Association of Timing of Surgery for Hip Fracture and Patient Outcomes FREE

The incidence of hip fracture is increasing in the United States, with more than 340 000 occurring in the year 2000.¹ In elderly patients, the 1-year mortality rate for hip fracture ranges from 14% to 36%.² Additionally, hip fracture is associated with poor functional outcomes.^{3 - 4}

Surgical repair is a key element in the management of hip fracture. Before surgery, most patients are confined to bed rest. In theory, delay in surgery and mobilization could affect functional and other outcomes by increasing bed rest–associated complications, including thromboembolism,⁵ urinary tract infections,⁶ atelectasis, and pressure ulcers.⁷ On the other hand, precipitous surgery and failing to stabilize medical problems could increase the risk of perioperative complications.

Whether early surgery is beneficial is a long-running controversy and is one of the most common clinical issues in the early acute management of these patients. A randomized trial of surgical timing in hip fracture has never been conducted, but the relationship of timing and outcomes has been examined in other studies.^{8 - 20} Many of these studies did not adjust for comorbid illness or other parameters that might be used to "select" patients for earlier surgery. Of the studies that controlled for comorbidity,^{11 ,14 ,16 - 17 ,19} some found that early surgery is associated with lower mortality, LOS, and complications. A more recent, large study,²⁰ however, found that surgery after 48 hours was not associated with increased mortality compared with surgery in 24 to 48 hours. Of note, virtually all studies evaluated the association of early surgery with mortality rather than functional outcomes. We examined the association of the timing of hip fracture surgery with mortality, functional outcomes, pain, length of stay (LOS), and complications.

Consecutive admissions to 4 hospitals in the New York City metropolitan area were screened for cases of hip fracture for 29 months ending December 1999. The hospitals included 1 academic medical center, an urban teaching hospital, an orthopedics hospital, and a suburban hospital. Exclusion criteria included patient age younger than 50 years, fractures that occurred as an inpatient, transfers from another hospital, multiple trauma, pathological fractures, distal and femoral shaft fractures, bilateral hip fractures, or previous fracture or surgery on the currently fractured site. The study was approved by the institutional review boards at the 4 participating hospitals. All patients provided verbal informed consent. Of the 1741 cases admitted, 23.4% met exclusionary criteria, 4.1% refused to participate, 2.6% were discharged before consent could be obtained, 0.6% had incomplete data, and 69.3% (n = 1206) were enrolled.

Trained research associates enrolled patients as early in the admission as possible (69% were enrolled on or before the day of surgery). Information on prefracture function, residential location, and history of dementia was collected from patients or their proxies (if the patient was delirious or cognitively impaired). Information on each patient's functional status for the 2 weeks prior to fracture was obtained by interview using the motor scale of the Functional Independence Measure (FIM),²¹ which consists of 13 items in 4 subscales of physical functioning (locomotion, self-care, toileting, and transferring). Each item was scored between 1 (for complete dependence) and 7 (for complete independence) using specific criteria. Medical records were reviewed throughout the hospital admission to collect information on comorbid medical problems, type of fracture, and other aspects of medical care.

Information was also collected on abnormal clinical findings (admission physical examination and laboratory findings) that are commonly available and used by clinicians to decide whether to delay surgery. This information was used to determine whether the patient had (1) a systolic blood pressure of 90 mm Hg or less; (2) a rate or rhythm abnormality (defined as atrial fibrillation or supraventricular tachycardia at a rate greater than 120/min, ventricular tachycardia, 3° block, or a rate of ≤45/min); (3) chest pain or myocardial infarction within 3 months; (4) poorly compensated heart failure (a chest x-ray consistent with congestive heart failure, the presence of dyspnea, abnormal lung findings [eg, rales, rhonchi, and decreased breath sounds] or an S₃ gallop); (5) an abnormal international normalized ratio of 1.4 or higher; and (6) a laboratory abnormality (sodium <125 or >155 mEq/L; potassium <2.5 or >6.1 mEq/L; bicarbonate <18 or >36 mEq/L; glucose >600 mg/dL [33.3 mmol/L]; serum urea nitrogen >50 mg/dL [17.8 mmol/L]; creatinine >2.5 mg/dL [221 µmol/L]; hemoglobin ≤7.5 g/dL; a pulse oximetry reading of <90%; PO₂ <60 mm Hg; or PCO₂ >55 mm Hg). Cutoff values were established by examining the relationship between the occurrence of early surgery and the range of values for each finding.

For the patients enrolled during the first 12 months, each patient was also seen 5 days a week in the hospital to collect additional information on pain and complications. At each visit, the patient was asked to assess pain severity for the previous 24 or 48 hours (on a 5-point scale ranging from no pain to very severe pain). The hospital course and medical record were reviewed at each visit, and complications were recorded. Major complications were defined as those that posed a threat to life or bodily functions and that typically are treated with parenteral medications, procedures, or intensive monitoring. Examples of major complications include (1) pneumonia if both respiratory symptoms and/or hypoxia were documented, and (2) arrhythmias if their occurrence increased the risk of ischemia or hemodynamic compromise. Nurses identifying complications were not aware of the study hypothesis. Physicians who categorized complications as to whether they were true complications or arose from a preexisting condition were not blinded as to surgical timing.

All patients were followed up, and information on functional status and mortality was obtained by telephone at 6 months. Additional deaths were identified from hospital records and vital statistics. Ascertainment of death or functional outcome was available for 94.0% of participants at 6 months.

We compared patients having surgery within 24 hours with those having surgery after 24 hours on the following outcomes: (1) mean pain scores over the first 5 hospital days; (2) number of days of severe and very severe pain over hospital days 1 to 5 (assessed by asking patients if they were experiencing no pain, or mild, moderate, or severe pain); (3) major postoperative complications; (4) LOS; (5) mortality through 6 months; (6) FIM locomotion (a 2-item subscale focusing on walking and climbing stairs) score at 6 months; (7) FIM self-care (a 6-item scale of self-care activities including bathing and dressing); and (8) FIM transferring (a 3-item scale focusing on transfers from the bed, toilet, and tub). The analyses of complications (n = 554) and pain (n = 487 patients able to report on pain) were limited to patients enrolled in the first 12 months. Other analyses involved data from all enrolled patients having surgery. We focused on pain over the first 5 hospital days because we were interested in cumulative pain burden rather than preoperative or postoperative pain that would have been difficult to compare between the early and later surgery groups. For the FIM measures, analyses were restricted to survivors. We excluded patients who were totally or maximally dependent on walking at baseline (n = 75) from analyses of FIM locomotion because the scale could not be responsive to worsening in these patients due to a "floor" effect. To test the sensitivity of our results to the exclusion of survivors, we also examined the effect on a combined measure of mortality or needing total assistance in locomotion.²²

Using ordinary least squares regression (for continuous outcomes), logistic regression (for binary outcomes), or Cox proportional hazards regression for our main analyses, we controlled for age, sex, nursing home residence, needing a proxy for consent, delirium on admission, prefracture FIM locomotion score, fracture type, hospitalization within 6 months, hospital site, day and time of admission, abnormal clinical findings, and history of diabetes, chronic obstructive pulmonary disease, stroke syndrome, dementia, cardiac disease, and hypertension. Odds ratios (ORs) were adjusted to approximate the relative risk.²³

We performed 2 types of supplementary or sensitivity analyses. First, we performed an analysis using propensity scores^{24 - 25} to match patients for whom the likelihood of having early surgery was similar. In the first step, we used stepwise logistic regression to generate a propensity score for having early surgery for each patient using the available variables. Each case of surgery within 24 hours was then matched with a case having later surgery based on the closest propensity score (within 10%) and closest age (when multiple matches were obtained). We compared patients having surgery within 24 hours to matching cases having surgery after 24 hours.

As our second supplementary analysis, we examined whether the results changed when we excluded patients who might not be candidates for early surgery because of markedly abnormal clinical findings or the need for additional time for preoperative evaluation. Thus, the restricted cohort excludes patients admitted with abnormal clinical findings, aortic stenosis, dementia, and end-stage renal disease on dialysis. All analyses were performed using STATA, release 7 (STATA, College Station, Tex). P<.05 was considered significant.

Twenty-eight patients (2.3%) did not have surgery and were excluded from the analyses. Of the remaining patients (n = 1178), 33.8% (n = 398) had surgery 24 hours or less after hospital arrival. The median time to surgery was 19 hours (interquartile range, 15-22 hours) in the early surgery group and 40.6 hours (interquartile range, 29-53 hours) in the late surgery group. Compared with patients who went to surgery within 24 hours, patients who had later surgery (n = 780) were less likely to have been admitted from nursing homes (P = .04) and were more likely to have poorly compensated heart failure (P<.001), abnormal international normalized ratio (P<.001), and other laboratory abnormalities (P = .02) (Table 1). Having surgery within 24 hours also varied by hospital site, by day of the week, and time of admission.

The overall unadjusted mortality was 8.2% at 2 months and 17.5% at 6 months (unadjusted hazard ratio for early surgery was 0.68; 95% confidence interval [CI], 0.48-0.97; P = .03). After adjustment for the factors shown in Table 2, earlier surgery was not associated with improved mortality (hazard ratio, 0.75; 95% CI, 0.52-1.08; P = .12).

Pain control was assessed in 487 of the 554 patients enrolled in the first 12 months. Patients enrolled in the first 12 months were similar to those enrolled later except that they were slightly less likely to have delirium on admission, had slightly lower admission FIM scores, were more likely to have a cardiac history, and be admitted in the previous 6 months (data available from authors on request). Patients who responded to questions on pain had better function, were younger, and were less likely to have dementia than those who did not (data available from authors on request). Compared with patients having later surgery, earlier surgery was associated with lower pain scores (difference of −0.24 points; 95% CI, −0.44 to −0.06) and fewer days of severe and very severe pain for the first 5 days of hospitalization (difference of −0.22 days; 95% CI, −0.41 to −0.03 [Figure 1]). Early surgery was also associated with shorter LOS by 1.94 days (P<.001). The differences in pain and LOS were accounted for by the delay in surgery; postoperative pain and LOS were equivalent. Early surgery was associated with improved FIM self-care (P = .04), but not fewer complications (P = .10). FIM locomotion scores did not differ at 6 months (difference of −0.04 points; 95% CI, –0.49 to 0.39). The relationship of surgical timing with FIM locomotion score was similar when we used different time frames (surgery within 24 hours, within 24-48 hours, or after 48 hours of arrival) for surgery (data available on request).

The logistic regression to derive the propensity score had a C statistic of 0.68. Of the cases who had surgery within 24 hours (n = 398), 373 (93.7%) were successfully matched. No significant differences were found between the 2 matched groups for the characteristics listed in Table 1. Earlier surgery was not associated with improved mortality (OR, 0.98; 95% CI, 0.63-1.50; P = .99) in propensity score–matched cases. Early surgery was associated with reduced pain and LOS (Table 3), but not complications or FIM functional status measures at 6 months.

In the restricted cohort, early surgery continued to be associated with reduced pain and LOS (Table 2). Number of days of severe and very severe pain was −0.30 days fewer in the early surgery group (95% CI, –0.50 to −0.08). Additionally, early surgery was associated with reduced major postoperative complications (P = .04). Early surgery was not associated with functional outcomes or mortality in the restricted cohort (Table 2).

Previous studies have yielded conflicting results on the relationship of early surgery for hip fracture and survival, and the relationship of early surgery and functional outcomes and pain is unknown. In this study, we found that compared with later surgery surgery in the first 24 hours was not associated with either improved or worsened survival and function at 6 months, a time frame during which most recovery will occur.^{3 - 4} On the other hand, early surgery was consistently associated with decreased LOS and less pain and probably with reduced major complications among patients who were medically stable at admission. In an earlier study,²⁶ we reported that clinical reasons (waiting for test results or for medical stabilization) were infrequent reasons for delayed surgery in patients operated on between 24 and 48 hours after admission. Instead, system problems (timely consultation or availability of the surgeon or operating room) accounted for most delays. Thus, it is feasible to improve surgical timing that could in turn translate to improved efficiency and reductions in severe pain.

In the case of mortality, our finding is consistent with those of a recent study of 8383 patients in which no association was found when mortality was compared in early (defined as 24-48 hours) and later surgery.²⁰ Patients with surgery in the first 24 hours were excluded; thus, our study goes beyond that analysis by considering surgical timing from the time of admission—a better test of the early surgery hypothesis. In the case of functional outcomes, several explanations exist for the null finding. First, the benefit could be small. For example, we hypothesized that reduced pain might translate to improved function based on an earlier study,²⁷ but the effect of earlier surgery on reduced pain may be too small or too short in duration to make a difference on function. Second, the functional benefit may be limited to a subgroup of patients and obscured in analyses focusing on the average patient. Third, the benefit from early surgery may be limited and short-lived if it is not followed up by timely mobilization, early rehabilitation,²⁸ and attention to postacute care.

This study is the largest study of hip fracture we know of that has detailed clinical information (beyond administrative data and medical records) on the hospital course, as well as information on functional outcomes through 6 months. Nevertheless, our study was limited by reliance on self-report for functional status, by reduced statistical power for selected outcomes, and by the observational study design.

In the case of measuring function, we selected our methods knowing that prefracture function is an important predictor of outcomes. Since observing function before the fracture is not feasible, a measure that involved reporting of functional status was needed. We selected what we considered to be the most appropriate interview measure given that a considerable fraction of hip fracture patients would need proxy respondents because of dementia or delirium. The literature on proxy respondents indicates that agreement between patients and proxies is greatest for questions that focus on discrete and observable tasks, as were the measures we used, relative to questions that may ask about perceived limitations in doing those tasks.

For functional outcomes, the 95% CI for the FIM locomotion scale showed that it is unlikely that early surgery improved or worsened locomotion by more than half a point, and we believe that clinically significant differences were excluded. For example, the difference in 6-month mortality between a prefracture FIM locomotion score of 6 vs 8 (2-point difference) was 21% vs 15%; the mortality difference for a half point would be extremely small. Furthermore, an unadjusted association between early surgery and improved mortality diminished after adjustment for other risk factors; the benefit was largely eliminated in analyses involving propensity score matching or a restricted cohort. Although we could not exclude a moderately large benefit of early surgery, the 95% CI (0.52-1.08 for the hazard ratio) in the main adjusted analysis indicates that anything more than a small increase in mortality from early surgery is unlikely. Finally, we showed that early surgery was consistently associated with shortened LOS and reduced pain. In addition, the number of fewer days of severe or very severe pain with early surgery (0.22-0.30 days or 5-7 fewer hours of severe pain on average) suggests that this is a clinically significant difference.

Although our study was observational, we attempted to control for selection in several ways. First, our analyses adjusted for a range of variables used by clinicians to select patients for early surgery. These measures go beyond those available from administrative data and include information on function collected from interviews. Second, we also used propensity score methods to match cases of early and late surgery. Finally, we repeated our analyses excluding patients who might not be appropriate candidates for early surgery. Given that a randomized trial of early vs delayed surgery is not likely to be done, we believe that this important clinical question can only be answered by careful observational research methods.

In conclusion, early surgery alone does not appear to have a beneficial effect on mortality or function for the average patient with hip fracture. However, early surgery was associated with less pain, reduced LOS, and probably fewer major complications among patients who were medically stable at admission. Further studies are needed that focus on the functional impact of early surgery on subgroups of patients. Additionally, research is needed on whether the theoretical benefits of early surgery on functional outcomes may be achieved when early surgery is combined with the timely provision of mobilization, rehabilitation, and the full range of postacute medical services. In the meantime, early surgery should be a goal for most medically stable patients with hip fracture, given that adverse events are unlikely and that pain, LOS, and possibly complications will be reduced.