Surgical outcomes in nonagenarian versus octogenarian patients: a propensity-matched analysis with implications for shared decision-making

Abstract

Background:

Although surgical outcomes among octogenarian patients are well documented, evidence for nonagenarian patients is limited. We sought to compare surgical outcomes between these age groups to guide clinical decision-making.

Methods:

We conducted a retrospective cohort study (2013 to 2023) with 1:1 propensity-score matching. We included patients aged 80 to 99 years who underwent general surgery. The primary outcome was 30-day all-cause mortality. Secondary outcomes included 90-day and 1-year mortality, functional status at last follow-up, complications (Clavien–Dindo classification), and hospital readmissions.

Results:

From 700 screened patients aged 80 to 99 years who underwent general surgery, 174 met inclusion criteria (73 nonagenarian and 101 octogenarian patients), yielding 73 matched pairs for analysis. Nonagenarian patients had significantly higher 30-day mortality (26.0% v. 9.6%, p = 0.02), 90-day mortality (49.3% v. 23.3%, p = 0.002), and 1-year mortality (75.3% v. 39.7%, p < 0.001) than octogenarian patients. At last follow-up (median 11 to 12 mo), poor functional status was observed in 34.2% of nonagenarian versus 23.3% of octogenarian patients. Hospital readmissions within 30 days occurred in 42.5% of nonagenarian versus 21.9% of octogenarian patients (p = 0.002). Despite propensity matching, the Fried frailty phenotype remained significantly imbalanced between groups (standardized mean difference 0.714, p < 0.001).

Conclusion:

Nonagenarian patients face substantially worse surgical outcomes than octogenarian patients, with nearly triple the 30-day mortality and high rates of functional impairment among survivors. The persistent frailty imbalance despite matching suggests inherent selection bias in surgical nonagenarians. Unlike octogenarians, for whom selective surgery may be justified, these findings support careful consideration of nonoperative management as the default approach for nonagenarians, with surgery reserved for highly select cases after comprehensive geriatric assessment and thorough shared decision-making with the patient.

The remarkable advances in modern medicine, public health, and societal development have fundamentally transformed human longevity. Life expectancy has dramatically increased because of multiple convergent factors such as improved hygiene and sanitation; enhanced accessibility to sports and recreational activities that promote healthier lifestyles; better work–life balance and decreased exposure to dangerous working conditions, reducing occupational hazards; and unprecedented access to medical knowledge and sophisticated diagnostic tools.¹ The evolution from invasive to minimally invasive treatments has further reduced procedural mortality while maintaining therapeutic efficacy.² Consequently, people aged 80 years or older represent the fastest-growing demographic segment globally. Currently, octogenarian people (aged 80 to 89 yr) comprise 4% to 5% of developed nations’ populations, while nonagenarian people (aged 90 to 99 yr) represent 0.5% to 1%, with projections indicating these figures will triple by 2050.^3,4

This demographic shift necessitates refined medical considerations. Different medical specialties demonstrate varying outcomes among older patients. Internal medicine manages chronic conditions with generally favourable results through optimization,⁵ while orthopedic surgery reports 20% 3-month mortality among nonagenarian patients with hip fractures, despite the potential for functional recovery.^6,7 General surgery presents particular challenges, as general anesthesia in nonagenarian patients correlates with substantial functional deconditioning, with up to 40% failing to return to baseline status, compared with 20% to 25% of octogenarian patients.^8,9

Current literature presents conflicting perspectives on surgical interventions in these older populations. Supporting studies demonstrate acceptable outcomes with careful patient selection,^10,11 while others report prohibitive 30-day mortality rates of 15% to 25% among nonagenarian patients, triple those observed among octogenarian patients.^12,13 Emergency procedures carry particularly poor prognoses, with 1-year mortality reaching 60% to 75% among nonagenarian patients, compared with 35% to 55% among octogenarian patients.^14,15

Previous research has established that well-selected octogenarian patients can achieve surgical outcomes comparable to septuagenarian patients, particularly when baseline functional status is preserved.^16,17 Building on this foundation, our study aimed to determine whether nonagenarian patients face categorically worse outcomes than octogenarian patients following general surgery, and whether preoperative functional status, cognitive function, and frailty assessment could predict outcomes in this vulnerable population,^18,19 to determine whether chronological age or physiologic status should drive surgical decision-making in the tenth decade of life.

Methods

Study design and setting

We conducted a retrospective cohort study at a university-affiliated hospital in central Holon, Israel, including all consecutive nonagenarian patients (aged 90 to 99 yr) and octogenarian patients (aged 80 to 89 yr) who underwent general surgery between Jan. 1, 2013, and Dec. 31, 2023. The study protocol adhered to Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklist for observational studies.²⁰

We hypothesized that, among patients aged 80 to 99 years who underwent general surgery, nonagenarian patients would have higher 30-day all-cause mortality (primary outcome) and worse secondary outcomes (including 90-day and 1-year mortality, complications, ICU admission, functional status at last follow-up, and 30-day readmissions) than octogenarian patients.

Patient selection and data collection

We retrospectively identified 700 patients through systematic query of electronic medical records using procedural codes for general surgical interventions from the 9th and 10th revisions of the International Classification of Diseases. We included all patients aged 80 to 99 years who underwent 1 of 4 types of general surgical procedures, namely cholecystectomy for biliary disease, appendectomy for acute appendicitis, right hemicolectomy for colonic pathology, and inguinal hernia repair. We selected these procedures as they represent common emergency and elective general surgical interventions in older populations. We excluded patients who underwent diagnostic procedures without therapeutic intervention, those who underwent procedures performed under local anesthesia only (given their fundamentally different physiologic stress and recovery patterns compared with general or regional anesthesia), those with incomplete medical records (< 80% of propensity matching variables), and patients lost to follow-up within 30 days of surgery. We included patients who underwent hernia repairs under spinal anesthesia. We included right hemicolectomies only as our institution refers left-sided emergency colon procedures to the regional colorectal centre.

We collected data on airway preparation, including preoperative respiratory physiotherapy and incentive spirometry training for high-risk patients; management of chronic obstructive pulmonary disease, including bronchodilator optimization and chest physiotherapy, when indicated by pulmonology consultation; and stable induction, defined as absence of hypotension (systolic blood pressure < 90 mm Hg) or substantial tachycardia (heart rate > 120 beats/min) during anesthetic induction requiring vasopressor support.

Primary outcome

The primary outcome was 30-day all-cause mortality, verified through cross-referencing hospital records with the National Population Registry by 2 independent reviewers (F.K., M.K.). Disagreements were resolved by consensus with a third reviewer (N.Z.).

Secondary outcomes

Secondary outcomes encompassed multiple domains reflecting the multidimensional impact of surgery on nonagenarian patients, including additional mortality variables, functional status at last follow-up, postoperative complications, hospital resource utilization, measures related to process of care, and readmission metrics.

We assessed in-hospital, 90-day, and 1-year mortality with exact time documentation to create comprehensive survival curves. We distinguished surgery-related deaths from other causes using detailed chart review and death certificate analysis. We verified 90-day and 1-year mortality through the same National Population Registry cross-referencing protocol used for 30-day mortality.

Recognizing that survival alone inadequately captures meaningful outcomes among nonagenarian patients, we assessed functional status at last available follow-up (median 11 mo, interquartile range [IQR] 6 to 14 mo). We classified functional status as good (return to baseline or minimal decline from baseline) or poor (substantial functional decline or institutional placement).

We assessed postoperative complications using the Clavien–Dindo classification system. Grade I or II complications were minor complications managed medically, including surgical site infections (United States Centers for Disease Control and Prevention criteria), pneumonia (clinical or radiographic diagnosis), and medical management of organ dysfunction. Grade III complications included those requiring procedural, endoscopic, or radiologic intervention. Grade IVa or IVb complications included single- or multiorgan dysfunction requiring intensive care unit (ICU) admission, including acute kidney injury (Kidney Disease: Improving Global Outcomes [KDIGO] criteria), myocardial infarction (elevated troponins with clinical correlation), respiratory failure requiring prolonged mechanical ventilation, thromboembolic events (pulmonary embolism or deep vein thrombosis), and cerebrovascular accidents. A grade V complication is defined as death. We also determined prolonged mechanical ventilation, defined as invasive mechanical ventilation via endotracheal intubation lasting more than 48 hours postoperatively. Data on noninvasive positive pressure ventilation use were not systematically collected in this retrospective study. We assessed the occurrence and duration of delirium using Confusion Assessment Method criteria. Unplanned returns to the operating room were documented separately.

For hospital resource utilization, we determined ICU admission rates and length of ICU stay (in days), total hospital length of stay (from surgical admission to discharge), and discharge disposition (home, rehabilitation facility, or skilled nursing facility).

Measures related to process of care focused on implementation of enhanced recovery elements, including early mobilization (ambulation within 24 h), timely removal of invasive devices (nasogastric tubes, urinary catheters, peripheral lines), progression of diet, participation in daily rehabilitation, multimodal pain management minimizing opioid exposure, and geriatrician involvement for polypharmacy management.

We determined hospital readmissions within 30 days of discharge, characterized by location (acute hospital v. rehabilitation facility) and primary reason (surgical complications, medical complications, or failure to thrive).

Other variables

Comprehensive geriatric assessment parameters

We systematically assessed functional status using validated Hebrew translations of the Katz Index for Activities of Daily Living (ADLs) (bathing, dressing, toileting, transferring, continence, feeding) and the Lawton Scale for Instrumental ADLs (telephone use, shopping, food preparation, housekeeping, laundry, transportation, medication management, finances).^21,22 Loss of 2 or more instrumental ADLs served as a benchmark for functional impairment, given its established association with increased postoperative complications and mortality.²³

Cognitive screening used Mini-Cog (3-word recall and clock drawing, with a cut-off < 3 indicating impairment) and Mini-Mental State Examination (MMSE) scores (cutoff < 24 suggesting dysfunction), when available.^24,25 For Hebrew-speaking patients, culturally adapted versions were employed.²⁶

Postoperative functional assessment

We determined functional status at last follow-up retrospectively through systematic review of outpatient clinic documentation, discharge summaries, and primary care reports. We classified patients as having good (return to baseline or minimal decline) or poor functional status (substantial decline from baseline or institutional placement) based on documented ADL performance.

We evaluated frailty retrospectively using modified Fried frailty phenotype criteria, adapted for chart review, including weight loss (unintentional weight loss > 5% in 6 mo or body mass index [BMI] < 18.5), documented exhaustion or fatigue in medical notes, low physical activity per therapy assessments, slowness (indicated by mobility aid use or gait speed < 0.8 m/s, when documented), and weakness (defined by documented grip strength < 20 kg for females or < 30 kg for males, or inability to rise from chair without assistance). We classified patients who met 3 or more criteria as frail.^27,28

We calculated the Charlson Comorbidity Index (CCI) using validated algorithms, with age-adjusted scores computed. ^29,30 We extracted American Society of Anesthesiologists (ASA) physical status classification from anesthesia records. We calculated the Emergency Surgery Acuity Score for nonelective procedures.³¹

We collected nutritional parameters, including serum albumin (cut-off < 3.5 g/dL), prealbumin (< 20 mg/dL), unintentional weight loss exceeding 5% within 6 months, and BMI. We calculated the Malnutrition Universal Screening Tool score retrospectively when data permitted.³²

We extracted complete medication lists, categorizing drugs using the Anatomic Therapeutic Chemical (ATC) classification system. We defined polypharmacy as 5 or more regular medications, with particular attention to Beers Criteria medications (anticoagulants, anticholinergics, benzodiazepines, and other central nervous system–active agents).³³ We calculated the Drug Burden Index when feasible.³⁴

Preoperative laboratory values included complete blood count, comprehensive metabolic panel, coagulation studies, and when available, inflammatory markers (C-reactive protein).

We identified deconditioning attributable to surgery through retrospective chart review based on documented functional decline in discharge summaries, physical therapy notes, or follow-up clinic documentation that specifically attributed functional deterioration to the surgical hospital admission rather than progression of underlying disease or medical complications. This included documentation of new mobility limitations, increased dependency in ADLs, or failure to progress in rehabilitation specifically related to postsurgical weakness or debility.

Ethics consultation

We collected data regarding ethics committee consultation, as in for cases involving complex decision-making or uncertainty regarding patient capacity. We also took note of documentation of patient willingness to proceed with surgery in the medical records, when available.

Data completion strategy

For incomplete retrospective data, we implemented a hierarchical approach to maximize information accuracy. We made direct telephone contact with patients without cognitive impairment (MMSE ≥ 24 or Mini-Cog ≥ 3) using structured questionnaires; for those with cognitive impairment, we contacted designated family members or caregivers. We also communicated with primary care physicians through the Clalit Health Services network, accessed the National Insurance Institute’s electronic database (Mobility Allowance and Long-Term Care Benefits assessments), and reviewed social services documentation through the national ALUMA system. This multisource approach achieved data completeness of more than 90% for key variables.

Propensity score matching

To balance measured baseline characteristics between age groups, we performed 1:1 propensity score matching between nonagenarian and octogenarian patients using nearest-neighbour matching without replacement.³⁵ We acknowledge this approach cannot eliminate selection bias inherent in surgical populations (i.e., only the “fittest” nonagenarian patients will undergo surgery) or address unmeasured confounding. We estimated the propensity score using multivariable logistic regression including sex, ASA class (III v. IV to V), surgical urgency (emergency or urgent v. elective), surgical category (cholecystectomy, appendectomy, right colectomy, or inguinal hernia repair), baseline functional status (independent v. dependent in ≥ 1 ADL), CCI, albumin level, and presence of cognitive impairment.

We performed matching with a caliper width of 0.2 standard deviations (SDs) of the logit of the propensity score. Standardized mean differences (SMDs) of less than 0.1 indicated adequate balance between groups.³⁶

Statistical analysis

We performed statistical analyses using SPSS version 28.0, R Studio version 2023.05.0+496 (with packages match it, survival, and table one), and GraphPad PRISM version 9.0 (for visualization).

We assessed continuous variables for normality using the Shapiro–Wilk test and Q-Q plots and expressed them as means and SDs or medians and IQRs accordingly. Categorical variables were presented as frequencies and percentages. We calculated between-group comparisons using the Student t test or the Mann–Whitney U test for continuous variables, and the χ² or Fisher exact test for categorical variables.

Survival analysis employed Kaplan–Meier curves with log-rank testing for univariate comparisons. We used Cox proportional hazards regression to identify independent predictors of mortality, with proportional hazards assumption verified using Schoenfeld residuals. Competing risk analysis using Fine–Gray models addressed discharge to hospice as a competing event.³⁷

We used logistic regression to determine risk factors for poor functional status at last follow-up and for major complications. We entered variables with p values less than 0.1 in univariate analysis into multivariable models using backward stepwise selection (retention p < 0.05). We assessed multi-collinearity using variance inflation factors (< 5 acceptable).

We evaluated model performance using C-statistics (area under receiver operating characteristic curve) and calibration via Hosmer–Lemeshow goodness-of-fit tests. Internal validation employed bootstrap resampling (1000 iterations) to assess optimism-corrected performance.

Subgroup analyses examined emergency versus elective procedures, specific surgical categories, patients with or without frailty, and patients with or without cognitive impairment. Interaction terms tested for effect modification.

We set the 2-tailed statistical significance at p less than 0.05. We analyzed missing data patterns using the Little Missing Completely at Random test. For variables with less than 10% missingness, we performed complete case analysis. Multiple imputation by chained equations (20 imputations) addressed variables with 10% to 30% missingness in sensitivity analyses.³⁸

Sample calculation indicated that 200 matched pairs would provide 80% power to detect a 10% absolute difference in 30-day mortality (assuming 15% baseline mortality in octogenarians) with an α of 0.05. The achieved sample provided 70% power for the observed difference in 30-day mortality but limited power for secondary outcomes and multivariable analyses.

Ethics approval

The study was approved by Wolfson Medical Center Institutional Review Board (0022-25-WOMC) and adhered to the Declaration of Helsinki principles.

Results

Baseline characteristics

Of 174 patients who met inclusion criteria, 73 (42.0%) were nonagenarian patients and 101 (58.0%) were octogenarian patients. Before matching, nonagenarian patients were more likely to undergo emergency surgery (38 [52.1%] v. 31 [30.7%], p < 0.001) and had higher frailty scores (median Fried score 2.0 v. 1.0, p < 0.001) (Table 1). After propensity matching, we analyzed 73 pairs. Despite matching, frailty remained imbalanced between groups (SMD 0.714) (Table 2). Emergency surgery rates among nonagenarian and octogenarian patients were similar after matching (38 [52.1%] v. 31 [42.5%], p = 0.3). General anesthesia was used for 63 (86.3%) nonagenarian patients and 68 (93.2%) octogenarian patients (Table 3).

Table 1.

Baseline characteristics before propensity score matching

Characteristic	No. (%) of patients*		p value	SMD
	Octogenarian patients n = 101	Nonagenarian patients n = 73
Age, yr, median (IQR)	84.32 (82.62–86.17)	92.00 (91.00–94.00)	< 0.001	–
Sex, male	48 (47.5)	35 (47.9)	> 0.9	0.008
BMI, median (IQR)	21.47 (19.49–24.17)	21.47 (19.88–24.17)	0.5	0.045
Cognition
Cognitive impairment (Mini-Cog 0)	40 (39.6)	30 (41.1)	> 0.9
Possible impairment (Mini-Cog 1–2)	20 (19.8)	14 (19.2)	> 0.9
Normal cognition (Mini-Cog 3)	41 (40.6)	29 (39.7)	> 0.9
Mini-Cog, median (IQR)	2.00 (0.00–3.00)	2.00 (0.00–3.00)	0.9	0.028
ASA class, median (IQR)	3.00 (3.00–4.00)	3.00 (3.00–4.00)	0.2	0.158
CCI, median (IQR)	7.00 (5.00–8.00)	7.00 (6.00–8.00)	0.4	0.089
Comorbidities
Ischemic heart disease	39 (38.6)	30 (41.1)	0.9	0.051
Diabetes mellitus	34 (33.7)	32 (43.8)	0.2	0.208
COPD	21 (20.8)	19 (26.0)	0.5	0.122
Hypertension	92 (91.1)	66 (90.4)	> 0.9	0.025
Hyperlipidemia	56 (55.4)	39 (53.4)	0.9	0.040
CRF or CKD	17 (16.8)	12 (16.4)	> 0.9	0.011
Fatty liver	40 (39.6)	25 (34.2)	0.5	0.110
Weight loss	18 (17.8)	16 (21.9)	0.6	0.103
NOAC or DOAC use	17 (16.8)	9 (12.3)	0.5	0.126
Antiaggregation therapy	42 (41.6)	29 (39.7)	0.9	0.038
Total no. of drugs, median (IQR)	4.00 (4.00–5.00)	4.00 (4.00–6.00)	0.2	0.142
Preoperative reassessment
Albumin < 3.5	25 (24.8)	20 (27.4)	0.7
Albumin, median (IQR)	3.50 (3.50–3.50)	3.50 (3.40–3.60)	0.6	0.082
Frailty
Robust (Fried 0)	6 (5.9)	0 (0.0)	< 0.001
Pre-frail (Fried 1–2)	89 (88.1)	51 (69.9)
Frail (Fried 3–5)	6 (5.9)	22 (30.1)
Fried frailty phenotype, median (IQR)	1.00 (1.00–1.00)	2.00 (1.00–3.00)	< 0.001	0.714
ADLs	15 (14.9)	26 (35.6)	< 0.001

ADLs = activities of daily living; ASA = American Society of Anesthesiologists; BMI = body mass index; CCI = Charlson Comorbidity Index; CKD = chronic kidney disease; COPD = chronic obstructive pulmonary disease; CRF = chronic renal failure; DOAC = direct oral anticoagulant; IQR = interquartile range; NOAC = novel oral anticoagulant; SMD = standardized mean difference.

^*Unless indicated otherwise.

Table 2.

Baseline characteristics after propensity score matching

Characteristic	No. (%) of patients*		p value	SMD
	Octogenarian patients n = 73	Nonagenarian patients n = 73
Age, yr, median (IQR)	84.95 (83.00–86.73)	92.00 (91.00–94.00)	< 0.001	–
Sex, male	36 (49.3)	35 (47.9)	> 0.9	0.027
BMI, median (IQR)	21.47 (19.49–24.17)	21.47 (19.88–24.17)	0.6	0.128
Mini-Cog, median (IQR)	2.00 (0.00–3.00)	2.00 (0.00–3.00)	0.5	0.121
ASA class, median (IQR)	3.00 (3.00–4.00)	3.00 (3.00–4.00)	0.9	< 0.001
CCI, median (IQR)	7.00 (5.00–8.00)	7.00 (6.00–8.00)	0.7	0.043
Comorbidities
Ischemic heart disease	29 (39.7)	30 (41.1)	> 0.9	0.028
Diabetes mellitus	25 (34.2)	32 (43.8)	0.3	0.198
COPD	17 (23.3)	19 (26.0)	0.8	0.064
Hypertension	69 (94.5)	66 (90.4)	0.5	0.156
Hyperlipidemia	44 (60.3)	39 (53.4)	0.5	0.139
CRF or CKD	12 (16.4)	12 (16.4)	> 0.9	< 0.001
Fatty liver	29 (39.7)	25 (34.2)	0.6	0.114
Weight loss	13 (17.8)	16 (21.9)	0.7	0.103
NOAC or DOAC use	12 (16.4)	9 (12.3)	0.7	0.131
Antiaggregation therapy	30 (41.1)	29 (39.7)	> 0.9	0.028
Total no. of drugs, median (IQR)	4.00 (4.00–5.00)	4.00 (4.00–6.00)	0.3	0.052
Preoperative reassessment
Albumin, median (IQR)	3.50 (3.50–3.50)	3.50 (3.40–3.60)	0.6	0.082
Fried frailty phenotype, median (IQR)	1.00 (1.00–1.00)	1.00 (1.00–3.00)	< 0.001	0.714

ASA = American Society of Anesthesiologists, BMI = body mass index; CCI = Charlson Comorbidity Index; CKD = chronic kidney disease; COPD = chronic obstructive pulmonary disease; CRF = chronic renal failure; DOAC = direct oral anticoagulant; IQR = interquartile range; NOAC = novel oral anticoagulant; SMD = standardized mean difference.

^*Unless indicated otherwise.

Table 3.

Operative characteristics after propensity score matching

Characteristic	No. (%) of patients*		p value	SMD
	Octogenarian patients n = 73	Nonagenarian patients n = 73
Emergency surgery	31 (42.5)	38 (52.1)	0.3	0.193
Airway preparation	23 (31.5)	18 (24.7)	0.5	0.153
CHF echocardiography performed	15 (20.5)	15 (20.5)	> 0.9	< 0.001
COPD management	24 (32.9)	18 (24.7)	0.4	0.182
General anesthesia	68 (93.2)	63 (86.3)	0.3	0.227
Intubation	65 (89.0)	65 (89.0)	> 0.9	< 0.001
Intubation conversion during surgery	3 (4.1)	4 (5.5)	0.6	0.167
Stable induction	63 (86.3)	65 (89.0)	0.8	0.083
Successful extubation	52 (71.2)	48 (65.8)	0.7	0.126
Surgery type			0.1	0.399
Cholecystectomy	30 (41.1)	26 (35.6)
Appendectomy	6 (8.2)	6 (8.2)
Right hemicolectomy	2 (2.7)	5 (6.8)
Inguinal hernia repair	35 (47.9)	36 (49.3)
Surgical duration, min, median (IQR)	138.01 (118.77–157.46)	143.95 (118.51–164.97)	0.5	0.118
Laparoscopic approach	31 (42.5)	32 (43.8)	> 0.9	0.028

CHF = congestive heart failure; COPD = chronic obstructive pulmonary disease; IQR = interquartile range; SMD = standardized mean difference.

^*Unless indicated otherwise.

Primary outcome

Thirty-day all-cause mortality was 19 (26.0%) among nonagenarian patients and 7 (9.6%) among octogenarian patients (p = 0.02) (Table 4 and Figure 1).

Table 4.

Postoperative outcomes after propensity score matching

Outcome	No. (%) of patients*		p value	SMD
	Octogenarian patients n = 73	Nonagenarian patients n = 73
ICU admission	3 (4.1)	10 (13.7)	0.03	0.33
Deconditioning after surgery	23 (31.5)	32 (43.8)	0.2	0.257
Discharge to facility	10 (13.7)	10 (13.7)	> 0.9	< 0.001
Follow-up visit attended	12 (16.4)	26 (35.6)	0.01	0.448
Good postoperative functional status	56 (76.7)	48 (65.8)	0.3	0.246
Length of follow-up, mo, median (IQR)	12.00 (10.00–14.00)	11.00 (6.00–14.00)	0.03	0.282
Hospital readmission within 30 d	16 (21.9)	31 (42.5)	0.002	0.671
Intraoperative death	2 (2.7)	5 (6.8)	0.4	0.193
Hospital stay, d, median (IQR)	5.00 (3.00–8.00)	5.00 (3.00–8.00)	0.8	0.117
Death from surgical complications	8 (16.0)	21 (36.8)	0.04	0.520
Mortality
In-hospital mortality	7 (9.6)	17 (23.3)	0.04	0.376
30-d mortality	7 (9.6)	19 (26.0)	0.02	0.440
90-d mortality	17 (23.3)	36 (49.3)	0.002	0.603
1-yr mortality	29 (39.7)	55 (75.3)	< 0.001	0.772

ICU = intensive care unit; IQR = interquartile range; SMD = standardized mean difference.

^*Unless indicated otherwise.

Fig. 1.

Kaplan–Meier survival curves showing 30-day mortality patterns in nonagenarian (n = 101) versus octogenarian (n = 73) patients, before propensity matching. Shaded areas represent 95% confidence intervals (CIs). Log-rank p = 0.4.

Secondary outcomes

Among both nonagenarian and octogenarian patients, mortality increased over time from 90 days (36 [49.3%] v. 17 [23.3%], p = 0.002) to 1 year after surgery (55 [75.3%] v. 29 [39.7%], p < 0.001) (Table 4 and Figure 1).

Admission to the ICU occurred in 10 (13.7%) nonagenarian patients and 3 (4.1%) octogenarian patients (p = 0.03). New institutional placement was required for 20 (27.4%) nonagenarian patients and 2 (2.7%) octogenarian patients (p < 0.001). Acute kidney injury occurred in 30 (41.1%) nonagenarian patients and 19 (26.0%) octogenarian patients (p = 0.08) (Table 5).

Table 5.

Postoperative complications after propensity score matching

Complication	No. (%) of patients*		p value	SMD
	Octogenarian patients n = 73	Nonagenarian patients n = 73
Prolonged intubation or ventilation	18 (24.7)	13 (17.8)	0.4	0.168
Surgical site infection	30 (41.1)	25 (34.2)	0.5	0.142
Pain killers requirement	30 (41.1)	32 (43.8)	0.9	0.055
Pneumonia	28 (38.4)	31 (42.5)	0.7	0.084
Acute kidney injury	19 (26.0)	30 (41.1)	0.08	0.323
Myocardial infarction	19 (26.0)	28 (38.4)	0.2	0.266
Pulmonary embolism	2 (2.7)	2 (2.7)	> 0.9	< 0.001
Delirium occurrence	6 (8.2)	11 (15.1)	0.3	0.215
New institutional placement	2 (2.7)	20 (27.4)	< 0.001	0.734
Daily rehabilitation	23 (31.5)	32 (43.8)	0.2	0.257
Delirium prevention measures	3 (4.1)	13 (17.8)	0.02	0.449
Oral intake start, d, median (IQR)	0.00 (0.00–2.00)	0.00 (0.00–0.00)	0.5	0.108
Venflon removal, d, median (IQR)	3.00 (2.00–5.00)	3.00 (2.00–5.00)	0.4	0.108

IQR = interquartile range; SMD = standardized mean difference.

^*Unless indicated otherwise.

At median 11-month follow-up, good functional status was achieved by 48 (65.8%) nonagenarian patients, compared with 56 (76.7%) octogenarian patients (p = 0.3) (Table 4).

Within 30 days, 31 (42.5%) nonagenarian patients and 16 (21.9%) octogenarian patients were readmitted (p = 0.002) (Table 4).

Survival analysis

Median survival was 18.5 (95% confidence interval [CI] 14.2 to 24.1) months for nonagenarian patients and 32.4 (95% CI 26.8 to 40.2) months for octogenarian patients (log-rank p < 0.001) (Figure 2).

Fig. 2.

Kaplan–Meier curves of survival probability over 48 months after general surgery among nonagenarian (median survival 18.5 mo, 95% confidence interval [CI] 14.2–24.1 mo) versus octogenarian patients (median survival 32.4 mo, 95% CI 26.8 to 40.2 mo) after propensity matching. Log-rank p < 0.001. The progressive divergence in survival curves demonstrates the sustained vulnerability of nonagenarian patients extending beyond the immediate perioperative period.

Predictors of mortality

In multivariable logistic regression for 30-day mortality, only BMI was significant (odds ratio 1.19, 95% CI 1.05 to 1.34). Frailty was excluded given its collinearity with age group (Table 6 and Table 7).

Table 6.

Univariate analysis of risk factors for 30-day mortality

Variable	OR (95% CI)
Age (per yr)	1.10 (0.99–1.23)
Sex, male	0.96 (0.39–2.36)
BMI (per unit)	1.22 (1.09–1.38)
ASA class (per level)	1.45 (0.80–2.64)
CCI (per point)	1.42 (1.13–1.81)
Ischemic heart disease	0.73 (0.28–1.81)
Diabetes mellitus	2.33 (0.95–5.89)
COPD	4.80 (1.80–12.74)

ASA = American Society of Anesthesiologists; BMI = body mass index; CCI = Charlson Comorbidity Index; CI = confidence interval; COPD = chronic obstructive pulmonary disease; OR = odds ratio.

Table 7.

Multivariable analysis of risk factors for 30-day mortality^*

Variable	OR (95% CI)
Age (per yr)	1.12 (0.99–1.27)
BMI (per unit)	1.19 (1.05–1.34)
CCI (per point)	1.21 (0.89–1.65)
Diabetes mellitus	2.19 (0.77–6.27)
COPD	2.66 (0.77–9.23)

BMI = body mass index; CCI = Charlson Comorbidity Index; CI = confidence interval; COPD = chronic obstructive pulmonary disease; OR = odds ratio.

^*Significant variables (p < 0.1) in univariate analysis were entered into the multivariable model.

Predictors of poor functional status

Given the small sample and event rate, multivariable analysis for functional status outcomes was not feasible. In univariate analysis, factors associated with poor functional status at last follow-up included age group (nonagenarian patients v. octogenarian patients: 34.2% v. 23.3%, p = 0.2), baseline frailty indicated by a Fried score of 3 or higher (45.2% v. 18.5%, p = 0.01), and emergency surgery (41.2% v. 22.7%, p = 0.048).

Ethics consultation

Consultation with an ethics committee was required for 15 (20.5%) nonagenarian patients and 3 (4.1%) octogenarian patients (p = 0.004). Patient willingness was documented for 62 (84.9%) nonagenarian patients and 60 (82.2%) octogenarian patients (p = 0.8).

Subgroup analysis

Among patients who underwent emergency procedures, 30-day mortality was 11 (28.9%) among 38 nonagenarian patients and 5 (16.1%) among 31 octogenarian patients (p = 0.2). Elective surgery had lower mortality rates (8/35 [22.9%] among nonagenarian patients v. 2/42 [4.8%] among octogenarian patients, p = 0.04).

Among frail patients (Fried ≥ 3), 30-day mortality was 9 (40.9%) among 22 nonagenarian patients and 1 (16.7%) among 6 octogenarian patients (p = 0.4). Nonfrail patients had lower mortality (10/51 [19.6%] nonagenarian patients v. 6/67 [9.0%] octogenarian patients, p = 0.1).

Of the procedures included in our study, right hemicolectomy had the highest mortality (3/8 [37.5%] among nonagenarian patients v. 2/8 [25.0%] among octogenarian patients), while inguinal hernia repair had the lowest (2/32 [6.3%] among nonagenarian patients v. 1/53 [1.9%] among octogenarian patients).

Detailed procedure-specific outcomes are presented in Appendix 1, Tables S1 to S4, available at www.canjsurg.ca/lookup/doi/10.1503/cjs.009525/tab-related-content.

Discussion

The surgical management of the oldest population represents one of modern surgery’s most pressing challenges. As life expectancy continues to rise globally, surgeons increasingly encounter nonagenarian patients requiring operative intervention, yet evidence guiding decision-making in this age group remains sparse. Although extensive literature supports surgical intervention in octogenarian patients, with outcomes often approaching those of younger cohorts,³⁹ the assumption that such evidence can be extrapolated to nonagenarian patients demands critical examination.

Our study of 73 propensity-matched pairs reveals a sobering reality. Nonagenarian patients who underwent general surgery had substantially worse outcomes than their octogenarian counterparts across every measured domain. The difference in 30-day mortality (26% among nonagenarian patients v. 9.6% among octogenarian patients) represents not merely a statistical difference but a fundamental divergence in surgical resilience. This gap widens progressively — reaching 49.3% and 23.3% at 90 days and 75% and 39.7% at 1 year for nonagenarian and octogenarian patients, respectively — suggesting that initial surgical stress triggers a cascade of vulnerability that extends far beyond the perioperative period, with 36% of nonagenarian patients’ deaths caused by surgical complications, compared with 16% among octogenarian patients.

Perhaps even more troubling is the functional trajectory of survivors. At last follow-up, 34.2% of nonagenarian patients had poor functional status, compared with 23.3% of octogenarian patients, with fewer than half returning to their baseline functional status. This finding challenges the traditional surgical paradigm where technical success equates to meaningful benefit. For many nonagenarian patients, survival may come at the cost of independence — a trade-off that fundamentally questions the goals of surgical intervention in this population. Notably, 52% of our cohort underwent emergency surgery when medical management had failed, and surgical intervention represented the only viable option. In these scenarios, our findings should not discourage necessary life-saving interventions but rather inform realistic prognostic discussions with patients and families about expected outcomes, including the high likelihood of functional decline even among survivors.

The contrast with established octogenarian outcomes proves particularly instructive. Several landmark studies have demonstrated the feasibility and benefit of complex procedures in octogenarian patients. Engoren and colleagues⁴⁰ reported good functional outcomes following cardiac surgery, while Chen and colleagues⁴¹ demonstrated that pancreaticoduodenectomy could be safely performed in this population, with outcomes comparable to those of younger patients. Kurazumi and colleagues⁴² showed acceptable mid-term survival following aortic arch surgery, and Tang and colleagues⁴³ documented quality-of-life improvements that justified surgical intervention. Our octogenarian cohort’s outcomes align with these findings, validating our methodology while highlighting the unique vulnerability of nonagenarian patients.

Emergency surgery amplifies the disparities between these age groups. The 28.9% mortality rate for emergency procedures among nonagenarian patients was nearly double that of octogenarian patients at 16.1%, consistent with Becher and colleagues’⁴⁴ observation that emergency surgery multiplies baseline risk by 2 to 3 in the oldest age groups. In our study, this disparity was particularly pronounced among patients who underwent right hemicolectomy, where 30-day mortality reached 37.5% in nonagenarian patients and 25.0% in octogenarian patients, while procedures like inguinal hernia repair showed more favourable but still concerning outcomes (6.3% v. 1.9%), aligning with systematic review findings that colorectal resections in patients older than 80 years have mortality rates of 0% to 5% with higher risks among nonagenarian patients.^45,46 Even elective procedures had mortality rates nearly fivefold higher among nonagenarian patients than octogenarian patients, suggesting that careful patient selection and optimal timing cannot fully mitigate age-related physiologic decline. This finding contradicts the conventional wisdom that elective surgery in well-prepared older patients carries acceptable risk.

The implications for preoperative assessment prove profound. Traditional risk stratification tools developed in younger populations fail to capture the multidimensional vulnerability of nonagenarian patients. Although ASA classification correlated with outcomes, it inadequately differentiated risk within our nonagenarian cohort, where 91.1% were classified as ASA III or higher. Similarly, the CCI, despite its widespread use, showed limited discriminatory power in this age group, where multiple comorbidities represented the norm rather than the exception.

Our analysis identified several key domains requiring systematic evaluation. Baseline functional status emerged as the strongest predictor of poor outcomes, with preexisting ADL dependence (≥ 2) affecting 35.6% of nonagenarian patients versus 14.9% of octogenarian patients. This finding aligns with Robinson and colleagues’⁴⁷ emphasis on functional assessment superseding chronological age in surgical decision-making. Frailty — assessed retrospectively using modified Fried criteria — stratified risk effectively, with a 30-day mortality of 39.1% among frail nonagenarian patients and 19.2% among nonfrail patients. These observations support calls for routine frailty screening in geriatric surgical candidates.⁴⁸

Cognitive function, often overlooked in surgical assessment, proved critically important. Mini-Cog scores below 3 predicted not only postoperative delirium, but also functional decline and failure to return home. This finding resonates with a systematic review linking preoperative cognitive impairment to adverse surgical outcomes across multiple domains.⁴⁹ The 41.1% prevalence of cognitive impairment in our nonagenarian cohort underscores the need for routine cognitive screening and tailored perioperative pathways for cognitively impaired patients.

Nutritional status emerged as a modifiable yet often neglected risk factor. Albumin levels below 3.5 g/dL were associated with increased mortality, consistent with extensive literature linking malnutrition to surgical complications.⁵⁰ However, the similar prevalence of hypoalbuminemia between nonagenarian and octogenarian patients (27.4% v. 24.8%) suggests that nutritional markers alone inadequately stratify risk between these age groups, although preoperative nutritional optimization remains important when time permits.

The concept of selective nonoperative management emerged as perhaps our study’s most important contribution. Unlike younger cohorts of older adults, among whom surgical benefits often justify risks, the combination of high mortality and persistent functional decline in nonagenarian patients demands fundamental reconsideration of treatment goals. The traditional surgical imperative — to operate when technically feasible — must yield to a more nuanced approach that prioritizes patient-centred outcomes and realistic expectations.

This paradigm shift aligns with evolving geriatric surgery principles emphasizing shared decision-making and goal-concordant care.⁵¹ For nonagenarian patients presenting with surgical pathology, the question transforms from “Can we operate?” to “Should we operate?” — recognizing that technical feasibility does not equate to patient benefit. Nonoperative management should be strongly considered for emergency presentations in nonagenarian patients with frailty, baseline functional dependence, or multiple highrisk features.

The implementation of comprehensive geriatric co-management models offers potential for optimization. Our data suggest that systematic approaches that incorporate geriatrician involvement, targeted prehabilitation, and enhanced recovery protocols might improve outcomes, although the magnitude of benefit remains uncertain in this age group. The American College of Surgeons’ Geriatric Surgery Verification Program provides a framework for such initiatives, although specific adaptations for nonagenarian patients require development.⁵²

Future research must address critical knowledge gaps. Development of nonagenarian-specific risk calculators incorporating multidimensional geriatric assessment could improve patient selection. Randomized controlled trials comparing operative and nonoperative management for specific conditions, although challenging to conduct, would provide definitive evidence. Investigation of enhanced recovery protocols tailored to nonagenarian physiology might identify modifiable factors to improve outcomes. Perhaps most importantly, qualitative research exploring nonagenarian and family perspectives on acceptable outcomes could inform shared decision-making frameworks.

The ethical dimensions of surgical decision-making with nonagenarian patients warrant explicit consideration. Our finding that 20.5% of nonagenarian patients required ethics committee involvement, compared with only 4.1% of octogenarian patients, reflects the complexity of balancing potential benefits against substantial risks in this age group. Patient willingness was documented for 84.9% of nonagenarian patients, family willingness was recorded for only 26.7% of all patients, suggesting incomplete shared decision-making processes. This discordance between high rates of patient consent and the severe outcomes observed — with 75% 1-year mortality and 34% poor functional status among survivors — highlights the critical importance of thorough preoperative counselling. Such counselling must address not just risk of death but also the high likelihood of functional decline and compromised quality of life, ensuring that consent is truly informed by realistic outcome expectations rather than optimistic hopes.

As global demographics shift toward an aging population, with nonagenarian people representing the fastest-growing age group, these questions gain urgency. By 2050, the nonagenarian population is projected to triple, making evidence-based guidelines for this cohort imperative.³ Our data suggest that extrapolating from octogenarian patients risks both nihilistic undertreatment and harmful over-treatment of nonagenarian patients.

Context is important when interpreting our findings. The background mortality rate during the ninth decade of life is around 80%, indicating substantially higher baseline risk of death in nonagenarian versus octogenarian patients. Our observed surgical mortality rates should be considered against these population-level differences. Future studies comparing surgical outcomes to matched nonsurgical cohorts would help distinguish surgery-attributable mortality from background age-related mortality.

Limitations

The retrospective design, although strengthened by propensity matching, could not eliminate unmeasured confounding. Selection bias undoubtedly influenced results, as the frailest nonagenarian patients likely never reached surgical evaluation. Moreover, we enrolled only patients who underwent surgery, not all those with surgical diagnoses. This creates selection bias that likely affects both age groups, as the frailest octogenarian patients may also have been managed nonoperatively. Additionally, patients reaching surgery may represent those for whom conservative management failed, potentially selecting for worse outcomes. A comprehensive study comparing operative and nonoperative management in all patients with surgical diagnoses would better characterize these selection effects. Our single-centre design may limit generalizability, although outcome alignment with national benchmarks provides reassurance. A critical limitation was our inability to adjust for frailty in multivariable analyses. Despite including frailty in our propensity score, substantial imbalance persisted, suggesting that nonagenarian patients selected for surgery differed fundamentally from octogenarian patients in frailty status. With only 26 deaths in our matched cohort, including both age and frailty would have violated the 10-events-per-variable rule, risking model overfitting. We prioritized age group as our primary exposure but acknowledge that residual frailty confounding may partially explain the observed disparities in mortality. This limitation underscores the challenge of disentangling chronological age from physiologic frailty in older surgical populations. Our study is underpowered, given that our power calculation indicated 200 matched pairs were needed and we achieved only 73 pairs (36% of target). This underpowering predisposes our results to statistical fragility, where small numbers of adverse outcomes may disproportionately influence findings. Several trending differences that failed to reach significance may have been clarified in an adequately powered study. The heterogeneity of surgical procedures, although reflecting real-world practice, complicates procedure-specific recommendations. Our functional outcome assessment was limited to categorizing patients as returning to baseline versus experiencing decline, without capturing potential improvements beyond baseline functional status. This may have underestimated surgical benefit in patients with severe preoperative functional limitations who could experience improvement postoperatively. For example, patients with mobility restrictions from symptomatic hernias or those with transfusion dependence from bleeding colonic pathology may have achieved better function after surgery than their compromised preoperative baseline. Future studies should include functional improvement as a distinct outcome category to fully characterize the potential benefits of surgery in this population. Moreover, our functional outcome assessment was limited to a single determination at variable follow-up intervals rather than standardized time points, preventing a true analysis of functional decline. The retrospective binary classification of functional status lacks the granularity of prospective serial assessments. The small sample limited our ability to perform robust multivariable analyses for secondary outcomes, including predictors of poor functional status.

Conclusion

Although surgical intervention in octogenarian patients showed outcomes justifying selective operative management, nonagenarian patients represent a categorically different risk profile, demanding fundamental reconsideration of surgical indications. The combination of high mortality and persistent functional decline in nonagenarian survivors challenges routine surgical intervention in this population. Comprehensive geriatric assessment may identify robust nonagenarian patients capable of tolerating surgery with acceptable outcomes, but our data support a default toward nonoperative management unless compelling evidence suggests otherwise. Perhaps, for nonagenarian patients, surgical wisdom lies not in technical capability but in recognizing when not to operate, acknowledging that the greatest service may be in helping patients navigate their final chapter with dignity rather than subjecting them to interventions unlikely to restore meaningful function. The challenge for modern surgery is developing evidence-based frameworks to ensure that treatment decisions align with realistic outcomes and patient-centred goals, recognizing that less may indeed be more for nonagenarian patients.