Introduction

Hip fractures are regarded as the most severe of osteoporotic fractures [1], and are among the most frequent reasons of admission to hospital and nursing facilities for elderly in western countries [2]. In a recent review, the total number of hip fractures in 18 countries was 983,000, with a projection of 1.87 million in 2050 [3]. Reports suggest that hip fracture leads to increased 1-year mortality [3] and partial disability [4, 5], leading to high health care and social costs [6].

Hip fracture patients often have a poor nutritional status [7, 8], with prevalence figures of malnutrition ranging from 7 to 26% depending on the evaluation tool used [7]. Previous reports indicate that malnutrition can be associated with impaired clinical outcome, lower mobility, functional disability and loss of independence [9,10,11], higher postoperative complication rate [11], higher hospital readmission rate [9], prolonged rehabilitation time [12], and increased mortality [9, 10]. However, many publications were not adjusted for confounders such as age, sex, American Society of Anesthesiologists (ASA) score and comorbidities. This limitation is especially relevant because hip fracture patients will only profit from nutritional intervention as far as their nutritional status is a true determinant of recovery.

We previously performed an open-label randomized clinical trial on the effect of intensive nutritional intervention in elderly after hip fracture [13]. The three-month intervention comprised oral nutritional support and a diet enriched in energy and protein, with close regular supervision by trained dieticians. Although this intensive nutritional intervention had a favourable short-term effect on nutritional intake and body weight, no effect was detected on any clinical or functional outcomes, including length of stay in hospital and rehabilitation clinics and 5-year mortality [13].

The database from this study contained extensive information on diet, demographic variables, comorbidities, mobility, cognition, anxiety, depression and health-related quality of life. This well-defined prospective dataset allowed us to perform a detailed post-hoc study on the prognostic value of nutritional status at baseline for surgical complications, length-of-stay in hospital and rehabilitation, hospital readmissions, functional recovery, new fractures and mortality. We hypothesized that nutritional status would be an independent prognostic factor for clinical and functional outcome after hip fracture surgery.

Methods

Study population and design

Study population and design have been described extensively elsewhere [13, 14]; a flow chart of screening is shown in Fig. 1. Eligible were individuals ≥55y admitted for surgical treatment of a fracture of the proximal femur (medial, lateral collum, intertrochanteric), based on a daily inventory of patients admitted to the surgical and orthopedic wards of three hospitals in the region of South-Limburg. Patients were enrolled within 5 days after surgical treatment of hip fracture. Exclusion criteria were: pathological or periprosthetic fractures, diseases of bone metabolism other than osteoporosis (e.g. M. Paget, primary/secondary bone tumors, hyperparathyroidism, M. Kahler), life expectancy <1y due to underlying disease (e.g. cancer), receiving oral nutritional support on medical prescription before hospital admission, inability to speak Dutch, living outside the region of Zuid-Limburg, being bedridden before the hip fracture, dementia or cognitive impairment (score < 7 on Abbreviated Mental Test), or unavailable for follow-up. Of 152 enrolled patients, 73 were randomly allocated to the intervention group and 79 patients to the control group [13]. Of the 152 patients enrolled, ten patients discontinued participation and 11 patients died during the 6-month study period, leaving 131 (63 intervention, 68 control) for analysis of functional outcome at 6 months in the present study. For length of stay, complications, new fractures and survival, data of all 152 patients were available for analysis.

Fig. 1
figure 1

Flow chart

Full size image

Baseline measurements were performed immediately after enrolment. Outcome assessment was performed at the patient’s home at 3 and 6 months following hip fracture. Details on the intervention have been described elsewhere (Wyers et al. 2018) Briefly, patients randomized to the intervention group received dietetic counseling and oral nutritional supplementation of Cubitan (N.V. Nutricia, Zoetermeer, The Netherlands). Cubitan is a high-protein oral nutritional supplement (ONS) containing 124 kcal and 8.8 g protein per 100 ml. Intervention patients received an amount of Cubitan designed to meet their individual protein requirements (max. 400 ml/d). When a patient’s food intake was sufficient to meet individual energy and protein requirements, the ONS was discontinued. The mean energy and protein intake from Cubitan in the intervention group amounted to 215 kcal/d and 21 g/d of protein at 1 week of intervention, and 97 kcal/d and 3.6 g/d of protein at 3 months (i.e. end of intervention). The control group received usual nutritional care as provided in the hospital, rehabilitation clinic or at home, i.e. dietetic advice and oral nutritional supplements only if prescribed by the medical doctor in charge. According to standard care after hip fracture in The Netherlands, all patients in both groups received physical and exercise therapy daily during hospitalization and during rehabilitation; patients discharged to their home received physiotherapy and exercise therapy aimed at full functionality in activities of daily living (ADL) and instrumental activities in daily living (IADL, such as household activities) for six months post-fracture, or longer if prescribed by the treating clinician.

The study was approved by the Medical Ethical Committee of Maastricht University Hospital and Maastricht University (06–3-098) and conducted according to the Declaration of Helsinki; it was registered prior to starting enrolment at www.clinicaltrial.gov as NCT00523575. All subjects gave written informed consent.

Prognostic factors: nutritional status and confounders

Nutritional status at baseline was assessed by the 18-item Mini Nutritional Assessment (MNA). The MNA is a validated rapid assessment tool to evaluate the risk of protein-energy malnutrition in frail elderly patients in clinics, hospitals, and nursing homes, in order to permit early nutritional intervention when needed. The MNA is composed of 18 simple measurements and brief questions on (1) weight, height, and weight loss; (2) lifestyle, medication and mobility (6 questions); (3) number of meals, food and fluid intake, and autonomy of feeding (8 questions); (4) self-perception of health and nutrition. The MNA was developed and validated in populations of elderly representing the whole spectrum from very active and healthy, very frail housebound, to individuals institutionalized for dementia. For validation, results of the MNA were compared with complete nutritional assessment including anthropometry, biochemical markers (albumin, C-reactive protein, prealbumin, alpha1-acid glycoprotein), functional status (grip strength) dietary intake and functional geriatric assessment. The sum of the MNA score distinguishes elderly patients with adequate nutritional status (MNA ≥ 24); at risk for malnutrition, MNA between 17 and 23.5; and protein-calorie undernutrition, MNA < 17) [15].

Body weight and height at baseline were patient-reported as patients were bedridden. Body mass index (BMI) was calculated as weight/height2 (kg/m2), and handgrip strength was measured by a Jamar hydraulic dynamometer (Saehan Corp., Masan, Korea). Mid-upper arm muscle area was calculated according to Frisancho [16]. Fat mass and fat-free mass were calculated from body weight and the sum of four skinfolds (triceps, subscapular, biceps, supra-iliacal) [17].

Demographic data and medical history were derived from hospital records and included age, sex, comorbidities, type of hip fracture, surgery and anaesthesia, American Society of Anesthesiologists (ASA) score, assessing the patient’s physical status before surgery. The ASA score classification is divided into five levels: (1) normally healthy; (2) mild systemic disease; (3) severe systemic disease that limits activity, but is not incapacitating; (4) incapacitating systemic disease that is a constant threat to life; (5) moribund, not expected to survive 24 h with or without treatment. Comorbidities were grouped according to the International Classification of Diseases (ICD-10).

Outcome assessment

Functional outcomes (including cognition, fatigue and health-related quality of life) were assessed by validated questionnaires. Baseline measurements referred to the situation just preceding the fracture, except for the Mini Mental State Examination (MMSE) which assessed cognitive state at the time of baseline assessment (i.e. ~ 3 days after fracture). Higher scores indicate better functional status for the MMSE [18] and the EuroQoL EQ-5D visual analogue scale (VAS) [19], and higher scores represent worse condition for the Groningen Activity Restriction Scale (GARS) [20], the five EQ-5D-3L subscales of the EuroQoL [19], the Hospital Anxiety and Depression Scale (HADS) [21], and the Checklist Individual Strength (CIS) measuring fatigue [22]. The CIS state was not administered at baseline as retrospective assessment of fatigue was considered invalid.

Information on surgery (duration, blood loss) and postoperative complications was obtained from hospital medical records. Length-of-stay in hospital and rehabilitation clinics as well as hospital readmissions were assessed until six months postoperatively. Hospital length of stay was calculated as primary hospital stay and as the sum of primary stay and all hospital readmissions. Date and type of new fractures at any skeletal site were extracted from both hospital records and general practitioners’ records to safeguard completeness of data. Survival status at 1y and 5y and date of decease were retrieved from hospital records and checked against general practitioners’ records and the Dutch Basic Registry of Persons, a database covering all inhabitants of the Netherlands.

Statistical analysis

Baseline values of relevant determinants and outcomes are reported by descriptive statistics, applying Student’s t test for continuous variables, Fisher’s exact test for categorical variables with 2 categories, and chi-square test for categorical variables with > 2 categories. For prognostic analyses, we first calculated Pearson correlation coefficients between all potentially prognostic variables and outcomes. The following potential determinants were included in these univariable correlation analyses: age, sex, nutritional status (MNA), type of housing (see Table 1), ASA score, extensive medical history using ICD scoring (see summary in Table 1), number of fractures in the past, history of orthopedic surgery, type of hip fracture, type of surgery, type of anaesthesia; hemoglobin, hematocrit, albumin, prealbumin, C-reactive protein, and (to check for potential bias by study design): number of days between hip fracture and surgery, number of days between surgery and baseline assessment. As a second step, all prognostic factors with a statistically significant univariate correlation with the outcome of interest were included as covariates in backward regression models to select, in addition to nutritional status, any other significant potential confounders. Finally, all prognostic factors with significant p-values in these stepwise regression models were included in a final multivariable model for each outcome of interest, including in addition the following ‘compulsory variables’: age, sex, ASA score and nutritional status (MNA), and a standard dummy variable for treatment group to ensure that all presented results in this paper refer to the ‘control’ situation (of note, neither treatment group nor a tentatively included interaction term between nutritional status and treatment group was statistically significant in any of the regression models).

Table 1 Baseline characteristics by nutritional status as determined by Mini Nutritional Assessment (MNA)
Full size table

Nutritional status was operationalized using the usual cut-off point of ≤23.5 on the MNA for subjects at risk of malnutrition or malnourished (together named “subjects with impaired nutritional status”) vs. ≥24.0 for well-nourished subjects. Where relevant for in-depth analyses and graphical presentation, we split up the former group into malnourished subjects (MNA-score < 17) and subjects at risk (17 ≤ MNA-score ≤ 23.5) [23]. Linear regression models were fitted for continuous outcomes, logistic regression models for dichotomous outcomes, and Cox regression models for time-dependent outcomes (i.e. length of stay and 5-year survival). For length-of-stay and mortality, Kaplan-Meier plots were fitted. For functional outcomes, standardized regression coefficients (RC) were calculated, and time curves with values at 0, 3 and 6 months were constructed in order to visualize significant differences by MNA categories over time; the overall equality of means and changes over time (0–3 months and 3–6 months) were tested by one-way analysis of variance (ANOVA). All p-values in this paper (except ANOVA, Figs. 2 and 3) and Kaplan-Meier plots are multivariable adjusted as described above; p-values < 0.05 were considered statistically significant.

Fig. 2
figure 2

Length of stay in hospital and rehabilitation clinics until 6 months after hip fracture, by nutritional status (Mini Nutritional Assessment, MNA). P-values were obtained by Cox regression, adjusted for age, sex and American Society of Aneshesiologists (ASA) score (see Methods). A. Overall test of equality of MNA categories: p = 0.190; malnourished vs. well-nourished: HR 0.53, 95% CI 0.26–1.07, p = 0.075; at risk vs. well-nourished: Hazard ratio (HR) 0.87, 95% Confidence interval (CI) 0.61–1.24, p = 0.452. B. Overall test of equality of MNA categories: p = 0.019; malnourished vs. well-nourished: HR 0.44, 95% CI 0.22–0.90, p = 0.023; at risk vs. well-nourished: HR 0.68, 95% CI 0.47–0.97, p = 0.034. C. Overall test of equality of MNA categories: p = 0.027; malnourished vs. well-nourished: HR 0.41, 95% CI 0.19–0.87, p = 0.020; at risk vs. well-nourished: HR 1.22, 95% CI 0.85–1.75, p = 0.287. D. Overall test of equality of MNA categories: p = 0.012. Malnourished vs. well-nourished: HR 0.34, 95% CI 0.16–0.70, p = 0.004; at risk vs. well-nourished: HR 1.05, 95% CI 0.74–1.51, p = 0.773

Full size image
Fig. 3
figure 3

Physical disability (Groningen Activity Restriction Scale) before and at 3 and 6 months after hip fracture, by nutritional status at baseline (MNA). ADL, Activities in Daily Living; IADL, Instrumental Activities of Daily Living. Higher scores represent worse condition. Error bars represent standard errors. Test of equality of MNA categories (ANOVA, unadjusted): p < 0.05 for ADL, IADL and GARS total at all time points, except ADL at 6 months (p = 0.068). Slope between MNA categories were not significantly different, except for change from 3 to 6 months in IADL (p = 0.006) and GARS total (p = 0.050)

Full size image

Results

Participants and baseline data

Baseline data of the 152 enrolled patients after hip fracture are shown in Table 1. At baseline, 65 patients (43%) were malnourished or at risk of malnutrition, and 87 patients (57%) were well-nourished. Of all patients, 85% lived independently at home before the fracture, with no significant differences by nutritional status. Mean age of patients malnourished or at risk was 78.3 years, vs. 74.9y in well-nourished patients (between-group difference: p = 0.03). Medical data indicated a high prevalence of comorbidities: in the total group, 55% had a history of cardiovascular disease, 20% had diabetes, 17% a disease of the nervous system, 11% osteoporosis, and 36% other diseases of the musculoskeletal system and/or connective tissue. Patients malnourished or at risk had a significantly higher prevalence of heart diseases, osteoporosis, COPD and asthma, gastro-intestinal disorders, and “other diseases of the musculoskeletal system and connective tissue” (ICD10: M00–99 except osteoporosis); they also tended to have a higher ASA score (p = 0.052). Fracture type, type of surgery and laboratory parameters at baseline were similar in subjects of different nutritional status categories. The average time between admission and surgery was 0.72 days (~ 20 h), and time between surgery and baseline assessment was 3.2 days, without significant difference between the nutritional categories.

Physical measures of nutritional status at baseline were highly significantly different between the nutritional status categories: patients who were malnourished or at risk of malnutrition weighed ~8.7 kg less than well-nourished patients (Table 1). This weight difference was due to both difference in fat mass (~4.9 kg) and in fat-free mass (~4.0 kg). Although weight loss in the 6 months preceding the fracture tended to be higher in patients malnourished or at risk than in well-nourished patients (− 1.8 kg vs. -0.2 kg, p = 0.058), this recent weight loss accounted for less than 20% of the total weight deficit in patients malnourished or at risk (i.e. -1.6 kg difference in weight loss out of the total weight deficit of − 8.7 kg). Patients malnourished or at risk had also significantly lower BMI, mid-upper arm circumference, mid-arm muscle area and handgrip strength (Table 1) than well-nourished patients.

Moreover, patients malnourished or at risk had significantly more physical restrictions (GARS-ADL, GARS-IADL and GARS-Total), a worse cognitive state (MMSE), a worse score on depression (HADS) as well as on most health-related quality of life indicators as measured by EuroQoL (Table 1).

Short-term clinical outcome

As shown in Table 2, patients who were malnourished or at risk of malnutrition had a shorter duration of surgery (64 vs. 71 min, adjusted difference: − 16 min, 95% CI-29 − 4 min, p = 0.010) and lost less blood during surgery (306 vs. 382 ml, adj. difference - 97 ml, 95% CI -172 − 22 ml, p = 0.012).

Table 2 Prognostic value of nutritional status for clinical outcome after hip fracture
Full size table

Post-operative complications occurred in 55.4% of patients malnourished or at-risk compared to 35.6% of well-nourished patients (adj. OR 2.00, 95% CI 1.01–3.98, p = 0.047). The mean number of complications per patient was 0.85 vs. 0.45 (adj. Difference 0.34, 95% CI 0.07–0.62, p = 0.016). OR’s for specific subtypes of complications were also consistently elevated but none reached statistical significance due to low frequency of specific complications. Pressure ulcers (grade 2, 3 or 4) occurred in only one well-nourished patient vs. five patients malnourished or at risk (OR 7.21, p = 0.104). When we split up the study population into three MNA-levels (malnourished vs. at-risk vs. well-nourished), the association was statistically significant: OR per MNA-level: 5.52 (95% CI 1.24–24.4, p = 0.025), implying a (5.52)2 = 30.4-fold odds (i.e. risk) of pressure ulcers in malnourished patients relative to well-nourished patients.

Two malnourished or at-risk patients and one well-nourished patient died in hospital, and one patient each died during rehabilitation, without significant difference between the groups (p = 0.944 and 0.847, respectively).

Out of 58 patients malnourished or at risk who before the hip fracture had lived at home, 27.6% were directly discharged back to their home from hospital, as compared to 52.4% of 84 home-dwelling well-nourished patients (adj. OR 0.41, 95% CI 0.18–0.98, p = 0.044). After rehabilitation, the proportion of home-dwelling patients discharged back to their home was 85.2% vs. 91.6% (p = 0.511), indicating that after rehabilitation the vast majority of both groups in our study population returned to their homes.

Overall, three times as many patients malnourished or at risk were readmitted to hospital one or more times after discharge (29.2%) compared to well-nourished patients (9.2%; adj. OR 4.59, 95% CI 1.70–12.4, p = 0.003); the average number of readmissions was almost four times higher (0.43 vs. 0.11, adj. difference 0.34, 95% CI 0.15–0.54, p = 0.001). This analysis was repeated excluding patients (n = 10) who had been readmitted for reasons directly related to hip fracture surgery (n = 5 malnourished / at risk, n = 5 well-nourished). In the 17 patients readmitted for reasons unrelated to hip fracture surgery (e.g. Crohn’s disease, diabetic foot, myocardial infarction, respiratory insufficiency), 14 (23.3%) were malnourished or at risk vs. 3 (3.7%) well-nourished (adj. OR 10.1, 95% CI 2.55–39.6, p = 0.001).

Length of stay

Kaplan Meier plots of length of stay in hospital and rehabilitation clinics are shown in Fig. 2A - D. Length of primary hospital stay (Fig. 2A) was not significantly different by nutritional status. However, total length of hospital stay including readmissions (Fig. 2B) was longer both in malnourished patients (adj. HR of being discharged from hospital: 0.44, 95% CI 0.22–0.90, p = 0.023) and patients at risk (adj. HR: 0.68, 95% CI 0.47–0.97, p = 0.034; for patients malnourished and at risk combined: adj. HR: 0.63, 96%CI 0.44–0.89, p = 0.008). Malnourished patients also had a significantly longer length of stay in rehabilitation clinics (Fig. 2C, HR 0.41, 95% CI 0.19–0.87, p = 0.020) and overall length of stay (Fig. 2D, HR 0.34, 95% CI 0.16–0.70, p = 0.004). In patients at risk of malnutrition, length of stay in rehabilitation and overall length of stay were not significantly different from well-nourished patients.

Functional outcome, fatigue and quality of life

Functional outcomes at 6 months after hip fracture are shown in Supplementary Table 1. Comparison with Table 1 shows that at 6 months, only cognition and general quality of life (EuroQoL EQ-5D VAS) had improved relative to baseline measurements. By contrast, all other functional outcomes had not fully recovered to pre-fracture values at 6 months after hip fracture. In multivariable analyses, functional outcomes at 6 months were highly significantly associated with their respective baseline values: standardized regression coefficients (comparable with correlations) varied from 0.30–0.83, with p-values mostly < 0.001 (Suppl. Table 1). In contrast, baseline nutritional status was not independently associated with functional recovery except for the EuroQoL subscale of mobility (standardized RC: − 0.182, p = 0.036) and the CIS subscale ‘motivation’ (standardized RC: 0.197, p = 0.045; Suppl. Table 1).

Curves over time (0, 3 and 6 months) of physical disability (GARS) by nutritional status (Fig. 3) show that the three nutritional status categories followed almost parallel trajectories from baseline until 6 months after hip fracture, suggesting that malnourished patients starting at a relatively low level of function retained their relative position throughout the 6-month observation period. The curves show that from 3 to 6 months after hip fracture, physical function partly improved in well-nourished patients and in patients at risk, though not completely back to baseline levels. In contrast, in malnourished patients, curves of GARS IADL (Fig. 3B) and overall physical disability (GARS total, Fig. 3C) showed an absolute deterioration between 3 and 6 months; statistical testing (ANOVA) confirmed a significant difference between the three MNA categories in slope of the curves of GARS IADL (p = 0.006) and GARS total (p = 0.050) from 3 to 6 months.

New fractures at 1 and 5 years after hip fracture

The proportion of patients with one or more new fractures during the first year after hip fracture was 3.1% in patients malnourished or at risk vs. 1.1% in well-nourished patients (adj. OR 1.79, 95% CI 0.14–23.0, p = 0.655, Table 2). After five years, new fractures had occurred in 16.9% of patients malnourished or at risk vs. 16.1% of well-nourished patients (adj. OR 0.87, 95% CI 0.34–2.24, p = 0.769). Specific subtypes of fractures were also unrelated to nutritional status (data not shown). Almost half the total population (n = 73, 48%) had had one or more bone fractures in the past, of whom 20 patients (13%) had ≥2 fractures. In multivariable analyses, of all studied prognostic factors, the number of previous fractures was the only significant predictor of new fractures over 5 years (OR 2.22, 95% CI 1.34–3.68, p = 0.002).

All-cause mortality at 6 months, 1 and 5 years after hip fracture

At 6 months after hip fracture, 12.3% (n = 8) of patients malnourished or at risk had died, compared to 3.4% (n = 3) in well-nourished patients, but the difference did not reach statistical significance (adj OR 3.11, 95% CI 0.55–17.5, p = 0.198, Table 2). At 1 year, mortality was 15.4% in patients malnourished or at risk, compared to 3.4% of well-nourished patients (adj. OR 4.30, 95% CI 0.90–20.6, p = 0.068), and at 5 years, 47.7% in patients malnourished or at risk vs. 19.5% of well-nourished patients (adj. OR 3.94, 95% CI 1.53–10.2, p = 0.005). Survival analysis using Cox regression showed that the adjusted hazard ratio of dying over 5 years after hip fracture was 2.48 (95% CI 1.33–4.59, p = 0.004) in patients malnourished or at risk, relative to well-nourished patients.

As shown in Fig. 4, survival decreased in the order well-nourished > at risk > malnourished. Relative to well-nourished patients, the adjusted HR of dying over 5 years after hip fracture was 3.72 (95% CI 1.41–9.81, p = 0.008) in malnourished patients, and 2.28 (95% CI 1.20–4.34, p = 0.012) in patients at risk of malnutrition.

Fig. 4
figure 4

Five-year survival in elderly subjects after hip fracture, by nutritional status at baseline (MNA). Overall test of equality of MNA categories: p = 0.008. Malnourished vs. well-nourished: HR 3.72, 95% CI 1.41–9.81, p = 0.008; at risk vs. well-nourished: HR 2.28, 95% CI 1.20–4.34, p = 0.012. P-values were derived from Cox regression, adjusted for age, sex, American Society of Anesthesiologists (ASA) score, and osteoarthritis (see Methods)

Full size image

Discussion

Summary of main findings

In this well-defined cohort, we confirmed the hypothesis that nutritional status is a strong and independent prognostic factor for clinical and functional outcome after hip fracture, after adjustment for age, sex, ASA score and other confounders. Impaired nutritional status was associated with a higher frequency of postoperative complications and hospital readmissions, a lower frequency of home discharge from hospital, longer total length of stay in hospital and rehabilitation, as well as with higher 5-year mortality. Furthermore, compared with well-nourished patients, patients with impaired nutritional status had a clear deficit in functional status at baseline which persisted at 3 and 6 months after hip fracture. However, for some outcomes our hypothesis was rejected: length of primary hospital stay (i.e. without hospital readmissions) and 5-year occurrence of new fractures were independent of baseline nutritional status, and duration of surgery and blood loss during surgery were less in patients at-risk or malnourished. Below, we discuss our main findings in the light of previous literature.

Study population

Body weight was 63.1 kg in patients malnourished or at risk and 71.8 kg well-nourished patients (Table 1), This is ~ 5 kg less than in a comparable study in 448 Dutch geriatric outpatients (age ~ 80 years) in Amsterdam, where MNA-screened malnourished patients and patients at risk weighed on average ~ 69 kg and well-nourished patients 76 kg. This was also reflected by lower BMI in the present study (malnourished/at risk 23.3 kg/m2, well-nourished 25.6 kg/m2) relative to the Amsterdam study (24–27 kg/m2, respectively) [24]. Thus, our hip fracture patients had considerably lower weight and BMI than a ~ 5 years older Dutch outpatient population.

Our data also suggest that at least part of the participants were sarcopenic, as shown by lower mid-arm muscle area, fat-free mass and handgrip strength in subjects malnourished or at risk. Preoperative blood values of hemoglobin and hematocrit were in the lower normal range both in subjects malnourished or at risk and in well-nourished subjects. Blood values of micronutrients, which were reported earlier (online supplemental materials to [13]) did not give an indication of pre-existing nutritional deficiencies.

Postoperative complications

Patients malnourished or at risk of malnutrition had twice the risk (OR) and number of postoperative complications compared to well-nourished patients. The OR for pressure ulcers was even ~ 5.5 for patients at risk of malnutrition and OR ~ 30 for malnourished patients, confirming an earlier report [25]. However, our findings contrast with several previous reports which did not show an association of nutritional status (MNA [26], MNA-SF [9], BMI [27] or sarcopenia [28] with postoperative complications.

Hospital readmissions, length of stay and home discharge

Relative to well-nourished patients, patients malnourished or at risk had a 4.6-fold odds of being readmitted to hospital, confirming one previous unadjusted study [9] and one adjusted study [29]. When we excluded patients readmitted to hospital for reasons directly related to hip fracture surgery, the OR even increased to ~ 10. Patients malnourished and patients at risk also had a longer total hospital stay including readmissions (Fig. 2B), and malnourished patients had longer stay in rehabilitation (Fig. 2C) and overall length of stay (Fig 2D). Previous (mainly unadjusted) studies either showed longer length of stay in malnourished patients [25, 29] or no association [26, 28].

The odds of home discharge from hospital was lower in malnourished or at risk patients relative to well-nourished patients (OR 0.41), confirming several literature reports [25, 30, 31].

Functional outcome

Analysis of functional outcomes showed several noteworthy findings. First, nutritional status was a strong determinant of baseline status of all functional parameters. Second, prospective analyses showed that functional outcomes at 6 months were strongly (mostly p < 0.001) associated with their respective baseline values. This finding was further substantiated by graphical presentations showing that the three MNA categories followed parallel trajectories of physical function (Fig. 3); in other words, subjects who started off with low physical function already before hip fracture kept their low position relative to other subjects. And third, although physical function showed slight improvement between 3 and 6 months after hip fracture in well-nourished and at risk patients, it did not completely recover to baseline values (Fig. 3). In fact, physical function in malnourished patients even deteriorated further between 3 and 6 months, especially with regard to IADL, i.e. preparing meals, laundry, making beds, shop**, light and heavy household activities.

Our findings further substantiate earlier reports showing unadjusted or adjusted associations of baseline malnutrition with mobility or ADL [10, 30,31,32,33,34] or only partial functional recovery after hip fracture [26, 35]. Based on our findings, it would appear that functional recovery after hip fracture is incomplete and is mainly determined by limiting factors which are already present before the fracture, such as nutritional status, age and ADL [36].

Long-term results: new fractures and mortality

Nutritional status was unrelated with new fractures over 5 years after hip fracture; new fractures were only predicted by the number of previous fractures. We found no previous studies on the association between nutritional status and new fractures after hip fracture.

Kaplan Meier plots (Fig. 4) showed a marked association between nutritional status and mortality, with an overall 5-year HR of 3.72 in malnourished patients and 2.28 in patients at risk relative to well-nourished patients. Many previous, mostly unadjusted, studies showed an association between malnutrition and mortality up to 12 months post-fracture [10, 29, 30, 37, 38]. In contrast, reports on long-term mortality are few. An unadjusted study [39] showed elevated 5-year mortality in hip fracture patients with baseline sarcopenia. Two studies showed higher unadjusted 3-year [9] or 5-year [40] mortality in malnourished patients, however in both studies the associations did not persist after adjustment for confounders. Thus, our study appears to be the first showing elevated 5-year mortality after hip fracture both in patients at risk and malnourished patients after full adjustment for confounders.

Course of surgery

Our hypothesis was rejected for duration and blood loss during surgery: patients malnourished or at risk had not longer, but shorter surgery duration; also they lost less blood. Both findings had high statistical significance (p ~ 0.01), minimizing the chance of false-positive results. Two earlier reports showed a positive association between BMI and surgery time in hip fracture [41] and total hip arthroplasty [42]. The surgical procedure in patients with higher BMI is thought to be technically more demanding [42]; amongst others, abundancy of adipose tissue may lead to a reduced access to the operative field and the need for longer incisions, greater force of retraction, and higher number of retractors required to achieve an acceptable exposure. Our finding of lower intraoperative blood loss in patients at risk or malnourished is not in line with previous reports of one unadjusted and two adjusted studies, which showed either null or negative association with BMI (e.g. [43, 44]) or sarcopenia [28].

Significance and implications of findings

Our study has important clinical implications. Our data demonstrate that malnutrition in hip fracture patients is a major independent determinant of functional status and clinical outcome. Despite intensive rehabilitation over six months post-fracture in rehabilitation clinics or at home, our study population did not recover to pre-fracture physical function. Moreover, as previously reported, intensive nutritional intervention in the intervention group led to short-term weight gain, but not to any better short- or long-term clinical or functional outcome relative to the control group [13].

Our combined data suggest that we should not wait with nutritional intervention until after hip fracture, but instead put more effort on prevention of malnutrition and functional decline at an earlier stage. Regular nutritional screening in elderly would be valuable in order to identify patients at risk of malnutrition and to intervene already long before hip fracture. This notion is in line with a recent report showing that dietary protein and calcium intake in middle-aged women was significantly associated with hip fracture risk over the subsequent 20 years [45]. A specially pressing issue is case-finding to close the gap for individuals of lower socio-economic status.

Also, better understanding of causes underlying undernutrition in elderly is needed. Although malnutrition and frailty are two different concepts, their phenotypes are closely interrelated: both are strongly associated with functional and cognitive impairment, multimorbidity, and multimedication, all of which are important contributors to both hip fracture risk and impaired post-fracture recovery.

Strengths and limitations

One strength of our study is the comprehensive dataset with detailed data on medical history and confounders, with complete follow-up until 5 years for new fractures and survival. Second, the MNA [46] was recommended by ESPEN in 2003 [47] and still complies well with the 2019 ESPEN GLIM criteria for the diagnosis of malnutrition. It is considered to be the most effective screening tool to assess the risk of malnutrition in elderly [48] and its superiority over other nutritional screening instruments was recently reconfirmed [10, 32].

Our study also has some limitations. First, due to the need of informed consent and a sufficient cognitive state to administer questionnaires, our study population comprised mostly home-dwelling patients (> 80%) in a relatively good pre-fracture condition. Although this might limit external validity, we would consider it quite unlikely that the very strong observed associations between nutritional status and outcome would disappear in a hip fracture population with greater variability in pre-fracture condition. Second, our study design did not allow us to identify causes of hip fracture or the question whether malnutrition contributed to hip fracture.

Conclusion

In conclusion, we have shown that nutritional status is a strong independent prognostic factor for postoperative complications, discharge to patients’ home after hospitalization, hospital readmissions, length of stay and 5-year survival, independent of confounders such as age, sex and ASA score. Moreover, functional recovery remained incomplete until 6 months after hip fracture in all MNA categories; in fact, in malnourished patients, physical function (especially IADL) even further deteriorated from 3 to 6 months post-fracture. We conclude that elderly care should focus on early prevention and identification of at-risk individuals long before hip fracture.