Postoperative Dehydration Is Associated with Frailty and Decreased Survival in Older Patients with Hip Fracture

Background: Hyperosmolar dehydration (HD) is a risk factor for severe complications in hip fracture in older patients. However, evidence for recommending screening of dehydration is insufficient and its relation with frailty and mortality is unclear. We tested the hypothesis that postoperative HD is associated with frailty and increased mortality. Methods: We recruited 625 older (>65 years) patients surgically treated for hip fracture and co-managed by an orthogeriatric team over one year in 2017. Pre- and postoperative HD (serum osmolarity > 300 mmol/L) was diagnosed. Frailty and associated mortality risk were assessed by the Multidimensional Prognostic Index (MPI). Results: The prevalence of preoperative HD was 20.4%. Compared with no-HD, MPI was similar in HD patients despite higher (p < 0.05) prevalence of polypharmacy, arterial hypertension, diabetes, chronic kidney disease and heart failure. After surgery the incidence of HD decreased to 16.5%, but increased (p = 0.003) in the MPI high-risk subgroup. Postoperative HD was associated with more complications and was an independent determinant of adjusted hospital length of stay (LOS) and of 60- to 365-days mortality. Conclusions: Older frail patients with hip fracture are prone to developing postoperative HD, which independently predicts prolonged hospital LOS and mortality. Systematically screening older patients for frailty and dehydration is advisable to customize hydration management in high-risk individuals.


Introduction
Preoperative dehydration is a well-recognized, potentially modifiable prognostic factor of in-hospital complications, hospital readmissions and of poor functional outcomes in older patients with hip fracture [1][2][3]. Previous studies have suggested that up to 50% of patients show signs of dehydration before and after hip fracture surgery [4][5][6]. Multiple factors potentially contribute to this condition, including increased urea production due to fracture-related bleeding and pre-renal azotemia secondary to insufficient fluid intake [7], as well as-most importantly-impaired renal capacity to preserve water and to govern the hemodynamic balance of fluids, a derangement typically associated with hemodynamic frailty [8].
In addition to dehydration, several modifiable and non-modifiable risk factors influence morbidity, mortality and functional recovery after hip fracture [2,3,9,10], which include surgical variables, as well as pre-fracture patient demographics, social, functional and medical conditions. The combination of the latter factors is best described by the frailty syndrome, which is notably characterized by a decline in multiple domains and by the presence of comorbidities and polypharmacy [11]. By ranking hip fracture patients according to frailty, it has been demonstrated that about 30% of them are frail [12] and they are predisposed to adverse clinical outcomes, including postoperative mortality, complications and prolonged length of hospital stay LOS [12,13]. Therefore, a standardized approach based on perioperative risk evaluation and treatment optimization, as in the integrated orthogeriatric care management, is mandatory, since it results in improved short-and long-term clinical outcomes, including morbidity, LOS, functional recovery and mortality [14][15][16][17].
Although clinical frailty scoring tools usually explore several domains, it is not clear whether in older patients with hip fracture the occurrence of pre-and postoperative dehydration can be predicted on the basis of the clinical frailty assessment performed at admission. In addition, the relative roles of pre-and postoperative dehydration on the burden of morbidity and mortality in these patients has been scarcely explored. Based on available literature we may hypothesize that frailer subjects are more prone to dehydration, which contributes to adverse clinical outcomes.
Therefore, the aims of this study were: (1) to describe the prevalence of postoperative dehydration as compared to the corresponding preoperative value in a cohort of patients acutely admitted for hip fracture; (2) to assess its associations with frailty and (3) its relationships with postoperative complications, length of hospital stay and short-mediumand long-term mortality.

Study Design, Setting and Population
A retrospective observational study was conducted in the University Hospital of Trieste, Italy. All consecutive patients aged ≥ 65 years admitted because of a femur fracture from 1 January to 31 December 2017 and in the charge of the orthogeriatric team (in which patients were co-managed by orthopedics and geriatricians) were considered for eligibility. Patients undergoing conservative treatment (no surgery) and those with incomplete laboratory data to compute preoperative serum osmolarity were excluded.

Study Variables
Clinical, functional and serum biochemistry data assessed during the hospital stay were collected from the patients' electronic medical records. Sociodemographic variables were also collected and two age categories (≤85 and >85 years) were created. Moreover, the chronic use of medications potentially affecting the hydration status was documented.
The Krahn and Khajuria equation (osmolarity = 1.86 × (Na + + K + ) + 1.15 × glucose + urea + 14; each component measured in mmol/L) [18], was used to calculate serum osmolarity. This equation has been validated in very different populations of older adults [19] and has demonstrated to be an accurate and objective diagnostic tool to assess hyperosmolar dehydration in older hospitalized medical patients [20]. Patients with calculated osmolarity of > 300 mmol/L were considered as having a hyperosmolar dehydration (HD) [21].
The eGFR was calculated through the equation developed by Levey et al. [22]. Severity of frailty was scored according to the Multidimensional Prognostic Index (MPI), an assessment tool including information on functional, nutritional, cognitive and social status, polypharmacy and comorbidities, which has been developed for hospitalized older subjects and has proved to predict short-and long-term adverse clinical outcomes including mortality [11,23,24]. The MPI was computed upon hospital admission based on the routinely collected Comprehensive Geriatric Assessment (CGA), establishing three risk levels: low (0.0-0.33); moderate (0.34-0.66); and severe (0.67-1.00) [24].
Both osmolarity and eGFR were computed pre-(the day before surgery) and postoperatively (the day after surgery), while GPS was available only in the preoperative phase since serum albumin was determined just once, upon patient admission.

Study End-Points
The primary study endpoint was all-cause mortality-defined as death from any cause within 12-months from hospital admission. Date of death was obtained from the electronic mortality register. Moreover, the possible independent association between postoperative HD and total hospital length of stay (LOS) was explored.
As a secondary endpoint, the association of postoperative HD with the occurrence of the following postoperative complications (defined according to combinations of clinical, laboratory and instrumental data) was documented: acute respiratory failure, pneumonia, delirium, blood glucose imbalance (i.e., occurrence of hyperglycemia) or electrolyte imbalance (i.e., occurrence of hyper-hyponatremia or hyper-hypokalemia), pressure ulcers, urinary tract infections, heart failure exacerbation, surgical wound infection and total number of transfused blood units. Finally, the association between the level of frailty and the modification of serum osmolarity between pre-and postoperative stages was explored.

Data Analysis
Since we recruited a convenience sample including all consecutive patients over one year, no formal sample size was calculated a priori. The number of deaths observed in the study population was sufficient to satisfy the pragmatic thumb rule of 1:10 predictors to events ratio in the most multivariable models.
Data distribution was evaluated using the Kolmogorov-Smirnov test. The continuous variables were presented as medians and interquartile ranges (IQRs). Unadjusted comparisons between the groups were analyzed via nonparametric Mann-Whitney's U test for independent samples or Wilcoxon test for paired samples, as appropriate. Nominal variables were shown as a number and percentage and were analyzed using contingency tables and χ test or Fisher's exact test, as appropriate.
The independent association between a postoperative HD condition and the patients' hospital LOS was investigated by forward stepwise multiple linear regression model, after controlling for study variables associated to the hospital LOS with a p-value of <0.1. Since the hospital LOS data had a skewed distribution, logarithmic transformation was performed to obtain a more approximately normal variable. Results were reported as β and p values.
The different time of death according to the occurrence of postoperative HD was explored. Observations were right-censored until 30, 60, 90, 180 or 365 days after hospital admission (i.e., known survival). Both unadjusted and adjusted analyses were performed. Crude evaluation was carried out by comparing Kaplan-Meier curves and differences in survival rates between groups were assessed with Log-rank Mantel-Cox test and presented with crude Kaplan-Meier curves. Adjusted comparisons were performed by fitting multivariable Cox proportional-hazards models with forward stepwise selection, adjusting for confounders potentially affecting serum osmolarity (i.e., eGFR) or known to be related to mortality in older acute inpatients. The results were presented as an adjusted proportional hazard ratio (HR) and 95% confidence intervals (CI); cumulative hazard adjusted curves were drawn, as well.
All statistical analyses were performed using the software IBM SPSS Statistics, version 24.0 (IBM Corp., Armonk, NY, USA) For all tests, an alpha level of p ≤ 0.05 was set for statistical significance.

Results
During the study period, 625 older patients were admitted to the hospital and managed by the orthogeriatric team because of a hip fracture. After excluding subjects not undergoing surgery and/or having insufficient data to compute preoperative serum osmolarity, 599 patients were included in the study population ( Figure 1). version 24.0 (IBM Corp., Armonk, NY, US) For all tests, an alpha level of p ≤0 for statistical significance.

Results
During the study period, 625 older patients were admitted to the ho managed by the orthogeriatric team because of a hip fracture. After excluding s undergoing surgery and/or having insufficient data to compute preopera osmolarity, 599 patients were included in the study population ( Figure 1).  Table 1 shows the main characteristics of the patients. They were mos (74.8%), of median age 84 years old. Median MPI value was 0.50 with 23.5%, 22.7% of patients in the low, moderate and severe risk groups, respectively represented comorbidity was arterial hypertension. Just under half of the pa on polypharmacy with the most prescribed drugs represented by anti-hyp neuroleptics and benzodiazepines, and diuretics. Among patients chronically diuretics, the most frequently prescribed medications were furosemide (n = 12 alone (n = 92, 72.4%) or associated with other diuretics (spironolactone: 27, 2 diuretics: 8, 6.3%), and hydrochlorothiazide (alone: 48, 25.3%; with amiloride Among laboratory parameters, the median-calculated serum osmolarity was 2 with a prevalence of preoperative hyperosmolar dehydration (HD) of 20.4% (n  Table 1 shows the main characteristics of the patients. They were mostly females (74.8%), of median age 84 years old. Median MPI value was 0.50 with 23.5%, 53.3% and 22.7% of patients in the low, moderate and severe risk groups, respectively. The most represented comorbidity was arterial hypertension. Just under half of the patients were on polypharmacy with the most prescribed drugs represented by anti-hypertensives, neuroleptics and benzodiazepines, and diuretics. Among patients chronically assuming diuretics, the most frequently prescribed medications were furosemide (n = 127, 66.8%)-alone (n = 92, 72.4%) or associated with other diuretics (spironolactone: 27, 21.3%; other diuretics: 8, 6.3%), and hydrochlorothiazide (alone: 48, 25.3%; with amiloride: 10, 5.3%). Among laboratory parameters, the median-calculated serum osmolarity was 295 mmol/L with a prevalence of preoperative hyperosmolar dehydration (HD) of 20.4% (n = 122).
Compared with the no-HD subgroup, HD patients showed a higher (p < 0.05) BMI, a greater prevalence of arterial hypertension, chronic kidney disease and heart failure and diabetes, with an increased percentage of polypharmacy and a higher rate of home therapy with diuretics. Moreover, HD patients presented preoperatively with poorer eGFR, lower haemoglobin levels and higher concentrations of serum urea, sodium, potassium and glucose exposing to a HD condition as represented by higher serum osmolarity. Conversely, the two subgroups did not differ in age, sex, living condition and MPI, as well as in the other considered comorbid and laboratory variables. Patients underwent surgery after a median time of 2.0 (IQR 1.0-3.0) days from admission. Data to compute postoperative osmolarity were available for 595 patients (99.3%). Overall, a statistically significant reduction in serum osmolarity was found after surgery (pre: 295.0, IQR 290.5-299.3 mmol/L; post: 293.4, IQR 288.8-297.7 mmol/L; p < 0.001), the incidence of postoperative HD being 16.5% (n = 99). The incidence of postoperative HD was higher in patients with chronic heart failure (no-HD: n = 63, 12.7%; HD: n = 30, 30.3%; p < 0.001), hypertensive (no-HD: n = 94, 19.0%; HD: n = 29, 29.3%; p = 0.020) or ischemic heart disease (no-HD: n = 83, 16.7%; HD: n = 29, 29.3%; p = 0.004), and chronic kidney disease (no-HD: n = 42, 8.5%; HD: n = 32, 32.3%; p < 0.001). Table 2 describes the comparisons between patients presenting or not a postoperative HD. No between-group difference was found according to the type of surgery. A statistically significant higher HD incidence was documented among patients who underwent surgery after two or more days from hospital admission. Overall, 259 patients (43.2%) underwent blood transfusions; a more frequent need for transfusions was documented among patients with postoperative HD. When considering the occurrence of complications, no patient presented a surgical wound infection. Individuals with postoperative HD showed a higher incidence of pneumonia, electrolyte imbalance, pressure ulcers and exacerbation of heart failure, while only a non-significant trend was documented toward a higher incidence of respiratory failure. No further statistically significant difference was documented when considering the association of postoperative HD with the other complications. After stratifying the study population according to MPI risk categories, in postoperative stage the HD prevalence decreased in low-and moderate-risk groups, and it increased among patients scored as MPI high-risk (Figure 2), with a statistically significant difference after surgery (p = 0.003).  Postoperative HD was a statistically significant independent determinant of hospital LOS in multiple linear regression analysis adjusted for sex, MPI, GPS and postoperative eGFR ( Table 3). All-cause mortality data within 12 months from hospital admission were available in 99.5% of cases (n = 596). The cumulative mortality rate was 4.0%, 8.1%, 9.1%, 13.3%, and 20.0% at 30, 60, 90, 180, and 365 days, respectively. Results of unadjusted and adjusted survival analyses are reported in Table 4; crude and adjusted 365 days survival curves are shown in Figure 3. In unadjusted analyses, the crude risk of death in all considered time intervals for patients with postoperative HD was significantly higher than that of those without this condition (Table 4).   All variables considered to adjust the regression models (i.e., age > 85, male sex, MPIsevere risk, and GPS-poor prognosis) showed a statistically significant association (p < 0.05) with one-year mortality, except for eGFR (p = 0.089). In multivariable analyses, patients with postoperative HD showed a statistically significant increased risk of death at 60-to 365 days, while the same predictive power was not confirmed for 30 days mortality. All variables considered to adjust the regression models (i.e., age > 85, male sex, MPIsevere risk, and GPS-poor prognosis) showed a statistically significant association (p < 0.05) with one-year mortality, except for eGFR (p = 0.089). In multivariable analyses, patients with postoperative HD showed a statistically significant increased risk of death at 60-to 365 days, while the same predictive power was not confirmed for 30 days mortality.

Discussion
In this retrospective observational study including a large cohort of older patients who underwent surgery for hip fracture we found that: (1) preoperative dehydration is associated with comorbidities and with polypharmacy but not with frailty; (2) in the postoperative phase the prevalence of dehydration is reduced in patients at low-and moderate-risk MPI categories, while it markedly increases in the high-risk MPI subgroup; (3) postoperative dehydration predisposes the patients to complications including pneumonia and acute heart failure; and (4) it is independently associated with increased length of hospital stay and with mortality in the medium-and long-term.
Screening for dehydration in the perioperative phase is not routine clinical practice. The issue is relevant, because 30-50% of older hip fracture patients preoperatively are affected by this condition [5,6] which has a direct negative impact on postoperative complications [6] and on rehabilitation [2,3]. In contrast, an adequate fluid management targeted at achieving hemodynamic stability results in a significant reduction of morbidity and mortality rates [26]. The diagnosis of dehydration is challenging, because most of clinical signs and laboratory tests are notably biased by low diagnostic accuracy [27]; however, at present plasma osmolarity is considered the most valuable parameter to assess hydration status and is currently recommended to identify dehydration in older patients [20,21]. By using serum osmolarity we found that 20.4% of study patients were dehydrated in the preoperative phase. This finding is in line with that reported by Loffel et al. [28] who used a composite score based on urine and blood tests to diagnose dehydration in older patients undergoing urological surgery while another study conducted a in hip fracture cohort reported a higher prevalence [6]. In addition, consistent with available literature, these findings demonstrate that some comorbidities, including renal failure and diabetes, are associated with dehydration in older people as well as polypharmacy and the use of diuretics [29,30]. In this study, preoperative dehydration was not related to frailty measured by the Multidimensional Prognostic Index which includes information on nutritional, functional, cognitive and social status as well as on polypharmacy and comorbidities [24]. This is in line with previous data from our group in a different cohort of older, acutely hospitalized patients, showing an independent contribution of dehydration and frailty to increased mortality [31].
In contrast, we found that frailty was directly associated with the occurrence of postoperative dehydration, suggesting that this condition-although not evident in the preoperative phase-may quickly develop postoperatively in MPI high-risk older subjects. In these patients, hip fracture surgery may represent an extrinsic stressor unsettling fluid balance and patenting hemodynamic frailty, which determines dehydration. Indeed, the incidence of postoperative acute heart failure exacerbation, a complication possibly associated with hemodynamic frailty, was higher in patients with HD as compared to those with no-HD. These findings are relevant, since most studies analyzing the impact of dehydration on morbidity, mortality and functional outcomes in hip fracture patients have been conducted using the preoperative data. One small study including 38 patients and using a combination of clinical parameters and serum osmolality to diagnose postoperative dehydration reported a higher prevalence (40%), however the relationships with clinical outcomes were not explored [4]. Similar data were reported in another study from the same group including 43 patients in which postoperative HD was diagnosed using a composite score combining urine and plasma parameters [5]. Regardless these discrepancies on the prevalence of postoperative HD which may be influenced by the protocols adopted for hemodynamic management of fluids in the perioperative period, our results underpin the concept that standard approaches to optimize hydration status during hip fracture surgery, that are highly effective in older fit patients, may be inadequate for frail subjects. Importantly, it should be noted that in the present investigation clinicians were unaware about patients' serum osmolarity, which was instead calculated retrospectively for research aims. We might speculate that the integration of serum osmolarity in the patient's preoperative assessment could have driven interventions aimed at better correcting preoperative HD and preventing its postoperative persistence or onset.
Postoperative HD was also an independent risk factor for increased hospital LOS along with severity of disease as defined by the Glasgow prognostic score. Based on previous literature [24,31,32], the selected cut offs for moderately-to-severely decreased renal function, GPS poor prognosis and MPI severe risk were, respectively, an eGFR <60 mL/min/1.73 m 2 , a GPS score of 2 and a MPI value of 0.67-1.00. These cut offs have been shown to correlate with adverse clinical outcomes, included prolonged LOS and in-creased mortality. Conversely, a more severe frailty level as described by MPI was predictive of a significantly shorter hospital LOS. Intuitively, while acute conditions have a direct impact on longer hospital LOS, uncomplicated frailty may not require prolonged hospitalization.
Finally, for the first time to our knowledge, we documented the effect of postoperative HD on short-, medium-and long-term mortality. Our findings demonstrate that all-cause mortality progressively increased over time up to 20% at one year in older hip fracture patients with postoperative HD, being significantly higher than that in patients without HD at 60, 90, 180 and 365 days. The effect of dehydration on mortality was independent of other factors which showed a correlation with mortality in bivariate analysis and were included in the final model, namely frailty, severity of disease and male sex. In contrast, mortality was not associated with postoperative HD at 30 days. This finding is in line with that reported by Hahn in a smaller group of patients [5].
We should however acknowledge that the study has some limitations, mainly related to the retrospective design, that should be considered when interpreting the results. First, we enrolled a convenience sample of 599 patients corresponding to the population admitted to the study setting because of a femur fracture over one year ago. Considering the one-year mortality in normo-and dehydrated subgroups, a post-hoc analysis revealed that a sample size of 680 subjects would have been needed to ensure an 80% power with an α-error of 0.05. Second, only a limited number of confounders were considered to adjust multivariable regression models because of the relatively low number of events observed within this study period: other variables not considered in this study could have affected the study outcomes. Moreover, a certain number of data to calculate Glasgow Prognostic Score were missing; the availability of data for the entire population could have led to different results. In addition, measured serum osmolarity was not available for many patients and for this reason calculated osmolarity was used. However, a direct measurement should be warranted. Finally, this study was conducted in a single center, which implies that ward-specific factors (e.g., staffing, nursing skill mix, organizational issues) may have influenced patient outcomes. A multicenter approach would be necessary to improve generalizability of our results.
Nonetheless, some findings of this study are worth mentioning because of their clinical implications. First, frail patients who are not dehydrated in the preoperative phase may develop dehydration postoperatively; therefore, older subjects scoring at high risk for mortality due to frailty should be strictly monitored even though dehydration is not apparent at admission. Second, the use of standard protocols to optimize the hydration status in the perioperative phase may be less efficient in frail patients, especially in the presence of heart or kidney failure, which impose limits on the possibility of expanding the circulating volume. Third, since postoperative dehydration is a marker of increased morbidity, hospital LOS and mortality, development of individualized approaches to hemodynamic handling in high-risk patients should be considered. The mortality benefit of a precision medicine approach to prevent undesired outcomes in these patients warrants further investigation.

Conclusions
Older frail patients with hip fracture are prone to developing postoperative hyperosmolar dehydration after surgery, especially when affected by chronic heart or kidney failure. The frailest subjects are particularly exposed to the risk of getting postoperative dehydration, which is associated with an increased likelihood of developing complications and independently predicts prolonged hospital LOS and mid-to long-term mortality. These findings strongly support the concept of systematically screening older patients for frailty and dehydration, and of developing individualized protocols to manage hydration status in patients at high risk.

Institutional Review Board Statement:
This research is part of the "MPI-Orthogeriatrics: predicting adverse clinical outcomes in fragile elderly with hip fracture" retrospective study, approved by the Regional Bioethics Committee of Friuli Venezia Giulia, Italy (protocol number: 087_2021H; date of approval: 23 November 2021). The research was conducted according to the Declaration of Helsinki.

Informed Consent Statement:
At hospital admission, all enrolled patients authorized the use of their clinical data for study purposes.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to privacy and ethical reasons.

Conflicts of Interest:
The authors declare no conflict of interest.