Abstract
Aims and background
Intra-operative hypothermia is a potentially modifiable peri-operative risk factor of uncertain clinical significance in hip fracture surgery. This study examines whether intra-operative hypothermia is associated with adverse peri-operative and mortality outcomes, and whether the nature of that association differs when hypothermia is quantified using a conventional binary threshold (core temperature <36.0 °C) vs a continuous cumulative burden metric [area under the curve (AUC) below 36.0 °C, expressed as °C·minutes].
Materials and methods
We conducted a retrospective cohort study of patients aged 55–89 years undergoing operative fixation or arthroplasty for hip fracture at a Level I Trauma Centre (2018–2025). Patients were classified by binary intra-operative hypothermia status, with 102 hypothermic patients and 102 randomly selected normothermic controls. Continuous intra-operative core temperature data were used to calculate hypothermia burden (AUC < 36.0 °C, °C·minutes). Outcomes included intra-operative blood transfusion (primary), estimated blood loss (EBL), surgical site infection (SSI), length of stay (LOS), 30 days readmission and mortality, and 1 year mortality. Associations were evaluated using logistic and linear regression models.
Results
Among 204 patients (mean age 73.4 ± 8.7 years; 58.8% female), binary hypothermia was associated with increased intra-operative transfusion [17.6 vs 5.9%; odds ratio (OR) 3.43, 95% confidence interval (CI): 1.30–9.04; p = 0.013]. Although transfused patients exhibited higher mean hypothermia burden (31.6 vs 17.5 °C·minutes), continuous modelling demonstrated no statistically significant association between hypothermia burden and transfusion (OR 1.08 per 10 °C·minutes; p = 0.097) or any secondary outcome. Hypothermia burden correlated modestly with longer operative duration and lower pre-operative haemoglobin but not with clinical endpoints.
Conclusions
An association between binary hypothermia and intra-operative transfusion was observed, but this association was not robust when hypothermia was quantified as a continuous cumulative burden. The relationship between intra-operative hypothermia and transfusion risk in hip fracture surgery therefore remains uncertain, and may reflect operative complexity or confounding rather than a direct causal effect. Studies with larger event numbers and prospective design are needed to clarify this relationship.
Clinical significance
These findings highlight a nuanced relationship between intra-operative hypothermia and transfusion risk in hip fracture surgery. The divergence between binary and continuous analyses may reflect the non-linear nature of thermal effects, limited statistical power for rare events, or confounding by operative complexity. Hypothermia remains an important peri-operative consideration for trauma surgeons and anaesthesiologists, but whether it acts as an independent modifiable driver of adverse outcomes or primarily as a marker of surgical complexity warrants further investigation. Continuous AUC-based quantification offers an alternative metric of hypothermia exposure and may improve cross-study comparability in future research.
Level of evidence
Level III, retrospective cohort study.
Keywords: Hip fracture, Hypothermia, Modelling, Transfusion
Introduction
During surgery, a patient's core body temperature typically decreases by 1–2 °C, initiating a cascade of homeostatic responses.1,2 Below 36 °C, sympathetic activation occurs, characterised by peripheral vasoconstriction, increased catecholamine release, and shivering when not suppressed by anaesthesia.2,3 These responses can translate to deleterious post-operative outcomes, as meta-analyses indicate that even mild hypothermia, a mean difference of 0.85 °C, increases blood loss by approximately 16% and transfusion risk by 22%.4–6 The related complications of intra-operative mismanagement, particularly in frail or elderly orthopaedic patients, are significant.4,5,7
The adverse effects of intra-operative hypothermia are observed both locally, through disrupted wound healing, and systemically.1,7 Retrospective analyses of patients undergoing total hip and knee arthroplasty have demonstrated that even a single intra-operative core temperature measurement below 36 °C is associated with increased blood loss and higher transfusion rates, findings corroborated by prospective studies.5,8,9 Notably, a substantial proportion of patients remain hypothermic for over an hour into the post-operative period, suggesting that intra-operative hypothermia has prolonged systemic effects.1,4,8,10
Although the adverse effects of peri-operative hypothermia are well established in elective arthroplasty, its clinical significance in orthopaedic trauma, particularly among elderly patients undergoing hip fracture fixation, remains poorly defined.5,11–13 While numerous risk factors for complications have been identified in this population, many are non-modifiable in the acute peri-operative setting.1,13–15 Beyond expediting time to surgery, few interventions have demonstrated meaningful improvement in prognosis, with 1 year mortality rates persistently exceeding 20–30% following traumatic hip fracture.16–18
Prior studies examining hypothermia in hip fracture surgery have largely defined exposure using a single temperature value or summary statistic; an approach that may obscure clinically distinct thermal patterns such as brief redistribution hypothermia vs sustained intra-operative thermal deficit.12,13,19,20 These approaches may not accurately capture the duration and temporal pattern of thermal fluctuation under anaesthesia, potentially limiting cross-study comparability and clarity regarding associations with clinical outcomes. Moreover, they raise an important question regarding the relative contribution of pre-operative, intra-operative, and post-operative thermal management to adverse outcomes. Current literature suggests that up to one-third of post-operative hip fracture fixation patients are hypothermic upon entering the recovery room despite active warming, and that post-operative hypothermia alone is associated with higher morbidity.10,20
Understanding the clinical significance of intra-operative hypothermia in hip fracture surgery is important, as this population has persistently high 1 year mortality and limited modifiable peri-operative risk factors. Whether hypothermia constitutes an independent driver of adverse outcomes, or whether conventional binary definitions consistently capture clinically meaningful thermal exposure, has not been rigorously examined in this population. This study evaluated a cohort of hip fracture patients managed with continuous high-precision core temperature monitoring. We quantified hypothermia as both a conventional binary variable (any temperature <36.0 °C) and a continuous cumulative burden metric [area under the curve (AUC) below 36.0 °C, expressed as °C·minutes], and examined associations with intra-operative and post-operative outcomes. The primary clinical question was whether intra-operative hypothermia is associated with increased transfusion requirement and other adverse outcomes, and whether the observed association is robust across modelling approaches or sensitive to how hypothermia is defined.
Materials and Methods
Study Design and Setting
This was a retrospective cohort study conducted at a single large urban academic medical centre and Level I Trauma Centre. Institutional Review Board (IRB) approval was obtained prior to data collection.
Patient Selection
Surgical and anaesthesia databases were queried using current procedural terminology (CPT) codes to identify patients aged 55–89 years who underwent operative fixation or arthroplasty for an index femoral neck, intertrochanteric, subtrochanteric, or acetabular fracture between January 1, 2018 and January 1, 2025. Current procedural terminology codes queried included: 27,244, 27,245, 27,236, 27,235, 27,125, 27,130, 27,506, and 27,507. All eligible cases were identified through chart review. Patients were initially classified based on binary intra-operative hypothermia status (any intra-operative core temperature <36.0 °C). All 102 eligible hypothermic patients meeting inclusion criteria were enrolled. To enable balanced comparison across binary exposure groups, a random sample of 102 normothermic patients was then selected from the remaining eligible population.
Demographic information and peri-operative variables such as age, sex, body mass index (BMI), pre-operative haemoglobin, duration of surgery, anaesthesia modality, fracture type, fixation method, use of intra-operative forced air warming devices (ex. Bair Hugger), antibiotic timing, and time from injury to surgery (pre-operative waiting time) were collected.
As a measure of overall pre-operative comorbidity burden, the American Society of Anaesthesiologists (ASA) Physical Status classification system was utilised as a reliable proxy of patient morbidity status and operative risk on a scale of 1–5, where 1 is normal health and 5 is moribund.21
Exclusion criteria included: Revision procedures, non-fracture arthroplasty, pathologic fractures, pregnancy, incarceration, incomplete intra-operative temperature data, or absence of high-precision core temperature monitoring (nasopharyngeal, tympanic, esophageal, or bladder probes).
Exposure
The primary exposure was intra-operative core temperature, recorded continuously using high-precision probes. We calculated the AUC for temperatures below 36 °C from the time of induction to procedure completion to quantify the degree and duration of hypothermia exposure. Because temperature measurements were recorded at consistent 1 minute intervals, minute-by-minute summation of per-minute deficits from 36.0 °C provided an accurate approximation of trapezoidal integration.22 A custom Python algorithm was developed to ensure replicable calculation (Supplementary Fig. 1). For descriptive purposes, hypothermia was also classified as a binary variable (any intra-operative temperature <36.0 °C).
Outcomes
Primary outcome: Intra-operative blood transfusion. Secondary outcomes included: Estimated blood loss (EBL); post-operative transfusion within 72 hours; deep or superficial surgical site infection (SSI); hospital length of stay (LOS); 30 days readmission; 30 days mortality; and 1 year mortality.
Statistical Analysis
Baseline characteristics were compared using Student's t-tests or χ2 tests as appropriate. Univariable logistic regression was used for binary outcomes and linear regression for continuous outcomes, with log-transformation for non-normally distributed variables. The primary continuous exposure (hypothermia AUC) was scaled per 10 °C·minutes for logistic models to improve interpretability. Spearman rank correlation was used for non-parametric association testing. Multivariable models adjusting for age, ASA class, fracture type, procedure type, surgical duration, and pre-operative haemoglobin were explored; because baseline characteristics were well balanced and null findings were consistent across unadjusted models, adjusted estimates are reported only where they materially altered conclusions. All analyses were conducted in R (version 4.4) and Stata (version 17) by a trained statistician. Two-sided p < 0.05 was considered statistically significant.
The primary analysis was powered around intra-operative transfusion, the outcome for which a signal was observed in prior binary analyses. With 204 patients and an observed transfusion rate of 11.8%, the study had adequate power to detect odds ratio (ORs) ≥ 2.5 at conventional thresholds. For rare outcomes (SSI, 30 days mortality), the study was underpowered to detect small effect sizes, and null findings for these endpoints should be interpreted as inconclusive.
Results
A total of 204 patients were included (mean age 73.4 ± 8.7 years; 58.8% female). Using the binary definition, 102 patients (50.0%) experienced intra-operative hypothermia, and 102 (50.0%) remained normothermic. Baseline demographics, injury characteristics, and pre-operative haemoglobin were similar between groups (p > 0.05). This balance was coincidental and not the result of additional matching (Table 1).
Table 1.
Patient demographics, peri-operative characteristics, and clinical outcomes stratified by binary hypothermia classification
| Variable | Overall (n = 204) | Hypothermia (n = 102) | Normothermia (n = 102) | p-Value (OR, 95% CI) |
|---|---|---|---|---|
| Patient demographics | ||||
| Age (years), mean (SD); median (IQR) | 73.4 (8.7); 73.0 (66.0–81.5) | 74.1 (8.6); 75.0 (66.0–83.0) | 72.7 (8.9); 72.0 (65.0–81.0) | 0.22 |
| Age ≥75 years, n (%) | 96 (47.1) | 52 (51.0) | 44 (43.1) | 0.26 |
| Female sex, n (%) | 120 (58.8) | 61 (59.8) | 59 (57.8) | 0.78 |
| BMI (kg/m2), mean (SD); median (IQR) | 24.6 (5.7); 23.9 (20.9–27.3) | 24.0 (5.5); 23.4 (20.2–26.2) | 25.3 (5.8); 24.4 (21.3–28.5) | 0.11 |
| Pre-operative haemoglobin (g/dL), mean (SD); median (IQR) | 10.8 (1.8); 10.9 (9.5–12.2) | 10.7 (1.9); 10.9 (9.2–12.1) | 10.9 (1.6); 10.9 (9.8–12.2) | 0.27 |
| ASA class, n (%) II/III/IV | 15 (7.4)/148 (72.5)/41 (20.1) | 8 (7.8)/70 (68.6)/24 (23.5) | 7 (6.9)/78 (76.5)/17 (16.7) | 0.43 |
| Fracture type, n (%) femoral neck/intertrochanteric/subtrochanteric/acetabular | 93 (45.6)/98 (48.0)/12 (5.9)/1 (0.5) | 51 (50.0)/43 (42.2)/7 (6.9)/1 (1.0) | 42 (41.2)/55 (53.9)/5 (4.9)/0 (0.0) | 0.30 |
| Operative characteristics | ||||
| Time to surgery (hours), mean (SD); median (IQR) | 42.0 (38.8); 30 (20–47) | 49.0 (48.8); 33 (22–54) | 35.0 (23.6); 29 (19–43) | 0.010* |
| Surgical duration (minutes), mean (SD); median (IQR) | 82.6 (39.8); 72.0 (54.5–107.0) | 87.8 (43.7); 81.0 (55.0–113.0) | 77.4 (34.9); 69.5 (54.0–97.0) | 0.061 |
| Intra-operative warming device used, n (%) | 138 (67.6) | 76 (74.5) | 62 (60.8) | 0.036* |
| Antibiotics within 1 hour of incision, n (%) | 193 (94.6) | 100 (98.0) | 93 (91.2) | 0.030* |
| Clinical outcomes | ||||
| EBL (mL), mean (SD); median (IQR) | 190.6 (152.5); 200 (100–250) | 198.1 (168.5); 200 (100–250) | 183.0 (134.9); 150 (100–250) | 0.48 |
| Intra-operative transfusion, n (%) | 24 (11.8) | 18 (17.6) | 6 (5.9) | 0.013* (3.43, 1.30–9.04) |
| Mean AUC, transfused vs non-transfused (°C·minute) | – | 31.6 vs 17.5 | – | – |
| Transfusion ≤72 hour postop, n (%) | 73 (35.8) | 37 (36.3) | 36 (35.3) | 0.88 (1.04, 0.59–1.85) |
| Hospital LOS (days), mean (SD); median (IQR) | 9.3 (8.2); 8 (5–11) | 9.5 (9.7); 8 (5–11) | 9.1 (6.3); 7 (5–11) | 0.71 |
| Deep SSI, n (%) | 5 (2.5) | 3 (2.9) | 2 (2.0) | 0.65 (1.52, 0.25–9.26) |
| Superficial SSI, n (%) | 2 (1.0) | 1 (1.0) | 1 (1.0) | 1.00 (1.00, 0.06–16.21) |
| 30 days readmission, n (%) | 35 (17.2) | 16 (15.7) | 19 (18.6) | 0.58 (0.81, 0.39–1.69) |
| 30 days mortality, n (%) | 9 (4.4) | 3 (2.9) | 6 (5.9) | 0.31 (0.48, 0.12–1.99) |
| 1 year mortality, n (%) | 21 (10.3) | 12 (11.8) | 9 (8.8) | 0.49 (1.38, 0.55–3.43) |
AUC, area under the curve; EBL, estimated blood loss; LOS, length of stay; SSI, surgical site infection. A *p < 0.05 denotes statistical significance. Continuous variables presented as mean (SD); median (IQR). Odds ratios (OR) and 95% confidence intervals (CI) for clinical outcomes presented with p-values. Mean AUC values (transfused vs non-transfused) apply to the full cohort irrespective of binary group assignment
Several peri-operative factors differed between groups. Hypothermic patients had a longer time to surgery (median 33 vs 29 hours; p = 0.010). Intra-operative warming device use was more frequent among hypothermic patients (74.5 vs 60.8%; p = 0.036). Antibiotic administration within 1 hour of incision was also more frequent in the hypothermia group (98.0 vs 91.2%; p = 0.030; Table 1).
Binary hypothermia classification was associated with a significantly higher rate of intra-operative transfusion [17.6 vs 5.9%; OR 3.43, 95% confidence interval (CI): 1.30–9.04; p = 0.013]. Among transfused patients, mean hypothermia burden was notably higher than among non-transfused patients (31.6 vs 17.5 °C·minutes), consistent with a directional dose relationship. No significant differences were observed between groups for EBL, post-operative transfusion, hospital LOS, SSI, 30 days readmission, 30 days mortality, or 1 year mortality (Table 1).
When hypothermia burden was analyzed as a continuous exposure, pre-operative haemoglobin demonstrated an inverse association with AUC (Spearman ρ = −0.15, p = 0.035; β = −3.7 °C·minutes per g/dL, p = 0.011) and surgical duration demonstrated a positive association (ρ = 0.19, p = 0.008; β = 0.19 °C·minute per minute, p = 0.004; (Table 2).
Table 2.
Associations between patient and peri-operative variables and hypothermia burden (AUC < 36 °C) as a continuous exposure
| Variable | n | Mean (SD) | Spearman ρ | Spearman p | Reg. β (SE) | Reg. p |
|---|---|---|---|---|---|---|
| Age (years) | 204 | 73.4 (8.7) | 0.11 | 0.11 | 0.44 (0.30) | 0.14 |
| BMI (kg/m2) | 204 | 24.6 (5.7) | −0.15 | 0.037* | −0.22 (0.46) | 0.62 |
| Pre-operative haemoglobin (g/dL) | 204 | 10.8 (1.8) | −0.15 | 0.035* | −3.7 (1.4) | 0.011* |
| Surgical duration (minutes) | 204 | 82.6 (39.8) | 0.19 | 0.008* | 0.19 (0.06) | 0.004* |
| EBL (mL) | 204 | 190.6 (152.5) | 0.08 | 0.28 | 0.023 (0.017) | 0.18 |
| Time to surgery (hours) | 204 | 42.0 (38.8) | 0.11 | 0.12 | 0.10 (0.07) | 0.12 |
| LOS (days) | 204 | 9.3 (8.2) | 0.05 | 0.50 | −0.11 (0.32) | 0.73 |
AUC, area under the curve (expressed as °C·minutes); EBL, estimated blood loss; LOS, length of stay; SE, standard error. A *p < 0.05 denotes statistical significance
Univariable regression analyses did not demonstrate significant associations between hypothermia burden (continuous AUC) and any peri-operative or clinical outcome (Table 3). Hypothermia burden was not significantly associated with intra-operative transfusion (OR 1.08 per 10 °C·minutes, 95% CI: 0.99–1.18; p = 0.097), EBL (β = 1.003 per AUC unit; p = 0.058), post-operative transfusion within 72 hours (OR 1.04, 95% CI: 0.96–1.12; p = 0.37), hospital LOS (β = −0.005 days per AUC unit; p = 0.73), 30 days readmission (OR 1.04, 95% CI: 0.95–1.14; p = 0.36), 30 days mortality (OR 0.97, 95% CI: 0.79–1.20; p = 0.80), or 1 year mortality (OR 1.06, 95% CI: 0.96–1.17; p = 0.24). The mean AUC elevation observed among transfused patients (31.6 vs 17.5 °C·minutes) did not translate to a significant dose–response in regression modelling, which may reflect the highly right-skewed AUC distribution and the large proportion of patients with zero AUC (Table 3).
Table 3.
Association between intra-operative hypothermia burden (AUC < 36 °C) and peri-operative and clinical outcomes
| Outcome | Model type | Effect estimate | 95% CI | p-value |
|---|---|---|---|---|
| Estimated blood loss | Linear regression (log-transformed) | β = 1.003 per 1 AUC unit | – | 0.058 |
| Intra-operative blood transfusion | Logistic regression | OR = 1.08 per 10 AUC units | 0.99–1.18 | 0.097 |
| Blood transfusion within 72 hours | Logistic regression | OR = 1.04 per 10 AUC units | 0.96–1.12 | 0.37 |
| Length of stay (days) | Linear regression | β = −0.005 per 1 AUC unit | – | 0.73 |
| 30 days readmission | Logistic regression | OR = 1.04 per 10 AUC units | 0.95–1.14 | 0.36 |
| 30 days mortality | Logistic regression | OR = 0.97 per 10 AUC units | 0.79–1.20 | 0.80 |
| 30 days readmission or mortality (composite) | Logistic regression | OR = 1.03 per 10 AUC units | 0.95–1.12 | 0.47 |
| 1 year mortality | Logistic regression | OR = 1.06 per 10 AUC units | 0.96–1.17 | 0.24 |
AUC, area under the curve (expressed as °C·minutes); CI, confidence interval; OR, odds ratio. Logistic regression outcomes reported per 10 AUC units; linear regression outcomes reported per 1 AUC unit. A *p < 0.05 denotes statistical significance
Discussion
This study examined whether intra-operative hypothermia is associated with adverse peri-operative and mortality outcomes in hip fracture surgery, and whether this association is consistent across binary and continuous exposure models. Binary classification identified a statistically significant association between hypothermia and intra-operative transfusion (OR 3.43; p = 0.013; Table 1), consistent with prior literature in this setting. However, when the same patients were analyzed using continuous AUC-based hypothermia burden, this signal attenuated and did not reach statistical significance (OR 1.08; p = 0.097), despite a numerically higher mean AUC among transfused patients (31.6 vs 17.5 °C·minutes; Table 3). No significant associations were identified between hypothermia burden and any secondary outcome.
The divergence between binary and continuous analyses may reflect several plausible explanations that cannot be distinguished with the current data. First, the relationship between hypothermia and transfusion risk may be threshold-based or non-linear, such that a binary threshold adequately captures clinically relevant exposure whereas an AUC model assumes a linear dose–response that may not hold across the full exposure range. Second, the continuous analysis may be underpowered: With 24 transfused patients, the study had limited ability to detect moderate continuous associations, and the borderline p-value for hypothermia burden (OR 1.08; p = 0.097) combined with the directionally higher AUC among transfused patients raises the possibility of a type II error rather than a true null effect. Third, the highly right-skewed and zero-inflated AUC distribution may have reduced regression efficiency. These explanations are not mutually exclusive, and the present data cannot adjudicate between them. The observed binary association should therefore not be dismissed on the basis of the continuous modelling result alone.
After controlling for patient and surgical demographics, our binary hypothermia analysis yielded results consistent with prior work examining intra-operative transfusion in hip fracture fixation, supporting the internal validity of our cohort. Charles-Lozoya et al. observed that hypothermic patients were more likely to be transfused, and that transfusion itself predicted 30 days mortality.19 Methodologically, however, multi-measurement modality designs raise a different causal structure than intra-operative AUC. Hypothermia measured at the end of surgery or in the post-anaesthesia care unit may reflect blood loss, need for transfusion, anaesthetic duration, or shock physiology, increasing the risk that hypothermia is a marker of severity rather than an independent driver.6,7,23 Our analysis of patient and potential confounding peri-operative variables with an AUC approach restricted to the intra-operative period may reduce this reverse-causation risk and help explain why observed associations attenuate when cumulative hypothermia exposure is modeled directly. In this context, the absence of a significant association between hypothermia burden and EBL, together with only a borderline dose-dependent relationship when modeled continuously (β = 1.003 per AUC unit, p = 0.058; Table 3), supports the interpretation that hypothermia burden may reflect operative complexity, with any relationship to blood loss potentially secondary rather than predictive.
The systematic review and meta-analysis by Mroczek et al. pooled studies with heterogeneous hypothermia definitions and timing of temperature measurement and concluded that peri-operative hypothermia was associated with increased 30 days mortality, a finding not replicated by the current study in either binary or continuous analysis (Tables 1 and 3).13 This discrepancy may be explained by the fact that emergency department hypothermia, as recorded by the pooled studies, may more accurately capture prehospital exposure and serve as a proxy for baseline physiologic reserve, and is supported by earlier studies.2,13 Mroczek et al. also acknowledge that the small number of eligible studies and heterogeneity in definitions and timing limit causal inference.13 Our findings therefore do not necessarily contradict the meta-analysis; rather, they affirm that the mortality signal in the existing peri-operative hypothermia literature may be driven more by pre-operative and emergency department hypothermia and systemic vulnerability than by incremental increases in cumulative intra-operative thermal deficit within the range seen under contemporary warming practices.1,2,13
Broadly, early operative management of hip fractures is associated with reduced mortality and peri-operative morbidity, with meta-analytic data demonstrating a 12% increase in serious adverse events for every 10 hours delay to surgery.24 Optimal thresholds for time to surgery remain poorly defined, in part due to confounding by indication, as sicker patients both require longer optimisation and experience higher mortality.25 The HIP-ATTACK randomised trial mitigated this bias and demonstrated reduced peri-operative complications, but no mortality difference, among patients undergoing accelerated surgery.25 Our findings also demonstrate an association between increased time to surgery and greater hypothermia burden, suggesting that peri-operative delay may contribute to thermal dysregulation in this population, however, causality cannot be inferred.22,25
Limitation(s)
This study has several important limitations. Its retrospective single-centre design precludes causal inference and may limit generalizability to institutions with differing warming protocols or anaesthetic practices. A critical statistical limitation is the risk of type II error (false negative) in the continuous analyses. With only 24 intra-operative transfusion events, the study was powered to detect ORs ≥ 2.5 using binary classification, but was likely underpowered to detect modest continuous associations, particularly given the skewed AUC distribution. The observed OR of 1.08 per 10 °C·minutes (p = 0.097) for transfusion, combined with the directionally higher AUC burden among transfused patients, is consistent with a true effect that the current sample size was insufficient to confirm rather than evidence of absence of effect. This distinction is clinically important: Non-significant findings for the continuous exposure should not be interpreted as evidence that hypothermia burden is unrelated to transfusion risk. The cohort was also underpowered for rare endpoints including SSI (3.5%) and 30 days mortality (4.4%); null findings for those endpoints should similarly be interpreted as inconclusive (Table 1). Furthermore, the AUC distribution in this cohort was highly right-skewed and zero-inflated, which may have violated linear regression assumptions and reduced the efficiency of the continuous model; alternative modelling approaches (e.g., spline regression, transformation of AUC) were not explored. Ambient operating room temperature and anaesthetic depth, both of which influence thermoregulation, were unavailable.1 Outcomes such as readmission and mortality may be underestimated if events occurred outside the health system. Finally, warming device use was more frequent in the hypothermic group, likely reflecting reactive application; without timing data, it is not possible to determine how much of the hypothermia burden preceded warming device initiation.
Conclusion
In conclusion, this retrospective cohort study of 204 hip fracture patients identified an association between binary hypothermia classification and intra-operative transfusion risk (OR 3.43; 95% CI: 1.30–9.04; p = 0.013), consistent with prior literature in this setting. However, when hypothermia was quantified as a continuous cumulative burden, this association attenuated and did not reach statistical significance (OR 1.08 per 10 °C·minutes; p = 0.097), and no significant associations were observed for any secondary outcome. These findings should be interpreted cautiously, given the limited number of outcome events and the substantial risk of type II error in the continuous analysis. The relationship between the magnitude of intra-operative hypothermia and transfusion risk in hip fracture surgery remains under characterised: The binary association is consistent with prior work and provides the foundation for continuous modelling approaches. While binary methodology is well supported in the literature, it may be sensitive to confounding by operative complexity or model specification rather than reflecting a linear dose–response. Whether hypothermia represents a modifiable driver of adverse outcomes in this population, or primarily a surrogate marker of longer and more complex operations, cannot be determined from these data alone. Hypothermia burden correlated modestly with surgical duration and pre-operative haemoglobin, which is consistent with both interpretations. Future prospective, multicentre studies with larger event numbers should evaluate hypothermia as a continuous exposure using pre-specified non-linear models and examine whether thermal management interventions translate into improved clinical outcomes in the hip fracture population.
Clinical Significance
This study identifies an association between binary hypothermia and intra-operative transfusion in hip fracture surgery that is consistent with prior literature, but demonstrates a divergence in association when hypothermia is modelled as a continuous cumulative burden. The clinical significance of this finding is uncertain. This discrepancy may indicate that brief temperature excursions below 36 °C carry less prognostic weight than sustained thermal deficits, but alternatively, it may also reflect limited statistical power given the small number of transfusion events in the continuous analysis. For trauma surgeons and anaesthesiologists managing hip fracture patients, a population with persistently high 1 year mortality and few modifiable peri-operative risk factors, these results do not provide sufficient evidence to conclude that intra-operative hypothermia is a non-modifiable or clinically unimportant exposure. Maintenance of normothermia remains a reasonable and low-risk peri-operative target. However, these findings highlight that the existing evidence base for hypothermia as an independent risk driver in this specific population remains limited, and that alternate definitions of hypothermia may elucidate different results. Future research using both continuous and threshold-based hypothermia metrics, larger event numbers, and prospective designs are needed to determine whether thermal management interventions offer meaningful clinical benefit in hip fracture patients.
Ethical Approval
This study received Institutional Review Board approval at the institution where it was conducted. This was a retrospective study that did not require patient consent. Institutional Review Board approval #32469.
Artificial Intelligence (AI) Disclosure
Artificial intelligence was not utilised for study design, data collection, data analysis, or manuscript writing.
Authors' Contributions
All the authors contributed to the final study design and final manuscript. Henry Morar, S S Ashok, and Saqib Rehman developed the initial study design. Henry Morar drafted the IRB. Henry Morar and Alex Stratton-Moore conducted data extraction. Henry Morar, Huaqing Zhao, and Kyle B Stump conducted data analysis and wrote the initial draft of the manuscript.
Acknowledgement
None.
Supplementary Materials
The supplementary Figure 1 is available on its journal website www.stlr.com
Orcid
Henry Morar https://orcid.org/0009-0006-2609-5978
Alex Stratton-Moore https://orcid.org/0009-0007-5430-2791
Kyle B Stump https://orcid.org/0009-0006-3878-5183
Huaqing Zhao https://orcid.org/0000-0002-0953-4768
S Sharon Ashok https://orcid.org/0000-0003-4060-1524
Saqib Rehman https://orcid.org/0000-0002-2504-4267
Footnotes
Source of support: Nil
Conflict of interest: Saqib Rehman receives royalties and consulting fees from Globus Medical. None of the other authors have relevant conflicts of interest.
Data Availability
Data are not publicly available and are not available on request as patient identifiers may not have been thoroughly cleared.
References
- 1.Sessler DI. Perioperative thermoregulation and heat balance. Lancet. 2016;387(10038):2655–2664. doi: 10.1016/S0140-6736(15)00981-2. [DOI] [PubMed] [Google Scholar]
- 2.Sessler DI. Thermoregulatory defense mechanisms. Crit Care Med. 2009;37(7 Suppl):S203–S210. doi: 10.1097/CCM.0b013e3181aa5568. [DOI] [PubMed] [Google Scholar]
- 3.Frank SM, Cattaneo CG, Wieneke-Brady MB, et al. Threshold for adrenomedullary activation and increased cardiac work during mild core hypothermia. Clin Sci (Lond) 2002;102(1):119–125. 11749669 [PubMed] [Google Scholar]
- 4.Pan P, Song K, Yao Y, et al. The impact of intraoperative hypothermia on blood loss and allogenic blood transfusion in total knee and hip arthroplasty: A retrospective study. Biomed Res Int. 2020;2020:1096743. doi: 10.1155/2020/1096743. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Schmied H, Kurz A, Sessler DI, et al. Mild hypothermia increases blood loss and transfusion requirements during total hip arthroplasty. Lancet. 1996;347(8997):289–292. doi: 10.1016/s0140-6736(96)90466-3. [DOI] [PubMed] [Google Scholar]
- 6.Rajagopalan S, Mascha E, Na J, et al. The effects of mild perioperative hypothermia on blood loss and transfusion requirement. Anesthesiology. 2008;108(1):71–77. doi: 10.1097/01.anes.0000296719.73450.52. [DOI] [PubMed] [Google Scholar]
- 7.Kurz A. Thermal care in the perioperative period. Best Pract Res Clin Anaesthesiol. 2008;22(1):39–62. doi: 10.1016/j.bpa.2007.10.004. [DOI] [PubMed] [Google Scholar]
- 8.Matos JR, McSwain JR, Wolf BJ, et al. Examination of intra-operative core temperature in joint arthroplasty: A single-institution prospective observational study. Int Orthop. 2018;42(11):2513–2519. doi: 10.1007/s00264-018-3967-y. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Frisch NB, Pepper AM, Rooney E, et al. Intraoperative hypothermia in total hip and knee arthroplasty. Orthopedics. 2017;40(1):56–63. doi: 10.3928/01477447-20161017-04. [DOI] [PubMed] [Google Scholar]
- 10.Kleimeyer JP, Harris AHS, Sanford J, et al. Incidence and risk factors for postoperative hypothermia after orthopaedic surgery. J Am Acad Orthop Surg. 2018;26(24):e497–e503. doi: 10.5435/JAAOS-D-16-00742. [DOI] [PubMed] [Google Scholar]
- 11.Goel R, Boissonneault A, Grissom H, et al. Impact of intraoperative hypothermia on transfusion requirements in patients with pelvic and acetabular trauma. J Orthop Trauma. 2021;35(12):632–636. doi: 10.1097/BOT.0000000000002136. [DOI] [PubMed] [Google Scholar]
- 12.Frisch NB, Pepper AM, Jildeh TR, et al. Intraoperative hypothermia during surgical fixation of hip fractures. Orthopedics. 2016;39(6):e1170–e1177. doi: 10.3928/01477447-20160811-04. [DOI] [PubMed] [Google Scholar]
- 13.Mroczek TJ, Prodromidis AD, Pearce A, et al. Perioperative hypothermia is associated with increased 30-day mortality in hip fracture patients in the United Kingdom: A systematic review and meta-analysis. J Orthop Trauma. 2022;36(7):343–348. doi: 10.1097/BOT.0000000000002332. [DOI] [PubMed] [Google Scholar]
- 14.Cao Q, Zhang L, Cai X, et al. Comparison of clinical characteristics and outcomes among different age groups in elderly patients with hip fracture surgery. PLoS One. 2025;20(10):e0333909. doi: 10.1371/journal.pone.0333909. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Lowe J, Mitchell SM, Agarwal S, et al. Traumatic hip fracture and primary elective total hip patients are not the same: A comparison of comorbidity burden, hospital course, postoperative complications, and cost of care analysis. J Orthop Trauma. 2020;34(11):583–588. doi: 10.1097/BOT.0000000000001800. [DOI] [PubMed] [Google Scholar]
- 16.Bhandari M, Swiontkowski M. Management of acute hip fracture. N Engl J Med. 2017;377(21):2053–2062. doi: 10.1056/NEJMcp1611090. [DOI] [PubMed] [Google Scholar]
- 17.Blanco JF, da Casa C, Pablos-Hernández C, et al. 30-day mortality after hip fracture surgery: Influence of postoperative factors. PLoS One. 2021;16(2):e0246963. doi: 10.1371/journal.pone.0246963. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Huette P, Abou-Arab O, Djebara AE, et al. Risk factors and mortality of patients undergoing hip fracture surgery: A one-year follow-up study. Sci Rep. 2020;10(1):9607. doi: 10.1038/s41598-020-66614-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Charles-Lozoya S, Cobos-Aguilar H, Manilla-Muñoz E, et al. Survival at 30 days in elderly patients with hip fracture surgery who were exposed to hypothermia. Medicine (Baltimore) 2021;100(39):e27339. doi: 10.1097/MD.0000000000027339. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Gurunathan U, Stonell C, Fulbrook P. Perioperative hypothermia during hip fracture surgery: An observational study. J Eval Clin Pract. 2017;23(4):762–766. doi: 10.1111/jep.12712. [DOI] [PubMed] [Google Scholar]
- 21.Hackett NJ, De Oliveira GS, Jain UK, et al. ASA class is a reliable independent predictor of medical complications and mortality following surgery. Int J Surg. 2015;18:184–190. doi: 10.1016/j.ijsu.2015.04.079. [DOI] [PubMed] [Google Scholar]
- 22.Sessler DI, Pei L, Li K, et al. Aggressive intraoperative warming vs routine thermal management during non-cardiac surgery (PROTECT): A multicentre, parallel group, superiority trial. Lancet. 2022;399(10337):1799–1808. doi: 10.1016/S0140-6736(22)00560-8. [DOI] [PubMed] [Google Scholar]
- 23.Melling AC, Ali B, Scott EM, et al. Effects of preoperative warming on the incidence of wound infection after clean surgery: A randomised controlled trial. Lancet. 2001;358(9285):876–880. doi: 10.1016/S0140-6736(01)06071-8. [DOI] [PubMed] [Google Scholar]
- 24.Kelly-Pettersson P, Samuelsson B, Muren O, et al. Waiting time to surgery is correlated with an increased risk of serious adverse events during hospital stay in patients with hip-fracture: A cohort study. Int J Nurs Stud. 2017;69:91–97. doi: 10.1016/j.ijnurstu.2017.02.003. [DOI] [PubMed] [Google Scholar]
- 25.Borgers A, Cuppens K, Janssen P, et al. A prospective comparative study between a cooling device and manual cooling after total knee arthroplasty. Acta Orthop Belg. 2020;86(2):287–293. 33418620 [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Data Availability Statement
Data are not publicly available and are not available on request as patient identifiers may not have been thoroughly cleared.