Objective
We sought to assess congruity between data abstracted from medical records with answers to self-administered questionnaires.
Study Design
This was a multicenter prospective nulliparous pregnancy cohort.
Results
A total of 1507 women enrolled. Analyses were reported for 1296 with medical record data and 3-month postpartum follow-up. There was near-perfect agreement (κ ≥0.80) between maternal report and abstracted data for reproductive history, induction/augmentation method, epidural/spinal analgesia, method of birth, perineal repair, infant birthweight, and gestation. Agreement was poor to moderate for maternal position in second stage and duration of pushing.
Conclusion
Maternal report of pregnancy, labor, and birth factors was very reliable and considered more accurate in relation to maternal position in labor and birth, smoking, prior terminations, and miscarriages. Use of routine birthing outcome summaries may introduce measurement error as hospitals differ in their definitions and reporting practices. Using primary data sources (eg, partograms) with clearly defined prespecified criteria will provide the most accurate obstetric exposure and outcome data.
Studies assessing the contribution of obstetric risk factors to postpartum maternal morbidities such as urinary incontinence, fecal incontinence, and persisting perineal pain commonly draw upon data obtained from self-administered questionnaires, birth outcome summaries, and/or data abstracted from medical records. Each of these sources of information is susceptible to measurement error limiting assessment of the possible causal pathways involving obstetric risk factors.
Maternal accounts of labor and birth events may be subject to recall bias due to poor recall or lack of knowledge of particular exposures. Medical records are often assumed to provide more accurate data as information is recorded prospectively ; however, several studies have highlighted potential sources of error when abstracting data from medical records. These include ambiguous or inconsistent recording of information, missing information, and information that is missed by data abstractors. Birth outcome summaries generated by health information staff are also prone to each of these types of errors, and may introduce further errors due to inaccurate coding. In addition, the definitions used to generate birth outcome summaries may not be sufficiently precise for research purposes, and/or may not match the precise definition of exposures of interest in a particular study.
Several studies have assessed congruity between data abstracted from medical records and women’s accounts of obstetric events. Martin assessed the congruity of 223 mothers’ reports of their childbearing experiences with their obstetric records and found good agreement with regard to method of birth, epidural analgesia, episiotomy, and infant birthweight, and poor concordance in relation to the number and timing of antenatal appointments, tests during pregnancy, acceleration of labor, duration of first- and second-stage labor, and previous terminations of pregnancy. Five main reasons for discrepancies were identified including mothers’ limited knowledge or understanding of procedures, problems of interpretation and definition, and underreporting of sensitive information (eg, terminations). This study also identified occasional inaccuracies or missing information in medical records, and instances where definitions applied in medical records did not match the definition of data items sought by the researchers.
Hewson and Bennett compared medical records and Australian women’s self-report of labor and birth events in a study involving 337 “low-risk” primiparous women. The authors found that both data sources were subject to variation from the actual events they were meant to represent. In all, 3% of abstracted data was found to be inaccurate. Substantial to excellent agreement was found once the medical records data were corrected for abstraction error. A more recent study by Elkadry et al compared data abstracted from medical records with interview data conducted between 0-21 years postpartum (median of 10 weeks after birth). They report good agreement for method of birth and episiotomy, moderate agreement for induction and augmentation, and poor agreement for anal sphincter injury. Mothers who were older, had more recent births, and had more years of education had more accurate recall of labor and birth events.
The Maternal Health Study is an Australian multicenter prospective nulliparous pregnancy cohort study that was designed to overcome some of these limitations in previous studies investigating the contribution of obstetric risk factors to postpartum pelvic floor disorders. The primary aims of the study were to: (1) measure the incidence and natural history (onset, severity, and duration) of maternal health problems, in particular urinary incontinence, fecal incontinence, perineal pain, and sexual health issues among women having their first child; and (2) explore the contribution of obstetric risk factors–in particular prolonged labor, method of birth, and degree of perineal trauma–to postpartum health problems. The purpose of this study was to assess congruity between data abstracted from hospital medical records for women participating in the Maternal Health Study with study participants’ own accounts of pregnancy and birth events collected via self-administered questionnaires.
Materials and Methods
Sample
Women who had registered to give birth at any of 6 metropolitan public hospitals with a mix of high- and low-risk perinatal services were recruited to the study from April 1, 2003, through Dec. 31, 2005. To be eligible for the study women needed to: be nulliparous (ie, have no prior live births or pregnancies ending in a stillbirth), have an estimated gestation of ≤24 weeks at enrollment (according to the date of last menstrual period or ultrasound estimated due date), be aged ≥18 years, and have sufficient English language proficiency to complete written questionnaires and telephone interviews. Follow-up in the first phase of the study was scheduled at 30-32 weeks’ gestation, and at 3, 6, 9, 12, and 18 months’ postpartum using a combination of self-administered questionnaires and computer-assisted telephone interviews, and was extended to include follow-up after second and subsequent births and when the index (first) child is aged 4.5 years.
Recruitment was by mailed invitation facilitated by participating hospitals. In 2 larger tertiary hospitals, research staff regularly attended antenatal registering clinics to give out information about the study. At another hospital research staff attended “early bird” childbirth education classes to present information about the study and invite women to take part. The invitation package included a copy of the baseline questionnaire, and was followed up by a single mailed reminder postcard. Women were invited to return the questionnaire, signed consent papers, and contact information in a reply-paid envelope addressed to the Maternal Health Study research team. Given the frequent follow-up and long-term commitment asked of women joining the study, coupled with the personal and potentially intrusive nature of aspects of the study dealing with bowel symptoms and sexual health issues, we anticipated an overall response fraction of between 20-40%. As the primary aims of the study were to assess: (1) changes in women’s health over the course of pregnancy and postpartum, and (2) risk factors for incident cases of urinary incontinence and fecal incontinence, a response fraction in this range was considered acceptable provided that high rates of retention were achieved in subsequent follow-up (≥80%).
Data on obstetric exposures
Maternal report of pregnancy and birth events (obstetric exposures) was collected via questionnaires completed in early pregnancy (≤24 weeks’ gestation) and at 3 months’ postpartum. Pilot testing of questions and questionnaire formats was carried out to ensure the baseline and 3-month follow-up questionnaires were straightforward for women to complete and to check that language and terms used to describe pregnancy complications and obstetric events were understood by women from diverse sociodemographic backgrounds. Most questions were precoded, and study participants were asked to tick the box matching the correct answer and/or to provide additional information. Data collected in the baseline questionnaire included: previous pregnancies (miscarriages and/or terminations); treatment for infertility; medical conditions such as diabetes, asthma, and hypertension; and smoking in pregnancy. The questionnaire at 3 months’ postpartum covered labor and birth events, including onset of labor, use of analgesia/anesthesia, method of birth, length of first- and second-stage labor, duration of pushing, maternal position during second-stage labor and the birth, and genital tract trauma. The questionnaire completed at 3 months’ postpartum also included questions about infant characteristics, including infant birthweight and gestation. Study questionnaires are available via the study website.
Data on prepregnancy medical conditions, pregnancy complications, and labor and birth events were abstracted from the hospital medical records using a detailed data abstraction protocol for participants who had provided written consent. The protocol provided precise definitions for all data items, and all staff abstracting data participated in training and completed pilot data abstraction forms before commencing formal data abstraction for the study. Data were obtained from primary sources (eg, progress notes, partogram) wherever possible. Secondary sources such as birth outcome summaries were only used when information was not available in progress notes, and/or as a supplementary source to assist with clarifying inconsistencies in the medical record. The protocol specified that 10% of records at each participating hospital would be randomly selected for independent review and cross-checking by 2 members of the research team. The purpose of this procedure was to identify reasons for discrepancies and identify any systematic sources of errors. Reasons for discrepancies were discussed at a weekly team meeting chaired by the principal investigator (S.J.B.) and additional clarification of definitions for individual data items was added to the data abstraction protocol if deemed necessary. In addition, a research midwife (M.F.) checked all data abstraction forms for consistency of coding and compliance with the data abstraction protocol. A section was provided in the data abstraction form for coders to record if information contained in the medical record was illegible or inconsistent with information elsewhere in the record.
Analysis
Analyses were restricted to women for whom data were abstracted from medical records and who had completed the baseline questionnaire in early pregnancy and the 3-month postpartum follow-up. All statistical analyses were performed using statistical software (Stata, version 11; Stata Corp, College Station, TX). The extent of agreement between data abstracted from medical records and women’s own accounts of pregnancy and birth events was assessed using a κ coefficient (expressed as percentage agreement corrected for chance agreement) with 95% confidence intervals (CIs). For ordered categorical variables, a weighted κ and correlation coefficient with 95% CI were calculated. The κ values of ≥0.80 were considered to indicate near-perfect agreement, 0.60–0.79 substantial agreement, 0.40–0.59 moderate agreement, and 0–0.39 poor agreement. For continuous variables such as birthweight, correlation between women’s reports and medical records was calculated with 95% CI. Stratified analyses were conducted to assess the extent of agreement across different categories of maternal age, income, education, and country of birth.
The study was approved by the following human research ethics committees: La Trobe University (2002/38); Royal Women’s Hospital, Melbourne (2002/23); Southern Health, Melbourne (2002-099B); Angliss Hospital, Melbourne; and Royal Children’s Hospital, Melbourne (27056A).
Results
Sample
Of 1537 women who enrolled in the study, 30 were excluded after enrollment (12 women had a miscarriage, 11 were not sufficiently fluent in English to participate, 5 were not nulliparous, and 2 terminated the pregnancy due to fetal abnormalities). The final sample comprised 1507 women. Exact response figures were difficult to obtain as hospitals identified potential participants and mailed the invitation package to women soon after registering. The researchers relied on hospital report of number of questionnaires mailed with no ability to assess wrongly addressed mail, duplicate mailings, and returns to sender or noneligible women. Over 6000 invitation packages were distributed and 1 reminder postcard was sent 2 weeks after the original invitation. Australian privacy legislation and the conditions of our ethics approval prevented us from following up nonresponders by telephone. Assuming that 80-90% of invitations were correctly addressed and sent to eligible women, we conservatively estimate that the final response fraction was in the range of 28–31%, but it may have been higher. Mean gestation at enrollment was 15.0 weeks (SD = 3.1; range, 6–24 weeks). The Maternal Health Study sample includes fewer young women and women of non-English-speaking background compared with confinements at the study hospitals during the study period. Further details regarding the study sample are available elsewhere.
The Figure shows a flowchart of participation and data abstraction. Of the 1507 participants, 95% (n = 1430) gave separate written consent for the research team to access their hospital medical record. Data were abstracted for 1328 of the 1351 women who gave birth at 1 of the 6 participating study hospitals, 88% of the total sample. All medical record data collection forms were reviewed by a research midwife (M.F.) for consistency of coding and compliance with the data abstraction protocol and 98 randomly selected medical records were double reviewed, ie, coded by 2 independent abstractors (7.4%).
The sample for this study comprised 1296 women with medical record data and 3-month postpartum questionnaire data ( Figure ). The social characteristics of these 1296 women and the 135 participants who completed the 3-month postpartum follow-up but for whom medical record data were not abstracted are shown in Table 1 . Women who were overseas born of a non-English-speaking background had higher odds of being in the group that did not have medical record data abstracted compared with women who were Australian born (odds ratio, 2.1; 95% CI, 1.5–3.0), as did women who were not working in early pregnancy (home duties, unemployed, or unable to work) and who earned <AU$20,001 in the last financial year ( Table 1 ).
| Characteristic | Medical records data abstracted n = 1296 | Medical records data not abstracted n = 135 | Crude OR a | |||
|---|---|---|---|---|---|---|
| n | % | n | % | OR | 95% CI | |
| Maternal age at birth, y | ||||||
| 18–24 | 155 | 12.0 | 21 | 15.6 | 1.63 | 0.94–2.83 |
| 25–29 | 357 | 27.5 | 43 | 31.9 | 1.45 | 0.93–2.26 |
| 30–34 | 517 | 39.9 | 43 | 31.9 | 1.0 (reference category) | |
| 35–39 | 224 | 17.3 | 21 | 15.6 | 1.13 | 0.65–1.94 |
| ≥40 | 43 | 3.3 | 7 | 5.2 | 1.96 | 0.83–4.61 |
| Relationship status | ||||||
| Married | 792 | 61.1 | 122 | 57.8 | 1.0 (reference category) | |
| Living with partner | 446 | 34.4 | 76 | 36.0 | 1.11 | 0.81–1.51 |
| Divorced/separated/single | 58 | 4.5 | 13 | 6.2 | 1.46 | 0.77–2.73 |
| Country of birth | ||||||
| Australia | 983 | 76.1 | 132 | 63.5 | 1.0 (reference category) | |
| Overseas, English speaking | 119 | 9.2 | 22 | 10.6 | 1.38 | 0.84–2.25 |
| Overseas, non-English speaking | 189 | 14.6 | 54 | 26.0 | 2.13 b | 1.50–3.03 |
| Highest education level | ||||||
| University degree | 603 | 46.7 | 86 | 41.1 | 1.0 (reference category) | |
| Certificate/diploma | 338 | 26.2 | 54 | 25.8 | 1.12 | 0.78–1.61 |
| Year 12 | 234 | 18.1 | 43 | 20.6 | 1.29 | 0.87–1.91 |
| <Year 12 | 115 | 8.9 | 26 | 12.4 | 1.59 | 0.98–2.57 |
| Employment status in early pregnancy | ||||||
| Paid work | 1071 | 84.0 | 150 | 72.5 | 1.0 (reference category) | |
| Home duties | 47 | 3.7 | 15 | 7.2 | 2.28 c | 1.24–4.18 |
| Study | 38 | 3.0 | 8 | 3.9 | 1.50 | 0.69–3.28 |
| Unemployed | 45 | 3.5 | 14 | 6.8 | 2.22 d | 1.19–4.14 |
| Not able to work | 60 | 4.7 | 16 | 7.7 | 1.90 d | 1.07–3.39 |
| Other | 14 | 1.1 | 4 | 1.9 | 2.04 | 0.66–6.28 |
| Income during last financial year ($AU) | ||||||
| ≤20,000 | 195 | 16.3 | 32 | 17.8 | 0.99 | 0.63–1.54 |
| 20,001-40,000 | 445 | 37.2 | 74 | 41.1 | 0.94 | 0.65–1.36 |
| 40,001-60,000 | 376 | 31.4 | 59 | 32.8 | 1.0 (reference category) | |
| >60,000 | 181 | 15.1 | 15 | 8.3 | 0.50 d | 0.28–0.89 |
| Total | 1296 | 100.0 | 211 | 100.0 | ||
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
