We used the Nationwide Inpatient Sample databases for 1990 through 2001. The Nationwide Inpatient Sample is a part of the Healthcare Cost and Utilization Project (HCUP), sponsored by the Agency for Healthcare Research and Quality [1
]. The Nationwide Inpatient Sample is the largest all-payer inpatient care database publicly available in the United States and contains between five and eight million records of inpatient stays per year from approximately 1000 hospitals. This represents a 20% stratified sample of community hospitals in the United States [1
]. To ensure maximal representation of US hospitals, the Nationwide Inpatient Sample included hospitals according to five important hospital characteristics: geographic region (Northeast, North Central, West, South); ownership (public, private not-for-profit, private investor-owned); location (urban, rural); teaching status (teaching hospital, nonteaching hospital); and bed size (small, medium, large). Additional details on these variables can be found at the HCUP web site [1
The HCUP assigned validation and quality assessment of these data sets to an independent contractor [2
]. The Nationwide Inpatient Sample also was validated extensively against the National Hospital Discharge Survey and confirmed to perform well for many estimates [23
The Nationwide Inpatient Sample database contains information on primary International Classification of Diseases, Ninth Edition, Clinical Modification (ICD-9-CM) diagnosis and procedure codes and 14 secondary diagnoses and procedure codes for each patient record. Each record in the data set represents one patient admission and has a unique identification number. We selected admissions with an ICD-9-CM primary diagnosis code for closed transcervical fracture of femoral neck (820.00, 820.01, 820.02, 820.09). We included patients in whom ORIF (79.35), HA (81.52), or THA (81.51) was designated as the primary procedure code, which represented a majority of patients (n = 163,209 of 164,093 patients with femoral neck fractures; 99.5% of patients). Other procedures such as internal fixation without reduction (including prophylactic internal fixation of bone, reinsertion of fixation device, revision of broken or displaced fixation device; n = 13,307) and closed reduction with internal fixation (n = 12,654) were used in fewer patients. We also did not include patients with closed base of femoral neck fractures because their management is different from that of patients with other femoral neck fractures (n = 13,058). Records in which the same patient had a primary procedure code for one of the three procedures (ORIF, HA, or THA) and also a secondary procedure code for one of the three procedures also were not included (n = 952). Our final analysis included 162,257 patients who had surgical repair of a closed transcervical femoral neck fracture between 1990 and 2001 (Fig. ).
The case inclusion schema is shown for patients with femoral neck fractures undergoing HA, THA, or ORIF. ICD-9-CM = International Classification of Diseases, 9th Revision, Clinical Modification.
We grouped the admissions into the three periods: Period I (1990–1993), Period II (1994–1997), and Period III (1998–2001), creating approximately equal intervals. Patient race-ethnicity was categorized in the Nationwide Inpatient Sample databases as white, black, Hispanic, and other (including Asian or Pacific Islander and Native American). The category “black” may include African-American and Caribbean-American patients. We categorized age into younger than 50, 50 to 64, 65 to 79, and 80 years or older based on distribution of procedure utilization within age groups. A majority of patients managed surgically for femoral neck fractures were female (75% to 78% in Periods I through III), white (91% to 94% in Periods I through III), and in age groups of 65 to 79 and 80 years or older (Table ).
Baseline characteristics of patients managed surgically, 1990–2001
Hospital teaching status for a given year was obtained from the American Hospital Association Annual Survey by the HCUP. A hospital was considered a teaching hospital if it had an American Medical Association-approved residency program, was a member of the Council of Teaching Hospitals, or had a ratio of full-time-equivalent interns and residents to beds of 0.25 or greater. We characterized hospitals as urban if they were in a metropolitan statistical area and rural if they were in a nonmetropolitan statistical area. We calculated hospital volume per year for THA and HA separately using unique hospital identifiers provided in the databases. This was performed by looking at all THA and HA procedures performed by a given hospital in the entire Nationwide Inpatient Sample data sets. Thus, THA and HA annual volumes were calculated for each of these respective procedures performed in a hospital for any indication (and not only for femoral neck fractures). Annual hospital volume of THA was divided into zero procedures, one to 24 procedures, 25 to 99 procedures, 100 to 199 procedures, and 200 or more procedures; and that of HA was divided into zero procedures, one to 14 procedures, 15 to 25 procedures, 25 to 100 procedures, and 100 or more procedures. This was based on the distribution of procedure utilization in our patient population. Annual THA and HA surgeon volume was calculated in a similar way. Because surgeon identifiers were missing in 46% of patient records, we could not determine whether a surgeon did not perform any THAs or HAs in a year or the information was missing. Therefore, for records with available surgeon identifiers, we divided annual surgeon volume of THA into one to nine procedures, 10 to 19 procedures, 20 to 49 procedures, and 50 or more procedures; and that of HA was divided into one to four procedures, five to nine procedures, 10 to 20 procedures, and 20 or more procedures. We performed analyses to determine if surgeon identifier was especially likely to be missing in particular patient or provider groups. These analyses indicated the missing data were distributed evenly across patient and hospital characteristics.
We described patient demographics, surgeon characteristics, and hospital characteristics across periods using proportions. The overall association between times and the distribution of ORIF, HA, and THA utilization was examined using the chi square test of independence. Additional analyses focused on identification of factors influencing the association between times and the distribution of ORIF, HA, and THA. Factors considered were age, hospital volume, and surgeon volume. We performed statistical analyses using Intercooled STATA for Windows (version 8.2; Stata Corp, College Station, TX) and SAS for Windows (version 9.1; SAS Institute Inc, Cary, NC).