The database was declared to the French National Commission on Computing and Liberty (CNIL) of the Assistance Publique-Hôpitaux de Paris (APHP) for this study (no. 20190822165316). This report follows the STROBE recommendations (Additional file 1) [15].
Study design, study setting and eligibility criteria
A nested case–control study was conducted in the unit for perioperative geriatric care of the Pitié Salpêtriere hospital, a university hospital of the greater Paris University Hospitals Group (Assistance Publique Hôpitaux de Paris, AP-HP). The unit for perioperative geriatric care is part of a dedicated orthogeriatric care pathway including coordination between the department of emergency medicine and surgery, department of anaesthesiology and critical care, department of orthopaedic surgery and department of rehabilitation. This dedicated orthogeriatric care pathway is defined as 1) an early alert from the emergency department (ED), 2) considering hip fracture as requiring surgery as soon as feasible (i.e., 24 h/day), 3) rapid transfer to the unit for perioperative geriatric care after surgery, and 4) rapid transfer of stable patients to a dedicated rehabilitation unit [16, 17].
The management strategy in this unit for perioperative geriatric care, previously described [16], focused on early mobilization with the aim of chair-sitting and walking within 24 and 48 h after arrival, respectively; pain management; the provision of air-filled mattresses for patients with pressure sores or at high risk of pressure sores; swallowing disorders detected using a systematic medical survey; detection of stool impaction and urinary retention using bedside ultrasonography; correction of anemia with transfusion of packed red blood cells (haemoglobin level threshold ≤ 10 g/dl from 2009 to 2011 and haemoglobin level threshold ≤ 8 g /dl or symptoms since January 2012) [18]; and detection of delirium, AF and malnutrition. To screen for POAF, electrocardiography was systematically performed on admission to the unit for perioperative geriatric care. Then, electrocardiography was performed in case of tachycardia or irregular rhythm on clinical examination performed at least twice a day or hypotension, dyspnea, malaise, syncope, chest pain, delirium, anemia (haemoglobin level ≤ 8 g/dl) or electrolyte imbalance.
From July 1, 2009 to December 31, 2019, all consecutive patients with hip fracture admitted to the unit for perioperative geriatric care were evaluated for eligibility. Patients were included if they were ≥ 70 years old and their primary presentation was hip fracture (first hospitalization after surgery in the unit for perioperative geriatric care). On average, 150 to 200 patients aged ≥ 70 years old with hip fracture undergo hip surgery annually at Pitié Salpêtriere hospital, 75% of whom will be hospitalized in our unit for perioperative geriatric care. We excluded patients with multiple or metastatic or periprosthetic fractures, a history of permanent AF before admission in the unit for perioperative geriatric care, post-operative supraventricular arrhythmia other than AF; and missing data (missing anesthesia records, missing data from the ED, no electrocardiogram before admission in the unit for perioperative geriatric care). Patients were followed until death or the end of hospitalization in the unit for perioperative geriatric care. Some patients had been included in previous studies [14, 16,17,18,19,20,21,22,23,24,25].
Outcomes
Our main outcome measure was the occurrence of POAF defined as new-onset AF in the immediate period after surgery (until the end of hospitalization in the unit for perioperative geriatric care). The occurrence of POAF was retrospectively adjudicated by 2 geriatricians (AR, PG) who independently reviewed medical charts (kappa = 0.948, 95% confidence interval [CI] 0.87–0.99). In case of disagreement, consensus was reached with a third independent senior expert (JB).
Data collection methods and variables
Since the opening of the unit for perioperative geriatric care in 2009, we have created a dedicated research database that is prospectively implemented by 3 senior geriatricians (JB, JCB, LZ), experts in orthogeriatrics, and that integrates all the data from the orthogeriatric care pathway for each patient.
The following variables were collected prospectively by interviewing patients, their family members or their physicians and pharmacists during the hospital stay and were defined as baseline characteristics before hip fracture: age, sex, home or nursing home living conditions, walking ability, previous medical history including cardiovascular and neurologic diseases, chronic medications, and type of fracture (radiological definition by an orthopedic surgeon).
Co-morbidity severity was assessed with the Charlson Comorbidity Index [26] because the index, among a number of other comorbidity indexes, has been found to predict mortality in this population [20]. Functional status was evaluated with an activities of daily living scale (the Katz ADL index) [27]. Repeated falls was defined as 2 or more falls in the previous year, chronic renal failure as Cockcroft creatinine clearance < 30 ml.min−1, hypokalemia as potassium level < 3.5 mmol/L and anemia as haemoglobin level < 12 g.dL−1 for women and 13 g.dL−1 for men.
During the perioperative period, we prospectively recorded the surgical treatment, the delay and duration of surgery, the anesthetic drugs used and all drugs and transfusions administered from the ED to the unit for perioperative geriatric care. Cardiovascular drugs taken at baseline and their suspension were recorded retrospectively.
After surgery, delays to first sitting and first walking, destination (home or rehabilitation) at discharge from the unit for perioperative geriatric care and length of stay in acute care were recorded. All postoperative complications during the acute care period were prospectively recorded.
Statistical analysis
The study is based on data for all available patients during the study period, and thus no a priori power calculation was conducted. Data are presented as mean ± SD or median (interquartile range) for continuous variables and number (percentage) for categorical variables. Comparison of quantitative variables involved unpaired Student t test or Mann–Whitney test depending on the normal distribution of data. Normality was assessed by graphical representation of the data distribution. Comparison of categorical variables involved chi-squared or Fisher’s exact test, as appropriate.
Patients with (cases) and without POAF (controls) were matched 1:5 on 5 baseline characteristics (age, hypertension, diabetes, coronary artery disease, cardiac failure). We selected those 5 factors from a literature review [4] and deliberations of a panel of 9 independent experts. This panel included 3 geriatricians, 4 cardiologists and 2 anesthesiologists who were all blinded to the research question at this time and to the other experts’ answers. We asked each expert the following standardized question: “According to you, what are the 9–10 main baseline predisposing factors of POAF among patients of 70 and more undergoing hip fracture surgery?” The 4 most frequently given answers were age (89%), diabetes (89%), hypertension (89%) and coronary artery disease (78%) (Additional file 2). Then, 4 other factors were mentioned with equal frequency: history of paroxysmal AF, chronic heart failure, chronic kidney disease, and valvular disease. We chose to keep chronic heart failure because it was most homogenously proposed by the 3 represented medical specialties and was previously described as a strong predisposing factor in this specific context [11, 28, 29]. This a priori and pragmatic method allows for improving the external validity and efficiency of the results and selecting a restricted number of baseline factors associated with POAF, in order to limit the number of patients excluded from the matching [30].
Then, using the matched dataset, we performed a conditional logistic regression analysis to assess independent modifiable variables present at admission in the unit for perioperative geriatric care that were associated with POAF; adjusted odds ratios (ORs) and their 95% CIs were calculated [30]. To avoid overestimation, a conservative approach was used [31, 32]: all variables with P < 0.20 on univariate analyses and all clinically relevant variables were included.
Finally, we performed a sensitivity analysis excluding patients with pre-existing AF before surgery (another definition of POAF).
Statistical analyses were performed with R v 1.4.1717 (package Matchlt). All p-values were two-tailed and p < 0.05 was considered statistically significant.