We investigated the relationship between a number of patient and management variables and mortality after surgery for fracture of the hip. Data relating to 18 817 patients were obtained from the Scottish Hip Fracture Audit database. We divided variables into two categories, depending on whether they were case-mix (age; gender; fracture type; pre-fracture residence; pre-fracture mobility and ASA scores) or management variables (time from fracture to surgery; time from admission to surgery; grade of surgical and anaesthetic staff undertaking the procedure and anaesthetic technique).
Multivariate logistic regression analysis showed that all case-mix variables were strongly associated with post-operative mortality, even when controlling for the effects of the remaining variables. Inclusion of the management variables into the case-mix base regression model provided no significant improvement to the model. Patient case-mix variables have the most significant effect on post-operative mortality and unfortunately such variables cannot be modified by pre-operative medical interventions.
Fracture involving the hip is a common and serious injury in the elderly, with a high associated mortality reported at between 14% and 47%.1–11 In Scotland, just over 6000 patients per year sustain a fracture of the hip, of which 95% are treated surgically.1 Mortality is one of the most important and frequently reported measures of outcome, and a number of patient and management2–11 variables have been studied to identify the principal determinants of mortality after surgery. The conclusions drawn from such studies are often inconsistent, and controversy continues regarding the optimal management of this common injury.
In this study we used the data from the Scottish Hip Fracture Audit database to identify patient and management variables associated with early post-operative mortality after surgery for fracture of the hip.
Patients and Methods
The Scottish Hip Fracture Audit database is a national prospective audit which collects data relating to patients admitted to hospital with a fracture of the hip. Data are collected by on-site, locally funded, dedicated audit coordinators. A standard core dataset was collected during the acute stay, and follow-up data were collected at 120 days by telephone or postal questionnaire. Accurate and complete data were ensured by standardised data collection procedures, subcontracted dual data entry and a monthly ongoing validation process.
The variables studied in this analysis were divided into two groups, depending upon whether they were case-mix (patient characteristics which cannot be adjusted) or management variables (aspects of management which we may potentially be able to control). Case-mix variables included age, gender, fracture type, pre-fracture residence, pre-fracture mobility, and the Society of Anesthesiologists (ASA)4,7,9 score. Management variables included time from fracture to operation, time from ward admission to operation, seniority of operating surgeon, seniority of administering anaesthetist, and anaesthetic type. Patients who were treated non-operatively were excluded from the analysis.
Statistical analysis was performed using SPSS version 13.0 (SPSS Inc., Chicago, Illinois). In order to control for the effects of potentially confounding variables we used multiple logistic regression models to determine which factors had an association with post-operative mortality. Case-mix and management variables were first subjected to univariate analysis to identify potential association with post-operative mortality, and examined with chi-squared testing. These variables were then combined in a base multivariate logistic regression model to assess the significance of individual variables when controlling for the others. We then assessed the case-mix-adjusted significance of management variables by offering them to this base model and calculated the odds ratio (OR) with the 95% confidence intervals (CI). Statistical significance was assumed when p < 0.05.
Data were collected at source from 22 acute orthopaedic units between January 1998 and December 2004. A total of 18 817 individuals were included in the study for whom the 30- and 120-day mortality was 7% and 18%, respectively. The relationship between mortality and each case-mix or management variable is summarised in Tables I⇓ and II⇓. Initial univariate analysis (Table I⇓) identified that all case-mix variables were strongly associated with early post-operative mortality, with a p-value of < 0.01 on chi-squared testing. However, with respect to the management variables, only the grade of anaesthetist appeared to have any significant association (Table II⇓).
Multivariate analysis confirmed that all case-mix variables continued to have a significant association with peri-operative mortality, even after controlling for the effects of the remaining variables (Table III⇓). Inclusion of the management variables in the case-mix base model indicated that only anaesthetic type for the 120-day mortality, and time from fracture to surgery for the 30-day mortality, had any association with peri-operative mortality (Table IV⇓).
Patients who sustain a fracture of the hip have higher reported age and/or gender-adjusted mortality rates than the general population.11 The factors responsible for this increased mortality have been subjected to extensive study, but the results have often been contradictory. Advancing age has been associated with increased mortality after hip fracture,2,9,12–15 but other variables such as co-morbidity may act as confounding variables; several authors have challenged the notion that age as an isolated variable affects the mortality rate after fracture of the hip.7,16 In this study age was found to be strongly associated with mortality after surgery for fracture of the hip even after controlling for the confounding effects of differences in case-mix variables between age groups. Indeed, individuals aged 90 years and above have a 120-day mortality of 28%, compared with only 5% for the 50- to 59-year old cohort (OR 7.95, 95% CI 5.49 to 11.5).
The ASA score is a validated and accepted means of documenting an individual’s health status before surgery.17 Patients with high ASA scores have higher reported postoperative mortality after surgery for fracture of the hip.10,18 This is consistent with the findings of our study, where the ASA score was strongly associated with peri-operative mortality. The effects of individual co-morbid conditions on peri-operative mortality (such as cognitive impairment or cardiac/respiratory pathology) were not specifically examined. There are several reasons for this. It is difficult to accurately collect such detailed data on a national basis. It is also difficult to then stratify the effect which a particular medical co-morbidity presents, as most medical conditions have a spectrum of severity which can affect post-operative mortality according to the severity of the condition. Because of this, we used the ASA scoring system as a surrogate measure to stratify pre-fracture co-morbidity. This scoring system is a practical means of documenting a co-morbidity effect and is an important element of the case mix, with higher scores associated with higher mortality after surgery for fracture of the hip.1 In this study we used multivariate analysis to control for the effects of confounding variables. Although such models have proved useful and have gained enormous popularity, they may also be treacherous, as there is no limit to the amount of data that can be included in the analyses. One problem faced by analysts of multivariate data is over-parameterisation. As such, by using the ASA score as a simple measure of pre-operative morbidity, we were able to limit the number of variables included in the multivariate regression model.
In Scotland, the ratio of the incidence of fractures of the hip in women to men is 3.4:1.1 Men undergoing surgery for fracture of the hip have been reported to have increased morbidity and mortality, a younger age at presentation, and poorer markers of pre-operative health status than age-matched women.1,19–23 However, several authors have reported that, when controlling for such variables, male gender is not associated with an increased peri-operative mortality after surgery for fracture of the hip.6,7,18 In this study women were found to have significantly lower early post-operative mortality, even after controlling for the effects of differences in the other case-mix variables between the genders (OR 0.49, 95% CI 0.45 to 0.54).
Poor pre-fracture function has been reported as having a strong association with increased mortality after surgery for hip fracture.24–27 We used pre-fracture residence and mobility as markers of pre-fracture function. Both factors as independent variables were found to be strongly associated with peri-operative mortality, even when controlling for potentially confounding variables such as age and medical co-morbidity.
The Scottish hip fracture audit has highlighted considerable variations in the grade of surgeon and anaesthetist providing treatment,1 but univariate analysis did not identify any significant association between the grade of operating surgeon and peri-operative mortality. However, patients anaesthetised by junior anaesthetic staff had significantly lower 30- and 120-day post-operative mortality. This finding may be the result of confounding factors such as more experienced anaesthetists undertaking greater numbers of procedures involving high-risk patients. When we controlled for differences in patient case mix using the multivariate model, the grade of neither anaesthetist nor surgeon was found to have any significant association with early post-operative mortality.
The impact of anaesthetic technique on outcome after surgery for fracture of the hip has been reported in a number of studies, the majority suggesting that regional anaesthesia is associated with a lower one-month mortality.28–30 Parker, Handoll and Griffiths31 performed a meta-analysis of 22 trials involving 2567 patients comparing regional anaesthesia with general anaesthesia in surgery for fracture of the hip. All trials were found to contain methodological flaws, and many did not reflect current anaesthetic practice. It was concluded that there was insufficient evidence available from the trials to rule out clinically important differences. We divided anaesthetic technique into two groups, depending on whether a general or a regional (spinal/ epidural) anaesthetic was administered. Univariate analysis did not identify any statistically significant difference in peri-operative mortality between the two groups. Anaesthetic technique was then entered into the multivariate model that controlled for all six case-mix variables (in addition to the grade of surgeon and anaesthetist). No association was found with 30-day mortality. However, regional anaesthesia was associated with a slightly higher 120-day mortality than was general anaesthesia (18% vs 17%, OR 1.10, 95% CI 1.01 to 1.20). This is the largest single study to date to analyse the issue of anaesthetic technique and mortality after surgery for fracture of the hip. The reported findings differ from those of the majority of previous studies, but it should be noted that modern techniques of general anaesthesia differ significantly from those used in many of the studies published in earlier years, and the difference in mortality between anaesthetic techniques is quite small, albeit statistically significant.29,30
Table I⇑ documents mortality according to fracture pattern. Extracapsular fractures in the elderly have a higher reported mortality than intracapsular, and we observed similar findings in this study.32 Although the number of tumour-related fractures was relatively small, such injuries were found to have a particularly high 120-day mortality (46%) compared with osteoporotic fractures.
The association between timing of surgery for fracture of the hip and peri-operative mortality remains one of the most controversial issues in the management of hip fracture patients, and reports published in the medical literature have been contradictory.33–42 The majority of studies examining the effect of delay to theatre and early post-operative mortality concentrate on the time from orthopaedic admission to surgery. One highly confounding factor that may affect such analyses, and which is rarely addressed, is the variable of time from fracture to surgery. It is important to distinguish between time from fracture and time from admission, as this may vary greatly depending on the availability of local resources, and is of particular relevance in regions where the geographical distribution of the population may mean that transfer to a specialist centre may be delayed for some time after the injury has been sustained. We have, therefore, accounted for both times in our analyses. In the univariate analysis, both time from fracture to surgery and time from admission to surgery had no significant relationship with early post-operative mortality. When these variables were entered into the regression model we found no association between time from admission to surgery and mortality at either 30 or 120 days. Time from fracture to surgery was found to affect 30-day but not 120-day mortality. However, there is no clear association between 30 day mortality and time from fracture to surgery, as mortality does not consistently increase or decrease with time, suggesting that this may represent a chance association (Table IV⇑). NHS Quality Improvement Scotland has adopted the standard that 98% of individuals admitted with a fracture of the hip (if medically fit) should undergo surgery within 24 hours of ‘safe operating time’, defined as 8am to 8pm seven days a week. This has the benefit that the patient is treated by a senior surgical team, and reduces repeated cycles of fasting for surgery.43 From a humanitarian point of view it is favourable to undertake surgery for fracture of the hip as soon as it is medically safe to do so. Several authors have suggested that patients who undergo surgery within 24 hours of admission have improved outcomes in terms of early post-operative mortality, but we have found no compelling evidence to support this statement.41,42 More than 60% of patients admitted with a fracture of the hip have significant co-existing medical pathology at the time of presentation, and we believe that it is more important that these frail patients are treated on dedicated, planned trauma lists after appropriate pre-operative assessment and medical optimisation.1
Using the logistic regression coefficients (B) listed in Table III⇑, it is possible to calculate the predicted mortality for an individual patient based upon their case-mix characteristics. The following formula may be used (where ‘e’ is the base of the natural logarithm (equal to 2.72) and ‘constant’ is related to the regression constant for 30-day and 120-day mortality (Table III⇑).
For example, the predicted 30-day mortality for a hip fracture patient aged 90 or over; ASA = 3; from a care home, assuming other base characteristics (male, intracapsular fracture, walked without aids before fracture) = 1/1 + e− (−4.79 + 0.80 + 0.53 + 1.96) = 18.2%.
In this study we have analysed the effects of a number of patient case-mix and management variables on early mortality after surgery for fracture of the hip. It is clear that the principal factors affecting mortality after such surgery are case-mix factors (age, ASA score, gender, fracture type, pre-fracture residence, and mobility). Unfortunately, such variables cannot be modified by pre-operative medical interventions.
A further opinion by Mr M. Parker is available with the electronic version of this article on our website at www.jbjs.org.uk
No benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article.
- Received May 27, 2008.
- Accepted June 11, 2008.
- © 2008 British Editorial Society of Bone and Joint Surgery