 |
|
 |
|
 |
 |
|
 |
 |
 |
 |
 |
 |
 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
* Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim Health Care, Boston, Massachusetts
Children’s Hospital, Boston, Massachusetts
 |
ABSTRACT |
|---|
 TOP ABSTRACT METHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Methods. Interrupted time series analysis was conducted on retrospective data from the automated medical records of a large Massachusetts health maintenance organization (HMO). A population of 20 366 mother–infant pairs with normal vaginal deliveries between October 1990 and March 1998 was identified. The interventions included a new HMO protocol in 1994 of 1 hospital overnight after delivery, plus a nurse home visit, then the Massachusetts’ 1996 minimum coverage law. Postpartum length of stay, clinical evaluation on day 3 or 4 of life, health center visits up to day 21, health center diagnoses of jaundice or feeding problems, bilirubin testing and test severity, rehospitalizations, and emergency department visits were measured.
Results. Postpartum stays <2 nights rose from 28% of newborns before implementation of the program to 70% immediately after implementation. Later, this rate fell from 66% before the mandate to 21% just after the law went into effect. Day 3 or 4 evaluation rose from 24.5% to 64% after the program, then dropped somewhat to 53% after the mandate. Controlling for longer-term trends in health center visits, implementation of the early discharge program was associated with approximately 1 extra visit for every 4 newborns within the first 21 days of life. The state mandate did not affect health center visit rates. Jaundice diagnoses were flat at 8% of newborns during the baseline, then rose to a constant 11% throughout the program and postmandate periods. Bilirubin testing of newborns also rose by 3.4 percentage points at the time of program implementation, and the proportion of tested newborns with results calling for at least consideration of phototherapy rose by 6 percentage points. Phototherapy use rose from a flat 1.8% to 2.4% of newborns after program implementation. Feeding problem diagnoses more than doubled at the time of program implementation and remained elevated after the mandate. Rehospitalizations overall and specifically for jaundice were constant over time, whereas more rare emergency department visits for jaundice dropped from 0.3% of newborns to 0 on program implementation.
Conclusions. Sudden increases in jaundice-related measures and identification of infant feeding problems were not associated with changes in length of stay in this setting. Instead, these increases seem to be the result of more frequent evaluation of newborns during the critical day 3 to 4 period and may also have been elevated by a new climate of concern about neonatal vulnerability. "Ascertainment bias" may have confounded findings in previous reports that raised concerns about the safety of early discharge.
Key Words: newborn infant • length of stay • health policy • neonatal jaundice • research design
Abbreviations: LOS, length of stay • HMO, health maintenance organization • AAP, American Academy of Pediatrics • TSB, total serum bilirubin • ED, emergency department
The clinical and political controversies over reductions in postpartum lengths of stay (LOSs) in the early to mid-1990s have been well documented.1–3 The design limitations and contradictory findings that have consistently plagued the literature assessing the safety of LOS reductions have also been ably described.4–8 Nevertheless, Congress and >40 states have passed legislation since 1995 aimed at discouraging early discharge.1,9,10
Unfortunately, some of the most worrisome reports linking adverse outcomes to LOS reductions have relied on weak methods.11–15 Methodological problems include the inability to control for potential confounding factors such as rising rates of breastfeeding and changes in medical practice other than hospital discharge timing.
Our primary goal was to examine the effects of LOS changes on key measures of jaundice and feeding-related problems in newborns using a strong, quasi-experimental design—an interrupted time series analysis16—evaluating the effects of a policy that resulted in a sharp decrease in LOS as well as sudden withdrawal of this policy. We also reexamine the conclusions from frequently cited previous reports in light of our findings and discuss future directions for research in this area.
|
METHODS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
In the United States, multiple case reports of severe breastfeeding malnutrition and hypernatremia14 and kernicterus15,18 added to widespread concerns. These authors concluded that there were genuine increases in these very rare outcomes during the early 1990s (although very small samples of 16 and 41 cases, respectively, informed their analyses). While acknowledging a range of possible contributing factors, the authors highlighted a likely role played by LOS reductions. Others19,20 have cautioned against overinterpretation of case reports, which lack denominators and are subject to unmeasurable fluctuations in clinical ascertainment and communication.
The Problem of Confounding
The LOS reductions of the 1990s did not occur in isolation but were accompanied in many settings by changes in service mix, secular increases in breastfeeding, and intensified attention to neonatal health issues. Each of these factors has the potential to lead to increases in ascertainment of newborn problems and thereby confound an apparent relationship with early discharge.
Early discharge programs frequently included enhancements of postdischarge services, such as added follow-up home visits, office visits in the first week, and/or lactation assistance.4,6,21–24 These newer arrangements contrast with "traditional" postdischarge care from previous years, when LOS was typically 2 or 3 postpartum nights and the first office visit was scheduled at 2 weeks.21,25 Such changes in the frequency and timing of clinical contacts could theoretically reduce adverse outcomes through earlier intervention, or these changes could result in more problem detection through increased observation.
Several recent studies have found that breastfeeding rates increased during the 1990s.26–28 In our own investigation,23 we found steady secular increases in breastfeeding initiation (and unchanging rates of continuation among breastfeeding initiators) between 1990 and 1998 that were unaffected by sharp changes in LOS as a result of an early discharge program and a subsequent minimum stay mandate. Breastfeeding, although considered beneficial on the whole, is strongly associated with jaundice, dehydration, and feeding problems among neonates.29–32 Increases in breastfeeding therefore could lead to increases in adverse outcomes, independent of LOS. It is also possible that problems such as jaundice could be detected more often when more mothers seek help for breastfeeding difficulties.
Finally, ascertainment of neonatal health problems could rise as a result of heightened awareness among clinicians. Within a single health system, this heightened awareness could result from program start-up activities.21,33,34 Debates in the popular media1–3 or in the medical literature25,35,36 about early discharge policies, as well as new treatment guidelines for hyperbilirubinemia in newborns,19,20,37 could also contribute to heightened awareness over a broader geographic area. The increased ascertainment might then manifest itself as an increase in diagnosis coding or treatment.
Our study takes advantage of a natural experiment in which all of these potential confounders (changes in service mix, increasing breastfeeding rates, and cultural stimuli) coincided with LOS reductions. Later, the LOS reductions were reversed by legislation, but the potential confounders remained in place. Our aim was to examine changes in measures of newborn jaundice and feeding problems over time and draw inferences about the relative influence of LOS versus other factors on these changes.
Study Design
The methods of our investigation are described in greater detail elsewhere.22,23 Using automated medical records and claims data from 7.5 years of births within a large Massachusetts health maintenance organization (HMO), we evaluated changes in measures of newborn jaundice and feeding problems coincident with an early discharge program and a subsequent state minimum-stay law. Changes in outcome variables were estimated with interrupted time series analyses, which compare the 3 time periods demarcated by the policy shifts while controlling for underlying trends.
Setting and Interventions
The setting was a large, staff-model HMO (formerly known as Harvard Community Health Plan) in eastern Massachusetts. During the third quarter of 1994, the HMO implemented an early-discharge program21–23 that set a new standard of 1 postpartum overnight after uncomplicated vaginal deliveries, subject to attending clinicians’ discretion. The program also featured a home visit by a nurse within 48 hours of discharge, enhanced prenatal classes and screening for social risk, professional workshops, and distribution of new discharge protocols.
A Massachusetts 48-hour minimum coverage mandate effectively ended the HMO program in February 1996, although shorter hospital stays were permissible if mothers gave consent and were offered home visitation. (This meant that many women could have a 2-night postpartum stay as well as a home visit if the hospital reported discharge at 47 hours or fewer.) Federal law effective January 1998 extended the mandate to other states and health plans but did not change the policy environment for this study population.
Study Population
We collected data on all delivering mothers and newborn members in the HMO between October 1990 and March 1998. This window of observation spans a 4-year baseline period as well as the entire early discharge program period and 2 years after the mandate. Eligible infants for the study had to be HMO members for at least 45 of their first 60 days of life and have at least 1 medical record in their first 7 days. We matched mothers to infants using personal identifiers and dates and locations of services. For the analyses in this article, we excluded mother–infant pairs whose deliveries involved cesarean sections, more than 4 nights’ postpartum LOS, or unequal mother and infant LOS, because only vaginal deliveries without serious complications were subject to the policy changes of interest. There were 20 366 unique mother–infant pairs for these analyses, or an average of 679 births per quarter.
Data Sources and Measures
LOS and Follow-Up Services
We determined initial LOS in postpartum hospital nights using mothers’ claims data (discharge date minus infant’s date of birth). Claims data also include information on home visits conducted by outside vendors. Information on home visits by staff nurses and on newborn health center visits and diagnoses was available from detailed HMO automated medical records.22,23,38 We determined whether newborns received clinical evaluation on day 3 or 4 after birth by combining information on timing of hospital discharge, home visits, and health center visits.22 We also looked at numbers of visits per newborn to the HMO health centers during the first 21 days of life, an interval that would capture most early occurrences of the diagnoses of interest but still be close enough to birth for outcomes to be plausibly affected by LOS policies. We repeated our analyses using 10-day and 45-day intervals and obtained similar results (not shown). The endpoints for each of these follow-up intervals fall between peaks of routinely scheduled visits and are therefore insensitive to slight variations in patient flow over time.
Outpatient Diagnoses
Newborns were classified by whether they received diagnoses of interest during any health center visit in the period between initial discharge and day 21 of life. Newborns with "jaundice" had at least 1 of 3 codes for jaundice and hyperbilirubinemia. For "feeding problems," we looked for any of 11 specific and nonspecific feeding problem or dehydration codes. (Dehydration was combined with feeding problems because its appearance was too rare for separate analysis, and it often overlapped with diagnosis of feeding problems.) Home visit records, which do not contain coded diagnoses, contribute to our analyses only insofar as infants were referred by nurses to health centers and diagnosed there. We disregarded diagnostic information that appeared in clinicians’ notes23 and records from telephone contacts.39 Diagnoses in the medical records have no bearing on compensation for HMO staff clinicians, so their validity was not threatened by "upcoding."
Bilirubin Testing and Test Severity
The automated medical records provided information on bilirubin tests conducted at health centers. Data on hospital-based bilirubin tests were not available. Test results were classified into age-adjusted "severity levels" on the basis of the management algorithm from the American Academy of Pediatrics (AAP) practice guideline.37 Because we had newborn age only in days (not hours, as in the AAP algorithm), we counted tests taken on day 1 or 2 after birth as occurring "at age = 25 to 48 hours," day 3 as "49 to 72 hours," and day 4 and later as ">72 hours." (This conversion overestimates severity for a small portion of cases but in a way expected to be constant over time.) More than 96% of first bilirubin tests were taken on day 3 or later. We defined severity level 1 as a total serum bilirubin (TSB) value corresponding with an AAP management recommendation of "consider phototherapy" for that age. Severity level 2 corresponded to the "phototherapy" recommendation, and much rarer levels 3 and 4 corresponded to indications for intensive treatments including exchange transfusion. (Level 0 applied to benign TSB values.) Tested newborns were classified according to their "worst" severity level in the 21 days postpartum. For 96% of tested newborns, the severity of the worst test was equivalent to the severity of the first test. Most tested newborns (55%) received only 1 test.
Phototherapy
We measured phototherapy use with an automated text search of newborn medical records for terms indicating phototherapy (eg, "lights," "Wallaby"). Contextual language around each term found was then checked manually to confirm receipt of phototherapy. Newborns were classified as "phototherapy use mentioned within 21 days" or "not mentioned," irrespective of service date, location, or intensity. Both home and hospital use of phototherapy is recorded in newborn medical records.
Hospital Services
Information on rehospitalizations after initial discharge and emergency department (ED) visits was obtained from newborns’ claims data. Primary diagnoses of feeding problems or jaundice were determined using International Classification of Diseases, Ninth Revision codes. Again, dehydration codes were combined with feeding-related issues because of low numbers.
Demographic Information
We obtained demographic information on mother–infant pairs such as race, maternal age and parity, Medicaid enrollment, and neighborhood income and education levels, as well as clinical measures at birth, to ensure that population characteristics were stable over time. These data came from medical records, membership files, 1990 US Census data, and some inpatient hospital records.22,23
Statistical Analyses
We aggregated all study measures into 3-month intervals based on date of birth (n = 30 quarters). We used segmented linear time-series regression models40–42 to estimate changes in all rates (or means) at the points of change in LOS policies. Our models included a constant term, a baseline time trend, and terms to estimate changes in the level and trend of outcome rates on program implementation or the state mandate. Transitional quarters—Q3`94 and Q1`96—were omitted because they were not clearly characterized by a single policy. Time series analyses were conducted using SAS Proc Autoreg.43 In earlier investigations,22,23 we determined that there were no discontinuities in trends for population characteristics that could potentially confound our results. Analyses in the present study were repeated with 2 additional terms—the percentage of primiparous mothers and the percentage "poor"—to control for slight quarterly variations in population demographics.22,23 The controlled and uncontrolled results were essentially identical, and we report the latter.
|
RESULTS |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
|
Jaundice Diagnoses, Bilirubin Testing, and Phototherapy
Our jaundice-related measures were internally consistent during all periods. Nine percent of all study newborns received postdischarge health center diagnoses of jaundice, and 76% of those also received outpatient bilirubin tests. Only 0.5% of newborns without outpatient health center jaundice diagnoses had an outpatient bilirubin test of severity level 1 or higher. Phototherapy use (sometimes occurring before hospital discharge) was recorded for 7.5% of newborns who had outpatient bilirubin tests of severity = 0, for 24.5% with severity = 1, and for 65.5% with severity = 2.
The percentage of newborns who received a jaundice diagnosis at a health center changed considerably over time (Fig 2A). The rate was stable at 7.6% during the baseline period, then increased suddenly by 3.0% of newborns at the beginning of the early discharge program (P < .0001), remaining at 10.6% throughout the later years of observation, with no changes associated with the mandate. (An unusually high observation at Q4`95 was contemporaneous with spikes in both pre- and postdischarge coding of "rule-out sepsis" that winter; this period is excluded from statistical analyses of all jaundice-related measures.)
|
As testing rates increased, the severity of test results also increased perceptibly (Fig 2B). The percentage of tested newborns who had TSB severities of level 1 or higher was constant at 28.7% during baseline and rose to 34.9% immediately after program implementation (P = .003). Likewise, the percentage at severity level 2 or higher rose from a constant 9.8% to 13.2% of tested newborns at the same time (P = .022). No changes were observed in either percentage at the time of the LOS mandate in early 1996.
Phototherapy use was estimated to be constant at 1.8% of newborns before program implementation, after which it rose to a constant 2.4% of newborns (P = .014). No significant change in phototherapy use was associated with the mandate.
Feeding Problem Diagnoses
The percentage of newborns who received a diagnosis of feeding problems (including dehydration) at a health center visit within 21 days was also considerably higher in later years (Fig 3). We estimated a long-term secular increase of 1.0 percentage points per year (P < .0001) throughout the study period, which might be expected given steadily rising rates of breastfeeding initiation in this population.23 The estimated rate of feeding problem diagnoses rose from 1.5% in late 1990 to 5.0% at Q2`94. We estimated a sudden increase of 5.2 percentage points over the underlying trend on program implementation and a smaller reduction of –2.5 percentage points after the mandate (P < .0001), a point when the rate of day 3 or 4 follow-up also declined.
|
Postdischarge Hospital-Based Events
Overall, the percentage of newborns who were rehospitalized during the first 21 days of life and the percentage who visited an ED (without rehospitalization) both were constant throughout the study, at 2.3% and 2.0%, respectively. The rate of rehospitalization for jaundice was also constant at 0.5%. Although ED visits for jaundice were rare, 30 newborns had them during the baseline period, and only 3 had them during the early discharge program and postmandate years combined. This virtual disappearance, a drop from a constant 0.3% to a constant 0.0% at Q4`94, was significant at P < .001. Rehospitalizations and ED visits for feeding/dehydration problems were too rare and unevenly distributed to permit reliable statistical analysis (n = 43 and 41, respectively).
|
DISCUSSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
Our evidence supports the hypothesis that reduced LOS in this HMO did not result in poorer health outcomes for newborns. If early discharge caused true increases in jaundice and feeding problems, then we would expect a reduction in related measures after the reversal of that policy. However, there was no such reduction. Therefore, we conclude that changes in diagnoses, testing, and therapy reflect changes in clinical ascertainment. Several phenomena that are likely to influence ascertainment occurred nearly simultaneously in this setting: increased newborn contact at home and at health centers, program implementation activities, 2 revised national guidelines,25,37 and increased public concern about perinatal care. We cannot determine the separate impacts of these factors. Nevertheless, ascertainment or heightened vigilance is the most likely explanation for our findings.
The sudden implementation of LOS policies within a stable HMO population allows us to distinguish policy effects from long-term secular trends. The off-on-off aspect of this natural experiment provided additional control, allowing us to distinguish effects specifically as a result of LOS (which reversed direction) from effects as a result of other changes, which endured. A simple analysis assessing only the first policy might have led to an erroneous conclusion that increased newborn jaundice and feeding difficulties were attributable to early discharge.
Our findings offer insights for the interpretation of previous reports. Studies11–15 that have raised concerns about early discharge by temporally associating LOS reductions with increases in adverse jaundice- or feeding-related events lacked our strengths in terms of discrete timing and reversal of interventions. Apparent increases in newborn problems within those studies may have resulted from scattered, incremental increases in ascertainment, especially at early follow-up visits that often took the place of traditional longer stays. The possibility exists that more attentive care created these apparent effects. Our data, from a unified system, clearly distinguish between long-term trends and sudden policy-related shifts in practice.
Of course, this study is not a replication of earlier research; the study settings and interventions are different. Although we believe that increases in breastfeeding within this setting did not contribute to sudden changes in diagnoses and testing (because the breastfeeding changes were gradual and long-term23), breastfeeding increases might have been a major factor in longer-term trends in jaundice- and feeding-related rehospitalizations elsewhere. In addition, the Canadian birth hospitalizations seem to have been generally longer, such that reductions there may have occurred around the more clinically important day 3 of life, in contrast to the shift in LOS from 2 days to 1 day commonly seen in the United States.8,9,22,23,44 Thus, LOS reductions and increased rehospitalizations in Canada might have been directly linked, despite the lack of strong causal evidence in published studies.
Unfortunately, even with our large study population, we did not have sufficient statistical power to resolve uncertainty about feeding-related hospital events, and we could not address kernicterus. In general, only very large public data sets permit detailed quantitative analysis of rehospitalizations,45–47 but their drawback is a lack of detailed information about outpatient care and private-sector interventions.
Rehospitalizations did not change perceptibly in this setting, but we did note other changes in closely related measures of practice and utilization (increased diagnosing, testing, and phototherapy; reduced ED visits for jaundice). In the extensive literature assessing early discharge,4–8 rehospitalizations have been the main outcome of interest. Others have pointed out that rehospitalizations are only a proxy measure for health status and that population health status could remain stable while rehospitalizations increase with an early discharge program simply because predictable problems begin to arise postdischarge rather than predischarge.5,29 It is worth pointing out that rehospitalizations are also a marker of clinical practice style and that changes in rehospitalization rates could as likely stem from early follow-up protocols as from initial discharge protocols. Health services research relies heavily on utilization measures as markers for true health outcomes, but when the subject of evaluation (eg, LOS at birth) is also a key utilization measure, evaluative research can become unavoidably circular.
The literature on early discharge is further complicated by the nature of jaundice and feeding problems. These are especially attractive targets for study because they are common enough to permit statistical analysis and are to some extent manageable, thus plausibly related to LOS, but this commonness and manageability means that treatment is relatively discretionary and varied,5,20,45 making evaluation via treatment proxies for health status (eg, rehospitalization) problematic.
The lack of a rehospitalization effect in this setting may indicate a lack of generalizability to other settings, which is our main limitation. The HMO that we studied has a reputation for high-quality health care.48 The patient population, although diverse,22,23 has higher levels of education and income than the United States overall. A conservative approach to early discharge, good continuity of care, and relative patient sophistication during all study years might have minimized the degree to which intensified follow-up identified additional cases requiring hospital treatment.
However, the special setting does not necessarily undermine our conclusions with respect to the effects of LOS versus ascertainment on other outcomes. Indeed, if a strong link between early discharge and rehospitalization were to be found in other settings, then our findings raise the question of whether standard 48-hour stays without a routine visit on day 3 or 4 represents neglectful care for some populations. This pattern of care was effectively favored by many states’ and the federal mandates, which have no provisions regarding clinical follow-up. In Massachusetts, the continuation of elevated home visit rates after the mandate seems to have been an unintended effect of the legislation.
|
CONCLUSION |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
We are especially grateful to Inna Dashevsky for data-set extraction.
|
FOOTNOTES |
|---|
Reprint requests to (D.R.-D.) Department of Ambulatory Care and Prevention, Harvard Medical School and Harvard Pilgrim Health Care, 133 Brookline Ave, 6th Fl, Boston, MA 02215. E-mail: drossdeg{at}hms.harvard.edu
|
REFERENCES |
|---|
|
TOPABSTRACTMETHODSRESULTSDISCUSSIONCONCLUSIONREFERENCES |
|---|
This article has been cited by other articles:
 |
|
 |
|
  W. H. Dow, D. M. Harris, and Z. Liu Differential Effectiveness in Patient Protection Laws: What Are the Causes? An Example from the Drive-Through Delivery Laws Journal of Health Politics Policy and Law, December 1, 2006; 31(6): 1107 - 1127. [Abstract] [PDF]
|
|
|
|
 |
|
 I. M. Paul, E. B. Lehman, C. S. Hollenbeak, and M. J. Maisels Preventable Newborn Readmissions Since Passage of the Newborns' and Mothers' Health Protection Act Pediatrics, December 1, 2006; 118(6): 2349 - 2358. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
A. Datar and N. Sood Impact of Postpartum Hospital-Stay Legislation on Newborn Length of Stay, Readmission, and Mortality in California Pediatrics, July 1, 2006; 118(1): 63 - 72. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G. K. Suresh and R. E. Clark Cost-Effectiveness of Strategies That Are Intended to Prevent Kernicterus in Newborn Infants Pediatrics, October 1, 2004; 114(4): 917 - 924. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
I. M. Paul, T. A. Phillips, M. D. Widome, and C. S. Hollenbeak Cost-Effectiveness of Postnatal Home Nursing Visits for Prevention of Hospital Care for Jaundice and Dehydration Pediatrics, October 1, 2004; 114(4): 1015 - 1022. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
||
|
||
Home | My Pediatrics | Journal Information | Current Issue | Past Issues | Subscriptions & Services | Contact Us Click here for faster international access © 2004 American Academy of Pediatrics. All rights reserved. |
||
|
||
