The accuracy of population health data for monitoring trends and outcomes among women with diabetes in pregnancy

https://doi.org/10.1016/j.diabres.2008.03.001Get rights and content

Abstract

Aim

To assess the accuracy of routinely collected population birth and hospital datasets in identifying maternal pregestational diabetes mellitus (PDM) and gestational diabetes mellitus (GDM).

Methods

Information on maternal diabetes status was obtained from the medical records of a random sample of 1200 women and compared with routinely collected, population-based birth and hospital data. PDM and GDM are reported in both databases. Sensitivity, specificity, positive predictive value (PPV), negative predictive value and the kappa statistic were determined.

Results

Medical records were available for 1184 of the 1200 women sampled. 0.3% of women were classified with PDM and 4.8% with GDM. ‘True’ PDM was under-reported and misclassified in the birth data, but all cases were reported in the hospital data. GDM was also more completely and more accurately reported in the hospital data than in the birth data. Diabetes requiring insulin was more likely to be reported than non-insulin dependent diabetes.

Conclusions

Hospital data were more sensitive and accurate (higher PPVs) than birth data and these measures were not improved by ascertaining diabetes from either of the two datasets. More severe forms of diabetes were more likely to be reported than less severe.

Introduction

Pregestational and gestational diabetes mellitus (PDM and GDM) are common complications of pregnancy, affecting around 0.3–0.6% and 2–9% of pregnancies respectively [1], [2], [3], [4], [5], [6]. Both forms of diabetes are associated with adverse outcomes for mothers and their infants [4], [7], [8], [9], [10], [11].

The population coverage and availability of routinely collected, population health data sets (PHDS) make PHDS a cost-efficient resource for estimating the prevalence of PDM and incidence of GDM in pregnant populations, and for assessing the health outcomes for these women and their infants [12], [13]. These type of data are used internationally to monitor diabetes and as the basis of research studies, usually without regard for the validity of the reporting [2], [10], [14], [15], [16]. There are, however, limitations relating to the completeness and validity of data in studies utilising single datasets [4], [12], [13]. For conditions like diabetes where information is available from more than one dataset, linkage of PHDS may reduce the problem of under-ascertainment, but this allows the possibility of discordant reports [17], [18], [19]. Only one recent study has assessed the accuracy and reliability of the reporting of diabetes in pregnancy in population health data [17]. The aim of this study was to assess the accuracy of the individual birth and hospital datasets alone, and either of these datasets, in identifying maternal PDM and GDM.

Section snippets

Data sources

New South Wales (NSW) is the most populous Australian state with a population of ∼6.8 million and 83,000 births per annum in over 100 hospitals. The population health data for this study were obtained from two NSW Department of Health routinely collected datasets, the NSW Midwives Data Collection (MDC) and the NSW Admitted Patient Data Collection (APDC). The MDC (referred to as ‘birth data’) is a legislated surveillance system covering all NSW births ≥20 weeks gestation or ≥400 g birthweight,

Results

Medical records were available for 1184 of the 1200 women sampled. Demographic and pregnancy characteristics of the weighted sample were not statistically different from the population [23]; 18.2% were aged 35 years or older, 39.9% were primiparous, 10.3% were admitted to rural hospitals, 44.3% to tertiary hospitals, and 93.3% had a gestational age of 37 weeks or more. Diabetes prevalence is reported in Table 1 and the accuracy of the diabetes coding in the individual birth and hospital

Discussion

This study reports the accuracy and reliability of contemporary birth and ICD10-coded hospital discharge data (individual and linked) in identifying diabetes during pregnancy. We have demonstrated that hospital data are more accurate and reliable than birth data, and that PDM is more reliably reported than GDM. Where available, hospital data should be used in preference to birth data to monitor population rates of diabetes during pregnancy.

Assessment of the validity of PHDS is important from

Conflict of interest statement

The authors declare that they have no conflict of interest.

Acknowledgements

We wish to acknowledge the help of Margie Pym in collecting the data. Jane Ford is supported by the Health Research and Outcomes Network, a National Health and Medical Research Council (NHMRC) Capacity Building Grant. Christine Roberts is a NHMRC Senior Research Fellow. This study was supported by a NHMRC Project Grant (#402498).

References (34)

  • L. Hoffman et al.

    Gestational diabetes mellitus–management guidelines. The Australasian Diabetes in Pregnancy Society

    Med. J. Aust.

    (1998)
  • American Diabetes Association

    Diagnosis and classification of diabetes mellitus

    Diab. Care

    (2007)
  • CEMACH, Confidential Enquiry into Maternal and Child Health, Pregnancy in Women with Type 1 and Type 2 Diabetes in...
  • I.M. Evers et al.

    Risk of complications of pregnancy in women with type 1 diabetes: nationwide prospective study in the Netherlands

    BMJ

    (2004)
  • C.A. Crowther et al.

    Effect of treatment of gestational diabetes mellitus on pregnancy outcomes

    N. Engl. J. Med.

    (2005)
  • A.W. Shand, J.C. Bell, A. McElduff, J. Morris, C.L. Roberts, Outcomes of pregnancies in women with pre-gestational...
  • R. Bell et al.

    Trends in prevalence and outcomes of pregnancy in women with pre-existing type I and type II diabetes

    BJOG

    (2008)
  • Cited by (51)

    • Maternal and perinatal outcomes associated with the use of renin-angiotensin system (RAS) blockers for chronic hypertension in early pregnancy

      2018, Pregnancy Hypertension
      Citation Excerpt :

      Given that identification of chronic hypertension from the perinatal record and hospital admission record associated with the delivery results in under ascertainment [17], diagnoses from prior perinatal and hospital records ((ICD10-AM: O10), chronic hypertension with superimposed preeclampsia (O11) and essential primary hypertension (I10)) was also used [17]. Other indications for which RAS blockers could have been prescribed were also identified, including chronic kidney disease, diabetic nephropathy, cardiac disease [18] and pre-existing diabetes [19] using recommended methods [18,19]. Women who had both chronic hypertension and another of these diagnoses were excluded.

    • Early-onset preeclampsia appears to discourage subsequent pregnancy but the risks may be overestimated

      2016, American Journal of Obstetrics and Gynecology
      Citation Excerpt :

      Although the risk of delivery per week increases steadily from 32 weeks gestation, the health risks that are associated with preterm birth decrease correspondingly.54-56 The strength of this study is the use of large, reliably collected longitudinally linked, population-based data24,26,57-60 over a recent 12-year period. Advantages of these data include the ability to link pregnancies longitudinally to the same women, multiple fields for recording diagnoses and procedures, and extensive validation of completeness and accuracy.

    View all citing articles on Scopus
    View full text