The determinants of mortality and morbidity during the year following fracture of the hip

Patients and Methods

This prospective multicentre observational study involved all consecutive patients with a fracture of the hip who were admitted to our institutions between January 2011 and April 2012. There were initially 606 patients. Inclusion criteria were age ≥ 50 years and a history of low-energy trauma. Patients with a pathological fracture (14) and those treated conservatively (24) were excluded. This left 568 patients available for review one-year post-operatively. The study had ethical approval and all patients gave informed consent. Sociodemographic data including level of education and living situation, height and weight, medical history and whether they were smokers, were collected. In those patients who were cognitively impaired, information was obtained from either the partner or a close relative. Pre-fracture mobility was recorded using a five-item scale that ranged from grade 1 (unable to walk) to grade 5 (no need for walking aids).¹⁹ The activities of daily living (ADL) index²⁰ was used to assess function. A score of 6 indicates full function and 1 indicates the most severe functional impairment. Cognitive status was assessed using the mini-mental state examination (MMSE)²¹ and comorbidities using the cumulative illness rating scale (CIRS).²² The MMSE consists of questions grouped into seven categories, each representing a different cognitive domain. The maximum score is 30 and any score ≥ 27 indicates normal cognition. The CIRS contains 13 items for different organ systems, and each system is rated in five levels of increasing impairment. Interviews were carried out by three authors (PR, EA, MG) who were unaware of the pattern of the fracture. The baseline characteristics of the patients are shown in Table I. All fractures were classified on anteroposterior (AP) and lateral radiographs by two orthopaedic surgeons (GGC, SC), who were unaware of the clinical characteristics of the patients, into femoral neck, trochanteric and subtrochanteric fractures. The pattern was further defined using the AO classification²³ for trochanteric fractures, and both the AO and the Garden²⁴ classifications for fractures of the femoral neck. The intra- and interobserver reliability of the classification of the fractures was tested using Cohen’s κ coefficient (Table II).

Pre-operative levels of haemoglobin (Hb), the American Society of Anaesthesiologists (ASA)²⁵ grade, the time to surgery, the type of surgery (arthroplasty or osteosynthesis), in-hospital general or hip-related complications and length of hospital stay were recorded at discharge. In-hospital complications were further divided into major and minor, based on the perceived impact on outcome and survival. Follow-up information was available for 552 patients (97.2%). A telephone interview at four and 12 months post-operatively was used to collect data from the patients or, where necessary, from their relatives or carers. Information was collected regarding possible implant and medical complications, as well as re-operations that took place in the previous time interval. When a complication or re-operation was reported, the family physician was contacted for further information. For each deceased patient, the precise date of death was obtained. A total of 16 patients were lost to follow-up and were excluded from the analysis.

Results

The classification of the fractures and surgical features of the patients are summarised in Table III. The 30-day, four-month and one-year rates of mortality were 4.3% (24/552), 11.4% (63/552) and 18.8% (104/552), respectively. The rate of mortality at 30 days included in-hospital deaths (11 patients) and deaths that occurred between discharge from hospital and 30 days after surgery (13 patients). The Kaplan–Meier survival curves for the two age groups showed an increased mortality in patients aged ≥ 80 years compared with the younger age group (Fig. 1). There was no significant difference in the 30-day rate of mortality between these two groups (< 80; 10/264, 3.8% vs > 80; 14/288, 4.9%; p = 0.537). However, there was a significantly increased rate of mortality one year post-operatively in the older age group (< 80; 33/264, 12.5% vs > 80; 71/288, 24.7%; p < 0.001). Table IV shows age-adjusted hazard ratios for single variables associated with mortality. There was a positive association between male gender and the 30-day and one-year rate of mortality. The determinants of mortality revealed by the multivariate Cox’s regression analysis are shown in Table V. Complications that had occurred within the previous time interval and the ASA grade represented the two most important predictors. In the model including patients aged < 80, the strongest predictors of 30-day and one-year mortality were ASA grade and the mobility four months post-operatively, respectively. In the model in which only patients aged ≥ 80 years were selected, complications that had occurred within the previous time interval represented by far the most important predictor of mortality at both time intervals. Complications and medical conditions following the fracture are shown in Table VI. In-hospital, four-month and one-year general complications occurred in 167 (29.4%), 91 (18.6%) and 30 (6.7%) patients, respectively. During their hospital stay, many developed several complications. Factors significantly influencing the development of complications and medical conditions are shown in Table VII. Using multivariate analysis, the occurrence of pressure sores was inversely related to the MMSE score (odds ratio (OR) 0.90; 95% confidence interval (CI) 0.87 to 0.94; p < 0.001) and to surgery performed within 72 hours (OR 0.53; 95% CI 0.30 to 0.93; p = 0.028), whereas it was directly associated with ASA grade (OR 2.41; 95% CI 1.40 to 4.14; p = 0.001). The MMSE score was the single most important predictor in this model. With the numbers available, no predictors of hip-related complications at any time were identified by multivariate analysis, even when patients treated with osteosynthesis or prosthesis were examined separately. Using multiple linear regression analysis, the length of stay was directly associated with in-hospital general complications (coefficient (c) = 1.47; p < 0.001), the use of a prosthesis vs osteosynthesis (c = 1.70; p < 0.001) and male gender (c = 1.97; p < 0.001), whereas surgery carried out within 72 hours (c= -4.17; p < 0.001) and the mean pre-operative Hb level (c = -0.43; p < 0.001) represented inverse predictors. The model accounted for 36% of the total variance in the length of hospital stay, with a short operative delay contributing 26%. A reoperation was reported by 15 patients (3.0%) at four months and by 13 (2.9%) one-year post-operatively. The BMI (OR 1.12; 95% CI 1.03 to 1.22; p = 0.009) was a weak direct predictor of the likelihood of re-operation, whereas the type of fracture and the type of surgery were not predictive. The pre-fracture treatment for osteoporosis was not associated with reported re-operations at follow-up.

Fig. 1

Graph showing survival curves after hip fracture in two age groups (< 80 years and ≥ 80).

Discussion

The rate of mortality in our cohort at the various time intervals was similar to that which has been previously reported.⁴ It was also closely linked to post-operative complications as previously described.^8,9,12,13,15As in the other studies,^3,6,10 increasing ASA grade played an independent detrimental role. The effect of patients’ mobility and ADL on mortality appeared only one year after the fracture and was not significantly associated with in-hospital deaths. Limited function prior to fracture has been associated with complications²⁶ and mortality.^1,2,27 We found that the odds of being alive one year after a fracture of the hip were lower for patients aged > 80 years than for those aged < 80 years. We also found a direct association between male gender and mortality in univariate analysis, but this was not confirmed using multivariate analysis. Both advanced age and male gender have been recognised as strong predictors of mortality after a fracture of the hip.⁴ However, not all studies agree on the effect of age^2,12,28 and male gender^29,30 in independently predicting mortality using a multivariate analysis. For example, others have found that men are more likely to have more severe medical comorbidities than women at the time of surgery.^15,31 Clearly, multivariate analysis investigating the predictors of mortality is dependent on the choice of risk factors that are included. Therefore, the selection of covariables not used in previous studies or the use of different cut-off values may help to explain the lack of an association between mortality and increasing age or male gender in the multivariate analysis. However, when we constructed separate models of Cox’s regression analysis to assess whether age influenced the effect estimates, the role of complications as predictors of mortality was enhanced in patients aged ≥ 80 years compared with the younger age group. Thus, even though increasing age was not directly related to increased mortality in our multivariate analysis, it may have set the stage for the direct effect of early or delayed complications on mortality in the first post-operative year following fracture, because of the alteration in the functional balance caused by the fracture in these patients.

Our cohort was comparable to others^6,9,11-13 in their rate of post-operative complications. We detected a decrease in the frequency of life-threatening conditions related to the fracture at four months and one year post-operatively. This may reflect the previous higher mortality owing to severe complications in the early post-operative period. Our multivariate analysis identified comorbidity to be a determinant of the likelihood of complications while in hospital and at four months post-operatively. These elderly patients may not have the functional reserve to withstand multiple insults, and this may lead to more frequent delayed medical complications and increased mortality. In earlier studies, a higher grade of ASA or CIRS has also been associated with an increased risk of early^6,12,15 and delayed³² complications. In our patients a lower MMSE score was positively associated with an increased risk of post-operative complications. Cognitive impairment has been associated with a poor functional outcome³³ and mortality^33,34 in patients after a fracture of the hip, perhaps because of interference with motivation and function.³³ Poor cognitive status, longer time to surgery and a higher ASA grade were also independent determinants of pressure sores in our patients, confirming that complications often result from a combination of the these risk factors rather than from a single cause. Previous studies have found similar relationships.^10,35-37

The cut-off of 72 hours for the delay to surgery was longer than in most other studies.^3,10,36 In our study, the delay was mainly attributable to secondary evidence of fracture, delayed presentations, unavailability of operating rooms and/or surgical personnel, and the patient’s comorbidity at the time of admission. Although mortality was not influenced, we detected a reduction in the risk of complications among patients who underwent earlier surgery, which is consistent with the literature.^10,36-38 These models using multivariate analysis were adjusted for the level of comorbidity. Therefore, independent of comorbidities leading to the surgical delay, the delay itself was a determinant of in-hospital and general complications four months post-operatively. Delayed surgery was also the most important independent factor prolonging the hospital stay. Earlier studies also indicated that the shorter the pre-operative waiting time, the shorter the post-operative stay.^37,38

Surgical factors, such as the type of fracture or specific treatment, have been under-reported in observational studies of fractures of the hip.¹³ The mortality and morbidity of our patients were not influenced by type of fracture (femoral neck or trochanteric) or surgical implant, thereby confirming previous reports.^2,5,6,13,27 Only the increasing AO grade of fracture represented a weak predictor of mortality one year post-operatively in our multivariate analysis.

This study has several limitations. First, the study population may not be typical of the elderly patients who sustain a fracture of the hip, as the hospitals were not selected for their case mix. Hence, conclusions based on these patients may not apply to all patients with a fracture of the hip in our geographical area. Nevertheless, the participating hospitals have large catchment areas that include urban and rural areas, with patients of diverse socioeconomic status. Second, the patients may have been unable to recall their pre-fracture status owing to diminished mental status. Third, no testing of reliability was performed on the accuracy of the replies from relatives and carers, which might alter the precision of assessments made during the interviews. In order to minimise this effect, validated instruments were administered during the interviews. Moreover, family physicians were contacted to collect data on any complications or re-operations that occurred during the study period.

The study has several strengths, however. We used a large cohort of consecutive patients with data collected prospectively. Patients lost to follow-up and those treated conservatively did not exceed 10% of the study population. Also, the use of validated tools permitted comparison with other studies. This makes ours one of the few prospective studies to report complications and medical conditions between admission and the situation one year post-operatively.