Patient information was obtained for 100% Medicare Provider Analysis and Review (MEDPAR) inpatient files, hospice files, the nursing home residential assessment minimum data set (MDS), and denominator files from January 1, 1999 to December 30, 2007. Information on site of death and the type of services used was based on the Residential History File (RHF) methodology.10
The RHF integrates Medicare administrative files including the denominator file and Medicare claims with the MDS to create an individualized historical record of service utilization and site of care including sequence of care sites, length of stay at each service location, and date and location at time of death.10
Beneficiaries were eligible if they were 75 years of age or older at the time of the admission for their first hip fracture, were enrolled in Medicare part A and part B, and were not enrolled in a managed care health plan (n=1,412,119). We excluded individuals admitted for their first hip fracture after June 30, 2007 to allow for a six-month observation period (n=68,580);, refer to for cohort inclusion criteria.
Hospice Cohort Inclusion and Exclusion Criteria
Incident hip fractures were identified using the primary International Classification of Diseases, 9th revision (ICD-9) diagnosis codes on the hospital admission (see Appendix). A five-year look-back period was used to exclude any beneficiaries with prior hip fracture. Suspect claims occurring after the date of death (n=5,798), hospitalization outside the U.S. (n= 299), and those with surgery codes indicating a revision of a previous hip fracture, bilateral hip surgery, or multiple trauma (n= 20,011) also were excluded.
Among those identified as having their first hip fracture, enrollment in hospice at the time of the fracture was identified. Hospice use was defined as any submitted claim for hospice in the 30 days prior to the hip fracture hospitalization. Individuals not receiving hospice in the 30 days prior to the hip fracture hospitalization were excluded (n=1,303,031). The final cohort consisted of 14,400 individuals receiving hospice who were hospitalized for their first hip fracture between January 1, 1999–June 30, 2007. If a second fracture occurred during the observation period, it was not included in the analysis.
Main Explanatory Variable
Surgical or non-operative treatment was determined based on ICD-9 procedure codes and/or a diagnosis related group (DRG) that indicated surgery (see Appendix). To ensure that the cohort was not biased by excluding those individuals who were not medically stable to have surgery during the initial hospital stay or those individuals who were transferred to another hospital for surgery, individuals having a hip fracture procedure within thirty days also were included in the surgical cohort.7
Medicare beneficiaries hospitalized with an ICD-9 diagnosis code for hip fracture (820.XX) and having either an ICD-9 procedure code or DRG code consistent with a hip fracture repair within 30 days of index hip fracture hospitalization were defined as the hip fracture surgical cohort.
Although pathologic fracture should not be coded with the 820.XX ICD-9 diagnosis code, the final cohort was reviewed to determine if any of the index event hip fractures were diagnosed as pathological fractures (ICD-9 diagnosis code 733.10); none were found.
Survival from hip fracture hospital admission (in days) to death was the outcome of interest. For those individuals alive at six months and those dying within the first six months, the type of service being used at six months or the time of death was identified: nursing home, skilled nursing facility, inpatient rehabilitation facility, home with home care, home with self-care, home on hospice, nursing home with hospice, death during initial hip fracture hospitalization, hospitalization (i.e., acute care hospital stay, emergency room visit, or observation), or other site of care (psychiatric hospital, long-term care hospital, outpatient services).
Sociodemographic characteristics used to describe the hospice cohort included: age, race (White, Black, and Hispanic), and gender. In order to describe what type of hospice services were being used at the time of the fracture, indicators of the location where hospice services were being received was included (nursing home or community). Hospital admission records were used to identify the type of hip fracture (femoral neck fracture, petrochanteric fracture), the occurrence of an intensive care unit (ICU) stay, length of hospitalization (days), comorbid diagnoses (cancer, cardiovascular disease, dementia, degenerative disease of the central nervous system, chronic obstructive pulmonary disease (COPD), acute myocardial infarction, and renal disease), year of the hip fracture, and the Elixhauser comorbidity measure.8
The means and distributions of patient characteristics and time trends are presented in . Using STATA 10.0 (StataCorp, LP, College Station, TX), six-month survival was calculated using Kaplan-Meier survival curves. A Cox proportional hazard model was estimated to test the effect of surgery on survival, adjusting for covariates and clustering within hospitals using the Huber-White correction.
Characteristics of Hospice Patients Experiencing a Hip Fracture