In-hospital waiting time to surgery and functional outcomes in geriatric hip fractures: a directed acyclic graph-based preplanned analysis from a prospective multicenter cohort study

Background:

The early recovery of hip function after hip fracture surgery values more attention, especially for patients with delayed surgery of longer than 48 h. We aim to evaluate the associations of in-hospital surgical waiting time with the functional outcomes [Harris Hip Score (HHS), Parker Mobility Score (PMS), and EuroQol 5 dimensions VAS (visual analogue scale) score (EQ-5D VAS)] in elderly patients who sustained hip fractures.

Materials and methods:

Data on sociodemographic and clinical factors were prospectively collected using a multicenter hip fracture registry system. Participants in the cohort underwent a 12-month follow-up investigation. After adjusting potential confounders identified by the directed acyclic graphs, the associations between surgical waiting time longer than 48 h and functional outcomes were estimated by log-binomial regression and multivariable linear regression models with generalized estimating equations.

Results:

Of 863 survival participants with available functional data at 12 months after surgery, an increased risk was obtained from receiving surgery after 48 h and the poor functional outcomes (HHS<80: relative risk (RR)=1.56, 95% CI: 1.00–2.51; PMS<7: RR=1.49, 95% CI: 1.13–2.01; EQ-5D VAS<80: RR=1.97, 95% CI: 1.57–2.47). In-hospital waiting time greater than 48 h were time-invariantly associated with lower PMS during recovery (−0.44 units 95% CI: −0.70 to −0.18). In addition, delayed surgery was time-varying associated with HHS and EQ-5D VAS.

Conclusions:

The associations between in-hospital waiting time and postoperative functional score suggest that delayed surgery can lead to poor functional outcomes, especially in patients waiting longer than 72 h from injury. Delayed surgery mainly impacted hip function and mobility recovery with a slower speed in early recovery of the first 3 months. More attention should be paid to mechanisms behind the associations between delayed surgery on general healthy status.

Introduction

Highlights

• Delayed surgery (>48 h) of hip fractures in elderly people might lead to poor functional outcomes (hip function, mobility, and general healthy status).

• The associations were time-varying during a 12-month follow-up, and early recovery in the first 3 months after surgery calls for more possible interventions.

• Long-delayed surgery (>72 h from injury) of hip fractures in elderly people likely has a higher risk of poor functional outcomes than short-delayed surgery (≤72 h from injury).

As the population ages, fragility hip fractures are becoming an increasingly urgent health care problem among the elderly with a substantial burden for individuals and society^1–3. The treatment generally focuses on optimizing surgery and postoperative recovery, including looking for the appropriate surgical timing. Previous researches have examined the effects of waiting time to surgery on postoperative adverse outcomes in elderly hip fractures and suggested the clinically practical surgical timing (etc. 48 h or 24 h)^4–7. Despite the well-known impact of delayed surgery on short-term and long-term mortality and complications after surgery, the relationships between surgical waiting time and postoperative functional outcomes are still unclear. In clinical practice, we mainly focus on in-hospital delaying, and considering surgery within 24–48 h of admission as timely surgery, although the time from injury to admission may vary, especially in those elderly people⁸.

Factors related to postoperative function in elderly patients are complicated and reported to interact with or have different roles in function recovery^9,10. Previous studies (both randomized controlled trial and cohort study, Supplemental Digital Content 1, http://links.lww.com/JS9/A289) have limitations on potential confounders identified; most were presented without the time before admission^11,12. It is also unclear whether the associations between delayed surgery and functional scores varied over time. Several studies with a limited sample size explored the impact of delayed surgery on functional outcomes in almost 6 months’ postoperative recovery phase after a hip fracture surgery, and reported inconsistent conclusions^13–15. However, the function score for the hip and general status are time-dependent. Whether surgical waiting time associated with hip function recovery varies across time needs to be investigated.

What is needed is to analyze these potential time-dependent relationships accurately and appropriately provide clinical suggestions for fracture recovery. Hence, we conducted a preplanned analysis with the aid of directed acyclic graphs (DAGs) to investigate the impact of surgical waiting time on the functional outcomes and examined their longitudinal associations in a cohort of hospitalized geriatric hip fracture patients from six hospitals in China.

Materials and methods

Data source: a multicenter prospective Chinese geriatric hip fractures cohort

We used data from a multicenter prospective Chinese geriatric hip fractures cohort study which was conducted between June 2016 and June 2019 in six tertiary hospitals. Participants who received hip fracture surgery and met the following criteria were enrolled: aged 65 years or older; no more than 24 h from fracture to emergency admission; low-energy hip fractures (femoral neck fracture or intertrochanteric fractures); American Society of Anesthesiologists (ASA) grades I–III and can withstand surgery; willing to receive hip surgery and a 12-month follow-up. The exclusions were: fractured in the past 6 months; complex trauma, malignancy, or pathologic fractures; presence of congestive heart failure or renal failure; surgery cannot be made within 7 days; had severe Parkinson disease, cognitive impairment or other mental illness; refused to participate. The cohort study was approved by the Medical Ethics Committee of each hospital and was registered at the Chinese Clinical Trial Registry (registration ID: ChiCTR-OOC-16008965). Data of sociodemographic (age, gender, marital status), clinical characteristics (fracture type, comorbidities, physical status, surgical information), and follow-up investigation (complications, mortality, functional outcomes) were prospectively collected and recorded using a multicenter registry system, which was accessible by all six centers. Comorbidities and physical status were measured using Charlson Comorbidity Index (CCI)¹⁶ and ASA grade. All participants signed and dated Institutional Review Boards/Ethics Committees-approved written informed consent. Patient data were de-identified and protected by privacy safeguards appropriate to local regulations. Here, we performed a preplanned analysis in survivors with available function scores at 12 months and reported the findings according to the STROCSS (strengthening the reporting of cohort, cross-sectional and case–control studies in surgery) criteria¹⁷, Supplemental Digital Content 2, http://links.lww.com/JS9/A290.

Surgical treatment and exposure of interest

Upon admission to the emergency department, all patients were clinically assessed by orthopedic surgeons. Patients were operated on when medically possible. To ensure the quality, operations were performed by surgeons with clinical experience as associate chief surgeons for more than 3 years (or surgeons who had completed at least 30 similar operations). The exposure of interest was the in-hospital waiting time for surgery (calculated as the time interval between admission and surgery). Receiving surgery within 48 h from admission has been recommended by evidence-based clinical guidelines as the standard care for hip fractures^8,18. Results of Hip Attack RCTs (randomized controlled trials) suggest that accelerated surgery did not significantly lower mortality risk or a composite of major complications compared with standard care⁶. Similar to previous studies, it was classified into a binary variable (delayed surgery or not: 0=equal or less than 48 h after admission, 1=more than 48 h after admission)^19,20.

Functional outcomes measurement

Questionnaires on Harris Hip Score (HHS), Parker Mobility Score (PMS), and EuroQol 5 dimensions (EQ-5D) were assessed at four time points-prefracture (on admission, based on recall), 3, 6, and 12 months after surgery by blinded, trained interviewers. As an outcome tool typically used to evaluate total hip replacements and hip fractures with excellent test–retest reliability of 0.94²¹, the HHS has a maximum of 100 points and comprises four domains: pain (1 item, 0–44 points); function (7 items, 0–47 points); absence of deformity (1 item, 4 points); and range of motion (2 items, 5 points). The higher HHS results indicate less dysfunction and better outcomes. It can be interpreted as a good or excellent outcome as equal to or higher than 80 points^21,22. PMS is a self-reported assessment tool (on a scale of 0–9) and has proven to be a valid and reliable measurement of hip fracture patient mobility, including three items (walk inside, walk outside, and go shopping or visit family)^23–25. A general score of 6 points or less indicates functional impairment in patients with hip fractures²⁶. EuroQol 5 dimensions 3 levels (EQ-5D-3L) questionnaire with a self-rated visual analogue scale (VAS) score (0 represents the worst imaginable health and 100 the best imaginable health) were applied^27,28. To measure the general health status, EQ-5D VAS instead of the index score was used²⁹. Because we have no aim at estimating quality-adjusted life years, and the index scores are not comparable internationally³⁰. According to published literature, we defined the poor outcomes for HHS and PMS as a score lower than 80 points and 7 points at 12 months, respectively^26,31. As no evidence-based minimal clinically important difference (MCID) has been determined for EQ-5D VAS, we defined its poor outcome as a score lower than 80 points at 12 months after surgery, the median EQ-5D VAS score in this cohort.

Data analyses

We summarized the baseline characteristics of the cohort using counts and percentages for categorical variables and means (SDs) for the continuous variables. We examined the differences between the two groups (≤48 h vs. >48 h) using a t test for continuous variables and χ ² test or Fisher exact test for categorical variables where appropriate.

By reviewing possible causal mechanisms reported by published studies on surgical waiting time and functional outcomes (Supplemental materials 1, Supplemental Digital Content 3, http://links.lww.com/JS9/A291), we constructed a DAGs framework to reveal the minimal sufficient adjustment sets (MSAS) of potential confounders (Supplemental materials 2, Supplemental Digital Content 3, http://links.lww.com/JS9/A291). All DAGs were created and analyzed with DAGitty 3.0 software³².

After adjusting potential confounders, the associations between surgical waiting time on functional outcomes were estimated. We applied log-binomial regression models to examine the associations (relative risk, RR) of surgical waiting time and poor functional outcomes, as we defined. We performed multivariable linear regressions with generalized estimating equations (GEEs) to examine the associations (differences between the means, MD) of postoperative function scores. The GEEs estimated the population-average associations of the exposure while accounting for the correlation among the repeated measures within an individual³³. To assess whether the strength of the associations varies with time, we added an interaction term between surgical waiting time and observation time in GEEs. Coefficients estimates were presented with 95% CIs.

To visualize the trends, we plotted the modeled trajectories of the two groups of four-time repeated measures (including prefracture) using the confounding-adjusted estimated marginal means³⁴ with 95% CIs in GEEs.

All analyses were performed with SAS 9.4 statistical software (SAS Institute, Cary, North Carolina, USA). Considering the existence of missing data attributed to uncompleted follow-up, multivariable analyses were repeated on 30-imputed data sets performing multivariate imputation by chained equations (MICE) as sensitivity analyses by the IVEware program running through SAS³⁵. To further explore the impact on functional outcomes from delayed status prolonged by the preadmission time, we also performed log-binomial regression models as subgroup analysis in short-delayed surgery and long-delayed surgery patients (receiving surgery within 72 h from injury or not). A P<0.05 was considered statistically significant. P values of multiple comparisons between groups in multivariable linear regression models were adjusted using Bonferroni correction.

Results

Patients’ characteristics

A total of 1960 registered patients received surgery for hip fracture from six centers, and 1524 patients who met the inclusion criteria were enrolled. The average age of all participants was 80.3 (7.8) years, with the majority being females (69.3%). Details of the characteristics of study participants were listed and compared between two groups (surgical waiting time ≤48 h vs. >48 h) in Table 1. Of these, 75.7% (1153/1524) completed the 12 months follow-up investigation (median: 378 days), and 863 survivors with available functional outcomes were included in the preplanned analysis (Fig. 1). In addition to the covariates in MSAS (age, gender, CCI, ASA grade, time to admission, prefracture function score and fracture type), we added marital status (married: 90.4% vs. 81.3%; P<0.001) and surgery method (arthroplasty: 27.3 vs. 43.2%; P<0.001) into the final confounding sets for multivariable analyses.

Table 1.

Baseline demographic and clinical characteristics of 1524 participants.

Characteristics	Total (n=1524)	≤48 h (n=447)	>48 h (n=1077)	P
Age, mean (SD), years	80.3 (7.8)	80.1 (8.1)	80.4 (7.7)	0.50
BMI, mean (SD), kg/m2	22.4 (3.4)	22.6 (3.1)	22.4 (3.4)	0.41
Gender, n (%)				0.15
Male	468 (30.7)	149 (33.3)	319 (29.6)
Female	1056 (69.3)	298 (66.7)	758 (70.4)
Married, n (%)				<0.001
No (never married, separated, divorced, or widowed)	244 (16.0)	43 (9.6)	201 (18.7)
Yes	1280 (84.0)	404 (90.4)	876 (81.3)
Smoking, n (%)				0.44
No	1468 (96.5)	433 (97.1)	1035 (96.3)
Yes	53 (3.5)	13 (2.9)	40 (3.7)
CCI, mean (SD)	4.4 (0.9)	4.3 (0.8)	4.5 (0.9)	<0.001
Fracture type, n (%)				<0.001
Femoral neck fracture	668 (43.8)	150 (33.6)	518 (48.1)
Intertrochanteric fractures	856 (56.2)	297 (66.4)	559 (51.9)
Fracture lateral, n (%)				0.93
Left	787 (51.6)	230 (51.5)	557 (51.7)
Right	737 (48.4)	217 (48.5)	520 (48.3)
ASA grade, n (%)				<0.001
≤2	1259 (82.6)	396 (88.6)	863 (80.1)
3	265 (17.4)	51 (11.4)	214 (19.9)
Time to surgery, mean (SD), h	85.5 (39.0)	43.9 (13.1)	100.8 (33.8)	<0.001
Time to admission, mean (SD), h	8.2 (6.2)	7.6 (6.5)	8.4 (6.0)	0.01
Surgical waiting time, mean (SD), h	75.0 (33.6)	36.5 (11.1)	89.5 (27.1)	<0.001
Surgery type, n (%)				<0.001
Internal fixation	937 (61.5)	325 (72.7)	612 (56.8)
Arthroplasty	587 (38.5)	122 (27.3)	465 (43.2)
Prefracture EQ-5D-3L, n (%)
Mobility (no problem)	1123 (79.4)	346 (84.2)	777 (77.4)	0.004
Self-care (no problem)	1274 (90.2)	377 (91.7)	897 (89.5)	0.21
Usual activities (no problem)	1261 (89.0)	374 (91.0)	887 (88.2)	0.12
Pain/discomfort (no problem)	1236 (87.6)	373 (91.0)	863 (86.2)	0.01
Anxiety/depression (no problem)	1322 (93.9)	386 (94.4)	936 (93.7)	0.63
Prefracture EQ-5D VAS, mean (SD)	89.2 (9.4)	91.9 (9.2)	88.1 (9.2)	<0.001
Prefracture HHS, mean (SD)	92.9 (10.3)	94.9 (8.8)	92.1 (10.7)	<0.001
Prefracture PMS, mean (SD)	7.6 (2.1)	7.7 (2.1)	7.5 (2.0)	0.09

CCI, Charlson Comorbidity Index; EQ-5D-3L, EuroQol 5 dimensions 3 levels; EQ-5D VAS, EuroQol 5 dimensions with Visual Analogue Scale; HHS, Hip Harris Score; PMS, Parker Mobility Score.

Figure 1.

Flowchart for participants including in the geriatric hip fracture cohort and this preplanned analysis. EQ-5D, EuroQol 5 dimensions; HHS, Harris Hip Score; PMS, Parker Mobility Score.

Surgical waiting time and poor functional outcomes at 12 months

Of 863 patients with available function scores at 12 months, 15.9% and 29.3% for HHS and PMS were defined as having poor outcomes. For EQ-5D VAS, the proportion was 44.9%. After adjusting for confounding sets, the results of log-binomial regression models showed receiving surgery after 48 h is associated with an increased risk of poor functional outcomes (HHS<80: RR=1.56, 95% CI: 1.00–2.51; PMS<7: RR=1.49, 95% CI: 1.13–2.01; EQ-5D VAS<80: RR=1.97, 95% CI: 1.57–2.47) (Table 2).

Table 2.

Multivariable analysis of receiving surgery after 48 h and poor functional outcomes at 12 months after surgery.

Function score	Binary outcome	N (%)	Crude RR (95% CI)	Adjusted RR (95% CI)	P
HHS	Poor: 0–79	137 (15.9)	2.03 (1.34–3.20)	1.56 (1.00–2.51)	0.05
	Good: 80–100	726 (84.1)	Ref.	Ref.
PMS	Poor: 0–6	252 (29.3)	1.88 (1.44–2.52)	1.49 (1.13–2.01)	0.01
	Good: 7–9	609 (70.7)	Ref.	Ref.
EQ-5D VAS	Poor: 0–79	381 (44.9)	2.18 (1.76–2.75)	1.97 (1.57–2.47)	<0.001
	Good: 80–100	467 (55.1)	Ref.	Ref.

EQ-5D VAS, EuroQol 5 dimensions with Visual Analogue Scale; HHS, Hip Harris Score; PMS, Parker Mobility Score; Ref., Reference; RR, relative risk.

Variables in confounding sets include age, marital status, Charlson Comorbidity Index (CCI), American Society of Anesthesiologists (ASA) grade, time to admission, prefracture function score, fracture type, and surgery method.

Surgical waiting time and function scores over time

The average function score decreased substantially at 3 months (HHS: 72.42, PMS: 4.50, EQ-5D VAS: 72.10), which then rebounded at 6 months (HHS: 81.55, PMS: 6.23, EQ-5D VAS: 76.83) and remained to recover slowly at 12 months (HHS: 85.75, PMS: 6.67, EQ-5D VAS: 78.75). To assess whether the strength of the associations was time-varying, an interaction term between surgical waiting time and time of observation was included in the multivariable models (Table 3). An invariant negative association between receiving surgery after 48 h and postoperative PMS (−0.44 units 95% CI: −0.70 to −0.18) was observed. However, we detected a P value of 0.02 and less than 0.001 for the interaction term, suggesting a time-variant association of receiving surgery after 48 h on postoperative HHS (3 months: −3.07 units 95% CI: −5.40 to −0.75; 6 months: −1.71 units 95% CI: −3.75 to 0.34; 12 months: −1.13 units 95% CI: −3.06 to 0.81) and EQ-5D VAS (3 months: −3.01 units 95% CI: −5.06 to −0.96; 6 months: −0.77 units 95% CI: −2.48 to 0.93; 12 months: −0.49 units 95% CI: −2.27 to 1.30).

Table 3.

Multivariable analysis of participants’ (N=863) postoperative functional scores for receiving surgery.

		≤48 h		>48 h
Function score	Time point after surgery	Mean (SD)	Estimated mean (SE)	Mean (SD)	Estimated mean (SE)	Adjusted MD (95% CI)	P
HHS	Time-invariant estimates	82.16 (10.85)	79.37 (0.73)	78.96 (13.43)	77.40 (0.59)	−1.97 (−3.21 to −0.73)	0.002
	Time-varying estimates						0.02
	3 months	75.37 (10.71)	72.61 (0.83)	71.10 (12.79)	69.53 (0.65)	−3.07 (−5.40 to −0.75)	0.002
	6 months	83.57 (9.07)	80.90 (0.74)	80.65 (12.14)	79.20 (0.63)	−1.71 (−3.75 to 0.34)	0.21
	12 months	87.39 (9.02)	84.59 (0.75)	85.01 (11.39)	83.46 (0.62)	−1.13 (−3.06 to 0.81)	1.00
PMS	Time-invariant estimates	6.19 (2.41)	5.51 (0.17)	5.63 (2.84)	5.07 (0.14)	−0.44 (−0.70 to −0.18)	0.001
	Time-varying estimates						0.91
	3 months	4.90 (2.58)	4.21 (0.20)	4.32 (2.95)	3.75 (0.15)	−0.46 (−1.00 to 0.08)	0.20
	6 months	6.60 (2.06)	5.93 (0.18)	6.07 (2.61)	5.51 (0.15)	−0.41 (−0.86 to 0.04)	0.11
	12 months	7.07 (1.98)	6.40 (0.17)	6.50 (2.46)	5.94 (0.15)	−0.45 (−0.88 to −0.02)	0.03
EQ-5D VAS	Time-invariant estimates	77.31 (8.70)	76.59 (0.60)	75.25 (10.40)	75.17 (0.44)	−1.42 (−2.44 to −0.41)	0.01
	Time-varying estimates						<0.001
	3 months	74.57 (9.07)	73.87 (0.70)	70.97 (11.96)	70.85 (0.53)	−3.01 (−5.06 to −0.96)	<0.001
	6 months	77.81 (8.03)	77.13 (0.64)	76.38 (9.05)	76.36 (0.48)	−0.77 (−2.48 to 0.93)	1.00
	12 months	79.56 (8.24)	78.78 (0.66)	78.38 (8.78)	78.29 (0.48)	−0.49 (−2.27 to 1.30)	1.00

EQ-5D VAS, EuroQol 5 dimensions with Visual Analogue Scale; HHS, Hip Harris Score; MD, mean differences; PMS, Parker Mobility Score.

The adjusted MD calculated by generalized estimating equation (GEE) models, including an interaction term between surgical waiting time and observation time (time-varying estimates), adjusting for time and confounding sets [age, marital status, Charlson Comorbidity Index (CCI), American Society of Anesthesiologists (ASA) grade, time to admission, prefracture function score, fracture type, and surgery method]. All P values derived from multiple-comparison used Bonferroni correction.

In addition, we explored the associations between surgical waiting time and specific HHS and detected the time-varying association was maximized at 3 months in hip motion range (−2.10 units 95% CI: −3.90 to −0.30) and hip function (−0.32 units 95% CI: −0.50 to −0.12) (Supplemental materials 3, Supplemental Digital Content 3, http://links.lww.com/JS9/A291). We also provided the specific levels in five domains of EQ-5D-3L in the two groups in each time point after surgery (Supplemental materials 4, Supplemental Digital Content 3, http://links.lww.com/JS9/A291).

The trajectories of the estimated marginal means of function scores at different time points were plotted after adjusting for the confounding sets (Fig. 2). Compared to patients with delayed surgery, those receiving surgery within 48 hours have more favorable trajectories.

Figure 2.

Modeled trajectories of function scores at estimated marginal means. (A) HHS; (B) PMS; (C) EQ-5D VAS. The modeled trajectories plotted four-time repeated measures (including prefracture status) in generalized estimating equations (GEEs) models, using the confounding-adjusted estimated marginal means with 95% CIs. Variables in confounding sets: age, marital status, Charlson Comorbidity Index (CCI), American Society of Anesthesiologists (ASA) grade, time to admission, prefracture function score, fracture type, and surgery method. Four time points: prefracture (recalled), 3 mo: 3 months after surgery, 6 mo: 6 months after surgery, 12 mo: 12 months after surgery. Blueline: receiving surgery in 48 h; redline: receiving surgery after 48 h. HHS, Hip Harris Score; EQ-5D VAS, EuroQol 5 dimensions with Visual Analogue Scale; PMS, Parker Mobility Score.

Sensitivity and subgroup analysis

We repeated all multivariable regression analyses on 30-imputed data sets using multivariate imputation by MICE for all 1524 enrolled participants and found similar results, suggesting that our findings were robust (Supplemental materials 5, Supplemental Digital Content 3, http://links.lww.com/JS9/A291). Subgroup analysis showed an increased risk of poor functional outcomes from short-delayed surgery (between 48 and 72 h from injury) to long-delayed surgery (after 72 h from injury). For EQ-5D VAS, the associations between delayed surgery and poor functional outcomes were obvious and stable (short-delayed: RR=1.80, 95% CI: 1.38–2.36; long-delayed: RR=2.07, 95% CI: 1.64–2.62). For HHS and PMS, statistically significant associations were only examined in long-delayed subgroup (HHS<80: RR=1.70, 95% CI: 1.08–2.77; PMS<7: RR=1.61, 95% CI: 1.22–2.20) (Supplemental materials 6, Supplemental Digital Content 3, http://links.lww.com/JS9/A291).

Discussion

In this multicenter geriatric hip fracture cohort, 54.2% (469/863) elderly patients suffered at least one poor functional outcome (HHS<80, PMS<7 or EQ. 5D VAS<80) at 12 months. Multivariable analyses indicated that surgical waiting time longer than 48 h was associated with poor functional outcomes and unfavored function trajectories. In addition, the negative association between delayed surgery and HHS was variant across time, especially on the specific domains of function and hip motion range.

Limited studies have investigated the impact of delayed surgery on function recovery in the elderly with hip fractures. Orosz et al. ¹¹. found that undergoing surgery longer than 24 h showed no effect on locomotion at 6 months, while Cohn et al. ¹⁵ found that surgery beyond 48 h compromises early short-term functional recovery regardless of patient comorbidity. Recently, a study has reported the effect of delayed hip replacement on postoperative HHS in 47-matched-elderly-patient-pairs with a femoral neck fracture¹³. Although restricted to the small sample size and the absence of potential confounders such as the prefracture HHS and ASA grade, the study still found that people with a delayed surgery (>21 days) recovered more slowly and had lower HHS than those who underwent surgery earlier in 12-month follow-up. A long waiting time for surgery is associated with delayed mobilization rehabilitation³⁶, resulting in muscle disuse of the affected hip and delayed postoperative functional recovery.

Consistent with these prior findings, our results support the negative associations between in-hospital waiting time and hip function recovery. We overcame some limitations of previous similar studies in confounding adjustment, which was absent of confounders such as prefracture function score and time to admission^13,37 or ignorance of the possible time-variant associations between delayed surgery and relative long-term functional recovery¹². After adjusting appropriately for confounders identified by DAGs, we found that in-hospital waiting time longer than 48 h is still associated with poor functional outcomes at postoperative 12 months, showing an increased risk of poor hip function (HHS<80, RR=1.83, 95% CI: 1.10–3.14; PMS<7, RR=1.89, 95% CI: 1.29–2.80) as well as poor general health status (EQ-5D VAS<80, RR=3.36, 95% CI: 2.38–4.79). Our data also suggest that in-hospital waiting time longer than 48 h would produce worse trajectories of function scores. The modeled trajectories of both surgical waiting time groups manifest slower after 6 months; this implies the importance of the short-term recovery period for older adults with hip fractures.

In our study, the time-invariant association of delayed surgery on PMS suggests that the mobility impairment due to longer in-hospital waiting time does not change over the first year after surgery. However, unlike the time-invariant impacts on PMS, the associations of HHS (function and hip motion range) and EQ-5D VAS were variant across 3, 6, and 12 months. The differences in recovery speed between the two groups were maximized in 3 months and then minimized with time. Notably, hip motion contributes to maintaining hip muscle strength. Previous studies showed that preoperative muscle strength affects functional recovery^38,39. For patients with delayed surgery, the reduction of hip motion resulted in the disuse of the musculature around the hip, which may account for the slow recovery speed of hip function and mobility. Similar results were found in specific EQ-5D-3L; the group with surgical waiting time longer than 48 h has a higher proportion of ‘severe problem’ after surgery in mobility (3 months: 35.7 vs. 19.8%; 6 months: 16.0 vs. 10.0%, 12 months: 12.5 vs. 6.0%) and lower proportion of ‘no problem’ in usual activities (3 months: 48.6 vs. 70.9%; 6 months: 80.3 vs. 87.0%; 12 months: 86.9 vs. 93.7%). Therefore, it may be helpful for patients whose surgery has to be delayed to apply the early intervention to avoid the slow functional recovery.

Several limitations of our study merit discussion. Firstly, various functional outcomes measures were used in the existing literatures^13,37; the absence of other general health indicators, such as preoperative activity level and activities of daily living may limit the extrapolation of our conclusions. More functional outcomes of general health status after hip fracture surgery need to be explored in the future. Secondly, in this preplanned analysis, our data were limited to survival patients who were contactable for follow-up. Multiple reasons may account for losing follow-ups, such as the frail elderly unable to attend clinics without additional caregivers or patients not knowing follow-up values, which inevitably leads to bias. In addition, of participants who completed the 12-month follow-up, 25.2% (290/1153) were dead or without available functional outcomes. Hence, we repeated the multivariable analyses using the MICE method (30-imputed data sets) for 1524 enrolled participants. We obtained similar results in the sensitivity analyses, suggesting the main findings are relatively robust. Further subgroup analysis found a trend of increasing risk from short-delayed surgery (within 72 h from injury) to long-delayed surgery (longer than 72 h from injury). The associations between long-delayed surgery and poor functional outcomes were obvious, with relatively narrow CIs and a P less than 0.05. For short-delayed surgery, the effects on hip function (HHS) and mobility (PMS) were close to 1 with a lower bound away from 1, suggesting the impairment of hip function and mobility from short-delayed surgery was limited. However, the effect on general health status (EQ-5D VAS) in short-delayed surgery was still obvious and stable (RR=1.80, 95% CI: 1.38–2.36). More attention should be paid to mechanisms behind the associations between delayed surgery on general health status.

Our findings provide helpful insight into the management of geriatric hip fractures. We highlight that delayed surgery of more than 48 h was associated with poor functional outcomes, including hip function, mobility, and general health status. Besides aiding clinicians, including physiotherapists, to make a better functional prognosis during triage for rehabilitation, our results underscore the importance of the early recovery of hip function in the first 3 months after surgery and call for more studies on possible interventions to maintain the strength of lower limb muscle in patients with delayed surgery.

In conclusion, the associations between in-hospital waiting time and postoperative functional score suggest that delayed surgery can lead to poor functional outcomes, especially in patients receiving surgery longer than 72 h from injury. Delayed surgery mainly impacted hip function and mobility recovery with a slower speed in early recovery of the first 3 months. Besides, more attention should be paid to mechanisms behind the associations between delayed surgery on general health status.

Ethical approval

The study was approved by the Ethics Committee of Shanghai Sixth People’s Hospital Affiliated to Shanghai Jiao Tong University School of Medicine, Approval No: 2016-43.

Sources of funding

This work was supported by the Shanghai Municipal Health Commission, Capacity building for multidisciplinary cooperation in diagnosis and treatment of elderly hip fractures, Shanghai Jiao Tong University School of Medicine (DLY201616), Shanghai Shen Kang Hospital Development Center, Clinical Research Plan of SHDC (SHDC2020CR6019), and Shanghai Sixth People's Hospital Affiliated to Shanghai Jiao Tong University School of Medicine (ynhg202004).

Author contribution

C.Z., S.C., and Q.C.: concept and design; W.J., X.L., H.H., Z.Y., X.C., Y.D., Q.W., B.K., and B.Q.: acquisition and quality control of data; Q.C., S.C., and K.F.: analysis and interpretation of data; K.F. and Q.C.: drafting of the manuscript; S.C., W.J., and X.L.: critical revision of the manuscript for important intellectual content; K.F. and Q.C.: statistical analysis; C.Z. and Q.C.: obtained funding; C.Z., Z.Y., X.C., Y.D., Q.W., and B.K.: supervision. Dr Changqing Zhang had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Drs Qianying Cai and Kai Fu share co-first authorship, and Drs Changqing Zhang and Shengbao Chen share co-senior authorship.

Conflicts of interest disclosure.

All authors declare no conflicts of interest and confirm that authors or their institutions have not received any payments or services in the past 36 months from a third party that could be perceived to influence, or give the appearance of potentially influencing, the submitted work.

Research registration unique identifying number (UIN)

1. Name of the registry: Chinese Clinical Trial Registry.

2. Unique identifying number or registration ID: ChiCTR-OOC-16008965.

3. Hyperlink to your specific registration (must be publicly accessible and will be checked): http://www.chictr.org.cn/showproj.aspx?proj=15064.

Guarantor

Changqing Zhang.

Provenance and peer review

Not commissioned, externally peer-reviewed.

Data availability statement

No additional data are available.

Supplementary Material

js9-109-1612-s001.docx ^{(34KB, docx)}

js9-109-1612-s002.pdf ^{(821.1KB, pdf)}

js9-109-1612-s003.docx ^{(1.4MB, docx)}