Data sources

The present study is based on two separate datasets. One dataset was compiled from nationwide Finnish register data, while the other includes materials from a specialized hospital for joint replacement (Coxa, Hospital for Joint Replacement, Tampere, Finland). These two datasets are later referred to as the national register series and the single center series respectively.

Both national register and single center series were created by combining data from several different sources. All the material was initially collected in electronic form (except for the records of the local infection register). Data from different sources were then combined using the unique personal identification numbers (see Gissler and Haukka 2004), date of operation and – when possible – the side of operated knee as match keys.

In the single center series, patient records were manually reviewed when contradicting or missing data was found in the electronic databases. Additionally, certain variables concerning the patients with infection in the single center series were collected from patient records using a custom case-record form and were then typed into electronic form.

7.1.1 The Finnish Arthroplasty Register

The Finnish Orthopaedic Association founded the Finnish Arthroplasty Register in 1980 (Paavolainen et al. 1991). Its purpose is to enable analyses of implant survival so that inferior prostheses and techniques can be detected, and to provide epidemiological data for health policy makers’ use (Puolakka et al. 2001, Rantanen et al. 2006).

In 1987 the responsibility for managing the register was transferred to the National Agency for Medicines, and currently the register is under the Ministry for

Social Affairs and Health. Since 1997 orthopedic surgeons have been obliged to make reports to the register on joint replacement operations (Puolakka et al. 2001).

Data on deaths and emigrated patients is routinely collected from the population register center data.

The unit of each register entry is a joint, and consequently, bilateral arthroplasty results in two entries in the register. Besides personal identification number and definition of the operated joint (hip, knee, other; laterality) the variables registered are the date of the operation, identification number of the operating hospital, indication for the operation, type and brand of the implanted femoral, tibial and patellar components (but not that of polyethylene tibial insert), fixation method of each component and use of bone grafts and prophylactic antimicrobial medication (see Puolakka et al. (2001) for the registration form).

For revision procedures, one predominant reason for revision is registered. Early postoperative complications can be reported on the regular notification form, and there is a separate form for reporting long-term complications. Revision surgeries and preceding operations are directly linked to each other.

The Finnish Arthroplasty Register has provided data for a number of studies on both hip and knee arthroplasty, and it gives feedback to the Finnish Orthopedic Society in its bi-annual yearbook on orthopedic endoprostheses (Rantanen et al.

2006, National Agency for Medicines 2008). In the mid-1990s, the coverage of the register was estimated by comparing its data to the records of the Finnish Hospital Discharge Register (see below). The Finnish Arthroplasty Register was found to detect approximately 95% of joint replacement operations performed yearly (Puolakka et al. 2001). So far, it has not been scientifically validated.

7.1.2 National Hospital Discharge Register

The Finnish national Hospital Discharge Register (of health care institutions) (later referred to as the Hospital Discharge Register), currently administered by the National Research and Development Centre for Welfare and Health (STAKES), was established in 1967 and is widely used in health research (Gissler and Haukka 2004). Its data are compiled from mandatory reports from health care institutions on discharges from hospitals. For each hospitalization requiring overnight stay

demographic data, identification number of the health care institution, dates of admission and discharge, primary and up to three subsidiary diagnoses and the codes of surgical procedures are registered among certain other patient administration -related variables.

It is estimated that the Hospital Discharge Register covers approximately 95% of all discharges. Its validity has been studied e.g. in the fields of cardiovascular disease, psychiatry and reproductive health and is considered good in general (Gissler and Haukka 2004). The validity of data on orthopedic diagnoses and procedures has not been assessed.

7.1.3 Hospital database “Tekoset”

The “Tekoset” database is a computerized registration system designed specifically to collect operative and outcomes data on endoprosthesis surgery (Lehto et al.

1999). The registration system covers the whole course of treatment, including preoperative, operative and follow-up data. Data is typed into the registration system in a prospective manner. Operative details are saved in the operating room by the assisting nurse and the operating surgeon approves the data entry.

The recorded operation-related variables include identification number of the operating surgeon, surgical approach and type, brand, size and fixation of each component, use of bone substitutes, list of additional procedures (osteotomies, soft tissue releases etc.), amount of blood loss, number of wound drains, wound closure method and the type of antimicrobial and antithrombotic treatment given perioperatively. Early postoperative complications are also recorded. Preoperative and follow-up information include measurement of joint function using the Knee Society ranking system (Insall et al. 1989) and analysis of radiographs according to a predefined scheme. All information recorded in the database can be exported to statistical software packages.

The validity of “Tekoset” data has not been evaluated. In this study, few discrepancies between Tekoset data and local hospital administration register entries were observed (data not shown), and all of them could be manually resolved.

7.1.4 Hospital infection register

The Coxa hospital collaborates with the local hospital infection register of Tampere University Hospital to perform infection surveillance. In addition, it belongs to the Finnish Hospital Infection Surveillance Program SIRO (Huotari et al. 2007a).

The local infection register is run by a team of trained nurses and is lead by a specialist in infectious diseases. Infection surveillance is prospective and active, and is based on NNIS methodology (Emori et al. 1991, Huotari et al. 2007a). Infections are classified according to the CDC criteria (Horan et al. 1992).

To identify SSI occurring after discharge patients are given questionnaires concerning problems related to wound healing upon discharge from the operating hospital. They are asked to return the questionnaires after a month even if they had not experienced any wound-related complications. In case of any problems related to the wound, the questionnaire is usually returned to the infection register by a doctor or a nurse who has clinically evaluated the patient. Over the study period the annual response rate in postdischarge surveillance varied between 71% and 83%.