Clinical management of hip fractures in elderly patients with dementia and postoperative 30‐day mortality: A population‐based cohort study

Abstract Objectives Patients with dementia have an increased 30‐day mortality after hip fracture. We investigated clinical management including time to surgery, out‐of‐hours admission and surgery, surgery on weekends, surgery volume per ward, and anesthesia technique for this excess mortality risk. Method This register‐ and population‐based study comprised 12,309 older adults (age 70+) admitted to hospital for a first‐time hip fracture in 2013–2014, of whom 11,318 underwent hip fracture surgery. Cox proportional hazards regression models were applied for the analysis. Results The overall postoperative 30‐day mortality was 11.4%. Patients with dementia had a 1.5 times increased mortality risk than those without (HR = 1.50 [95% CI 1.31–1.72]). We observed no time‐to‐surgery difference by patient dementia status; additionally, the excess mortality risk in patients with dementia could not be explained by the clinical management factors we examined. Conclusions Increased mortality in patients with dementia could not be explained by the measured preoperative clinical management. Suboptimal handling of postoperative complication and rehabilitation are to be investigated for their role in the witnessed increased mortality for patients with dementia.

People with dementia are up to three times more likely than their cognitively intact counterparts to experience a hip fracture, often via accidents including falls (Friedman, Menzies, Bukata, Mendelson, & Kates, 2010;Petersen et al., 2018). After hip fracture, patients with dementia have a higher risk of dying within 30 days after surgical treatment compared with other hip fracture patients (Bai et al., 2018;Liu, Wang, & Xiao, 2018), and the reasons for this are not yet fully understood. Although this higher mortality may be linked to the combined effects of dementia, comorbidities, and intake of multiple medications Smith, Pelpola, Ball, Ong, and Myint (2014), clinical management of hip fracture surgery in hospital settings may play a role.
Studies of these clinical management factors in hip fracture patients with dementia for mortality risk are limited and have been conducted in different healthcare systems and are heterogenic in study methodology Yonezawa, Yamazaki, Atsumi, and Obara (2009) ;Petersen, Jorgensen, Hansen, & Duus, 2006;Johansen et al., 2017).
In Denmark, national clinical guidelines for hip fracture patient care are implemented in routine daily clinical practice, presumably regardless of patients' dementia status, but in real life, discrepancies may exist. Hence, this setting was apt for our study to investigate whether differences in the abovementioned clinical management factors may explain the excess postoperative 30-day mortality risk for hip fracture patients with dementia.

| Study design and population
This was a population-based study using data from Danish national registers linked using the unique civil registration (CPR) number assigned to each individual at birth and to persons who hold a Danish residence permit upon immigration (Pedersen, 2011).
The study population was all residents in Denmark aged 70 or older (70+) who had a first-time hip fracture admission to a hospital in Denmark between January 1, 2013, and December 31, 2014(2013. We excluded individuals from the study if they met any of the following criteria: (a) history of hip fracture in 2008-2012, (b) died before the index date of hip fracture admission, and (c) emigration on the index date. In addition, we also excluded individuals who immigrated to Denmark later than December 31, 2007, due to potentially incomplete information on hip fracture, chronic diseases, and medication in the registers for the period of 2008 to 2012 ( Figure 1).
Records of hip fracture and hip fracture surgery were extracted from the Danish National Patient Register (DNPR) (Lynge, Sandegaard, & Rebolj, 2011). The date of hospital admission for a first-time hip fracture in 2013-2014 was defined as the index date.
The eligible participants were followed for 30 days from both the

Key points
• Patients with dementia were at a higher risk of dying within 30 days after hip fracture surgery.
• This excess postoperative 30-day mortality in patients with dementia could not be explained by the measured preoperative clinical management.
F I G U R E 1 Flowchart of the study participant selection date of the hip fracture admission (n = 12,309) and the date of the hip fracture surgery (n = 11,318) for mortality assessment.
International Classification of Diseases, 10th Revision (ICD-10) codes were used to define hip fractures as fracture of femoral head or neck (S720), pertrochanteric fracture (S721), or subtrochanteric fracture (S722) in 2013-2014 from the DNPR. Additionally, ICD-10 code M809B for osteoporotic hip fracture was used together with hip fracture codes (S72*) for inclusion. Using the same hip fracture diagnosis codes, an individual's history of hip fracture records in [2008][2009][2010][2011][2012] was also traced in the register for exclusion purpose. Both primary and secondary diagnoses were included with any type of inpatient, outpatient, or emergency department registration from Danish hospitals. The Nordic Medico-Statistical Committee codes for arthroplasty (KNFB*) and osteosynthesis (KNFJ*) were used to extract potential hip fracture surgery performed on eligible study participants.

| 30-day mortality
We defined 30-day postoperative mortality as death within the first 30 days after hip fracture surgery (in or out of hospital, all-cause).
We obtained information on death (all-cause) including the date of death from the CPR Pedersen, (2011), which records an individual's vital status and date of death.
Anatomical Therapeutic Chemical (ATC) codes were used to identify dementia-related medications prescribed in Denmark (ATC codes: N06DA01, N06DA02, N06DA03, N06DA04, N06DX01) (Pottegard et al., 2017;Schmidt et al., 2015). The prescription registrations were used in this study in order to capture patients with dementia but missing a diagnosis in the DNPR or PCRR.
We traced the history of an individual's dementia diagnosis and dementia-related medications for the period of 2008-2014. For this study, the identified date of dementia had to be at least one day before the index date of the hip fracture hospital admission registered in 2013-2014.
We grouped these diseases into four categories: 0, 1, 2, or 3 or more (3+) comorbid diseases, in any combination, based on the number of diseases present on the index date of hospital admission for hip fracture, as well as on the date of the hip fracture surgery. Denmark contains 100 pills, and the usual dose is 1 pill a day). We grouped usage of these medications into three categories: 0, 1, 2, or 3 and more (3+), in any combination.

Time to surgery
We defined TTS as the time from the hospital admission for hip fracture to the time hip fracture surgery was performed and measured in hours. There were a few cases with a zero-time registration between admission and surgery. If the two events occurred on the same day, then we set the TTS equal to 0.5 days, that is, 12 hrs.

Out-of-hours admission and surgery
We defined out of hours as any admission and surgery in the time period of 16:01-07:59.

Surgery on weekend
Any surgery performed on Saturdays, Sundays, and on Danish public holidays in 2013-2014 was defined as surgery on weekends.

Anesthesia technique
Anesthesia technique used for hip fracture surgery was divided into two groups: general anesthesia (NAAC*), and regional anesthesia (NAAD*) includes both spinal (NAAD1*) and epidural anesthesia (NAAD0*). Since very few patients were operated on with local anesthetic and/or conscious sedation (NAAB*), we categorized them with the regional anesthesia group.

Surgery volume per ward
Surgery volume per ward (identified using the six-digit code for each hospital) was measured by the total number of registered hip fracture surgeries performed (only arthroplasty (KNFB*) and osteosynthesis (KNFJ*) surgeries were included) in 2013-2014. We only counted one procedure in cases where several procedures were recorded for the same patient in the study period.

| Other covariates
The study participants' age, sex, marital status, immigration, and emigration status were all extracted from the CPR Pedersen, (2011).
The participants' education information was extracted from the Population Education Register (Jensen & Rasmussen, 2011).
The patients were divided into three age groups: 70-79, 80-90, or 90 years and older (90+). Marital status was grouped into four categories: married (including partnership or cohabitation), divorced (including dissolved partnership), widowed (including partner died), or single (no cohabitation). Education level was grouped according to the individual's highest education level in years: low education (<10 years), medium education (10-12 years), or higher education (>12 years).

We also included American Society of Anesthesiologists Physical
Status Classification System codes (ASA class) scale as a covariate (Daabiss, 2011). Given that ASA class scale of 5 indicates a patient who is not expected to survive 24 hr with or without surgery for the present study, we included a scale of 1-4 to group patients: A lower number indicates patients who are more fit for surgery, while a higher number indicates patients less fit for surgery. In the DNPR, ASA class is recorded using codes EZA*.

| Statistical analysis
We described the age group, sex, education, marital status, presence of comorbid diseases, and use of sedative medication by the index date of hospital admission for hip fracture in patients with and without dementia. We also described the characteristics of hip fracture surgeries in 2013-2014, including clinical management for hip fracture surgery by dementia status.
The 30-day postsurgery mortality was time to event. Patients were censored at the end of the study or at their date of emigration.
However, for the TTS, patients were also censored if they died before the hip fracture surgery. We applied Cox proportional hazards regression (Cox regression) to estimate association between dementia and the study outcomes.
For the association between dementia and TTS, the multivariable Cox regression was adjusted for age group, sex, education, marital status, comorbid diseases, sedative medications, and out-of-hours hospital admissions for hip fracture.
For the association between dementia and 30-day postsurgery mortality, we applied three Cox regression models. Model A was univariable analysis of dementia for 30-day mortality after hip fracture surgery. Model B was Model A adjusted for confounders (age group, sex, education, marital status, comorbidity, and intake of sedative medications). Model C was Model B adjusted for clinical management factors including out-of-hours admissions, out-of-hours surgery, surgery on weekends, anesthesia technique, and surgery volume per ward. Additionally, ASA class scores, type of surgery, and type of hip fracture were all adjusted for in Model C.
The risk estimations are presented as hazard ratios (HRs) with corresponding 95% confidence intervals (95% CIs). All statistical analysis was performed using STATA version 14.0 (StataCorp LP).

| Ethical consideration
Complying with European data protection rules, the Danish Data Protection Agency registered the project (Journal no. 2016-41-4674). According to Danish law, review by an ethics board or patient consent is not required for purely register-based studies.

| Population characteristics at hospital admission for hip fracture
A total of 12,309 first-time hip fracture hospital admissions were identified in the period 2013-2014 among older people (age 70+) in Denmark, and among them, 1,853 (15.1%) patients had a diagnosis of dementia. There were a higher proportion of older, female, and divorced or widowed patients, and more patients with comorbid chronic diseases and taking sedative medications in the dementia group than in the nondementia group (Table 1). More than half (53.6%) of hip fracture patients were admitted out-of-hours, but no significant difference by dementia status was found.

| Characteristics of hip fracture surgeries
Among the 11,318 patients who underwent a hip fracture surgery, half (51.7%) had a fracture of the femoral head or neck; nearly twothirds of the total study population received osteosynthesis surgery; and more than one-thirds (37.8%) of the total surgeries were performed out of normal business hours, and 29.4% on weekends (Table 3).
There was no significant difference in clinical management of hip fracture surgery, including out-of-hours admission, surgery on weekends, and surgery volume per ward, between patients with and without dementia. However, patients with dementia were more likely to undergo an operation out of hours and receive general anesthesia than those without dementia (p < .05). In multivariable analysis adjusted for confounders including age group, sex, education, marital status, comorbidities, and sedative medications, patients with dementia had a risk of dying within 30 days after hip fracture surgery that was 1.5 times higher than the A sensitivity analysis the using Charlson Comorbidity Index (Charlson, Pompei, Ales, & MacKenzie, 1987) instead of the comorbidity we defined did not change the estimates for mortality risk for the other factors.

| Comparison with existing studies
To the best of our knowledge, our study is the first population-based study to investigate dementia for postoperative 30-day mortality in Denmark by taking into account the clinical management factors investigated in this study. Other studies, although in line with our postoperative mortality risk estimations in patients with dementia, lack clinical management factors for comparison (Bai et al., 2018;Pedersen, Ehrenstein, & Szepligeti, 2017).
Managing patients with dementia has great challenges in clinical practice Krupic, Eisler, Skoldenberg, and Fatahi (2016). Due to the nature of the disease, communication with dementia patients and recognizing their wants and needs can be difficult (Judd, 2017;Tarazona-Santabalbina et al., 2015). It often requires of staff extra time and specialized knowledge of dementia Klestil et al., (2018).
These requirements may be difficult to meet in surgery departments, for example, due to heavy workloads and an inadequate supply of personnel with expertise in dementia care, and this situation may worsen as the population ages (Simunovic et al., 2010). Furthermore, patients undergoing hip fracture surgery include individuals with dementia at all stages of severity. Among those with moderate-to-severe dementia, nearly 30% to 90% of them manifest behavioral and psychological symptoms, and such symptoms may play a modifying role with hip fracture for clinical management (Muller-Spahn, 2003).
All these factors mentioned previously may lead to low priority in clinical practice. However, we found no significant difference in these Furthermore, we found that among all patients whose anesthesia technique was recorded in the DNPR, those who received regional anesthesia had an 18% reduction in 30-day postsurgery mortality compared with general anesthesia. A systematic review and me-

ta-analysis by Guay et al published by the Cochrane Library in 2016
examined 11 studies and concluded that there was no difference in mortality at 30 days between these two techniques, but they also concluded that the low quality of the evidence calls for more randomized clinical trial contributions (Guay et al., 2016). It should be noted that in our study population, among patients who underwent mation; however, this missing information showed no effect on postoperative 30-day mortality.
Our results revealed that being older, male, comorbid with more chronic diseases, taking more sedative medications, and having a higher ASA class score were more important prognostic factors for 30-day mortality than the clinical management aspects we assessed. These factors are often present in patients with dementia.
Therefore, when admitting a hip fracture patient, proper screening should take place to identify these prognostic factors and guide extra attention to mitigation of same: Cautious care may improve survival rate in this population.

| Study strengths and limitations
Our study has a number of important strengths. In this populationbased cohort study, all variable data were extracted from Danish national registers. The CPR number-linked information tracked from the registers resulted in no loss to follow-up within our cohorts and minimized selection bias.

TA B L E 4 (Continued)
be improved upon (Seitz, Gill, & Austin, 2016). Also, delirium, which often accompanies dementia, may explain part of the excess mortality (Mosk et al., 2017). Further studies should focus on the role of postoperative clinical management in the excess mortality after surgery for dementia patients, and these studies should not be purely register-based.
Severity of dementia is linked to functional recovery and mortality after hip fracture (Tarazona-Santabalbina et al., 2015). Given that the DNPR lacks validation of dementia severity registration (Phung et al., 2007), we adjusted using the number of days since identification of dementia as a proxy, and found it had no influ-

| CON CLUS ION
Among elderly patients with a first-time hip fracture in 2013-2014, we found an increased 30-day mortality in patients with dementia.
However, the increased postoperative 30-day mortality was not explained by the surveyed clinical factors. Future research should focus on delirium and postoperative complications and clinical care in order to further examine the increased mortality.

CO N FLI C T O F I NTE R E S T
None declared.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.1823.

DATA AVA I L A B I L I T Y S TAT E M E N T
For legal reasons, the data that support the findings of this study cannot be made publicly available. However, with permission, researchers can obtain access to the data from Statistics Denmark.