SEARCH

SEARCH BY CITATION

Keywords:

  • Economics;
  • healthcare costs;
  • medical complications;
  • obesity;
  • pregnancy

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Objective

To investigate the impact of maternal body mass index (BMI, kg/m2) on clinical complications, inpatient admissions, and additional short-term costs to the National Health Service (NHS) in Scotland.

Design

Retrospective cohort study using an unselected population database.

Setting

Obstetric units in Scotland, 2003–2010.

Population

A total of 124 280 singleton deliveries in 109 592 women with a maternal BMI recorded prior to 16 weeks of gestation.

Methods

Population-based retrospective cohort study of singleton deliveries, with multivariable analysis used to assess short-term morbidity and health service costs.

Main outcome measures

Maternal and offspring outcomes, number and duration of hospital admissions, and healthcare costs.

Results

Using multivariable analysis, in comparison with women of normal weight, women who were overweight, obese, or severely obese had an increased risk of essential hypertension [1.87 (1.18–2.96), 11.90 (7.18–19.72), and 36.10 (18.33–71.10)], pregnancy-induced hypertension [1.76 (1.60–1.95), 2.98 (2.65–3.36), and 4.48 (3.57–5.63)], gestational diabetes [3.39 (2.30–4.99), 11.90 (7.54–18.79), and 67.40 (37.84–120.03)], emergency caesarean section [1.94 (1.71–2.21), 3.40 (2.91–3.96), and 14.34 (9.38–21.94)], and elective caesarean section [2.06 (1.84–2.30), 4.61 (4.06–5.24), and 17.92 (13.20–24.34)]. Compared with women of normal weight, women who were underweight, overweight, obese, or severely obese were associated with an 8, 16, 45, and 88% increase in the number of admissions, respectively, and women who were overweight, obese, or severely obese were associated with a 4, 9, and 12% increase in the duration of stay (all < 0.001). The additional maternity costs [mean (95% CI), adjusted analyses] for women who were underweight, overweight, obese, or severely obese were £102.27 (£48.49–156.06), £59.89 (£41.61–78.17), £202.46 (£178.61–226.31), and £350.75 (£284.82–416.69), respectively.

Conclusions

Maternal BMI influences maternal and neonatal morbidity, the number and duration of maternal and neonatal admissions, and health service costs.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Maternal obesity is defined as a body mass index (BMI) > 30 kg/m2. It affects more that 30% of the antenatal population in the UK and is the most common co-morbidity of pregnancy.[1] Maternal obesity is associated with an increased risk of maternal and offspring morbidity and mortality during pregnancy, with complications including gestational diabetes, hypertensive disorders (including pre-eclampsia), thromboembolic complications,[2-4] induction of labour, operative delivery[3, 5] late stillbirth,[6, 7] birth injury,[8] and admission to the neonatal unit.[9] Although these associations have been well reported, most studies report the consequences of maternal obesity for pregnancy in geographically confined cohorts that may not be generalisable to the population level. Furthermore, very few studies have estimated the impact that high (and low) BMI has on hospital admissions and short-term costs to the health service. This contrasts with the situation outside pregnancy where the clinical and economic burden of obesity-related diseases unrelated to pregnancy has been extensively modelled, with costs projected to rise by £1.9–2.0 billion/year and $48–66 billion/year by 2020 in the UK and USA, respectively.[10, 11] The paucity of this population-level data about the clinical and economic burden of obesity for pregnancy limits the ability of policy makers to make informed decisions about the design and delivery of maternity services for an increasingly complex obstetric population. Furthermore, it makes evaluation of the cost-effectiveness of the current public health strategies to reduce the prevalence and clinical consequences of obesity in women of reproductive age very difficult.[12-15]

The objective of our study was therefore to investigate the impact of high (and low) maternal BMI on clinical complications and inpatient admissions, and to estimate the additional short-term costs to the National Health Service (NHS) at the level of the Scottish population.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Admissions data set

We obtained approval for the study from the Caldicott Guardians and from the Privacy and Advisory Committee at the Information Services Division (ISD) of NHS Scotland. Anonymised data were retrieved on all singleton deliveries in Scotland, between January 2003 and February 2010, from the Scottish Morbidity Records 2 (SMR02) held at the ISD of NHS Scotland. An SMR02 record is completed at the time of discharge of every pregnant woman from a Scottish maternity hospital, and is estimated to be more than 98% complete.[16] The record contains basic demographic information about the woman and detailed information about the nature and indication for the hospital admission.

The following demographic information was extracted from SMR02 for every pregnant woman: parity, maternal age, weight, height, smoking status, and deprivation category. For women with more than one pregnancy during the study period, information was recorded for each pregnancy. Maternal deprivation was categorised using the Scottish Carstairs 2001 quintiles, with 1 being the least deprived and 5 being the most deprived.[17] Maternal BMI was calculated (weight/height2, kg/m2) using maternal height and weight values recorded before 16 weeks of gestation. Women were excluded from the analysis if height and weight values were recorded after 16 weeks of gestation to eliminate any incorrect categorisation of maternal BMI, which might arise from physiological changes in weight with gestation. Height values were deemed valid within the range 122–220 cm, and similarly weights were considered valid within the range 35–140 kg. Parity and maternal age were deemed valid if they were <20 and >8 years, respectively. Delivery gestation and offspring birthweight were deemed valid if they were ≤44 weeks of gestation and ≤6 kg, respectively. For women who contributed more than one pregnancy to the data set, subsequent pregnancies were excluded if the inter-pregnancy interval was less than 120 days. Women were grouped into five categories according to their BMI, using World Health Organization (WHO) criteria: BMI < 18.5 kg/m2 (underweight); BMI 18.5 < 25 kg/m2 (normal); BMI 25 < 30 kg/m2 (overweight); BMI 30 < 40 kg/m2 (obese); and BMI ≥ 40 kg/m2 (severely obese).[18] The reference group used for all analyses were women with a normal BMI (18.5 < 25 kg/m2).

Clinical outcomes were extracted from SMR02 and coded using the tenth revision of the International Statistical Classification of Diseases and Related Health Problems (ICD-10) criteria.[19] Maternal outcomes, included pre-existing essential hypertension, pregnancy-induced hypertension, eclampsia, diabetes (pre-existing and gestational), ventouse/forceps delivery, caesarean section (emergency and elective), third- or fourth-degree tears, and induction of labour. Offspring outcomes included preterm birth (spontaneous and iatrogenic or induced), stillbirth and neonatal death. Spontaneous preterm birth was defined as a birth between 24 and 36 weeks of gestation in which the onset of labour was spontaneous, and delivery was either vaginal or by emergency caesarean section. An induced or iatrogenic preterm birth was defined as a birth between 24 and 36 weeks of gestation after induction of labour, or by elective caesarean section, with no evidence of induced delivery.

The following information was extracted from SMR02 about each inpatient or day-case admission: hospital of admission, length of stay, type of admission, and healthcare provider. The Office of Population, Censuses, and Surveys Classification of Surgical Operations and Procedures (OPCS) codes and/or ICD-10 diagnosis codes were used to classify the clinical indication for admission, as appropriate.[19]

Each admission was linked to a pregnancy and each pregnancy was linked to an individual woman and birth record, with some women experiencing two or more pregnancies and/or multiple admissions within the study period.

Analysis of clinical outcomes

Descriptive statistics for maternal age at admission, smoking status during pregnancy, and the Scottish Carstairs 2001 deprivation quintiles were summarised by BMI group.[17]

As pregnancy outcomes within individual women are correlated, standard statistical methods that assume independence between observations were not considered appropriate for the analysis. To account for the nesting of pregnancies within women, all pregnancies for an individual woman were therefore considered simultaneously within hierarchical regression models for each outcome of interest.[20-22] Multinomial logistic regression was used to examine the association between maternal BMI and various demographic characteristics, including smoking and the Carstairs quintile. To estimate the association between maternal BMI and binary clinical outcomes (e.g. gestational diabetes), random-effects logistic regression models were used. The relationship between maternal BMI group and each of the maternal, perinatal, and neonatal complications was expressed as an odds ratio (OR) and 95% confidence interval (95% CI) before and after adjustment for maternal age, smoking, and Carstairs quintile.[23]

To estimate the effect of variables, including maternal BMI, maternal age, smoking, and Carstairs quintile, on the continuous outcome of birthweight, a random-effects mixed linear regression model was used. The birthweight data were expressed as the mean (SE) birthweights of babies born to women who were underweight, overweight, obese, or severely obese, compared with women of normal BMI.

To confirm the results of the random-effects model, the population-averaged model was used in sensitivity analyses. The population-averaged model is widely used to analyse clustered data. It uses an exchangeable structure for the working correlation matrix, and gives a lower estimate than the random-effects model.[24, 25] The resultant population-averaged estimated risk for a specific maternal BMI category is the average of the estimated risks from the random-effects model for each maternal BMI category across the women.

Costing admissions

Initially, hospital and specialty-specific unit costs, published by ISD,[26] were used to estimate the cost of each hospital admission based on the admitting hospital, type of care provided (consultant or midwifery lead), and length of stay. The approach described by McDougall and Buchanan was followed when applying these unit costs.[27] This approach captures variation in admission costs, within specialties, driven by duration of stay and type of care provided, but is less sensitive to variation in cost driven by the intensity of resource use required to manage admissions for different clinical indications.

As an alternative approach, each maternal admission was also assigned a Department of Health Healthcare Resource Group (HRG4) code, based on the OPCS procedure and ICD-10 codes assigned for each admission,[28] and the NHS reference cost for the relevant HRG was assigned.[14] This method captures variation in admission costs driven by the intensity of resource use required for specific clinical procedures and indications, but is less influenced by duration of stay. However, adjustments to admission costs were made where the length of stay exceeded a defined costing threshold (known as the trim-point) for each HRG.[28] In these cases the cost per excess bed day was applied per admission day over and above the trim-point.[14]

To calculate the total maternal inpatient and day-case costs associated with each pregnancy, the individual costs of each admission associated with each pregnancy were then summed (for both the ISD and HRG approaches).

Accurate costing of neonatal admissions was not possible because the reason for, and duration of, any neonatal stay was not reliably recorded. As such, neonatal costs have been excluded from this analysis. However, neonatal admission (<48 hours and ≥48 hours) was included as a clinical outcome.

Analysis of cost data

The mean number of admissions, length of stay, and associated mean costs were summarised across BMI groups. A simple one-way anova was used to test for associations between maternal BMI and each of these outcome variables. The Bonferroni method was used to adjust the P values for multiple testing.[29]

Generalised linear models (GLMs) were used to assess the effect of BMI on maternal admission costs, with and without adjustment for covariates (Table 5).[30] The results are expressed as the estimated changes in mean admission costs associated with having a low or high BMI, compared with having a BMI in the normal range. A gamma distribution with identity link function was specified for each GLM model to account for the skewed nature of the cost data, with the exception of model 6 (Table 5), which used a Gaussian distribution because of problems with convergence. To account for the lack of independence between consecutive pregnancies within individual women, a robust cluster variance estimator was used to derive standard errors and confidence intervals for the estimated changes in admission costs.[31] A random-effects model was also assessed as an approach for dealing with repeated observations within women, and was found to produce similar results to those of the GLM model. As the GLM method is better suited to dealing with the skewed nature of the cost data, this approach was adopted.

The unadjusted effect of BMI on maternal admission costs was estimated by regressing total ISD and HRG costs (per pregnancy) on maternal BMI category. Following this, the cost models (ISD and HRG) were adjusted for the sociodemographic indicators of age, smoking, deprivation, hospital of delivery, and parity. The effects of BMI on cost were then adjusted for the adverse clinical outcomes found to be associated with low and high BMI. This was to assess the extent to which these clinical outcomes account for the observed increased admission costs associated with low and high BMI.

Finally, to investigate whether the increased maternity costs associated with low and high BMI are attributable to more frequent and/or longer hospital admissions, Poisson regression models were specified for number of admissions and total number of admission days. The total number of admissions was included as a covariate in the latter model to separate out the effect of BMI on the duration of stay per admission. The outputs from these models are reported as incidence-rate ratios (IRRs).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Descriptives for data set

Over the study period, there were a total of 394 014 singleton deliveries nested within 304 198 women. Of the 124 467 pregnancies that fulfilled the validation criteria, including having a maternal BMI calculated using valid height and weight measurements taken prior to 16 weeks of gestation, 187 pregnancies were further excluded because the inter-pregnancy interval was less than 120 days (Figure 1). The final data set used for analysis therefore comprised 124 280 deliveries nested in 109 592 women, equating to 32% of singleton deliveries in Scotland during the study period.

image

Figure 1. Flow diagram for study.

Download figure to PowerPoint

The data set included 95 638 (87.3%) women with a single pregnancy, 13 242 (12.1%) women with two pregnancies, 691 (0.6%) women with three pregnancies, 20 (0.02%) women with four pregnancies, and a single woman with five pregnancies. We found that 76.3% of women with a normal BMI, 64.1% of women who were overweight, 74.9% of women who were obese, and 62.3% of women who were severely obese in the index pregnancy remained in the same BMI group in their next pregnancy.

The demographics for the final study population are presented in Table 1. The prevalence of smoking was highest in underweight women (< 0.001). Underweight and severely obese women had a higher proportion of women in the highest deprivation quintile and a lower proportion of women in the lowest deprivation quintile, compared with women of normal weight, or those who were overweight or obese (< 0.001). Maternal age at delivery increased with increasing BMI category (< 0.001).

Table 1. Distribution of demographics characteristics across maternal BMI group for study population
Demographic characteristicMaternal BMI group (kg/m2)
≤18.49 (n = 3607)18.5–24.99 (n = 61 232)25–29.99 (n = 35 087)30–39.99 (n = 21 634)≥40 (n = 2720)
N (%)N (%)N (%)N (%)N (%)
Ever-smoked during pregnancy 1266 (35.1)13 176 (21.5)6842 (19.5)4315 (20.0)458 (16.8)
Highest Carstairs quintile of any pregnancy
1 least deprived411 (11.4)10 893 (17.8)5732 (16.3)2884 (13.3)315 (11.6)
2538 (14.9)11 513 (18.8)6298 (18.0)3848 (17.8)44 (16.2)
3729 (20.2)11 556 (18.9)6641 (18.9)3961 (18.3)53 (19.6)
4779 (21.6)12 739 (20.8)7786 (22.2)5085 (23.5)599 (22.0)
5 most deprived1,142 (31.7)14 319 (23.4)8507 (24.3)5765 (26.7)824 (30.3)
Missing8 (0.2)212 (0.4)123 (0.4)91 (0.4)8 (0.3)
Minimum maternal age at delivery, mean (SD) 24.9 (5.9)28.4 (6.0)29.3 (5.9)29.1 (5.8)29.4 (5.8)
Maximum maternal age at delivery, mean (SD) 25.6 (5.9)29.0 (6.1)29.9 (5.8)29.8 (5.7)30.0 (5.7)

Clinical complications and maternal BMI

Table 2 demonstrates the prevalence of maternal, delivery, and neonatal complications expressed as a proportion of the total number of pregnancies across maternal BMI group. Unadjusted ORs are shown in Table S1. Adjusted ORs (adjusted for maternal age, deprivation, and smoking) for complications are shown in Table 3. There was a dose-dependant increase in the odds of developing many complications, including essential hypertension, pregnancy-induced hypertension, pre-existing diabetes, gestational diabetes, other maternal diseases, induction of labour, emergency and elective caesarean section, admission to the neonatal unit, and iatrogenic preterm birth, with increasing maternal BMI. The greatest increases in adjusted ORs of complications associated with women who were overweight, obese, or severely obese were observed for pre-existing diabetes, gestational diabetes, essential hypertension, and elective and emergency caesarean section. The only complication that demonstrated a dose-dependant decrease in adjusted odds from overweight to obese and severely obese was forceps and ventouse delivery.

Table 2. Maternal complications (numbers shown relate to number of pregnancies not to number of women) for maternal complications across maternal BMI groups
 Maternal BMI group (kg/m2)
≤18.49 (n = 3607)18.50–24.99 (n = 61 232)25.00–29.99 (n = 35 087)30.00–39.99 (n = 21 634)≥40.00 (n = 2720)
n (%)n (%)n (%)n (%)n (%)
  1. *Other: breech and other modes of delivery.

  2. **Distribution of tears, among those with vaginal delivery (i.e. spontaneous vertex delivery, forceps delivery or ventouse delivery).

Maternal complication
Essential hypertension2 (0.1)80 (0.1)80 (0.2)166 (0.8)47 (1.7)
Pregnancy induced hypertension62 (1.7)1592 (2.6)1403 (4.0)1268 (5.9)211 (7.8)
Eclampsia2 (0.1)25 (0.0)20 (0.1)6 (0.0)4 (0.1)
Placenta praevia8 (0.2)99 (0.2)66 (0.2)51 (0.2)4 (0.1)
Abruption30 (0.8)546 (0.9)271 (0.8)172 (0.8)23 (0.8)
Antepartum haemorrhage23 (0.6)430 (0.7)215 (0.6)128 (0.6)16 (0.6)
Pre-existing diabetes4 (0.1)135 (0.2)157 (0.4)128 (0.6)28 (1.0)
Gestational diabetes1 (0.0)84 (0.1)131 (0.4)205 (0.9)82 (3.0)
Any other disease424 (11.8)5728 (9.4)3515 (10.0)2938 13.6)529 (19.4)
Induction of labour691 (19.2)12 734 (20.8)8615 (24.6)6151 (28.4)863 (31.7)
Mode of delivery
Normal2482 (68.8)39 529 (64.6)21 071 (60.1)12 357 (57.1)1302 (47.9)
Ventouse/forceps511 (14.2)8233 (13.4)4133 (11.8)1932 (8.9)203 (7.5)
Elective caesarean237 (6.6)5017 (8.2)3902 (11.1)3035 (14.0)543 (20.0)
Emergency caesarean339 (9.4)7772 (12.7)5630 (16.0)4074 (18.8)651 (23.9)
Other*38 (1.1)682 (1.1)351 (1.0)236 (1.1)21 (0.8)
Tears **
1st–4th-degree or vaginal or cervical laceration1197 (41.0)22 265 (48.2)12 450 (50.9)7124 (51.2)736 (50.2)
Neonatal complication
Birthweight mean (SD)3126.8 (568.0)3365.1 (558.6)3476.9 (577.1)3523.6 (605.9)3555.3 (667.6)
Apgar ≤ 5 (>5 = ref.)29 (0.8)516 (0.9)280 (0.8)176 (0.9)26 (1.0)
Stillbirth14 (0.4)242 (0.4)148 (0.4)122 (0.6)20 (0.7)
Neonatal death4 (0.1)26 (0.0)17 (0.0)12 (0.1)2 (0.1)
Admission to neonatal unit (no admission = ref)
≤ 48 hours138 (3.8)2172 (3.5)1338 (3.8)950 (4.4)147 (5.4)
> 48 hours240 (6.7)2619 (4.3)1533 (4.4)1115 (5.2)182 (6.7)
Preterm birth
Spontaneous236 (6.5)2676 (4.4)1350 (3.8)975 (4.5)139 (5.1)
Iatrogenic75 (2.1)770 (1.3)541 (1.5)386 (1.8)68 (2.5)
Table 3. Adjusted odds ratios for maternal, delivery, and neonatal complications in the study population
 Maternal BMI group (kg/m2)
≤ 18.49 OR (95% CI)18.50–24.99 OR (95% CI)25.00–29.99 OR (95% CI)30.00–39.99 OR (95% CI)≥40.00 OR (95% CI)
  1. Adjusted for maternal age, deprivation, and smoking. Odds ratios significant at < 0.05 (highlighted in bold).

Maternal complication
Essential hypertension0.60 (0.10–3.82)1.00 1.87 (1.18–2.96) 11.90 (7.18–19.72) 36.10 (18.33–71.10)
Pregnancy-induced hypertension 0.63 (0.46–0.86) 1.00 1.76 (1.60–1.95) 2.98 (2.65–3.36) 4.48 (3.57–5.63)
Eclampsia1.49 (0.35–6.34)1.001.38 (0.76–2.49)0.67 (0.28–1.64) 3.50 (1.21–10.11)
Placenta praevia1.70 (0.82–3.52)1.001.08 (0.80–1.48)1.38 (0.98–1.94)0.85 (0.31–2.31)
Abruption0.89 (0.60–1.30)1.000.86 (0.74–1.01)0.89 (0.74–1.06)0.95 (0.61–1.46)
Antepartum haemorrhage0.88 (0.57–1.35)1.000.87 (0.73–1.03)0.83 (0.68–1.02)0.82 (0.49–1.37)
Pre-existing diabetes0.33 (0.07–1.53)1.00 2.54 (1.79–3.61) 3.76 (2.59–5.47) 7.71 (4.04–14.71)
Gestational diabetes0.22 (0.02–2.02)1.00 3.39 (2.30–4.99) 11.90 (7.54–18.79) 67.40 (37.84–120.03)
Other maternal diseases 1.31 (1.12–1.53) 1.00 1.13 (1.06–1.21) 1.86 (1.73–2.00) 3.60 (3.07–4.22)
Delivery complication
Induction of labour 0.85 (0.77–0.94) 1.00 1.30 (1.25–1.35) 1.64 (1.57–1.72) 1.97 (1.78–2.18)
Mode of delivery (normal = ref.)
Ventouse/forceps1.04 (0.93–1.16)1.00 0.93 (0.89–0.98) 0.73 (0.68–0.77) 0.72 (0.61–0.85)
Elective caesarean section1.08 (0.78–1.50)1.00 2.06 (1.84–2.30) 4.61 (4.06–5.24) 17.92 (13.20–24.34)
Emergency caesarean section 0.52 (0.36–0.74) 1.00 1.94 (1.71–2.21) 3.40 (2.91–3.96) 14.34 (9.38–21.94)
Tears (none = ref.)
1st–4th-degree or vaginal or cervical laceration 0.82 (0.74–0.90) 1.00 1.12 (1.07–1.16) 1.16 (1.10–1.21) 1.09 (0.96–1.24)
Neonatal complications
Apgar ≤ 5 (>5 = ref.)0.94 (0.65–1.38)1.000.96 (0.83–1.11)0.99 (0.83–1.17)1.22 (0.82–1.81)
Stillbirth0.92 (0.53–1.60)1.001.07 (0.87–1.32) 1.43 (1.14–1.79) 1.91 (1.19–3.07)
Neonatal death2.14 (0.74–6.20)1.001.14 (0.62–2.11)1.26 (0.64–2.51)1.65 (0.39–6.97)
Admission to neonatal unit (no admission = ref.)
≤48 hours1.10 (0.91–1.33)1.001.07 (0.99–1.15) 1.25 (1.15–1.36) 1.58 (1.31–1.90)
>48 hours 1.66 (1.39–1.97) 1.001.04 (0.96–1.12) 1.28 (1.17–1.40) 1.87 (1.53–2.28)
Preterm birth
Spontaneous 1.56 (1.31–1.87) 1.00 0.85 (0.79–0.93) 1.03 (0.94–1.14)1.25 (1.00–1.55)
Iatrogenic 1.74 (1.31–2.31) 1.00 1.25 (1.10–1.42) 1.45 (1.26–1.68) 2.12 (1.57–2.86)

After adjusting for maternal age, deprivation, and smoking, the mean birthweight of babies born to women who were underweight was lower [mean (SE), –181 g (9 g)], and the mean birthweights of babies born to women who were overweight, obese, or severely obese were higher [+104 g (4), +155 g (5 g), and +179 g (13), respectively], compared with women with a normal BMI (all < 0.001; data not shown).

Although the effect sizes calculated from the population-averaged models were generally smaller in magnitude than the effect sizes from the random-effect models, the relationship between maternal BMI and clinical complications was consistent between the two models (data not shown).

Relationship between maternal BMI, inpatient admissions, and costs

The mean number of admissions, inpatient days, and costs by maternal BMI group are presented in Table 4. The mean number of maternal admissions was higher and the duration of admissions longer for women who were underweight, overweight, obese, or severely obese, compared with women of normal weight (all < 0.01). This is reflected in the higher maternal admission costs observed for women in these BMI categories using both the ISD and HRG costs.

Table 4. Number of admissions, inpatient days, and costs (£) by maternal BMI group for inpatient data set
 Maternal BMI group (kg/m2)P (one-way anova)
≤ 18.4918.50–24.9925.00–29.9930.00–39.99≥ 40.00
  1. *Values expressed as mean (SD).

  2. **P value for post-hoc pairwise comparison with normal BMI group <0.01 (P values for post-hoc comparisons adjusted for multiple testing using the Bonferroni method).

Maternal admissions (n)*1.40 (2.10)**1.14 (1.83)**1.30 (2.05)**1.64 (2.40)**2.09 (2.87)**<0.01
Admission days (n)*3.59 (4.45)**3.31 (3.60)3.47 (3.71)**3.79 (4.06)**4.29 (4.15)**<0.01
Maternal admission costs (£)
ISD model (mean, SD)*3519.15 (2851.32)**3206.57 (2350.68)3344.46 (2449.99)**3677.23 (2737.81)**4091.27 (3014.08)**<0.01
HRG model (£)*2250.46 (1765.40)**2103.47 (1505.36)2273.80 (1630.21)**2554.05 (1819.26)**2927.14 (1983.95)**<0.01

Table 5 presents the estimated mean increases in maternal admission costs associated with having a low or high BMI, compared with a normal BMI, derived from the general linear models. Based on the unadjusted models (Table 5: models 1 and 4), the estimated increased admission costs range from £312 to £145 for women who were underweight, from £138 to £170 for women who were overweight, from £472 to £451 for women who were obese, and from £885 to £823 for women who were severely obese, using ISD and HRG cost data, respectively.

Table 5. Additional maternal admission costs
Additional costs (mean, 95% CI)Maternal BMI (kg/m2)
≤ 18.4918.50–24.9925.00–29.9930.00–39.99≥ 40.00
  1. ***P < 0.001.

  2. a

    Unadjusted costs.

  3. b

    Adjusted for sociodemographics.

  4. c

    Adjusted for sociodemographics and clinical complications.

ISD costing model (£)
Model 1a312.58 (215.98–409.18)***137.89 (105.54–170.23)***470.65 (428.40–512.92)***884.70 (768.40–1001.00)***
Model 2b168.87 (83.63–254.12)***157.37 (129.25–185.49)***472.34 (434.45–510.23)***864.75 (766.64–962.86)***
Model 3c145.35 (71.13–219.58)***58.55 (34.07–83.04)***234.76 (201.08–268.44)***428.69 (342.39–514.99)***
HRG costing model (£)
Model 4a146.99 (86.69–207.29)***170.34 (149.04–191.64)***450.58 (422.58–478.59)***823.68 (746.82–900.53)***
Model 5b91.40 (35.91–146.90)***149.97 (130.85–169.09)***399.08 (373.16 425.01)***754.93 (683.35–826.51)***
Model 6c102.27 (48.49–156.06)***59.89 (41.61–78.17)***202.46 (178.61–226.31)***350.75 (284.82–416.69)***

Following adjustment for maternal age, smoking, deprivation, and parity (Table 5: models 2 and 5), the estimated increased admission costs are reduced somewhat, although the observed J-shaped relationship between cost and BMI category remains. Further adjustment for adverse clinical outcomes associated with maternal BMI was found to substantially attenuate the observed relationship between BMI and admission costs (Table 5: models 3 and 6), suggesting that these adverse clinical outcomes account for ~50% of the increased admission costs associated having a low or high BMI. However, the effect of low or high BMI on admission costs remains positive and significant following this adjustment.

To determine whether the increased costs associated with low or high maternal BMI were attributable more to the number of admissions or longer durations of stay (per admission), incidence rate ratios for number of admissions and admission days were estimated using Poisson regression analysis (Table 6). Underweight, overweight, obesity, and severe obesity were associated with an 8, 16, 45, and 88% increase in the incidence of admissions, compared with normal weight. After controlling for the total number of admissions, overweight, obesity, and severe obesity were associated with a 4, 9, and 12% increase in the duration of hospital admission. Increases in the frequency of admission appear to contribute more to the observed increased costs associated with high maternal BMI than does the duration of stay following admission.

Table 6. Incidence risk ratios for number and duration of visits by maternal BMI
 Maternal BMI (kg/m2)
≤ 18.49 OR (95% CI)18.50–24.9925.00–29.99 OR (95% CI)30.00–39.99 OR (95% CI)≥ 40.00 OR (95% CI)
  1. Data are expressed as incidence risk ratios: *P < 0.5; **< 0.01; ***< 0.001.

Number of visits
Antenatal1.09*** (1.04–1.15)1.01.16*** (1.14–1.18)1.45*** (1.42–1.48)1.87*** (1.78–1.97)
Postnatal0.84* (0.69–1.01)1.01.22*** (1.14–1.31)1.47*** (1.46–1.58)2.26*** (1.86–2.74)
Total1.08** (1.03–1.14)1.01.16*** (1.14–1.18)1.45*** (1.42–1.48)1.89*** (1.80–1.98)
Duration of stay
Antenatal1.09* (0.99–1.18)1.01.07*** (1.04–1.11)1.18*** (1.13–1.22)1.08 (0.94–1.24)
Postnatal1.01 (0.98–1.05)1.01.03*** (1.02–1.04)1.06*** (1.05–1.08)1.15*** (1.12–1.19)
Total1.04* (0.99–1.08)1.01.04*** (1.03–1.05)1.09*** (1.07–1.11)1.12*** (1.08–1.17)

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Our data demonstrate that both high and low maternal BMI are associated with increased risk of complications during pregnancy, increased numbers and duration of maternal admissions, and increased costs to the health service. The increase in inpatient costs remain statistically significant after adjusting for both sociodemographic factors and clinical complications associated with BMI.

Main findings

Our study demonstrates a strong association between low or high maternal BMI and adverse outcome. Although our study does not prove causality, it does lend support to a causal relationship between low or high BMI and risk of complications. Applying the Hill criteria for assessing causality,[32] we found that: first, consistent with other studies, there is a strong temporal gradient of increasing risk of adverse outcomes with maternal BMI;[2-5, 7-10, 33] second, the association between high BMI and adverse outcomes, such as gestational diabetes, is biologically plausible and coherent with the natural history of these diseases in non-pregnant populations; and third, weight loss either prior to pregnancy or between pregnancies reduces the risk of complications, including gestational diabetes and fetal macrosomia .[34, 35] Although the present study is not able to demonstrate a temporal relationship between BMI and disease, it does provide evidence supporting a causal relationship between high or low BMI and complications.

Although studies have estimated the costs of obesity in non-pregnant populations,[11, 36, 37] there is a paucity of information about pregnancy. Published studies are limited by small sample size,[38] undercoding of maternal obesity in the databases used for analysis,[39] are based on costs derived from different models of health care,[6, 38, 39] or have not adjusted costs for confounding factors.

Our study demonstrates that maternity costs are significantly higher in women with low or high BMI compared with women with a normal BMI. Whereas 50% of the increased maternity costs can be explained by an increased incidence of clinical complications, the observed relationship remains significant after adjustment for these complications. At a population level, even if clinical complications associated with maternal obesity were reduced, maternal obesity would therefore still have a significant impact on healthcare resources.

In Scotland and other developed countries, considerable monies are being directed towards local and national strategies aimed at reducing the prevalence of obesity.[12, 13, 40] Using the adjusted increased maternal admission costs associated with overweight and obesity (Table 5: model 2), and extrapolating this to the 807 776 live births in the UK in 2010–2011, it is possible to estimate the impact of a decrease in the proportion of overweight and obese pregnant women on annual admission costs for the UK. For example, with a 2.5-percentage point decrease in the proportion of both overweight and obese pregnant women in Scotland – and a corresponding 5-percentage point increase in the proportion of normal weight women – the anticipated cost savings to the UK for inpatient admissions alone would be £12 702 278. Further savings would be realised through reductions in neonatal admissions and potentially longer-term improvements in health. To maximise potential healthcare savings, a dual approach of targeting a reduction in the prevalence of maternal overweight and obesity and its associated complications is therefore required.

Strengths and weaknesses

A strength of the present study is that we used sensitivity analyses to further understand the relationship between maternal BMI, clinical complications, and hospital costs. The relationship between maternal BMI and clinical complications remained consistent across different models, thus strengthening the results of the primary clinical analyses. Furthermore, after adjustment for sociodemographic indicators and adverse clinical outcomes found to be associated with maternal BMI, the influence of BMI on costs also remained. Although the reason for this is not clear, it is likely that BMI independently affects the numbers of admissions and duration of stay associated with given complications. Alternatively, this finding may be driven by an association between BMI and other ‘minor’ conditions not included in the models.

Although our study may not be representative of other settings, we took care to ensure that our resource use and cost findings are generalisable to populations outside Scotland. First, we estimated admission costs using the English NHS reference cost for the HRGs assigned to each admission, as well as using Scottish ISD speciality costs. By using two different methods to cost admissions, this broadens the generalisability of our findings. Second, although there are always difficulties in generalising cost data because of differences in the organisation of health systems and price structures, the key issue here is not the precise costs, but the direction and magnitude of the effect in relation to BMI groupings, and third we have presented the impact of BMI on the number and duration of admissions in order to improve generalisability outside the UK.

The use of an unselected population-based data set does, however, have some weaknesses.

First, routinely collected data relies on reliable coding of variables to minimize sources of bias. In the current study, 32% of singleton deliveries during the study period were eligible for inclusion into the study: the remaining 68% were ineligible, mainly because of maternal height and weight not being recorded, being recorded after 16 weeks of gestation, or being invalid. Although only 32% of deliveries were included in the final data set, the proportions of women in different BMI categories was similar to those reported in another contemporaneous Scottish maternity cohort.[41] This increases our confidence that our results are generalisable to the Scottish maternity population as a whole.

Second, we used maternal BMI as a measure of pathological obesity. Recent studies suggest that abdominal obesity may be a more important risk factor for vascular and metabolic disease than general adiposity.[42] Alternative anthropometric measures of abdominal obesity, such as waist circumference, waist/hip ratio, bioimpedance, and skin-fold thickness have been proposed as alternatives to, or additions to, BMI in predicting disease in non-pregnant populations.[43] However, these measures are not validated, are impractical, and are unreliable in pregnancy, and are not recorded at a population level. Although BMI has its limitations, it is therefore the only practical option for large epidemiological population-based studies.

Third, ISD does not record outpatient visits and clinic attendances. Given that high or low BMIs are associated with an increased risk of complications that require increased outpatient monitoring, it is likely that our study underestimated the total maternity costs associated with high BMI.

Fourth, because of the lack of precise data for the duration of neonatal admissions, we were unable to estimate neonatal costs precisely, and finally, the study may not be representative of other settings, which have different participant demographics, ethnicity, and models of maternity care.

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

In conclusion, our study demonstrates that maternal obesity places a considerable clinical and economic burden on women, healthcare providers, and the NHS. Given that the level of obesity continues to rise in the maternity population, there is an urgent need for research to ascertain which factor(s) link maternal obesity with adverse pregnancy outcomes, so that appropriately targeted interventions can be developed. For example, if the link between maternal obesity and the increased risk of pre-eclampsia was better understood, and interventions could be targeted to reduce this risk, this could result in a reduction in the numbers of iatrogenic preterm deliveries in women who are obese. This would have considerable benefits for women and their offspring, and significant economic benefits for the NHS.

Disclosure of interests

The authors have nothing to declare.

Contribution to authorship

F.C.D. designed and secured funding for the study, assisted with data analysis, and drafted and revised the article. She is the guarantor for the work. P.N. and G.S. undertook the economic analyses, assisted with data cleansing, and drafted and revised the article. A.E.R. and A.J.L. undertook the clinical analyses, assisted with data cleansing, and drafted and revised the article. A.D. and C.M. extracted the initial data set, assisted with data cleansing, and commented on the revised article. J.E.N., S.B., and T.M. assisted with study interpretation and commented on the revised article.

Details of ethics approval

The Privacy and Advisory Committee of the ISD approved the study. Caldicott approval was also obtained to enable data transfer between Edinburgh and Aberdeen for analysis.

Funding

This study was funded by a Chief Scientist Office Grant (CZG/2/471).

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

This study was funded by a Chief Scientist Office Grant (CZG/2/471).

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information
Commentary on ‘Associations between maternal body mass index during pregnancy, short-term morbidity, and increased health service costs: a population-based study’

Given the worldwide obesity epidemic, 30–40% of women currently giving birth in industrialised nations have a body mass index (BMI, kg/m2) that categorises them as being overweight (BMI 25.0–29.9) or obese (BMI ≥30) by World Health Organization (WHO) definitions. Obesity is also increasing in the developing world, although low weight remains a significant problem in several economically deprived populations in these countries. Being underweight has long been recognised as a risk factor for adverse pregnancy outcomes, including: intrauterine growth restriction, low birthweight, and premature labour. Only recently have the adverse effects of obesity on maternal, fetal, neonatal, and transgenerational health been recognised. These include: increased hypertensive disorders, gestational diabetes, failed induction, delivery by caesarean section, postpartum haemorrhage, infection, thromboembolism, maternal death, miscarriage, congenital abnormalities, premature delivery, macrosomia, birth injury, stillbirth, neonatal intensive care unit (NICU) admission, and neonatal, childhood, and adolescent obesity (Vasudevan et al., Arch Dis Child Fetal Neonatal Ed 2011;96:F378–F382). These effects increase with increasing BMI, are amplified by pre-existing maternal disease and excessive pregnancy weight gain, and differ in various racial and ethnic groups. Studies addressing this issue have met with mixed results, with some limitation of maternal weight gain and inconsistent reductions in gestational hypertension, gestational diabetes, and caesarean delivery, but little effect on neonatal outcomes (Oteng-Ntim et al., BMC Med 2012;10:47). In this edition, Denison and associates present data regarding adverse maternal and neonatal outcomes associated with mothers being underweight, overweight, obese, and severely obese, in comparison with being normal weight by the WHO BMI classifications. The authors used linear regression to account for maternal demographic variables, but they did not account for the effects of pre-existing diseases. They included maternal hospital costs, but not neonatal or other maternal costs. The data regarding maternal complications among women who were underweight, overweight, obese, and severely obese, and their infants, are not new, and confirm the work of multiple previous investigators; however, the data regarding costs are both novel and comprehensive. Only limited research has addressed the economic costs associated with obesity in pregnancy (Rowlands et al., Semin Fetal Neonatal Med 2010;15:94–9), and only one recent publication has reported such detailed obstetrical cost data associated with abnormal BMI classifications in a large geopolitical entity (Watson et al., Aust NZ J Obstet Gynaecol 2013;53:243–49). Although the absolute numbers varied, very similar trends were detected in both studies.

The article is unlikely to change the immediate clinical care provided by practicing obstetricians; however, the information may ultimately exert profound effects on clinical practice, as it may provide governmental entities with the rational for funding urgently needed clinical efforts to develop programmes to combat and treat the epidemic of obesity among women of reproductive age and the associated obstetrical and neonatal complications. These may include legislative initiatives to redirect food priorities and establish public educational and promotional programmes.

  • JP Lavin, Jr

  • Department of Obstetrics and Gynecology, Akron General Medical Center, Akron, OH, USA

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information
FilenameFormatSizeDescription
bjo12443-sup-0001-TableS1.pdfapplication/PDF81KTable S1. Unadjusted ORs for maternal, delivery, and neonatal complications in the study population.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.