Materials and Methods
We used data that were collected between 2008 and 2010 for the Healthcare Cost and Utilization Project’s State Inpatient Database (SID). Forty-six states participate in SID, and together, they encompass approximately 97% of all annual hospital discharges in the United States. The SID contains the universe of inpatient discharges in participating states and a uniform set of variables to facilitate multistate comparisons. We selected all states with >80,000 annual live births (ie, the mean number of 2008-2010 births in the 50 states and District of Columbia from 2008-2010) and race and ethnicity data that were reported to SID for at least 70% of the state’s annual live births. The 7 states that met these criteria were Arizona, California, Florida, Michigan, New Jersey, New York, and North Carolina. Because this analysis involved the use of publicly available data that lacked direct personal identifiers, it was determined to be research that did not involve human subjects.
Our analytic sample consisted of 2008-2010 delivery hospitalizations in Arizona, California, Florida, Michigan, and New Jersey; 2008 and 2009 delivery hospitalizations in New York (2010 data were not available at the time of analysis and manuscript development), and 2010 delivery hospitalizations in North Carolina (race/ethnicity data were not reported for 2008 and 2009). Delivery hospitalizations were identified with the use of a previously published International Classification of Diseases, 9th Revision (ICD-9-CM)–based algorithm that incorporates diagnostic codes for an outcome of delivery, diagnosis-related group delivery codes, and procedure codes for selected delivery-related procedures. To ensure consistency among states, all delivery hospitalization records were restricted to the first 15 ICD-9-CM diagnoses and 15 ICD-9-CM procedures that were listed. We excluded hospitals with <30 deliveries in a given year, hospitals for which >50% of delivery records lacked data on race/ethnicity, and remaining records with a missing or invalid value for the uniform race/ethnicity variable. These exclusions yielded a total analytic sample of 3,476,392 delivery hospitalizations that represented 88.9-95.3% of all state- and year-specific live births in Arizona, California, Florida, New Jersey, New York, and North Carolina and 72.3-72.9% of year-specific live births in Michigan ( Table 1 ).
| State | Year | ||||||||
|---|---|---|---|---|---|---|---|---|---|
| 2008 | 2009 | 2010 | |||||||
| Live births (a) | Deliveries in dataset (b) | Ratio (a:b) | Live births (a) | Deliveries in dataset (b) | Ratio (a:b) | Live births (a) | Deliveries in dataset (b) | Ratio (a:b) | |
| Arizona | 99,442 | 92,860 | 93.4 | 92,798 | 87,048 | 93.8 | 87,477 | 82,156 | 93.9 |
| California | 551,779 | 503,663 | 91.3 | 527,020 | 481,049 | 91.3 | 510,198 | 463,362 | 90.8 |
| Florida | 231,445 | 219,641 | 94.9 | 221,394 | 210,894 | 95.3 | 214,590 | 203,659 | 94.9 |
| Michigan | 121,127 | 87,544 | 72.3 | 117,294 | 85,510 | 72.9 | 114,531 | 82,960 | 72.4 |
| New Jersey | 112,710 | 104,216 | 92.5 | 110,331 | 102,237 | 92.7 | 106,922 | 99,958 | 93.5 |
| New York | 250,383 | 231,121 | 92.3 | 248,110 | 229,694 | 92.6 | N/A | ||
| North Carolina | N/A | N/A | 122,350 | 108,820 | 88.9 | ||||
Delivery hospitalizations with severe morbidity were identified with condition-specific ICD-9-CM codes as per the algorithm proposed by Callaghan et al ; algorithm development methods are explained elsewhere. This algorithm was created based on previously proposed conceptual models for the identification of severe maternal morbidity in the United States and is largely based on indicators of organ system failure, thus in line with practices currently supported by the World Health Organization. In their recent analysis of severe maternal morbidity in the United States, Callaghan et al showed that blood transfusion was the most common indicator of severe morbidity and that rates of blood transfusion were considerably higher than rates that were calculated for the other 24 indicators of severe morbidity in their algorithm (list provided in Table 2 ). The ICD-9-CM code for blood transfusion does not include information about the number of units transfused; hence, the inclusion of women whose only indication of severe morbidity was a code for transfusion may inflate the rate of severe morbidity. Therefore, we examined both the rates of severe morbidity using women who had at least 1 indication of severe morbidity that included blood transfusion as the numerator and rates of severe morbidity with the use of women who had at least 1 indication of severe morbidity but without blood transfusion, thus excluding deliveries with blood transfusion as the only indicator of severe morbidity.
| Severe maternal morbidity indicator | Non-Hispanic White (n = 1,485,280) | Non-Hispanic Black (n = 434,431) | Hispanic (n = 1,140,592) | Asian/Pacific Islander (n = 247,852) | American Indian/Alaska Native (n = 20,535) | All a (N = 3,476,392) |
|---|---|---|---|---|---|---|
| Blood transfusion | 78.94 ± 0.73 | 187.03 ± 2.06 b | 104.31 ± 0.95 b | 97.92 ± 1.98 b | 168.49 ± 8.98 b | 104.40 ± 0.55 |
| Disseminated intravascular coagulation | 12.81 ± 0.29 | 21.96 ± 0.71 b | 14.11 ± 0.35 b | 16.10 ± 0.81 b | 17.53 ± 2.92 b | 14.84 ± 0.21 |
| Heart failure during procedure or surgery | 8.93 ± 0.25 | 23.87 ± 0.74 b | 10.51 ± 0.30 b | 10.89 ± 0.66 b | 10.71 ± 2.28 | 11.69 ± 0.18 |
| Hysterectomy | 7.64 ± 0.23 | 13.12 ± 0.55 b | 9.43 ± 0.29 b | 12.67 ± 0.71 b | 7.31 ± 1.89 | 9.36 ± 0.16 |
| Cardio monitoring | 1.42 ± 0.10 | 29.28 ± 0.82 b | 5.67 ± 0.22 b | 3.07 ± 0.35 b | 8.28 ± 2.01 b | 7.67 ± 0.15 |
| Ventilation | 5.44 ± 0.19 | 15.54 ± 0.60 b | 6.90 ± 0.25 b | 7.30 ± 0.54 b | 12.66 ± 2.48 b | 7.50 ± 0.15 |
| Operations on heart and pericardium | 7.11 ± 0.20 | 9.55 ± 0.47 b | 6.69 ± 0.24 | 8.15 ± 0.57 | 9.74 ± 2.18 | 7.36 ± 0.15 |
| Adult respiratory distress syndrome | 5.82 ± 0.20 | 13.44 ± 0.56 b | 6.26 ± 0.23 | 6.70 ± 0.52 | 11.20 ± 2.33 b | 7.10 ± 0.14 |
| Eclampsia | 3.63 ± 0.16 | 10.36 ± 0.49 b | 5.44 ± 0.22 b | 3.23 ± 0.36 | 7.31 ± 1.89 | 5.15 ± 0.12 |
| Acute renal failure | 4.07 ± 0.17 | 11.65 ± 0.52 b | 3.67 ± 0.18 | 4.80 ± 0.44 | 10.23 ± 2.23 b | 5.04 ± 0.12 |
| Shock | 2.71 ± 0.14 | 4.83 ± 0.33 b | 3.09 ± 0.16 | 5.04 ± 0.45 b | 6.82 ± 1.82 b | 3.33 ± 0.10 |
| Sepsis | 2.56 ± 0.13 | 5.96 ± 0.37 b | 3.26 ± 0.17 b | 2.95 ± 0.35 | 5.84 ± 1.78 | 3.31 ± 0.10 |
| Puerperal cerebrovascular disorders | 1.90 ± 0.11 | 3.59 ± 0.29 b | 2.07 ± 0.14 | 1.65 ± 0.26 | N/A | 2.16 ± 0.08 |
| Pulmonary edema | 1.32 ± 0.09 | 2.60 ± 0.25 b | 1.17 ± 0.10 | 2.46 ± 0.32 b | N/A | 1.55 ± 0.07 |
| Thrombotic embolism | 1.29 ± 0.10 | 3.38 ± 0.28 b | 1.11 ± 0.10 | 0.81 ± 0.18 | N/A | 1.46 ± 0.06 |
| Severe anesthesia complications | 0.82 ± 0.07 | 1.96 ± 0.21 b | 1.02 ± 0.10 | 0.69 ± 0.17 | N/A | 1.01 ± 0.05 |
| Conversion of cardiac rhythm | 0.81 ± 0.07 | 1.82 ± 0.21 b | 0.93 ± 0.09 | 0.97 ± 0.20 | N/A | 0.96 ± 0.05 |
| Sickle cell anemia with crisis | N/A | 6.91 ± 0.40 | N/A | N/A | N/A | 0.95 ± 0.05 |
| Cardiac arrest/ventricular fibrillation | 0.57 ± 0.06 | 1.50 ± 0.19 b | 0.67 ± 0.08 | 0.53 ± 0.15 | N/A | 0.72 ± 0.05 |
| Internal injuries of thorax, abdomen, and pelvis | 0.39 ± 0.05 | 0.92 ± 0.15 b | 0.47 ± 0.06 | N/A | N/A | 0.49 ± 0.04 |
| Amniotic fluid embolism | 0.28 ± 0.04 | 0.65 ± 0.12 b | 0.37 ± 0.06 | 0.85 ± 0.19 | N/A | 0.40 ± 0.03 |
| Temporary tracheostomy | 0.26 ± 0.04 | 0.69 ± 0.13 b | 0.31 ± 0.05 | N/A | N/A | 0.31 ± 0.03 |
| Acute myocardial infarction | 0.20 ± 0.04 | 0.46 ± 0.10 | 0.20 ± 0.04 | N/A | N/A | 0.25 ± 0.03 |
| Aneurysm | 0.10 ± 0.03 | N/A | 0.08 ± 0.03 | N/A | N/A | 0.10 ± 0.02 |
| Intracranial injuries | N/A | N/A | N/A | N/A | N/A | 0.06 ± 0.01 |
| In-hospital death | 0.51 ± 0.06 | 1.61 ± 0.19 b | 0.83 ± 0.09 b | 0.77 ± 0.18 | N/A | 0.78 ± 0.05 |
| Severe maternal morbidity b | 113.93 ± 0.87 | 284.26 ± 2.52 b | 145.28 ± 1.12 b | 131.97 ± 2.29 b | 225.47 ± 10.36 b | 150.68 ± 0.65 |
| Severe maternal morbidity without blood transfusion c | 48.06 ± 0.57 | 131.67 ± 1.73 b | 57.75 ± 0.71 b | 55.52 ± 1.49 b | 75.97 ± 6.06 b | 64.26 ± 0.43 |
a Includes other/multiple race category
b Statistically significantly different from the corresponding rate in non-Hispanic white women at P < .005
We fitted Poisson regression models for our 2 outcomes of interest (severe morbidity with and without blood transfusion) to explore relationships among key patient sociodemographic characteristics, medical characteristics, and severe maternal morbidity. Models were adjusted for the following elements: race/ethnicity that was categorized as non-Hispanic white, non-Hispanic black, Hispanic, API, American Indian/Alaska Native (AI/AN), and other/multiple race; the maternal age in 5-year groups between 15 and 49 years; insurance coverage that was categorized as Medicaid, private, self-pay, and other; state-level household income quartile for patient ZIP code; the presence (yes/no) of chronic medical conditions that included preexisting diabetes mellitus, chronic hypertensive disease, chronic heart disease, chronic respiratory disease, chronic renal disease, chronic liver disease, and HIV/AIDS (corresponding ICD-9-CM codes provided in Table 3 ); state, and year. All regression models in this analysis were adjusted for clustering at the hospital level with the use of generalized estimating equations, and all analyses were conducted with Stata software (version 12; StataCorp, College Station, TX).
| Condition | International Classification of Diseases, 9th Revision , diagnosis codes |
|---|---|
| Preexisting diabetes mellitus | 249.x, 250.x, 648.0x |
| Chronic hypertensive disease | 401.x-405.x, 642.0x, 642.1x, 642.2x, 642.7x |
| Chronic heart disease | 412.x-414.x, 394.x-397.x, 424.x, 428.22, 428.23, 428.32, 428.33, 428.42, 428.43 |
| Chronic respiratory disease | 491.x-496.x |
| Chronic renal disease | 581.x-583.x, 585.x, 587.x, 588.x, 646.2x |
| Chronic liver disease | 571.x, 572.x |
| HIV/AIDS | 042.x, V08.x |
Results
Of the 3,476,392 delivery hospitalizations in our sample, 42.7% were among non-Hispanic white women, 12.5% among non-Hispanic black women, 32.8% among Hispanic women, 7.1% among API women, 0.6% among AI/AN women, and 4.3% among other or multiple race/ethnicity groups (χ 2 , P < .05). Among all racial/ethnic groups, blood transfusion was the most common indicator of severe morbidity; the rates of blood transfusion ranged from approximately 79 to 187 per 10,000 delivery hospitalizations among non-Hispanic white and non-Hispanic black women, respectively ( Table 2 ). Racial/ethnic-specific morbidity rates exceeded 10 per 10,000 delivery hospitalizations for the following 8 specific conditions or procedures: rates of disseminated intravascular coagulation among all groups; heart failure during procedure or surgery among non-Hispanic black, Hispanic, API, and AI/AN women; acute renal failure, acute respiratory distress syndrome (ARDS), and ventilation among non-Hispanic black and AI/AN women; hysterectomy among non-Hispanic black and API women; eclampsia and cardio monitoring among non-Hispanic black women. Rates of specific severe morbidity indicators that were statistically significantly different at a probability value of < .05 than corresponding rates in non-Hispanic white women are shown in Table 2 ; of note, rates of 22 of the 25 specific severe morbidity indicators were significantly higher (all P < .05) in non-Hispanic black than in non-Hispanic white women.
Overall, for every 10,000 delivery hospitalizations, there were 151 hospitalizations (approximately 114 in non-Hispanic white women and 284 in non-Hispanic black women) with at least 1 severe complication and 64 hospitalizations (approximately 48 in non-Hispanic white women and 132 in non-Hispanic black women) with at least 1 severe complication, except for blood transfusion. Rates of severe morbidity measured with and without blood transfusion were lowest among non-Hispanic white women; unadjusted rates of severe morbidity with and without blood transfusion were 2.5 and 2.7, 2.0 and 1.6, 1.3 and 1.2, 1.2 and 1.2 times higher for black non-Hispanic, AI/AN, Hispanic and API women, respectively (all P < .001; data not shown). After we controlled for potential confounders ( Tables 4 and 5 ), we found that non-Hispanic black women had 2.1 ( P < .001) and 2.4 ( P < .001) times higher rates of severe morbidity that were measured with and without blood transfusion, respectively, relative to non-Hispanic white women; similarly, AI/AN women had 1.7 ( P < .001) and 1.3 ( P = .002) times higher rates of severe morbidity measured with and without blood transfusion, respectively. Hispanic and API women had 1.3 ( P < .001) and 1.2 ( P < .001) times, respectively, higher rates of severe morbidity, irrespective of its measurement with and without blood transfusion, than non-Hispanic white women. As expected, the rates of severe morbidity increased significantly with age and appeared to be higher for women <20 years old when compared with women 20-24 years old. Women with private insurance and those living in higher income neighborhoods had lower rates of severe morbidity compared with their counterparts. The presence of chronic medical conditions was associated with higher rates of severe morbidity, especially for women with chronic heart disease, chronic renal disease, and chronic liver disease; the presence of HIV/AIDS infection doubled the rates of severe morbidity that was measured with and without blood transfusion.
