Study population

This retrospective cohort study was conducted using the Nationwide Readmissions Database (NRD), from the United States Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project. The NRD is an all-payor administrative database designed to represent short-stay inpatient admissions in the United States. A total of 28 states participated in the NRD in 2020, including 59.7% of the population and 58.7% of hospitalizations. Weighting and stratification variables are provided in the NRD dataset to allow estimating national rates of hospitalizations using the NRD sample. The data include patient demographics, hospital characteristics, International Classification of Diseases Diagnosis and Procedure codes, discharge disposition, length of stay, and inpatient charges. While it is possible to follow readmissions for a patient, we limited analyses to the delivery hospitalization.

Deliveries at or beyond 24 weeks gestation with hospital discharges between 10/2015 and 12/2020 were included. The gestational age of 24 weeks was chosen because it is traditionally considered the limit of viability. Pregnant patients were identified based on procedural coding with the International Classification of Diseases 10^th revision, Clinical Modification (ICD-10-CM) (S1 Table). The starting date was selected given this was when ICD-10 was adopted for clinical use in the United States. Weeks of gestation was determined using the Z3A.xx series of codes. Diagnosis of HCV was identified for each patient using the ICD-10-CM codes for carrier of viral hepatitis C (Z22.52), chronic viral hepatitis C (B18.2), unspecified viral hepatitis C without hepatic coma (B19.20), or unspecified viral hepatitis C with coma (B19.21). Diagnosis of ICP was determined by whether patients had ICD-10 codes for both liver and biliary tract disorder in pregnancy (O26.6*) and obstruction of bile duct (K83.1). We consulted with an experienced inpatient obstetrics coder at our institution in developing these criteria.

Outcomes

The primary outcome for this analysis was occurrence of severe maternal morbidity (SMM) during the delivery hospitalization, based on criteria and the ICD-10 codes published by the Centers for Disease Control and Prevention (CDC) [11]. Our secondary outcomes included acute respiratory distress, acute kidney injury, sepsis, GDM, cesarean delivery, preterm birth defined as birth at less than 37 weeks gestational age, and hospital length of stay.

Statistical analysis

The NRD dataset includes weighting and stratification variables to allow for estimating national rates of hospitalizations. For comparing demographics, comorbidities, hospital characteristics, and SMM in each patient cohort, weighted linear regressions were used for continuous variables and weighted chi-square testing was used for binary and categorical variables as appropriate. Weighted logistic (all outcomes except length of stay) and negative binomial (length of stay) regression analyses were used to evaluate the association between HCV and ICP status and outcomes. Models were adjusted for age, primary payer, median household income by ZIP code, hospital type and size, year, discharge quarter, and clinical co-morbidities as identified by the expanded obstetric comorbidity index [12]. Missing values were only present for the ZIP code income (0.4%) and primary payer (0.06%) variables; given the low prevalence of missing values and complications of a weighted dataset, modal value imputation was performed. We assessed for synergistic or antagonistic effects by including an interaction term between ICP and HCV, which was retained in the final statistical models if it was statistically significant for the given outcome. Statistical analyses were performed in SAS Version 9.4 (SAS Institute, Cary, NC) and Stata Statistical Software version 16.1 (StataCorp, College Station, Texas). A two-sided alpha level of 0.05 was pre-specified as statistically significant. Because the NRD dataset is a Limited Data Set as defined by the Health Insurance Portability and Accountability Act, this study was ruled exempt from review by the Duke University Health Institutional Review Board (Pro00106911). As a Limited Data Set, the NRD dataset does not provide information sufficient to contact and obtain informed consent from participants (and in fact this would be prohibited by the Data Use Agreement allowing access to the data) and thus informed consent was not obtained.

Impact of HCV on maternal and obstetric outcomes

There is mixed data on the association between HCV and preterm birth, Our findings that HCV is associated with an increased odds of preterm birth and cesarean section are consistent with prior studies that have shown an increased likelihood of preterm birth in this patient population [13,14], Another study demonstrated no association between HCV and preterm birth, in contrast to prior literature. Authors suggest that this discrepancy may be due to differences in definition of preterm birth used or use of chart abstraction for measurement of HCV as opposed to direct diagnostic testing [15]. Additionally, some prior studies have also shown that patients with HCV were more likely to undergo cesarean section [13]; however, cesarean section is not specifically recommended in patients with HCV as it has not been shown to decrease perinatal transmission [16].

With respect to maternal morbidity, extant literature suggests an association between HCV and hypertensive disorders of pregnancy [14,17], but previous studies have not evaluated overall SMM, Thus, our observation of the correlation between HCV and increased odds of overall SMM is novel. Interestingly, in our cohort HCV was associated with a decreased odds of GDM. We know that HCV interferes with the insulin-signaling pathway through multiple mechanisms, can infect the beta islet cells, is associated with circulating micro-RNAs, and has been linked to type II diabetes mellitus [18]. Thus, it has been proposed that HCV infection could predispose patients to developing GDM [17,19]. Despite that proposed mechanism, findings in the literature are mixed. Reddick et al [20] queried the US Nationwide Inpatient Sample, and found increased adjusted odds for GDM among those with HCV. Notably, their data was from 1995–2005, so differences in coding or practice patterns may have contributed to the difference in outcomes. In contrast, a more recent inquiry of the Nationwide Inpatient Sample by Chen et al. using data from 2012–2018 found results in line with our study, observing lower odds of GDM among patients with HCV infection [14]. As our results add to an already heterogeneous picture, further epidemiologic and mechanistic study of HCV and GDM is needed.

Impact of ICP on maternal outcomes

Although ICP has a well-established association with increased risk of poor fetal outcomes, it has been found to be a generally benign disease in terms of maternal outcomes. However, in some cases, especially in those with genetic alterations in bile composition and familial intrahepatic cholestasis of pregnancy, patients have a progressive course with chronic cholestasis, cholelithiasis portal hypertension, and liver failure [21]. It has also been associated with possible potentiation of the development of GDM through alterations in glucose homeostasis [22]. We found that there was also an increased odds of acute kidney injury, but there have been no studies, to our knowledge, that investigate or demonstrate a relationship between ICP and acute kidney injury. One case study described a patient who had ICP secondary to acute renal insufficiency in the setting of pyelonephritis; however, further research is needed to identify the mechanism behind this process [23]. While this may be attributed to under-ascertainment of cofounders that would lead to increased odds of kidney disease, current research has largely focused on fetal and neonatal outcomes as a result of ICP, and further research is needed to investigate the ICP effects on SMM, specifically whether there may be a relationship between ICP and kidney disease.

Additive effects of ICP and HCV on maternal and obstetric outcomes

We found that having both HCV and ICP leads to an additive increased odds of SMM, preterm birth, and increased length of stay when compared to having either HCV, ICP, or neither. This is an important finding given that HCV is associated with increased odds of ICP. Many physiologic mechanisms have been proposed for how HCV leads to predisposition to ICP, with some research suggesting HCV impairs bile transportation, [24] potentially facilitating the occurrence of ICP at an earlier gestation age [8]. Moreover, certain mutations of biliary transporters, specifically the ABCB11 genotype, increase bile acids in infected patients [25] and this sub-group of HCV patients may be even more susceptible to ICP. Although the exact mechanism underlying the association between HCV and ICP has yet to be demonstrated, this study demonstrates that the mechanism is likely one that leads to additive effects rather than synergistic effects.

Strengths and limitations

Our study is strengthened by the large, all-payor nature of the source dataset. Given that we are limited in what variables we can use by the database, we cannot know the serotype of HCV to stratify outcomes by serotype, or bile acid levels to characterize the severity of ICP. It was also impossible to distinguish whether the patients received treatment for either disease that could have altered outcomes. However, the inability to control for these factors likely makes our results generalizable to patients with all types of HCV and severities of ICP. There are several limitations to administrative databases including reliability of coding data and lack of adequate control variables [26], but we are reassured on the former in that the rates of HCV and ICP are similar to the published literature in this cohort, as the reported prevalence of HCV in the Maternal Fetal Medicine Units Network is 0.3%. Given that HCV alone can cause cholestasis, it is possible some of the patients identified as having both HCV and ICP were misdiagnosed with ICP and rather had elevated bile acids and cholestasis attributable to HCV alone. This is a particular limitation in this dataset, as we cannot access clinical data beyond ICD-10 codes to assess how these diagnoses were made or if cholestasis resolved following pregnancy, as it would be expected were it caused by ICP.

While the models were adjusted for confounding factors using the Leonard comorbidity index, it is possible that HCV in this model is more of a “risk marker” rather than “risk predictor,” as there are many social comorbidities that are strongly correlated with HCV, but undercoded in administrative data. It is also important that we clarify the nature of an additive effect in a logistic regression model, as was used here. In a logistic regression, covariates have a constant additive effect on the natural log of the odds of the event (or alternatively, a constant multiplicative effect on the odds of the event). We selected logistic regression given its known good performance properties for binary outcomes, whether common or rare, but it is possible that an alternative regression parameterization would produce different results. Finally, although ICP is known to increase risk of stillbirth, we were unable to assess this outcome well using this dataset. Because ICP management in the United States is designed to potentially reduce the rates of stillbirth through antenatal testing and planned preterm and early term deliveries, it would be challenging to ascertain whether differences present reflect underlying biological differences or differences in processes of care.

Conclusions

This study demonstrates that having both HCV and ICP leads to an increased odds of SMM, preterm birth, and increased length of stay when compared to having either HCV, ICP, or both. These data may be useful in counseling patients regarding their increased risk of adverse outcomes when ICP presents in association with HCV versus ICP alone.