BACKGROUND AND OBJECTIVES:

Very low birth weight (VLBW) infants are at high risk for morbidities beyond the neonatal period and ongoing use of health care. Specific morbidities have been studied; however, a comprehensive landscape of medical complexity in VLBW infants has not been fully described. We sought to (1) describe the prevalence of complex chronic conditions (CCCs) and (2) determine the association of demographic, hospital, and clinical factors with CCCs and CCCs or death.

METHODS:

This retrospective cross-sectional analysis of discharge data from the Kids’ Inpatient Database (2009–2012) included infants with a birth weight <1500 g and complete demographics. Outcomes included having CCCs or having either CCCs or dying. Analyses were weighted; univariate and multiple logistic regression models were used to estimate unadjusted and adjusted odds ratios. A dominance analysis with Cox-Snell R2 determined the relative contribution of demographic, hospital, and clinical factors to the outcomes.

RESULTS:

Among our weighted cohort of >78 000 VLBW infants, >50% had CCCs or died. After adjustments, the prevalence of CCCs or CCCs or death differed by sex, race and ethnicity, hospital location, US region, receipt of surgery, transfer status, and birth weight. Clinical factors accounted for the highest proportion of the model’s ability to predict CCCs and CCCs or death at 93.3% and 96.3%, respectively, whereas demographic factors were 11.5% and 2.3% and hospital factors were 5.2% and 1.4%, respectively.

CONCLUSIONS:

In this nationally representative analysis, medical complexity is high among VLBW infants. Varying contributions of demographic, hospital, and clinical factors in predicting medical complexity offer opportunities to investigate future interventions to improve care delivery and patient outcomes.

Very low birth weight (VLBW) infants, defined as infants weighing <1500 g at birth, have experienced an increase in survival over the past several decades because of advances in maternal and neonatal care.1,2  However, with this increased survival, there has also been an increase in the number of children living with significant morbidities that often persist well beyond the neonatal period.3,4  It is known that VLBW infants accrue higher health care costs during early childhood and that infants who are admitted to the NICU have higher ongoing health care needs than the general population.57  Although there is a large body of literature on longitudinal outcomes of VLBW infants, most researchers focus on developmental outcomes or a single organ system810  with little research regarding the overall medical condition of these children after the neonatal period. Specifically, there is a lack of systematic identification and characterization of medically complex neonates during the birth hospitalization.

Children with medical complexity is a term that encompasses children with the highest level of pediatric health care use. This population represents <1% of all children; however, it accounts for almost a third of total pediatric health care spending.1113  Systematically identifying these children has remained a challenge, with various schemes and definitions being used in the literature. The gold standard identification of children with medical complexity includes assessment of high medical and family needs, functional limitations, high health care resource use, and underlying chronic conditions that are severe and/or associated with medical fragility.14  A widely used and readily available way to identify these children on a population level is by using a tested and comprehensive list of diagnoses that generally fit these criteria, specifically the published list of complex chronic conditions (CCCs).4,15  Developed initially to identify causes of death in pediatric patients, applications of this scheme using diagnostic codes to identify children with medical complexity have been well studied in the pediatric population as a whole and have been shown to be associated with increased hospital admission, hospital days, hospital charges, and an overall increase in the medical complexity of hospitalized children in the United States.3,16  Additionally, having >1 CCC has been identified as an indicator for higher medical complexity, including need for care coordination and multiple subspecialists.3,13  In many of these studies, however, newborns are excluded or are included only with older children; thus, the scope of medical complexity, particularly the prevalence of CCC, in the neonatal population remains largely unknown.

Given that VLBW infants are known to be at risk for high health care use both in the neonatal period and beyond, yet the full landscape of medical complexity in this population is poorly understood, we sought to (1) describe the prevalence of medical complexity, specifically the presence of CCCs, in a nationally representative sample of VLBW infants and (2) determine the association of specific demographic, hospital, and clinical factors with CCCs and with either CCCs or death.

We performed a retrospective cross-sectional analysis of hospital discharge data from the Healthcare Cost and Utilization Project Kids’ Inpatient Database (KID) from 2009 to 2012. The KID is the largest multistate, nationally representative database for pediatric hospitalizations. It includes randomly sampled discharges from 4200 acute care hospitals in 44 states and is released every 3 years. The data set is weighted for each discharge to produce national estimates of pediatric inpatient resource use,17  and the prevalence of VLBW infants is similar to the Centers of Disease Control and Prevention National Vital Statistics Reports.18  The data set includes variables regarding infant sociodemographic information, diagnoses, procedures, length of stay, and hospital characteristics. Survey data are weighted to account for the complex survey design, nonresponse, and noncoverage. This project was reviewed by our university’s institutional review board and deemed exempt.

Our cohort selection is summarized in Figure 1. We identified infants admitted during birth hospitalization who had an International Classification of Diseases, Ninth Revision (ICD-9) code indicating a birth weight of <1500 g and age at admission of 0 years. We verified our cohort by excluding any infants who had a diagnosis-related group code associated with term birth or uncomplicated neonates.19  Entries in the KID are not linked; each hospital admission is considered a unique entry. To address this and capture each infant only once and exclude any readmissions, we included only infants admitted during birth hospitalization, either born in the hospital of their entry or transferred during birth hospitalization, and whose final disposition was routine discharge from the hospital, discharge from the hospital to home health, or death. Infants with a final disposition of being transferred out to another institution were excluded. This allowed us to capture each infant only once, namely, during birth hospitalization at the hospital from which they were discharged. We also excluded infants with missing data for infant age, sex, race and ethnicity, insurance type, infant zip code location (urban or rural), household income quartile, teaching hospital status, and hospital size.

FIGURE 1

Cohort selection.

FIGURE 1

Cohort selection.

Close modal

Our primary outcomes were having medical complexity, defined as having a CCC, and the combined outcome of CCC or death, with diagnosis occurring at any point during the birth hospitalization. This latter composite outcome was included to capture infants who may have died before receiving a formal diagnosis that would indicate a CCC. CCCs are identified by using ICD-9 codes indicating conditions expected to last >12 months involving either several organ systems or 1 organ system severely enough to require specialty pediatric care and likely a period of hospitalization4,15  and used extensively to describe children with medical complexity.3,13,20  The comprehensive system of CCCs includes diagnoses in 11 categories including cardiovascular, respiratory, neuromuscular, renal, gastrointestinal, hematologic or immunologic, metabolic, other congenital or genetic, malignancy, technology dependence, and conditions arising in the neonatal period.4  We excluded CCC diagnoses that were reflective of only birth weight or gestational age.

We analyzed covariates in 3 groups including demographic, hospital, and clinical factors. Demographic variables included sex, race and ethnicity (non-Hispanic white [NHW], non-Hispanic Black [NHB], Hispanic, or other), insurance type (self-pay or no insurance, Medicaid or Medicare, or private insurance), patient zip code location (large metropolitan or fringe, medium or small metropolitan, or very small metropolitan or rural), and household income quartile. Hospital characteristics included hospital region (defined by US census region, ie, northeast, Midwest, south, or west), size (small, medium, or large) type (government, private and nonprofit, or private and investor owned), and teaching status (rural nonteaching, urban teaching, or urban nonteaching). Clinical characteristics included whether the infant was outborn, median number of procedures, having a procedure that occurred in the operating room (OR), gestational age, birth weight, final disposition, length of stay, and total charges.

#### Descriptive, Univariate, and Bivariate Analyses

We determined the prevalence of CCCs and CCCs or death in the cohort. All analyses were weighted to estimate the national and regional prevalence of CCCs. Univariate and multiple survey logistic regression models were used to estimate unadjusted and adjusted odds ratios for VLBW infants having CCCs and CCCs or death. In the multivariable models, odds ratios were adjusted by using covariates described above.

#### Modeling the Associations of Demographic, Hospital, and Clinical Factors on CCCs and CCCs or Death

To further model the relationship between the outcomes (CCCs or CCCs/death) and candidate predictors, we examined the 3 groups of predictors: demographic, hospital, and clinical. We first evaluated the correlations between variables and outcomes of interest within each group and next conducted multiple logistic regression with the final model containing all 3 groups. The C-statistic, a metric used to compare multiple models, was used to evaluate the final model’s goodness of fit. We performed a dominance analysis with Cox-Snell R2 to quantify the relative contribution of each of the 3 groups of variables to the model’s total explanatory value.21

All analyses were conducted by using SAS version 9.4 (SAS Institute, Inc, Cary, NC) and SUDAAN 11.03 (Research Triangle Institute International, Research Triangle Park, NC).

The cohort included a weighted total of 78 683 VLBW infants. The weighted prevalence of demographic, hospital, and clinical characteristics for the study population is summarized in Table 1. Overall, 37.5% of VLBW infants had CCCs (95% confidence interval [CI] 37.1%–37.9%), whereas 50.8% of VLBW infants experienced the combined outcome of CCC or death (95% CI 50.4%–51.2%). In addition, 10.3% of VLBW infants had ≥2 CCCs (95% CI 10.1%–10.6%) (Supplemental Table 3).

TABLE 1

Weighted Prevalence of Demographic, Hospital, and Clinical Level Characteristics for Overall Cohort and Infants With CCCs and CCCs or Death

CharacteristicOverallCCCCCC or Death
Weighted nWeighted Prevalence, % (95% CI)Weighted nWeighted Prevalence, % (95% CI)Weighted nWeighted Prevalence, % (95% CI)
Demographic level
Sex
Male 39 707 50.5 (50.1–50.9) 15 653 53.1 (52.4–53.8) 21 564 53.9 (53.4–54.5)
Female 38 975 49.5 (49.1–49.9) 13 825 46.9 (46.2–47.6) 18 427 46.1 (45.5–46.6)
Race and ethnicity
NHW 32 902 41.8 (41.4–42.2) 12 456 42.3 (41.6–42.9) 16 377 41.0 (40.4–41.5)
NHB 22 153 28.2 (27.8–28.5) 8194 27.8 (27.2–28.4) 11 608 29.0 (28.5–29.6)
Hispanic 14 449 18.4 (18.1–18.7) 5406 18.3 (17.8–18.9) 7361 18.4 (18.0–18.9)
Other 9179 11.7 (11.4–11.9) 3422 11.6 (11.2–12.0) 4645 11.6 (11.3–12.0)
Insurance type
Medicare or Medicaid 42 173 53.6 (53.2–54.0) 16 160 54.8 (54.2–55.5) 21 332 53.3 (52.8–53.9)
Private 31 425 39.9 (39.5–40.3) 11 716 39.7 (39.1–40.4) 15 645 39.1 (38.6–39.7)
Self-pay or no insurance 5085 6.5 (6.3–6.7) 1602 5.4 (5.1–5.7) 3014 7.5 (7.2–7.8)
Patient zip code location
Large metropolitan or fringe 48 553 61.7 (61.3–62.1) 18 242 61.9 (61.2–62.5) 24 656 61.7 (61.1–62.2)
Medium or small metropolitan 20 895 26.6 (26.2–26.9) 7741 26.3 (25.7–26.9) 10 577 26.4 (25.9–27.0)
Very small, rural 9234 11.7 (11.5–12.0) 3494 11.9 (11.4–12.3) 4758 11.9 (11.5–12.3)
Household income quartile for patient zip code, percentile
0–25th 26 238 33.3 (33.0–33.7) 9789 33.2 (32.6–33.8) 13 577 34.0 (33.4–34.5)
26th–50th 19 534 24.8 (24.5–25.2) 7273 24.7 (24.1–25.3) 9903 24.8 (24.3–25.3)
51st–75th 18 187 23.1 (22.8–23.5) 6908 23.4 (22.9–24.0) 9241 23.1 (22.6–23.6)
76th–100th 14 724 18.7 (18.4–19.0) 5508 18.7 (18.2–19.2) 7269 18.2 (17.7–18.6)
Hospital level
Hospital region
Northeast 13 498 17.2 (16.9–17.5) 4810 16.3 (15.8–16.8) 6746 16.9 (16.5–17.3)
Midwest 13 749 17.5 (17.2–17.8) 5339 18.1 (17.6–18.6) 7174 17.9 (17.5–18.4)
South 34 077 43.3 (42.9–43.7) 13 168 44.7 (44.0–45.3) 17 573 43.9 (43.4–44.5)
West 17 359 22.1 (21.7–22.4) 6160 20.9 (20.4–21.4) 8498 21.3 (20.8–21.7)
Hospital size
Small 4733 6.0 (5.8–6.2) 1880 6.4 (6.0–6.7) 2525 6.3 (6.0–6.6)
Medium 16 687 21.2 (20.9–21.5) 6191 21.0 (20.5–21.6) 8562 21.4 (20.9–21.9)
Large 57 263 72.8 (72.4–73.1) 21 407 72.6 (72.0–73.2) 28 904 72.3 (71.8–72.8)
Hospital type
Government, nonfederal 10 534 13.4 (13.1–13.7) 4114 14.0 (13.5–14.4) 5547 13.9 (13.5–14.3)
Private nonprofit 59 051 75.0 (74.7–75.4) 22 010 74.7 (74.1–75.2) 29 820 74.6 (74.1–75.1)
Private investor owned 9098 11.6 (11.3–11.8) 3354 11.4 (11.0–11.8) 4624 11.6 (11.2–11.9)
Teaching and location status of hospital
Rural 1428 1.8 (1.7–1.9) 309 1.0 (0.9–1.2) 901 2.3 (2.1–2.4)
Urban nonteaching 20 500 26.1 (25.7–26.4) 6572 22.3 (21.7–22.9) 9633 24.1 (23.6–24.6)
Urban teaching 56 755 72.1 (71.8–72.5) 22 597 76.7 (76.1–77.2) 29 457 73.7 (73.2–74.2)
Clinical level
Patient transfer status
Inborn 64 764 82.3 (82.0–82.6) 22 105 75.0 (74.4–75.6) 31 973 80.0 (79.5–80.4)
Transferred in 13 918 17.7 (17.4–18.0) 7373 25.0 (24.4–25.6) 8018 20.0 (19.6–20.5)
Median No. procedures (IQR) 78 683 5.1 (2.0–7.9) 29 478 6.8 (4.7–10.4) 29 590 5.5 (1.5–9.2)
OR procedure
No 58 598 74.5 (74.1–74.8) 18 634 63.2 (62.6–63.9) 28 835 72.1 (71.6–72.6)
Yes 20 085 25.5 (25.2–25.9) 10 844 36.8 (36.1–37.4) 11 156 27.9 (27.4–28.4)
Gestational age, wk
<24 9803 12.5 (12.2–12.7) 2057 7.0 (6.6–7.3) 9453 23.6 (23.2–24.1)
24 4882 6.2 (6.0–6.4) 3299 11.2 (10.8–11.6) 4215 10.5 (10.2–10.9)
25–26 11 614 14.8 (14.5–15.1) 7378 25.0 (24.5–25.6) 8268 20.7 (20.2–21.1)
27–28 15 797 20.1 (19.8–20.4) 7155 24.3 (23.7–24.8) 7708 19.3 (18.8–19.7)
29–30 17 231 21.9 (21.6–22.2) 5001 17.0 (16.5–17.5) 5262 13.2 (12.8–13.5)
31–32 11 055 14.1 (13.8–14.3) 2383 8.1 (7.7–8.5) 2533 6.3 (6.1–6.6)
≥33 6404 8.1 (7.9–8.4) 1056 3.6 (3.3–3.8) 1223 3.1 (2.9–3.3)
Unknown or missing 1897 2.4 (2.3–2.5) 1149 3.9 (3.6–4.2) 1329 3.3 (3.1–3.5)
Birth wt, g
<500 7403 9.4 (9.2–9.6) 1056 3.6 (3.3–3.8) 6890 17.2 (16.8–17.7)
500–999 28 366 36.1 (35.7–36.4) 16 175 54.9 (54.2–55.5) 20 039 50.1 (49.5–50.7)
1000–1499 42 914 54.5 (54.1–54.9) 12 247 41.5 (40.9–42.2) 13 062 32.7 (32.1–33.2)
Final disposition
Routine 52 364 66.6 (66.2–66.9) 19 907 67.5 (66.9–68.2) 19 907 49.8 (49.2–50.4)
HHC 11 349 14.4 (14.1–14.7) 5200 17.6 (17.1–18.1) 5200 13.0 (12.6–13.4)
Died in the hospital 14 852 18.9 (18.6–19.2) 4339 14.7 (14.3–15.2) 14 852 37.1 (36.6–37.7)
Other 118 0.1 (0.1–0.2) 32 0.1 (0.1–0.2) 32 0.1 (0.1–0.1)
Length of stay, median (IQR), d 78 653 45.4 (22.4–71.9) 29 448 68.0 (39.0–95.9) 39 961 48.3 (1.0–85.1)
Total charges, median (IQR), $75 783 207 097 (81 890–425 369) 28 304 354 043 (167 474–655 044) 38 081 231 466 (28 399–532 201) CharacteristicOverallCCCCCC or Death Weighted nWeighted Prevalence, % (95% CI)Weighted nWeighted Prevalence, % (95% CI)Weighted nWeighted Prevalence, % (95% CI) Demographic level Sex Male 39 707 50.5 (50.1–50.9) 15 653 53.1 (52.4–53.8) 21 564 53.9 (53.4–54.5) Female 38 975 49.5 (49.1–49.9) 13 825 46.9 (46.2–47.6) 18 427 46.1 (45.5–46.6) Race and ethnicity NHW 32 902 41.8 (41.4–42.2) 12 456 42.3 (41.6–42.9) 16 377 41.0 (40.4–41.5) NHB 22 153 28.2 (27.8–28.5) 8194 27.8 (27.2–28.4) 11 608 29.0 (28.5–29.6) Hispanic 14 449 18.4 (18.1–18.7) 5406 18.3 (17.8–18.9) 7361 18.4 (18.0–18.9) Other 9179 11.7 (11.4–11.9) 3422 11.6 (11.2–12.0) 4645 11.6 (11.3–12.0) Insurance type Medicare or Medicaid 42 173 53.6 (53.2–54.0) 16 160 54.8 (54.2–55.5) 21 332 53.3 (52.8–53.9) Private 31 425 39.9 (39.5–40.3) 11 716 39.7 (39.1–40.4) 15 645 39.1 (38.6–39.7) Self-pay or no insurance 5085 6.5 (6.3–6.7) 1602 5.4 (5.1–5.7) 3014 7.5 (7.2–7.8) Patient zip code location Large metropolitan or fringe 48 553 61.7 (61.3–62.1) 18 242 61.9 (61.2–62.5) 24 656 61.7 (61.1–62.2) Medium or small metropolitan 20 895 26.6 (26.2–26.9) 7741 26.3 (25.7–26.9) 10 577 26.4 (25.9–27.0) Very small, rural 9234 11.7 (11.5–12.0) 3494 11.9 (11.4–12.3) 4758 11.9 (11.5–12.3) Household income quartile for patient zip code, percentile 0–25th 26 238 33.3 (33.0–33.7) 9789 33.2 (32.6–33.8) 13 577 34.0 (33.4–34.5) 26th–50th 19 534 24.8 (24.5–25.2) 7273 24.7 (24.1–25.3) 9903 24.8 (24.3–25.3) 51st–75th 18 187 23.1 (22.8–23.5) 6908 23.4 (22.9–24.0) 9241 23.1 (22.6–23.6) 76th–100th 14 724 18.7 (18.4–19.0) 5508 18.7 (18.2–19.2) 7269 18.2 (17.7–18.6) Hospital level Hospital region Northeast 13 498 17.2 (16.9–17.5) 4810 16.3 (15.8–16.8) 6746 16.9 (16.5–17.3) Midwest 13 749 17.5 (17.2–17.8) 5339 18.1 (17.6–18.6) 7174 17.9 (17.5–18.4) South 34 077 43.3 (42.9–43.7) 13 168 44.7 (44.0–45.3) 17 573 43.9 (43.4–44.5) West 17 359 22.1 (21.7–22.4) 6160 20.9 (20.4–21.4) 8498 21.3 (20.8–21.7) Hospital size Small 4733 6.0 (5.8–6.2) 1880 6.4 (6.0–6.7) 2525 6.3 (6.0–6.6) Medium 16 687 21.2 (20.9–21.5) 6191 21.0 (20.5–21.6) 8562 21.4 (20.9–21.9) Large 57 263 72.8 (72.4–73.1) 21 407 72.6 (72.0–73.2) 28 904 72.3 (71.8–72.8) Hospital type Government, nonfederal 10 534 13.4 (13.1–13.7) 4114 14.0 (13.5–14.4) 5547 13.9 (13.5–14.3) Private nonprofit 59 051 75.0 (74.7–75.4) 22 010 74.7 (74.1–75.2) 29 820 74.6 (74.1–75.1) Private investor owned 9098 11.6 (11.3–11.8) 3354 11.4 (11.0–11.8) 4624 11.6 (11.2–11.9) Teaching and location status of hospital Rural 1428 1.8 (1.7–1.9) 309 1.0 (0.9–1.2) 901 2.3 (2.1–2.4) Urban nonteaching 20 500 26.1 (25.7–26.4) 6572 22.3 (21.7–22.9) 9633 24.1 (23.6–24.6) Urban teaching 56 755 72.1 (71.8–72.5) 22 597 76.7 (76.1–77.2) 29 457 73.7 (73.2–74.2) Clinical level Patient transfer status Inborn 64 764 82.3 (82.0–82.6) 22 105 75.0 (74.4–75.6) 31 973 80.0 (79.5–80.4) Transferred in 13 918 17.7 (17.4–18.0) 7373 25.0 (24.4–25.6) 8018 20.0 (19.6–20.5) Median No. procedures (IQR) 78 683 5.1 (2.0–7.9) 29 478 6.8 (4.7–10.4) 29 590 5.5 (1.5–9.2) OR procedure No 58 598 74.5 (74.1–74.8) 18 634 63.2 (62.6–63.9) 28 835 72.1 (71.6–72.6) Yes 20 085 25.5 (25.2–25.9) 10 844 36.8 (36.1–37.4) 11 156 27.9 (27.4–28.4) Gestational age, wk <24 9803 12.5 (12.2–12.7) 2057 7.0 (6.6–7.3) 9453 23.6 (23.2–24.1) 24 4882 6.2 (6.0–6.4) 3299 11.2 (10.8–11.6) 4215 10.5 (10.2–10.9) 25–26 11 614 14.8 (14.5–15.1) 7378 25.0 (24.5–25.6) 8268 20.7 (20.2–21.1) 27–28 15 797 20.1 (19.8–20.4) 7155 24.3 (23.7–24.8) 7708 19.3 (18.8–19.7) 29–30 17 231 21.9 (21.6–22.2) 5001 17.0 (16.5–17.5) 5262 13.2 (12.8–13.5) 31–32 11 055 14.1 (13.8–14.3) 2383 8.1 (7.7–8.5) 2533 6.3 (6.1–6.6) ≥33 6404 8.1 (7.9–8.4) 1056 3.6 (3.3–3.8) 1223 3.1 (2.9–3.3) Unknown or missing 1897 2.4 (2.3–2.5) 1149 3.9 (3.6–4.2) 1329 3.3 (3.1–3.5) Birth wt, g <500 7403 9.4 (9.2–9.6) 1056 3.6 (3.3–3.8) 6890 17.2 (16.8–17.7) 500–999 28 366 36.1 (35.7–36.4) 16 175 54.9 (54.2–55.5) 20 039 50.1 (49.5–50.7) 1000–1499 42 914 54.5 (54.1–54.9) 12 247 41.5 (40.9–42.2) 13 062 32.7 (32.1–33.2) Final disposition Routine 52 364 66.6 (66.2–66.9) 19 907 67.5 (66.9–68.2) 19 907 49.8 (49.2–50.4) HHC 11 349 14.4 (14.1–14.7) 5200 17.6 (17.1–18.1) 5200 13.0 (12.6–13.4) Died in the hospital 14 852 18.9 (18.6–19.2) 4339 14.7 (14.3–15.2) 14 852 37.1 (36.6–37.7) Other 118 0.1 (0.1–0.2) 32 0.1 (0.1–0.2) 32 0.1 (0.1–0.1) Length of stay, median (IQR), d 78 653 45.4 (22.4–71.9) 29 448 68.0 (39.0–95.9) 39 961 48.3 (1.0–85.1) Total charges, median (IQR),$ 75 783 207 097 (81 890–425 369) 28 304 354 043 (167 474–655 044) 38 081 231 466 (28 399–532 201)

IQR, interquartile range.

FIGURE 2

A, Adjusted odds ratios and 95% CIs for the outcome CCCs. The data are adjusted for all other variables listed in addition to birth weight, transfer status, and OR procedure. Ref, reference group.

FIGURE 2

A, Adjusted odds ratios and 95% CIs for the outcome CCCs. The data are adjusted for all other variables listed in addition to birth weight, transfer status, and OR procedure. Ref, reference group.

Close modal

When modeling the 3 groups of factors (demographic, hospital, and clinical) on the outcome of CCCs, we found that clinical factors accounted for the highest proportion of the model’s ability to predict CCCs at 93.3%, whereas demographic factors were 11.5% and hospital factors were 5.2% (Table 2). When demographic, clinical, and hospital factors were all included, the C-statistic for the model was 0.71, which indicates a good to strong model.22

TABLE 2

Model Performance and Variable Percentage Contribution to the Model by Demographic, Hospital, and Clinical Characteristics

OutcomeC-StatisticTotal R2Percentage of Contribution, %
DemographicHospitalClinical
CCC 0.71 0.134 1.5 5.2 93.3
CCC or death 0.77 0.219 2.3 1.4 96.3
OutcomeC-StatisticTotal R2Percentage of Contribution, %
DemographicHospitalClinical
CCC 0.71 0.134 1.5 5.2 93.3
CCC or death 0.77 0.219 2.3 1.4 96.3

Likewise, when measuring the contribution of each group of factors to the outcome CCCs or death, we again found that clinical factors accounted for the highest proportion of the model’s ability to predict CCCs or death at 96.3%; however, demographic factors accounted for 2.3% and hospital factors accounted for just 1.4% with a C-statistic of 0.77 (Table 2).

In this population-based study, we show that the prevalence of medical complexity or death is high among VLBW infants, with more than half of our cohort having the outcome of CCCs or death. In addition, increased medical complexity as demonstrated by ≥2 CCCs was present in >10% of the cohort. Certain characteristics, namely, urban hospital location, US region, and race and ethnicity, are associated with higher odds of having CCCs or CCCs/death. We also demonstrated varying contributions of demographic, clinical, and hospital factors to the likelihood of CCCs or CCCs/death.

Better characterization of both the overall rates and various factors that may lead to differences or disparities in the prevalence of medical complexity among VLBW infants is important for several reasons. Because the overall medical complexity of pediatric patients has increased over the past several decades, further characterization and early identification of this population has become a priority for both clinicians and health care payers.3  Applications of the CCC scheme have been well studied in the pediatric population as a whole and this definition of children with medical complexity has been shown to be associated with increased hospital admission, hospital days, hospital charges, and an overall increase in the level of complexity of hospitalized children in the United States.3,16  However, in many of these studies, neonates or admissions to the NICU specifically are excluded; thus, this study addresses an important gap in the literature describing the scope of CCCs in the neonatal population. In addition, medical complexity has implications for services outside the hospital including home health care (HHC).12,13,16  Although HHC represents the most complex patients, children who are discharged with HHC have decreased hospital use in the year after discharge compared with similar children who are not discharged with HHC,23  highlighting the potential cost savings of HHC. Despite this and the increasing need for these services, HHC has not kept up with demand, and this lack of HHC availability has been shown to delay hospital discharge.24  Additionally, studies have not been used to look at the differences in HHC attainment specifically by NICU families or stratified by the characteristics that we found significant. More work is needed to assess the impact of VLBW infants with medical complexity on the supply and demand of home nursing, and we postulate that there are significant disparities in the attainment of HHC services. Given the implications of children with medical complexity on both in-hospital and posthospital resource use, the data in this study are an important starting point for discussion surrounding differences and disparities in the vulnerable VLBW population.

We found that VLBW infants receiving care in rural hospitals had higher odds of CCCs or death compared with those infants receiving care at urban teaching hospitals. We suspect that finding lower odds of CCCs but higher odds of CCCs or death in rural hospitals is because of the high transfer rate if these vulnerable patients survive. Given that children with CCCs have a high reliance on subspecialty pediatric care, this care is more likely to be regionalized.25  However, these findings are important given that ∼15% of women deliver in rural hospitals each year.26  In previous work, researchers have shown that rural mothers are younger, more likely to have Medicaid, and more likely to experience negative obstetrical markers of care quality including higher rates of non–medically indicated induction of labor, lower rates of vaginal birth after cesarean delivery, and higher rates of cesarean delivery.26  Access to care also varies from urban to rural locations, with rural areas with lower median family income having an increased odds of obstetrical unit closure and more than half of all rural US counties having no hospital obstetric services available.27,28  Unsurprisingly, these disparities in access to maternal care translate to neonatal outcomes, with preterm birth in a non–level III or higher NICU and lower hospital patient volumes being associated with significantly higher odds of neonatal death.29,30  Our data echo these findings because they indicate that infants born in rural areas are more likely to have CCCs or die, potentially because of rural centers lacking the resources to provide intensive care for complex infants.

In addition to the disparities seen in rural versus urban care, we demonstrated disparities by US region. We found that infants in the Midwest and south had higher odds of both CCCs and CCCs or death compared with infants in the northeast. Maternal mortality is also highest in the south, with rates as high as 36.4 and 45.9 deaths per 100 000 for women in Alabama and Arkansas, respectively, compared with maternal mortality in United States as a whole at 17.4 deaths per 100 000 live births in 2018.31  Similarly, the rate of preterm birth varies across regions, with 10.9% and 10.1% of infants in the south and Midwest, respectively, being born preterm, compared with 9.2% of infants in the northeast being born preterm.32  Regionally stratified data for neonatal medical outcomes are lacking, and our data suggest that differences persist beyond pregnancy and childbirth and into the neonatal period.

With regard to differences among racial and ethnic groups, our findings of NHB infants having lower odds of CCCs and CCCs or death compared with NHW infants was surprising given that it has been well established that the rate of preterm birth is persistently elevated among NHB mothers.33  In recent studies, researchers have shown similar trends with decreased morbidities such as bronchopulmonary dysplasia in NHB infants,34  and this is postulated to be due to a variety of factors that affect the data that are used, namely, that dating for NHB mothers may be more inconsistent.35  Particular methods have been proposed to account for these discrepancies, including the fetus-at-risk model, which assesses for outcomes at a given gestational age among both delivered and ongoing (not delivered) pregnancies.36  Thus, our findings of lower odds for CCCs and CCCs or death in NHB infants is likely reflective of the disparity in the preterm birth rate. These infants appear to have a lower prevalence of medical complexity, but in fact if the preterm birth disparity did not exist, these infants would have a lower risk of VLBW status. Further stratification is needed to assess if these findings persist. Previous work has also revealed that substantial disparities exist in previable infant loss, with NHB women having significantly higher odds of fetal demise or stillbirth compared with NHW women.3740  Our data set only includes live-born infants who had a hospital admission; thus, we were unable to appropriately capture this specific population.

Lastly, in addition to assessing patient characteristics leading to increased odds of medical complexity, we also demonstrated the relative contribution of demographic-, clinical-, and hospital characteristics to our modeling of the outcomes of interest. Demographic and clinical factors can be seen as difficult to modify; however, recent quality improvement efforts have highlighted opportunities to understand and address deficits in NICU care for particular populations. Although decreasing the preterm birth rate may seem difficult to achieve in the short-term, efforts can be directed toward strategies shown to improve the near-term outcomes of our youngest infants, like the provision of breast milk to prevent necrotizing enterocolitis and avoiding long-term invasive respiratory support to decrease lung injury.4145  In addition, given that racial and ethnic disparities have previously been demonstrated in the quality of care received in the NICU, our data also support the need to address social determinants of health to narrow the disparities seen at the patient and clinical level.46

Likewise, hospital factors are an area that we see as potential for intervention because they accounted for >5% of the model’s ability to predict having CCCs. Hospital characteristics included in our analysis were region, size, type, and teaching status. Implementing perinatal regionalization programs is correlated with improvements in perinatal outcomes,47  and our analysis would suggest that this may also be true for medical complexity among our VLBW population. However, challenges in delivering the highest quality of neonatal care include disparities in timely access to both obstetrical and infant care in addition to regional and interhospital variation in the care of complex neonates.48  Studies have also revealed that management of complex pediatric conditions at specialized hospitals may improve outcomes and lower mortality.25  Additionally, hospital academic affiliation has been shown to be an important predictor of maternal outcomes, with better clinical outcomes occurring among patients at academic institutions compared with community hospitals49,50  and increased maternal and neonatal complications at lower-volume centers.51,52  In addition to size and teaching status of hospitals, the level of NICU care has been shown to be important in specific morbidities. For instance, the risk of developing bronchopulmonary dysplasia is higher for infants cared for in level II NICUs compared with level IV NICUs.53  As suggested in previous studies looking at regionalization of neonatal care, our data reveal that hospital factors are associated with the outcome of CCCs or death among VLBW infants and that finding ways to deliver the highest level of care to all potentially complex neonates is crucial.

Our study does have several limitations. As is the case with all retrospective cohort studies, we were limited by the data elements collected, and causation cannot be assessed nor implied. Identification of our cohort and assessment of outcomes were based on ICD-9 codes, so we are limited by the reliability of administrative data. The variable of race and ethnicity was taken from the infants’ medical record and we were unable to verify accuracy; additionally, a number of entries had missing information for this variable. In addition, we were unable to link patients to future encounters and therefore cannot assess ongoing health care use. Our models’ R2 of >0.7 indicate good to strong models; however, given the retrospective and observational nature of the study, other variables likely exist that we were unable to capture and may contribute to the unexplained variance that remains. Given this analysis only captured the portion of the birth hospitalization at the hospital from which infants were discharged, variables such as length of stay and total charges may be underestimated. The classification scheme of CCCs was not developed for the neonatal population, and despite being updated to include neonatal diagnoses,4  future work creating a neonatal-specific classification would be useful.

Despite these limitations, this large, nationally representative analysis provides novel information about the prevalence of CCCs and CCCs or death among VLBW infants and the associated demographic, clinical, and hospital characteristics. This population is known to contain high users of health care resources and represents an opportunity to enhance care coordination, strengthen HHC infrastructure, and subsequently improve clinical an potentially quality of life outcomes. Further work is needed to assess how to mitigate the differences that we have demonstrated across practice locations and US regions. Lastly, better understanding the differences and potential disparities is essential to providing ongoing high-level, equitable care to the already vulnerable population of VLBW infants.

Drs Hannan, Bourque, and Hwang conceptualized and designed the study, coordinated and supervised data analysis, drafted the initial manuscript, and reviewed and revised the manuscript; Ms Palmer and Ms Tong created the analysis plan, conducted the initial analyses, and reviewed and revised the manuscript; and all authors approved the final manuscript as submitted and agree to be accountable for all aspects of the work.

FUNDING: No external funding.

1
Kaiser
JR
,
Tilford
JM
,
Simpson
PM
,
Salhab
WA
,
Rosenfeld
CR
.
Hospital survival of very-low-birth-weight neonates from 1977 to 2000
.
J Perinatol
.
2004
;
24
(
6
):
343
350
2
Wilson-Costello
D
,
Friedman
H
,
Minich
N
,
Fanaroff
AA
,
Hack
M
,
Background
A
.
Improved survival rates with increased neurodevelopmental disability for extremely low birth weight infants in the 1990s
.
Pediatrics
.
2005
;
115
(
4
):
997
1003
3
Burns
KH
,
Casey
PH
,
Lyle
RE
,
Bird
TM
,
Fussell
JJ
,
Robbins
JM
.
Increasing prevalence of medically complex children in US hospitals
.
Pediatrics
.
2010
;
126
(
4
):
638
646
4
Feudtner
C
,
Feinstein
JA
,
Zhong
W
,
Hall
M
,
Dai
D
.
Pediatric complex chronic conditions classification system version 2: updated for ICD-10 and complex medical technology dependence and transplantation
.
BMC Pediatr
.
2014
;
14
:
199
5
W
,
Lee
G
,
Lin
K
,
Lantos
J
.
Changes in mortality for extremely low birth weight infants in the 1990s: implications for treatment decisions and resource use
.
Pediatrics
.
2004
;
113
(
5
):
1223
1229
6
Petrini
JR
,
Dias
T
,
McCormick
MC
,
Massolo
ML
,
Green
NS
,
Escobar
GJ
.
Increased risk of adverse neurological development for late preterm infants
.
J Pediatr
.
2009
;
154
(
2
):
169
176
7
Gray
JE
,
McCormick
MC
,
Richardson
DK
,
Ringer
S
.
Normal birth weight intensive care unit survivors: outcome assessment
.
Pediatrics
.
1996
;
97
(
6
pt 1
):
832
838
8
Msall
ME
,
Tremont
MR
.
Measuring functional outcomes after prematurity: developmental impact of very low birth weight and extremely low birth weight status on childhood disability
.
Ment Retard Dev Disabil Res Rev
.
2002
;
8
(
4
):
258
272
9
Stille
CJ
,
Antonelli
RC
.
Coordination of care for children with special health care needs
.
Curr Opin Pediatr
.
2004
;
16
(
6
):
700
705
10
McCormick
MC
.
Long-term follow-up of infants discharged from neonatal intensive care units
.
JAMA
.
1989
;
261
(
12
):
1767
1772
11
Srivastava
R
,
Stone
BL
,
Murphy
NA
.
Hospitalist care of the medically complex child
.
Pediatr Clin North Am
.
2005
;
52
(
4
):
1165
1187, x
12
Elixhauser
A
,
Machlin
SR
,
Zodet
MW
, et al
.
Health care for children and youth in the United States: 2001 annual report on access, utilization, quality, and expenditures
.
Ambul Pediatr
.
2002
;
2
(
6
):
419
437
13
Cohen
E
,
Berry
JG
,
Camacho
X
,
Anderson
G
,
Wodchis
W
,
Guttmann
A
.
Patterns and costs of health care use of children with medical complexity
.
Pediatrics
.
2012
;
130
(
6
). Available at: www.pediatrics.org/cgi/content/full/130/6/e1463
14
Cohen
E
,
Kuo
DZ
,
Agrawal
R
, et al
.
Children with medical complexity: an emerging population for clinical and research initiatives
.
Pediatrics
.
2011
;
127
(
3
):
529
538
15
Feudtner
C
,
Christakis
DA
,
Connell
FA
.
Pediatric deaths attributable to complex chronic conditions: a population-based study of Washington State, 1980-1997
.
Pediatrics
.
2000
;
106
(
1 pt 2
):
205
209
16
Simon
TD
,
Berry
J
,
Feudtner
C
, et al
.
Children with complex chronic conditions in inpatient hospital settings in the United States
.
Pediatrics
.
2010
;
126
(
4
):
647
655
17
Agency for Healthcare Research and Quality
. Introduction to the HCUP Kids’ Inpatient Database (KID). 2012. Available at: https://www.hcup-us.ahrq.gov/db/nation/kid/kid_2012_introduction.jsp. Accessed: July 3, 2018.
18
Martin
JA
,
Hamilton
BE
,
Osterman
MJ
,
Curtin
SC
,
Matthews
TJ
.
Births: final data for 2012
.
Natl Vital Stat Rep
.
2013
;
62
(
9
):
1
68
19
Hatch
LD
,
Scott
TA
,
Walsh
WF
,
Goldin
AB
,
Blakely
ML
,
Patrick
SW
.
National and regional trends in gastrostomy in very low birth weight infants in the USA: 2000-2012
.
J Perinatol
.
2018
;
38
(
9
):
1270
1276
20
Berry
JG
,
Hall
M
,
Cohen
E
,
O’Neill
M
,
Feudtner
C
.
Ways to identify children with medical complexity and the importance of why
.
J Pediatr
.
2015
;
167
(
2
):
229
237
21
Azen
R
,
Traxel
N
.
Using dominance analysis to determine predictor importance in logistic regression
.
J Educ Behav Stat
.
2009
;
34
(
3
):
319
347
22
Hosmer
D
,
Lemeshow
S
.
Applied Logistic Regression
. 2nd ed.
New York, NY
:
John Wiley & Sons, Inc
;
2000
23
Gay
JC
,
Thurm
CW
,
Hall
M
, et al
.
Home health nursing care and hospital use for medically complex children
.
Pediatrics
.
2016
;
138
(
5
):
e20160530
24
Maynard
R
,
Christensen
E
,
R
, et al
.
Home health care availability and discharge delays in children with medical complexity
.
Pediatrics
.
2019
;
143
(
1
):
e20181951
25
Lorch
SA
,
Myers
S
,
Carr
B
.
The regionalization of pediatric health care
.
Pediatrics
.
2010
;
126
(
6
):
1182
1190
26
Kozhimannil
KB
,
Hung
P
,
S
,
Casey
M
,
Moscovice
I
.
Rural-urban differences in obstetric care, 2002-2010, and implications for the future
.
Med Care
.
2014
;
52
(
1
):
4
9
27
Hung
P
,
Henning-Smith
CE
,
Casey
MM
,
Kozhimannil
KB
.
Access to obstetric services in rural counties still declining, with 9 percent losing services, 2004-14
.
Health Aff (Millwood)
.
2017
;
36
(
9
):
1663
1671
28
Hung
P
,
Kozhimannil
KB
,
Casey
MM
,
Moscovice
IS
.
Why are obstetric units in rural hospitals closing their doors?
Health Serv Res
.
2016
;
51
(
4
):
1546
1560
29
Phibbs
CS
,
Baker
LC
,
Caughey
AB
,
Danielsen
B
,
Schmitt
SK
,
Phibbs
RH
.
Level and volume of neonatal intensive care and mortality in very-low-birth-weight infants
.
N Engl J Med
.
2007
;
356
(
21
):
2165
2175
30
Kastenberg
ZJ
,
Lee
HC
,
Profit
J
,
Gould
JB
,
Sylvester
KG
.
Effect of deregionalized care on mortality in very low-birth-weight infants with necrotizing enterocolitis
.
JAMA Pediatr
.
2015
;
169
(
1
):
26
32
31
Centers for Disease Control and Prevention
US Department of Health and Human Services
. Maternal mortality by state. 2018. Available at: https://www.cdc.gov/nchs/maternal-mortality/MMR-2018-State-Data-508.pdf. Accessed: June 18, 2020
32
Centers for Disease Control and Prevention
March of Dimes Foundation
. National Center for Health Statistics, final natality data. 2018. Available at: www.marchofdimes/org/peristats/peristats.aspx. Accessed October 20, 2020
33
Hamilton
BE
,
Martin
JA
,
Osterman
MJ
,
Curtain
SC
.
Births: preliminary data for 2014
.
Natl Vital Stat Rep
.
2015
;
64
(
6
):
1
19
34
Ryan
RM
,
Feng
R
,
Bazacliu
C
, et al
;
Prematurity and Respiratory Outcome Program (PROP) Investigators
.
Black race is associated with a lower risk of bronchopulmonary dysplasia
.
J Pediatr
.
2019
;
207
:
130
135.e2
35
Burris
HH
,
Hwang
SS
,
Collins
JW
Jr
,
Kirpalani
H
,
Wright
CJ
.
Re-conceptualizing associations between race and morbidities of extreme prematurity
.
J Pediatr
.
2019
;
207
:
10
14.e1
36
Janevic
T
,
Zeitlin
J
,
Auger
N
, et al
.
Association of race/ethnicity with very preterm neonatal morbidities
.
JAMA Pediatr
.
2018
;
172
(
11
):
1061
1069
37
Williams
,
Wallace
M
,
Nobles
C
,
Mendola
P
.
Racial residential segregation and racial disparities in stillbirth in the United States
.
Health Place
.
2018
;
51
:
208
216
38
DeFranco
EA
,
Hall
ES
,
Muglia
LJ
.
Racial disparity in previable birth
.
Am J Obstet Gynecol
.
2016
;
214
(
3
):
394.e1
394.e7
39
Zhang
S
,
Cardarelli
K
,
Shim
R
,
Ye
J
,
Booker
KL
,
Rust
G
.
Racial disparities in economic and clinical outcomes of pregnancy among Medicaid recipients
.
Matern Child Health J
.
2013
;
17
(
8
):
1518
1525
40
Rowland Hogue
CJ
,
Silver
RM
.
Racial and ethnic disparities in United States: stillbirth rates: trends, risk factors, and research needs
.
Semin Perinatol
.
2011
;
35
(
4
):
221
233
41
Lechner
BE
,
Vohr
BR
.
Neurodevelopmental outcomes of preterm infants fed human milk: a systematic review
.
Clin Perinatol
.
2017
;
44
(
1
):
69
83
42
Liu
J
,
Parker
MG
,
Lu
T
, et al
.
Racial and ethnic disparities in human milk intake at neonatal intensive care unit discharge among very low birth weight infants in California
.
J Pediatr
.
2019
;
218
:
49
56.e3
43
Abman
SH
,
Collaco
JM
,
Shepherd
EG
, et al
;
Bronchopulmonary Dysplasia Collaborative
.
Interdisciplinary care of children with severe bronchopulmonary dysplasia
.
J Pediatr
.
2017
;
181
:
12
28.e1
44
Sullivan
S
,
Schanler
RJ
,
Kim
JH
, et al
.
An exclusively human milk-based diet is associated with a lower rate of necrotizing enterocolitis than a diet of human milk and bovine milk-based products
.
J Pediatr
.
2010
;
156
(
4
):
562
567.e1
45
Parker
MG
,
Burnham
LA
,
Melvin
P
, et al
.
Addressing disparities in mother’s milk for VLBW infants through statewide quality improvement
.
Pediatrics
.
2019
;
144
(
1
):
e20183809
46
Profit
J
,
Gould
JB
,
Bennett
M
, et al
.
Racial/ethnic disparity in NICU quality of care delivery
.
Pediatrics
.
2017
;
140
(
3
):
e20170918
47
Rashidian
A
,
Omidvari
AH
,
Vali
Y
, et al
.
The effectiveness of regionalization of perinatal care services--a systematic review
.
Public Health
.
2014
;
128
(
10
):
872
885
48
Bourque
SL
,
Hwang
SS
.
Underuse versus overuse of neonatal intensive care: what is the right amount?
J Pediatr
.
2018
;
192
:
5
6
49
Garcia
FAR
,
Miller
HB
,
Huggins
GR
,
Gordon
TA
.
Effect of academic affiliation and obstetric volume on clinical outcome and cost of childbirth
.
Obstet Gynecol
.
2001
;
97
(
4
):
567
576
50
Snyder
CC
,
Wolfe
KB
,
Loftin
RW
,
Tabbah
S
,
Lewis
DF
,
Defranco
EA
.
The influence of hospital type on induction of labor and mode of delivery
.
Am J Obstet Gynecol
.
2011
;
205
(
4
):
346.e1
346.e4
51
Bartels
DB
,
Wypij
D
,
Wenzlaff
P
,
Dammann
O
,
Poets
CF
.
Hospital volume and neonatal mortality among very low birth weight infants
.
Pediatrics
.
2006
;
117
(
6
):
2206
2214
52
Kyser
KL
,
Lu
X
,
Santillan
DA
, et al
.
The association between hospital obstetrical volume and maternal postpartum complications
.
Am J Obstet Gynecol
.
2012
;
207
(
1
):
42.e1
42.e17
53
Lapcharoensap
W
,
Gage
SC
,
Kan
P
, et al
.
Hospital variation and risk factors for bronchopulmonary dysplasia in a population-based cohort
.
JAMA Pediatr
.
2015
;
169
(
2
):
e143676

## Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.