality assurance are overseen by the ANZICS Centre for Outcome and Resource Evaluation.
We included all patients ≥ 90 years of age who were admitted to the ICU after TKA between 2010 and 2024. Eligible patients were identified according to primary diagnosis and operative codes from the Acute Physiology and Chronic Health Evaluation III (APACHE III)–J system. Only patients with Hb measurements available within the first 24 h of postoperative ICU admission and NDI linkage were included in the survival analysis.
Anaemia was classified according to WHO criteria and stratified into three (i) no anaemia (Hb > 130 g/L for men or > 120 g/L for women), (ii) mild anaemia (Hb = 111–130 g/L for men or 111–120 g/L for women), and (iii) moderate to severe anaemia (Hb ≤ 110 g/L for both sexes). The highest Hb value recorded within the first 24 h after ICU admission was used to determine anaemia status. As sensitivity analyses, mortality models were repeated using the lowest Hb value recorded within the same 24-h window, and early Hb variation was quantified as the absolute difference between the highest and lowest Hb values during this period.
Demographic variables included age, sex, ICU admission type (elective vs non-elective), and source of admission. Illness severity was assessed with the APACHE III and Sequential Organ Failure Assessment (SOFA) scores.
The primary outcome was all-cause mortality, assessed over short-term (< 365 days) and long-term (≥ 365 days) follow-up periods after ICU admission. Secondary outcomes included ICU and hospital length of stay (LOS), illness severity (APACHE III and SOFA scores), and early postoperative physiological and biochemical markers (e.g., Hb, albumin, urea, creatinine, bicarbonate, systolic blood pressure, and mean arterial pressure). Additional variables included comorbidities, surgical urgency, delirium, treatment limitation orders, and ICU admission source. Outcomes were examined according to anaemia severity at ICU admission, classified into normal, mild, or moderate/severe categories according to WHO Hb thresholds.
All statistical analyses were performed in R software, version 4.4.3 (R Foundation for Statistical Computing, Vienna, Austria). Patients were stratified into the following anaemia severity categories according to WHO Hb non-anaemia (≥ 130 g/L for men or ≥ 120 g/L for women), mild anaemia (110–129 g/L for men or 110–119 g/L for women), or severe anaemia (< 110 g/L for both sexes).
Baseline characteristics were compared across anaemia groups with descriptive statistics. The distribution of continuous variables was assessed with the Shapiro–Wilk test, histograms, and Q–Q plots. Because most continuous variables exhibited non-normal distributions, they were summarised as medians with interquartile ranges and compared with the Kruskal–Wallis test. Categorical variables are presented as counts and percentages, and comparisons were conducted with Pearson’s χ^2^ test or Fisher’s exact test, as appropriate.
Time-to-event outcomes were evaluated with Cox proportional hazards regression models implemented via the survival package. The proportional hazards assumption was assessed for each covariate with scaled Schoenfeld residuals. Because of observed global and covariate-specific violations, follow-up time was segmented into two predefined intervals (0–365 days and > 365 days), and separate Cox models were fitted for each interval. After stratification, the proportional hazards assumption was reassessed and confirmed for all covariates within each time interval. Hazard ratios (HRs) with 95% confidence intervals (CIs) are reported for both the unadjusted (univariate) and adjusted cohorts. Kaplan–Meier survival curves were constructed and compared with the log-rank test, and visualised with the survminer package.
Non-linear associations between haemoglobin variability, defined as the difference between highest and lowest haemoglobin values during admission and mortality were examined using restricted cubic spline functions within Cox proportional hazards models. Splines with four knots were used. Hazard ratios with 95% confidence intervals were estimated relative to ΔHb = 0.
LOS outcomes for ICU and hospital admissions were analysed with multivariable linear regression models with log-transformed LOS as the dependent variable to address right-skewness. Anaemia group served as the main exposure, and models were adjusted for clinically relevant covariates. Regression diagnostics were performed to verify assumptions of residual normality, multicollinearity, linearity, and homoscedasticity. Results are presented as exponentiated beta coefficients with 95% CIs, representing geometric mean ratios across anaemia groups. Analyses were conducted with the stats package in R. All statistical tests were two-tailed, and a p-value < 0.05 was considered statistically significant. Missing data were minimal (< 5% for all included variables) and were handled with complete-case analysis.
The participants’ baseline characteristics, stratified by anaemia severity at ICU admission, are presented in Table 1. Of the 493 patients included, 110 (22.3%) were classified as non-anaemic, 148 (30.0%) were classified as mildly anaemic, and 235 (47.7%) were classified as severely anaemic. The median age was similar across 91.5 years (interquartile range 90.7–92.7) in the non-anaemia group, 91.6 (90.8–92.9) in the mild anaemia group, and 92.0 (90.7–93.4) in the severe anaemia group (p = 0.386).Table 1Baseline demographics of nonagenarians undergoing arthroscopyVariableNon-anaemiaN = 110Mild anaemiaN = 148Severe anaemiaN = 235p-valueAge (years)91.5 (90.7, 92.7)91.6 (90.8, 92.9)92.0 (90.7, 93.4)0.386Hospital Stay Duration (hours)191.4 (145.0, 287.0)188.2 (137.5, 306.4)219.4 (147.5, 376.5)0.037ICU Stay Duration (hours)22.0 (19.8, 25.8)23.2 (19.9, 29.4)24.0 (19.5, 45.8)0.041Clinical Frailty Scale4 (3,6)4 (3,6)4 (3,5)0.022GCS (Score 3–15)15 (14,15)15 (14,15)15 (14,15)0.620APACHE III54.0 (47.0, 62.0)57.0 (49.0, 63.5)57.0 (51.0, 65.0)0.012SOFA2.0 (1.0, 3.0)2.0 (1.0, 3.0)2.0 (1.0, 4.0)0.344Survival (days)818.0 (12.9, 1,518.8)478.3 (12.0, 1,272.1)444.8 (14.0, 1,150.5)0.127Lactate (mmol/L)1.4 (1.0, 2.2)1.1 (0.9, 1.7)1.0 (0.8, 1.6)0.106Maximum Heart Rate (beats/min)85.0 (75.0, 95.0)85.0 (75.0, 99.0)85.0 (75.0, 96.0)0.942Minimum Heart Rate (beats/min)60.0 (54.0, 70.0)60.0 (52.0, 65.0)60.0 (55.0, 70.0)0.623Maximum Respiratory Rate (breaths/min)22.0 (20.0, 24.0)22.0 (20.0, 24.0)22.0 (20.0, 25.0)0.933Minimum Respiratory Rate (breaths/min)12.0 (11.0, 14.0)12.0 (10.0, 14.0)12.0 (10.0, 14.0)0.473Maximum Systolic BP (mmHg)160.0 (145.0, 174.0)153.0 (140.0, 170.0)150.0 (140.0, 165.0)0.037Minimum Systolic BP (mmHg)105.0 (95.0, 120.0)105.0 (92.0, 117.5)100.0 (90.0, 110.0)0.081Maximum Diastolic BP (mmHg)72.0 (65.0, 85.0)70.0 (60.0, 80.0)70.0 (60.0, 75.0)< 0.001Minimum Diastolic BP (mmHg)50.0 (45.0, 57.0)50.0 (40.5, 55.0)48.0 (40.0, 55.0)0.037Maximum Mean Arterial Pressure (mmHg)101.0 (93.0, 115.0)97.5 (89.0, 107.0)96.0 (87.0, 106.0)< 0.001Minimum Mean Arterial Pressure (mmHg)68.0 (61.0, 76.0)67.0 (61.0, 76.0)65.0 (60.0, 72.0)0.050Maximum Sodium (mmol/L)140.0 (137.0, 141.0)139.0 (136.0, 140.5)139.0 (136.0, 140.0)0.260Minimum Sodium (mmol/L)137.0 (134.0, 139.0)136.0 (134.0, 139.0)137.0 (134.0, 139.0)0.339Maximum Potassium (mmol/L)4.6 (4.3, 4.9)4.5 (4.3, 4.8)4.6 (4.3, 4.8)0.451Minimum Potassium (mmol/L)4.2 (3.9, 4.5)4.2 (4.0, 4.5)4.3 (4.0, 4.6)0.669Maximum HCO3 (mmol/L)26.0 (24.0, 28.0)25.0 (23.8, 27.0)25.0 (23.0, 27.0)0.026Minimum HCO3 (mmol/L)24.0 (22.0, 25.0)24.0 (22.0, 25.0)24.0 (21.7, 26.0)0.979Maximum Creatinine (umol/L)80.0 (67.0, 100.5)94.0 (76.0, 116.0)89.0 (69.0, 115.0)0.001Minimum Creatinine (mumol/L)73.0 (60.0, 86.5)85.0 (69.0, 105.0)80.0 (61.0, 101.0)0.001Maximum Haematocrit (L/L)0.4 (0.4, 0.4)0.4 (0.3, 0.4)0.3 (0.3, 0.3)< 0.001Minimum Haematocrit (L/L)0.4 (0.3, 0.4)0.3 (0.3, 0.3)0.3 (0.3, 0.3)< 0.001Maximum Haemoglobin (g/L)131 (124, 139)116 (113, 120)101 (94, 105)< 0.001Minimum Haemoglobin (g/L)116 (107, 125)106 (101, 114)92 (84, 99)< 0.001Maximum WCC (10^9^/L)11.7 (9.1, 13.5)10.7 (8.9, 13.9)10.6 (8.5, 12.8)0.128Minimum WCC (10^9^/L)8.1 (6.5, 10.3)8.6 (6.8, 10.6)8.5 (6.8, 10.8)0.573Maximum Platelets (10^9^/L)204.5 (173.0, 254.0)201.0 (163.0, 240.0)210.0 (163.0, 256.0)0.437Minimum Platelets (10^9^/L)188.0 (149.0, 233.0)181.5 (148.0, 217.0)197.0 (151.0, 227.0)0.228Maximum Glucose (mmol/L)8.4 (7.3, 10.0)7.9 (6.9, 9.8)8.0 (6.7, 9.4)0.073Minimum Glucose (mmol/L)6.7 (5.9, 7.4)6.4 (5.6, 7.1)6.2 (5.5, 7.2)0.059PaO2 (mmHg)91.0 (72.5, 123.0)107.0 (78.0, 143.0)97.0 (76.0, 134.0)0.181PaCO2 (mmHg)44.0 (37.0, 48.0)42.0 (38.0, 46.0)41.0 (38.0, 45.0)0.233pH7.4 (7.3, 7.4)7.4 (7.3, 7.4)7.4 (7.4, 7.4)0.501Urea (mmol/L)7.1 (5.7, 9.3)8.3 (6.5, 10.8)8.4 (6.3, 10.5)0.009Urine Output (mL)1,300.0 (1,000.0, 1,715.0)1,365.0 (1,000.0, 1,800.0)1,335.0 (999.0, 1,700.0)0.979Bilirubin (µmol/L)9.0 (7.0, 12.0)9.0 (7.0, 14.0)9.0 (6.0, 12.0)0.166Albumin (g/L)33.0 (30.0, 38.0)31.0 (28.0, 34.0)30.0 (27.0, 33.0)< 0.001VariableCategoryNon-anaemiaN = 110Mild anaemiaN = 148Severe anaemiaN = 235p-valueCountryAustralia110 (100%)145 (98%)230 (98%)0.310CountryNew Zealand0 (0%)3 (2.0%)5 (2.1%)0.310Hospital SourceHome/hospital in the home106 (97%)141 (95%)213 (92%)0.486Hospital SourceOther acute hospital (not ICU/ED)0 (0%)3 (2.0%)7 (3.0%)0.486Hospital SourceNursing home/chronic care/palliative care2 (1.8%)3 (2.0%)8 (3.4%)0.486Hospital SourceRehabilitation1 (0.9%)0 (0%)1 (0.4%)0.486Hospital SourceOther hospital ED0 (0%)1 (0.7%)3 (1.3%)0.486SexF75 (68%)59 (40%)144 (61%) < 0.001SexM35 (32%)89 (60%)91 (39%) < 0.001ICU SourceOT/recovery109 (99%)146 (99%)232 (99%)0.942ICU SourceEmergency department0 (0%)1 (0.7%)1 (0.4%)0.942ICU SourceWard/coronary care/other HDU1 (0.9%)1 (0.7%)2 (0.9%)0.942ReadmittedYes1 (0.9%)0 (0%)3 (1.3%)0.395Died in ICUNo109 (100%)147 (100%)234 (100%)> 0.99Died in HospitalYes0 (0%)1 (0.7%)3 (1.3%)0.463Chronic Respiratory DiseaseYes2 (1.8%)13 (8.8%)9 (3.8%)0.021Chronic Cardiovascular DiseaseYes15 (14%)40 (27%)40 (17%)0.012Chronic Liver DiseaseNo1 (0.9%)1 (0.7%)2 (0.9%)0.974Chronic Renal DiseaseYes2 (1.8%)7 (4.7%)13 (5.5%)0.292Immunological DiseaseYes0 (0%)0 (0%)2 (0.9%)0.332Immunosuppressive TherapyYes1 (0.9%)2 (1.4%)2 (0.9%)0.886AIDSNo110 (100%)148 (100%)235 (100%)> 0.99Hepatic FailureNo110 (100%)148 (100%)235 (100%)> 0.99LymphomaNo110 (100%)148 (100%)235 (100%)> 0.99MetastasisYes1 (0.9%)2 (1.4%)4 (1.7%)0.842LeukaemiaYes0 (0%)0 (0%)1 (0.4%)0.576ImmunosuppressionYes1 (0.9%)2 (1.4%)2 (0.9%)0.886CirrhosisNo1 (0.9%)1 (0.7%)2 (0.9%)0.974Insulin-Dependent Diabetes MellitusYes0 (0%)0 (0%)1 (1.4%)0.600Treatment LimitationFull active management (without treatment limitation)106 (96%)138 (93%)222 (95%)0.535Treatment LimitationTreatment limitation order4 (3.6%)10 (6.8%)12 (5.1%)0.535DeliriumYes2 (3.0%)3 (4.0%)4 (3.3%)0.942Acute Renal FailureYes1 (1.0%)1 (0.7%)2 (0.9%)0.976Elective SurgeryYes107 (97%)138 (93%)220 (94%)0.312AliveYes88 (80%)110 (74%)166 (71%)0.180Hospital OutcomeDied0 (0%)1 (0.7%)3 (1.3%)0.507Hospital OutcomeHome43 (40%)57 (39%)110 (47%)0.507Hospital OutcomeNursing home/chronic care/palliative care23 (21%)31 (21%)51 (22%)0.507Hospital OutcomeOther acute hospital10 (9.3%)10 (6.8%)19 (8.1%)0.507Hospital OutcomeRehabilitation32 (30%)48 (33%)51 (22%)0.507Hospital OutcomeOther0 (0%)0 (0%)1 (0.4%)0.507ICU OutcomeSurvived ICU: home/ward/other109 (100%)146 (99%)234 (100%)0.310ICU OutcomeTransferred to another ICU in the same hospital or a different hospital0 (0%)1 (0.7%)0 (0%)0.310Data are presented as median (interquartile range) for continuous variables and number (percentage) for categorical variables. Comparisons between groups were performed with the Kruskal–Wallis test for continuous variables, and the chi-squared test or Fisher’s exact test for categorical variables, as appropriate. Anaemia was classified according to World Health Organization haemoglobin non-anaemia (≥ 130 g/L for men or ≥ 120 g/L for women), mild anaemia (110–129 g/L for men or 110–119 g/L for women), or severe anaemia (< 110 g/L for both sexes)
Patients with severe anaemia had greater illness severity and resource utilisation than those in the other groups. The median APACHE III scores increased with anaemia 54.0 (47.0–62.0) in the non-anaemia group, 57.0 (49.0–63.5) in the mild anaemia group, and 57.0 (51.0–65.0) in the severe anaemia group (p = 0.012). ICU LOS was longer in the severe anaemia group (24.0 h [19.5–45.8]) than the non-anaemia group (22.0 h [19.8–25.8; p = 0.041]). Hospital LOS was 191.4 h (145.0–287.0) in the non-anaemia group and 219.4 h (147.5–376.5) in the severe anaemia group (p = 0.037).
Several biomarkers demonstrated statistically significant trends across anaemia categories. Albumin levels declined with increasing anaemia 33 g/L (30–38) in the non-anaemia group, 31 g/L (28–34) in the mild anaemia group, and 30 g/L (27–33) in the severe anaemia group (p < 0.001). Urea levels were highest in the anaemia 7.1 mmol/L (5.7–9.3) in the non-anaemia group, 8.3 mmol/L (6.5–10.8) in the mild anaemia group, and 8.4 mmol/L (6.3–10.5) in the severe anaemia group (p = 0.009). Maximum creatinine levels also rose progressively with anaemia 80 µmol/L (67–100.5) in the non-anaemia group, 94 µmol/L (76–116) in the mild anaemia group, and 89 µmol/L (69–115) in the severe anaemia group (p = 0.001).
Patients with anaemia also had lower arterial blood pressure measurements than those without anaemia. The maximum systolic blood pressure was 160 mmHg (145–174) in the non-anaemia group, 153 (140–170) in the mild anaemia group, and 150 (140–165) in the severe anaemia group (p = 0.038). The maximum mean arterial pressure similarly declined from 101 mmHg (93–115) to 96 mmHg (87–106) across the spectrum (p < 0.001). Bicarbonate levels showed a similar downward trend (maximum HCO₃: 26 mmol/L [24–28] in the non-anaemia group, with respect to the 25 mmol/L [23–27] in the severe anaemia group; p = 0.026), thus potentially reflecting metabolic compromise.
The sex distribution varied the men were more frequently anaemic than the women (non-anaemia 32% male; mild anaemia 60% male; severe anaemia 39% male; p < 0.001). Chronic cardiovascular disease was more prevalent in the anaemia groups (non-anaemia: 14%; mild 27%; severe 17%; p = 0.012), as was chronic respiratory disease (non-anaemia: 1.8%; mild 8.8%; severe 3.8%; p = 0.021).
No significant differences were observed in ICU admission source, readmission rates, or ICU mortality across groups. Elective surgery predominated in all groups (> 93%). Hospital discharge destinations, delirium incidence, and treatment limitation orders did not significantly differ by anaemia status.
Interval-specific mortality rates stratified by anaemia severity are summarised in Table 2. Within the first 365 days after ICU admission, mortality was 13.6% in the non-anaemic group, 7.1% in the mild anaemia group, and 17.7% in the severe anaemia group. Beyond 365 days, mortality increased across all 24.2% in the non-anaemia group, 42.3% in the mild anaemia group, and 40.2% in the severe anaemia group. Notably, although mild anaemia was associated with the lowest early mortality, it exhibited the highest long-term mortality. Severe anaemia was associated with persistently elevated mortality across both time intervals.Table 2Mortality by anaemia group and time intervalTime groupAnaemia groupnDeathsMortality (%)95% CI lower95% CI upper< 365 daysNormal44613.65.227.4< 365 daysMildly anaemic7057.12.415.9< 365 daysSeverely anaemic1132017.711.226.0≥ 365 daysNormal661624.214.536.4≥ 365 daysMildly anaemic783342.331.254.0≥ 365 daysSeverely anaemic1224940.231.449.4Data are stratified by anaemia severity and survival time interval. Deaths represent in-hospital mortality within each specified time frame from ICU admission. Percentages reflect the proportion of deaths within each anaemia category. Anaemia was classified according to admission haemoglobin levels according to WHO sex-specific cutoffs
Figure 1 displays Kaplan–Meier survival curves stratified by anaemia severity at ICU admission. Over the entire follow-up period, the survival probabilities significantly differed across anaemia groups (log-rank p = 0.0091). Patients with severe anaemia experienced the steepest decline in survival, and were followed by those with mild anaemia, whereas patients without anaemia consistently demonstrated the highest survival probability. These findings visually reinforced the prognostic gradient observed across anaemia severity and provided unadjusted support for the association between anaemia and long-term mortality risk.Fig. 1Kaplan–Meier survival curve by anaemia severity at ICU admission
The association between anaemia severity and mortality across early and late follow-up periods is summarised in Table 3. In the early post-ICU period (< 365 days), neither mild nor severe anaemia was significantly associated with mortality in univariate or multivariate models. The adjusted hazard ratio (HR) with respect to non-anaemia was 0.268 (95% CI 0.054–1.320; p = 0.106) for mild anaemia and 0.821 (95% CI 0.237–2.852; p = 0.757) for severe anaemia. Other covariates, including age, sex, APACHE III score, and surgical urgency, were not significantly associated with short-term mortality in this time frame.Table 3Association between anaemia severity and mortality, stratified by time intervalVariableUnivariate HR (95% CI)Univariate p-valueMultivariate HR (95% CI)Multivariate p-value< 365 days Mildly Anaemic (vs Normal)0.505 (0.153–1.670)0.2620.268 (0.054–1.320)0.106 Severely Anaemic (vs Normal)1.180 (0.468–2.960)0.7290.821 (0.237–2.852)0.757 Age1.040 (0.868–1.260)0.6420.876 (0.669–1.147)0.336 Male Sex1.140 (0.552–2.370)0.7191.986 (0.784–5.033)0.148 APACHE III1.010 (0.991–1.040)0.2161.019 (0.989–1.051)0.224 Non-elective Surgery0.712 (0.212–2.390)0.5830.629 (0.124–3.203)0.577\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ \ge
In contrast, during the late follow-up period (≥ 365 days), both mild and severe anaemia were independently associated with significantly elevated mortality. After adjustment for confounders, the HR for mild anaemia was 2.082 (95% CI 1.050–4.127; *p* = 0.036), and that for severe anaemia was 1.976 (95% CI 1.048–3.726; *p* = 0.035), with respect to the non-anaemia group. A higher APACHE III score showed a trend toward elevated risk (HR 1.021; 95% CI 0.997–1.046; *p* = 0.083), whereas non-elective surgery was associated with diminished risk of long-term mortality (HR 0.445; 95% CI 0.135–1.468; *p* = 0.184), although this association was not statistically significant. Age and sex did not demonstrate significant associations in the adjusted long-term model. ### Sensitivity analysis – lowest haemoglobin and haemoglobin variation To assess the robustness of the primary findings to alternative definitions of anaemia severity, a sensitivity analysis was performed using the lowest haemoglobin value within the first 24 h of ICU admission. Cox regression analyses stratified by time interval (< 365 days and ≥ 365 days) are presented in Supplementary Table 1. Within the first 365 days, neither mild anaemia (adjusted HR 0.507, 95% CI 0.146–1.763) nor severe anaemia (adjusted HR 1.394, 95% CI 0.552–3.521) was significantly associated with mortality, whereas beyond 365 days both mild anaemia (adjusted HR 2.110, 95% CI 1.133–3.930) and severe anaemia (adjusted HR 2.145, 95% CI 1.209–3.807) were independently associated with increased mortality, consistent with the general pattern observed in the primary analysis. In adjusted restricted cubic spline analyses, haemoglobin variation was not clearly associated with mortality across the observed range (Supplementary Fig. 1). Although point estimates suggested a modest inverse trend, confidence intervals were wide throughout, indicating substantial uncertainty and no robust or clinically meaningful association between haemoglobin variation and mortality after adjustment. ### ICU and hospital LOS Associations between anaemia severity and hospital and ICU LOS are summarised in Table 4. In the unadjusted analyses, severe anaemia was significantly associated with a prolonged hospital stay (Exp[β] 1.23, 95% CI 1.05–1.45; *p* = 0.011) 23% longer than observed in the non-anaemia group. However, this association was attenuated in multivariable models adjusting for age, sex, APACHE III score, and surgical urgency (Exp[β] 1.15, 95% CI 0.98–1.34; *p* = 0.088). Mild anaemia was not associated with hospital LOS in either model.Table 4Linear association between anaemia and hospital/ICU length of stayOutcomePredictorUnivariate (Exp[β], 95% CI)Univariate *p*-valueMultivariate (Exp[β], 95% CI)Multivariate *p*-valueHospital Length of StayAnaemia: normalRef-Ref-Anaemia: mild1.024 (0.860, 1.221)0.7870.952 (0.799, 1.133)0.577Anaemia: severe1.232 (1.050, 1.447)0.0111.147 (0.980, 1.343)0.088Age1.034 (1.002, 1.067)0.0371.029 (0.998, 1.061)0.071Male sex1.060 (0.934, 1.203)0.3701.086 (0.957, 1.231)0.200APACHE III1.015 (1.009, 1.020)< 0.001*1.013 (1.007, 1.018)< 0.001*Elective surgery0.626 (0.480, 0.815)0.0010.694 (0.533, 0.904)0.007ICU Length of StayAnaemia: normalRef-Ref-Anaemia: mild1.085 (0.953, 1.235)0.2201.036 (0.908, 1.182)0.597Anaemia: severe1.215 (1.079, 1.369)0.0011.176 (1.044, 1.324)0.008Age1.022 (0.998, 1.046)0.0701.019 (0.996, 1.043)0.102Male sex1.069 (0.973, 1.173)0.1651.083 (0.985, 1.191)0.101APACHE III1.005 (1.001, 1.010)0.0131.004 (0.999, 1.008)0.086Elective surgery0.701 (0.576, 0.853)< 0.001*0.733 (0.601, 0.893)0.002Linear regression was performed on log-transformed hospital and ICU length of stay, with coefficients exponentiated to present results as multiplicative effects (Exp[β]) on the original scale. Values represent the ratios of geometric means with corresponding 95% confidence intervals (CI). The reference group for anaemia status is non-anaemia. The *p*-values were derived from Wald tests; statistically significant results (*p* < 0.05) are denoted with an asterisk*. All models were adjusted for age, sex, APACHE III score, and elective surgery status in the multivariate analysis* In contrast, even after adjustment, severe anaemia remained independently associated with a prolonged ICU stay (Exp[β] 1.18, 95% CI 1.04–1.32; *p* = 0.008) 18% longer than observed in the non-anaemia group. Mild anaemia was not significantly associated with ICU stay duration. Across both outcomes, elective surgery was independently associated with shorter hospital (Exp[β] 0.694, 95% CI 0.533–0.904; *p* = 0.007) and ICU (Exp[β] 0.733, 95% CI 0.601–0.893; *p* = 0.002) LOS (30.6% and 26.7% shorter, respectively, than observed for non-elective surgery). In contrast, higher APACHE III scores were significantly associated with longer hospital stays (Exp[β] 1.013, 95% CI 1.007–1.018; *p* < 0.001), although this association was not statistically significant after adjustment (Exp[β] 1.004, 95% CI 0.999–1.008; *p* = 0.086). ## Discussion ### Key findings In this large binational cohort of nonagenarians undergoing TKA, anaemia at ICU admission was highly prevalent, affecting approximately four in five patients, and was associated with enhanced illness severity and resource utilisation. Although anaemia was not significantly associated with short-term mortality, both mild and moderate to severe anaemia independently predicted higher long-term mortality beyond 1 year. Severe anaemia was further associated with prolonged ICU and hospital stays, even after adjustment for confounders. Collectively, these findings underscored the prognostic importance of early postoperative anaemia and indicated that its clinical impact extends well beyond the immediate postoperative course. ### Relationship of the findings to the literature Despite the high prevalence of postoperative anaemia, particularly among surgical patients, few studies have examined its consequences beyond 1 year. A 2018 consensus statement on postoperative anaemia management underscored the paucity of evidence addressing outcomes in the recovery phase beyond several months (Muñoz et al. 2017).A large retrospective analysis of 142,510 surgical procedures within the US Veterans Health Administration has indicated that severe postoperative anaemia (nadir Hb < 10 g/dL) is strongly associated with diminished long-term survival, and was the second strongest independent predictor of mortality after age in fully adjusted models over 12 years of follow-up (Kougias et al. 2019). Similarly, studies in critically ill post-surgical patients have demonstrated that severe anaemia within the first week of ICU admission confers a 16–18% increase in 1-year mortality risk, even after propensity adjustment, and corroborating evidence has been provided by additional propensity score–based analyses (Wu et al. 2022; Lin et al. 2023). Collectively, these findings highlight postoperative anaemia as a potentially modifiable risk factor with long-term implications. The present study extends this limited body of work by characterising the prognostic importance of anaemia at ICU admission after elective orthopaedic surgery in nonagenarians. ### Clinical implications The consistent association between postoperative anaemia at ICU admission and elevated long-term mortality underscores the need for better perioperative and postoperative strategies to prevent and manage anaemia. Given its observed association with long-term mortality, postoperative anaemia may serve as a useful prognostic marker in elderly surgical patients admitted to intensive care. Optimising Hb levels in the preoperative phase through targeted interventions might not only facilitate short-term recovery but also influence long-term survival. Furthermore, the trajectory of Hb recovery after surgery might serve as both a prognostic marker and a potential therapeutic target in elderly surgical patients. ### Directions for future research Future research should delineate the mechanisms through which early postoperative anaemia contributes to long-term mortality in older surgical patients, particularly those undergoing elective orthopaedic procedures. Prospective trials are needed to determine whether targeted perioperative interventions, including preoperative Hb optimisation, nutritional support, and enhanced postoperative surveillance, might improve survival and recovery in this population. In addition, patient-centred studies should examine how anaemia affects frailty, functional trajectories, and quality of life among nonagenarians. As the number of patients of very advanced age undergoing elective surgery continues to increase, the development of evidence-based guidelines for the assessment and management of anaemia in this vulnerable group is a critical priority. ### Strengths and limitations This study has several strengths. To our knowledge, it is among the first and largest investigations to evaluate the prognostic value of anaemia among critically ill nonagenarians undergoing TKA. The use of a high-quality, binational ICU database encompassing more than 98% of Australian and 68% of New Zealand ICUs ensured excellent data completeness, clinical granularity, and real-world generalisability among high-income health systems. The incorporation of validated severity scores, objective biochemical measures, and long-term follow-up enhanced the robustness of the findings. In addition, stratifying anaemia according to WHO criteria and applying time-specific mortality modelling provided clinically relevant nuance beyond the binary anaemia definitions commonly used in prior literature. Several study limitations warrant consideration. First, the retrospective observational design precluded causal inference, and residual confounding might have persisted despite statistical adjustment. Second, the study population was drawn exclusively from high-resource health systems in Australia and New Zealand, thus potentially limiting the generalisability of the findings to settings with different perioperative practices. Third, the absence of data on perioperative transfusions, iron supplementation, or other haematinic therapies restricted assessment of anaemia management and its potential influence on outcomes. Fourth, detailed data on postoperative anaemia management, including transfusion practices (thresholds, targets, and volumes) and the use of iron or other haematinic therapies, were unavailable, limiting evaluation of how treatment strategies may have modified the observed associations. Moreover, although this binational registry captured all eligible patients over the study period, the number of mortality events was relatively low, particularly within individual anaemia categories. This may have limited statistical power for detecting modest associations and reduced the precision of adjusted estimates, especially for short-term mortality. Finally, this study was restricted to nonagenarians and centenarians admitted to the ICU following total knee arthroplasty, representing a highly selected subgroup with increased physiological vulnerability and institutional variation in postoperative care. Consequently, the findings may not be generalizable to younger patients, those not requiring ICU admission, or broader elective arthroplasty populations. ## Conclusion In this large, multicentre cohort of nonagenarians admitted to intensive care after TKA, anaemia at ICU admission was highly prevalent, affecting nearly four in five patients, and was associated with elevated illness severity and prolonged resource utilisation. Although anaemia did not significantly affect short-term survival, both mild and moderate to severe anaemia independently predicted elevated long-term mortality beyond 1 year. Overall, postoperative haemoglobin levels were closely associated with longer-term outcomes in the oldest surgical patients and may offer clinically relevant insight when considering recovery and longer-term follow-up in this growing, high-risk population. ## Supplementary Information Supplementary Material Supplementary Table 1. Association between anaemia severity and mortality, stratified by time interval using lowest haemoglobin. Supplementary Figure 1. Association between haemoglobin variation and mortality.