Introduction
Recommendations for HIV screening of pregnant women and treatment of pregnant women with HIV infection have led to a dramatic decrease in the rate of mother-to-child transmission of HIV to <2% in the United States. Recent US guidelines recommend combination antiretroviral (ARV) therapy (cART) as the standard of care for all HIV-infected individuals, and such therapy is particularly critical for pregnant women. With the development and widespread use of cART, HIV-infected women are living longer, healthier lives. From 2000 through 2006, estimated rates of childbearing increased among HIV-infected women and more such women desired pregnancy.
Despite the benefits of cART for the treatment and prevention of HIV, there are concerns about potential adverse effects to the pregnant woman, fetus, and newborn. ARV drugs have been linked with hyperglycemia, diabetes mellitus, lactic acidosis, hepatic steatosis, and renal and hepatic toxicity, although studies of their effects in pregnancy have generated conflicting results. Some studies have shown an increased risk for preeclampsia among HIV-infected women receiving ARV drugs compared to HIV-uninfected women, but these findings have been inconsistent. In addition, there are conflicting findings on ARV drugs and risk of preterm delivery. Many of these studies of pregnancy outcomes among women with HIV infection are limited by relatively small sample size, retrospective design, and no adjustment for possible confounding factors.
In a previous nationwide analysis of trends of hospitalizations among HIV-infected pregnant women in the United States from 1994 through 2003, we described higher rates of morbidity and adverse obstetric outcomes among HIV-infected, compared with HIV-uninfected, pregnant women. Of note, during that time, there was no increase in the rate of preterm labor, preterm delivery, or preeclampsia/hypertension (HTN) among pregnant HIV-infected women in the United States. Since this analysis was published, additional ARVs have become available, recommended first-line cART regimens during pregnancy have changed, and most HIV-infected women who know their status now receive cART prenatally and during labor and delivery. Given these changes, the objectives of our analysis were to use more recent data from the Healthcare Cost and Utilization Project (HCUP), obtained during a period of widespread cART availability, to examine trends of adverse outcomes that could be related to HIV infection or its treatments among pregnant women in the United States. Monitoring these trends is essential, as recommendations for this special population continue to evolve with the progress of HIV treatment and prevention options.
Materials and Methods
We obtained US hospital discharge data from the HCUP Nationwide Inpatient Sample (NIS). HCUP is a family of health care databases and related software tools sponsored by the Agency for Healthcare Research and Quality in partnership with state-level data collection organizations. The NIS is a 20%-stratified probability sample of US community hospitals defined by the American Hospital Association as nonfederal, short-term (average length of stay <30 days) general and specialty hospitals with facilities open to the public. Hospitals were sampled based on 5 strata: rural/urban location, number of beds, region, teaching status, and ownership. In each year spanning 2004 through 2011, NIS included inpatient stays from about 1000 hospitals with 7-8 million discharge records annually that were weighted to enable production of national estimates. The NIS is one of the largest collections of all-payer inpatient data and includes hospital charge and length-of-stay data. Discharge data include International Classification of Diseases, Ninth Revision, Clinical Modification ( ICD-9-CM ) diagnosis and procedure codes, as well as Diagnosis-Related Groups and Clinical Classification Software (CCS) diagnostic codes for ICD-9-CM , a uniform and standardized diagnosis and procedure categorization scheme.
We analyzed NIS data from 2004 through 2011, a period in which the number of states included in the NIS increased from 37-46. Our analysis was restricted to antenatal and delivery pregnancy hospitalizations. The women referred to in this study include girls and women 15–49 years of age. We identified hospitalizations of pregnant women ( ICD-9-CM codes 640-677, V22-V24, V27-V28, 792.3), and then identified those resulting in delivery ( ICD-9-CM codes 650, V27; ICD-9-CM procedure codes 72.0-72.9, 73.22, 73.59, 73.6, 74.0-74.2, 74.4, 74.99; and Diagnosis-Related Group codes 370-375). We excluded those for ectopic/molar pregnancy or spontaneous abortion ( ICD-9-CM codes 630, 631, 633, 632, 634-639; procedure codes 69.01, 69.51, 74.91, 75.0).
The ICD-9-CM codes used in this analysis to identify hospitalizations complicated by preterm labor or rupture of membranes, preterm delivery, gestational diabetes mellitus (GDM), preeclampsia/HTN disorders of pregnancy, antepartum hemorrhage, major puerperal infection, and HIV have been described previously. The definitions of certain morbidities and outcomes used in the previous analysis were modified as follows: (1) liver disorders of pregnancy, including hepatic steatosis, lactic acidosis, and elevated liver function tests (276.2, 571.8, 646.7, 790.4, 573.3, 573.9); (2) sexually transmitted infections and pelvic inflammatory disease (647.0-647.2, 090.0-099.9, 054.5, 054.7, 079.88, 614.0-614.5, 614.7-614.9, 615, 616); (3) bacterial infections (CCS codes 2, 3); (4) viral, mycoses, parasitic, and other infections (CCS codes 4, 6, 7, 8, 77, 78, 126); (5) influenza and pulmonary complications (480-488); and (6) urinary tract infections (646.6, 590, 595, 597, 599.0). Finally, we included the following conditions for use as adjustment variables: anemia (648.2, 280-285), hepatitis B and C (070.2-070.3, 070.41, 070.42, 070.44, 070.51, 070.52, 070.54, 070.7), drug use (304, 305.2-305.9), alcohol abuse (303.0, 303.9, 305.0, 291.0, 291.4, 291.81), and smoking (305.1, 649.0).
We used Student t test and the Wald χ 2 test with a significance level of .05 to compare demographic and hospital characteristics and outcomes, as well as length of hospitalization and hospital charges, of hospitalized HIV-infected vs HIV-uninfected pregnant women. We compared the following patient or hospital characteristics: age distribution, expected primary payer, median household income quartile based on patient ZIP code, urban/rural hospital, geographic region of the hospital, total and mean length of hospital stay, total and mean hospital charges for hospitalization, and inpatient mortality rate. All hospital charge data are adjusted for inflation and presented in 2011 dollars.
For 2004 through 2011, we compared odds of adverse outcomes among hospitalizations of HIV-infected, vs HIV-uninfected, women using multivariable logistic regression, adjusting for demographic variables and relevant comorbidities. To assess the trend in absolute number of delivery hospitalizations for HIV-infected women per year in the study period, we used a weighted least squares regression designed specifically for use with survey data. To assess trends in outcomes, we compared data from 2007 through 2011 with that from 2004 (reference year), adjusting for maternal age, expected payer, and delivery status in logistic or linear regression models, according to outcome type. Models with interaction terms were run to assess whether the magnitude of changes over time differed according to HIV status. Subanalyses were conducted to examine whether the associations between HIV and outcomes–including hospital charges, length of stay, and bacterial infection–were modified by hospitalization type (delivery or antenatal). All statistical analyses were conducted with survey procedures using software (SAS, Version 9.3; SAS Institute, Cary, NC). As a secondary analysis of a deidentified data set, this study was exempt from institutional review board approval.
Materials and Methods
We obtained US hospital discharge data from the HCUP Nationwide Inpatient Sample (NIS). HCUP is a family of health care databases and related software tools sponsored by the Agency for Healthcare Research and Quality in partnership with state-level data collection organizations. The NIS is a 20%-stratified probability sample of US community hospitals defined by the American Hospital Association as nonfederal, short-term (average length of stay <30 days) general and specialty hospitals with facilities open to the public. Hospitals were sampled based on 5 strata: rural/urban location, number of beds, region, teaching status, and ownership. In each year spanning 2004 through 2011, NIS included inpatient stays from about 1000 hospitals with 7-8 million discharge records annually that were weighted to enable production of national estimates. The NIS is one of the largest collections of all-payer inpatient data and includes hospital charge and length-of-stay data. Discharge data include International Classification of Diseases, Ninth Revision, Clinical Modification ( ICD-9-CM ) diagnosis and procedure codes, as well as Diagnosis-Related Groups and Clinical Classification Software (CCS) diagnostic codes for ICD-9-CM , a uniform and standardized diagnosis and procedure categorization scheme.
We analyzed NIS data from 2004 through 2011, a period in which the number of states included in the NIS increased from 37-46. Our analysis was restricted to antenatal and delivery pregnancy hospitalizations. The women referred to in this study include girls and women 15–49 years of age. We identified hospitalizations of pregnant women ( ICD-9-CM codes 640-677, V22-V24, V27-V28, 792.3), and then identified those resulting in delivery ( ICD-9-CM codes 650, V27; ICD-9-CM procedure codes 72.0-72.9, 73.22, 73.59, 73.6, 74.0-74.2, 74.4, 74.99; and Diagnosis-Related Group codes 370-375). We excluded those for ectopic/molar pregnancy or spontaneous abortion ( ICD-9-CM codes 630, 631, 633, 632, 634-639; procedure codes 69.01, 69.51, 74.91, 75.0).
The ICD-9-CM codes used in this analysis to identify hospitalizations complicated by preterm labor or rupture of membranes, preterm delivery, gestational diabetes mellitus (GDM), preeclampsia/HTN disorders of pregnancy, antepartum hemorrhage, major puerperal infection, and HIV have been described previously. The definitions of certain morbidities and outcomes used in the previous analysis were modified as follows: (1) liver disorders of pregnancy, including hepatic steatosis, lactic acidosis, and elevated liver function tests (276.2, 571.8, 646.7, 790.4, 573.3, 573.9); (2) sexually transmitted infections and pelvic inflammatory disease (647.0-647.2, 090.0-099.9, 054.5, 054.7, 079.88, 614.0-614.5, 614.7-614.9, 615, 616); (3) bacterial infections (CCS codes 2, 3); (4) viral, mycoses, parasitic, and other infections (CCS codes 4, 6, 7, 8, 77, 78, 126); (5) influenza and pulmonary complications (480-488); and (6) urinary tract infections (646.6, 590, 595, 597, 599.0). Finally, we included the following conditions for use as adjustment variables: anemia (648.2, 280-285), hepatitis B and C (070.2-070.3, 070.41, 070.42, 070.44, 070.51, 070.52, 070.54, 070.7), drug use (304, 305.2-305.9), alcohol abuse (303.0, 303.9, 305.0, 291.0, 291.4, 291.81), and smoking (305.1, 649.0).
We used Student t test and the Wald χ 2 test with a significance level of .05 to compare demographic and hospital characteristics and outcomes, as well as length of hospitalization and hospital charges, of hospitalized HIV-infected vs HIV-uninfected pregnant women. We compared the following patient or hospital characteristics: age distribution, expected primary payer, median household income quartile based on patient ZIP code, urban/rural hospital, geographic region of the hospital, total and mean length of hospital stay, total and mean hospital charges for hospitalization, and inpatient mortality rate. All hospital charge data are adjusted for inflation and presented in 2011 dollars.
For 2004 through 2011, we compared odds of adverse outcomes among hospitalizations of HIV-infected, vs HIV-uninfected, women using multivariable logistic regression, adjusting for demographic variables and relevant comorbidities. To assess the trend in absolute number of delivery hospitalizations for HIV-infected women per year in the study period, we used a weighted least squares regression designed specifically for use with survey data. To assess trends in outcomes, we compared data from 2007 through 2011 with that from 2004 (reference year), adjusting for maternal age, expected payer, and delivery status in logistic or linear regression models, according to outcome type. Models with interaction terms were run to assess whether the magnitude of changes over time differed according to HIV status. Subanalyses were conducted to examine whether the associations between HIV and outcomes–including hospital charges, length of stay, and bacterial infection–were modified by hospitalization type (delivery or antenatal). All statistical analyses were conducted with survey procedures using software (SAS, Version 9.3; SAS Institute, Cary, NC). As a secondary analysis of a deidentified data set, this study was exempt from institutional review board approval.
Results
The number of hospitalizations and deliveries among HIV-infected pregnant women did not increase over the study period (trend test P value: .569). There were an estimated 7107 hospitalizations of pregnant HIV-infected women in 2004, 6837 in 2007, and 4751 in 2011. The 2011 estimate is 33% lower than in 2004; in comparison, the corresponding estimate for HIV-uninfected women was 11% lower in 2011 than in 2004. There were an estimated 5339 delivery hospitalizations for pregnant HIV-infected women in 2004, 5397 in 2007, and 3855 in 2011. The 2011 estimate is 28% lower than in 2004; the corresponding estimate for HIV-uninfected women was 9% lower in 2011 than in 2004.
From 2004 through 2011, hospitalizations of HIV-infected pregnant women represented an estimated 0.11-0.15% of all hospitalizations of pregnant women in the United States. The number of hospitalizations during pregnancy per delivery remained higher among HIV-infected women compared with HIV-uninfected women in both 2004 and 2011 (1.33 vs 1.11 in 2004 [ P < .001]; 1.23 vs 1.10 in 2011 [ P < .001]) ( Table 1 ). Hospitalizations of pregnant HIV-infected women were longer (mean length, 3.64 vs 2.68 days in 2011, P < .001) and incurred higher hospital charges (mean charges, $22,980 vs $14,362 in 2011, P < .001) compared to HIV-uninfected women in both 2004 and 2011 ( Table 1 ). A subanalysis by hospitalization type (antenatal/delivery) showed that mean hospital charges were higher and mean length of stay was longer for both delivery and antenatal hospitalizations of HIV-infected, compared with HIV-uninfected, pregnant women. HIV infection did not have a significantly different effect on length of stay or hospital charges by hospitalization type (antenatal vs delivery) (data not shown).
| Patient/hospital characteristics | 2004 | 2011 | ||||
|---|---|---|---|---|---|---|
| HIV infected, n = 7107 | HIV uninfected, n = 4,675,615 | P a | HIV infected, n = 4751 | HIV uninfected, n = 4,180,200 | P a | |
| Age, y | N (%) | N (%) | N (%) | N (%) | ||
| 15–19 | 456 (6.4) | 488,041 (10.4) | .007 | 245 (5.2) | 359,049 (8.6) | .004 |
| 20–34 | 5622 (79.1) | 3,497,064 (74.8) | 3631 (76.4) | 3,210,178 (76.8) | ||
| 35–49 | 1030 (14.5) | 690,509 (14.8) | 875 (18.4) | 610,973 (14.6) | ||
| Expected primary payer | ||||||
| Public | 5155 (73.7) | 1,886,156 (40.4) | <.001 | 3500 (73.8) | 1,865,653 (44.7) | <.001 |
| Private | 1345 (19.2) | 2,514,553 (53.9) | 916 (19.3) | 2,080,438 (49.9) | ||
| Other | 496 (7.1) | 266,189 (5.7) | 326 (6.9) | 226,869 (5.4) | ||
| Median household income quartile for patient ZIP code | ||||||
| 1–Lowest income | 4052 (58.3) | 1,300,572 (28.3) | <.001 | 1905 (49.0) | 1,122,647 (27.3) | <.001 |
| 2 | 1455 (20.9) | 1,120,376 (24.4) | 910 (23.4) | 1,003,369 (24.4) | ||
| 3 | 920 (13.2) | 1,015,807 (22.1) | 699 (18.0) | 1,097,657 (26.7) | ||
| 4–Highest income | 526 (7.6) | 1,160,767 (25.2) | 374 (9.6) b | 894,705 (21.7) | ||
| Location of hospital | ||||||
| Rural | 320 (4.5) | 525,432 (11.2) | <.001 | 168 (3.6) | 474,446 (11.5) | <.001 |
| Urban | 6788 (95.5) | 4,150,183 (88.8) | 4491 (96.4) | 3,640,248 (88.5) | ||
| Region | ||||||
| Northeast | 1640 (23.1) | 822,720 (17.6) | <.001 | 1422 (29.9) | 690,264 (16.5) | <.001 |
| Midwest | 609 (8.6) b | 952,302 (20.4) | 712 (15.0) | 858,384 (20.5) | ||
| South | 4580 (64.4) | 1,703,296 (36.4) | 2457 (51.7) | 1,580,577 (37.8) | ||
| West | 278 (3.9) b | 1,197,296 (25.6) | 159 (3.4) | 1,050,974 (25.1) | ||
| Died during hospitalization | 0 (0) | 565 (0.01) | – | – c | 494 (0.01) | .182 |
| Mean length of stay, d | 3.63 | 2.68 | <.001 | 3.64 | 2.68 | <.001 |
| Mean hospital charges | $14,877 | $10,758 | <.001 | $22,980 | $14,362 | <.001 |
| Hospitalization/delivery ratio | 1.33 | 1.11 | <.001 | 1.23 | 1.10 | <.001 |
b Estimate should be used with caution-relative SE is >30%
Hospitalizations of HIV-infected pregnant women were more likely than those of HIV-uninfected pregnant women to be among older women ( P = .004 in 2011), be covered by public insurance (eg, Medicaid or Medicare) ( P < .001), have listed residences in ZIP codes with lower median household incomes ( P < .001), occur in urban hospitals ( P < .001), and occur in the South and Northeast ( P < .001). Frequency of death during hospitalization was not different among HIV-infected and HIV-uninfected pregnant women in 2011 ( P = .182). No statistical comparison was possible for deaths during hospitalization in 2004 given no recorded deaths among HIV-infected pregnant women. Following the general trends seen in the HIV-uninfected group, mean hospital charges increased over the study period for the HIV-infected group, while length of stay did not change from 2004 through 2011.
Adverse outcomes among HIV-infected pregnant women
The odds of a diagnosis of several of the adverse outcomes examined remained significantly higher among hospitalizations of HIV-infected pregnant women compared to HIV-uninfected pregnant women during 2004 through 2011 ( Table 2 ). In 2011, this included preterm labor and preterm delivery, preeclampsia and HTN disorders of pregnancy, sexually transmitted infections and pelvic inflammatory disease, bacterial infections, nonbacterial infections (including viral infections, mycoses, and parasitic infections), and influenza. A subanalysis showed that the effect of HIV infection status on the odds of bacterial infections was modified by hospitalization type (delivery/antenatal) ( P < .001). Among antenatal hospitalizations, the odds of bacterial infections did not differ by HIV status. However, among delivery hospitalizations, odds of bacterial infection were significantly higher for HIV-infected, compared with HIV-uninfected, pregnant women (adjusted odds ratio [aOR], 2.36; 95% confidence interval [CI], 1.71–3.25). In 2004, the odds of GDM were significantly lower among pregnancy hospitalizations for HIV-infected women compared to HIV-uninfected women. However, the rate of GDM in pregnancy hospitalizations of HIV-infected women increased significantly in 2011, such that there was no difference in the odds of GDM by HIV status in 2011. The odds of antepartum hemorrhage in hospitalizations of HIV-infected pregnant women, compared to hospitalizations of HIV-uninfected pregnant women, were significantly lower in 2004 but borderline higher in 2011. There was no difference in the odds of liver disorders according to HIV status in either year studied.
