healthmedicinet

Data

The study is based on the data collected within the framework of the ATLANTIC-DIP
collaborative which has been described previously 4]. In brief, ATLANTIC-DIP is a network of antenatal centers along the Irish Atlantic
seaboard serving a population of approximately 500,000 people. This regional area
can be considered as broadly representative of the whole population of Ireland 21]. Pregnant women were offered screening at 24–28 weeks’ gestation using a 75-g oral
glucose tolerance test (OGTT) with fasting, 1-h, and 2-h values. In total, 5,500 completed
the screening. Henceforth, we refer to the pregnancy in this period as an index pregnancy.
Of those screened, 12.4 % had GDM by IADPSG criteria 4].

In a subsequent follow-up study 22], 270 women with GDM (GDM group) and 388 women with NGT (NGT group) in the index pregnancy
returned for a follow-up screen using a 75-g OGTT 2 to 5 years after the index pregnancy.
In the current study, we build on this follow-up study by administering a postal questionnaire
(in March 2013) to all of these women. This questionnaire contained questions on HRQOL
assessed via the EQ-5D-3 L (with VAS component), healthcare service use, and socio-economic
characteristics. The VAS component valued the HRQOL on a scale anchored from 0 (the
worst health one can imagine) to 100 (the best health one can imagine). In total,
342 women responded to the questionnaire (52 % of the original sample) and were included
in the analysis: 231 of these women had NGT and 111 had GDM in the index pregnancy.
The participants of the study gave informed consent in writing at every stage of the
study, i.e. initial and follow-up screening, and follow-up survey. The study was approved
by the Irish Health Service Executive Clinical Research Ethics Committee (reference
C.A.537).

Outcomes

HRQOL was assessed via patient responses to EQ-5D-3 L. This is a multi-attribute instrument
to measure HRQOL based on five health dimensions: mobility, self-care, usual activities,
pain and discomfort, and anxiety and depression 23]. It also contains an instrument for a direct measurement of the self-reported HRQOL
via VAS. In this study, we focus on the self-reported HRQOL via VAS only. The multi-attribute
generic instrument EQ-5D-3 L produced highly homogenous results which in general exceeded
the self-reported status (see Additional file 1). Hence, we found this generic instrument to be insensitive for detecting HRQOL differences
in women affected by GDM.

Analysis showed that the HRQOL measurement had a skewed distribution which could not
be corrected by logarithmic transformation. Therefore, the utility decrement (henceforth
UD), representing the distance to perfect health was applied. The UD is measured as:

(1)

The distribution of UD was still found to significantly differ from the normal based
on a Shapiro-Wilk test, as was the HRQOL. However, its properties (unbounded on logarithmic
scale) preconditioned its use for further analysis in the regression models.

Covariates

For the multivariate analysis, a set of covariates were chosen to investigate the
impact of a diagnosis of GDM on HRQOL. Selection of covariates was based in part on
previous studies 10] and in part on that which was available in the dataset. The detailed definition of
the variables may be found in Additional file 2. Briefly, the main groups of factors were the indication of acquired abnormal glucose
tolerance after the pregnancy, age, life-style indicators (BMI, fruit and vegetable
intake, exercising 30 min a day), risky behavior in terms of alcohol and tobacco use,
mode of delivery in the index pregnancy, subsequent miscarriages, and a set of socio-economic
indicators (income, employment and cohabiting status). The two indicators, possession
of a ‘medical card’ and private health insurance, were introduced in the analysis
as they might have impacted on health service utilization, and consequently on HRQOL,
and may represent socio-economic status as well 24], 25]. In Ireland, a medical card is issued based on low income, age, or financial hardship
as a result of a medical condition and entitles the holder to free health care. Private
health insurance, depending on the level chosen, can cover some health care costs
and is usually purchased by people with high income. We also controlled for the time
elapsed after the index delivery before the HRQOL measurement.

Statistical analysis

A series of univariate analyses were undertaken to examine if the samples differed
by the characteristics mentioned above. This consisted of independent two-tailed t-tests
for continuous variables, and ?
²
tests for categorical variables. For the multivariate analyses, the choice of estimation
approach was informed by the nature of the dependent variable, i.e. utility decrement.
Specifically, a generalized linear model with log link and Gaussian family of the
error term distribution was selected (the distribution of the utility decrement is
intrinsically continuous and, therefore, Gaussian family was preferred, although Modified
Park’s test favored Poisson’s distribution). The coefficients in the model represent
the effect on the utility decrement on the logarithmic scale and, therefore, have
opposite effect on the HRQOL, i.e. positive coefficients mean reduction in HRQOL and
negative – increase in HRQOL.

The women in our study were selected from a number of antenatal study centers. In
order to detect explicitly if there were differences between the study centers, we
applied center-specific constant effects. Statistical significance was explored at
three levels (0.05, 0.01, and 0.001). All analyses were performed using STATA 12 software.