Abstract
Background
Autopsy rates in neonatal intensive care unit (NICU) patients who died are declining worldwide. Postmortem magnetic resonance imaging (MRI) is suggested as adjunct to or substitute for autopsy.
Objective
The aim of this paper was to determine the additional diagnostic value of autopsy in NICU patients and whether autopsy findings were potentially detectable using postmortem MRI.
Methods
From 2008 to 2015, 298 infants died during admission to our NICU. Permission for unrestricted, nonforensic autopsy was obtained in 100 (33.6%) of these 298 infants. Retrospectively, autopsy reports and medical records of NICU patients were compared. Additional autopsy findings were graded according to the Goldman system, grading the clinical relevance of additional findings. In addition, the potential detectability of these additional findings on postmortem MRI was assessed.
Results
Additional findings obtained by autopsy were found in 48% of the cases, divided into major (Goldman I/II, 24%) and minor (Goldman III/IV, 24%) additional findings. Major additional findings were significantly more often found in patients with a lower gestational age, and minor additional findings in patients with a higher postnatal age at death. Of all patients with additional findings determined by autopsy, 56.3% would most likely not have been detected using postmortem MRI.
Conclusions
Our results emphasize the still very important role of autopsy in the NICU setting and show that conventional autopsy could probably not be completely substituted by postmortem MRI.
Key Words: Neonatal intensive care, Postmortem magnetic resonance imaging, Autopsy, Mortality
Introduction
Autopsy of newborn infants can provide vital information to clinicians and families about the cause of death and the accuracy of antemortem clinical diagnosis [1]. Since clinical manifestations of conditions in newborn infants are often nonspecific [2], unintentionally delayed, wrong, or missed diagnoses are still inevitable. In the literature, the latter are defined as diagnostic errors [3]. Patients admitted to an intensive care unit appear to be more prone to diagnostic errors [4, 5, 6]. The detection of diagnostic errors is of great importance for clinicians to evaluate the accuracy of diagnostic methods and to improve medical education. For families, outcomes of autopsy can be a meaningful component of the bereavement process and provides information about the risk of recurrence in future children [7].
Autopsy data are most frequently used as the gold standard compared with clinical diagnoses to identify missed diagnoses. Although there are many potential advantages, declining autopsy rates in deceased neonatal intensive care unit (NICU) patients have been recorded in different countries [8, 9, 10]. This can be accounted for by clinicians' fear of confronting families; an inability to adequately explain the value of autopsy; or barriers for families, e.g., conflicts with religious or cultural beliefs, delay of the funeral, and fear of disfigurement [11, 12]. In addition, advances in new methods of diagnosis are being made [13]. Particularly the value of postmortem magnetic resonance imaging (MRI) as adjunct to or substitute for autopsy is increasingly being studied [7, 14]. Postmortem MRI was unable to detect focal or microscopic injury, functional changes, and some focal forms of brain injury, due to the lower resolution and nature of MRI. In 25% of the deceased patients, postmortem MRI findings did not agree with the cause of death determined by autopsy [15].
Although adult and pediatric intensive care unit patients seem to have a higher risk of diagnostic errors [5, 6], limited information is available on diagnostic errors in NICU patients in particular [2, 4]. Even less is known about the clinical relevance of additional findings, missed due to the use of postmortem MRI instead of autopsy. At present, postmortem MRI is not a standard procedure in our unit. The goals of this study were twofold: first, to determine the additional diagnostic value of autopsy in neonates admitted to a NICU, and second, to determine whether autopsy findings were potentially detectable using postmortem MRI.
Methods
Study Design and Population
A retrospective cohort study of autopsy findings and corresponding medical records of NICU patients was conducted at the Wilhelmina Children's Hospital of the University Medical Center Utrecht, The Netherlands. From 2008 to 2015, 4,785 patients were admitted to the NICU of whom 298 patients died. We included all NICU patients who underwent autopsy at the department of pathology of our institute (n = 117, 39.3%). Patients who did not die at the NICU (n = 3) and patients who underwent limited (n = 13) or forensic (n = 1) autopsies were excluded, leaving a total of 100 eligible patients for this study. Patient characteristics of the study cohort are listed in Table 1.
Table 1.
Sex, n | |
Male/female | 51/49 |
Median gestational age at birth (IQR), weeks | 37.3 (8.0) |
Median birth weight (IQR), g | 2,675 (2,011) |
Median postnatal age at death (IQR), days | 4 (8) |
Median Apgar score (IQR) | |
1 min | 4 (6) |
5 min | 6 (5) |
Type of pregnancy, n | |
Singleton | 84 |
Twin | 14 |
Triplet | 2 |
Mechanical ventilation, n | 96 |
Sepsis, n | 18 |
Seizures, n | 50 |
Inotropes, n | 65 |
Congenital malformations diagnosed before death, n | 38 |
Parental informed consent for autopsy was obtained in all cases. The Ethical Committee of the University Medical Center, Utrecht, The Netherlands, waived the need for informed consent for this study according to Dutch legislation reporting anonymized data.
Evaluation of Reports
Clinical diagnoses and causes of death were obtained from the electronic medical records and discharge letters of the patients. Autopsy diagnoses were obtained from the completed autopsy reports. Autopsies were performed by an experienced pathologist (P.G.J.N.) specialized in perinatal and pediatric pathology using a standardized procedure according to Khong and Malcomson [16]. The latter includes external examination and open dissection with macroscopic and histological examination of all organs. External examination also included photography of the whole body from a frontal and lateral view, and of external abnormalities if present. Additionally, a whole-body radiograph was performed in most cases to detect skeletal deformities. The weight of each organ was assessed against the means for gestational age, and tissue samples of each organ were collected for histological examination. The brain was removed and fixed in formalin prior to histological examination [16].
The causes of death were classified according to the main causes of newborn deaths: perinatal asphyxia, congenital/genetic disorder, prematurity, infection, or ‘other’. Both clinical and autopsy diagnoses were classified in 2 ways: by organ system (respiratory, circulatory, nervous system, gastrointestinal, urinary, or “other”) and by type of condition (infection, congenital malformation, or acquired injury). Discrepancies between clinical and autopsy diagnoses were assessed by an experienced neonatologist (F.G.) and a pathologist (P.G.J.N.) and classified using a classification system by Goldman et al. [1]. This system divides additional findings determined by autopsy into major (class I and II) and minor (class III and IV) additional findings. Major additional findings contain either principal underlying diseases or primary causes of death. Minor additional findings contain related diagnoses, contributing causes, antecedent conditions, or other important conditions. The Goldman criteria are listed in the online supplementary Table (for all online suppl. material, see www.karger.com/doi/10.1159/000493003) [1]. The most relevant additional finding was used to classify each patient.
To determine whether the additional findings were potentially detectable using postmortem MRI, the most relevant finding of each patient was reviewed by a pediatric radiologist experienced in postmortem MRI (M.H.L.) and scored as yes, doubtful, or no.
Statistical Analysis
Statistical analysis was performed to investigate the association between different patient characteristics and the occurrence of missed diagnoses (agreement, minor, or major) using SPSS Statistics 22.0 (SPSS Inc., Chicago, IL, USA). Gestational age, birth weight, sex, Apgar score (after 1 and 5 min), postnatal age at death, and type of pregnancy (singleton/twin/triplet) were analyzed. Continuous variables were analyzed using the Kruskal-Wallis analysis. Significance was set at 0.05. A significant Kruskal-Wallis test was followed by the Bonferroni post hoc method. Binary variables were analyzed using the χ2 test.
Results
During the time interval of this study, 289 NICU patients died (6.0%), and 117 autopsies were performed, of which 100 autopsies were included. The clinical data of the patients are described in Table 1. The leading cause of death of the patients in the study cohort was perinatal asphyxia (45 patients), followed by congenital/genetic disorders (27 patients), complications of prematurity, such as severe respiratory insufficiency or necrotizing enterocolitis (13 patients), infection (11 patients), and other (4 patients). Additional findings determined by autopsy were obtained in 48 patients, half of which were major findings (24 patients). Of these 24 major findings, 2 were classified as class 1 errors. In 1 patient, an infection with cytomegalovirus was determined in the kidneys, which had not been diagnosed during admission. In another patient, meconium peritonitis with candida infection based on a perforation of the terminal ileum was found at autopsy. This had not been diagnosed on antemortem abdominal radiology. Class II findings were found in the residual 22/24 patients with major findings. Minor additional findings were obtained in 24 patients, containing 10 cases of class III and 14 cases of class IV findings. The remaining 52 patients had complete agreement between clinical and autopsy findings. The numbers of additional findings (class I–IV) per organ system and per type of condition are provided in Tables 2 and 3. Additional findings were most frequently found in the circulatory system (35%), followed by the central nervous system (23%). A common class II finding in the circulatory system was hypoplasia of the ductus venosus (6/9 patients). An example of a class II finding in the central nervous system was a bleeding in the cerebellum, brain stem, and spinal cord, which was not diagnosed using cranial ultrasound. Besides additional findings, congenital malformations were noted in 52% of the cases.
Table 2.
Organ system | Class I | Class II | Class III | Class IV | Total |
---|---|---|---|---|---|
Respiratory | 0 | 2 | 2 | 3 | 7 |
Circulatory | 0 | 9 | 1 | 7 | 17 |
Central nervous | 0 | 5 | 4 | 2 | 11 |
Gastrointestinal | 1 | 1 | 0 | 0 | 2 |
Urinary | 1 | 2 | 0 | 1 | 4 |
Other | 0 | 3 | 3 | 1 | 7 |
Table 3.
Type of condition | Class I | Class II | Class III | Class IV | Total |
---|---|---|---|---|---|
Infection | 1 | 4 | 1 | 3 | 9 |
Congenital malformation | 0 | 13 | 4 | 8 | 25 |
Acquired injury | 1 | 5 | 5 | 3 | 14 |
Furthermore, 24 out of the 48 patients with additional findings had more than 1 finding, of whom14 patients had 2 and 10 patients had 3 additional findings in total.
Patient Characteristics and the Occurrence of Additional Findings
The occurrence of major additional findings is significantly more common at a lower gestational age (agreement 36.8 ± 4.8 weeks, minor 34.4 ± 5.4 weeks, major 34.2 ± 4.9 weeks, p < 0.05). Furthermore, minor additional findings were seen in infants with a higher postnatal age at death (agreement 7.3 ± 1.7 days, minor 18.4 ± 5.6, major 10.6 ± 4.3 days, p < 0.05). Clinical data, such as birth weight, sex, Apgar score (after 1 and 5 min), and type of pregnancy (singleton/twin/triplet), did not have a significant effect on the occurrence of additional findings.
Potential Detection of Additional Findings Using Postmortem MRI
Of all additional findings, 27/48 (56.3%) would not have been detected using postmortem MRI. Whether or not additional findings determined by autopsy would have been detected using postmortem MRI is summarized in Table 4. Abnormalities that were unlikely to be detectable by MRI include hypoplasia of the ductus venosus, pneumonia, candida and cytomegalovirus infections, atrial septal defect, and papillary muscle infarction.
Table 4.
Detectable using postmortem MRI? |
|||
---|---|---|---|
yes | doubtful | no | |
Goldman class | |||
I | 0 | 1 | 1 |
II | 8 | 0 | 14 |
III | 5 | 1 | 4 |
IV | 5 | 1 | 8 |
Total | 18 | 3 | 27 |
Discussion
In this retrospective cohort study, unrestricted, nonforensic autopsy was performed in 33.6% of the infants who died during their admission to our NICU. Reports in other NICUs range from 20.3 to 82% [2, 4]. Additional findings determined by autopsy (class I–IV) were encountered in 48% of the cases, consisting of major (24%) and minor (24%) additional findings, also referred to as diagnostic errors. Major findings included class I (2/24) and class II findings (22/24). Minor findings included class III (10/24) and class IV findings (14/24). Congenital malformations were found to be the most frequent type of condition causing the additional finding. This coincides with the fact that congenital/genetic disorders were the second most common cause of death in our study cohort. The higher postnatal age at death in patients with minor additional findings demonstrates that these findings were apparently difficult to detect clinically.
A systematic review of 6 studies by Custer et al. [4] reporting diagnostic errors in the NICU according to Goldman et al. [1] showed major errors in 19.2% (class I, 3.7%; class II, 15.5%). However, not all included studies reported class III and IV errors. Class III errors were reported in 5 studies and were identified in 434/1,168 cases (37.2%), class IV errors were reported in 3 studies and identified in 97/621 cases (15.6%). To compare the outcome of this study with our study, the 95% confidence intervals (CI) of our results was calculated according to a method described by Newcombe [17]. Class I findings were found in 2% (95% CI: 0.55, 7), class II findings in 22% (95% CI: 15, 31.1), class III findings in 10% (95% CI: 5.5, 17.4), and class IV findings in 14% (95% CI: 8.5, 22.1). This shows that our findings are in the same range as those of the study by Custer et al. [4]; however, less class III findings were found. This could be explained by the fact that some included studies by Custer et al. [4] reported multiple errors per patient, which results in a slightly higher percentage of error.
The findings of our study are relevant, since even small congenital malformations may lead to the diagnosis of a syndrome and improved genetic diagnostics and prepregnancy counselling, even in the era of whole-exome sequencing. Recently, it has been suggested that appropriate postmortem MRI of neonates who died could replace autopsy, but it has been emphasized in both papers that histological information is lost [7, 14]. The fact that in our cohort > 55% of the additional findings determined by autopsy would likely not have been detected using postmortem MRI further underscores the still very important role of autopsy in the NICU setting.
There are some limitations of this study that need to be addressed. First, there might be a bias in the selection of cases for autopsy. Of all deceased patients, 33.6% underwent unrestricted, nonforensic autopsy. Parents of infants with uncertain antemortem diagnoses are potentially more likely to grant permission for autopsy, which could cause a selection bias. Therefore, the number of additional findings in the present study might be higher than the number of findings in all deceased NICU patients [18, 19]. However, minor Goldman findings were present even after a more prolonged NICU stay, suggesting that these additional findings could have been found in deceased patients without permission for autopsy as well. Secondly, no postmortem MRI was performed, so only an assessment of potentially missed autopsy findings could be made. The assessment of the potentially missed findings was made, however, by a pediatric radiologist (M.H.L.) with vast experience in postmortem MRI.
Future Directions
Since autopsy rates are declining worldwide, partly due to parents who withhold consent for autopsies, and because postmortem MRI alone cannot provide a histological diagnosis, a new minimally invasive autopsy method is recently being studied [14, 20, 21]. The latter generally involves postmortem MRI with ancillary tests, e.g., external examination of the body, placental histopathological examination, blood sampling and tissue sampling by percutaneous or endoscopic biopsy, which may be supported by image guidance, directed to any abnormal lesion identified with the MRI [20, 22]. A study by the MARIAS collaborative group [14] found that the cause of death or major pathological lesion detected by minimally invasive autopsy, without histological tissue sampling by percutaneous biopsy, was concordant with conventional autopsy in 81.0% of all neonates aged 1 month or younger. This new approach implies minimal cosmetic consequences compared to conventional autopsy and might thus be acceptable in selected cases in which parents decline a conventional autopsy. Further studies are needed before this technique can be implemented on a larger scale.
In conclusion, our study emphasizes the still very important role of autopsy in infants who died during admission to the NICU and shows that conventional autopsy could not be completely substituted by postmortem MRI at present. Clinicians should be aware of the importance of autopsy and the limits of postmortem MRI when informing parents.
Disclosure Statement
The authors declare that there are no conflicts of interest to disclose.
Supplementary Material
Acknowledgements
The authors thank the Stichting Neonatale Neurologie Utrecht for the financial support to publish this paper.
References
- 1.Goldman L, Sayson R, Robbins S, Cohn LH, Bettmann M, Weisberg M. The value of the autopsy in three medical eras. N Engl J Med. 1983 Apr;308((17)):1000–5. doi: 10.1056/NEJM198304283081704. [DOI] [PubMed] [Google Scholar]
- 2.Widmann R, Caduff R, Giudici L, Zhong Q, Vogetseder A, Arlettaz R, et al. Value of postmortem studies in deceased neonatal and pediatric intensive care unit patients. Virchows Arch. 2017 Feb;470((2)):217–23. doi: 10.1007/s00428-016-2056-0. [DOI] [PubMed] [Google Scholar]
- 3.Graber ML, Franklin N, Gordon R. Diagnostic error in internal medicine. Arch Intern Med. 2005 Jul;165((13)):1493–9. doi: 10.1001/archinte.165.13.1493. [DOI] [PubMed] [Google Scholar]
- 4.Custer JW, Winters BD, Goode V, Robinson KA, Yang T, Pronovost PJ, et al. Diagnostic errors in the pediatric and neonatal ICU: a systematic review. Pediatr Crit Care Med. 2015 Jan;16((1)):29–36. doi: 10.1097/PCC.0000000000000274. [DOI] [PubMed] [Google Scholar]
- 5.Goldstein B, Metlay L, Cox C, Rubenstein JS. Association of pre mortem diagnosis and autopsy findings in pediatric intensive care unit versus emergency department versus ward patients. Crit Care Med. 1996 Apr;24((4)):683–6. doi: 10.1097/00003246-199604000-00022. [DOI] [PubMed] [Google Scholar]
- 6.Winters B, Custer J, Galvagno SM, Jr, Colantuoni E, Kapoor SG, Lee H, et al. Diagnostic errors in the intensive care unit: a systematic review of autopsy studies. BMJ Qual Saf. 2012 Nov;21((11)):894–902. doi: 10.1136/bmjqs-2012-000803. [DOI] [PubMed] [Google Scholar]
- 7.Griffiths PD, Paley MN, Whitby EH. Post-mortem MRI as an adjunct to fetal or neonatal autopsy. Lancet. 2005 Apr;365((9466)):1271–3. doi: 10.1016/S0140-6736(05)74816-9. [DOI] [PubMed] [Google Scholar]
- 8.Adappa R, Paranjothy S, Roberts Z, Cartlidge PH. Perinatal and infant autopsy. Arch Dis Child Fetal Neonatal Ed. 2007 Jan;92((1)):F49–50. doi: 10.1136/adc.2005.091447. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Brodlie M, Laing IA, Keeling JW, McKenzie KJ. Ten years of neonatal autopsies in tertiary referral centre: retrospective study. BMJ. 2002 Mar;324((7340)):761–3. doi: 10.1136/bmj.324.7340.761. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Swinton CH, Weiner J, Okah FA. The neonatal autopsy: can it be revived? Am J Perinatol. 2013 Oct;30((9)):739–44. doi: 10.1055/s-0032-1332798. [DOI] [PubMed] [Google Scholar]
- 11.Breeze AC, Statham H, Hackett GA, Jessop FA, Lees CC. Perinatal postmortems: what is important to parents and how do they decide? Birth. 2012 Mar;39((1)):57–64. doi: 10.1111/j.1523-536X.2011.00513.x. [DOI] [PubMed] [Google Scholar]
- 12.Sherwood SJ, Start RD. Asking relatives for permission for a post mortem examination. Postgrad Med J. 1995 May;71((835)):269–72. doi: 10.1136/pgmj.71.835.269. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.O'Grady G. Death of the teaching autopsy. BMJ. 2003 Oct;327((7418)):802–3. doi: 10.1136/bmj.327.7418.802. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Thayyil S, Sebire NJ, Chitty LS, Wade A, Chong W, Olsen O, et al. MARIAS collaborative group Post-mortem MRI versus conventional autopsy in fetuses and children: a prospective validation study. Lancet. 2013 Jul;382((9888)):223–33. doi: 10.1016/S0140-6736(13)60134-8. [DOI] [PubMed] [Google Scholar]
- 15.Leadbetter KZ, Vesoulis ZA, White FV, Schmidt RE, Khanna G, Shimony JS, et al. The role of post-mortem MRI in the neonatal intensive care unit. J Perinatol. 2017 Jan;37((1)):98–103. doi: 10.1038/jp.2016.156. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Khong TY, Malcomson RD, editors. Keeling's Fetal and Neonatal Pathology. Springer; 2015. [Google Scholar]
- 17.Newcombe RG. Two-sided confidence intervals for the single proportion: comparison of seven methods. Stat Med. 1998 Apr;17((8)):857–72. doi: 10.1002/(sici)1097-0258(19980430)17:8<857::aid-sim777>3.0.co;2-e. [DOI] [PubMed] [Google Scholar]
- 18.Podbregar M, Voga G, Krivec B, Skale R, Parežnik R, Gabršček L. Should we confirm our clinical diagnostic certainty by autopsies? Intensive Care Med. 2001 Nov;27((11)):1750–5. doi: 10.1007/s00134-001-1129-x. [DOI] [PubMed] [Google Scholar]
- 19.Shojania KG, Burton EC, McDonald KM, Goldman L. Overestimation of clinical diagnostic performance caused by low necropsy rates. Qual Saf Health Care. 2005 Dec;14((6)):408–13. doi: 10.1136/qshc.2004.011973. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Sebire NJ, Weber MA, Thayyil S, Mushtaq I, Taylor A, Chitty LS. Minimally invasive perinatal autopsies using magnetic resonance imaging and endoscopic postmortem examination (“keyhole autopsy”): feasibility and initial experience. J Matern Fetal Neonatal Med. 2012 May;25((5)):513–8. doi: 10.3109/14767058.2011.601368. [DOI] [PubMed] [Google Scholar]
- 21.Whitby EH, Offiah AC, Cohen MC. Initial experiences of a minimally invasive autopsy service. A report of the techniques and observations in the first 11 cases. Pediatr Dev Pathol. 2015 Jan-Feb;18((1)):24–9. doi: 10.2350/14-06-1503-OA.1. [DOI] [PubMed] [Google Scholar]
- 22.Addison S, Arthurs OJ, Thayyil S. Post-mortem MRI as an alternative to non-forensic autopsy in foetuses and children: from research into clinical practice. Br J Radiol. 2014 Apr;87((1036)):20130621. doi: 10.1259/bjr.20130621. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.