Persistent inaccuracies in completion of medical certificates of stillbirth: A cross‐sectional study

Abstract Background The UK Medical Certificate of Stillbirth (MCS) records information relevant to the cause of stillbirth of infants ≥24 weeks’ gestation. A cross‐sectional audit demonstrated widespread inaccuracies in MCS completion in 2009 in North West England. A repeat study was conducted to assess whether practice had improved following introduction of a regional care pathway. Methods 266 MCS issued in 14 North West England obstetric units during 2015 were studied retrospectively. Cause of death was assigned following review of information available at the time of MCS completion. This was compared to that documented on the MCS, and to data from 2009. Results Twenty‐three certificates were excluded (20 inadequate data, 3 late miscarriages). 118/243 (49%) MCS contained major errors. Agreement between the MCS and adjudicated cause of stillbirth was fair (Kappa 0.31; 95% CI 0.24, 0.38) and unchanged from 2009 (0.29). In 2015, excluding 34 terminations of pregnancy, the proportion of MCSs documenting “unexplained” stillbirths (113/211; 54%) was reduced compared to 2009 (158/213; 74%); causality could be assigned after case note review in 78% cases. Recognition of fetal growth restriction (FGR) as a cause of stillbirth improved (2015: 30/211; 14% vs 2009: 1/213; 0.5%), although 71% cases were missed. 47% MCSs following termination of pregnancy documented an iatrogenic primary cause of death. Conclusions Completion of MCSs remains inaccurate, particularly in recognition of FGR as a cause of stillbirth. Detailed case note review before issuing the MCS could dramatically improve the usefulness of included information; evaluation of practitioner education programmes/internal feedback systems are recommended.


| INTRODUC TI ON
In the United Kingdom (UK), the Medical Certificate of Stillbirth is issued to the parent(s) of a baby born without signs of life after 24 completed weeks' gestation (see Figure S1 for an example). 1 Similar certification systems are used in high-income countries including the United States of America (USA), Australia, New Zealand, and Ireland. The document is issued within days of birth, recording information relevant to the baby's demise. Hence, information from investigations such as autopsy or placental histopathology is not available at the time of issue. Unlike neonatal, child and adult deaths, in which a Coroner can instigate measures to determine the cause of death where unclear, stillbirths are not currently governed by the UK Coronial law, although this is being reviewed. 2 Autopsy and placental histopathology can reveal the cause of stillbirth, or provide additional information that alters the management of the mother's future pregnancies/health in up to 50% of cases. 3 Due to incomplete uptake of these investigations (autopsy: 48.1%, placental histology: 88.8%), 4 for many parents the Medical Certificate of Stillbirth is the principal source of information regarding the cause of their baby's death. From a public health perspective, the UK Office of National Statistics uses data from Medical Certificates of Stillbirth when describing key characteristics of stillbirths nationally. These are used to assist Governmental evidence-based policy decisions. 5 In 2016, the UK Government committed to halving stillbirth rates by 2025. 6 Therefore, the accuracy of data provided by Medical Certificates of Stillbirth is increasingly important to identify appropriate interventions to prevent avoidable stillbirths.
We conducted a regional cross-sectional audit of the accuracy  To minimise inflation of gestational age contributing to overdiagnosis of FGR, the birthweight centile, adjusted for maternal characteristics was calculated based on the date fetal demise was confirmed. In accordance with a recent international consensus statement, FGR was considered to be present if the baby's birthweight was <3 rd customised centile, or two or more of the following were present: (a) AC/EFW <10 th centile; (b) AC/EFW crossing >2 quartiles on non-customized growth centiles; (c) Cerebral perfusion ratio <5 th centile or umbilical artery pulsatility index >95 th centile. 13 In addition, the diagnosis of FGR at borderline centiles (3-10) was strengthened by evidence of placental insufficiency (abnormal fetoplacental Doppler results, ultrasound confirmed oligohydramnios or a markedly small 13 or visibly grossly infarcted placenta). 14 Where birthweight was >3 rd and <10 th centile but did not fulfil the criteria above an alternative diagnosis was assigned where judged to be clinical significant eg placental abruption, chorioamnionitis, and intrapartum hypoxia.

| ME THODS
Data extractors were provided with a copy of the ReCoDe classification of stillbirth. 15 This classification was chosen for i) consistency with the previous study, ii) low rate of "unexplained" stillbirths, and iii) recognition of FGR as a primary cause of death. The cause of stillbirth was then assigned by reviewers according to ReCoDe categories, based only on information that was available at the time of Medical Certificate of Stillbirth completion. For example, autopsy or placental histopathology results (usually unavailable for up to 6 weeks after birth) were not taken into account. As ReCoDe is a hierarchical classification system this promotes the importance of some abnormalities (eg FGR) over other potentially causal events eg intrapartum hypoxia. However, data extractors were asked to consider the relevant contributions of competing relevant conditions at death, rather than obeying the strict hierarchy of described conditions. For example, where FGR was noted in the presence of an alternative potential cause of death (for example significant placental abruption), data extractors were asked to consider which was likely to have been the main condition leading to fetal demise.
Furthermore, secondary analysis using the ReCoDe-R system was also performed. 16 This demotes FGR in the hierarchy of relevant conditions at death to prioritise other identifiable causes and has been shown to more than halve the proportion of stillbirths attributed to FGR.
Where termination of pregnancy had occurred, data extractors were instructed to assign an iatrogenic primary cause of death (as medical intervention directly led to the fetus' demise at that specific point in time), with recognition of the condition leading to the decision for termination to be acknowledged as a secondary cause, under either "other diseases or conditions in fetus" or "other maternal diseases or conditions affecting fetus" ( Figure S1).

| Statistical analysis
Data for gestation at stillbirth and primary cause of stillbirth were compared between that documented on the Medical Certificate of Stillbirth and that expected from case note review. Accuracy was quantified using the Kappa statistic; agreement was categorised as nil (0)   The majority of stillbirths in 2015 were diagnosed antepartum (

| Trends in causes of stillbirth as documented and adjudicated
Between 2009 and 2015, there was an increase in reporting of fetal causes (predominantly FGR) and decrease in reporting of "unexplained" causes on the Medical Certificate of Stillbirth (Table 2).
However, these causes remained under-and over-reported, respectively. Other broad categories of cause of stillbirth remained stable except for an increase in reporting of intrapartum asphyxia (55%, 11/20 of stillbirths in one hospital, of which 64%, [7/11] cases were adjudicated to be due to FGR).  (Table S1).

| Accuracy of Medical Certificate of Stillbirth completion
In the 34 terminations of pregnancy in 2015, 47% (n = 16) Medical Certificates of Stillbirth reflected the iatrogenic nature of the infant's death; 44% (n = 15) were erroneously classified as being primarily caused by congenital abnormality that instigated the decision for termination.

| Principal findings
In this follow-up study, we have shown that although there has been modest improvement, the previously reported inaccuracies in Medical Certificates of Stillbirth have persisted. 7 This reinforces the need for an ongoing perinatal surveillance process, as conclusions drawn from Medical Certificate of Stillbirth data alone inadequately report the true incidence of potentially avoidable causes of stillbirth, such as FGR. In this, and the preceding study, FGR has been determined to be the most common adjudicated cause of stillbirth.

| Strengths of the study
The major strength of this study is the systematic examination of all stillbirths from a variety of hospital settings. The total identified cases in the audit period matches that reported to the MBRRACE-UK perinatal surveillance programme from constituent hospitals in the same time period; 4 92.0% of these cases were reviewed in detail. Secondly, we demonstrate that even with minimal specific training clinicians can extract appropriate data to reach an informed conclusion regarding cause of stillbirth. Finally, by applying the same methodology in two consecutive studies, we are able to assess changes over time in the reporting, and prevalence, of various causes of stillbirth.

| Limitations of the data
The major weakness is that we were unable to examine the accuracy of secondary information included on the Medical Certificate of Stillbirth. This could be improved by routine inclusion of a photocopy of the Medical Certificate of Stillbirth in the maternal notes.
Secondly, it was necessary to exclude 23 cases from the study population. This may have minimised the reported error rate.
Another source of bias is likely to arise from use of date of confirmation of fetal death as a proxy for the date of actual fetal death within the study. If this could be determined, it is likely that the accuracy of reported gestational age would be worse, and that the proportion of FGR-attributed deaths in the "borderline" cases (centiles 5-10) minimised, although steps were taken to reinforce the diagnosis in these cases. No gold-standard method for ascertaining actual date of fetal death is described; indeed one proposed algorithm could only be implemented in 47% of cases in the author's own study. 19 Thus date of confirmation of fetal death is the only fixed time point that could be used.
Finally, use of the ReCoDe classification system, 15 with its hierarchical design, to assign cause of fetal death may have influenced the proportion of stillbirths attributed to FGR. 16

| Interpretation
Although some progress has been made (reduction in the proportion of Medical Certificates of Stillbirth with "unexplained" causes and improvement in baseline recognition of FGR) it is concerning that in the last 6 years there has been no significant improvement in agreement between initial and reviewed causes of stillbirth, with a similar proportion of major errors, and a significant burden of undocumented/unrecognised FGR. This is despite regional, 8 national, 9 and international 10 initiatives focussed on the antenatal detection and management of FGR. In one participating unit, the mechanism of death (perinatal asphyxia) was substituted for the cause (FGR); while both may reflect underlying placental insufficiency, overlooking FGR in this context may adversely affect care in a subsequent pregnancy.
The inaccuracy of documented gestational age was affected by use of delivery date (rather than the date on which fetal death was confirmed) to calculate gestational age at stillbirth, an error also noted in the USA. 23 This was particularly influential where co-twin survival resulted in up to 64 days between confirmation of death and birth. Although not part of the primary analysis, it was concerning that the iatrogenic nature of death is not accurately recorded in almost 50% of termination of pregnancy cases; substitution of the congenital abnormality as the cause of death may lead to an overestimate of this disorder as a cause for stillbirth.
We propose that to improve the accuracy of data reported on Causes are categorised according to the Relevant Condition at Death (ReCoDe) classification system. 14 Although fetal growth restriction (FGR) is being acknowledged as a cause of stillbirth more frequently than in 2009, there is a persistent failure to identify all cases of stillbirth primarily due to FGR. Furthermore, inappropriate classification of stillbirths as "unexplained" (equivalent to classifications I1 and I2) persists, although at a lower rate than in 2009. There was an increase in the reported rate of asphyxia as primary cause of stillbirth between 2009 and 2015, resulting in an increase in the proportion of stillbirths being classified as "intrapartum." On review, the majority of these stillbirths occurred in compromised, FGR babies. Key: MCS = Medical Certificate of Stillbirth, Adjudicated = adjudicated cause of stillbirth after review of medical records.

TA B L E 2 (Continued)
The contemporaneous Midlands and North of England Stillbirth Study (MiNESS) 35 describes a strikingly similar distribution of causes of death to that described here. This gives confidence in the validity/generalisability of the results presented here. MiNESS reports a greater proportion of stillbirths attributed to placental insufficiency; this is expected given that they had access to placental histopathology data. Thus, we believe it is clear that substantial improvements in the accuracy of Medical Certificates of Stillbirth can be made with structured case review alone.