The reliability and validity of self-reported reproductive history and obstetric morbidity amongst Birth to Ten mothers in Soweto

Objective: To assess whether self-reports of reproductive history and obstetric morbidity provide an accurate basis for clinical decision-making. Setting, participants and methods: Self-reports of mater­ nal age and reproductive history, together with clinical measurements of five medical disorders, were abstracted from the obstetric notes of 517 mothers whose children were enrolled in the Birth to Ten study. These data were compared to self-reported information collected by inter­ view during the Birth to Ten study. Findings: The reliability of self-reported age and gravidity was high (R=0.810-0.993), yet self-reports of previous mis­ carriages, terminations, prematureand stillbirths were only fairly reliable (Kappa=0.48-0.50). Self-reported dia­ betes and high blood pressure had specificities of more than 95% for glycosuria, hypertension and pre-eclamp­ sia. However, the specificity of self-reported oedema for hypertensive disorders and the specificity of self-reported urinary tract infection for STD seropositivity were only around 65%. Conclusions: The modest reliability and limited validity of self-reported obstetric morbidity undermines the clini­ cal utility of this information. Recommendations: These results strengthen the case for providing mothers with “Home-based Maternal Records” to facilitate access to accurate obstetric information dur­ ing subsequent clinical consultations. Opsomming D o e ls te llin g : Om vas te ste l o f se lf-aan g em eld e reproduktiewe geskiedenis en verloskundige morbiditeit ‘n akkurate basis vir kliniese besluitneming bied. M etode: M oeders se se lfv e rk la a rd e ouderdom en reproduktiewe geskiedenis, plus die kliniese meting van vy f s iek teaan d o en in g s is g e tran sk rib ee r v an af die verloskundige notas van 517 moeders wie se kinders in die Birth to Ten-studie geregistreer is. Hierdie gegewens is vergelyk met die self-aangemelde inligting wat deur middel van onderhoude verkry is in die Birth to Ten-studie. Resultate: Die betroubaarheid van die selfverklaarde ouderdom en aantal swangerskappe was hoog (R=0.8100.993), maar self-aanm eldings van vorige miskrame, b ee in d ig in g s van sw an g ersk ap p e , en v roeeen doodgeboortes was slegs redelik betroubaar (Kappa=0.480.50). Self-aangemelde diabetes en hoe bloeddruk het spesifisiteite van meer as 95% vir glucosuria, hipertensie en p re -ek lam p sia , m aar d ie sp e s if is ite it van selfaangemelde edeem vir hipertensie en die spesifisiteit vir se lf-aan g em eld e u rie n w e g in fe k s ie s v ir sek su ee loordraagbare seropositiwiteit was net om en by 65%. Gevolgtrekking: Die beskeie betroubaarheid en beperkte geldigheid van self-aangemelde verloskundige morbiditeit ondermyn die kliniese bruikbaarheid van hierdie inligting. Aanbevelings: Die resultate versterk die argument dat moeders “Tuisgebaseerde Moederrekords” by hulle moet hou sodat toegang to t m eer akkurate verloskundige inligting in kliniese opvolgbesoeke makliker beskikbaar is.


Introduction
Self-reported maternal information plays an im portant role in antenatal care by helping clinicians identify mothers who are at increased risk of poor obstetric outcomes (Carroll et al. 1988;Peoples-Sheps et al. 1991;Essén et al. 1994).Usually these self-reports are the only source o f inform ation on sociodemographic and behavioural risk factors, such as m a ternal age, social support and tobacco consumption (Harris et al. 1997).They may also be the only source of information on clinical risk factors, such as previous reproductive history and obstetric morbidity, when m edical records are unavailable (Essén et al. 1994).
In general, maternal reports o f sociodemographic and behav ioural risk factors seem to be very accurate (Tilley et al. 1985;Fox et al. 1989), although prestige bias can undermine the accuracy of sensitive information, such as smoking behav iour (Fox et al. 1989) and past terminations (Joffe and Grisso 1985).Nevertheless, such inaccuracies are less important than poor recall o f clinical risk factors, since these play a more important role in the selection o f appropriate antenatal and perinatal care (Hall et al. 1980;Essén et al. 1994).Indeed, the accuracy o f self-reported obstetric morbidity tends to be poor (Casey et al. 1992), and while maternal reports usually include a greater num ber of minor symptoms, such as bleed ing and vaginal discharge, than their medical records (Joffe and Grisson 1985), mothers often omit more serious condi tions and medical interventions (Tilley et al. 1985).
Inaccurate self-reports of obstetric morbidity severely under mine the provision o f appropriate m aternity care wherevér there are inadequate resources for maintaining access to ac curate medical records.Despite the introduction o f free pri mary health care for pregnant women (and to some extent, as a result of free care: M cCoy 1996; Hutchings et al. 2001) public antenatal services within many o f South A frica's ur ban and peri-urban townships remain fragmented and diffi cult to coordinate (Yach et al. 1991).The aim of the present study was therefore to assess the reliability and validity of self-reported reproductive history and obstetric m orbidity am ongst m others who delivered children at B aragw anath H o sp ital (su b seq u en tly ren am ed C hris H ani H ospital).Baragwanath Hospital provides specialist maternity services for Soweto and surrounding peri-urban areas, where self-reports of obstetric morbidity are likely to play an important role in the provision of antenatal care.We com pared selfreported obstetric morbidity, collected during interviews with the mothers of children enrolled in a longitudinal birth co hort study (Birth to Ten: Yach et al. 1991;Richter et al. 1995), with self-reported data and clinical m easurements recorded in their obstetric notes.

Methods
Birth to Ten is a longitudinal birth cohort study which set out to enrol all singleton births to mothers resident in the Soweto-Johannesburg m etropole during a 7-w eek period from 23rd April to 8 th June 1990 (Richter et al. 1995).A total of 5456 singleton births took place over this period, and 2 1 2 0 o f these occurred at Baragwanath Hospital.A subsequent search of the medical records department located obstetric notes for 517 Birth to Ten mothers and these comprise the sample included in the analyses that follow.Antenatal record charts contained within their obstetric notes provided self-reports of: maternal age, gravidity and previous obstetric problems including mis carriages, terminations, premature-and still-births.Clinical measurements of blood pressure, proteinuria and glycosuria taken during routine antenatal care were also recorded on the antenatal record charts, while seropositivity for sexually-transm itted disease (STD) was recorded on the neonatal record charts o f any child that had been admitted for postnatal care.These data provided objective measures of five clinical disor ders and the putative 'gold standards' against which the va lidity of self-reported obstetric morbidity could be assessed.All five of these disorders have been identified as important risk factors for poor obstetric outcome in Africa (Mati 1994): 1. Pre-existing (essential) hypertension was defined as a diastolic blood pressure of 90 mmHg or more recorded at book ing, during the first visit to the antenatal clinic (after Sweet 1992;Ventura et al. 1992).
2. Pregnancy-induced hypertension (PIH) was defined as a rise in diastolic blood pressure o f 20 mmHg or more from that recorded at booking, on at least two occasions during the course o f pregnancy (after Ventura et al. 1992).
3. Pre-eclampsia was diagnosed when PIH was accompanied by proteinuria (Davey and M acGillivray 1988).Proteinuria was defined as the presence of more than a trace of protein in maternal urine samples examined on two or more visits to the antenatal clinic.
4. Glycosuria was used as an indicator of gestational diabetes (Sweet 1992) and was defined as the presence o f more than a trace o f glucose in maternal urine samples examined on two or more visits to the antenatal clinic.
5. Seropositivity fo r sexually-transm itted infections (W R / VDRL) was determ ined from the cover of neonatal record charts for any mothers whose child had been adm itted for neonatal care.
Self-reports of maternal age, reproductive history and obstet ric morbidity were obtained at enrolm ent into the Birth to Ten study, using detailed interviews conducted during ante natal care or shortly after delivery.Among other questions, each mother was asked whether she had ever been pregnant before, and whether previous pregnancies had ended in m is carriages, terminations, premature-or still-births.Each mother was also asked whether she had experienced "swelling of your feet", "high blood pressure", "diabetes" or a "urinary infec tion" during her current (Birth to Ten) pregnancy, and whether she had ever been treated for "any sexually-transmitted dis ease" .Self-reports of obstetric morbidity related to each of the five disorders (1. to 5. above) recorded on antenatal charts were thereby defined as: Oedema ("swelling o f your feet" -1., 2. and 3. above); Hypertension ("high blood pressure" -1., 2. and 3. above); Pre-existing or gestational diabetes ("dia betes" -4.) above; and Urinary tract infection (both past: "any sexually transmitted disease" ; and present: urinary in fection" -5.above).
All results are presented as mean with standard errors (SEM) in parentheses.Standard Chi-squared (x2) and paired t-tests were used to assess the significance of the results (Sokal and Rohlf 1981).Ethical permission for the Birth to Ten study was obtained through a Human Subjects Clearance issued by the University o f the W itwatersrand.

Results
Only 313 (60.5%) of the 517 obstetric notes examined in the present study contained complete antenatal record cards.A c cording to sociodemographic information archived in the Birth to Ten database, these obstetric notes belonged to women who had attained a significantly higher mean educational stand ard (5.70(1.73)South African Standard Grade) than women w h o se a n te n a ta l re c o rd s w ere m issin g o r in c o m p le te (5.16( 1.67) SA Standard Grade, t= 2.423, p<0.05).However, there were no significant differences in maternal age, gravidity or access to medical insurance between the two groups.Us ing the clinical data abstracted from these 313 antenatal record cards it was possible to determine that 5.3% of the women exhibited pre-existing hypertension, and o f the 295 women who made more than one visit to the antenatal clinic, 2 . 1 % exhibited glycosuria, 12.7% developed PIH and 3.2% had proteinuria, while only 1 . 1 % displayed the symptoms of pre eclampsia (i.e. both PIH and proteinuria).
Fewer than 40% (200) of the 517 obstetric notes contained neonatal record cards, but there were no significant differ ences in the age, gravidity, educational standard attained or medical insurance status of mothers whose children had been admitted for neonatal care and those whose children had not.From the clinical data contained in these records it appeared that 7 (4.4%) of the 160 women whose STD status was known were seropositive for one or more STDs .
Due to the staggered nature o f enrolment into Birth to Ten (Richter et al. 1995), only half (255) o f the women examined in the present study had been interviewed about their obstet ric morbidity.Those who were interviewed had made signifi cantly more visits to their antenatal clinic (6.6(2.81))than those who were not interviewed (5.78(3.26);?=2.405,/j<0.05).However, there were no significant differences in the preva lence o f hypertension, pre-eclam psia, glycosuria or STD in fection between women who had been interviewed and those who had not.
To assess the reliability of self-reported reproductive histo ries, the self-reports o f maternal age, gravidity and prior re productive outcomes recorded in the antenatal record cards were compared to those collected during the Birth to Ten in terviews.For the two continuous variables (maternal age and gravidity) the unreliability (Sr2 ) o f duplicate self-reports was calculated using the technique o f M arks et al. (1988), where Sr2= ^jij2/ 2 n , the sum o f the squared differences between selfreported data collected on two different occasions.The reli ability coefficient, R , was then calculated as the proportion o f the total sample variance (s2) that remained after subtract ing unreliability (R= s2 -S 2 Is2 ).The unreliability o f selfreported maternal age was 1.16 years and the associated reli ability coefficient (R ) was 0.993.Likewise, the unreliability of self-reported gravidity was 0.25 pregnancies and the asso ciated reliability coefficient (R) was 0.810.For the categori cal variables (whether or not previous pregnancies ended in miscarriages, terminations, premature-or still-births) the re liability o f m aternal reports was calculated using C ohen's Kappa, which accounts for chance agreem ent between dupli cate self-reports (see Casey et al. 1992).Reliability was only fair (Casey et al. 1992) for duplicate reports o f previous pre mature births (K= 0.50), and previous pregnancies that ended in term inations, stillbirths and neonatal deaths (K= 0.48).Likewise, when asked to give the year in which their previ ous pregnancy took place, only 80.0% o f m ultigravid moth ers gave the same answer in their antenatal and Birth to Ten interviews.
To assess the validity of self-reported obstetric morbidity the proportion o f accurately identified positive (sensitivity) and negative (specificity) diagnoses (based on data abstracted from hospital records: see Casey et al. 1992) for each o f the five clinical conditions (pre-existing hypertension, PIH, pre-eclampsia, glycosuria and STD seropositivity) were calculated using self-reports o f obstetric morbidity (oedema, hyperten sion, diabetes, urinary tract infection and treatment for STDs) collected during the Birth to Ten interviews.The results of these analyses are displayed in Table 1, which shows that the apparent sensitivity o f self-reports for most obstetric disor ders was around 50%.However, the diagnosed prevalence of all five obstetric conditions was too low to provide accurate estimates o f sensitivity for all but one measure (oedema; with a sensitivity of 40% for pre-eclampsia) of self-reported ob stetric morbidity.Nevertheless, estimates of specificity were generally high, particularly for self-reported hypertension and diabetes which had specificities o f more than 95% for (pre existing or pregnancy-induced) hypertension or pre-eclam p sia and g ly c o su ria resp ectiv ely .S ince o ste n sib ly high specificities are more likely to occur by chance for rarer diag noses such as these, the lower specificity of commoner condi tions, such as self-reported oedema for (pre-exiting or preg nancy-induced) hypertension or pre-eclam psia (64.3% to 66.3%) and self-reported urinary tract infection for STD se ropositivity (65.7%), provide a better indication o f the lim ited specificity of self-reported obstetric morbidity.

Discussion
The impressive reliability o f self-reported gravidity observed in the present study is similar to that described in England (Joffe and Grisso 1985) and the United States (Tilley et al. 1985), and supports the view that mothers accurately recall "significant" life events, such as pregnancy and childbirth (Hewson and Bennett 1987).However, self-reports of past miscarriages, terminations, premature-and still-births, were only modestly reliable.This suggests either that mothers were less willing to disclose unsuccessful outcomes of past preg nancies during interviews with Birth to Ten researchers than they were during consultations with antenatal staff, or that these outcomes constitute less "significant" life events, at least within the context o f the Birth to Ten interviews (Joffe and Grisso 1985).Furthermore, the results of the present study confirm that the sensitivity and specificity o f self-reported obstetric morbidity tend to be poor (Casey et al. 1992).In fact, since hospital records are often incomplete (Tilley et al. 1985;Laurell et al. 1994;Harris et al. 1997), inconsistent (Joffe andGrisso 1985;Harris et al. 1997), inaccurate (Hewson and Bennett 1987), and difficult to abstract (Horwitz and Yu 1984;Harris et al. 1997), it is likely that the sensitivity of self-reported obstetric morbidity might have been even lower than that estimated in the present study.This does not necessarily mean that the mothers themselves were responsible for the limited validity o f their self-reported obstetric morbidity.To start with, some o f the self-reported symptoms examined in the present study provide somewhat less than perfect indicators o f the clinical conditions diag nosed.For example, oedem a may not always accompany pre existing hypertension, PIH or pre-eclam psia (D avey and M acGillivray 1988), while urinary tract infection and previ ous treatment for STDs might not accurately reflect current STD seropositivity (Essén et al. 1994).Nevertheless, assum ing that the hospital records were carefully com piled and ac curately abstracted (a generous assumption at best: Ellison et al. 1997), the self-reports of hypertension and diabetes should have provided reasonable estimates o f clinically diagnosed hypertensive disorders and glycosuria.In fact, these self-re ported conditions had the highest sensitivities and specificities observed in the present study, although the self-reports of high blood pressure appeared to be insensitive for PIH.
While it is tempting to conclude that the limited accuracy of self-reported obstetric morbidity might partly reflect the mod est educational attainm ent o f women in Soweto, the Birth to Ten mothers whose obstetric notes contained antenatal record cards had actually attained a slightly higher level o f school ing than those w hose notes did not.Indeed, com parably (in)accurate self-reports of obstetric morbidity have been ob served among ostensibly better educated mothers in the United States (Tilley et al. 1985), and even those with some tertiary education (Casey et al. 1992).This suggests that mothers eve rywhere are generally unable or unwilling to divulge infor mation about their obstetric health.In part this might reflect the perceived sensitivity of past and current medical events (Hewson and Bennett 1987), in part the inadequacy o f com munication between clinicians and mothers (Tilley et al. 1985;Matshidze et al. 1998).In the present study, for example, the large proportion of mothers who answered "don't know" ex acerbated the limited specificity of self-reported urinary tract infections for diagnoses o f STD infection (see Table 1).Like wise, all but one of the mothers who had been routinely tested for STDs during antenatal care reported that they did not know whether they had ever been treated for STDs.W hile it is pos sible that all of the mothers who tested negative for STDs during antenatal care were neither consulted beforehand nor inform ed thereafter, it is also likely that the social stigma attached to STD (Casey et al. 1992) dissuaded STD-positive mothers from disclosing this information to researchers who were uninvolved with their medical care.
Either way, the limited validity of self-reported obstetric m or bidity inevitably undermines the clinical utility of any infor mation obtained.It might therefore seem appropriate to avoid using self-reports of maternal morbidity wherever possible.This is often impractical, for a num ber o f reasons.Maternal medical records and obstetric notes from previous pregnan cies are often unavailable for consultation by antenatal clinic staff, even where record-keeping facilities are good (Lovell et al. 1987).Likewise, clinicians in the delivery ward can not refer to the medical records of mothers who did not attend antenatal care or those who present for delivery without medi cal records (Elboum e et al. 1987;Laurell et al. 1994).In view of the large num ber of obstetric notes examined in the present study which did not contain antenatal record cards, it ap pears that staff in the delivery ward at Baragwanath Hospital were often unable to consult the antenatal records of Birth to Ten mothers.W hatever the reason, there seems little alterna tive but to rely on maternal reports of obsteric morbidity for those mothers who find themselves in similar circumstances during clinical consultations.
It is therefore clearly important to improve the validity of self-reported obstetric morbidity, through better com m unica tion betw een antenatal clinicians and the m others in their care (Elbourne et al. 1987;M atshidze et al. 1998) and by ensuring greater access to education and health information for women (Shah et al. 1993).For this reason mothers should be provided with their own medical records, or at least a sum mary thereof, along the lines of the "Home-based Maternal Records" advocated by the World Health Organisation (WHO 1994).These records have the twin benefits of facilitating access to each m other's past medical history (wherever she may be), and empowering mothers as active consumers (rather than passive recipients) of health care services (Shah et al. 1993).Research conducted at Alexandra Health Centre and U niversity Clinic in Johannesburg found that patient-held records not only reduced the workload of medical records staff, but also reduced the amount o f time required to register pa tients and increased the proportion of consultations in which clinicians had access to patients' previous medical records (Daviaud et al. 1996).Table 1 .The sensitivity and specificity of questions relating to self-reported obstetric morbidity when compared to diagnoses of five obstetric conditions abstracted from antenatal, obstetric and neonatal records at Baragwanath Hospital.'The number of mothers who correctly reported the presence of each diagnosed condition excludes those who responded "don't know" Specificity (%) : inum ber of negative cases correctlv identified bv maternal self-reportsi x 1 0 0 [total num ber of negative cases diagnosed in hospital records] 'The number of mothers who correctly reported the absence o f each diagnosed condition excludes those who responded "don't know" T 'he lower denom inator reflects the smaller num ber of mothers who provided self-reports of treatment for STD.
tal cards -W hat should they contain?J Trop Pediatr 40: 130-132 FO X N L; SE X T O N M ; H E B E L J R E T A L 1989: The reliability o f self-reports o f smoking and alcohol consum p tion by pregnant women.Addiction Behav 14: 187-195 H A L L M H ; C H N G P K ; M A C G IL L IV R A Y I 1980: Is routine antenatal care worthwhile?Lancet 2: 78-80 H A R R IS H E ; E L L IS O N G T H ; H O LLID A Y M E T AL 1997: M ethodological considerations in the design of an ob stetric database abstracted from medical records.Meth In f M ed 36: 191-200.H E W SO N D & B E N N E T T A 1987: Childbirth research data: M edical records or w om en's reports?Am J Epidemiol 125: 484-491 H O R W IT Z R I & YU E C 1984: Assessing the reliability of epidemiologic data obtained from medical records.J C hrD is 37: 825-831 H U T C H IN G S L; E L L IS O N G T H & JA S S P 2000: The impact o f free health care on antenatal attendance and subse quent birth outcome at Alexandra Health Centre in Johan nesburg.Ann Hum Biol 28: 107-8 J O F F E M & G R IS S O JA 1985: Comparison o f ante-natal hospital records with retrospective interviewing.J Biosoc Sci 17: 113-119 L A U R E L L L; ESSÉN B & L A U R E L L L E T A L 1994: A study of the use o f the N icaraguan antenatal card.J Trop Pediatr 40: 133-136 L O V E L L A; Z A N D ER L I & JA M E S C E E T AL 1987: The St T hom as's H ospital m aternity case notes study: A randomised controlled trial to assess the effects o f giving ex pectant m others their own m aternity case notes.P aediatr Perinat Epidem iol 1: 57-66 M A R K S G C ; H A B IC H T J -P & M U E L L E R W H 1988. Reliability, dependability, and precision o f anthropom etric measurements.The second National Health and Nutrition Ex amination Survey 1976-1980.Am J Epidem iol 130: 578-587 M A T SH ID Z E K P; R IC H T E R L M & E L L IS O N G T H E T AL 1998: Caesarean section rates in South Africa: E vi dence o f bias am ong different 'Population G roups'.Ethn Health 3: 71-79.M ATI K G 1994: Antenatal care.In: Nasah; Mati, Kasonde (eds) Contem porary Issues in M aternal Health Care in A f rica.Harwood Academic Publishers, Luxembourg.M C C O Y D 1996: Free health care for pregnant women and children under six in South Africa: An impact assessment.Health Systems Trust, Durban.P E O P L E S -S H E P S M D ; K A L S B E E K W D & S IE G E L E E T A L 1991: Pre-natal records: A national survey o f con tent.Am J Obstet Gynecol 164: 514-21.R IC H T E R LM ; YACH D & C A M E R O N N E T A L 1995: Enrolment into Birth to Ten (BTT): population and sample characteristics.Paediatr Perinat Epidemiol 9: 109-120 SH A H PM ; SELW YN B J & SH A H K E T A L 1993: Evalu ation of the home-based maternal record: a W HO collabora tive study.Bull WHO 71: 535-548.SO K A L R R & R O H L F F J 1981: Biometry.WH Freeman and Co, San Francisco.S W E E T B R 1992: Bailliêre's M idw ives' Dictionary (8 th Edition).Bailliêre Tindall, London T IL L E Y BC; BA RN ES AB & B E R G S T R A L H E E T AL 1985: A comparison of pregnancy history recall and medical records.Implications for retrospective studies.Am J Epidemiol 121: 269-281 V EN TU RA SJ; M A R TIN JA & T A FFE L SM E T AL 1994: Advance report o f final natality statistics, 1992.Month Vital Stat Rep 43: 1-88 W H O (W orld H ealth O rganisation) 1994: Home-based M a ternal Records.World Health Organisation, Geneva YACH D; C A M E R O N N & PA D A Y A CH EE N E T A L 1991: Birth to ten: child health in South Africa in the 1990s.R ationale and m ethods o f a birth cohort study.P aediatr Perinat Epidemiol 5: 211-233 cases correctly identified bv maternal self-reportsi x 100 [total number of positive cases diagnosed in hospital records] Clinical measures o f obstetric morbidity: