Recruitment and randomization
This study included patients with femoral neck fracture aged ≥70 years, consecutively admitted to the orthopedic department at the Umeå University Hospital, Sweden, between May 2000 and December 2002, and the study was designed according to the CONSORT guidelines .
In Sweden different surgery methods are used depending on the displacement of the femoral neck fracture. In the present study patients with undisplaced fracture were operated on using internal fixation (IF) and patients with displaced fracture were operated on using hemiarthroplasty (HAP). If patients had severe rheumatoid arthritis, severe hip osteoarthritis, or pathological fracture they were excluded, by the surgeon on duty, because of the need for a different surgery method, such as total hip arthroplasty (THA). Patients with severe renal failure were excluded, by the anesthesiologist, because of their morbidity. Patients being bedridden before the fracture occurred were also excluded.
In the emergency room the patients were asked both in writing and orally if they were willing to participate in the study. The next of kin was always asked prior to the inclusion in patients with cognitive impairment. The patients or their next of kin could at any time decline participation. A total of 258 patients met the inclusion criteria; 11 patients declined to participate and 48 patients were not invited to participate because they had sustained the fracture in the hospital or the inclusion routines failed (Fig. 1). These 59 patients were more likely to be men (p = 0.033) and living in their own house/apartment (p = 0.009), but there was no difference in age (p=0.354) compared to the participating patients. The remaining 199 patients (Table 1) consented to participate. All patients received the same preoperative treatment.
Patients were randomized, to postoperative care in a geriatric ward with a special intervention program or to conventional care in an orthopedic ward, in opaque sealed envelopes. The lots in the envelopes were sequentially numbered. All participants received this envelope while in the emergency room but the envelope was not opened until immediately before surgery to ensure that all patients received similar preoperative treatment. Persons not involved in the study performed these procedures.
The randomization was stratified according to the operation methods used in the study. Depending on the degree of dislocation, the patients were treated with IF using two hook-pins (Swemac Ortopedica, Linköping, Sweden) (n=38 intervention vs n = 31 control) or with bipolar hemiarthroplasty (Link, Hamburg, Germany) (n = 57 vs 54). Basocervical fractures (n = 7 vs 10) were operated on using a dynamic hip screw (DHS, Stratec Medical, Oberdorf, Switzerland) and one had a resection of the femoral head due to a deterioration in medical status and one died before surgery (both were in the control group).
The intervention ward was a geriatric unit specializing in geriatric orthopedic patients. The staff worked in teams to apply comprehensive geriatric assessments, management, and rehabilitation [29, 30]. Active prevention, detection, and treatment of postoperative complications such as falls, delirium, pain, and decubitus ulcers was systematically implemented daily during the hospitalization (Table 2). The staffing at the intervention ward were 1.07 nurses/aides per bed.
The control ward was a specialist orthopedic unit following the conventional postoperative routines. A geriatric unit, specializing in general geriatric patients, was used for those who needed longer rehabilitation (n = 40). The staffing at the orthopedic unit was 1.01 nurses/aides per bed and 1.07 for the geriatric control ward. The main content of both the intervention program and the conventional care is described in Table 2.
The staffs on the intervention and control wards were not aware of the nature of the present study.
Two registered nurses were employed and performed the assessments during hospitalization.
Medical, social, and functional data were collected from the patients, relatives, staff, and medical records on admission. Complications during hospitalization, including falls, length of stay, morbidity, and mortality, were systematically registered in the medical and nursing records. Nurses are obliged by law to document any falls in the records . A fall was defined as an incident when the patient unintentionally came to rest on the floor and included syncopal falls. Numbers of falls and time lapse to first fall after admission were calculated. The Abbreviated Injury Scale (AIS)  was used to classify the injuries resulting from a fall. The maximum injury (MAIS) connected with each incident was recorded.
A few days after surgery, patients were assessed and interviewed regarding their cognitive status using the Mini Mental State Examination (MMSE) . The modified Organic Brain Syndrome Scale (OBS Scale)  was used to assess cognitive, perceptual, emotional, and personality characteristics as well as fluctuations in clinical states. Mental state changes were also documented from medical records. Depression during hospitalization was diagnosed due to current treatment with antidepressants and depression screened using the Geriatric Depression Scale (GDS-15)  in combination with depressive symptoms observed and registered by the OBS Scale. The patients’ vision and hearing were assessed by their ability to read 3-mm block letters with or without glasses, and their ability to hear a normal speaking voice from a distance of 1 m. Activities of daily living (ADL) prior to the fracture were measured retrospectively using the Staircase of ADL .
A geriatrician, unaware of study group allocation, analyzed all assessments and documentation, after the study was finished, for completion of the final diagnoses according to the same criteria for all patients.
The Ethics Committee of the Faculty of Medicine at Umeå University approved the study (§ 00-137).
The sample size was calculated to detect a 50% reduction of number of fallers between the intervention and control groups at a significance level of 0.050, based on our previous multifactorial fall intervention study in institutional care . Student’s t-test, Pearson’s χ2 test, and the Mann-Whitney U test were performed to analyze group differences regarding basic characteristics and postoperative complications.
We analyzed outcomes on an intention to treat basis. The incidence of falls between intervention and control groups was compared in three ways. First, an unadjusted comparison using Pearson’s χ2 and Fisher’s exact test regarding number of patients who fell and injuries. Second, the fall incidence rate was compared between intervention and control groups by calculating the fall incidence rate ratio (IRR) using a negative binomial regression, with adjustment for observation time and for overdispersion. Negative binomial regression (Nbreg) is a generalization of the Poisson regression model and is recommended for evaluating the efficacy of fall prevention programs . Third, a Cox regression was used to compare the time lapse to first fall between groups (hazard rate ratio, HRR). The difference in fall risk between groups was further illustrated by a Kaplan-Meier graph.
Basic characteristics that differed between the intervention and the control groups, corresponding to a p value <0.150 (depression, antidepressants, and dementia, Table 1), were considered as covariates in the Poisson (Nbreg) and the Cox regression models. However, the inclusion of these variables had only marginal effects on the log-likelihood values of the models as well as on the IRR and HRR values and standard errors for the group allocation variable (intervention or control). In addition, none of the variables showed significant effects on the dependent variable and are therefore not included in the Poisson (Nbreg) and Cox regression analyses.
Pearson’s χ2 test and Fisher’s exact test were also used to analyze the associations between falls and days with delirium between the groups.
All calculations were carried out using SPSS v 11.0 and STATA 9 statistical software for Macintosh. A p value <0.050 was considered statistically significant.