ABSTRACT
Sickle cell disease (SCD), a monogenic disorder, exhibits variable severity due to genetic and environmental modifiers. Prior studies have proposed hemolytic and vaso‐occlusive sub‐phenotypes based on limited data. We used latent class analysis (LCA) to identify complication‐based sub‐phenotypes in a large longitudinal cohort. From California administrative databases (1991–2019), we assembled a cohort of 7636 SCD patients (53% female). Disease‐related complications (e.g., vaso‐occlusive episodes [VOE], acute chest syndrome [ACS], avascular necrosis [AVN], chronic kidney disease [CKD], pulmonary hypertension, stroke, gallbladder disease, sepsis, obstructive lung disease, venous‐thromboembolism [VTE], and leg ulcers) were extracted via ICD‐9/10 codes. LCA models, stratified by sex, identified classes; we repeated the analysis in a younger sub‐cohort (≤ 25 years; n = 2210). Males had higher cumulative incidences for recurrent ACS, AVN, gallbladder disease, obstructive lung disease (age 20), and leg ulcers/CKD (age 40). LCA revealed four classes in both sexes: Vaso‐Occlusive (high VOE/ACS/AVN; 23% males, 15% females), Hemolytic (high CKD/pulmonary hypertension/stroke; 8% males, 13% females), Overlap (both patterns; 8% each), and Low Complications (62%–63%). In the younger cohort, classes were Vaso‐Occlusive (8%), Hemolytic (3%), ACS‐dominant (12%), and Low Complications (78%). Vaso‐Occlusive, Hemolytic, and Overlap classes had elevated mortality (HR 1.3–2.4, p < 0.05) versus Low Complications; in youth, Hemolytic had the worst survival (HR 7.8, p < 0.001). LCA confirms the presence of vaso‐occlusive and hemolytic sub‐phenotypes, with low‐complication groups predominant. Age‐specific differences suggest evolving phenotypes, which can inform future targeted therapies and trials.
Keywords: complications, latent class analysis, longitudinal, phenotypes, sickle cell disease
1. Introduction
Sickle cell disease (SCD) arises from a single point mutation in the beta‐globin gene that results in an abnormal hemoglobin molecule [1]. Under low oxygen conditions, erythrocytes containing sickle hemoglobin (HbS) irreversibly elongate, aggregate into dense heterocellular sheets, and occlude the microvasculature. These events trigger a cascade of hemolysis, severe pain, and tissue damage due to a lack of blood flow. Chronic hemolytic anemia and recurrent vaso‐occlusion manifest as pain, acute and chronic organ damage, and early death in affected individuals [2].
Although SCD is a monogenic disorder, its clinical severity is variable due to genetic and environmental modifiers [3, 4]. For example, co‐inherited genetic variants, such as those causing α‐thalassemia, reduce hemoglobin concentration per red blood cell (RBC), thereby decreasing HbS polymerization and hemolysis rates [5]. Increased HbF production mitigates RBC sickling and is a key therapeutic strategy for SCD [6].
The complex pathophysiology of SCD varies among individuals and can change over time within the same person. SCD severity and related complications are not fully explained by co‐inheritance of α‐thalassemia or differences in HbF levels. Others have proposed two overlapping yet distinct clinical sub‐phenotypes: (i) hemolysis‐endothelial dysfunction, characterized by lower baseline hemoglobin, ischemic stroke, pulmonary hypertension, leg ulcers, and priapism; and (ii) vaso‐occlusion and cell–cell adhesion, which present as recurrent pain episodes, acute chest syndrome (ACS), and avascular necrosis (AVN) [2, 7]. These sub‐phenotypes have been based on clinical data from single‐institution, cross‐sectional studies and review articles but have never been validated in large population studies.
The Sickle Cell Data Collection (SCDC) program estimates that 7000 people with SCD currently live in California [8]. Using administrative data from the California Department of Health Care Access and Information (HCAI), we created a longitudinal cohort of Californians with SCD that allows tracking each individual across multiple healthcare facilities statewide. Latent class analysis (LCA), a widely used statistical method in the social sciences, identifies unobserved subgroups within a population based on the distribution of categorical variables. We hypothesized that LCA could identify clinically distinct clusters of disease‐related complications within this longitudinal SCD cohort. We aimed to determine whether SCD clinical complications cluster into previously described subphenotypes and whether these groups are associated with different mortality rates.
2. Methods
2.1. Study Design and Population
In this longitudinal cohort study, we analyzed complications in Californians with SCD using linked administrative data from the Patient Discharge Database (PDD; 1991–2019), Emergency Department records (ED; 2005–2019), and the Comprehensive Death File (available through 2017), as previously described [9]. Individuals were longitudinally tracked across all reporting facilities using an encrypted record linkage number.
SCD cases were defined as: 2 admissions with SCD as the principal diagnosis, or one admission with SCD as the principal diagnosis and two additional admissions with SCD as a secondary diagnosis [10, 11, 12, 13, 14, 15, 16]. The first encounter recorded in either the PDD or ED database was used to determine sex. The patients' ages at discharge were recorded on all admissions.
A younger sub‐cohort (≤ 25 years old) comprised patients born after 1991, with complications assessed only during admissions with age at discharge ≤ 25 years.
2.2. Clinical Complications
Clinical complications were extracted from the HCAI PDD and ED databases. The following complications were selected based on their relevance to SCD: recurrent acute chest syndrome (ACS) episodes, frequent vaso‐occlusive pain episodes (VOE), avascular necrosis (AVN), pulmonary hypertension, chronic kidney disease (CKD), leg ulcers, liver disease, stroke, priapism, sepsis, obstructive lung disease, essential hypertension, gallbladder disease, and venous‐thromboembolism (VTE). Recurrent ACS was defined as two or more episodes during the total follow‐up time. We identified frequent VOEs as ≥ 3 hospitalizations for pain episodes per year over the total follow‐up period [17]. This definition was used for the younger cohort, but was restricted to admissions with age at discharge ≤ 25 years. All clinical complications were identified using International Classification of Diseases, Clinical Modification (ICD‐9/10‐CM) codes, 9th/10th revisions (Table S1); confirmatory laboratory, imaging, or procedure data were not available. We identified the age at discharge of the first occurrence of each complication except for recurrent ACS, which required a second event.
Although SCD genotype‐specific ICD‐9/10‐CM codes were available, substantial longitudinal discordance in genotype coding was observed across admissions. Therefore, individuals were classified as administratively defined HbSS if: all genotype‐specific SCD codes indicated HbSS, > 50% of all admissions included an HbSS code, or > 50% of admissions containing an SCD genotype code included HbSS.
2.3. Statistical Analysis
We used descriptive analysis to show the distribution of complications among SCD patients by sex. We calculated the cumulative incidence functions of each complication on the age scale, accounting for the competing risk of death using Fine and Gray methodology [18] and stratified by sex. Person‐time was calculated from age 0 (birth) to the age at discharge for each complication, at death, or at the last known discharge, whichever occurred first.
We employed LCA to identify unobserved subgroups (latent classes) of SCD patients based on patterns of clinical complications. LCA is a statistical method that assumes the population consists of distinct subgroups, where the observed variables (complications) are conditionally independent given class (or group) membership. Using an iterative approach with 7 models varying in the number of complications and the number of classes included, we selected the best model based on a combination of statistical fit and clinical interpretability. Statistical fit was assessed using indices such as the Bayesian Information Criterion (BIC), Akaike Information Criterion (AIC), and G2. Lower BIC and AIC indicate better model fit [19, 20] (Table S4). Because measurement invariance by sex was rejected, final LCA models were stratified by sex [20]. The class assignment used the highest posterior probability. Classification quality was assessed by the proportion of patients with posterior probability < 0.5 and by the average posterior probability per class (target ≥ 0.7). To evaluate potential sub‐phenotypes, the distribution of complications was plotted by LCA class to determine whether the LCA classes were distinct sub‐phenotypes (non‐parallel lines) or a single class with varying degrees of severity (parallel lines) [21].
To explore differences in complication patterns by age, we repeated the LCA separately for younger patients (≤ 25 years). Measurement invariance held for the younger cohort; therefore, we did not stratify by sex.
Kaplan Meier curves and the Cox proportional hazards model were used to examine the relationship between latent class membership and mortality. Kaplan Meier curves were estimated on the age scale and stratified by sex. The Cox proportional hazards model used the Lexis expansion methodology to account for age‐at‐risk [22] and sex. Class membership was treated as a categorical predictor, with the Low Complication class as the referent group.
A two‐sided p‐value of < 0.05 was considered statistically significant for all hypothesis tests. The study was approved by the California Health and Human Services Agency Committee for the Protection of Human Subjects and the University of California, Davis Institutional Review Boards. As the study used de‐identified data, a waiver of informed consent was granted. It was performed in accordance with the Declaration of Helsinki.
3. Results
3.1. Patient Characteristics
The study cohort consisted of 7636 individuals with SCD in California, of whom 4070 (53.3%) were females, observed between 1991 and 2019. Characteristics are summarized in Table 1. Follow‐up time was shorter in males than in females (median of 12.0 [IQR 4.0–20.0] vs. 14.0 [IQR 5.0–22.0] years). The younger cohort included 2210 individuals with SCD aged ≤ 25 years, of whom 50.9% were male (n = 1125). The median age of entry in this cohort was 2.0 years (IQR 0.0–7.0), and the median follow‐up was 18.0 years (IQR 10.0–23.0).
TABLE 1.
Baseline characteristics of SCD patients, 1991–2019.
| Characteristics | Total | Males | Females | p |
|---|---|---|---|---|
|
N (col %) (N = 7636) |
N (col %) (N = 3566) |
N (col %) (N = 4070) |
||
| SCD complications | ||||
| Recurrent ACS | 2745 (35.9) | 1363 (38.2) | 1382 (34.0) | < 0.001 |
| AVN | 1874 (24.5) | 911 (25.5) | 963 (23.7) | 0.056 |
| CKD | 973 (12.7) | 475 (13.3) | 498 (12.2) | 0.156 |
| Essential hypertension | 2588 (33.9) | 1162 (32.6) | 1426 (35.0) | 0.024 |
| Gallbladder disease | 2240 (29.3) | 1041 (29.2) | 1199 (29.5) | 0.798 |
| Leg ulcers | 603 (7.9) | 332 (9.3) | 271 (6.7) | < 0.001 |
| Liver disease | 678 (8.9) | 297 (8.3) | 381 (9.4) | 0.114 |
| Obstructive lung disease | 2635 (34.5) | 1131 (31.7) | 1504 (37.0) | < 0.001 |
| Priapism | — | 412 (11.6) | — | NA |
| Pulmonary hypertension | 1224 (16.0) | 555 (15.6) | 669 (16.4) | 0.299 |
| Sepsis | 2983 (39.1) | 1357 (38.1) | 1626 (40.0) | 0.090 |
| Stroke | 746 (9.8) | 319 (8.9) | 427 (10.5) | 0.023 |
| Frequent VOE | 2124 (27.8) | 982 (27.5) | 1142 (28.1) | 0.612 |
| VTE | 1011 (13.2) | 427 (12.0) | 584 (14.3) | 0.002 |
| Number of complications‐ALL | 0.046 | |||
| None | 1228 (16.1) | 561 (15.7) | 667 (16.4) | |
| 1–2 Complications | 2588 (33.9) | 1167 (32.7) | 1421 (34.9) | |
| 3–5 Complications | 2370 (31.0) | 1160 (32.5) | 1210 (29.7) | |
| 6–14 Complications | 1450 (19.0) | 678 (19.0) | 772 (19.0) | |
| Number of complications‐LCA a | < 0.001 | |||
| None | 2406 (31.5) | 1036 (29.1) | 1370 (33.7) | |
| 1–2 Complications | 3282 (43.0) | 1550 (43.5) | 1732 (42.6) | |
| 3–5 Complications | 1763 (23.1) | 879 (24.6) | 884 (21.7) | |
| 6–9 Complications | 185 (2.4) | 101 (2.8) | 84 (2.1) | |
| Number of years follow‐up in HCAI | < 0.001 | |||
| Mean ± SD | 13.2 ± 9.3 | 12.4 ± 9.2 | 13.9 ± 9.3 | |
| Median (IQR) | 13.0 (4.0–21.0) | 12.0 (4.0–20.0) | 14.0 (5.0–22.0) | |
Note: Recurrent acute chest syndrome (ACS) defined as ≥ 2 ACS admissions, Frequent vaso‐occlusive episodes (VOE) defined as average ≥ 3 admissions with crisis per year.
Abbreviations: AVN, avascular necrosis; CKD, chronic kidney disease; IQR, interquartile range; SD, standard deviation; VTE, venous thromboembolism.
Includes Recurrent ACS, Frequent VOE, AVN, Pulmonary Hypertension, CKD, Stroke, Liver Disease, Leg Ulcers, and Priapism.
3.2. Cumulative Incidences
The cumulative incidences (CMI) for each complication by sex, accounting for the competing risk of death, at the ages of 20, 30, 40, and 60 are shown in Table 2, with CMI curves shown in Figure S1. Some complications have an early onset, and males have a higher CMI than females. At the age of 20, the CMI of recurrent ACS was 23.8% for males and 18.0% for females (p < 0.001), and for obstructive lung disease at 19.5% for males and 16.6% for females (p = 0.042). The CMI at age 20 was higher among males for both gallbladder disease (males‐14.3% vs. females‐12.6%; p = 0.046) and AVN (males‐9.1% vs. females‐6.5%; p < 0.001). We found no differences in CMI by sex for sepsis or stroke.
TABLE 2.
Cumulative incidence of complications among Californians with sickle cell disease by sex.
| Complication | Age | Males CMI (95% CI) | Females CMI (95% CI) | p |
|---|---|---|---|---|
| Recurrent ACS a | 20 | 23.8% (22.4, 25.2) | 18.0% (16.8, 19.2) | < 0.001 |
| Recurrent ACS a | 30 | 37.9% (36.2, 39.6) | 29.9% (28.4, 31.3) | |
| Recurrent ACS a | 40 | 48.5% (46.7, 50.3) | 40.4% (38.7, 42.1) | |
| Recurrent ACS a | 60 | 64.6% (62.3, 66.8) | 58.2% (56.1, 60.3) | |
| Obstructive lung disease | 20 | 19.5% (18.2, 20.8) | 16.6% (15.5, 17.8) | 0.042 |
| Obstructive lung disease | 30 | 25.3% (23.8, 26.8) | 24.5% (23.1, 25.8) | |
| Obstructive lung disease | 40 | 30.2% (28.6, 31.9) | 32.7% (31.2, 34.3) | |
| Obstructive lung disease | 60 | 41.0% (38.7, 43.2) | 48.8% (46.7, 50.8) | |
| Gallbladder disease | 20 | 14.3% (13.2, 15.5) | 12.6% (11.6, 13.7) | 0.046 |
| Gallbladder disease | 30 | 23.3% (21.8, 24.8) | 21.3% (20.0, 22.6) | |
| Gallbladder disease | 40 | 30.1% (28.4, 31.9) | 28.5% (26.9, 30.0) | |
| Gallbladder disease | 60 | 42.0% (39.6, 44.3) | 39.5% (37.5, 41.6) | |
| AVN | 20 | 9.11% (8.16, 10.1) | 6.46% (5.70, 7.27) | < 0.001 |
| AVN | 30 | 17.7% (16.4, 19.1) | 14.0% (12.9, 15.2) | |
| AVN | 40 | 26.8% (25.1, 28.5) | 22.2% (20.8, 23.7) | |
| AVN | 60 | 41.6% (39.0, 44.2) | 36.7% (34.5, 38.9) | |
| Sepsis | 20 | 14.8% (13.6, 16.0) | 12.5% (11.5, 13.6) | 0.156 |
| Sepsis | 30 | 25.7% (24.2, 27.3) | 24.7% (23.3, 26.1) | |
| Sepsis | 40 | 36.7% (34.9, 38.5) | 35.0% (33.4, 36.6) | |
| Sepsis | 60 | 55.9% (53.3, 58.4) | 55.6% (53.4, 57.8) | |
| Stroke | 20 | 2.84% (2.32, 3.44) | 2.48% (2.02, 3.00) | 0.765 |
| Stroke | 30 | 5.03% (4.29, 5.84) | 4.74% (4.08, 5.47) | |
| Stroke | 40 | 7.71% (6.71, 8.79) | 7.54% (6.63, 8.52) | |
| Stroke | 60 | 16.9% (14.8, 19.1) | 18.8% (16.9, 20.8) | |
| Priapism | 20 | 5.83% (5.07, 6.66) | — | — |
| Priapism | 30 | 10.2% (9.12, 11.3) | — | |
| Priapism | 40 | 13.3% (12.1, 14.6) | — | |
| Priapism | 60 | 15.6% (14.1, 17.2) | — | |
| Leg ulcer | 20 | 0.87% (0.60, 1.25) | 0.80% (0.55, 1.13) | < 0.001 |
| Leg ulcer | 30 | 3.54% (2.91, 4.26) | 2.80% (2.28, 3.40) | |
| Leg ulcer | 40 | 9.09% (7.93, 10.3) | 5.97% (5.12, 6.90) | |
| Leg ulcer | 60 | 20.4% (18.1, 22.9) | 12.0% (10.5, 13.6) | |
| CKD | 20 | 0.66% (0.42, 1.00) | 0.40% (0.24, 0.66) | < 0.001 |
| CKD | 30 | 3.24% (2.62, 3.95) | 2.20% (1.75, 2.74) | |
| CKD | 40 | 9.06% (7.89, 10.3) | 5.85% (5.01, 6.78) | |
| CKD | 60 | 32.3% (29.4, 35.3) | 24.0% (21.8, 26.4) | |
| Essential hypertension | 20 | 6.96% (6.13, 7.86) | 4.24% (3.63, 4.91) | 0.123 |
| Essential hypertension | 30 | 16.9% (15.6, 18.3) | 12.8% (11.7, 13.9) | |
| Essential hypertension | 40 | 28.9% (27.1, 30.7) | 26.2% (24.6, 27.8) | |
| Essential hypertension | 60 | 58.3% (55.5, 60.9) | 60.9% (58.5, 63.2) | |
| Pulmonary hypertension | 20 | 2.95% (2.41, 3.58) | 2.07% (1.65, 2.56) | 0.089 |
| Pulmonary hypertension | 30 | 7.98% (7.03, 9.01) | 7.24% (6.40, 8.15) | |
| Pulmonary hypertension | 40 | 14.4% (13.1, 15.9) | 13.1% (11.9, 14.4) | |
| Pulmonary hypertension | 60 | 31.5% (28.8, 34.3) | 29.0% (26.8, 31.2) | |
| VTE | 20 | 2.78% (2.27, 3.39) | 2.52% (2.05, 3.05) | 0.285 |
| VTE | 30 | 6.45% (5.61, 7.38) | 7.34% (6.50, 8.24) | |
| VTE | 40 | 12.7% (11.4, 14.1) | 12.6% (11.4, 13.8) | |
| VTE | 60 | 22.3% (20.0, 24.7) | 23.7% (21.8, 25.7) | |
| Liver disease | 20 | 1.75% (1.35, 2.24) | 1.45% (1.11, 1.87) | 0.730 |
| Liver disease | 30 | 4.14% (3.46, 4.91) | 3.55% (2.98, 4.21) | |
| Liver disease | 40 | 6.96% (5.99, 8.01) | 7.08% (6.19, 8.06) | |
| Liver disease | 60 | 16.8% (14.7, 19.0) | 18.0% (16.1, 20.0) |
Abbreviations: AVN, avascular necrosis; CKD, chronic kidney disease; VTE, venous thromboembolism.
Recurrent acute chest syndrome (ACS)‐ cumulative incidence at age of 2nd ACS admission.
Complications such as leg ulcers, CKD, essential hypertension, pulmonary hypertension, VTE, and liver disease occurred with notably higher frequency in the older age groups, and males also had a higher CMI than females. At age 40, the CMI for leg ulcers and CKD were similar, approximately 9% for males compared to 6% for females (p < 0.001 for both). We found no differences in CMI by sex for essential hypertension, pulmonary hypertension, VTE, and liver disease, but CMI at age 40 ranged from 28.9% for essential hypertension to 7% for liver disease.
3.3. Latent Class Analysis (LCA)
Among males with SCD, four distinct classes were identified in the final model. Table S2a shows model selection. Figure 1a shows the distribution of complications across classes. The “vaso‐occlusive class” accounted for 22.8% of the cohort, who had complications consistent with the previously published vaso‐occlusive sub‐phenotype. Individuals in this class had moderately frequent VOE (54.2%), high recurrent ACS (75.7%), high AVN (60.1%), but lower frequency of pulmonary hypertension (21.4%) and priapism (16.7%). A total of 7.7% of males were in the “hemolytic class,” and their common complications were consistent with the proposed hemolytic sub‐phenotype i.e., high relative frequency of CKD (95.6%), pulmonary hypertension (48.4%), and stroke (24.4%), but a lower frequency of frequent VOE (4.7%) and priapism (7.6%). The “overlap class” comprised 7.9% of participants, with complications included an overlap of vaso‐occlusive and hemolytic pathophysiology. Individuals in this class had very frequent VOE (97.5%), high recurrent ACS (99.7%), high AVN (73.9%), high frequency of pulmonary hypertension (75.3%), and priapism (33.2%) but lower frequency of stroke (16.6%). Much of the cohort fell into the “low complications” class (61.5%), with low frequencies of most complications; the highest were frequent VOE (11.5%) and recurrent ACS (15.1%). The average posterior probabilities were ≥ 0.7 for all classes (vaso‐occlusive‐ 0.720, hemolytic‐ 0.797, overlap‐ 0.765, low complications‐ 0.862), and the potential of misclassification (final posterior probability of individuals < 0.50) was 1.4%. The overlap and vaso‐occlusive classes demonstrated the highest proportion of individuals with administratively defined HbSS (Figure S2). Follow‐up time was similar among classes vaso‐occlusive, hemolytic, and overlap (median of 16.0 [IQR 8.0–24.0], 17.0 [IQR 10.0–24.0], and 21.0 [IQR 15.0–27.0] years, respectively). Individuals in the low complication class had less follow‐up time (median 8.0 [IQR 2.0–17.0] years).
FIGURE 1.
Distribution of complications among Californians with sickle cell disease by LCA class and sex. (A) Males. (B) Females. LCA: Latent class analysis: Recurrent acute chest syndrome (ACS) defined as ≥ 2 ACS admissions, Frequent vaso‐occlusive episodes (VOE) defined as average ≥ 3 admissions with crisis per year, AVN‐avascular necrosis, CKD‐chronic kidney disease. Overlap describes a class of patients with both vaso‐occlusive and hemolytic characteristics. [Color figure can be viewed at wileyonlinelibrary.com]
Among females with SCD, four distinct classes were again identified in the final model (Table S2b). Figure 1b shows the distribution of complications across classes. The “vaso‐occlusive class” comprised 15.3% of the cohort with frequent VOE (69.1%), recurrent ACS (83.0%), and AVN (63.7%), but a lower frequency of pulmonary hypertension (13.7%). The “hemolytic class” comprised 13.1% of the female cohort and had a high frequency of CKD (55.4%), which was lower than what was observed in the male cohort. They also had a high percentage of pulmonary hypertension (56.3%) and stroke (30.6%), but a lower frequency of VOE (13.0%) and leg ulcers (15.4%). As in males, there was an “overlap class” that included both vaso‐occlusive and hemolytic complications. The overlap class included 8.2% of the cohort and had high frequent VOE (91.9%), recurrent ACS (99.4%), AVN (76.9%), pulmonary hypertension (85.3%), and stroke (30.9%). As in males, most of the female cohort fell into the “low complication” class (63.4%), with a low incidence of most complications; the highest was frequent VOE (13.1%) and recurrent ACS (11.2%). The average posterior probabilities were ≥ 0.7 for all classes, except the vaso‐occlusive class (vaso‐occlusive‐ 0.665, hemolytic‐ 0.719, overlap‐ 0.760, low complications‐ 0.873), and the potential of misclassification was 4.8%. Similar patterns were observed among females, with enrichment of administratively defined HbSS within the overlap and vaso‐occlusive classes (Figure S2). Follow‐up time was similar among the vaso‐occlusive, hemolytic, and overlap classes (16.0 [IQR 8.0–23.0], 17.0 [IQR 10.0–24.0], and 21.0 [IQR 15.0–27.0] years, respectively) but was shorter in the low complication class (8.0 [IQR 2.0–17.0] years).
In the younger SCD cohort, we identified four classes with distinct complication profiles compared with the entire cohort, as detailed in the model selection criteria (Table S3). Among the 2210 individuals with SCD under 25 years old, the “vaso‐occlusive class” comprised 7.6% of the cohort and exhibited frequent VOE (95.8%), recurrent ACS (73.8%), and AVN (100%). The “hemolytic class” comprised 3.2% of the cohort and showed a higher frequency of pulmonary hypertension (41.4%), CKD (34.3%), leg ulcers (17.1%), liver disease (54.3%), and stroke (27.1%). Unlike the entire cohort, the third class, designated the “ACS class” (11.5%), was characterized by recurrent ACS (100%), less frequent VOE (35.3%), and moderate pulmonary hypertension (27.1%). Lastly, most patients (77.7%) were in the “low complication” class. Figure 2 illustrates the distribution of complications by class. The average posterior probabilities were vaso‐occlusive (0.577), hemolytic (0.754), overlap (0.798), and low complications (0.956), and the potential for misclassification was 5.7%. Similar to the entire cohort, follow‐up time was similar among the vaso‐occlusive, hemolytic, and ACS classes (median of 19.0 [IQR 11.0–25.0], 17.0 [IQR 11.0–24.0], and 23.0 [IQR 16.0–27.0] years, respectively), but was shorter in the low complication class (median 10.0 [IQR 3.0–19.0] years).
FIGURE 2.
Distribution of complications by LCA class and sex among young Californians with sickle cell disease patients' admissions ≤ 25 years of age. LCA: Latent class analysis: Recurrent acute chest syndrome (ACS) defined as ≥ 2 ACS admissions, Frequent vaso‐occlusive episodes (VOE) defined as average ≥ 3 admissions with crisis per year, AVN‐avascular necrosis, CKD‐chronic kidney disease. [Color figure can be viewed at wileyonlinelibrary.com]
3.4. Mortality
There were 634 deaths in males with SCD (17.8%) through the end of 2017. Figure 3A shows that the vaso‐occlusive (HR 2.11, 95% CI 1.73–2.56, p < 0.001), hemolytic (HR 1.30, 95% CI 1.02–1.64, p = 0.032), or overlap classes (HR 2.04, 95% CI 1.61–2.59, p < 0.001) were associated with increased mortality compared to the low complication class.
FIGURE 3.
Association of LCA class with overall mortality among Californians with sickle cell disease by sex compared to the Low Complication LCA class. (A) Males. (B) Females. Overlap describes a class of patients with both vaso‐occlusive and hemolytic characteristics. Cox proportional hazard regression models were adjusted for age‐at‐risk. (A) Males‐ Significant differences between Vaso Occlusive and Hemolytic classes (p < 0.001), Vaso‐occlusive and Overlap (p = 0.001), and Hemolytic and Overlap classes (p < 0.001). (B) Females‐ Significant differences between Hemolytic and Overlap classes (p = 0.004). [Color figure can be viewed at wileyonlinelibrary.com]
There were 711 deaths in females with SCD (17.5%) through the end of 2017. Similar to what we observed in the male cohort, vaso‐occlusive (HR 2.13, 95% CI 1.72–2.63, p < 0.001), hemolytic (HR 1.70, 95% CI 1.40–2.06, p < 0.001), or overlap (HR 2.39, 95% CI 1.92–2.97, p < 0.0001) classes were associated with increased mortality compared to the low complication class in females (Figure 3b).
Overall mortality was low in the younger SCD cohort (N = 59, 2.7%), but, unlike the entire cohort, there were no significant differences between the vaso‐occlusive and low‐complication classes (HR 1.60, 95% CI 0.69–3.71, p = 0.270), as shown in Figure 4. The worst survival was seen in the hemolytic class (HR 7.81, 95% CI 4.00–15.24, p < 0.001) compared to the low complication class.
FIGURE 4.
Association of LCA class with overall mortality among young Californians with sickle cell disease patients' admissions ≤ 25 years of age compared to the Low Complication LCA class. ACS: acute chest syndrome. Cox proportional hazard regression model was adjusted for age‐at‐risk and sex. Significant differences between Vaso‐Occlusive and Hemolytic classes (p < 0.001) and Hemolytic and ACS classes (p = 0.001). [Color figure can be viewed at wileyonlinelibrary.com]
4. Discussion
Using our longitudinal SCD cohort, we determined the cumulative incidence of disease‐related complications and found that males generally have a higher CMI than females for recurrent ACS, AVN, gallbladder disease, and obstructive lung disease at age 20, as well as leg ulcers and CKD at age 40. No sex difference was noted for sepsis, stroke, essential hypertension, pulmonary hypertension, liver disease, or VTE. These complications were incorporated into LCA models, and we identified four distinct complication classes: vaso‐occlusive, hemolytic, overlap, and low‐complication. A separate analysis of patients aged 25 years or younger yielded similar classes, but the “overlap” class was replaced by an “ACS dominant” class. In the entire cohort, the vaso‐occlusive, hemolytic, and overlap classes were associated with worse overall survival than the low‐complication class. In the younger cohort, mortality was highest in the hemolytic class.
Our findings reveal both similarities and notable differences in the estimated CMI for complications compared with historical data from the Cooperative Study of Sickle Cell Disease (CSSCD), the most comprehensive longitudinal study of SCD natural history, encompassing 4085 participants from 1978 to 1998. The CSSCD found that 44% of 1085 patients with at least one episode of ACS experienced two or more episodes by age 20, though exact CMI was not specified [23]. Our study observed a recurrent ACS CMI of 23.8% in males by age 20, lower than the CSSCD, potentially due to modern interventions such as prophylactic penicillin, pneumococcal vaccination, and more widespread use of incentive spirometry during hospitalizations for vaso‐occlusive pain episodes [24] [25, 26]. The CSSCD reported an AVN prevalence of approximately 6% in patients under 25 years, with peak incidence between 30 and 40 [27]. Similarly, our study found a cumulative AVN incidence of 9.11% by age 20, nearly doubling to 17.7% by age 30. Regarding CKD, a contemporary cohort of 535 adults with HbSS (median baseline age 22 years) followed for approximately 5 years reported a CKD CMI of 7.3% [28]. Our findings align closely, with a CKD CMI ranging from 2.2% to 9.1% between ages 30 and 40, underscoring a progressive risk with age. These comparisons highlight evolving disease patterns and the impact of improved preventative strategies in contemporary SCD management.
Using LCA, we applied an unbiased statistical method to identify distinct subgroups within a California cohort of individuals with SCD based on accumulated complications, without pre‐specifying which complications would cluster together. This approach, which assigns all participants and complications to the model to reveal patterns over time, identified four sub‐phenotypic classes consistent with established mechanisms of SCD pathogenesis without sex‐based differences. The vaso‐occlusive class was characterized by complications linked to vaso‐occlusion, including ACS, VOE, and AVN [2, 29]. The hemolytic class was defined by complications associated with hemolysis and vasculopathy, such as CKD, leg ulcers, liver disease, pulmonary hypertension, and stroke [7]. The overlap class showed moderately high rates of both vaso‐occlusive and hemolytic complications. The low‐complication class included most patients and had overall low complication rates.
Priapism unexpectedly clustered with vaso‐occlusive complications rather than hemolytic complications, contrary to the sub‐phenotype originally proposed by Kato, Steinberg, and Gladwin [7]. This finding is likely influenced by significant under‐ascertainment in administrative data. Our observed prevalence of priapism was 11.6%, significantly lower than the 33%–45% reported in prospective studies and patient‐reported data [30]. Males with SCD frequently do not spontaneously report priapism episodes unless specifically elicited, and many stuttering or self‐resolving episodes are managed outpatient or at home without hospitalization [31, 32]. Consequently, only severe episodes requiring emergency or inpatient care are captured by ICD codes in this dataset. This limitation likely biases priapism toward clustering with other complications that result in hospitalization (e.g., VOE or ACS), rather than reflecting true pathophysiologic clustering. This is an important limitation of our data source that influences the classification of priapism within our sub‐phenotypic categories.
Although the statistical model suggested stratifying the LCA by sex, minimal differences were observed between male and female participants across the identified classes. Females in the vaso‐occlusive class exhibited slightly higher rates of ACS (83% vs. 76%) and frequent VOE (69% vs. 54%) than males. Conversely, males in the hemolytic class showed higher rates of CKD (96% vs. 55%) and leg ulcers (35% vs. 15%) than females. These findings align with a cross‐sectional study reporting higher odds of recurrent VOE in females and increased leg ulcer incidence in males with SCD [33], suggesting potential sex‐specific patterns in complication profiles that warrant further investigation.
When limited to younger SCD patients, LCA identified sub‐phenotypic classes that differed from those observed when all ages were included. The vaso‐occlusive class was characterized by high rates of frequent VOE and AVN and low rates of other complications. Like the entire cohort, the hemolytic class had high rates of pulmonary hypertension, CKD, leg ulcers, liver disease, and stroke, but a lower frequency of priapism in males. The ACS class was characterized by individuals who all had recurrent ACS, with moderate rates of frequent VOE, AVN, pulmonary hypertension, and stroke. This group also had high rates of priapism. Differences in class structure may reflect differences in the causes of complications in younger patients. In children and young adults, ACS is more commonly triggered by infections or obstructive lung disease [34, 35, 36, 37, 38]. In contrast, ACS in older adults is often associated with VOE and linked to pulmonary fat and bone marrow emboli [39]. These distinct mechanisms likely contribute to the differences in phenotypic classes observed between the younger and the entire cohorts of SCD patients. Some complications also do not typically develop until later in life (e.g., CKD, VTE), and patient class assignment likely changes with age.
Over the past 40 years, life expectancy for people with SCD in the United States has increased from 28 years to approximately 45 years [17, 40, 41, 42]. Cause of death is shifting away from acute infectious complications toward chronic end‐organ damage, including pulmonary, liver, and renal disease [40, 41]. As expected, males and females in LCA classes with higher complication rates had greater mortality than those in the low‐complication class. In both males and females, the highest risk for death was observed in the vaso‐occlusive and overlap classes, which had higher mortality than the hemolytic class. This is initially counterintuitive given the hemolytic class's higher frequency of CKD and liver disease. However, this could be explained by coding practices that can make acute events appear more prominent in aggregate data. While chronic end‐organ damage is becoming more common as patients age, acute complications (such as ACS or VOE) often act as the trigger and are more visibly coded [40, 43].
Most patients in our cohort were classified into the low‐complication class. This class had shorter follow‐up times, which may partly explain the lower observed complication rates, and the proportion in this class may decrease with longer observation periods. Additionally, because the study identified patients through emergency department visits or hospitalizations, complications managed exclusively in the outpatient setting may be underestimated. Nevertheless, most patients in this contemporary cohort exhibited low rates of recorded complications.
A notable finding is that a substantial proportion of individuals in the low‐complication class (65.7% among males and 60.0% among females) were classified as administratively defined HbSS. Recent data have also challenged the notion that compound heterozygous genotypes invariably confer a benign phenotype [44, 45]. Thus, while the low‐complication class is likely enriched for milder genotypes, it does not fully account for the observed genotype.
This study is limited by the inherent constraints of administrative claims data. We lacked access to laboratory values, imaging reports, or procedure codes that could confirm diagnoses according to gold‐standard clinical criteria (e.g., echocardiographic tricuspid regurgitant velocity or right heart catheterization for PH, radiographic confirmation for AVN or ACS, or laboratory confirmation of CKD). We also lacked detailed information on disease‐modifying therapies, such as hydroxyurea use or chronic transfusion therapy. Despite these limitations, the cumulative incidence of key chronic complications (Figure S1), such as AVN and CKD, align closely with prior cohort studies [27, 28] supporting the validity of our findings within the constraints of the dataset.
LCA confirmed distinct complication classes in a large, longitudinal cohort of patients with SCD in California, characterized by vaso‐occlusive complications (VOE, ACS, and AVN), hemolytic complications (CKD, pulmonary hypertension, stroke), or an overlap of both, thereby supporting the existence of clinically meaningful SCD sub‐phenotypes. Notably, a substantial subgroup exhibited low overall complication rates. Although this group is likely enriched for milder compound heterozygous genotypes (HbSC and HbSB+ thalassemia), this does not fully explain this phenotype. Other factors, such as contemporary disease‐modifying therapies, under ascertainment of complications, and shorter follow‐up time, likely contribute. These findings suggest that clinically significant end‐organ damage is not inevitable for all patients with SCD in the current era. They underscore the need for refined discussions about disease‐modifying and/or transformative therapies as well as stratified trial designs to optimize outcomes for diverse SCD sub‐phenotypes.
Author Contributions
Conception and study design: Ted Wun. Data analysis and interpretation: Shaina M. Willen, Ann M. Brunson, Oyebimpe O. Adesina, and Ted Wun. Statistical expertise: Ann M. Brunson and Ted Wun. Drafting the manuscript: Shaina M. Willen. Critical revision for important intellectual content: Shaina M. Willen, Ann M. Brunson, Oyebimpe O. Adesina, and Ted Wun. Final approval of the version to be published: Shaina M. Willen, Ann M. Brunson, Oyebimpe O. Adesina, and Ted Wun. Agreement to be accountable for all aspects of the work: Shaina M. Willen, Ann M. Brunson, Oyebimpe O. Adesina, and Ted Wun.
Funding
This work was supported by the American Society of Hematology (ASH).
Conflicts of Interest
T.W. served as a consultant for Pfizer until November 2024. The other authors declare no conflicts of interest.
Supporting information
Table S1: International Classification of Diseases, Clinical Modification using 9th (ICD‐9‐CM) and 10th (ICD‐10‐CM) revisions for all clinical complications.
Table S2: Cumulative incidence of complications among Californians with sickle cell disease by sex.
Figure S1: Cumulative incidence of complications among Californians with sickle cell disease by sex, using age as the scale.
Table S3: Latent Class Analysis (LCA) model fit statistics for (A) Males and (B) Females.
Figure S2: Proportion of SCD individuals classified as administratively defined HbSS across latent classes by sex.
Table S4: Latent Class Analysis (LCA) model fit statistics among Californians with sickle cell disease patients' admissions ≤ 25 years of age.
Acknowledgments
This study used de‐identified data from the California Department of Health Care Access and Information (HCAI). The findings and conclusions are those of the authors and do not necessarily represent HCAI's views. We thank the California Sickle Cell Data Collection (SCDC) Program for population estimates and the California Health and Human Services Agency Committee for the Protection of Human Subjects and the University of California, Davis Institutional Review Board for study approval. This work was supported by an American Society of Hematology (ASH) Scholar Award (SMW).
Data Availability Statement
De‐identified data is available from the corresponding author upon request, subject to a data use agreement where permissible.
References
- 1. Steinberg M. H., “Sickle Cell Anemia, the First Molecular Disease: Overview of Molecular Etiology, Pathophysiology, and Therapeutic Approaches,” ScientificWorldJournal 8 (2008): 1295–1324, 10.1100/tsw.2008.157. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Kato G. J., Gladwin M. T., and Steinberg M. H., “Deconstructing Sickle Cell Disease: Reappraisal of the Role of Hemolysis in the Development of Clinical Subphenotypes,” Blood Reviews 21, no. 1 (2007): 37–47, 10.1016/j.blre.2006.07.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Habara A. and Steinberg M. H., “Minireview: Genetic Basis of Heterogeneity and Severity in Sickle Cell Disease,” Experimental Biology and Medicine (Maywood, N.J.) 241, no. 7 (2016): 689–696, 10.1177/1535370216636726. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Sebastiani P., Solovieff N., Hartley S. W., et al., “Genetic Modifiers of the Severity of Sickle Cell Anemia Identified Through a Genome‐Wide Association Study,” American Journal of Hematology 85, no. 1 (2010): 29–35, 10.1002/ajh.21572. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. de Ceulaer K., Higgs D. R., Weatherall D. J., Hayes R. J., Serjeant B. E., and Serjeant G. R., “Alpha‐Thalassemia Reduces the Hemolytic Rate in Homozygous Sickle‐Cell Disease,” New England Journal of Medicine 309, no. 3 (1983): 189–190, 10.1056/nejm198307213090320. [DOI] [PubMed] [Google Scholar]
- 6. Akinsheye I., Alsultan A., Solovieff N., et al., “Fetal Hemoglobin in Sickle Cell Anemia,” Blood 118, no. 1 (2011): 19–27, 10.1182/blood-2011-03-325258. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Kato G. J., Steinberg M. H., and Gladwin M. T., “Intravascular Hemolysis and the Pathophysiology of Sickle Cell Disease,” Journal of Clinical Investigation 127, no. 3 (2017): 750–760, 10.1172/JCI89741. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Snyder A. B., Lakshmanan S., Hulihan M. M., et al., “Surveillance for Sickle Cell Disease ‐ Sickle Cell Data Collection Program, Two States, 2004–2018,” MMWR Surveillance Summaries 71, no. 9 (2022): 1–18, 10.15585/mmwr.ss7109a1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Brunson A., Lei A., Rosenberg A. S., White R. H., Keegan T., and Wun T., “Increased Incidence of VTE in Sickle Cell Disease Patients: Risk Factors, Recurrence and Impact on Mortality,” British Journal of Haematology 178, no. 2 (2017): 319–326, 10.1111/bjh.14655. [DOI] [PubMed] [Google Scholar]
- 10. Oluwole O., Brunson A. M., Adesina O. O., et al., “Rates of Strokes in Californians With Sickle Cell Disease in the Post‐STOP Era,” Blood 144, no. 24 (2024): 2517–2527, 10.1182/blood.2023023031. [DOI] [PubMed] [Google Scholar]
- 11. Adesina O. O., Brunson A., Fisch S. C., et al., “Pregnancy Outcomes in Women With Sickle Cell Disease in California,” American Journal of Hematology 98, no. 3 (2023): 440–448, 10.1002/ajh.26818. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Brunson A., Keegan T. H. M., Mahajan A., Paulukonis S., and Wun T., “Cancer Specific Survival in Patients With Sickle Cell Disease,” British Journal of Haematology 185, no. 1 (2019): 128–132, 10.1111/bjh.15687. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Hariharan N., Brunson A., Mahajan A., Keegan T. H. M., and Wun T., “Bleeding in Patients With Sickle Cell Disease: A Population‐Based Study,” Blood Advances 4, no. 5 (2020): 793–802, 10.1182/bloodadvances.2019000940. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14. Brunson A., Keegan T. H. M., Bang H., Mahajan A., Paulukonis S., and Wun T., “Increased Risk of Leukemia Among Sickle Cell Disease Patients in California,” Blood 130, no. 13 (2017): 1597–1599, 10.1182/blood-2017-05-783233. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Brunson A., Keegan T., Mahajan A., White R., and Wun T., “High Incidence of Venous Thromboembolism Recurrence in Patients With Sickle Cell Disease,” American Journal of Hematology 94, no. 8 (2019): 862–870, 10.1002/ajh.25508. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Adesina O., Brunson A., Keegan T. H. M., and Wun T., “Osteonecrosis of the Femoral Head in Sickle Cell Disease: Prevalence, Comorbidities, and Surgical Outcomes in California,” Blood Advances 1, no. 16 (2017): 1287–1295, 10.1182/bloodadvances.2017005256. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Platt O. S., Brambilla D. J., Rosse W. F., et al., “Mortality in Sickle Cell Disease ‐ Life Expectancy and Risk Factors for Early Death,” New England Journal of Medicine 330, no. 23 (1994): 1639–1644, 10.1056/NEJM199406093302303. [DOI] [PubMed] [Google Scholar]
- 18. Lin G., Ying S., and Gordon J., “Analyzing Survival Data With Competing Risks Using SAS Software Presented at: SAS Global Forum 2012,” 2012. Cary, NC, Session Statistics and Data Analysis.
- 19. Weller B., Bowen N., and Faubert S., “Latent Class Analysis: A Guide to Best Practice,” Journal of Black Psychology 46 (2020): 287–311, 10.1177/0095798420930932. [DOI] [Google Scholar]
- 20. Lanza S. T., Collins L. M., Lemmon D. R., and Schafer J. L., “PROC LCA: A SAS Procedure for Latent Class Analysis,” Structural Equation Modeling : A Multidisciplinary Journal 14, no. 4 (2007): 671–694, 10.1080/10705510701575602. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Sinha P., Calfee C. S., and Delucchi K. L., “Practitioner's Guide to Latent Class Analysis: Methodological Considerations and Common Pitfalls,” Critical Care Medicine 49, no. 1 (2021): e63–e79, 10.1097/ccm.0000000000004710. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Hong L. S. and Sarah L., “Lexis Expansion‐ Age at Risk Adjustment for Survival Analysis. Presented at: PhUSE 2013,” 2013, Bournemouth, England.
- 23. Castro O., Brambilla D. J., Thorington B., et al., “The Acute Chest Syndrome in Sickle Cell Disease: Incidence and Risk Factors. The Cooperative Study of Sickle Cell Disease,” Blood 84, no. 2 (1994): 643–649. [PubMed] [Google Scholar]
- 24. Bellet P. S., Kalinyak K. A., Shukla R., Gelfand M. J., and Rucknagel D. L., “Incentive Spirometry to Prevent Acute Pulmonary Complications in Sickle Cell Diseases,” New England Journal of Medicine 333, no. 11 (1995): 699–703, 10.1056/nejm199509143331104. [DOI] [PubMed] [Google Scholar]
- 25. Vichinsky E., Neumayr L., and Earles A., “Causes and Outcomes of Acute Chest Syndrome in Sickle Cell Disease,” New England Journal of Medicine 342, no. 25 (2000): 1855–1866. [DOI] [PubMed] [Google Scholar]
- 26. Klings E. S. and Steinberg M. H., “Acute Chest Syndrome of Sickle Cell Disease: Genetics, Risk Factors, Prognosis, and Management,” Expert Review of Hematology 15, no. 2 (2022): 117–125, 10.1080/17474086.2022.2041410. [DOI] [PubMed] [Google Scholar]
- 27. Milner P. F., Kraus A. P., Sebes J. I., et al., “Sickle Cell Disease as a Cause of Osteonecrosis of the Femoral Head,” New England Journal of Medicine 325, no. 21 (1991): 1476–1481, 10.1056/nejm199111213252104. [DOI] [PubMed] [Google Scholar]
- 28. Roger C., Lionnet F., Mattioni S., et al., “Risk Factors for CKD Stage II Onset in a Prospective Cohort of Homozygous Sickle Cell Adults,” American Journal of Hematology 96, no. 9 (2021): 1147–1155, 10.1002/ajh.26264. [DOI] [PubMed] [Google Scholar]
- 29. Zhang D., Xu C., Manwani D., and Frenette P. S., “Neutrophils, Platelets, and Inflammatory Pathways at the Nexus of Sickle Cell Disease Pathophysiology,” Blood 127, no. 7 (2016): 801–809, 10.1182/blood-2015-09-618538. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Idris I. M., Abba A., Galadanci J. A., et al., “Men With Sickle Cell Disease Experience Greater Sexual Dysfunction When Compared With Men Without Sickle Cell Disease,” Blood Advances 4, no. 14 (2020): 3277–3283, 10.1182/bloodadvances.2020002062. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Akakpo W. and Burnett A. L., “Update on the Prevalence and Pathophysiology of Sickle Cell Priapism: A Narrative Review,” International Journal of Impotence Research 38 (2025): 507–510, 10.1038/s41443-025-01026-5. [DOI] [PubMed] [Google Scholar]
- 32. Idris I. M., Burnett A. L., and DeBaun M. R., “Epidemiology and Treatment of Priapism in Sickle Cell Disease,” Hematology. American Society of Hematology. Education Program 2022, no. 1 (2022): 450–458, 10.1182/hematology.2022000380. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Masese R. V., Bulgin D., Knisely M. R., et al., “Sex‐Based Differences in the Manifestations and Complications of Sickle Cell Disease: Report From the Sickle Cell Disease Implementation Consortium,” PLoS One 16, no. 10 (2021): e0258638, 10.1371/journal.pone.0258638. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34. Anim S. O., Strunk R. C., and Debaun M. R., “Asthma Morbidity and Treatment in Children With Sickle Cell Disease,” 2011. [DOI] [PMC free article] [PubMed]
- 35. Assad Z., Valtuille Z., Rybak A., et al., “Unique Changes in the Incidence of Acute Chest Syndrome in Children With Sickle‐Cell Disease Unravel the Role of Respiratory Pathogens: A Time‐Series Analysis,” Chest 165 (2023): 150–160, 10.1016/j.chest.2023.07.4219. [DOI] [PubMed] [Google Scholar]
- 36. Knight‐Madden J. M., Forrester T. S., Lewis N., and Greenough A., “Asthma in Children With Sickle Cell Disease and Its Association With Acute Chest Syndrome,” Thorax 60, no. 3 (2005): 206–210, 10.1136/thx.2004.029165. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 37. Patterson G. D., Mashegu H., Rutherford J., et al., “Recurrent Acute Chest Syndrome in Pediatric Sickle Cell Disease: Clinical Features and Risk Factors,” Journal of Pediatric Hematology/Oncology 40, no. 1 (2018): 51–55, 10.1097/mph.0000000000001012. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38. Vichinsky E. P., Styles L. A., Colangelo L. H., Wright E. C., Castro O., and Nickerson B., “Acute Chest Syndrome in Sickle Cell Disease: Clinical Presentation and Course,” Blood 89, no. 5 (1997): 1787–1792. [PubMed] [Google Scholar]
- 39. Sysol J. R. and Machado R., “Sickle Cell Disease and Acute Chest Syndrome: Epidemiology, Diagnosis, Management, Outcomes,” in Hematologic Abnormalities and Acute Lung Syndromes (Springer International Publishing, 2016), 67–87. [Google Scholar]
- 40. Payne A. B., Mehal J. M., Chapman C., et al., “Trends in Sickle Cell Disease‐Related Mortality in the United States, 1979 to 2017,” Annals of Emergency Medicine 76, no. 3s (2020): S28–s36, 10.1016/j.annemergmed.2020.08.009. [DOI] [PubMed] [Google Scholar]
- 41. Chaturvedi S. and DeBaun M. R., “Evolution of Sickle Cell Disease From a Life‐Threatening Disease of Children to a Chronic Disease of Adults: The Last 40 Years,” American Journal of Hematology 91, no. 1 (2016): 5–14, 10.1002/ajh.24235. [DOI] [PubMed] [Google Scholar]
- 42. Debaun M. R., Ghafuri D. L., Rodeghier M., et al., “Decreased Median Survival of Adults With Sickle Cell Disease After Adjusting for Left Truncation Bias: A Pooled Analysis,” Blood 133, no. 6 (2019): 615–617, 10.1182/blood-2018-10-880575. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 43. Nwora C. K., Carroll C. P., Prince E. J., and Pecker L., “Mortality Review of Young Adults With Sickle Cell Disease,” Journal of the National Medical Association 117, no. 1 (2025): 18, 10.1016/j.jnma.2025.08.037. [DOI] [Google Scholar]
- 44. Attah A., Balaji D., Njoku M., et al., “Hemoglobin S Beta Thalessemia Plus: Clinical Analysis From the Georgia Comprehensive Sickle Cell Center at Grady Memorial Hospital, Atlanta, Georgia,” Blood 146 (2025): 411, 10.1182/blood-2025-411.40233322 [DOI] [Google Scholar]
- 45. Nelson M., Noisette L., Pugh N., et al., “The Clinical Spectrum of HbSC Sickle Cell Disease‐Not a Benign Condition,” British Journal of Haematology 205, no. 2 (2024): 653–663, 10.1111/bjh.19523. [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.
Supplementary Materials
Table S1: International Classification of Diseases, Clinical Modification using 9th (ICD‐9‐CM) and 10th (ICD‐10‐CM) revisions for all clinical complications.
Table S2: Cumulative incidence of complications among Californians with sickle cell disease by sex.
Figure S1: Cumulative incidence of complications among Californians with sickle cell disease by sex, using age as the scale.
Table S3: Latent Class Analysis (LCA) model fit statistics for (A) Males and (B) Females.
Figure S2: Proportion of SCD individuals classified as administratively defined HbSS across latent classes by sex.
Table S4: Latent Class Analysis (LCA) model fit statistics among Californians with sickle cell disease patients' admissions ≤ 25 years of age.
Data Availability Statement
De‐identified data is available from the corresponding author upon request, subject to a data use agreement where permissible.