Quality of care and 30 day mortality among patients with hip fractures: a nationwide cohort study

The Danish National Health Service provides tax-supported health care for all inhabitants of Denmark, including free access to hospital care. DNIP is a nationwide initiative to document, monitor and improve the quality of treatment and care provided by the Danish health care system. The project started in 2000 and focuses on developing and implementing evidence-based indicators related to health care structure, process, and outcomes [23, 25]. Hip fracture is one of eight specific diseases monitored in the DNIP by clinical indicators and quality standards. All patients ≥ 65 years admitted to Danish hospitals with a hip fracture, i.e., medial, pertrochanteric or subtrochanteric femur fracture are eligible for inclusion in the DNIP database. All Danish hospital departments (except from hospitals in the Copenhagen Hospital Corporation) caring for patients with hip fracture participated in the project during the study period. Thus a total of 28 hospitals (32 departments, more than 90% of all hospital departments caring for patients with hip fractures) reported data on hip fracture to DNIP in 2006. Each participating hospital department in the DNIP receives on a continuous basis their own results on the proportion of their patients meeting a number of quality of care criteria (see below). In the DNIP database, individual-level data for hip fracture patients from all hospitals are stored, including prospectively collected patient characteristics and the type and number of quality of care criteria each patient met during hospital stay.

We identified all patients with a hip fracture in the DNIP database who had a discharge date between August 16, 2005 and August 15, 2006 (N = 6,456). We excluded patients with multiple hip fractures during this study period to include only the first recorded hip fracture (n = 6,336). We excluded a few patients (n = 4) whose admission date was erroneously recorded to be later than the hip fracture operation or discharge date, and patients with more than one year time span between hospital admission and discharge (n = 2). Finally, we excluded patients with missing data for all five quality of care criteria (n = 64). Our study cohort thus comprised 6,266 patients with incident hip fracture.

A Danish national expert panel including physicians, nurses, physiotherapists, and occupational therapists identified five quality of care criteria to be measured during hospitalization with a hip fracture. The choice of criteria was based on a systematic search of the scientific literature and considerations regarding the feasibility of collecting the required data in routine clinical settings [25]. The five specific criteria are: early assessment (within 2 days after admission) of the patient's nutritional risk, systematic pain assessment during mobilization of the patient, assessment of Activities of Daily Living (ADL) before the fracture, assessment of ADL before discharge, and initiation of treatment to prevent future osteoporotic fractures. Assessment of nutritional risk was done according to updated nationwide guidelines. Systematic pain assessment had to be done using a pain scale with documented validity. Assessment of ADL before the fracture and again before discharge had to be done using a specific test to assess functional disability. Initiation of treatment to prevent future osteoporotic fractures was defined as ordination of any anti-osteoporotic medications or hip protectors. For two criteria, systematic pain assessment and prevention of future osteoporotic fractures, it was possible to classify patients as non-eligible: for instance if the patient had dementia or hemiplegia and therefore could not state their level of pain during mobilization, or if the patient already received anti-osteoporotic medication or had contraindications for treatment.

During hospitalization, hospital staff collected data on if the five quality of care criteria were met, as well as demographic characteristics and a range of prognostic covariates for each patient using a standardized form. After hospital discharge the computerized patient data were entered into the nationwide central DNIP database by the hospital staff.

Information on mortality during the study's follow-up period was obtained through linkage with the Danish Civil Registration System, which since 1968 has maintained electronic records of changes in vital status and migration for the entire Danish population [26].

At the time of hospital admission, data was collected on the following patient characteristics: age, gender, living situation (living together with another adult, living alone in one's own home, or other including living in a nursing home or other institution), alcohol intake (≤ 14/21 or >14/21 drinks per week for women/men, respectively), and smoking habits (current, former, never), respectively.

During the hospitalization, data was collected on the following disease-related covariates: type of fracture (medial, pertrochanteric, subtrochanteric), fracture displacement (displaced, undisplaced), delay before surgery (< 23 hours, 24-47 hours, 48-71 hours, and more than 72 hours), the American Society of Anaesthesiologists' (ASA) classification score before surgery (healthy, mild systemic disease, severe but not incapacitating systemic disease, incapacitating and life-threatening systemic disease, or moribund patient not expected to survive for 24 h), and type of surgery (osteosynthesis, hemi alloplastic, total alloplastic, other operation), respectively.

To adjust for the individual burden of comorbidity in our mortality analyses, we computed Charlson comorbidity index scores for all patients with hip fracture using records of all hospital discharge diagnoses made prior to the admission date in the Danish hospital discharge registry. This registry was established in 1977 and collects data on all hospitalizations, including dates of admission and discharge, surgical procedure(s) performed, and up to 20 discharge diagnoses assigned by the treating physician and coded according to the International Classification of Diseases (8^th revision (ICD-8) until the end of 1993, and 10^th revision (ICD-10) thereafter).

The Charlson comorbidity [27] index covers 19 major disease categories, weighted according to their prognostic impact on patient survival, and has been validated for use with discharge registry data in the ICD databases for prediction of mortality [28]. We defined three levels of comorbidity for each patient, based on the complete hospital discharge history since 1994, as follows: 'low' for patients with no recorded underlying diseases included in the Charlson index; 'medium' (score 1-2); and 'high' (score >2).

The study was approved by The Danish Data Protection Agency (record no. 2007-41-0073).

We used logistic regression to compute odds ratios (ORs) with 95% confidence intervals (CIs) for mortality at day 30 after hip fracture according to type and number of quality of care criteria met, adjusting for age, sex, the patient's living situation, and the Charlson comorbidity index score. In the second model we additionally adjusted for acute disease-related variables, i.e., type of fracture, fracture displacement status, ASA classification score, delay before surgery, and type of surgery, respectively. Because alcohol and smoking data was missing for 28.3% and 29.1% of the patients, respectively, the full model was run both with and without inclusion of alcohol and smoking data.

In all adjusted analysis, we used a random effect model to correct for possible clustering by hospital, because other and unmeasured quality of care characteristics at the health care provider level might be associated with patient mortality [29].

We first examined the association between each individual quality of care criterion met and 30 day mortality. Patients were excluded from the analysis if there was missing data if the specific quality of care criterion was met (ranging from 0.4% of patients for assessment of nutritional risk or ADL before the fracture to 15.4% for systematic pain assessment) or if the criterion was found non-eligible to the patient (i.e. if the patient had contraindications). Thus, the number of patients included in the analyses of the specific quality of care criteria differed. We ran analyses both for the entire patient cohort with use of separate categories for missing data on prognostic characteristics, and for a cohort of patients who had complete data on all prognostic characteristics. We also examined whether the quality of care criteria were independently associated with 30 day mortality by including all of the criteria into one model and thereby mutually adjusting them. This analysis was only possible to do among patients who were eligible for all quality of care criteria.

Thirdly, we did a number of subanalyses to test the robustness of our findings. To reduce confounding by contraindication in patients near end-of-life we reran analyses with exclusion of patients who died during hospitalization. We also ran analyses for 30 day post-discharge mortality among all patients discharged alive, and for 180 days of follow-up both post-admission and post-discharge. All statistical analyses were performed with STATA™ software (version 8.2).