Early surgery for patients with a fracture of the hip decreases 30-day mortality

Patients and Methods

The senior author (MJP) co-ordinates and delivers the care for patients sustaining a fracture of the hip within Peterborough and Stamford NHS trust. Between 1989 and 2013, data pertaining to this care were prospectively collected using a standard proforma. The collection of data began on admission and continued until one year from injury.

The data included patient age, gender and ASA grade.³ Mobility was scored using a scale from 0 (immobile) to 9 points (independently mobile).¹⁴ Residence was divided into patients living in their own home and those living in institutional accommodation. Mental function was calculated using a standard Mini Mental Test Score (MMTS)¹⁵ as recommended by NICE.² Fractures were classified as intra- or extracapsular. The type of operation was defined as either fixation or replacement. Extracapsular fractures were treated with an intramedullary nail or sliding hip screw. Intracapsular fractures were treated with cannulated screws or a plate and sliding screw device, and replacement was with a hemiarthroplasty or total hip arthroplasty. Delay to surgery was calculated from the time of admission to the Emergency Department (ED) to the time of operation. This time was divided into: one to six hours, seven to 12 hours, 13 to 18 hours, 19 to 24 hours, 25 to 36 hours, 37 to 48 hours and 49 to 72 hours. The outcome which was assessed was the 30-day mortality.

Patients aged < 60 years, those with pathological fractures and patients treated non-operatively were excluded from the study. As we were examining the effects of early surgery on survival, we excluded patients waiting >  72 hours for surgery. We also excluded patients delayed for correctable medical reasons when the surgical or anaesthetic team felt optimisation would significantly improve their chance of surviving an anaesthetic. These mirrored the immediately correctable reasons identified by NICE.²

Results

A total of 7872 fractures of the hip were managed during the study period. After applying the exclusion criteria, 6656 remained. Of these, 18 patients were lost to follow-up leaving 6638 patients for analysis (Fig. 1). The reasons for delay were due to an operating theatre or theatre team being unavailable and most commonly reflected an excessive trauma workload. A MMTS was not recorded pre-operatively for 430 patients and 24 did not have an ASA grade recorded. The mean age of the patients was 82 years (60 to 106), 5406 (78%) were female. Table I shows the characteristics of the patients for each delay to surgery group. Age, ASA grade, MMTS and pre-fracture mobility differed significantly between the groups.

Fig. 1

Flow chart demonstrating the exclusion criteria.

Age, gender, pre-fracture mobility, MMTS, fracture type, ASA grade, pre-fracture residence and delay to surgery were significantly correlated with 30-day mortality on univariate analysis (Table II). The type of operation was not significantly associated and was excluded from the final model. The variable inflation factor was < 2 for all independent variables in the final model suggesting that there was no significant collinearity between them.

Age, gender, pre-fracture mobility, MMTS and ASA grade were statistically significant predictors of 30-day mortality on backward multivariate binary logistic regression. Pre-fracture residence and delay to surgery thresholds were not significantly associated with 30-day mortality after correcting for confounding factors. The Hosmer–Lemeshow test denoted a good model fit (chi-squared test, chi = 5.468, p = 0.707). The variables included in the model predicted 13.1% of the variance in 30-day mortality (Nagelkerke r-square = 0.131). Overall the model strongly rejected the null hypothesis that none of the factors are associated with 30-day mortality (p < 0.001). The validity of the model was upheld with 402 events (deaths within 30 days of admission). This was above the required ten per variable included in the regression model.

The adjusted odds ratios (ORs) with 95% confidence intervals (CIs) for each independent variable in the final model are shown in Table III. The model suggests that ASA grade is the strongest predictor of 30-day mortality with every increase in ASA grade increasing the odds of mortality by 2.52 (95% CI 2.01 to 3.04; p < 0.001). Each additional year in age increased the odds by 1.04 (95% CI 1.03 to 1.06; p < 0.001). Sustaining an extracapsular fracture also increased mortality with an OR of 1.31 (95% CI 1.05 to 1.64; p = 0.019). Female gender, increasing MMTS and mobility score decreased the 30-day mortality with an OR of 0.725 (95% CI 0.56 to 0.94; p = 0.014), 0.906 (95% CI 0.87 to 0.94; p < 0.001) and 0.925 (95% CI 0.87 to 0.98; p = 0.014), respectively.

Surgery within 12 hours appeared to improve survival significantly, after adjusting for confounders, with patients undergoing surgery after 12 hours having a 1.59 (95% CI 1.11 to 2.29) increased odds of death within 30 days (p = 0.013). While there were benefits for the other thresholds examined they were not statistically significant (Table IV).

Discussion

The incidence of fragility fracture of the hip is rising globally¹⁶ and there is some evidence that within this group of patients the incidence of unstable peri-trochanteric fractures is increasing.¹⁷ Whilst mortality remains high, the medical community will endeavour to understand and improve it. A number of guidelines recommend that the balance between prolonged medical optimisation and early surgery should favour the latter. It has been suggested by various bodies that 24 hours,¹⁸ 48 hours¹⁹ and most recently suggested by NICE, 36 hours, is the acceptable limit of the delay from admission to operation.² This figure lacks substantiated evidence. Two recent meta-analyses have loosely recommended early surgery as beneficial but have acknowledged that a lack of prospective studies and poor adjustment for confounding factors prevent definite conclusions, especially when related to 30-day mortality.^8,9 In 2013, Uzoigwe et al²⁰ published results of a retrospective review of prospectively collected data, which examined the effects of early surgery on in-patient mortality. They found that sequential 12-hour thresholds significantly improved in-patient mortality after adjusting for evidence-based confounders. The 12-hour threshold conferred the greatest benefit; this is much earlier than other published studies examining the effects of delayed surgery on early mortality.^21,22

Our study validates these results with a larger sample size and a greater range of relevant confounders. Where possible, each of these confounders has been recorded in a continuous fashion rather than dichotomously; this reduced the significance of our results but produced a statistically robust regression model. We believe our study is also more widely applicable as it excludes patients aged < 60 years and so represents the fragility fracture group examined by the NHFD. It also examines 30-day mortality, which is easily interpreted and less likely to be influenced by local factors such as hospital discharge policies.

Our results support a growing body of evidence that the ASA grade is a powerful predictor of short-term mortality following a fracture of the hip (p < 0.001). Due to its wide spread use and ease of interpretation, we believe it should represent the minimum level of confounding adjustment for studies examining survival after a fracture of the hip. We accept that it is a subjective measure but have found that studies struggle to produce an effective method of stratifying specific co-morbidities against survival.²¹ We found increasing age (p < 0.001) and extracapsular fracture (p = 0.019) predict 30-day mortality. Protective factors are female gender (p = 0.014), increasing MMTS (p < 0.001) and an increasing mobility score (p = 0.014). These are easily reproducible, commonly recorded (as recommended by NICE) and should be included in future survival analyses.

Our aim was to examine the effects of early surgery on 30-day mortality in patients sustaining a fracture of the hip. Our model did not demonstrate an overall benefit of six hourly early watersheds (six hours, 12 hours, 18 hours and 24 hours), nor did it find a significant benefit of later watersheds on survival (36 hours or 48 hours) that previous studies have demonstrated. This is probably due to our comprehensive adjustment for confounding variables. Our study did support the findings by Uzoigwe et al²⁰ that early surgery before 12 hours does significantly decrease short-term mortality in patients sustaining a fracture of the hip (p = 0.013) but not that immediate surgery (< 6 hours) confers a significant survival benefit.

We acknowledge that our study had missing data (24 missing ASA grade and 430 missing MMTS score), lost 18 patients to follow-up and that data from a single centre may not reflect national trends. Another limitation is that our model predicted 13.1% in the variability of mortality, suggesting that there are more factors affecting survival yet to be elucidated. Furthermore, despite the recent surge of resources and streamlining of services towards the care of these patients, most trusts will be unable to provide surgery within 12 hours for most of them. This makes it difficult to incorporate our results into a guideline for the acceptable operative delay in patients with a fracture of the hip. Our results support the benefits of early surgery and justify the allocation of resources to this vulnerable group.

Ethical reasons have precluded any effective randomised control trials looking at the effect of operative delay on survival in patients who sustain a fracture of the hip. Our results will inform future studies assessing whether early surgery confers a survival benefit. As proposed by The Hip Fracture Accelerated Surgical Treatment and Care Track (HIP ATTACK) Investigators,²³ early surgery versus routine care would appear to be an ethically appropriate method of examining this question, which still has considerable clinical equipoise.⁹ This remains an issue of great importance as displacing patients with other injuries from the trauma list without definitive evidence of a benefit may not be ethically justified.