Safety of Early Discharge for Medicaid Newborns FREE

Rising health care costs and other recent market forces have brought significant pressures on hospitals to discharge healthy newborns and their mothers earlier than in the past. This movement to earlier discharge occurred rapidly and with limited assessment of its safety.^1- 2 As a result, there have been documented increases in malnutrition and hyperbilirubinemia (including kernicterus) in breastfeeding infants, infants of primiparous women, and, particularly, immature infants.^3- 5 Because most studies on early discharge of newborns have focused on middle-class, commercially insured populations,⁶ the impact of early discharge on the high-risk indigent and Medicaid-eligible populations, for whom linkages to primary care may be less well established,^7- 9 is unclear.

Some small studies have suggested that early discharge is likely to be safe for Medicaid patients with strict criteria and careful follow-up,^10- 13 but concern remains about high rates of noncompliance with clinic visits in low-income populations.^10,14 Because neonatal rehospitalization rates are low,¹ few studies have had the statistical power to detect clinically significant effects on neonatal rehospitalization.

Changes in discharge policies for newborns due to market pressures since 1992 allowed us to use this natural experiment to evaluate the impact of changing postnatal lengths of stay. The present study examines the effect of changes in length of stay on immediate rehospitalization rates for newborns insured by Medicaid.

We conducted a retrospective cohort study using Medicaid claims data linked to vital statistics files from the state of Ohio for fiscal years 1992 through 1995. This combined database provided information on sociodemographic characteristics of newborns and mothers, along with date of birth, date of discharge, postdischarge health care use, diagnoses and procedures performed during the birth hospitalization, and subsequent hospitalizations.

The source population for this study was all neonates born in Ohio from July 1, 1991, to June 15, 1995 (Table 1). The last 2 weeks of June 1995 were eliminated due to incomplete data. From this larger group, a subset of newborns and their mothers were retained if they were Ohio Medicaid recipients, had a valid Medicaid birth claim, and were enrolled in the Medicaid program for at least 30 days after birth. Approximately 12% of the neonates were born to mothers enrolled in Medicaid health maintenance organization plans and were excluded from the analysis due to incomplete reporting. Healthy full-term newborns were identified by 3 criteria: diagnosis related group 391 (normal newborn), birth weight greater than 2000 g, and gestational age greater than 37 weeks. The birth weight criterion of 2000 g was chosen to include all likely healthy newborns cared for in the normal nursery even if of somewhat lower birth weight. The study group was further divided by cesarean or vaginal birth.

The primary outcomes were rehospitalization rates within 7 and 14 days of discharge. Demographic, maternal, prenatal, and neonatal characteristics were examined. Prenatal care inadequacy was assessed using self-reported prenatal visit information available from birth certificates.¹⁵ Because many women may visit a clinic or physician only for confirmation of pregnancy, a conservative estimate of 2 or fewer visits was chosen to represent minimal or absent prenatal care.

Length of stay was calculated as the difference between date of birth and the date of the last claim for the hospital stay. A modified length-of-stay variable, labeled "short stay," was defined as discharge within 1 day of vaginal birth or within 2 days of cesarean birth. Postdischarge health care use was measured by emergency department, primary care, and home health visits within 14 days of discharge.

Regional variations in length of stay, short stay, primary care visits, and rehospitalization rates over time were determined for 6 previously described perinatal service/education regions of the state.¹⁶

The primary diagnoses at rehospitalization and changes in rehospitalization rates for the most common diagnoses over the study period were also examined.

We hypothesized that there would be a shift toward shorter length of stay and that this would be associated with an increase in rehospitalization rates. A preliminary study in 1 Ohio county revealed a rehospitalization rate of 1.2% within 7 days and 2.4% within 14 days of discharge for healthy full-term neonates. A sample of 39,000 patients was deemed necessary to detect a 20% change in the rehospitalization rate within 7 days of discharge with an α of .05 and 90% power. A sample of 19,000 was required to detect a 20% change in rehospitalization within 14 days of discharge with 90% power at α=.05 level.

Rehospitalization rates within 7 and 14 days of discharge were calculated. The unit of analysis was the patient, so that only the first rehospitalization within the specified period of time was measured. Mantel-Haenszel χ² tests (analysis of trend for categorical variables) and t tests (continuous variables) were performed. Univariate analyses were performed to compare patient and maternal characteristics of neonates rehospitalized within 7 and 14 days of discharge with those not rehospitalized. We then developed a multivariate logistic regression model for each primary outcome of interest. The primary independent variable was whether the hospitalization was a short stay. Potential confounders in the model included maternal age, race, education, parity, prenatal care adequacy, year of birth, gestational age, birth weight, route of delivery, 5-minute Apgar score, and region of birth. Odds ratios (ORs) and 95% confidence intervals (CIs) were calculated for the likelihood of being readmitted within 7 or 14 days of discharge.

To explore the possibility that the change over time in the rate of short stays was different among the 6 regions of the state, 2 logistic models of short stay were created. The first contained birth year and region of birth (represented by binary dummy variables), and the second also included an interaction variable for birth year and region. To determine if the 2 models were significantly different, the log-likelihoods of the 2 models were computed and significance determined from the χ² distribution. Similar models were created to determine if the 6 regions of the state differed in change in rehospitalization rates over the study period. All statistical analyses were performed using PC-SAS software (Release 6.12, SAS Institute Inc, Cary, NC).

There were 623,266 births recorded in Ohio between July 1, 1991, and June 15, 1995. Of these, 194,748 neonates and their mothers were identified as Medicaid recipients with valid claims. From that group, 102,678 healthy full-term newborns (diagnosis related group 391, gestational age greater than 37 weeks, and birth weight greater than 2000 g) were still receiving Medicaid 30 days after birth, forming the final study cohort.

An examination of maternal characteristics (Table 1) showed a small, but statistically significant, increase over time in the proportion of women who were white, primiparous, had 2 or fewer prenatal visits, or completed high school and a decrease in the proportion of married mothers. Gestational age, birth weight, and route of delivery remained fairly stable over the period of study. There were no statistically significant differences in these demographic variables between the 6 regions.

The mean (SD) hospital length of stay for newborns decreased in each of the years examined, from 2.2 (1.0) days in 1991 to 1.6 (0.9) days in 1995, a decrease of 27% (P<.001). In 1991, 20% of newborns were discharged 1 day after birth and 76% after 2 days. By 1995, 54% were discharged by 1 day and 88% by 2 days after birth. There was a concomitant and statistically significant change in the percentage of newborns with short stays (Table 1). The percentage of newborns discharged after a short stay increased from 21.0% in 1991 to 59.8% in 1995, an increase of 185% (P<.001). The increase in the short stay rate was statistically significant for both vaginal and cesarean deliveries (P<.001).

There was considerable regional variation in length of stay for newborns born between 1991 and 1995. Mean length of stay over the course of the study was 1.9 days. It ranged from a low of 1.8 days in the southwest region of the state to 2.1 days in the east central area. Mean length of stay decreased in each year of the study in every region. The southwest region showed the greatest change, a decrease of 43.5% (Table 2), while the east central region showed the least change, decreasing 17.4%.

The percentage of newborns discharged after a short stay (≤1 day after vaginal delivery or ≤2 days after cesarean birth) varied widely between regions and over the course of the study (Figure 1) but increased each year in every region (P<.001). The southwest region (Table 2) showed the greatest increase (553%). The northeast region had the smallest percentage increase (117%). When modeled with the birth region–year of birth interaction variable, the change in short stay over time was found to differ significantly across the 6 regions for both vaginal and cesarean births (P<.001).

The proportion of newborns who received a primary care visit within 14 days of discharge increased in each year of the study (Table 3), with an overall increase of 117% (P=.001). The proportion of newborns who received a home health visit within 14 days of discharge, while still low, also increased significantly (P=.001). There was a small, but statistically significant, increase in the proportion of newborns who visited the emergency department within 14 days of discharge (P=.001).

The increased proportion of neonates receiving a primary care visit within 14 days of discharge was seen in each region (Figure 2). The proportion of newborns receiving a home health visit within 14 days of discharge differed significantly among the 6 regions (P=.001). The proportion having a primary care or emergency department visit did not vary significantly.

In refutation of our hypothesis, there was a statistically significant 23% decrease in rehospitalization rates for healthy full-term newborns during the course of the study (Table 4). Rehospitalization rates within 7 days of discharge decreased from 1.3% in 1991 to 1.0% in 1995 (P=.01). The average rehospitalization rate within 7 days of discharge during the study period was 1.1%. Similarly, rehospitalization rates within 14 days of discharge decreased by 19%, from 2.1% in 1991 to 1.7% in 1995 (P=.03). The average rate of rehospitalization within 14 days of discharge over the 4 years was 1.8%. In this population, 3.3% of the rehospitalized newborns were hospitalized more than once within 14 days of the original discharge.

Five of the 6 regions showed decreasing rehospitalization rates within 7 days (Table 4), but these trends were not statistically significant. These same 5 regions also had a decrease in rehospitalization rates within 14 days of discharge. For 1 region, the decrease was statistically significant (P=.04). The 6 regions did not significantly differ in their rate of decline in rehospitalization rates.

Mothers of newborns readmitted within 7 days of discharge were significantly more likely to be married, primiparous, and white (P<.001 for all). Rehospitalized newborns were more likely to have been delivered vaginally (P<.001), have lower 5-minute Apgar scores (8.98 vs. 9.05, P<.001), shorter mean gestations (39.2 vs 39.5 weeks, P<.001) and shorter mean length of stays (1.86 vs 1.94 days, P<.005). Maternal age, maternal education, adequacy of prenatal care, birth weight, and short stay rates were not statistically significant factors for rehospitalization by 7 days after discharge.

Mothers of newborns readmitted within 14 days of discharge were significantly more likely to be primiparous (P=.02), white (P=.01) and younger (mean age, 22.4 vs 22.8 y, P<.005). Rehospitalized newborns were more likely to have been delivered vaginally (P<.002), have shorter gestations (39.3 vs 39.5 weeks, P<.001), lower birth weights (3345 vs 3367 g, P=.04), lower 5-minute Apgar scores (9.00 vs 9.04, P<.001) and shorter lengths of stay (1.89 vs 1.94 days, P=.04). Maternal marital status, education, mean parity, prenatal care adequacy, and short stay rates were not significantly associated with rehospitalization by 14 days after discharge.

After adjustment for other covariates, maternal factors significantly associated with an increased risk of newborn rehospitalization within 7 days of discharge (Table 5) included primiparity (OR, 1.52 [95% CI, 1.26-1.83]), white race (OR, 1.34 [95% CI, 1.13-1.59]), and being married (OR, 1.30 [95% CI, 1.14-1.48]). Factors associated with a decrease in the risk of rehospitalization included lower 5-minute Apgar score (OR, 0.78 [95% CI, 0.71-0.87]), birth in the southeast region of the state (OR, 0.76 [95% CI, 0.61-0.95]), cesarean birth (OR, 0.68 [95% CI, 0.56-0.82]), and singleton birth (OR, 0.53 [95% CI, 0.32-0.88]). In addition, the odds of rehospitalization decreased by 9% for each ensuing year of birth (OR, 0.91 [95% CI, 0.87-0.96]) and by 24% for each additional week of gestation (OR, 0.76 [95% CI, 0.72-0.80]). Maternal age, mean parity, adequacy of prenatal care, maternal education, birth weight, short stay rate, and birth in other regions of the state were not statistically significant predictors of 7-day rehospitalization rates.

Similarly, adjusting for all factors, white race (OR, 1.21 [95% CI, 1.07-1.37]), primiparity (OR, 1.17 [95% CI, 1.02-1.34]), and being married (OR, 1.14 [95% CI, 1.03-1.26]) were associated with a statistically significant increase in the risk of rehospitalization within 14 days of discharge. The odds of rehospitalization decreased 5% for each succeeding year of birth (OR, 0.95 [95% CI, 0.91-0.99]). Birth in the southeast region of the state (OR, 0.82 [95% CI, 0.69-0.97]), higher 5-minute Apgar score (OR, 0.82 [95% CI, 0.75-0.89]), and cesarean birth (OR, 0.82 [95% CI, 0.71-0.94]) were associated with a significantly decreased risk of rehospitalization. Also, the odds of rehospitalization decreased 1% for each 1-year increase in maternal age (OR, 0.99 [95% CI, 0.97-1.00]) and 16% for each additional week of gestation (OR, 0.84 [95% CI, 0.81-0.88]). Birth weight, singleton birth, parity, adequate prenatal care, cesarean birth, short stay, maternal education, and birth in other regions of the state were not statistically significant factors for rehospitalization within 14 days of discharge.

When the results were stratified by race, birth in the southeast region and age were no longer statistically significant factors for rehospitalization within 14 days of discharge for white women. Primiparity, marriage, year of birth, shorter gestation, cesarean birth, lower 5-minute Apgar score, along with the new factors of singleton birth and birth in the east central region, were statistically significant. For nonwhite mothers, only birth in the northeast region was a statistically significant factor.

When the results were stratified by age, for mothers younger than 20 years, longer gestation, later year of birth, and higher 5-minute Apgar score were associated with a statistically significant decrease in the risk of rehospitalization within 14 days of discharge. For mothers older than 20 years, marriage, white race, primiparity, lower 5-minute Apgar score, shorter gestation, birth in the southeast region, and cesarean birth continued to be statistically significant factors for rehospitalization within 14 days of discharge.

Most rehospitalizations were due to jaundice, which accounted for 42% and 26% of hospitalizations within 7 days and 14 days of discharge, respectively (Table 6). Readmission rates for jaundice did not change significantly during the 4 years of the study. However, readmission rates for jaundice within both 7 and 14 days of discharge did differ significantly among the 6 regions of the state (P<.001).

The second most common reason for rehospitalization was respiratory problems (8% and 10% of hospitalizations within 7 and 14 days of discharge, respectively). Other diagnoses associated with rehospitalization included fever, infections, disorders of the digestive system, bronchiolitis, dehydration, and feeding problems. Only rehospitalizations for bronchiolitis increased significantly over the course of the study (P=.009.)

Unlike previous reports, our analysis of full-term Medicaid births in Ohio showed that a marked trend toward earlier discharge of newborns from 1991 to 1995 was not associated with greater rates of rehospitalization for newborns during the same period. In fact, rehospitalizations for Medicaid newborns decreased significantly during this time, both within 7 and 14 days after discharge. While previous studies lacked the power to detect statistically significant differences due to small numbers of rehospitalizations, our study sample of 102,678 would have allowed us to detect a 13% change in readmission rates (from 1.3% to 1.13%) with 90% power.

Our rehospitalization rates of 1.1% within 7 days and 1.8% within 14 days of discharge are similar to those reported elsewhere.^5,10,17- 19 The lack of an increase in rehospitalizations seen in our study is similar to results reported for Medicaid patients in other states.^19- 20 Using Maryland data from 1989 to 1992, Fox and Kanarek¹⁹ showed that the odds of readmission for healthy Medicaid newborns with shorter lengths of stay were similar to those with longer lengths of stay. On the other hand, in a large population-based study from Ontario,⁵ shorter lengths of stay were associated temporally with more rehospitalizations. The Ontario study included all newborns and, because readmission data were not linked to birth hospital data, adjustment for confounds that might affect readmission rates was not done. In a case-control study of newborns in Washington State,²¹ risk of readmission by 7, 14, and 28 days increased following early discharge. The groups at greatest risk were those born to young or primiparous women and to mothers with premature rupture of membranes. Thus, the risk factors for readmission identified in the Washington study are similar to those seen in our study.

The primary reasons for rehospitalization in all of the aforementioned studies were jaundice, breastfeeding failure, or dehydration. The risk for clinically significant jaundice is known to be greater in breastfed newborns compared with formula-fed newborns.^3,10,17,20 Historically, breastfeeding rates are low in the Medicaid population.^22- 23 The overall breastfeeding rate in the United States in the early postpartum period has been reported to be 54%, while the rate for low-income mothers is only 32%.²⁴ Although breastfeeding information was not available to us, the likely low breastfeeding rate in our population may have played an important role in the low rehospitalization rates that we observed.

Another significant reason for the lower rehospitalization rates may have been better postdischarge coordination of care. Primary care visits and home health care visits increased dramatically and there was a small, but statistically significant, increase in visits to the emergency department. This shift toward ambulatory care, along with low rates of breastfeeding, may explain the reduced rates of rehospitalization we observed even in the face of decreasing hospital lengths of stay.

Several studies have explored regional variations within a single state for a single payer and patient population. Spong and Hulet²⁵ found significant variation in length of stay and in the proportion of newborns who have a short stay in different regions of the country. Our data show similar variations and may reflect the influence of local market factors. For example, in 1991, a powerful consortium of businesses in southwestern Ohio entered into a collaborative partnership with area hospitals to lower the cost of employee health care.^26- 27 Benchmarking spurred competition among the hospitals, led to a dramatic reduction in health care costs, and resulted in a coordinated newborn early discharge program. Regional rates of readmissions may also reflect local variation in the patterns and type of follow-up care. Although the demographic characteristics of the population were similar among the 6 perinatal regions, the pattern of home visits did vary significantly and may be an important factor.

Studies such as ours, using administrative data, are limited by the lack of detail necessary to estimate length of stay in hours, to minimize the likelihood of misclassification of short stay. To address this possible problem, we examined length of stay for vaginal deliveries in a single hospital, comparing short stay as it was defined in the present study with short stay as determined by the number of hours between birth and discharge. This analysis revealed that only 4.9% of those who should have been classified as short stay were not. An additional 2.5% were classified as short stay, but should not have been. Thus, we believe that misclassification errors were unlikely to influence our findings.

In summary, in this population-based study of Medicaid newborns in Ohio, significant decreases in postnatal length of stay were not associated with an increase in rehospitalizations. It is important to state that reduced readmissions do not necessarily imply better or comparable health of newborns, especially when the same market forces influencing length of stay may discourage rehospitalization. Rehospitalization may represent only 1 possible outcome of early discharge; a shift to outpatient care, differences in breastfeeding rates, and changing patterns of postdischarge care, such as home visits and use of home phototherapy, may have played a role. Additional studies should be undertaken to determine the reasons for these findings and to confirm them in other low-income populations.