Timing of surgical intervention following traumatic femoral fractures in Northern Tanzania: a hospital-based retrospective cohort study

Abstract

Objectives

Delayed surgical fixation of femoral fractures is associated with increased morbidity and mortality; however, delays remain common in low-income and middle-income countries. This study aimed to determine time to surgery and identify factors associated with time to surgery among patients with traumatic femoral fractures at a tertiary referral hospital in northern Tanzania.

Design

Retrospective cohort study using routinely collected data.

Setting

Kilimanjaro Christian Medical Centre, a zonal referral hospital in northern Tanzania.

Participants

2040 patients with traumatic femoral fractures treated surgically between January 2018 and December 2024. Patients with pathological fractures, femoral neck fractures treated with total hip replacement, prior femoral fractures, revision surgeries or spinal cord injuries were excluded.

Outcome measure

The primary outcome was time to surgery, defined as the number of days from injury to surgery. Weibull proportional hazards regression models were used to identify predictors of time to surgery, with directed acyclic graphs informing covariate selection. We examined 11 pre-specified factors potentially associated with time to surgery, including health insurance, co-morbidities, haemoglobin level (g/dL), associated injuries, American Society of Anesthesiologists (ASA) class, fracture location, fracture morphology, fracture status, implant type, mode of implant application and anaesthesia type.

Results

The median (IQR) time to surgery among 2040 patients was 14 (7–24) days. Of the 11 factors examined, 7 were associated with time to surgery, 2 had no clear association and 2 showed no evidence of an association. Factors associated with shorter time to surgery included health insurance coverage (HR 1.83, 95% CI 1.71 to 1.96), higher haemoglobin levels (HR per unit increase, 1.02, 95% CI 1.01 to 1.04), middle and distal fracture locations, simpler fracture morphology and specific implant types. Factors associated with longer delays included associated injuries (HR 0.87, 95% CI 0.79 to 0.96), open fractures (HR 0.82, 95% CI 0.71 to 0.95) and higher ASA classes (HR range 0.64–0.88 for classes II–IV).

Conclusions

Time to surgical fixation following femoral fractures was prolonged, with a median delay of 14 days. This study identified factors that may be related to surgical timing, although causality could not be firmly demonstrated. Interventions aimed at improving financial access to care and perioperative optimisation may reduce delays in this setting.

STRENGTHS AND LIMITATIONS OF THIS STUDY.

• This hospital-based retrospective cohort study modelled time to surgery using a wide range of characteristics.

• The study included children, adolescents and adults with traumatic femoral fractures over 7 years, providing findings that are relevant to trauma care in Tanzania and similar low-income and middle-income country settings.

• The use of a directed acyclic graph-informed covariate selection strengthened the analytical approach by identifying a minimally sufficient adjustment set for confounding.

• Important health system factors, including operating theatre availability, surgeon staffing and implant stock-outs were not measured, and patients with missing time-to-surgery data were excluded, which may have introduced selection bias.

• As this was a single-centre, hospital-based study conducted at a tertiary referral hospital that manages a high proportion of patients with severe injuries, findings may not be generalisable to other healthcare settings.

Introduction

Femoral fractures represent a substantial and growing global health burden. Between one and three million femoral fractures occur worldwide each year, with the highest incidence observed among children aged 5–14 years and young to middle-aged adults aged 15–59 years.¹ Low-income and middle-income countries (LMICs) bear a disproportionate share of this burden, with reported incidence rates ranging from 15.5 to 44.8 per 100 000 population.¹ Projections suggest that the global incidence of femoral fractures will rise to approximately 6 million cases annually by 2050, with nearly 70% occurring in Asia, Latin America, the Middle East and Africa.²

In sub-Saharan Africa (SSA), the incidence of femoral fractures is increasing, reflecting rapid urbanisation, motorisation and limited injury-prevention infrastructure. In South Africa, femoral fractures accounted for approximately 19.3 cases per 100 000 population per year, with femoral neck fractures in young adults often related to high-energy trauma, constituting 3%–10% of all hip fractures.^{3 4} Studies from West Africa indicate that femoral fractures are the second most common fracture type among patients with multiple injuries.^{5 6} Across East Africa, road traffic accidents and high-energy falls predominate, particularly among socio-economically active young adults, as reported in Kenya, Malawi and northern Uganda.^7–9 In Tanzania, the reported incidence of femoral shaft fractures ranges from 2.1 to 18.4 per 100 000 population per year, with road traffic accidents, particularly those involving motorcycles, as the principal cause.¹⁰ Femoral fractures constitute approximately 39% of orthopaedic admissions at Kilimanjaro Christian Medical Centre (KCMC), disproportionately affecting young adults and predominantly resulting from road traffic accidents and falls.¹¹

Femoral fractures are associated with substantial morbidity and mortality across all age groups, including both short-term and long-term disability.^12–14 These injuries may be accompanied by life-threatening complications. Systemic complications include thromboembolism, pulmonary compromise and haemodynamic instability.^{12 15} Management strategies include non-operative treatment with skeletal traction and surgical fixation, with intramedullary nailing widely regarded as the gold standard for definitive fixation.^{16 17}

A robust body of evidence demonstrates that early definitive fixation, typically within 24–48 hours of admission, is associated with reduced complications, lower mortality and shorter hospital length of stay.^18–23 Despite these benefits, delays to surgery persist globally and are influenced by patient-related factors such as age, comorbidities and injury severity, as well as system-level factors including theatre availability, hospital capacity and prioritisation protocols.^24–26 While timely surgery is routinely achieved in high-income countries,^{15 22 25 27} patients in LMICs experience substantial delays. Fewer than 13% of patients receive definitive fixation within 48 hours, and more than three-quarters undergo surgery after 72 hours,²⁶26 with reported delays exceeding 11 days in Uganda.²⁸

These delays are driven by multifactorial barriers, including delayed care-seeking, transportation challenges, referral inefficiencies and resource constraints.^{23 29} Despite the high burden of femoral fractures in Tanzania, determinants of surgical timing remain poorly characterised in East Africa. Most existing studies are descriptive and do not explicitly address causal relationships and time to surgery. In this study, we investigated 11 pre-specified factors of interest that could potentially be associated with time to surgery based on clinical considerations and an a priori causal framework. These factors included having health insurance, comorbidities, haemoglobin level, associated injuries, American Society of Anesthesiologists (ASA) classification, fracture location, fracture status, fracture morphology, implant type, implant application and anaesthesia type. We used a directed acyclic graph (DAG) to identify appropriate covariates for adjustment and minimise confounding. Therefore, this study aimed to assess the association between these pre-specified factors and time from injury to surgery among patients with traumatic femoral fractures treated at KCMC zonal referral hospital, northern Tanzania, between 2018 and 2024.

Methods

Study design and settings

We conducted a retrospective cohort study using routinely collected clinical cohort data collected between 1 January 2018 and 31 December 2024 at KCMC zonal referral and consultant hospital located in Kilimanjaro region in Moshi Municipality, northern Tanzania. The region has a total of seven districts, including the Moshi Municipality and Moshi rural district. As a major referral centre for trauma care, KCMC serves patients from the Kilimanjaro and neighbouring regions, including Tanga, Arusha, Manyara and Singida. This means that the main catchment area of the hospital is the local population.

Study population

The study population comprised all patients with traumatic femoral fractures who underwent surgery at KCMC during the study period. Patients were eligible for inclusion if they sustained a traumatic femoral fracture and received operative management at the institution. Patients with missing information on time to surgery, pathological femoral fractures, femoral neck fractures managed with total hip replacement, a prior history of femoral fractures or femoral fractures associated with spinal cord injury were excluded (figure 1).

Figure 1. Total flowchart of the study population.

Data collection and study procedures

Following admission to the orthopaedic wards, patients with femoral fractures were identified by clinical physiotherapists through hospital records. Data were collected by physiotherapists while the patient was in the orthopaedic ward. Data were collected by a paper-based questionnaire on sociodemographic characteristics, injury-related factors, hospital admission details and treatment characteristics. Data were initially recorded on standardised paper case report forms and subsequently entered into an electronic database. For potentially sensitive variables, such as income, participants were given the option to decline disclosure.

Measures

The primary outcome was time to surgery, defined as the interval between the date of injury (the earliest date of injury on 1 January 2018) and the date of definitive surgical fixation, measured in days (the latest date of surgical fixation on 22 December 2024). The date of injury was a calendar date on which the injury occurred, as documented in the patient’s electronic record or reported by the patient. The date of surgery was the calendar date on which the definitive surgical procedure was performed, as recorded in the operation notes. Survival time was defined as the number of days from the date of injury (time zero) to the date of surgery, calculated by subtracting the date of injury from the date of surgery.

The primary exposure variables were selected using DAGs. The following information was collected: health insurance (no or yes), comorbidities (no or yes), haemoglobin level (g/dL), associated injury (no or yes), ASA class (I, healthy patient; II, mild systemic disease; III, severe systemic disease; IV, severe systemic disease that is a constant threat to life) assigned by anaesthesiologists and obtained from patients’ electronic records and fracture location (proximal, mid and distal) and fracture morphology (transverse, comminuted, spiral, oblique and segmental) obtained from radiological examination (X-ray) and reported by radiologists in the patients’ electronic records; Arbeitsgemeinschaft für Osteosynthesefragen/Orthopaedic Trauma Association classification was not used. Fracture status (open or closed) and fracture side (right or left) were obtained from physical clinical examination. Primary exposures included the type of intramedullary implant (proximal femoral nails, S & K nails, plates and screws, hemiarthroplasty, dynamic hip screws (DHS) and cannulated hip screws (CHS) and distal condylar plates (DCP)), mode of implant application (antegrade, retrograde, lateral incision or anterolateral incision) and anaesthesia type (spinal or general) obtained from patients’ electronic records reported by the anaesthesiologist and orthopaedic surgeon who performed the surgery. Preoperative anaesthetic assessment was performed to determine fitness for surgery and the anaesthetic plan; patients requiring medical optimisation may experience delays. Similarly, following admission and radiographic assessment, the treating surgeons determined the appropriate implant required before surgery. Implant availability, cost and insurance coverage could therefore influence the timing of surgery. Potentially confounding variables included age in years, sex (male or female), education (none, primary, secondary or tertiary), residence (rural or urban), marital status (married or not married), employment (employed or not employed), income per month (US$) and mechanism of injury (motor vehicle crash, fall from height or assault) and nutritional status. Body mass index (BMI) was used to assess nutritional status, calculated as weight (kg) divided by height squared (m²) for adults (≥20 years); for children aged <10 years, BMI-for-age z-scores were derived using WHO Anthro and for adolescents aged 10–19 years using WHO AnthroPlus. BMI was categorised as underweight, normal weight, overweight and obesity.

Directed acyclic graph–based confounder selection

A DAG was developed a priori based on published evidence on the determinants of the time to surgery, previous studies and clinical knowledge to represent the hypothesised causal relationships among patient, clinical and health system factors and time to surgery.^{30 31} The DAG was constructed using the DAGitty package in R V.4.4.2 and was used to identify the minimally sufficient adjustment sets required to estimate the total effect of each primary exposure while minimising confounding, over-adjustment and adjustment for mediators or colliders.³² A simplified DAG can show the hypothesised causal relationships relevant to confounder selection. Some background causal relationships between covariates were omitted for visual clarity (figure 2). Separate regression models were then fitted using the appropriate adjustment set for each exposure.

Figure 2. Plot of the relationship between variables (DAG). This DAG is a simplified representation and that some background relationships were omitted for visual clarity. ASA, American Society of Anesthesiologists; DAG, directed acyclic graph.

Because the relationship between variables differed across exposures, each primary exposure had its minimally sufficient adjustment set. Potential confounders identified by the DAG included age, residence, income, mechanism of injury, associated injuries, fracture morphology, fracture location, implant type and ASA class, depending on the exposure evaluated. For exposures with more than one minimally sufficient adjustment set (eg, associated injuries, ASA class, fracture location, fracture morphology, implant type and anaesthesia type), sensitivity analyses were performed using alternative adjustment sets, and the model with the lowest Akaike information criterion (AIC) and Bayesian information criterion (BIC) was selected. For comparison, a conventional multivariable model was also fitted, including all primary exposures and covariates.

Statistical analysis

All statistical analyses were performed using Stata V.15 (StataCorp, College Station, TX, USA). Time to surgery was summarised for the overall study population using the median and IQR owing to its non-normal distribution. The median time to surgery with corresponding IQRs was also estimated across categories of explanatory variables. Frequencies and proportions summarised categorical variables. Continuous variables were summarised using mean (SD) for symmetrical variables and median (IQR) for non-symmetrical variables. The entry date was the date of injury, and the exit date was the date of surgery. Surgery rates were calculated as the number of surgeries divided by the total person-time at risk and expressed per 100 person-days of follow-up. Surgery rates were also estimated across exposure groups for comparison. Survival analysis was conducted using the Cox proportional hazards (PH) model and three parametric models: exponential, Weibull and Gompertz. The PH assumption for the Cox models was assessed using Schoenfeld residuals; both global and covariate-specific tests were examined. AIC and BIC were used for model selection, specifically between the Cox PH models, exponential models, Weibull PH models and Gompertz models. On average, Cox models had AIC=23 124.28 and BIC=25 314.72; exponential models, AIC=5692.19 and BIC=5724.54; Weibull models, AIC=5592.21 and BIC=5630.11; and Gompertz, AIC=5661.66 and BIC=5699.55. Weibull PH models had the lowest AIC and BIC and thus were used in survival analysis modelling. The PH assumption in the Weibull model was assessed using time-dependent covariates by including interactions between predictors and log(time) and evaluating them with Wald tests. There was no evidence of violation of the PH assumption (Wald test, p>0.05). Univariable and multivariable Weibull regression models were fitted to estimate crude and adjusted HRs and the corresponding 95% CIs. In the multivariable models, each model included the primary exposure of interest and the corresponding DAG-derived adjustment variables. No data-driven variable selection procedures were used. Variables included in the DAG were considered according to their temporal relationship with the outcome and the clinical pathway leading from injury to surgery. Missing data were handled through complete case analysis (observations with missing values were excluded from regression models). Given the low proportion of missing data (<5%), this approach was considered appropriate.³³ All statistical tests were two-sided, and statistical significance was assessed at an alpha level of 0.05.

Patient and public involvement

None.

Results

Participant characteristics

A total of 2040 femoral fractures were analysed. The median (IQR) age of the study participants was 43.5 (27–63) years, 1496 (73.3%) were males, 1052 (52.5%) had primary education, 1429 (70.6%) were residing in rural areas, 1161 (57.7%) were employed, 443 (44.2%) were insured and 236 (12.9%) were overweight. Of all participants, 1135 (55.7%) were involved in a motor vehicle crash. The median time from injury to surgery was 14 (IQR 7–24) days. Overall, 80 (3.9%) of femoral fractures underwent surgery within 2 days, 160 (7.8%) within 3 days, 550 (26.9%) within 7 days and 1490 (73.1%) more than 7 days after injury. The overall surgery rate was 4.81 (95% CI 4.60 to 5.02) per 100 person-days. The median time to surgery was 15 days in those aged >64 years, 13 days in females, 11 days in those with secondary or higher education, 14 days in those residing in rural areas and 11 days in those with health insurance. In the crude analysis, those aged 3–14 years (HR 1.54, 95% CI 1.22 to 1.96), those with higher income (HR 0.99, 95% CI 0.99 to 1.00) and those with insurance (HR 1.85, 95% CI 1.76 to 1.95) had higher hazards of surgery (table 1).

Table 1. Baseline characteristics, median time to surgery, surgery rate and univariate analysis of time-to-surgery determinants in traumatic femoral fractures (n=2040).

Characteristics	Total (%)	Median time to surgery (IQR) (days)	Surgery rate (95% CI) per 100 person-days	Crude HR (95% CI)	P value
Age (years)
Median (IQR)	43.5 (27.0–63.0)
3–14	79 (3.9)	9 (6–17)	7.96 (6.38 to 9.93)	1.54 (1.22 to 1.96)	<0.001
15–24	313 (15.3)	12 (7–24)	4.55 (4.08 to 5.09)	0.86 (0.75 to 0.99)	0.048
25–44	651 (31.9)	13 (7–25)	5.00 (4.63 to 5.40)	0.96 (0.84 to 1.08)	0.465
45–64	529 (25.9)	17 (8–27)	4.19 (3.86 to 4.57)	0.79 (0.70 to 0.90)	<0.001
>64	468 (22.9)	15 (8–22)	5.21 (4.76 to 5.71)	1.00 (ref)
Per 1-year increase				0.98 (0.91 to 0.99)	0.038
Sex
Male	1496 (73.3)	14 (7–24)	4.71 (4.48 to 4.95)	1.00 (ref)
Female	544 (26.7)	13 (6–25)	5.09 (4.68 to 5.54)	1.08 (0.98 to 1.19)	0.109
Educational level
None	299 (14.9)	16 (9–24)	4.95 (4.42 to 5.55)	1.00 (ref)
Primary	1052 (52.5)	15 (7–25)	4.60 (4.33 to 4.89)	0.93 (0.82 to 1.06)	0.255
Secondary/tertiary	652 (32.6)	11 (7–24)	5.07 (4.69 to 5.47)	1.02 (0.89 to 1.17)	0.762
Missing=37
Residence
Rural	1429 (70.6)	14 (7–24)	4.73 (4.49 to 4.98)	1.00 (ref)
Urban	595 (29.4)	13 (7–25)	4.95 (4.56 to 5.36)	1.05 (0.95 to 1.15)	0.358
Missing=16
Employment status
Employed	1161 (57.7)	14 (6–24)	4.71 (4.45 to 4.99)	1.00 (ref)
Not employed	852 (42.3)	14 (8–24)	4.92 (4.60 to 5.26)	1.05 (0.96 to 1.14)	0.319
Missing=27
Income/month (US$)
Median (IQR)	60 (20–160)
≤60	621 (55.8)	15 (8–25)	5.17 (4.78 to 5.59)	1.00 (ref)
>60	492 (44.2)	15 (7–28)	3.69 (3.38 to 4.04)	0.71 (0.63 to 0.81)	<0.001
Per 1 US$ increase				0.99 (0.99 to 1.00)	0.025
Missing=7
Health insurance
No	1590 (78.2)	15 (7–25)	4.97 (4.74 to 5.22)	1.00 (ref)
Yes	443 (21.8)	11 (6–20)	4.28 (3.90 to 4.69)	1.85 (1.76 to 1.95)	0.003
Missing=7
Nutritional status (BMI)
Normal	1191 (65.2)	14 (7–23)	5.19 (4.91 to 5.50)	1.00 (ref)
Underweight	334 (18.3)	16 (8–29)	4.14 (3.72 to 4.61)	0.79 (0.70 to 0.89)	<0.001
Overweight	236 (12.9)	11 (6–23)	4.54 (4.00 to 5.17)	0.87 (0.76 to 1.00)	0.056
Obesity	65 (3.6)	11 (5–20)	6.08 (4.77 to 7.75)	1.17 (0.91 to 1.50)	0.213
Missing=214
Mechanism of injury
MVC	1135 (55.7)	14 (7–25)	4.66 (4.40 to 4.94)	1.00 (ref)
Fall from height	829 (40.6)	14 (7–23)	5.01 (4.67 to 5.36)	1.07 (0.98 to 1.18)	0.114
Assault	76 (3.7)	12 (8–25)	4.89 (3.91 to 6.12)	1.05 (0.83 to 1.32)	0.673

BMI, body mass index; MVC, motor vehicle crash.

Clinical characteristics

Table 2 presents the clinical characteristics of the study population. 256 (12.6%) had comorbidities, the mean (SD) haemoglobin level was 11.9 (2.1) g/dL, 812 (39.8%) had associated injuries, 1046 (53.1%) were ASA class I, 1022 (50.5%) had proximal femoral fractures, 1814 (89.7%) had closed femoral fractures, 652 (32.3%) had transverse femoral fractures, 1060 (52.4%) were managed surgically by SIGN and Küntscher (S & K) nails, 1239 (61.4%) had antegrade nailing and 1796 (89.5%) had spinal anaesthesia. The median time to surgery was 16 days with comorbidities, 17 days with anaemia, 16 days with other injuries, 25 days with ASA class IV, 15 days with proximal femoral fractures and 19 days with segmental femoral fractures, 17 days in those who underwent hemiarthroplasty and 16 days in those who had anterolateral incision of the femoral fracture. In the crude analysis, higher haemoglobin level; mid-femoral, distal, spiral and oblique femoral fractures; S & K nailing; plates and screws; hemiarthroplasty; DHS and CHS; retrograde nailing; and lateral incision were associated with earlier surgery, whereas ASA class (II–IV) and open fractures were associated with delayed surgery.

Table 2. Clinical characteristics, median time to surgery, surgery rate and univariate analysis of time-to-surgery determinants in traumatic femoral fractures (n=2040).

Characteristics	Total (%)	Median time to surgery (IQR) (days)	Surgery rate (95% CI) per 100 person-days	Crude HR (95% CI)	P value
Comorbidities
No	1784 (87.4)	14 (7–24)	4.77 (4.55 to 4.99)	1.00 (ref)
Yes	256 (12.6)	16 (8–24)	5.09 (4.51 to 5.76)	1.07 (0.94 to 1.22)	0.299
Haemoglobin level (g/dL)
Mean (SD)	11.9 (2.1)
Normal	990 (50.5)	12 (6–21)	4.72 (4.44 to 5.03)	1.00 (ref)
Anaemic	971 (49.5)	17 (9–25)	4.84 (4.54 to 5.15)	1.02 (0.94 to 1.12)	0.589
Per 1 g/dL increase				1.02 (1.01 to 1.04)	0.034
Missing=79
Associated injuries
No	1228 (60.2)	13 (7–22)	4.84 (4.58 to 5.12)	1.00 (ref)
Yes	812 (39.8)	16 (8–27)	4.75 (4.43 to 5.09)	0.98 (0.89 to 1.07)	0.676
ASA classification
I	1046 (53.1)	12 (6–21)	5.19 (4.88 to 5.51)	1.00 (ref)
II	696 (35.4)	17 (8–26)	4.49 (4.18 to 4.84)	0.87 (0.79 to 0.95)	0.003
III	182 (9.2)	17 (10–28)	4.33 (3.74 to 5.01)	0.83 (0.71 to 0.98)	0.025
IV	45 (2.3)	25 (12–35)	3.59 (2.69 to 4.82)	0.69 (0.51 to 0.93)	0.016
Missing=71
Fracture location
Proximal	1022 (50.5)	15 (8–25)	4.23 (3.98 to 4.49)	1.00 (ref)
Middle	430 (21.3)	12 (7–22)	5.96 (5.42 to 6.55)	1.43 (1.27 to 1.59)	<0.001
Distal	570 (28.2)	14 (6–24)	5.26 (4.84 to 5.71)	1.25 (1.13 to 1.39)	<0.001
Missing=18
Fracture status
Closed	1814 (89.7)	13 (7–23)	4.89 (4.68 to 5.13)	1.00 (ref)
Open	208 (10.3)	22 (15–33)	4.02 (3.51 to 4.61)	0.82 (0.71 to 0.95)	0.008
Missing=18
Fracture morphology
Transverse	652 (32.3)	13 (7–23)	4.47 (4.14 to 4.83)	1.00 (ref)
Comminuted	712 (35.3)	16 (7–25)	4.70 (4.37 to 5.06)	1.06 (0.95 to 1.18)	0.311
Spiral	197 (9.8)	12 (6–23)	6.22 (5.41 to 7.15)	1.41 (1.20 to 1.66)	<0.001
Oblique	363 (18.0)	12 (7–22)	5.36 (4.83 to 5.94)	1.21 (1.06 to 1.37)	0.004
Segmental	93 (4.6)	19 (11–29)	3.74 (3.06 to 4.59)	0.84 (0.67 to 1.04)	0.114
Missing=23
Implant type
PFN	528 (26.1)	15 (8–23)	3.84 (3.53 to 4.18)	1.00 (ref)
S & K nails	1060 (52.4)	13 (7–24)	5.25 (4.94 to 5.57)	1.39 (1.26 to 1.55)	<0.001
Plates and screws	89 (4.4)	10 (6–18)	7.61 (6.19 to 9.37)	2.06 (1.64 to 2.59)	<0.001
Hemiarthroplasty	136 (6.7)	17 (9–34)	4.66 (3.94 to 5.51)	1.24 (1.03 to 1.49)	0.026
DHS and CHS	96 (4.7)	15 (5–27)	5.90 (4.83 to 7.21)	1.58 (1.27 to 1.97)	<0.001
DCP	113 (5.6)	16 (7–28)	4.49 (3.74 to 5.41)	1.19 (0.97 to 1.46)	0.098
Missing=18
Implant application
Antegrade	1239 (61.4)	15 (8–25)	4.37 (4.13 to 4.62)	1.00 (ref)
Retrograde	583 (28.9)	12 (6–23)	5.73 (5.28 to 6.21)	1.33 (1.20 to 1.46)	<0.001
Lateral incision	81 (4.0)	10 (5–18)	7.89 (6.34 to 9.81)	1.85 (1.47 to 2.31)	<0.001
Anterolateral incision	115 (5.7)	16 (6–28)	4.52 (3.77 to 5.43)	1.04 (0.86 to 1.26)	0.694
Missing=22
Anaesthesia type
General	210 (10.5)	14 (8–30)	4.27 (3.73 to 4.89)	1.00 (ref)
Spinal	1796 (89.5)	14 (7–24)	4.92 (4.69 to 5.15)	1.15 (0.99 to 1.33)	0.052
Missing=34

Comorbidities included hypertension, diabetes mellitus and chronic kidney disease.

ASA, American Society of Anesthesiologists; CHS, cannulated hip screws; DCP, distal condylar plates; DHS, dynamic hip screws; PFN, proximal femoral nails; S & K nails, SIGN and Küntscher nails.

Factors associated with time to surgery for femoral fractures

Findings from the multivariable Weibull PH regression analysis based on the minimal adjustment sets from the DAG for each primary exposure of surgical timing in traumatic femoral fractures are shown in table 3. Adjusted Weibull regression analysis revealed that patients with insurance (HR 1.83, 95% CI 1.71 to 1.96), higher haemoglobin levels (HR 1.02, 95% CI 1.01 to 1.04), mid-femoral fractures (HR 1.49, 95% CI 1.32 to 1.69), distal femoral fractures (HR 1.28, 95% CI 1.14 to 1.43), spiral fractures (HR 1.36, 95% CI 1.16 to 1.61), oblique fractures (HR 1.26, 95% CI 1.10 to 1.44), implant types such as S & K nails (HR 1.55, 95% CI 1.37 to 1.75), plates and screws (HR 2.29, 95% CI 1.81 to 2.91), hemiarthroplasty (HR 1.34, 95% CI 1.10 to 1.64), DHS and CHS (HR 1.73, 95% CI 1.38 to 2.17), DCP (HR 1.27, 95% CI 1.03 to 1.57) and retrograde implant application (HR 1.20, 95% CI 1.07 to 1.36) were associated with earlier surgery. In contrast, patients with associated/other injuries (HR 0.87, 95% CI 0.79 to 0.96), open femoral fractures (HR 0.82, 95% CI 0.71 to 0.95) and increasing ASA classification with ASA classes II (HR 0.88, 95% CI 0.80 to 0.97), III (HR 0.79, 95% CI 0.67 to 0.93), and IV (HR 0.64, 95% CI 0.47 to 0.87) were associated with delayed surgery indicating longer waiting times.

Table 3. Weibull proportional hazards regression analysis of the time-to-surgery determinants for traumatic femoral fractures at KCMC zonal referral hospital (n=2040).

Variables	Multivariable Weibull PH model*		Multivariable Weibull PH models with minimal covariate adjustments based on DAG analysis†
	HR (95% CI)	P value	Adjustment set	HR (95% CI)	P value
Health insurance			Income and place of residence
No	1.00 (ref)			1.00 (ref)
Yes	0.77 (0.62 to 0.95)	0.012		1.83 (1.71 to 1.96)	0.013
Comorbidities			Age (years)
No	1.00 (ref)			1.00 (ref)
Yes	1.29 (0.97 to 1.71)	0.077		1.09 (0.95 to 1.26)	0.195
Haemoglobin level (g/dL)	1.02 (0.99 to 1.06)	0.229	Associated injury and fracture morphology	1.02 (1.01 to 1.04)	0.016
Associated injuries			ASA class, fracture morphology and implant type
No	1.00 (ref)			1.00 (ref)
Yes	0.89 (0.75 to 1.07)	0.241		0.87 (0.79 to 0.96)	0.007
ASA classification			Associated injury, fracture morphology and implant type
I	1.00 (ref)			1.00 (ref)
II	0.85 (0.73 to 0.99)	0.05		0.88 (0.80 to 0.97)	0.013
III	0.75 (0.57 to 0.98)	0.038		0.79 (0.67 to 0.93)	0.004
IV	0.42 (0.23 to 0.79)	0.008		0.64 (0.47 to 0.87)	0.004
Fracture location			Associated injury and fracture morphology
Proximal	1.00 (ref)			1.00 (ref)
Middle	1.07 (0.81 to 1.40)	0.64		1.49 (1.32 to 1.69)	<0.001
Distal	1.29 (0.95 to 1.76)	0.096		1.28 (1.14 to 1.43)	<0.001
Fracture status			No adjustment set selected
Closed	1.00 (ref)			1.00 (ref)
Open	0.86 (0.67 to 1.09)	0.201		0.82 (0.71 to 0.95)	0.008
Fracture morphology			ASA class, associated injury and fracture location
Transverse	1.00 (ref)			1.00 (ref)
Comminuted	1.15 (0.97 to 1.36)	0.115		1.05 (0.94 to 1.18)	0.358
Spiral	1.44 (1.09 to 1.91)	0.01		1.36 (1.16 to 1.61)	<0.001
Oblique	1.58 (1.26 to 1.98)	<0.001		1.26 (1.10 to 1.44)	0.001
Segmental	0.71 (0.51 to 0.98)	0.042		0.86 (0.69 to 1.08)	0.2
Implant type			ASA class, associated injury and fracture morphology
PFN	1.00 (ref)			1.00 (ref)
S & K nails	1.69 (1.31 to 2.19)	<0.001		1.55 (1.37 to 1.75)	<0.001
Plates and screws	2.54 (0.89 to 7.29)	0.082		2.29 (1.81 to 2.91)	<0.001
Hemiarthroplasty	1.28 (0.92 to 1.78)	0.147		1.34 (1.10 to 1.64)	0.004
DHS and CHS	1.97 (1.41 to 2.77)	<0.001		1.73 (1.38 to 2.17)	<0.001
DCP	1.93 (0.19 to 18.75)	0.572		1.27 (1.03 to 1.57)	0.027
Implant application			Implant type
Antegrade	1.00 (ref)			1.00 (ref)
Retrograde	1.03 (0.81 to 1.31)	0.821		1.20 (1.07 to 1.36)	0.002
Lateral incision	1.32 (0.38 to 4.49)	0.662		1.67 (0.73 to 3.84)	0.227
Anterolateral incision	0.84 (0.09 to 8.08)	0.877		1.46 (0.29 to 7.23)	0.645
Anaesthesia type			ASA class
General	1.00 (ref)			1.00 (ref)
Spinal	0.89 (0.68 to 1.17)	0.418		1.08 (0.93 to 1.26)	0.291

Comorbidities included hypertension, diabetes mellitus and chronic kidney disease.

^*Multivariable Weibull proportional hazards regression analysis adjusted for all covariates/variables regardless of the adjustment sets.

^†Multivariable Weibull proportional hazards models adjusted for the minimal sufficient adjustment set based on the direct acyclic diagram analysis.

ASA, American Society of Anesthesiologists; CHS, cannulated hip screws; DCP, distal condylar plate; DHS, dynamic hip screws; KCMC, Kilimanjaro Christian Medical Centre; PFN, proximal femoral nail; PH, proportional hazards; S & K nail, SIGN and Küntscher nail.

On the classical analysis, health insurance was associated with delay in surgery (HR 0.77), whereas on the DAG-based analysis, it was associated with earlier surgery (HR 1.83). In both analyses, haemoglobin was associated with earlier surgery; it was not significant in the classical analysis (p=0.229) but was significant in the DAG-based analysis. Associated injuries were associated with delayed surgery in both analyses; although they were not significant in the classical analysis, they were significant in the DAG-based analysis. Fracture location and status were not significant in the classical analysis but were significant in the DAG-based analysis. Segmental fractures were associated with delayed surgery in both analyses; they were significant in the classical analysis but were not significant in the DAG-based analysis. Plates and screws, hemiarthroplasty and DCP were not significant in the classical analysis but were significant in the DAG-based analysis. In the classical analysis, the mode of implantation was not associated with time to surgery; however, retrograde nailing was associated with earlier surgery in the DAG-based analysis. The anaesthesia type was not associated with time to surgery in both analyses. Results based on the classical analysis (adjusting for variables that are not confounders) may lead to biased estimates.

Discussion

This study assessed the time to surgery and factors associated with time to surgery among patients with traumatic femoral fractures treated at KCMC zonal referral hospital in Northern Tanzania between 2018 and 2024. The median time to surgery was 14 days. Factors associated with earlier surgery included health insurance coverage, higher haemoglobin levels, middle and distal fracture locations, simpler fracture morphology and specific implant types, whereas associated injuries, open fractures and higher ASA class were associated with longer time to surgery.

These findings indicate substantial delays compared with the 24–48 hours often recommended for early fixation of femoral fractures in settings with established trauma systems.^{20 21} These findings also highlight that surgical timing in this setting reflects a complex interaction among patient condition, injury characteristics and health system factors. The observed median time to surgery was comparable to delays reported from other LMICs. A study conducted in Uganda on orthopaedic surgical task-shifting and its impact on time to surgery in western Uganda reported 11 days as that time to surgery.²⁶26 Although dDelayed fracture fixation is associated with increased morbidity, prolonged disability and excess mortality among adults and older patients,^{26 29} the consequences of delayed fixation in the present cohort could not be directly evaluated because outcomes such as mortality and major complications were not recorded. The prolonged waiting time observed in this study reflects the challenges of providing timely orthopaedic trauma care in a setting where the demand for surgical services continues to increase alongside limitations in healthcare resources.

An important finding of this study is that patients with more complex clinical presentations appeared to experience longer delays despite potentially greater clinical need for timely fixation. Patients with associated injuries, open fractures and higher ASA class had longer times to surgery, likely reflecting the need for additional resuscitation, optimisation, specialist assessment and infection prevention measures before definitive fixation. This creates an important clinical challenge; patients who may benefit substantially from early stabilisation may also require more preparation before surgery. Improving timely access for these patients requires balancing appropriate perioperative optimisation with avoidance of unnecessary delays.

Health insurance coverage was associated with earlier surgery, highlighting the continued influence of financial access on surgical timing. In settings where patients and families may be required to contribute directly to treatment costs, delays related to financial clearance, payment arrangements or access to necessary resources may postpone definitive care. Pre-payment through health insurance schemes may facilitate earlier access by reducing financial barriers and administrative delays. Although national insurance enrolment has expanded in Tanzania, effective access to insured emergency and trauma care remains uneven. Similar findings have been reported across SSA, where out-of-pocket payments, delayed insurance authorisation and informal costs continue to impede timely surgical care.³⁴ These findings suggest that progress towards universal health coverage should be assessed not only by insurance enrolment but also by whether patients can access timely and appropriate surgical care when needed.

Patient-level and clinical factors remain important contributors to surgical timing. Increasing ASA classification demonstrated an association with delayed surgery, emphasising the role of perioperative risk assessment and optimisation. In tertiary referral hospitals in Tanzania, limited access to laboratory diagnostics, blood products and specialist consultations may contribute to longer preparation periods for patients with greater physiological risk. Although anaemia, comorbidities and associated injuries were associated with delayed surgery in descriptive analyses, much of their effect may have been represented by ASA classification in multivariable models. ASA classification provides a broad assessment of overall physiological status and perioperative risk, which may influence readiness for surgery.³⁵

Fracture characteristics were also associated with surgical timing. Open fractures and more complex fracture patterns may require additional management considerations, including infection prevention, soft tissue assessment, staged treatment planning and availability of appropriate implants. These challenges may be amplified in resource-constrained environments where implant supply chains and operating theatre capacity remain limited.²⁹ Conversely, mid-femoral and distal femoral fractures and simpler fracture morphologies were associated with earlier surgery, potentially reflecting greater feasibility of standard fixation techniques and shorter operative preparation requirements within constrained surgical environments.

The association among implant type, implant application and surgical timing highlights the importance of health system and organisational factors in determining access to definitive fixation, which should be interpreted with caution. Implant availability, surgeon familiarity with specific techniques, theatre efficiency, operating room scheduling and referral processes may all influence the timing of surgery. Although these factors were not directly measured in this study, they represent important potential contributors to surgical delay. Evidence from LMICs suggests that delays in surgical care are often related not only to patient factors but also to health system capacity and organisation.²⁸ Addressing these challenges will likely require improvements in operating theatre availability, implant procurement systems, staffing capacity, referral coordination and perioperative pathways.

The findings also have implications for the broader development of trauma care in Tanzania and SSA. Orthopaedic trauma services in the region continue to evolve through the expansion of specialist training, improved surgical capacity and strengthening of referral systems. However, the delays observed in this study demonstrate that substantial challenges remain before timely fixation comparable with high-resource trauma systems can be consistently achieved. Reducing avoidable delays will likely require coordinated interventions addressing both patient access barriers and health-system limitations, including financial protection, surgical infrastructure, workforce capacity and availability of essential resources.

This study has several strengths. First, it included a large retrospective cohort of children, adolescents and adults with traumatic femoral fractures treated over a 7-year period at a major tertiary referral hospital in northern Tanzania, providing important evidence from a setting where data on timing of femoral fracture surgery remain limited. Second, it evaluated a broad range of sociodemographic, clinical and fracture-related characteristics associated with surgical timing. Finally, the use of a DAG-based approach for confounder selection strengthened the analytical framework by allowing adjustment based on predefined causal assumptions rather than solely statistical associations, reducing the risk of inappropriate adjustment, over-adjustment and collider bias.

Several limitations should be considered when interpreting these findings. Because this was a retrospective observational study, causal relationships cannot be established, and the observed associations should not be interpreted as evidence that reported factors directly cause surgical delay. Although the analysis accounted for important measured confounders, residual confounding from unmeasured factors remains possible. In particular, health-system factors such as operating theatre availability, implant stock-outs, surgeon and anaesthesia staffing, operating room scheduling and competing emergency surgical priorities were not collected. These factors may explain part of the observed variation in surgical timing and should be incorporated into future prospective studies. Patients with missing information on the time to surgery were excluded from the analysis. Comparison of baseline characteristics between excluded and included patients showed no clinically meaningful differences, suggesting that the risk of selection bias may be limited. However, some degree of bias due to missing data cannot be completely excluded. This study was conducted at a single tertiary referral hospital, which may limit generalisability to district hospitals or settings with different healthcare structures and resources. However, the findings may be relevant to other tertiary referral hospitals in SSA with similar referral pathways, patient populations and resource constraints. Differences in healthcare financing systems, surgical capacity and local organisation of trauma services should be considered when applying these findings to other settings. Finally, important clinical outcomes associated with delayed fixation, including mortality, fat embolism syndrome, deep vein thrombosis, pulmonary embolism, postoperative complications and functional outcomes, were not routinely collected in this study. Therefore, we were unable to evaluate whether delayed surgery was associated with these clinically important outcomes. Future prospective studies with standardised collection of perioperative complications, mortality outcomes and longer-term functional follow-up are needed to better understand the consequences of delayed femoral fracture fixation.

Conclusion

In this study, the median time to surgery was 14 days. The timing of surgery was associated with patient, clinical and health system factors. Health insurance coverage, higher haemoglobin levels, mid-femoral and distal femoral fracture location, simpler femoral fracture morphology and specific implant types were associated with earlier surgery. In contrast, associated injuries, open fractures and higher ASA classes (II–IV) were associated with a longer time to surgery. The observed associations suggest the potential roles of health system factors in surgical timing in this setting. Strategies to improve financial access and perioperative optimisation warrant further investigation to determine whether they may reduce delays to surgery. Future prospective studies should evaluate the impact of surgical delays on functional outcomes, complications and quality of life among patients with femoral fractures to better inform clinical practice and health policy.

Data availability statement

Data are available upon reasonable request.