On the use of administrative databases in health care analyses

Abstract

The aim of this thesis is to demonstrate the potential of secondary administrative data sources in analyses of health care, its changes over time and its costs. The Swedish Cancer Registry the National Inpatient Registry and Social Insurance Board Registries of early retirement pensions, local databases and surveys were used.

Relative survival after colon cancer among 41,700 patients 1973-1990 improved substantially in all catchment area categories, the initial difference between categories decreasing with time. Analyses considering length of follow-up, time period, catchment area category and age at diagnosis confirmed this. The development is best explained by improved perioperative care.

Treatment of benign prostatic hyperplasia has changed dramatically. Cross-sectional annual data on health care 1987 to 1994 were used for calculation of direct costs. Drug sales increased rapidly from 1992, while conventional transurethral resections of the prostate decreased. Total costs showed a slight decrease.

Inflammatory bowel diseases (IBD) include Crohn's disease and ulcerative colitis. All types of health care, sick-leave and early retirement for IBD in 1994 were collected. Ambulatory care was concentrated to hospitals and specialists in internal medicine. One-fourth of the patients accounted for 48 per cent of hospital admissions. Aminosalicylates and steroids dominated medication. Sick leave episodes were on average 6 weeks long. Early retirements were uncommon but lasted 14 years on average. Morbidity accounted for 68 per cent of total costs. Among direct costs, admissions took 58 per cent. Neither complications nor surveillance added much to total costs.

Appendectomy in Sweden was assessed by data from the Inpatient Registry 1989-1993. For 60,306 patients operated on, diagnostic accuracy, incidence rate of appendicitis and perforated appendicitis, weekday of admission and hospital category were analysed. The age-standardised incidence rate of appendectomy decreased by 9.8 per cent in women and 4.1 per cent in men. With an almost constant number of patients discharged with an end diagnosis of appendicitis, diagnostic accuracy increased. This was more pronounced in women than men, seen in all hospital categories and higher for those admitted during periods of low capacity such as weekends and holidays. Perforated appendicitis was constant among women and increased non-significantly in men.

The introduction of laparoscopic operations brought an unprecedented increase in cholecystectomies. All 19,432 patients operated on in 1989 and 1994 recorded in the national Inpatient Registry were analysed. As a proxy variable indicating adverse outcome, the risk of experiencing a long admission was analysed. The odds ratio increased with age and acuteness of gallstone disease. The absolute number of patients operated on with acute gallstone disease changed little. Stratification revealed that their risk of a long admission was increased both in 1989 and 1994, particularly among women. Patients with chronic gallstone disease had no increased risk.

Secondary data sources in studies of health care permit analyses on the national level at low cost by reuse of data already available. Their retrospective nature and high case ascertainment allow studies which cannot be performed from small datasets. Most of their limitations can be managed by study design or in the analyses.