Opioid Use Among HIV-Positive Pregnant Women and the Risk for Maternal–Fetal Complications

Ngoc H. Nguyen, PharmD; Erika N. Le, PharmD; Vanessa O. Mbah, PharmD; Emily B. Welsh, PharmD; Rana Daas, BS; Kiara K. Spooner, DrPH, MPH; Jason L. Salemi, PhD, MPH; Omonike A. Olaleye, PhD, MPH; Hamisu M. Salihu, MD, PhD


South Med J. 2020;113(6):292-297. 

In This Article


National Inpatient Sample (NIS) data on pregnancy-related hospital discharges from January 1, 2002 to December 31, 2014 were analyzed. The NIS is part of a family of healthcare databases developed for the Healthcare and Cost Utilization Project (HCUP). HCUP stratifies participating nonfederal community hospitals by five characteristics: rural/urban location, number of beds, geographic region, teaching status, and ownership. A systematic random sampling technique is then used to randomly select 20% of the hospitals from each stratum[17] to ensure the geographic representativeness of the sample. Beginning in 2012, however, the NIS sampling strategy was changed to allow for a sample collected from all of the participating hospitals. For each sampled hospital, all of the inpatient hospitalization records are included in the NIS, and HCUP provides discharge-level sampling weights so that national frequency and prevalence estimates take into account this two-stage cluster sampling design. The NIS contains approximately 7 million inpatient hospitalizations each year (36 million when weighted), and the number has progressively grown as other states join the consortium.

The dataset used in this study encompasses pregnancy-related hospitalizations among women between the ages of 15 and 49 identified using the NEOMAT indicator. The HCUP NEOMAT variable represents maternal diagnosis records on the basis of the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) diagnosis and procedure codes for pregnancy and delivery.[18] To assess the primary exposure of the study population, we scanned ICD-9-CM codes (the principal diagnosis and up to 24 secondary diagnoses) in the discharge record of each woman for an indication of opioid use during pregnancy, HIV diagnosis, and selected maternal–fetal clinical outcomes. We identified HIV-positive pregnant women and opioids users using the following ICD-9-CM codes: for HIV (042, 079.53, and V08) and for opioid use (304.0, 304.7, 305.5, 965.00, 965.01, 965.02, 965.09, E850.0, E850.1, E850.2, E935.0, E935.1, and E935.2).

Individual-level sociodemographic factors reported in the NIS database were extracted and used as part of the analysis. Maternal age in years was classified into three categories: 15 to 24, 25 to 34, and 35 to 49. Self-reported race and ethnicity data were collected under the classifications of non-Hispanic white, non-Hispanic black, Hispanic, or other. Median household income was classified into quartiles based on the patient's ZIP code, serving as a proxy for socioeconomic status. Primary insurance payers were grouped into three categories: government (Medicare/Medicaid), private (commercial carriers, health maintenance organizations, and preferred provider organizations), and other (eg, self-pay, charity). Hospital characteristics also were considered, including US census region (northeast, midwest, south, or west) and hospital type (rural, urban teaching, or urban nonteaching).

Maternal outcomes (and the corresponding ICD-9-CM codes) considered in the study were alcohol use (291, 303, 305.0, 357.5, 425.5, 535.3, 571.0, 571.1, 571.2, 571.3, 760.71, 980.0, E860.0, and V11.3), tobacco use (305.1, 649.0, and 989.84), depression (296.2x, 296.3x, 300.4, 301.12, 309.0, 309.1, and 311), and sepsis (038x, 670.2x, 785.52, 995.91, and 995.92). Fetal complications considered included any abortion, henceforth referred to as "any abortive pregnancy" (630.x, 631.x, 634.x, 637.x, and 635.x–636.x), spontaneous abortion (631.x, 634.x, and 637.x), early onset of labor/delivery (644.2x), and poor fetal growth (656.5x).

Statistical Analysis

We used joinpoint regression to construct the temporal trends in the use of opioids among HIV-positive compared with HIV-negative pregnant women during the period of study. Joinpoint regression is of utility in defining key periods in time that characterize changes in the rate of events over time.[19,20] The iterative model-building process was initiated by fitting the yearly rate data to a straight line with no joinpoints, which assumed a single trend that best described the data points. Then a joinpoint, reflecting a change in the trend, was added to the model, and a Monte Carlo permutation test assessed the improvement in model fit. The process continued until a final model with an optimal (best-fitting) number of joinpoints was selected, with each joinpoint indicating a change in the trend and an annual percentage change estimated to characterize how the rate was changing within each distinct trend segment.

We calculated descriptive statistics, including rates, percentages, and ratios to analyze the relation between opioid use, HIV diagnosis, and maternal and fetal outcomes. Multivariable survey logistic regression then was used to generate adjusted odds ratios that quantified the magnitude of the association between exposure status (opioid use among HIV-positive pregnant women) and the outcomes of spontaneous abortion, any abortive pregnancy, early-onset delivery, and poor fetal growth. Statistical analyses were performed with SAS version 9.4 (SAS Institute, Cary, NC); we assumed a 5% type I error rate for all of the hypothesis tests (two-sided). Because of the deidentified, publicly available nature of NIS data, the analyses performed for this study were considered exempt by the Baylor College of Medicine institutional review board. It is pertinent to mention that some of the results of our analyses contained small numbers that must be suppressed in accordance with guidelines set forth by HCUP. The reason is to prevent possible identification of these individuals, and in such cases, we described the findings in text without displaying the actual values.