Abstract
Background
Previous studies identified microscopic changes associated with intrauterine retention of stillbirths based on clinical time of death. The objective of this study was to utilize unsupervised machine learning (not reliant on subjective measures) to identify features associated with time from death to delivery.
Methods
Data were derived from the Stillbirth Collaborative Research Network. Features were chosen a priori for entry into hierarchical cluster analysis, including fetal and placental changes.
Results
A four-cluster solution (coefficient = 0.983) correlated with relative time periods of “no retention,” “mild retention,” “moderate retention,” and “severe retention.” Loss of nuclear basophilia within fetal organs were found at varying rates among these clusters.
Conclusions
Hierarchical cluster analysis is able to classify stillbirths based on histopathological changes, roughly correlating to length of intrauterine retention. Such clusters, which rely solely on objective fetal and placental findings, can help clinicians more accurately assess the interval from death to delivery.
Keywords: Perinatal pathology, Time of Death, Stillbirth/Intrauterine Fetal Demise, Unsupervised Machine Learning
Introduction
Stillbirth rates are stable worldwide despite decreases in infant mortality; this results in a constant burden on providers to explain these unfortunate events (1). This is heightened in the setting of medicolegal claims, which are common following a stillbirth as patients seek to heal after this devastating event (2–4). Postmortem pathology has significant implications in legal cases since 25–60% of stillbirths do not have a clear etiology; knowing time of death can help elucidate these details (5).
Our current methods for determining time of death are based on knowledge of last known alive, usually derived from subjective reports of fetal movements and rarely from fetal heart rate or ultrasonographic evidence (6). Consensus has been established regarding pathologic findings indicative of differing lengths of intrauterine retention, but these findings are largely correlated to subjective measures (7). Algorithms have also been developed to estimate time from death to delivery based on clinical evidence (8).
Therefore, we sought to minimize such subjectivity by utilizing unsupervised machine learning to identify which features were associated with clinically relevant interval from time of death to delivery. This allows for grouping of relevant variables without bias of subjective last fetal movement. Such information is of great utility as it provides objective evidence of time from fetal death to delivery.
Methods
Data
The Stillbirth Collaborative Research Network is a comprehensive study of stillbirths and livebirths conducted from 2006 to 2009. Study procedures were approved by the Institutional Review Board at each clinical site and the Data Analysis Center. Written, informed consent was collected from each participant; all procedures have been described previously (9).
Data were derived from the complete study population; participants who delivered a singleton stillbirth of gestational age (GA) > 20 weeks and consented to complete postmortem examination were included. Clinical information was extracted via chart review and interview.
Postmortem examination
Study pathologists were trained in standard postmortem and placental examinations (10,11). Pathologists determined cause of death according to the Initial Causes of Fetal Death system (12). Histologic features were reported based on those determined by Genest et al (13–15). Loss of nuclear basophilia in the following areas was included: renal cortical tubules (isolated = 4 hours, all cells = 4 weeks), liver (isolated = 24 hours, all cells = 96 hours), myocardium (inner half = 24 hours, outer half = 48 hours), bronchial epithelium (isolated = 96 hours), tracheal cartilage (isolated = 1 week), gastrointestinal tract (all cells = 1 week), and adrenal glands (all cells = 1 week) (15). Placental findings included: villous intravascular karyorrhexis = 6+ hours; luminal obliterations of stem villi (multifocal = 2+ days, extensive = 2+ weeks); and extensive fibrosis of terminal villi = 2+ weeks (13). Genest et al. (14) further estimated fetal death based on the grade of maceration: I = desquamation ≥1 cm and/or brown-red discoloration of umbilical stump; II = desquamation involving face/abdomen/back; III = desquamation involving ≥5% of body surface; IV = brown skin discoloration; and V = mummification.
Statistical analyses
Lesions were entered into hierarchical cluster analysis. Using this method, we identified distinct groups of stillbirths based on hierarchical relationships among lesions, thus constructing groups intended to correlate with increasing intrauterine retention time (16). Clusters were derived using agglomerative (“bottom-up”) or divisive (“top-down”) clustering, according to which method resulted in the highest clustering coefficient and were merged according to optimal cluster number (elbow method) (17–20) Clustering was completed in R (version 4.1.3); tested linkage methods for clustering included average, single, complete, and ward.
Once clusters were derived, sample characteristics were compared across and between clusters. Kruskall-Wallis and Mann-Whitney tests were used for continuous data; Chi-Square tests were used for categorical data. Significance was set at p < 0.05.
Results
Three-hundred and seventy-nine (379) patients met inclusion criteria for these analyses. A four-cluster agglomerative solution using the ward method yielded the highest clustering coefficient (c = 0.983). Comparisons across clusters are illustrated in Tables 1–3.
Table 1:
Maternal characteristics
| Cluster 1 (n = 122) |
Cluster 2 (n = 39) |
Cluster 3 (n = 106) |
Cluster 4 (n = 112) |
p-value | Post-hoc comparisons | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 v 2 | 1 v 3 | 1 v 4 | 2 v 3 | 2 v 4 | 3 v 4 | ||||||
| Age (years) | 26.4 ± 6.4 | 29.5 ± 6.3 | 27.6 ± 6.9 | 28.4 ± 6.9 | 0.047 | 0.012 | 0.2 | 0.04 | 0.14 | 0.3 | 0.4 |
| Race (minority) | 60 (49.2) | 14 (35.9) | 42 (39.6) | 36 (32.1) | 0.059 | 0.15 | 0.15 | 0.008 | 0.7 | 0.7 | 0.2 |
| Ethnicity (Hispanic) | 34 (27.9) | 17 (43.6) | 37 (34.9) | 42 (37.8) | 0.2 | 0.07 | 0.3 | 0.1 | 0.3 | 0.5 | 0.7 |
| Education (years) | 12.5 ± 3.0 | 13.6 ± 2.5 | 12.8 ± 3.1 | 13.2 ± 3.0 | 0.15 | 0.05 | 0.5 | 0.09 | 0.13 | 0.5 | 0.3 |
| Pre-pregnancy history | |||||||||||
| Diabetes | 6 (5.4) | 2 (5.3) | 8 (8.1) | 5 (4.7) | 0.8 | 0.9 | 0.4 | 0.8 | 0.7 | 0.9 | 0.3 |
| Hypertension | 15 (13.4) | 5 (13.2) | 16 (16.2) | 13 (12.1) | 0.9 | 0.9 | 0.6 | 0.8 | 0.7 | 0.9 | 0.4 |
| Prenatal history | |||||||||||
| Had prenatal care | 99 (88.4) | 37 (97.4) | 94 (94.9) | 101 (95.3) | 0.14 | 0.12 | 0.09 | 0.065 | 0.9 | 0.9 | 0.9 |
| Reduced fetal movement | 41 (37.3) | 18 (47.4) | 66 (66.7) | 71 (67.6) | <0.001 | <0.001 | <0.001 | <0.001 | 0.038 | 0.027 | 0.9 |
| Alcohol in pregnancy | 46 (41.1) | 13 (34.2) | 41 (41.4) | 38 (35.5) | 0.7 | 0.5 | 0.9 | 0.4 | 0.4 | 0.9 | 0.4 |
| Smoking in pregnancy | 29 (25.9) | 4 (10.5) | 19 (19.2) | 13 (12.1) | 0.035 | 0.048 | 0.2 | 0.01 | 0.2 | 0.9 | 0.2 |
All values given as mean ± standard deviation (continuous measures) or number (percentage of cluster) (categorical measures). P-values derived from Kruskall-Wallis or Chi-Square tests (continuous and categorical, respectively) for omnibus tests and Mann-Whitney and Chi-Square tests for post-hoc comparisons. P-values < 0.05 are considered significant (bolded).
Table 3:
Histologic findings
| Cluster 1 | Cluster 2 | Cluster 3 | Cluster 4 | p-value | Post-hoc comparisons | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 v 2 | 1 v 3 | 1 v 4 | 2 v 3 | 2 v 4 | 3 v 4 | ||||||
| Fetal findings (loss of nuclear basophilia) – clustering measures | |||||||||||
| Renal cells | |||||||||||
| Isolated | 1 (0.8) | 19 (48.7) | 61 (57.5) | 0 (0.0) | <0.001 | <0.001 | <0.001 | 0.9 | 0.3 | <0.001 | <0.001 |
| Majority | 1 (0.8) | 10 (25.6) | 16 (15.1) | 106 (94.6) | <0.001 | <0.001 | <0.001 | <0.001 | 0.14 | <0.001 | <0.001 |
| Hepatocytes | |||||||||||
| Isolated | 0 (0.0) | 38 (97.4) | 4 (3.8) | 0 (0.0) | <0.001 | <0.001 | 0.045 | -- | <0.001 | <0.001 | 0.054 |
| Majority | 0 (0.0) | 38 (97.4) | 4 (3.8) | 0 (0.0) | <0.001 | <0.001 | 0.045 | -- | <0.001 | <0.001 | 0.054 |
| Myocytes | |||||||||||
| Inner | 0 (0.0) | 15 (38.5) | 70 (66.0) | 112 (100.0) | <0.001 | <0.001 | <0.001 | <0.001 | 0.003 | <0.001 | <0.001 |
| Outer | 0 (0.0) | 6 (15.4) | 65 (61.3) | 103 (92.0) | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 |
| Bronchial epithelium | 4 (3.3) | 5 (12.8) | 54 (50.9) | 78 (69.6) | <0.001 | 0.039 | <0.001 | <0.001 | <0.001 | <0.001 | 0.005 |
| GI tract (majority) | 1 (0.8) | 26 (66.7) | 96 (90.6) | 109 (97.3) | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | 0.035 |
| Adrenal glands (majority) | 3 (2.5) | 21 (53.8) | 92 (86.8) | 108 (96.4) | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | <0.001 | 0.01 |
| Placental findings | |||||||||||
| Villous karyorrhexis | 12 (9.8) | 15 (38.5) | 34 (32.1) | 63 (56.2) | <0.001 | <0.001 | <0.001 | <0.001 | 0.5 | 0.05 | <0.001 |
| Fetal blood cell karyorrhexis | 9 (7.4) | 8 (20.5) | 21 (19.8) | 37 (33.0) | <0.001 | 0.033 | 0.006 | <0.001 | 0.9 | 0.14 | 0.027 |
| Chorionic vascular lesions | 29 (23.8) | 17 (43.6) | 20 (18.9) | 46 (41.1) | <0.001 | 0.017 | 0.4 | 0.005 | 0.002 | 0.8 | <0.001 |
| Disc calcifications | 1 (0.8) | 1 (2.6) | 4 (3.8) | 3 (2.7) | 0.5 | 0.4 | 0.2 | 0.4 | 0.9 | 0.9 | 0.7 |
| Umbilical cord | |||||||||||
| Degenerative changes | 7 (5.7) | 10 (25.6) | 13 (12.3) | 22 (19.6) | 0.002 | 0.001 | 0.082 | 0.001 | 0.05 | 0.4 | 0.14 |
| Wharton’s substance calcifications | 0 (0.0) | 0 (0.0) | 0 (0.0) | 1 (0.9) | 0.7 | -- | -- | 0.5 | -- | 0.9 | 0.9 |
All values given number (percentage of cluster). P-values derived from Chi-Square tests. P-values < 0.05 are considered significant (bolded).
Characterizing clusters (Table 4)
Table 4:
Cluster Characterization (Summary)
| Cluster 1 | Cluster 2 | Cluster 3 | Cluster 4 | |
|---|---|---|---|---|
| Cluster name | No retention | Mild retention | Moderate retention | Severe retention |
| Fetal lesions (loss of nuclear basophilia)* | None | Isolated/majority renal cells, isolated/majority hepatocytes, inner myocytes, GI tract, adrenal glands | Isolated renal cells, inner/outer myocytes, bronchial epithelium, GI tract, adrenal glands | Majority renal cells, inner/outer myocytes, bronchial epithelium, GI tract, adrenal glands |
| Placental lesions* | Chorionic vascular lesions | Villous & fetal blood cell karyorrhexis, chorionic vascular lesions | Villous karyorrhexis | Villous & fetal blood cell karyorrhexis, chorionic vascular lesions |
| Umbilical cord lesions* | None | Degenerative changes | None | None |
| Maceration Grade IV-V | 4.9% | 5.1% | 12.3% | 28.6% |
| Reduced fetal movement | 37.3% | 47.4% | 66.7% | 67.6% |
| Cause of death | ||||
| Intrapartum | 36.4% | 17.9% | 4.7% | 3.6% |
| Obstetric | 59.5% | 56.4% | 27.4% | 25.0% |
Cluster 1 had limited post-mortem changes and the statistically highest rate of intrapartum cause of death, lowest GA at delivery, and lowest rates of grade IV-V maceration and reported reduced fetal movement. This group likely represents stillbirths with limited to no intrauterine retention (“No retention” cluster).
Cluster 2 had loss of nuclear basophilia within the GI tract, adrenal glands, and hepatocytes and higher rates of degenerative umbilical cord changes relative to cluster 1. Compared to cluster 1, this group had similar rates of maceration grade IV-V and obstetric cause of death but had significantly lower rates of intrapartum cause of death. Cluster 2 had significantly higher rates of reduced fetal movement compared to cluster 1 and lower compared to clusters 3 and 4. This group likely represents a short length of intrauterine retention that is longer than cluster 1 but shorter than 3 and 4 (“Mild retention” cluster).
Cluster 3 had intermediate rates of histologic changes; loss of nuclear basophilia was observed in the GI tract, adrenal glands, isolated renal cells, inner/outer myocytes, and bronchial epithelium. This group also had intermediate rates of maceration grade IV-V, high rates of reduced fetal movement, and low rates of intrapartum cause of death. This group likely represents stillbirths with prolonged, though not the longest, intrauterine retention (“Moderate retention” cluster).
Cluster 4 had significant post-mortem changes, including complete inner myocyte loss of nuclear basophilia and high rates of loss of nuclear basophilia in the GI tract, adrenal glands, outer myocytes, and renal cells. This group also had statistically highest rates of maceration grade IV-V and reduced fetal movement and lowest rates of intrapartum cause of death. This group likely represents stillbirths with the longest length of intrauterine retention compared to the other three clusters (“Severe retention” cluster).
Discussion
In this study, we utilized hierarchical cluster analysis to identify four groups of stillbirths according to histologic features with the following relative intervals: No retention, Mild retention, Moderate retention, and Severe retention. Individuals in the clusters were well-correlated, indicated by the agglomerative clustering coefficient of 0.983. These clusters highlight that objective groups of stillbirths of differing intrauterine retention times can be derived without using subjective last known normal. These findings indicate which histologic findings are indeed associated with different times from death to delivery, building on those previously established (7,13–15,21). Isolated loss of renal tubular nuclear basophilia (≥4 hours) was a finding that increased in prevalence from Mild to Moderate retention. Further validating this finding, only majority renal cell loss of nuclear basophilia (≥4 weeks) was observed in Severe retention. Loss of nuclear basophilia in inner versus outer myocytes, (≥24 and ≥48 hours) followed a similar trend from Mild to Severe retention.
The clinical utility of such an approach lies in the ability for pathologies to match features found on postmortem examination with the relative cluster. Though there are overlapping features among clusters, those features listed above that are unique to clusters can signify a fetus may have been retained for that relative period of time. Furthermore, by relying solely on objective findings, including fetal and placental microscopic examination, pathologists and clinicians can determine relative time of retention without use of subjective fetal movements.
Several findings were not consistent with consensus, most strikingly loss of nuclear basophilia in hepatocytes (both isolated and majority, ≥96 hours) only in the Mild retention cluster. If such a finding were indeed related to intrauterine retention, we would expect it to be present in all longer retention clusters. Loss of nuclear basophilia is relatively uncommon in this cohort (11.1% of the total cohort); these cases may represent a subset of fetuses who experienced intrauterine hepatic changes affecting both fetal liver size and postmortem hepatic degeneration. The original studies by Genest et al. did not thoroughly assess maternal conditions such as gestational diabetes or acute fatty liver of pregnancy or fetal myelopoiesis, anemia, biliary atresia, or infection, which may also contribute (15,22,23). Future external validation of these clusters will help elucidate the cause of this unexpected finding.
Placental lesions in our clusters may correlate with both intrauterine retention times and cause of death. Intravascular karyorrhexis in Mild through Severe clusters, correlated well with consensus, with reported retention ≥6 hours. Vascular karyorrhexis and chorionic vascular lesions both can be attributed to intrauterine retention and fetal vascular malperfusion, which may result from umbilical cord accidents (24). Our moderate and severe retention clusters had the highest rates of umbilical-cord related cause of death, and the severe retention cluster had the highest rate of villous karyorrhexis; these findings may be related in attributing cause of death.
Our study has several strengths, principally the quality of our pathological data, derived from a study with standardized training and protocols, which allows us to state our findings with confidence. Our methodology is singular in being able to objectively identify findings correlated with different lengths of intrauterine retention. We also are cognizant of the limitations of this study; while we were able to correlate clusters with clinical variables and estimate time from death to delivery, we are unable to give an exact time frame from our analyses. We would have liked to have utilized an objective measurement to estimate this interval, such as determining the difference between clinical GA at delivery and GA according to postmortem examination; however, we do not have this capability in our data at this time. We also did not assess further confounding factors such as organ weights, fetal conditions, and maternal factors that may influence findings. Such factors will be included in further studies of these clusters.
In conclusion, we were able to identify four clusters of stillbirths stratified by time from death to delivery based on histopathologic findings using hierarchical cluster analysis. This information will be of great clinical utility for those caring for and identifying causes of these patients’ stillbirths. Further analysis will seek to correlate clusters with quantifiable time of death utilizing methods from adult forensic sciences (25).
Table 2:
Stillbirth characteristics
| Cluster 1 | Cluster 2 | Cluster 3 | Cluster 4 | p-value | Post-hoc comparisons | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 v 2 | 1 v 3 | 1 v 4 | 2 v 3 | 2 v 4 | 3 v 4 | ||||||
| Gestational age at delivery | 26.6 ± 6.7 | 28.4 ± 7.0 | 31.1 ± 6.3 | 29.1 ± 6.5 | <0.001 | 0.12 | <0.001 | 0.001 | 0.03 | 0.5 | 0.033 |
| Vaginal delivery | 98 (80.3) | 33 (84.6) | 90 (84.9) | 99 (88.4) | 0.4 | 0.5 | 0.4 | 0.1 | 0.9 | 0.6 | 0.4 |
| External exam | |||||||||||
| Weight (g) | 1142 ± 1148 | 1328 ± 1102 | 1653 ± 1174 | 1308 ± 1229 | 0.001 | 0.2 | <0.001 | 0.6 | 0.13 | 0.5 | 0.005 |
| Sex (female) | 50 (41.0) | 19 (48.7) | 46 (43.4) | 53 (47.3) | 0.7 | 0.4 | 0.7 | 0.3 | 0.6 | 0.9 | 0.6 |
| Maceration grade IV-V | 6 (4.9) | 2 (5.1) | 13 (12.3) | 32 (28.6) | <0.001 | 0.9 | 0.8 | <0.001 | 0.4 | 0.003 | 0.003 |
| Cause of death (INCODE) | |||||||||||
| Obstetric | 72 (59.5) | 22 (56.4) | 29 (27.4) | 28 (25.0) | <0.001 | 0.7 | <0.001 | <0.001 | 0.001 | <0.001 | 0.7 |
| Intrapartum | 44 (36.4) | 7 (17.9) | 5 (4.7) | 4 (3.6) | <0.001 | <0.001 | <0.001 | <0.001 | 0.017 | 0.007 | 0.7 |
| Umbilical cord | 12 (9.9) | 7 (17.9) | 27 (25.5) | 27 (24.1) | 0.011 | 0.3 | 0.002 | 0.004 | 0.3 | 0.4 | 0.8 |
| Placental | 54 (44.6) | 21 (53.8) | 69 (65.1) | 65 (58.0) | 0.018 | 0.3 | 0.002 | 0.041 | 0.2 | 0.6 | 0.3 |
| Genetic/syndromic | 33 (27.3) | 9 (23.1) | 27 (25.5) | 22 (19.6) | 0.6 | 0.6 | 0.8 | 0.2 | 0.8 | 0.6 | 0.3 |
| Hypertensive disorders | 19 (15.7) | 2 (5.1) | 17 (16.0) | 18 (16.1) | 0.4 | 0.09 | 0.9 | 0.9 | 0.08 | 0.08 | 0.9 |
| Maternal medical complication | 43 (35.5) | 12 (30.8) | 37 (34.9) | 36 (32.1) | 0.9 | 0.6 | 0.9 | 0.6 | 0.6 | 0.9 | 0.7 |
| Infection | 43 (35.5) | 13 (33.3) | 35 (33.0) | 43 (38.4) | 0.9 | 0.8 | 0.7 | 0.7 | 0.9 | 0.6 | 0.4 |
All values given as mean ± standard deviation (continuous measures) or number (percentage of cluster) (categorical measures). P-values derived from Kruskall-Wallis or Chi-Square tests (continuous and categorical, respectively) for omnibus tests and Mann-Whitney and Chi-Square tests for post-hoc comparisons. P-values < 0.05 are considered significant (bolded).
Acknowledgements:
This manuscript is based on material previously presented at the Society for Pediatric Pathology Fall 2022 Meeting, held on October 7-9, 2022, in Rochester, NY.
Funding details:
This work was supported by grant funding from the Eunice Kennedy Shriver National Institute of Child Health and Human Development: U10-HD045953 Brown University, Rhode Island; U10-HD045925 Emory University, Georgia; U10-HD045952 University of Texas Medical Branch at Galveston, Texas; U10-HDO45955 University of Texas Health Sciences Center at San Antonio, Texas; U10-HD045944 University of Utah Health Sciences Center, Utah; and U01-HD045954 RTI International, RTP. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.
Footnotes
Conflict of interest: The authors have no financial or other conflicts of interest to disclose.
References
- 1.Hug L, You D, Blencowe H, Mishra A, Wang Z, Fix MJ, Wakefield J, Moran AC, Gaigbe-Togbe V, Suzuki E, et al. Global, regional, and national estimates and trends in stillbirths from 2000 to 2019: a systematic assessment. The Lancet [Internet]. 2021. [cited 2023 Feb 6];398:772–785. doi: 10.1016/S0140-6736(21)01112-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Marchetti D, Belviso M, Fulcheri E. A case of stillbirth: the importance of placental investigation in medico-legal practice. Am J Forensic Med Pathol [Internet]. 2009. [cited 2023 Feb 7];30:64–68. doi: 10.1097/PAF.0B013E318187387E. [DOI] [PubMed] [Google Scholar]
- 3.Glaberson W After Stillbirth, Courts Try to Put a Price on a Mother’s Anguish. Academic Medical Professionals Insurance Exchange Risk Retention Group. 2011. [Google Scholar]
- 4.Sawicki NN. A Malpractice-Based Duty to Disclose the Risk of Stillbirth: A Response to Lens. Iowa Law Review Online [Internet]. 2021. [cited 2023 Feb 7];106. [Google Scholar]
- 5.Fretts RC. Etiology and prevention of stillbirth. Am J Obstet Gynecol. 2005;193:1923–1935. doi: doi: 10.1016/j.ajog.2005.03.074. [DOI] [PubMed] [Google Scholar]
- 6.Heazell AEP, Frøen JF. Methods of fetal movement counting and the detection of fetal compromise. https://doi.org/101080/01443610801912618 [Internet]. 2009. [cited 2023 Feb 7];28:147–154. doi: 10.1080/01443610801912618. [DOI] [PubMed] [Google Scholar]
- 7.Paternoster M, Perrino M, Travaglino A, Raffone A, Saccone G, Zullo F, D’Armiento FP, Buccelli C, Niola M, D’Armiento M. Parameters for estimating the time of death at perinatal autopsy of stillborn fetuses: a systematic review. Int J Legal Med [Internet]. 2019. [cited 2023 Feb 7];133:483–489. doi: 10.1007/S00414-019-01999-1/TABLES/3. [DOI] [PubMed] [Google Scholar]
- 8.Conway DL, Hansen NI, Dudley DJ, Parker CB, Reddy UM, Silver RM, Bukowsk R, Pinar H, Stoll BJ, Varner MW, et al. An algorithm for the estimation of gestational Age at the time of fetal death. Paediatr Perinat Epidemiol. 2013;27:145–157. doi: 10.1111/ppe.12037. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Parker CB, Hogue CJR, Koch MA, Willinger M, Reddy UM, Thorsten VR, Dudley DJ, Silver RM, Coustan D, Saade GR, et al. Stillbirth Collaborative Research Network: design, methods and recruitment experience. Paediatr Perinat Epidemiol [Internet]. 2011. [cited 2022 Jan 19];25:425–435. doi: 10.1111/J.1365-3016.2011.01218.X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Pinar H, Koch M, Hawkins H, Heim-Hall J, Abramowsky C, Thorsten V, Carpenter M, Zhou H, Reddy U. The stillbirth collaborative research network postmortem examination protocol. Am J Perinatol [Internet]. 2012. [cited 2022 Jan 19];29:187–202. doi: 10.1055/S-0031-1284228. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Pinar H, Koch MA, Hawkins H, Heim-Hall J, Shehata B, Thorsten VR, Carpenter M, Lowichik A, Reddy UM. The Stillbirth Collaborative Research Network (SCRN) Placental and Umbilical Cord Examination Protocol. Am J Perinatol [Internet]. 2011. [cited 2022 Jan 19];28:781. doi: 10.1055/S-0031-1281509. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Dudley DJ, Goldenberg R, Conway D, Silver RM, Saade GR, Varner MW, Pinar H, Coustan D, Bukowski R, Stoll B, et al. A new system for determining the causes of stillbirth. Obstetrics and gynecology [Internet]. 2010. [cited 2022 Feb 6];116:254–260. doi: 10.1097/AOG.0B013E3181E7D975. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Genest DR. Estimating the time of death in stillborn fetuses: II. Histologic evaluation of the placenta; a study of 71 stillborns. Obstetrics & Gynecology. 1992;80:585–592. [PubMed] [Google Scholar]
- 14.Genest DR, Singer DB. Estimating the time of death in stillborn fetuses: III. External fetal examination; a study of 86 stillborns. Obstetrics & Gynecology. 1992;80:593–600. [PubMed] [Google Scholar]
- 15.Genest DR, Williams MA, Greene MF. Estimating the time of death in stillborn fetuses: I. Histologic evaluation of fetal organs; an autopsy study of 150 stillborns. Obstetrics & Gynecology. 1992;80:575–584. [PubMed] [Google Scholar]
- 16.Köhn H-F, Hubert LJ. Hierarchical Cluster Analysis. Wiley StatsRef: Statistics Reference Online [Internet]. 2015. [cited 2023 Feb 5];1–13. doi: 10.1002/9781118445112.STAT02449.PUB2. [DOI] [Google Scholar]
- 17.Patnaik AK, Bhuyan PK, Krishna Rao K v. Divisive Analysis (DIANA) of hierarchical clustering and GPS data for level of service criteria of urban streets. Alexandria Engineering Journal. 2016;55:407–418. doi: 10.1016/J.AEJ.2015.11.003. [DOI] [Google Scholar]
- 18.Müllner D Modern hierarchical, agglomerative clustering algorithms. 2011. [cited 2023 Feb 5]; doi: 10.48550/arxiv.1109.2378. [DOI] [Google Scholar]
- 19.Shi C, Wei B, Wei S, Wang W, Liu H, Liu J. A quantitative discriminant method of elbow point for the optimal number of clusters in clustering algorithm. EURASIP J Wirel Commun Netw [Internet]. 2021. [cited 2021 May 12];2021:1–16. doi: 10.1186/s13638-021-01910-w. [DOI] [Google Scholar]
- 20.Marutho D, Hendra Handaka S, Wijaya E, Muljono. The Determination of Cluster Number at k-Mean Using Elbow Method and Purity Evaluation on Headline News. Proceedings - 2018 International Seminar on Application for Technology of Information and Communication: Creative Technology for Human Life, iSemantic 2018. Institute of Electrical and Electronics Engineers Inc.; 2018. p. 533–538. [Google Scholar]
- 21.Jacques SM, Qureshi F, Johnson A, Alkatib AA, Kmak DC. Estimation of time of fetal death in the second trimester by placental histopathological examination. Pediatr Dev Pathol [Internet]. 2003. [cited 2023 Feb 5];6:226–232. doi: 10.1007/S10024-003-8089-9. [DOI] [PubMed] [Google Scholar]
- 22.Roberts AB, Mitchell J, Murphy C, Koyaa H, Cundy T. Fetal liver length in diabetic pregnancy. Am J Obstet Gynecol [Internet]. 1994. [cited 2023 Apr 23];170:1308–1312. doi: 10.1016/S0002-9378(94)70147-4. [DOI] [PubMed] [Google Scholar]
- 23.Jensen KK, Oh K, Patel N, Narasimhan ER, Ku AS, Sohaey R. Fetal hepatomegaly: Causes and associations. Radiographics [Internet]. 2020. [cited 2023 Apr 23];40:589–604. doi: 10.1148/RG.2020190114/ASSET/IMAGES/LARGE/RG.2020190114.FIG11B.JPEG. [DOI] [PubMed] [Google Scholar]
- 24.Redline RW, Ravishankar S. Fetal vascular malperfusion, an update. APMIS [Internet]. 2018. [cited 2021 Jul 22];126:561–569. doi: 10.1111/APM.12849. [DOI] [PubMed] [Google Scholar]
- 25.Madea B Methods for determining time of death. Forensic Sci Med Pathol [Internet]. 2016. [cited 2023 Feb 10];12:451–485. doi: 10.1007/S12024-016-9776-Y/TABLES/15. [DOI] [PubMed] [Google Scholar]