Abstract

Methods

In this study, we used deidentified administrative data, which does not meet the definition of human subjects’ research and, by policy, did not require Children’s Hospital of Philadelphia Institutional Review Board review.

Patient Demographic, Clinical, and Hospitalization Characteristics

The primary exposure of interest was state of residence. Patient characteristics included age, sex, type(s) of CCC(s), and number of CCCs (Table 1). Hospitalization characteristics included type (surgical or medical), length of stay (LOS), and year (Table 1).

TABLE 1

Sample Characteristics and Adjusted Odds Ratios of Receiving Posthospitalization Home Nursing

	Overall (N = 88139)	No Home Nursing (n = 79413)	Home Nursing (n = 8726)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Age, y
<2	19663	16843 (21.2%)	2820 (32.3%)	Reference	Reference
2–5	14043	12681 (16.0%)	1362 (15.6%)	0.64 (0.60 to 0.69)	0.63 (0.59 to 0.68)
6–11	17481	15834 (19.9%)	1647 (18.9%)	0.62 (0.58 to 0.66)	0.64 (0.60 to 0.69)
12–15	18762	17237 (21.7%)	1525 (17.5%)	0.53 (0.49 to 0.56)	0.60 (0.56 to 0.65)
16–18	18190	16818 (21.2%)	1372 (15.7%)	0.49 (0.46 to 0.52)	0.55 (0.51 to 0.59)
Sex
Male	45756	40929 (51.5%)	4827 (55.3%)	Reference	Reference
Female	42383	38484 (48.5%)	3899 (44.7%)	0.86 (0.82 to 0.90)	0.91 (0.86 to 0.95)
Year
2013	30910	27463 (34.6%)	3447 (39.5%)	Reference	Reference
2014	26897	24180 (30.4%)	2717 (31.1%)	0.90 (0.85 to 0.94)	0.88 (0.83 to 0.93)
2015	15816	14471 (18.2%)	1345 (15.4%)	0.74 (0.69 to 0.79)	0.75 (0.70 to 0.81)
2016	14516	13299 (16.7%)	1217 (13.9%)	0.73 (0.68 to 0.78)	0.71 (0.66 to 0.76)
Hospitalization type
Surgical	33709	30711 (38.7%)	2998 (34.4%)	Reference	Reference
Medical	54430	48702 (61.3%)	5728 (65.6%)	1.20 (1.15 to 1.26)	1.55 (1.47 to 1.64)
Mean hospital days	5.31 (SD 0.04)	4.69 (SD 0.03)	10.91(SD 0.21)	1.03 per day (1.03 to 1.04)	1.03 per day (1.02 to 1.03)
>1 CCC	20751	16831 (21.2%)	3920 (44.9%)	3.03 (2.90 to 3.17)	1.38 (1.25 to 1.52)
CCC type(s)
Cardiovasculara,b	21392	19275 (24.3%)	2117 (24.3%)	1.00 (0.95 to 1.05)	1.18 (1.09 to 1.28)
Neuromusculara,b	20639	18547 (23.4%)	2092 (24.0%)	1.03 (0.98 to 1.09)	1.20 (1.10 to 1.30)
Congenital or geneticc	15785	14376 (18.1%)	1409 (16.1%)	0.87 (0.82 to 0.92)	1.04 (0.96 to 1.14)
Technology dependence	15402	12944 (16.3%)	2458 (28.2%)	2.01 (1.91 to 2.11)	1.78 (1.67 to 1.89)
Gastrointestinal	12661	10532 (13.3%)	2129 (24.4%)	2.11 (2.00 to 2.23)	2.10 (1.93 to 2.28)
Hematology or immunology	10194	9047 (11.4%)	1147 (13.1%)	1.18 (1.10 to 1.26)	1.31 (1.19 to 1.44)
Malignancy	9312	7566 (9.5%)	1746 (20.0%)	2.38 (2.24 to 2.52)	3.34 (3.05 to 3.66)
Metabolica,b	8728	7865 (9.9%)	863 (9.9%)	1.00 (0.93 to 1.08)	1.18 (1.07 to 1.31)
Renal	6773	6149 (7.7%)	624 (7.2%)	0.92 (0.84 to 1.00)	1.12 (1.01 to 1.25)
Respiratory	6204	4667 (5.9%)	1537 (17.6%)	3.42 (3.22 to 3.65)	2.80 (2.55 to 3.06)
Neonatal	2329	1785 (2.2%)	544 (6.2%)	2.89 (2.62 to 3.19)	1.54 (1.35 to 1.75)
Transplant	1906	1536 (1.9%)	370 (4.2%)	2.25 (2.00 to 2.52)	1.62 (1.41 to 1.85)

Open in a separate window

For state of residence, see Supplemental Table 3. OR, odds ratio.

^aPearson’s χ² P > .05.

^bUnadjusted OR P > .05.

^cAdjusted OR P > .05.

Results

A total of 88139 unique patient hospitalizations were analyzed (Table 1). The mean age was 8.7 years (95% CI: 8.7 to 8.8), with a fairly uniform distribution across age categories. The mean length of hospitalization was 5.3 days (95% CI: 5.2 to 5.4). Patients had a mean of 1.3 CCCs (95% CI: 1.29 to 1.30). The most common CCCs were neuromuscular disorders (23.4%) and cardiovascular disorders (24.3%). Posthospitalization home nursing was provided after 8726 (9.9%) admissions.

In the unadjusted bivariate analysis (Table 1) receipt of home nursing was associated with patient age, patient sex, year, admission type (surgical versus medical), each additional hospital day, having >1 CCC, and most CCC types (P < .001). In the multivariable adjusted analysis (Table 1), the same associations were identified. The addition of state of residence improved model fit significantly (likelihood ratio test, P < .001). Across states with an adequate sample size, the average adjusted marginal probability of receiving any posthospital home nursing across the states of residence ranged from 3.4% to 19.2% (Fig 1, Supplemental Table 3).

Open in a separate window

FIGURE 1

Adjusted marginal probability of receiving posthospitalization home nursing by state (among states with at least 20 observations). The widths of the 95% CIs are largely due to the sample size in each state, with smaller states having wider CIs.

Regarding the quantity of home nursing days delivered to the 8726 recipients of home nursing, the overall mean days of home nursing delivered was 15.4 days (95% CI: 15.2 to 15.6) per child. Age, year, having >1 CCC, LOS, and state of residence were included in a quantile regression model of median days of home nursing among home nursing recipients. On the basis of marginal effects at the mean value of other predictor variables, the adjusted median predicted days of home nursing (across the 42 states with >20 observations with home nursing) ranged from 6.6 to 24.5 days (Fig 2, Supplemental Table 4).

Open in a separate window

FIGURE 2

Adjusted predicted median number of days of home nursing among recipients by state (among states with at least 20 observations). The widths of the 95% CIs are largely due to the sample size in each state, with smaller states having wider CIs. Values for the adjusted 95% CIs <0 have been truncated.

Some common combinations of ages and conditions were selected to illustrate the degree of difference between states regarding receipt of any home nursing and the quantity of home nursing days (Table 2). Within those categories, the (unadjusted) proportion of patients receiving any home nursing and, among patients who did receive home nursing, the median number of days received in the 30 days after hospital discharge were identified. For instance, in the fifth highest state (among states with >20 observations), 41.0% of children aged 0 to 2 years with technology dependence received home nursing, in contrast to 16.8% of similar children in the fifth lowest state. In this population, the median days of home nursing among recipients in the fifth lowest state was 7 days, compared with 25 days in the fifth highest state. There was a 2.5-fold difference in the proportion of children aged 16 to 19 with malignancies who receive home nursing between the fifth highest and the fifth lowest states.

Quantile regression provided a single framework for examining state-level variation in both receipt of any home nursing and the quantity of days of home nursing distributed. In our sample, the 95th percentile of observed number of days of home nursing (range: 0–30 days) exhibited the maximal diversity between states. Although the 95th percentile of days of home nursing across the entire sample was 14 days, the 95th percentile for some states was 0 days, and for other states, it was 30 days, with other states at points between. Therefore the 95th percentile was selected as the maximal diversity point of comparison across states, encompassing both home nursing receipt and quantity. Age, year, having >1 CCC, LOS, and state of residence were included in a quantile regression model of the 95th percentile of home nursing days across the entire sample. On the basis of marginal effects at the mean value of other predictor variables, the adjusted 95th percentile predicted days of home nursing ranging from 6.8 to 22.6 days (Fig 3, Supplemental Table 5).

Open in a separate window

FIGURE 3

Adjusted predicted 95th percentile number of days of home nursing by state (among states with at least 20 observations). The widths of the 95% CIs are largely due to the sample size in each state, with smaller states having wider CIs. Values for the adjusted 95% CI <0 have been truncated.

Discussion

After accounting for clinical and hospitalization characteristics, we found that significant state-to-state variability exists in the proportion and quantity of posthospitalization home nursing received by commercially insured CMC in this large national cohort. There is more than a fivefold difference in the adjusted marginal probability of receiving home nursing between states. Ranking of the 95th percentile of home nursing days allows for a reproducible assessment of both probability and quantity of home nursing delivery.

Our results are consistent with those of previous studies. Our unadjusted national receipt of posthospitalization home nursing (~10%) is concordant with estimates from a single-center chart audit study which identified an overall home nursing receipt of 7.5% (25% in children who were technology dependent).¹⁸ Similarly, an analysis of the Agency for Healthcare Research and Quality’s Healthcare Cost and Utilization Project Kids’ Inpatient Database revealed that 7.6% of children with at least 1 CCC had a hospital discharge disposition code indicating discharge to home nursing, with wide state variability.¹⁹ Both studies included patients of all insurance statuses (Medicaid, private, uninsured, and dual coverage).

These findings should be interpreted in light of the study’s strengths and limitations. Strengths include an updated national perspective on home nursing services based on actual claims from a well-curated data source. Whereas in previous work rates of home nursing receipt have been discussed, using claims data allowed us to examine the quantity of home nursing delivered.

This is a study of commercially insured children, who make up ~30% of all CMC.^20,21 A majority of children who receive home nursing are state Medicaid waiver beneficiaries. Children with commercial or private insurance may have lower odds of receiving home nursing than children with public insurance¹⁹; therefore, these findings may not be generalizable to children with other types of insurance. Despite these theoretical differences, our overall estimates are in line with data from samples that included both Medicaid-insured and commercially insured patients.^18,19 Interactions between state Medicaid policies (including Medicaid expansion efforts) and commercial insurance benefits have not been characterized.

The identifier for home nursing used in this analysis does not differentiate between skilled nursing visits and private-duty nursing shifts. Skilled nursing visits (for tasks such as wound care, medication administration, or line care) last on the order of minutes to hours, whereas private-duty nursing shifts provide hours up to a full day of patient care. Across states, we cannot determine if fewer days of home nursing was offset by more hours per day. Although there are, to our knowledge, no data to suggest that such a trade-off is occurring, this possibility should be examined in future work.

Analysis was limited by the availability of patient and clinical information within this administrative data set. The data set does not include information about whether families self-paid for additional home care services. The desired or optimal number of days of home nursing could not be ascertained in this administrative data set. Race, ethnicity, or family socioeconomic factors are not included in the data set, precluding analysis of these potentially important sources of disparity. No information about hospital or facility or clinician was available.

Keeping the study’s strengths and limitations in mind, we identified 5 aspects of these findings that warrant discussion. First, to a limited degree, clinical and demographic factors influence the distribution of posthospitalization home nursing to commercially insured CMC, all in accordance with previously identified associations. Younger children with multiple CCCs and longer, more costly hospital stays were most likely to receive home nursing.^19,25 Families rely on home nursing to safely care for these children who are medically fragile. For some, hospital discharge may be contingent on securing appropriate home nursing care.⁷ Yet despite evidence of the need for home nursing, no standards exist to guide distribution, and thus there remain disparities in who receives home nursing and how much home nursing they receive.

Second, pediatric posthospitalization home nursing is infrequent. Even in this sample of CMC with commercial insurance, the proportion of children receiving home nursing was far smaller than the estimated 29.2% of older adult Medicare beneficiaries who are referred to posthospital home health care.²⁶ A dearth of trained pediatric home nurses with skills in providing care to children who are technology dependent, wide variation in state coverage programs for children, and the ability to rely on informal labor provided by family caregivers are potential contributors to the relative paucity of pediatric home nursing.²² Although the optimal rate of home nursing distribution is not known, this difference suggests that there may be opportunity to provide more support for families of CMC.

Third, the substantial variation across states, both in terms of providing any home nursing and the amount of home nursing distributed, raises the question of why. Proposed drivers of these differences include regional differences in referral patterns, the availability of home nursing services, scope-of-practice regulation differences in who is eligible to provide home nursing, payer coverage of services, home nurse wage rates, and the accessibility of inpatient rehabilitation and skilled nursing facilities (which are alternatives to home health care).^22,26

Fourth, there are observed differences in state distribution, as measured by proportion versus by quantity; in some states, home nursing is more common, but the median days of home nursing is lower and vice versa. It is possible that because we do not distinguish between episodic-visiting nursing and long-term private-duty nursing, state variation may in part reflect differences in the balance between these 2 types of services. Nonetheless, in light of the well-documented shortages and state-level differences in licensure and reimbursement,²² it is important to consider whether CMC who qualify for and seek posthospital home nursing are receiving adequate coverage or whether these differences reflect gaps in home nursing shifts that are difficult to fill. Ultimately, adequacy of home nursing distribution is measured both by the proportion and quantity of home nursing delivered, which is why a combined metric is important.

Fifth, although this is a study of commercially insured CMC, these state-level differences suggest opportunities for policy and programmatic approaches to increase access to home nursing services beyond expansion of Medicaid. In adults, a county-level analysis of home health agencies revealed that in addition to sociodemographic factors, there was an association between increased number of agencies and increased home nursing reimbursement rates.²⁷ Although the same association has not been explored in pediatrics, there are known regional disparities in both availability of qualified pediatric home nurses^7,19,22 and Medicaid reimbursement rates for home nursing.^28–30

In future work, researchers should evaluate potential drivers of the observed variation among states and over time regarding the provision of home nursing, including more direct regulatory and market-based mechanisms and, potentially, more complex policy mechanisms. For example, the relative size of the home nursing labor supply in any given state may be directly due to the restrictiveness of that state’s scope-of-practice regulations or due to the local prevailing differential of wages between home and hospital nursing.²² At the level of policy, levels of home nursing among commercially insured children might be driven by Medicaid policies but, potentially, in divergent directions. On the one hand, more expansive Medicaid coverage policies regarding home nursing might increase the pool of available home nurses in an area, allowing for improved access to home nursing among both Medicaid-insured and commercially insured children. On the other hand, in states with more expansive Medicaid programs, because children would be more likely to have Medicaid coverage, commercial insurance plans might be more restrictive regarding the provision of home nursing. All of these factors and others that could potentially influence the receipt of home nursing could be examined by using our reproducible state-level comparison methodology.

Conclusions

By quantifying and rank-ordering state performance, we hope to motivate and inform efforts to ensure appropriate access to home nursing. The degree of variability in home nursing delivery, after adjustment for clinical and demographic factors, suggests that state-level factors may contribute to delivery of home nursing and that there may be opportunities for improvement in access to this vital service.

Glossary

Footnotes

References