Declining induced abortion rate in Finland: data quality of the Finnish abortion register

Int J Epidemiol. 1996 Apr;25(2):376-80. doi: 10.1093/ije/25.2.376.

Abstract

Background: Induced abortion rates have declined in Finland since 1973. A possible explanation offered has been that of deteriorating data collection.

Methods: To assess the completeness of the Register, we compared the information from a consecutive sample of hospital records (N = 482) to the Finnish Abortion Register in 18 hospitals in three counties. A smaller consecutive sample (N = 345) was collected from the same hospitals to assess the validity of the Register information.

Results: Only five abortions (1 percent) found in the hospitals were not reported in the Abortion Register. A total of 95 percent of all the length of pregnancy (definition problems), the classification of the abortion procedure, and social class (out-of-date classifications). Furthermore, early complications were poorly reported.

Conclusions: The data from the Finnish Abortion Register are a reliable source for monitoring trends in the abortion rate and its variation by subgroups, but are an unreliable source for the study of the medical aspects of induced abortion.

PIP: The Development Centre for Welfare and Health in Helsinki, Finland, collected data on 345 abortions performed in 18 hospitals from three counties in November 1993. The data sources were medical records and the Abortion Register. The woman's personal identification number was used to link the data from the medical records to those from the Abortion Register. The Centre aimed to learn whether deteriorating data collection explains the decline in induced abortion rates in Finland since 1973. The record completeness rate was 99%. 95% of the information in the Abortion Register matched that in the medical records. 7 of 22 variables had good validity levels (=or+ 99% agreement). 11 had satisfactory validity levels (90-98.9%). Variables with unsatisfactory validity levels were woman's contraception after the abortion (89%), type of abortion procedure (88%), social class (80%), and gestation length (64%). Agreement on gestation length improved when using ultrasound instead of last menstruation (81%). Only 9 of 32 women with abortion complications that were in the medical records were also in the Abortion Register. Yet four of these complications were correctly reported for a data quality rate of 13%. Data quality indexes ranged from 75% to 82%. These findings show that the Abortion Register is a reliable source to monitor abortion trends and its variation by subgroups and that it is not reliable for studying the medical aspects of induced abortion.

MeSH terms

  • Abortion, Induced / classification
  • Abortion, Induced / statistics & numerical data*
  • Abortion, Induced / trends*
  • Adolescent
  • Adult
  • Bias
  • Data Collection / standards*
  • Female
  • Finland / epidemiology
  • Gestational Age
  • Humans
  • Medical Records / standards
  • Middle Aged
  • Pregnancy
  • Registries / standards*
  • Reproducibility of Results
  • Social Class