Identifying Hemolytic Disease of the Fetus and Newborn within a Large Integrated Health Care System

Abstract

Objective

This study aims to identify hemolytic disease of the fetus and newborn (HDFN) pregnancies using electronic health records (EHRs) from a large integrated health care system.

Study Design

A retrospective cohort study was performed among pregnant patients receiving obstetrical care at Kaiser Permanente Southern California health care system between January 1, 2008, and June 30, 2022. Using structured (diagnostic/procedural codes, medication, and laboratory records) and unstructured (clinical notes analyzed via natural language processing) data abstracted from EHRs, we extracted HDFN-specific “indicators” (maternal positive antibody test and abnormal antibody titer, maternal/infant HDFN diagnosis and blood transfusion, hydrops fetalis, infant intravenous immunoglobulin [IVIG] treatment, jaundice/phototherapy, and first administrated Rho[D] Immune Globulin) to identify potential HDFN pregnancies. Chart reviews and adjudication were then performed on select combinations of indicators for case ascertainment. HDFN due to ABO alloimmunization alone was excluded. The HDFN frequency and proportion of each combination were fully analyzed.

Results

Among the 464,711 eligible pregnancies, a total of 136 pregnancies were confirmed as HDFN pregnancies. The percentage of the HDFN-specific indicators ranged from 0.02% (infant IVIG treatment) to 34.53% (infant jaundice/phototherapy) among the eligible pregnancies, and 32.35% (infant IVIG treatment) to 100% (maternal positive antibody test) among the 136 confirmed HDFN pregnancies. Four combination groups of four indicators, four combination groups of five indicators, and the unique combination of six indicators showed 100% of HDFN pregnancies, while 80.88% of confirmed HDFN pregnancies had the indicator combination of maternal positive antibody test, maternal/infant HDFN diagnosis, and infant jaundice/phototherapy.

Conclusion

We successfully identified HDFN pregnancies by leveraging a combination of medical indicators extracted from structured and unstructured data that may be used in future pharmacoepidemiologic studies. Traditional indicators (positive antibody test results, high titers, and clinical diagnosis codes) alone did not accurately identify HDFN pregnancies, highlighting an unmet need for improved practices in HDFN coding.

Key Points

• A case ascertainment method was developed to identify HDFN from structured and unstructured data.

• The method used in this study may be used in future pharmacoepidemiologic studies.

• The study highlighted an unmet need for improved practices in HDFN coding.

Keywords

hemolytic disease of the fetus and newborn - antibody test - titer - transfusion - intravenous immunoglobulin - jaundice - phototherapy - electronic health record - natural language processing

Antenatal and perinatal morbidity and mortality caused by rare diseases present important challenges to the health care system around the world.[1] Hemolytic disease of the fetus and newborn (HDFN), also known as alloimmunization HDFN or erythroblastosis fetalis/neonatorum, is a rare red blood cell disorder caused by maternal–fetal red cell antigen incompatibility during pregnancy. It is the most common cause of severe hemolytic complications in the fetus and neonate.[2] [3] Several antibodies, including anti-D, c, C, E, and Kell alloimmunization, are known to cause HDFN; however, anti-D and Kell alloimmunization are most commonly associated with severe HDFN.[4]

HDFN occurs in approximately 3 to 80 per 100,000 pregnancies in the United States annually.[3] [5] Within the last five decades, the incidence of HDFN has dramatically declined due to the availability of prophylactic immunoglobulins and the establishment of antibody screening programs.[6] [7] However, HDFN remains a common cause of morbidity and mortality in many developing countries without routine screening or alloimmunization prevention programs.[8] [9] The severity of HDFN varies and can range from mild immune reactions to severe fetal anemia, hydrops, or death.[4] [10] [11] Infants with HDFN can exhibit many issues, from hyperbilirubinemia to cholestasis and kernicterus, requiring specific management according to their postnatal complications.[12] [13] Clinically, HDFN can be detected through fetal anemia assessments performed via ultrasound evaluation of the middle cerebral artery peak systolic velocity.[14] If there are clinically significant indications of anemia or other issues, the fetus is often treated by intrauterine transfusion (IUT).[15] Intravenous immunoglobulin (IVIG) is administered in select cases to delay the gestational age at first IUT.[16]

While the prevalence of HDFN in fetuses and newborns has decreased,[6] [7] [17] better clarification of the epidemiology and postnatal complications of HDFN is still needed.[5] [18] [19] [20] The first critical step required is to effectively identify HDFN-affected pregnancies from administrative data or electronic health records (EHRs). However, effective and accurate ascertainment of HDFN cases is still very challenging because there is no single unique data structure to identify HDFN. Instead, a set of clinical information embedded in both structured and unstructured (clinical notes) data needs to be considered together for HDFN case ascertainment. Extracting information from clinical notes requires specific techniques for data processing and analyses. Natural language processing (NLP), a field of computer-based methods aimed at standardizing and analyzing free text, processes unstructured data through information extraction residing in natural language and semantic representation learning for information retrieval, classifications, and predictions.[21] The NLP techniques have been successfully used to extract or classify various medical conditions.[22] [23] [24] The purpose of the study is to develop an effective algorithm and process to accurately identify HDFN pregnancies from EHRs or claims records for clinical and pharmacoepidemiologic research.

Materials and Methods

Kaiser Permanente Southern California (KPSC) is a large integrated health care system providing comprehensive medical services to over 4.8 million racially and ethnically diverse members across 15 large medical centers area with more than 250 medical offices scattered throughout the Southern California region.[25] KPSC health care staff in both outpatient and inpatient clinical settings utilize an EHR based on an Epic® platform, KP Health Connect (KPHC), that is accessible to multiple health care providers at the same time. KPHC is a highly sophisticated integrated health information and care management system designed to enhance the quality of patient care. The data are collected in real time and can be instantly accessed by authorized clinicians and researchers. KPHC contains individual-level structured data (including diagnosis codes, procedure codes, medications, immunization records, laboratory test results, and pregnancy episodes and outcomes) and unstructured data (including free-text clinical notes, radiology and pathology reports, as well as imaging and videos) covering all medical visits across all health care settings (i.e., outpatient, inpatient, emergency department, virtual encounters, etc.). Clinical care of KPSC members provided by external contracted providers is captured in the EHR through reimbursement claim requests. The study protocol was reviewed and approved by the KPSC Institutional Review Board with a waiver of the requirement for informed consent. Only authorized persons were given access and permission to perform all analyses.

We conducted a retrospective cohort study among pregnant patients receiving obstetrical care at KPSC from January 1, 2008, to June 30, 2022. Pregnancies without membership, who underwent an elective abortion for non-medical reasons and lacked complete information due to the procedure being performed at an outside KPSC facility, or HDFN due to ABO alloimmunization alone were excluded from further analysis. However, abortions performed for medical reasons were required to be performed at KPSC facilities, and as such, were included in the study analysis. Our strategy to accurately identify HDFN among these pregnancies consisted of an iterative investigatory approach that utilized both structured (diagnosis codes, laboratory test results, and procedure codes or treatments) and unstructured (clinical notes) data extracted from the KPSC EHR system, to narrow down to potential HDFN pregnancies, and then ascertain true cases by full manual chart reviews and adjudication using a sampling approach as summarized in [Table 1].

Table 1 Chart review samples of pregnancies and selected criteria for each phase used for identifying hemolytic disease of the fetus and newborn

Phase	Sample selection criteria	Sample size	Confirmed HDFN cases
Pre-phase	• Unspecified HDFN-related diagnosis or DRG codes  ▪ ICD-9: 773.2  ▪ ICD-10: P55.8, P55.9, O36.0990, O36.1990, O36.20 × 0, R76.0, O36.82  ▪ DRG: 817, 818, or 819, but no alloimmunization code and intrauterine transfusion code  ▪ Rho(D) Immune Globulin administration	81	5
Phase 1	• Without a positive antibody test and/or abnormal titer • At least one indication of HDFN codes, transfusion, IVIG, hydrops fetalis, jaundice/phototherapy	51	9
Phase 2	• With a positive antibody test result and abnormal titer • At least one indication of HDFN code, transfusion, IVIG, hydrops fetalis, jaundice/phototherapy	207	91
Phase 3	• Infant had a positive antibody laboratory test result regardless of other indicators • Mother had a positive antibody laboratory test result, indicator of jaundice/phototherapy, and at least one HDFN code, transfusion, and IVIG • The mother had a positive antibody laboratory test result and fetal death or neonatal death • The mother had a positive antibody laboratory test result and any two abnormal HGB, HCT, and reticulocyte laboratory results	277	23
Phase 4	• The combination group (sample >5) with confirmed HDFN for selected review samples in phase 3	91	8
Overall		707	136

Abbreviations: DRG, diagnosis-related group; HCT, hematocrit; HDFN, hemolytic disease of the fetus and newborn; HGB, hemoglobin; ICD, International Classification of Diseases; IVIG, intravenous immunoglobulin.

Selection of Potential Structured Codes for Hemolytic Disease of the Fetus and Newborn

Our initial strategy was to determine potential structured codes that could be used to identify HDFN. A set of relevant diagnosis codes or diagnosis-related group (DRG) codes was assembled by the study team, as summarized in [Supplementary Table S1] (available in the online version only). Specific HDFN codes were used as indicators for potential HDFN. A chart review was not performed to confirm the proper use of these codes at this stage. For less-specific HDFN codes, a sample of up to five cases (randomly selected if there were more than five cases) for each unspecified HDFN code was selected for manual chart review by trained abstractors and adjudicated by the study's maternal–fetal medicine specialist, Dr. Michael J. Fassett. Codes that had no confirmed HDFN case upon chart review were excluded. In addition, we randomly selected 10 pregnancies with Rho(D) Immune Globulin administration and alloimmunization (five from the mother and five from the infant) for full chart review. Investigation of these cases showed the importance of the timing of Rho(D) Immune Globulin administration. Therefore, Rho(D) Immune Globulin injection was leveraged as an indicator to determine/exclude HDFN.

Identification of Hemolytic Disease of the Fetus and Newborn Indicators

With guidance from the study's maternal–fetal medicine specialist, we assembled evidence including maternal antibody tests, maternal antibody titers, HDFN diagnosis codes, IUT, blood transfusion, hyperbilirubinemia, jaundice, phototherapy, etc., to identify potential HDFN cases. We then built a comprehensive data process using combinations of HDFN indicators to identify potential HDFN pregnancies for full chart review and case ascertainment. We first utilized structured data in the KPSC EHR system as follows:

1. Antibody test flag: Indirect Coombs laboratory test results performed during the pregnancy window were identified for each pregnancy based on the laboratory procedure codes listed in [Supplementary Table S2] (available in the online version only). The flag was deemed as 1 if at least one of the indirect Coombs results was positive, otherwise, it defaulted to zero.

2. Abnormal titer flag: The titer results of several antibodies of interest listed in [Supplementary Table S3] (available in the online version only) were identified. The flag was defined as 1 for titers above the critical level (anti-Kell ≥4; other antibodies ≥8), otherwise, it was set as zero.

3. HDFN diagnosis code flag: If at least one HDFN-related diagnosis code listed in [Supplementary Table S1] (available in the online version only; as described in the above subsection) was used during the pregnancy window, the patient was flagged with a value of 1, otherwise, it was set as zero.

4. Transfusion flag: If the mother or infant had at least one transfusion performed (based on the procedure codes listed in [Supplementary Table S4] [available in the online version only]) during the pregnancy window for each pregnancy, the patient was flagged as 1; otherwise, it defaulted to zero.

5. Combined jaundice/phototherapy flag: If an infant had at least one code for hyperbilirubinemia, jaundice, or phototherapy based on diagnosis or procedure codes listed in [Supplementary Table S5] (available in the online version only), it was flagged as 1; otherwise, it was set as zero.

6. Hydrops flag: If at least one code for hydrops fetalis (based on the diagnosis codes listed in [Supplementary Table S6] [available in the online version only]) was found during the pregnancy window, the patient was flagged as 1, otherwise, it was set as zero.

7. First Rho(D) Immune Globulin injection flag: If the first Rho(D) Immune Globulin injection was 3 days prior to delivery, it was coded as 1 and coded as 2 if it was within 3 days of the delivery, otherwise, it was coded as zero.

In addition, we utilized NLP to search for evidence of these indicators from the clinical notes to supplement the information from structured data through the following process:

1. Limit the NLP process to pregnancies with a positive indirect Coombs test. This significantly reduced the volume of clinical notes for processing.

2. Extract clinical notes associated with encounters that occurred in the pregnancy window among pregnancies with a positive indirect Coombs test.

3. Search for terms of antibodies listed in [Supplementary Table S3] (available in the online version only) and retrieve the corresponding titer results from the clinical notes associated with each pregnancy. The retrieved results were also used to determine the abnormal titer flag to supplement the structured data results.

4. Identify the following evidence from the clinical notes associated with each pregnancy (both mother and infant). Any negated, general, historical, and uncertain descriptions were excluded.

   • Positive description of HDFN terms in [Supplementary Table S7] (available in the online version only). The identified positive terms were used to supplement the HDFN diagnosis code flag.

   • Positive description of IUT or blood transfusion terms in [Supplementary Table S7] (available in the online version only). The identified positive terms were used to supplement the transfusion flag.

   • Positive description of hyperbilirubinemia or phototherapy terms in [Supplementary Table S7] (available in the online version only). The identified positive terms were used to supplement the combined jaundice/phototherapy flag.

   • Positive description of hydrops terms in [Supplementary Table S7] (available in the online version only). The identified positive terms were used to supplement the hydrops flag.

   • Positive description of IVIG terms in [Supplementary Table S7] (available in the online version only). The identified positive terms were used to create an IVIG treatment flag.

Chart Review Selection for Hemolytic Disease of the Fetus and Newborn Case Ascertainment

Since HDFN is a rare medical condition, we narrowed down the study cohort by using combinations of select indicators to identify patients who were highly likely to be HDFN candidates for case ascertainment through multiple phases of chart reviews. The selection of chart review cases for each phase was dependent upon the findings in previous phases. According to the chart-reviewed findings for potential HDFN diagnosis and clinical guidance, a confirmed HDFN search should first check whether the mother had a positive antibody test and the corresponding titer results.

Pre-phase: The medical records of 81 pregnancies were manually chart-reviewed and adjudicated as described in the above subsection of HDFN potential code selection ([Table 1]). These charts reviewed pregnancies through a combination of the eight indicators that were used to guide the next phase of sample selection.

Phase 1: Per the HDFN chart review findings during the pre-phase and clinical guidance during adjudication, confirmed HDFN required the mother to have at least one positive antibody test and abnormal titer results. A set of 51 pregnancies without positive antibody test and/or abnormal titer but with at least one indication of an HDFN diagnosis code, transfusion, IVIG, hydrops fetalis, and/or jaundice/phototherapy was selected for full chart review and adjudication ([Table 1]). The chart review findings were used to determine the inclusion/exclusion of these indicator combinations.

Phase 2: This phase was focused on the high likelihood of HDFN pregnancy candidates with a positive antibody test result and abnormal titer. Per the pre-phase and phase 1 findings, the assumption was that a patient would need to have at least a positive antibody test result and abnormal titer with at least another HDFN indicator to be a true HDFN case. The more flagged indicators, the higher the chance of true HDFN cases. Indicators used were positive antibody tests and abnormal antibody titer results where the pregnancy had additional indicators, including an HDFN code (mother or baby), transfusion, IVIG, hydrops fetalis, and/or jaundice/phototherapy. A total of 207 were sampled, and all were chart-reviewed and adjudicated ([Table 1]).

Phase 3: Based on the results of phase 2 and the guidance of the study's maternal–fetal medicine specialist, we selected a sample of patients with at least one of the following criteria for full chart review and adjudication from the remaining unreviewed pregnancy records.

• The infant had a positive antibody laboratory test result regardless of other indicators.

• The mother had a positive antibody laboratory test result, an indicator of jaundice/phototherapy, and at least one HDFN code, transfusion, and IVIG.

• The mother had a positive antibody laboratory test result and fetal death or neonatal death.

• The mother had a positive antibody laboratory test result and two abnormal hemoglobin (HGB), hematocrit (HCT), and reticulocyte laboratory results.

If the number within the combination group was over 5, then five cases were randomly selected from the corresponding group. With these above conditions, a total of 277 pregnancy records were pulled for full chart review and adjudication ([Table 1]).

Phase 4: This was the continuation of phase 3 for combination groups (sample >5), where a random sample of cases was reviewed, and at least one true HDFN case was found. All pregnancies within that group were fully chart reviewed and adjudicated (a total of 91 pregnancies; [Table 1]).

Chart Review and Adjudication Process

For each pregnancy selected to determine HDFN status, relevant notes from the infant's chart and the mother's EHR were thoroughly reviewed by trained abstractors. In the infant's chart, blood test results such as ABO typing, direct antibody test, HGB level, HCT level, reticulocyte count, and total bilirubin level were recorded. In addition to laboratory results, any mention of phototherapy, jaundice, anemia, blood transfusion, and/or HDFN diagnosis found in the history and physical exam notes, OB-GYN office visit notes, inpatient progress notes, discharge summaries, and well-baby visits were recorded. In the mother's EHR, blood test results such as ABO typing, direct Coombs test results, antibody screenings, and titer results, as well as any mention of alloimmunization, middle cerebral artery (MCA) peak systolic velocity findings, Rho(D) Immune Globulin administrations, and/or IUT, were recorded. After compiling notes from both the mother's and infant's EHR, the HDFN status of the infant was determined on a case-by-case basis, and the details were summarized in [Supplementary Table S8] (available in the online version only).

Results

A total of 464,711 pregnancies among the initial 572,328 pregnancies were eligible for the study after the exclusion criteria were applied ([Fig. 1]). Of these eligible pregnancies, 136 pregnancies (29.3 per 100,000 pregnancies) were confirmed as HDFN cases by chart review and adjudication ([Table 1]). The distribution of these indicators among the eligible pregnancies and confirmed HDFN pregnancies was summarized in [Table 2]. Of the 464,711 pregnancies, the indicators for maternal positive antibody test (2.62%), maternal abnormal titer (0.08%), maternal/infant HDFN diagnosis (0.46%), maternal/infant transfusion treatment (0.66%), infant IVIG treatment (0.02%), and infant jaundice/phototherapy (34.53%) were identified. The rate of HDFN pregnancies among each indicator was 1.12%, 26.63%, 5.30%, 2.01%, 61.11%, and 0.08% for maternal positive antibody test, maternal abnormal titer, maternal/infant HDFN diagnosis, maternal/infant transfusion treatment, infant IVIG treatment, and infant jaundice/phototherapy, respectively, while the rate of these indicators among the confirmed HDFN pregnancies was 100% for maternal positive antibody test, 75.00% for maternal abnormal titer, 83.82% for maternal/infant HDFN diagnosis, 45.59% for maternal/infant transfusion treatment, 32.35% for infant IVIG treatment, and 95.59% for infant jaundice/phototherapy.

Fig. 1 Flowchart of hemolytic disease of the fetus and newborn (HDFN) pregnancies.

Table 2 The rate of confirmed hemolytic disease of the fetus and newborn pregnancies by indication of positive antibody test, abnormal titer, hemolytic disease of the fetus and newborn diagnosis codes, transfusion, intravenous immunoglobulin, jaundice/phototherapy

Indicators[a]	Total pregnancies (%)[b]	Confirmed HDFN pregnancies	HDFN rate (%)	Percentage of HDFN pregnancies among all confirmed HDFN pregnancies
One indicator
A	12,157 (2.62)	136	1.12	100.00
B	383 (0.08)	102	26.63	75.00
C	2,152 (0.46)	114	5.30	83.82
D	3,086 (0.66)	62	2.01	45.59
E	72 (0.02)	44	61.11	32.35
F	160,456 (34.53)	130	0.08	95.59
Two indicators
A, B	383 (0.08)	102	26.63	75.00
A, C	971 (0.21)	114	11.74	83.82
A, D	330 (0.07)	62	18.79	45.59
A, E	72 (0.02)	44	61.11	32.35
A, F	5,027 (1.08)	130	2.59	95.59
B, C	162 (0.03)	87	53.70	63.97
B, D	73 (0.02)	50	68.49	36.76
B, E	35 (0.01)	34	97.14	25.00
B, F	216 (0.05)	97	44.91	71.32
C, D	85 (0.02)	51	60.00	37.50
C, E	43 (0.01)	36	83.72	26.47
C, F	1,064 (0.23)	110	10.34	80.88
D, E	47 (0.01)	31	65.96	22.79
D, F	2,683 (0.58)	57	2.12	41.91
E, F	68 (0.01)	44	64.71	32.35
Three indicators
A, B, C	162 (0.03)	87	53.70	63.97
A, B, D	73 (0.02)	50	68.49	36.76
A, B, E	35 (0.01)	34	97.14	25.00
A, B, F	216 (0.05)	97	44.91	71.32
A, C, D	76 (0.02)	51	67.11	37.50
A, C, E	43 (0.01)	36	83.72	26.47
A, C, F	507 (0.11)	110	21.70	80.88
A, D, E	47 (0.01)	31	65.96	22.79
A, D, F	253 (0.05)	57	22.53	41.91
A, E, F	68 (0.01)	44	65.71	32.35
B, C, D	47 (0.01)	43	91.49	31.62
B, C, E	30 (0.01)	30	100.00	22.06
B, C, F	115 (0.02)	84	73.04	61.76
B, D, E	24 (0.01)	24	100.00	17.65
B, D, F	54 (0.01)	45	83.33	33.09
B, E, F	35 (0.01)	34	97.14	25.00
C, D, E	29 (0.01)	26	89.66	19.12
C, D, F	72 (0.02)	48	66.67	35.29
C, E, F	42 (0.01)	36	85.71	26.71
D, E, F	46 (0.01)	31	67.39	22.79
Four indicators
A, B, C, D	47 (0.01)	43	91.49	31.62
A, B, C, E	30 (0.01)	30	100.00	22.06
A, B, C, F	115 (0.02)	84	73.04	61.76
A, B, D, E	24 (0.01)	24	100.00	17.65
A, B, D, F	54 (0.01)	45	83.33	33.09
A, B, E, F	35 (0.01)	34	97.14	25.00
A, C, D, E	29 (0.01)	26	89.66	19.12
A, C, D, F	63 (0.01)	48	76.19	35.29
A, D, E, F	42 (0.01)	36	85.71	26.47
B, C, D, E	22 (0.00)	22	100.00	16.18
B, C, D, F	42 (0.01)	40	95.24	29.41
B, C, E, F	30 (0.01)	30	100.00	22.06
B, D, E, F	24 (0.01)	24	100.00	17.65
C, D, E, F	29 (0.01)	26	89.66	19.12
Five indicators
A, B, C, D, E	22 (0.00)	22	100.00	16.18
A, B, C, D, F	42 (0.01)	40	95.24	29.41
A, B, C, E, F	30 (0.01)	30	100.00	22.06
A, B, D, E, F	24 (0.01)	24	100.00	17.65
A, C, D, E, F	29 (0.01)	26	89.66	19.12
B, C, D, E, F	22 (0.00)	22	100.00	16.18
Six indicators
A, B, C, D, E, F	22 (0.00)	22	100.00	16.18

Abbreviation: HDFN, hemolytic disease of the fetus and newborn; IVIG, intravenous immunoglobulin.

^a A, positive maternal antibody test; B, abnormal maternal titer; C, maternal/infant HDFN diagnosis; D, maternal/infant transfusion; E, infant IVIG treatment; F, infant jaundice/phototherapy.

^b The percentage among the total study cohort pregnancies.

For combinations with multiple indicators, the more indicators that were flagged, the higher the chance of corresponding pregnancies being HDFN cases. The highest rate of HDFN pregnancies with two combined indicators numerically calculated was maternal positive antibody test and infant jaundice/phototherapy (97.14%), while the combination of maternal positive antibody test and infant jaundice/phototherapy had the highest rate (95.59%) among the confirmed HDFN pregnancies. Among the three indicator combinations, the combination of maternal abnormal titer, maternal/infant HDFN diagnosis, and infant IVIG treatment, and the combination of maternal abnormal titer, maternal/infant transfusion treatment, and infant IVIG treatment numerically calculated were 100% HDFN pregnancies while the combination of maternal positive antibody test, maternal/infant HDFN diagnosis and infant jaundice/phototherapy had the highest rate among the confirmed HDFN pregnancies (80.88%). For the combinations of more than three indicators, four combinations of four indicators, four combinations of five indicators, and the unique combination of six indicators were 100% HDFN pregnancies ([Table 2]). However, the corresponding rate of the combination with more than three indicators among the confirmed HDFN pregnancies was significantly reduced to a lower value. The detailed distribution of total pregnancies with or without confirmed HDFN pregnancies by the values of eight indicators was presented in [Supplementary Tables S9] and [S10] [available in the online version only], respectively.

Discussion

This study represents the first HDFN case ascertainment approach that combined structured and unstructured data from inpatient, outpatient, and emergency services as well as pharmacy, laboratory, and radiology databases. Using a large integrated health care system, our study developed an algorithm and process to effectively determine a set of clinical indicators including maternal positive antibody test, maternal abnormal antibody titer, maternal/infant HDFN diagnosis, maternal/infant blood transfusion, infant IVIG treatment, infant jaundice/phototherapy in identifying potential HDFN pregnancies and then ascertained through an extensive chart review and adjudication process. A total of 136 HDFN pregnancies among nearly a half million pregnancies within the KPSC system were successfully identified through this process. The HDFN rate was 29.3 per 100,000 pregnancies, which is consistent with and within previously reported prevalence rate ranges.[3] [5]

When testing the rate of HDFN pregnancies corresponding to specific indicators including maternal positive antibody test and abnormal titer, maternal/infant HDFN diagnosis and transfusion treatment, infant IVIG treatment as well as infant jaundice/phototherapy, we observed highly variable HDFN rates across indicators, with higher rates for maternal positive antibody test, infant jaundice/phototherapy, and maternal/infant HDFN diagnosis followed by maternal abnormal titer, maternal/infant transfusion treatment, and infant IVIG treatment.

Due to being a rare medical condition, accurate identification of HDFN among a large volume of pregnant population from EHRs poses a great challenge and requires comprehensive data abstraction, chart review, and adjudication by clinical staff. The HDFN identification based on a single clinical information, such as a diagnosis code, could lead to either inaccurate or incomplete determination. For example, only 5.30% of HDFN diagnoses were ascertained as true HDFN, and 16.18% of HDFN cases did not have an HDFN diagnosis. Therefore, the appropriate approach was to leverage the combination of multiple HDFN-related diagnoses, treatments, laboratory assessments, etc., to capture these potential HDFN cases; then, these identified cases were further ascertained and confirmed via a comprehensive chart review and adjudication process. Combining multiple indicators from both structured and unstructured medical records significantly improved true HDFN case ascertainment, reducing potential biases.

Our study acknowledges several limitations. First, the completeness and accuracy of the extracted HDFN-related indicators depended on the information documented in the EHRs. Incomplete documentation of information could lead to bias. Second, the chart review process was more focused on HDFN case ascertainment rather than the confirmation of identifying each indicator. Third, HDFN pregnancies were identified by a computerized algorithm-assisted chart review process rather than a data automation process. The process first selected these highly likely HDFN pregnancies from the entire study pregnancy cohort, then was fully chart reviewed to determine the corresponding status of HDFN. Lastly, although case selections from these indicator combinations were driven by an iterative chart review findings approach to minimize the omission of HDFN cases, the developed process cannot guarantee the inclusion of all HDFN pregnancies within the study cohort. However, the number of missed cases should be minimal, if any.

In conclusion, we successfully identified HDFN pregnancies by leveraging the combination of medical indicators extracted from structured and unstructured data that may be used in future pharmacoepidemiologic studies. Traditional indicators (positive antibody test results, high titers, and clinical diagnosis codes) alone did not accurately identify HDFN pregnancies, highlighting an unmet need for improved practices in HDFN coding.

Conflict of Interest

None declared.

Acknowledgments

Authors would like to thank Evo Alemao for his scientific insights and Mary Malek for her assistance in the manual chart reviews of the EHRs. The authors thank the patients of KPSC for helping to improve care through the use of information collected through our EHR systems.

• References

• 1 Pegoraro V, Urbinati D, Visser GHA. et al. Hemolytic disease of the fetus and newborn due to Rh(D) incompatibility: A preventable disease that still produces significant morbidity and mortality in children. PLoS One 2020; 15 (07) e0235807
• 2 Jackson ME, Baker JM. Hemolytic disease of the fetus and newborn: historical and current state. Clin Lab Med 2021; 41 (01) 133-151
• 3 Delaney M, Matthews DC. Hemolytic disease of the fetus and newborn: managing the mother, fetus, and newborn. Hematology (Am Soc Hematol Educ Program) 2015; 2015: 146-151
• 4 Lee BK, Ploner A, Zhang Z, Gryfelt G, Wikman A, Reilly M. Constructing a population-based research database from routine maternal screening records: a resource for studying alloimmunization in pregnant women. PLoS One 2011; 6 (11) e27619
• 5 Geaghan SM. Diagnostic laboratory technologies for the fetus and neonate with isoimmunization. Semin Perinatol 2011; 35 (03) 148-154
• 6 Fung Kee Fung K, Eason E, Crane J. et al; Maternal-Fetal Medicine Committee, Genetics Committee. Prevention of Rh alloimmunization. J Obstet Gynaecol Can 2003; 25 (09) 765-773
• 7 Aitken SL, Tichy EM. Rh(O)D immune globulin products for prevention of alloimmunization during pregnancy. Am J Health Syst Pharm 2015; 72 (04) 267-276
• 8 Osaro E, Charles AT. Rh isoimmunization in Sub-Saharan Africa indicates need for universal access to anti-RhD immunoglobulin and effective management of D-negative pregnancies. Int J Womens Health 2010; 2: 429-437
• 9 Ayenew AA. Prevalence of rhesus D-negative blood type and the challenges of rhesus D immunoprophylaxis among obstetric population in Ethiopia: a systematic review and meta-analysis. Matern Health Neonatol Perinatol 2021; 7 (01) 8
• 10 de Haas M, Thurik FF, Koelewijn JM, van der Schoot CE. Haemolytic disease of the fetus and newborn. Vox Sang 2015; 109 (02) 99-113
• 11 Nassar GN, Wehbe C. Erythroblastosis fetalis. In: StatPearls. Treasure Island (FL): StatPearls Publishing. Copyright © 2021, StatPearls Publishing LLC; 2021
• 12 Rath ME, Smits-Wintjens VE, Walther FJ, Lopriore E. Hematological morbidity and management in neonates with hemolytic disease due to red cell alloimmunization. Early Hum Dev 2011; 87 (09) 583-588
• 13 Ree IMC, Smits-Wintjens VEHJ, van der Bom JG, van Klink JMM, Oepkes D, Lopriore E. Neonatal management and outcome in alloimmune hemolytic disease. Expert Rev Hematol 2017; 10 (07) 607-616
• 14 Hendrickson JE, Delaney M. Hemolytic disease of the fetus and newborn: modern practice and future investigations. Transfus Med Rev 2016; 30 (04) 159-164
• 15 Moise Jr KJ, Argoti PS. Management and prevention of red cell alloimmunization in pregnancy: a systematic review. Obstet Gynecol 2012; 120 (05) 1132-1139
• 16 Ruma MS, Moise Jr KJ, Kim E. et al. Combined plasmapheresis and intravenous immune globulin for the treatment of severe maternal red cell alloimmunization. Am J Obstet Gynecol 2007; 196 (02) 138.e1-138.e6
• 17 Yu D, Ling LE, Krumme AA, Tjoa ML, Moise Jr KJ. Live birth prevalence of hemolytic disease of the fetus and newborn in the United States from 1996 to 2010. AJOG Glob Rep 2023; 3 (02) 100203
• 18 Rahimi-Levene N, Chezar J, Yahalom V. Israeli HDFN Study Group Investigators. Red blood cell alloimmunization prevalence and hemolytic disease of the fetus and newborn in Israel: a retrospective study. Transfusion 2020; 60 (11) 2684-2690
• 19 Hall V, Avulakunta ID. Hemolytic diseases of the newborn. In: StatPearls. Treasure Island (FL): StatPearls Publishing. Copyright © 2021, StatPearls Publishing LLC; 2021
• 20 Chávez GF, Mulinare J, Edmonds LD. Epidemiology of Rh hemolytic disease of the newborn in the United States. JAMA 1991; 265 (24) 3270-3274
• 21 Friedman C, Alderson PO, Austin JH, Cimino JJ, Johnson SB. A general natural-language text processor for clinical radiology. J Am Med Inform Assoc 1994; 1 (02) 161-174
• 22 Crowley RS, Castine M, Mitchell K, Chavan G, McSherry T, Feldman M. caTIES: a grid based system for coding and retrieval of surgical pathology reports and tissue specimens in support of translational research. J Am Med Inform Assoc 2010; 17 (03) 253-264
• 23 Xie F, Khadka N, Fassett MJ. et al. Identification of preterm labor evaluation visits and extraction of cervical length measures from electronic health records within a large integrated health care system: algorithm development and validation. JMIR Med Inform 2022; 10 (09) e37896
• 24 Xie F, Mercado C, Kim SS. et al. Identifying spontaneous abortion from clinical notes within a large integrated healthcare system. SN Comput Sci 2022; 3 (04) 268
• 25 Koebnick C, Langer-Gould AM, Gould MK. et al. Sociodemographic characteristics of members of a large, integrated health care system: comparison with US Census Bureau data. Perm J 2012; 16 (03) 37-41

Publication History

Received: 18 March 2024

Accepted: 15 October 2024

Article published online:
12 November 2024