Agreement between hospital records and maternal recall of mode of delivery: Evidence from 12 391 deliveries in the UK Millennium Cohort Study
M Quigley, National Perinatal Epidemiology Unit, University of Oxford, Old Road Campus, Headington, Oxford OX3 7LF, UK. Email firstname.lastname@example.org
Objective The objective of this study was to measure the agreement between hospital records and maternal reporting of mode of delivery in a representative UK sample.
Design Population-based survey (Millennium Cohort Study).
Population A total of 12 391 singleton infants born in 2000–2002.
Methods Mothers were interviewed when infants were approximately 9 months old. Information was collected by interview on many obstetric and perinatal factors including mode of delivery. Record linkage to the mother’s delivery hospital records was undertaken in those who gave consent (90%). A matching record was found for 83%. Maternal report and hospital records were compared using mode of delivery classified into three (normal, assisted and caesarean) and six groups. Factors associated with disagreement between the two data sources were identified.
Main outcome measure Proportion of records in which there was agreement between the two data sources.
Results Agreement between maternal report and hospital records was at least 94% using six mode of delivery groups and 98% using three groups. Much of the disagreement (57–63%, depending on country) was between forceps and ventouse, and between planned and emergency caesarean. Disagreement was more common in women whose babies were first born and in women not born in the UK.
Conclusion Our study confirms that maternal reporting of mode of delivery is highly reliable. This is important for clinical staff caring for women and those conducting epidemiological studies. Additional data sources may be necessary to gather reliable data from ethnic minority women, particularly those born outside the UK, or to distinguish forceps from ventouse, or planned from emergency caesarean section.
Maternal recall of perinatal events is often relied upon when a clinician elicits a patient history, or when investigators conduct research. It is assumed that maternal reporting of mode of delivery is accurate, for example, a woman is unlikely to forget having had a caesarean section. In the few studies that have assessed the reliability of maternal reporting of mode of delivery, it has been accurately reported.1–5 However, these studies are based on broad classifications of mode of delivery such as caesarean section versus ‘not caesarean section’; two studies also classified mode of delivery as forceps versus ‘not forceps’ and another distinguished between operative and nonoperative vaginal deliveries. We identified no studies, which assessed whether women accurately distinguish between elective and emergency caesarean, or between forceps and vacuum deliveries, and included the whole range of options for mode of delivery.
The present study aimed to measure agreement between hospital records and maternal reporting of mode of delivery in a representative sample of 12 391 singleton infants born in the UK in 2000–2002. Mode of delivery was classified into three broad and six specific categories. Perinatal and socio-demographic factors associated with disagreement between the two data sources were identified.
Millennium cohort study
The Millennium Cohort Study (MCS) is a nationally representative UK longitudinal study of 18 819 infants born in the UK.6 A random two-stage sample of all infants born in England and Wales between September 2000 and August 2001, and in Scotland and Northern Ireland between November 2000 and January 2002, who were alive and living in the UK at 9 months of age was drawn from child benefit registers. Stratified sampling at electoral ward level, with oversampling of ethnic minority and disadvantages areas, ensured adequate representation of such areas. This was the first of the large UK cohorts to include non-English speaking families. The interview response rate was 85%.6
Parents were interviewed for the first time (sweep 1) when most infants were aged 9 months, and detailed information was collected on a range of socio-demographic and health factors. The MCS does not cover births where the infant died within the first 9–10 months of age, but these constituted only about 0.6% of all births. In those women who were natural mothers (99.7% of the MCS mothers), we attempted to obtain data from their delivery hospital records using record linkage methods. Ninety percent of the MCS mothers consented to this linkage.7
Record linkage to hospital records
In all four countries in the UK, hospitals submit their patient records to the relevant country’s health services department. Hospitals have high submission rates for patient records although some fields in the records are not well completed by some hospitals.8 Moreover, details of deliveries in women who gave birth at home or in private hospitals will have been less likely to be submitted to the central government department. These constituted an estimated 2 and 0.5% of deliveries in England, respectively.8 Those giving birth at home will, of necessity, be spontaneous vaginal delivery.
Full details of the record linkage process and results have been described elsewhere (C. Hockley et al., unpubl. obs.). In brief, the variables used for record linkage varied by country: all countries used mother’s date of birth, postcode at birth and hospital at birth; England, Wales and Scotland also used additional variables such as baby’s date of birth, baby’s sex, baby’s birthweight, mother’s name and baby’s name. In England, Wales and Northern Ireland, record linkage was performed using a deterministic method of matching, whereby a match was only accepted if there was complete agreement on a given combination of common variables. Matching in England used differing combinations of the whole and partial variables provided. This catered for missing and incomplete data yet still allowed for as good a match as possible. Scotland adopted a probabilistic method of matching entailing comparisons with all possible pairs of records and making a decision as to whether they belonged to the same individual. Every time an item of identifying information was the same on the two records, the probability that they applied to the same person increased.
A matching record was found in 83% of those who gave consent, and this varied by country. Among singletons, a matching record was found for 74% of records in England, 85% in Scotland, 86% in Wales and 58% in Northern Ireland (C. Hockley et al., unpubl. obs.).
Classification of mode of delivery
Mode of delivery was ascertained in the MCS interview using the question ‘What type of delivery did you have? Was it …’, at which point the categories listed in Table 1 were read out by the interviewer. This question was one of many in the section on pregnancy, labour and delivery. In the hospital records, mode of delivery was recorded in the maternity tail part of the hospital record (as a single code) and in the general record (under the operation codes listed; these are coded using the Office of Population Censuses and Surveys tabular list of surgical operations and procedures, 4th revision). In each country, we ascertained mode of delivery using the most completed of the two parts of the hospital records; this was the maternity tail in Scotland and the general record in the other three countries.
Table 1. Classification of mode of delivery
|Normal||A normal delivery||Spontaneous vertex, spontaneous other cephalic||R201, R213, R232, R238, R239, R249|
|Forceps||Assisted with forceps||Low forceps, not breech; other forceps, not breech||R211, R212, R213, R214, R215, R218, R219|
|Ventouse||Assisted vacuum extraction||Ventouse, vacuum extraction||R191, R198, R199, R221, R222, R223, R228, R229|
|Assisted breech||Assisted breech||Assisted breech||R202|
|Elective caesarean||A planned caesarean||Elective caesarean||R171, R172, R178, R179|
|Emergency caesarean||An emergency caesarean||Emergency caesarean||R181, R182, R188, R189|
Mode of delivery was analysed using six groups (normal, forceps, ventouse, assisted breech, elective caesarean and emergency caesarean) or three groups (normal, assisted [forceps/ventouse/assisted breech] and caesarean section [elective/emergency]). For consistency with other studies, we also analysed the mode of delivery using the two groups: caesarean and not caesarean. The small number of women where mode of delivery was classified as ‘other’ (i.e. categories or codes not listed in Table 1) or not known were excluded from the analysis (Table 2).
Table 2. Agreement between MCS and hospital records for mode of delivery by country
|Total after excluding 24 hospitals**||7337||2326||1978||1103|
|Classifiable, n (%)***||7140 (97.3)||2253 (96.9)||1973 (99.7)||1025 (92.9)|
|Agreement, n (%)****||6766 (94.8)||2153 (95.6)||1901 (96.4)||968 (94.4)|
|Agreement, n (%)*****||7004 (98.1)||2215 (98.3)||1942 (98.4)||1005 (98.0)|
|Agreement, n (%)******||7113 (99.6)||2245 (99.6)||1968 (99.7)||1024 (99.9)|
Methods for assessing agreement between the MCS and the hospital records
The analysis of agreement is restricted to MCS natural mothers who gave consent and had singleton deliveries, and whose mode of delivery was not classified as other or not known. Agreement was assessed by estimating the proportion of women in whom the mode of delivery in the MCS and hospital records agreed. This was performed separately using six, three and two groups for mode of delivery. Where there was less than perfect agreement, we explored whether the records being compared may not be true matches. Here, we estimated the probability that each match could have occurred by chance. These probabilities varied between 0.0001 and 0.0201 according to the combination of matching variables, so it was assumed that the vast majority of matches were true.
Then, we identified hospital records where mode of delivery appeared to be incorrect. For example, some hospitals are known to be poor at completing and submitting their records, particularly the maternity tail.8 We excluded from the analysis those hospitals, which either had mode of delivery coded as other on at least 30% of records, or mode of delivery coded as other or ‘missing’ on at least 50% of records. This resulted in 24 hospitals (all from England) being excluded from the analysis (n = 1121 women). Finally, we used logistic regression to explore whether any maternal characteristics were associated with disagreement. The factors explored were age at delivery, marital status, occupation, education, ethnic group, country of birth, language spoken at home, first-born baby and ward type. All analyses allowed for the clustered, stratified sample using the ‘survey commands’ in Stata version 9 (Stata Corporation, College Station, Texas, USA).
A matching record was found for 13 865 singleton deliveries, and mode of delivery was classifiable in 12 391 deliveries. Agreement between MCS and hospital records was high in all four countries: at least 94% using six groups for mode of delivery; at least 98% using three groups and at least 99% using two groups (Table 2). Among the records where there was any disagreement, the majority of discrepancies were between forceps and ventouse and between planned and emergency caesarean section for the women in England (236/374 = 63.1%), Wales (62/100 = 62%), Scotland (41/72 = 56.9%) and Northern Ireland (36/57 = 63.2%).
Table 3 illustrates the disagreement in England only: compared with the hospital records, the MCS reported slightly more normal deliveries (69.9 versus 69.5%) and more emergency caesarean sections (12.1 versus 11.9%), and fewer assisted deliveries (Table 3). Of the 1466 women in England who reported having had a caesarean section, there were only 19 for whom a caesarean was not reported in the hospital records; and of the 5659 women who reported not having had a caesarean, there were eight for whom a caesarean was reported in the hospital records. However, the rate of disagreement was much higher for the indication for caesarean section: of the 1447 women in England in whom caesarean section was reported by the woman and the records, there were 155 discrepancies (10.7%; in italics in Table 3) between planned and emergency caesarean.
Table 3. Pattern of disagreement between MCS and hospital records in England
|Assisted breech||4||0||2||9||0||0||15 (0.2)|
|Planned caesarean section||3||2||3||0||522||73||603 (8.4)|
|Emergency caesarean section||9||1||1||0||82||770||863 (12.1)|
|Total, n (%)||4962 (69.5)||218 (3.0)||496 (6.9)||9 (0.1)||608 (8.5)||847 (11.9)||7140|
When we explored what factors were associated with disagreement (using three groups), there was no statistically significant association between disagreement and age at delivery, marital status, occupation, education and ward type. However, in all four countries, disagreement was significantly higher in deliveries where the baby was first born (Table 4). For example, in all four countries, the percentage disagreement was about 2.5% in deliveries where the baby was first born compared with about 1% in other deliveries. In addition, disagreement was higher in women who reported that English was not the only language spoken at home (except in Wales, where Welsh was usually the only other language spoken at home), in certain ethnic groups (in England and Scotland only) and where the mother was not born in the UK (in England and Scotland only, where the percentage disagreement was 1.4% in those born in the UK, 3.5% in woman in England who were born outside the UK and 5.1% in women in Scotland who were born outside the UK).
Table 4. Factors associated with disagreement between MCS and hospital records by country
| No||0.9 (3645)||1.2 (1275)||1.0 (1065)||1.3 (609)|
| Yes||2.3 (2814)||2.5 (940)||2.1 (852)||2.8 (399)|
| P value||<0.001||0.050||0.031||0.056|
|Languages spoken at home|
| English only||1.4 (5926)||1.7 (1994)||1.4 (1902)||1.8 (994)|
| English and other||3.3 (439)||1.8 (169)||4.5 (28)||6.0 (13)|
| Other only||4.4 (133)||3.2 (55)||8.4 (9)||20.0 (4)|
| P value||<0.001||0.74||0.084||0.002|
|Mother’s ethnic group|
| White||1.3 (5824)||1.7 (2178)||1.4 (1897)||1.9 (1007)|
| Mixed||0.0 (69)||0.0 (7)||0.0 (3)||—|
| Indian||5.1 (107)||0.0 (3)||15.5 (8)||—|
| Pakistani||3.1 (168)||0.0 (5)||7.2 (10)||—|
| Bangladeshi||3.3 (59)||0.0 (6)||− (0)||—|
| Black Caribbean||2.1 (69)||0.0 (2)||0.0 (7)||—|
| Black African||2.3 (84)||0.0 (3)||0.0 (1)||—|
| Other||7.6 (104)||5.8 (11)||0.0 (12)||0.0 (4)|
| P value||<0.001||0.95||0.049||0.78|
|Mother’s country of birth|
| UK||1.4 (5852)||1.7 (2091)||1.4 (1866)||Not available|
| Not UK||3.5 (645)||1.6 (126)||5.1 (72)|
When these variables were fitted in a multivariate logistic regression model (Table 5), these associations remained: disagreement was significantly more common when the baby was first born (odds ratio [OR] equals about 2 in all four countries); when the mother was not born in the UK (OR = 1.92 in England and 4.01 in Scotland); and when English was not the language spoken at home (OR = 2.49 in England).
Table 5. Odds ratios for association between disagreement and selected factor by country
| Yes||2.64 (1.80–3.86)||2.15 (0.98–4.70)||2.16 (1.06–4.40)||2.09 (0.96–4.52)|
|Languages spoken at home|
| English only||1||—||—||—|
| English and other||1.46 (0.94–2.25)|
| Other only||2.49 (1.15–5.40)|
|Mother’s country of birth|
| Not UK||1.92 (1.16–3.20)||4.01 (1.31–12.34)|
In all four countries, agreement between MCS and hospital records was at least 94% using six groups for mode of delivery. Much of the disagreement was between forceps and ventouse, and between planned and emergency caesarean. When agreement was assessed between the three groups of mode of delivery (normal, assisted, caesarean), the amount of disagreement reduced to 2%, and disagreement was more common in women whose babies were first born and in women not born in the UK.
Our findings are based on a large, representative study from the UK. Record linkage occurred in the vast majority of the MCS: those who gave consent to linkage (90% of MCS mothers); natural mothers (99.7% of MCS mothers) and those who delivered in NHS hospitals (estimated to be 97.5% of MCS mothers). Consent was lower in women from ethnic minorities and those who reported that English was not the only language spoken at home.7 Therefore, our study may have slightly underestimated the amount of disagreement between hospital records and maternal recall.
A potential limitation of our study is that we used the hospital records as the gold standard, but we cannot be sure that the hospital records are completely accurate. Hence, possible explanations for less than perfect agreement include matching errors, errors in the hospital records and errors in the MCS records such as those due to maternal inaccuracy. The probability of false matches was explored, and these suggest that most matches are likely to be true, although we cannot rule out some false matches (C. Hockley et al., unpubl. obs.). Some probable errors in the hospital records were expected, at least in the data from England;8 hospitals known to be poor at completing the hospital maternity tails were identified and excluded from the analysis, and this did not change the results. As with all routinely collected datasets, errors associated with data collection, coding and entry are likely to have occurred.
The presence of maternal recall or reporting bias in the MCS records was explored by identifying whether disagreement between records was associated with perinatal and socio-demographic characteristics. The fact that disagreement was strongly associated with two maternal characteristics, country of birth and first-born baby, suggests that maternal misreporting is the most likely cause of disagreement in most cases. Ethnicity and parity were associated with poor recall of labour and delivery events in a North American study.5 Women experiencing their first delivery may be less familiar with perinatal events and terminology than women who have given birth previously. Similarly, women from ethnic groups (Indian, Pakistani and Bangladeshi), particularly those born outside the UK or who do not speak English at home, have been shown to be more likely to book late for antenatal care;9 this may result in less prenatal information, which in turn may affect their understanding or recall of perinatal events.
Our finding of more than 99% agreement for caesarean versus ‘not caesarean’ is consistent with studies conducted in the 1980s and 1990s in Australia and North America, which observed 98.3% agreement in 356 women3 and 100% agreement in 397 women,1 69 women,2 and 754 women.4 However, these studies did not assess maternal reporting of more specific groupings of mode of delivery. Among the women who reported having had a caesarean section in our study, there were discrepancies between planned and emergency caesarean for 10.7% of women. This is consistent with a recent North American study in which 11% (10/88) of women who were delivered abdominally did not know the indication for their caesarean section.5 Clearly, the definition of emergency caesarean is open to misunderstanding, particularly when it often refers to an ‘in-labour caesarean’ that happens because of failure to progress rather than a more life-threatening event such as severe haemorrhage.
Our large study, which is representative of contemporary UK, confirms that maternal reporting of mode of delivery is highly reliable. This is an important finding in the context of clinical staff caring for women and those conducting epidemiological studies. However, in settings that require minimal misclassification between forceps and ventouse or between elective and emergency caesarean section, additional data sources may be needed to ascertain mode of delivery more precisely. Similarly, in settings that are focussing on ethnic minority groups where many women were born outside the UK, additional data sources may also be required. The National Institute for Clinical Excellence guideline on postnatal care recommends that women should be offered an opportunity to talk about their birth experiences and ask questions about the care they received during labour.10 Implementation of this guideline would result in a better understanding and reporting of mode of delivery and other perinatal events.
This work was undertaken by the National Perinatal Epidemiology Unit, which receives funding from the Department of Health. The views expressed in this publication are those of the authors and not necessarily those of the Department of Health.
Conflict of interest