Hip fracture constitutes a worldwide public health problem among the elderly . The most common cause of hip fractures is low-energy trauma combined with osteoporosis, and is most frequently seen in elderly women.
Several studies have demonstrated an increased risk of mortality in hip fracture patients [2-6]. The highest risk of mortality is seen in the immediate post fracture period , but increased mortality is still seen 12 month post fracture [2, 8]. Over the first 12 months, the crude mortality ranges from 27.6% to 40.5 % in men and from 15.8% to 23.3% in women . The mortality rate depends on a variety of factors including age, gender, co-morbidity and pre-fracture functional state .
Numerous studies have explored the impact of different factors on the outcome following a hip fracture, such as admission blood tests, co-morbidities, clinical score systems etc. [3, 10-23].
If we are able to identify high risk patients, treatment can be optimized and outcome might be improved. If such identification can be performed by the use of simple clinical score systems, it will be an easily accessible tool for adjusting and specializing treatment. It would furthermore be an inexpensive and quickly performed assessment, which would be highly useful for clinicians.
In this metaanalysis we chose to focus exclusively on simple bedside tests and their ability to predict mortality in hip fracture patients. Thus, we chose score systems containing parameters directly observable in the patient or parameters derived from questionnaires. Score systems containing paraclinical data such as blood tests or requiring the use of highly technological equipment in general were thus excluded.
Three clinical score systems, the American Society of Anesthesiologist score (ASA), Charlson Comorbidity Index (CCI) and dementia score (MMSE: Mini Mental State Examination) (Table 1) met the inclusion criteria, and the aim of this metaanalysis was to assess the association of these score systems with mortality in hip fracture patients.
The PRISMA guidelines for metaanalyses were followed .
A search for articles in the Medline Database using PubMed was conducted on November 13, 2011. The string, used for this search, consisted of three concepts, as shown below:
I. (“hip fracture” OR “hip fractures” OR (“femoral neck” AND fractur*) OR (*trochant* AND fractur*) OR (“collum femoris” AND fractur*) OR (intertrochant* AND fractur*) OR “proximal femoral fracture*”)
II. (mortality OR death OR fatal* OR survival)
III. (ASA OR “ASA score” OR “American Society of Anesthesiologists”).
The first concept was designed to retrieve articles relating to hip fractures and the second to retrieve those relating to mortality or survival.
The last concept was constructed to search for the three score systems individually, thus resulting in three different search strings, one for each score. ASA is shown as an example:
“hip fracture” OR “hip fractures” OR (“femoral neck” AND fracture*) OR (*trochant* AND fracture*) OR (“collum femoris” AND fracture*) OR (intertrochant* AND fracture*) OR “proximal femoral fracture*”) AND (mortality OR death OR fatal* OR survival) AND (ASA OR “ASA score” OR “American Society of Anesthesiologists”)
Third concept strings for CCI and dementia: CCI: (Charlson OR “Charlson co-morbidity” OR “co-morbidity score”). Dementia: (dementia).
Articles were chosen through an iterative process; headlines and abstracts were screened and articles found relevant were reviewed in full text version. The reference lists of the selected articles were screened, and the service “related citations” in PubMed was applied, to find additional publications and additional words for the search strings. The selection process was repeated three times. The search yielded a total of 315 publications.
The publications were further sorted based on the inclusion criteria (English language; full-text editions electronically available at the Royal Library of Denmark; study population larger than 100 subjects; trial population of mixed gender; trial duration less than or equal to two years).
Figure 1 shows the search which resulted in the inclusion of thirteen studies [3, 10, 12-22]. Essential characteristics, including clinical score system, number of studies included, number of study participants and average days of follow-up, are listed in Table 2 (n = 14, some of the included studies evaluated more than one clinical score system).
Odds ratios (OR) with matching 95% confidence intervals (CI) were calculated for the individual studies as described by Bland and Altman , using a contingency 2 χ2 table. p-values were calculated using χ2. p < 0.05 was considered significant.
The calculations were conducted with Microsoft Office Excel 2007.
Review Manager (RevMan), developed by the Cochrane Collaboration, was used to carry out the systematic review . The standard errors and ORs were plotted in a funnel plot to create a visual aid for the detection of any systematic heterogeneity or publication bias, also known as Egger’s test .
The I2 statistic was used to evaluate the degree of dissimilarity between the individual studies. I2 describes the percentage of inconsistency between studies that is due to heterogeneity rather than sampling error . I2 > 50% was used as a threshold, indicating high heterogeneity.
Studies were analyzed using the fixed effect model when I2 < 50% and the random effect model when I2 ≥ 50%.
Table 2 shows the characteristics of the studies in the analyses, including first author, clinical score system, important demographics, length of follow-up, design of studies, ORs and p-values.
In the funnel plots, the data were distributed equally around the vertical axis, implying no presence of publication bias (not shown).
In both the ASA and CCI analyses, statistical dissimilarities were absent, evidenced by I2 = 0%. This indicates full homogeneity, and a fixed effect analysis was therefore chosen.
The dementia analysis showed high statistical heterogeneity with an I2 of 64%. This suggests that the studies might not have a common effect size. Differences in outcome are more likely to result from variance than from pure chance. Thus the random effect model was used. With this model, each study is weighted equally. It will further incorporate greater variability, resulting in wider confidence intervals .
Besides statistical heterogeneity, the studies are to some extent different with regard to design and the type of personnel carrying out the tests.
Seven studies were included in the analysis of the association of a high ASA score with mortality (Figure 2). 56.4% of the patients had a high ASA score of three or above. The results showed no statistical heterogeneity, I2 = 0, and the 95% confidence intervals (95% CI) overlapped implying that the effects are similar . All seven studies had ORs above one and, with related p-values all below 0.0001, they were statistically significant. The ORs ranged from 2.63 (95% CI 1.74-3.97) in the study by Holvik et al to 4.94 (95% CI 2.33-10.50) in the study by Talsnes et al, resulting in a total OR of 3.07 with a narrow 95% CI (2.78-3.38), (p < 0.0001). Thus patients who scored three or more have an approx. three times increased risk of dying compared with those who scored 1-2.
Three studies were included in the analysis dealing with the Charlson Comorbidity Index. 38.6% of the patients scored one or more on the CCI (Figure 2). The results showed no statistical heterogeneity, I2 = 0. All studies had an OR above one, but only the studies by Franzo et al and Nielsen et al showed statistical significance with a p value below 0.0001.
The study by Kirkland et al, in contrast, had a p value of 0.2, which is above the established 0.05 threshold. However, the pooled OR of 2.05 with a narrow 95% CI (1.79-2.34) showed statistical significance (p < 0.00001) for a 2.05 times increased risk of dying if co-morbidity is present.
Four studies were included in the analysis of the association of dementia with mortality (Figure 2). 41.2% of the patients had dementia on admission. The results showed a high statistical heterogeneity, I2=64%, which is why the random effect model was used. All four studies had ORs above one and showed statistical significance with p values < 0.05. The ORs ranged from 1.64 (95% CI 1.11-2.43) in the study by Bergeron et al to 4.71 (95% CI 2.26-9.85) in the study by Hershkovitz et al, resulting in a statistically significant (p = 0.0001) pooled OR of 2.73 (95% CI 1.64-4.57).
This metaanalysis has found all three clinical score systems to be of prognostic value for mortality in hip fracture patients.
The ASA analysis demonstrated an excess mortality three times greater in patients with an ASA score of 3-5 compared with those who score 1-2. The statistically significant (p < 0.0001) OR was 3.07 with a narrow 95% CI (2.78-3.38).
The CCI analysis showed an excess mortality in patients who score 1+ compared with patients who have no co-morbidities present. The statistically significant (p < 0.0001) OR of this study is 2.05 with a narrow 95% CI (1.79-2.34).
The patients with premorbid dementia demonstrated an almost three times greater mortality compared with non-demented patients, with the statistically significant (p = 0.0001) OR of 2.73 (95% CI 1.64-4.57).
However, the analyses of the CCI and dementia do not comprise many studies. It would, of course, be preferable to re-analyze the two score systems and their relation with mortality in hip fracture patients when more studies are available. Furthermore, different methods were used for the assessment of dementia, MMSE or journal extraction. It would be useful for future studies if the same diagnostic methods were used and were well documented in the papers as well.
With the help of these three bed-side clinical score systems, the clinicians can perform a rapid and inexpensive assessment of their patients and and increased risk of dying without needing paraclinical tests to optimize care and treatment.
In this metaanalysis we were interested in investigating the sole effect of the clinical score systems on the assessment of mortality. Nevertheless, the clinical score systems can easily be combined with other types of risk assessment methods, thereby potentially improving the assessment. Some studies have alone focused on the effect of paraclinical tests, others have combined paraclinical tests with clinical score systems. Markers, such as haemoglobin, albumin, total leukocyte count, creatinine and parathyroid hormone in admission blood tests, can be used as significant outcome predictors in hip fracture patients [11, 23].
Other scoring systems with a proven predictive value do exist, but they require additional investigations such as blood samples, X-rays etc.
The Estimation of Physiologic Ability and Surgical Stress (E-PASS) scoring system is comprised of a preoperative risk score, a surgical stress score, and a comprehensive risk score based on the preoperative risk score and surgical stress score . It seems useful for predicting postoperative risk, estimating costs and for comparison of outcome in patients following surgery of hip fractures [30, 31].
The Nottingham Hip Fracture Score (NHFS) is a scoring system validated for the prediction of 30-day mortality after hip fracture surgery . It is made up of seven independent predictors of mortality that have been incorporated into a risk score: age (66–85 and ≥ 86 years); sex (male); number of co-morbidities (≥ 2), admission mini-mental test score (≤ 6 out of 10), admission haemoglobin concentration (≤ 10 g × dl-1), living in an institution; and the presence of malignancy . The NHFS uses data that are easily and routinely collected from all patients presenting with hip fractures. The NHFS can be calculated on admission to hospital .
The physiologic and operative severity score for the enumeration of mortality and morbidity (POSSUM) system predicts mortality and morbidity in surgical patients using physiologic and operative factors . This system has also been presented in a modified version for orthopedic use that has been validated for patients with femoral neck fractures . The POSSUM requires surgical and anaesthetic data  and in some studies the orthopedic POSSUM scoring system has been shown to over-predict mortality in patients with hip fractures [36, 37]. However, it has been pointed out that this may be due to the method of analysis used in these studies .
These scoring systems, however, remain complex and it can be difficult to single out the effect of each of the individual components comprised by the systems. This metaanalysis therefore focused on individual clinical scoring systems allowing for differentiation of care, e.g. in assigning the patients to specialized orthogeriatric units.
The functional outcome in hip fracture patients can further be optimized by combining the orthopaedic surgeon’s and the geriatrician’s expertise this has yielded a significant reduction of the overall length of hospital stay while still maintaining a high quality of care [38, 39].
Furthermore, ongoing adjustments in treatment may be conducted on the basis of the simple clinical score systems.
Moreover, the rate of postoperative complications and mortality may be reduced by using an optimized hip fracture treatment, i.e. a fast-track programme, where improvements such as a more efficient pain treatment and reduction of time to surgery are applied .
Factors such as functionality (New Mobility Score, Cumulated Ambulation Score), type of fracture, type of surgery, time to surgery and the geriatric follow-up are other possible predictors of mortality in hip fracture patients [41, 42].
A future study of interest is one that looks at all factors: paraclinical parameters, bedside clinical scoring systems, co-morbidities, logistical parameters such as waiting times for surgery and medication consumption and the correlation with mortality in hip fracture patients in a large patient material.
Correspondence: Debbie Norring-Agerskov, Nordre Fasanvej 155B, 2000 Frederiksberg, Denmark. E-mail: firstname.lastname@example.org
Accepted: 29 May 2013
Conflicts of interest:None. Disclosure forms provided by the authors are available with the full text of this article at www.danmedj.dk
Cooper C, Campion G, Melton LJ, 3rd. Hip fractures in the elderly: a world-wide projection. Osteoporos Int 1992;2:285-9.
Johnell O, Kanis JA, Oden A et al. Mortality after osteoporotic fractures. Osteoporos Int 2004;15:38-42.
Beringer TR, Clarke J, Elliott JR et al. Outcome following proximal femoral fracture in Northern Ireland. Ulster Med J 2006;75:200-6.
Keene GS, Parker MJ, Pryor GA. Mortality and morbidity after hip fractures. BMJ 1993; 307:1248-50.
Gordon PC, The probability of death following a fracture of the hip. Can Med Assoc J 1971;105:47-51 passim.
Abrahamsen B, van Staa T, Ariely R et al. Excess mortality following hip fracture: a systematic epidemiological review. Osteoporos Int 2009;20:1633-50.
Vestergaard P, Rejnmark L, Mosekilde L. Has mortality after a hip fracture increased? J Am Geriatr Soc 2007;55:1720-6.
Vidal EI, Coeli CM, Pinheiro RS et al. Mortality within 1 year after hip fracture surgical repair in the elderly according to postoperative period: a probabilistic record linkage study in Brazil. Osteoporos Int 2006;17:1569-76.
Morin S, Lix LM, Azimaee M et al. Mortality rates after incident non-traumatic fractures in older men and women. Osteoporos Int 2011;22:2439-48.
Maggi S, Siviero P, Wetle T et al. A multicenter survey on profile of care for hip fracture: predictors of mortality and disability. Osteoporos Int 2010;21:223-31.
Laulund AS, Lauritzen JB, Duus BR et al. Routine blood tests as predictors of mortality in hip fracture patients. Injury 2012;43:1014-20.
Bergeron E, Lavoie A, Belcaid A et al. Should patients with isolated hip fractures be included in trauma registries? J Trauma 2005;58:793-7.
Hershkovitz A, Polatov I, Beloosesky Y et al. Factors affecting mortality of frail hip-fractured elderly patients. Arch Gerontol Geriatr 2010;51:113-6.
Olofsson B, Stenvall M, Lundstrom M et al. Mental status and surgical methods in patients with femoral neck fracture. Orthop Nurs 2009;28:305-13.
Pretto M, Spirig R, Kaelin R et al. Outcomes of elderly hip fracture patients in the Swiss healthcare system: a survey prior to the implementation of DRGs and prior to the implementation of a geriatric fracture centre. Swiss Med Wkly 2010;140:w13086.
Franzo A, Francescutti C, Simon G. Risk factors correlated with post-operative mortality for hip fracture surgery in the elderly: a population-based approach. Eur J Epidemiol 2005;20:985-91.
Kirkland LL, Kashiwagi DT, Burton MC et al. The Charlson Comorbidity Index Score as a predictor of 30-day mortality after hip fracture surgery. Am J Med Qual 2011;26:461-7.
Elliott J, Beringer T, Kee F et al. Predicting survival after treatment for fracture of the proximal femur and the effect of delays to surgery. J Clin Epidemiol 2003;56:788-95.
Holvik K, Ranhoff AH, Martinsen MI et al. Predictors of mortality in older hip fracture inpatients admitted to an orthogeriatric unit in Oslo, Norway. J Aging Health 2010;22:1114-31.
Soderqvist A, Ekstrom W, Ponzer S et al. Prediction of mortality in elderly patients with hip fractures: a two-year prospective study of 1,944 patients. Gerontology 2009;55:496-504.
Talsnes O, Hjelmstedt F, Dahl OE et al. Clinical and biochemical prediction of early fatal outcome following hip fracture in the elderly. Int Orthop 2011;35:903-7.
Nielsen KA, Jensen NC, Jensen CM et al. Quality of care and 30 day mortality among patients with hip fractures: a nationwide cohort study. BMC Health Serv Res 2009;9:186.
Mosfeldt M, Pedersen OB, Riis T et al. Value of routine blood tests for prediction of mortality risk in hip fracture patients. Acta Orthop 2012;83:31-5.
Moher D, Altman DG, Liberati A et al. PRISMA statement. Epidemiology 2011;22:128; author reply 128.
Bland JM, Altman DG. Statistics notes. The odds ratio. BMJ 2000;320:1468.
Review Manager (RevMan) [computer programme]. Version 5.1. Copenhagen: The Nordic Cochrane Centre, 2011.
Sterne JA, Egger M. Funnel plots for detecting bias in meta-analysis: guidelines on choice of axis. J Clin Epidemiol 2001;54:1046-55.
Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med 2002;21:1539-58.
Walker E, Hernandez AV, Kattan MW. Meta-analysis: its strengths and limitations. Cleve Clin J Med 2008;75:431-9.
Hirose J, Mizuta H, Ide J et al. E-PASS for predicting postoperative risk with hip fracture: a multicenter study. Clin Orthop Relat Res 2008;466:2833-41.
Hirose J, Mizuta H, Ide J et al. Evaluation of estimation of physiologic ability and surgical stress (E-PASS) to predict the postoperative risk for hip fracture in elder patients. Arch Orthop Trauma Surg 2008;128:1447-52.
Maxwell MJ, Moran CG, Moppett IK. Development and validation of a preoperative scoring system to predict 30 day mortality in patients undergoing hip fracture surgery. Br J Anaesth 2008;101:511-7.
Wiles MD, Moran CG, Sahota O et al. Nottingham Hip Fracture Score as a predictor of one year mortality in patients undergoing surgical repair of fractured neck of femur. Br J Anaesth 2011;106:501-4.
Copeland GP, Jones D, Walters M. POSSUM: a scoring system for surgical audit. Br J Surg 1991;78:355-60.
Wright DM, Blanckley S, Stewart GJ et al. The use of orthopaedic POSSUM as an audit tool for fractured neck of femur. Injury 2008;39:430-5.
Ramanathan TS, Moppett IK, Wenn R et al. POSSUM scoring for patients with fractured neck of femur. Br J Anaesth 2005;94:430-3.
Young W, Seigne R, Bright S et al. Audit of morbidity and mortality following neck of femur fracture using the POSSUM scoring system. N Z Med J 2006;119:U1986.
Adunsky A, Lusky A, Arad M et al. A comparative study of rehabilitation outcomes of elderly hip fracture patients: the advantage of a comprehensive orthogeriatric approach. J Gerontol A Biol Sci Med Sci 2003;58:542-7.
Gonzalez-Montalvo JI, Alarcon T, Mauleon JL et al. The orthogeriatric unit for acute patients: a new model of care that improves efficiency in the management of patients with hip fracture. Hip Int 2010;20:229-35.
Pedersen SJ, Borgbjerg FM, Schousboe B et al. A comprehensive hip fracture program reduces complication rates and mortality. J Am Geriatr Soc 2008;56:1831-8.
Gregersen M, Zintchouk D, Hougaard K et al. Interdisciplinary geriatric intervention among nursing home residents with hip fracture reduces mortality. Ugeskr Læger 2010;172:1902-7.
Daugaard CL, Jørgensen HL, Riis T et al. Is mortality after hip fracture associated with surgical delay or admission during weekends and public holidays? A retrospective study of 38,020 patients. Acta Orthop 2012;83:609-13.