Postpartum length of stay has been decreasing during the past 30 years, from about 7 days in the 1960s to 4 days in the 1970s to about 2 days in the early 1990s.1 There is concern that shorter length of stay might adversely affect maternal and neonatal outcomes.2–7 With these concerns in mind, the Newborns' and Mothers' Health Protection Act of 1996 was enacted into law January 1, 1998 in the state of Utah.
The Newborns' and Mothers' Health Protection Act prohibits payers from restricting “benefits for any hospital length of stay in connection with childbirth for the mother or newborn child, after a normal vaginal delivery, to less than 48 hours.”8 According to Mandl et al,9 this legislation is founded on the philosophy that length of postdelivery hospital stay should be based on the unique characteristics of each mother and her newborn child and that timing of the hospital discharge should be made by the attending provider in consultation with the mother.
Although the Newborns' and Mothers' Health Protection Act legislation became law on January 1, 1998, few studies to date have examined whether implementation of the legislation has affected maternal or neonatal outcomes. Before implementation of the Newborns' and Mothers' Health Protection Act legislation, most research studied the effect of early postpartum discharge on maternal and neonatal outcomes. Much of that research found that early postpartum discharge was not associated with poorer maternal or neonatal outcomes.10–20 However, other recent research suggested that postpartum stays shorter than 30 hours might negatively affect neonatal morbidity and mortality.21–23 Most research examining the effect of postpartum length of stay on maternal and neonatal outcomes has been conducted with healthy neonatal populations receiving prenatal care. There is less conclusive evidence regarding the effect of reduced length of stay for unhealthy populations receiving insufficient prenatal care.24
Few studies have examined the effect of Newborns' and Mothers' Health Protection Act legislation on patient satisfaction. Earlier studies that included women who expressed interest in voluntary postpartum discharge found that women with early discharge were more satisfied than women with longer hospital stays.10,11,25,26 Conversely, when early discharge is mandated, women with longer stays were often more satisfied than those with early discharge.27
Given the paucity of research examining the effect of the Newborns' and Mothers' Health Protection Act on patient outcomes, the primary objective of this study was to evaluate the effect of the Newborns' and Mothers' Health Protection Act on maternal and neonatal length of stay, maternal patient satisfaction, and maternal and neonatal clinical outcomes as measured by hospital readmission, emergency room utilization, and total mother-infant hospitalization cost, adjusting for sociodemographic, health system, and clinical characteristics.
MATERIALS AND METHODS
Patients and Setting
All mother-infant clinical information was obtained through an electronic enterprise data warehouse. The electronic enterprise data warehouse is one of the most comprehensive in the managed care industry in terms of its ability to track inpatient, outpatient, and laboratory information. The system contains data collected from admitting departments, accounts receivable, and medical records; and it summarizes inpatient length of stay, emergency room visits, urgent care visits, and outpatient registrations. These clinical and financial data have been published previously.28–30
We conducted a retrospective, observational study of 18,023 mother-infant dyads at three urban hospitals in the greater Salt Lake City area before and after implementation of the Newborns' and Mothers' Health Protection Act legislation. The three urban hospitals included a primary, secondary, and tertiary care hospital (with a teaching hospital affiliation) that together perform about 10,000 deliveries annually. These hospitals are part of a vertically integrated, not-for-profit health care system, Intermountain Health Care, which includes 23 hospitals and 60 outpatient clinics in Utah, Wyoming, and Idaho, with 100,000 inpatient admissions and 4 million outpatient encounters annually, representing 50% of regional health care delivery. Physician relationships with Intermountain Health Care vary from employed and contractual health plan providers to physicians with courtesy admitting privileges only.
Mother-infant dyads were eligible for inclusion in the study if they delivered between July 1, 1996 and June 30, 1999 and had discharge diagnoses of normal vaginal delivery or normal newborn without complications (All Patient Refined Diagnosis Related Groups, Version 15, codes 541, 542, or 560 [mother] and 640 [newborn]). All Patient Refined Diagnosis Related Groups was selected as a screening because it includes a severity of illness measure for mothers and neonates. The severity of illness scores are assigned based on the following criteria: severity of illness of secondary diagnoses, maternal age, principal diagnosis, and the presence of certain nonoperative procedures.
Patients were excluded for the following reasons: major or extreme maternal or neonatal severity of illness (score ≥ 3), gestational age younger than 37 weeks, twin or higher-order multiple birth, neonatal birth weight of 2499 g or less, or maternal or neonatal postpartum length of stay less than 11 hours or more than 100 hours. Common diagnoses within major or extreme maternal or neonatal severity of illness included maternal comorbidities (eg, diabetes), fetal conditions developed during birth (eg, jaundice), and trauma during delivery (eg, fetal distress). These exclusion criteria are similar to those found in previously published research and are indicative of a non-normal maternal and neonatal population.9,18
Clinical and Financial Outcomes
Maternal length of stay was measured in hours from the time of delivery until hospital discharge. Neonatal length of stay included the time from birth until hospital discharge, measured in hours.
Using computer-assisted telephone interview survey methods, an independent research firm (with over 20 years of experience) surveyed each mother regarding their satisfaction with length of stay. Two survey methodologies were incorporated throughout the study. An effort was made to contact every mother who delivered at each of the three study hospitals (census sampling) from July 1, 1996 through December 31, 1998. Because of reduced resources available to conduct satisfaction research, the survey methodology changed to a random sample of 30% of the maternal population (random sampling) after January 1, 1999. Funding for this satisfaction research was provided by Intermountain Health Care.
Mothers were asked to rate their total hospital length of stay as “much too short, somewhat too short, about right, somewhat too long, or much too long.” The satisfaction question was pilot-tested on six focus groups and is similar to other questions reported in previous research.31,32 These data were then entered into the patient's electronic medical record.
Of the 18,024 mothers in the study, 14,872 were eligible for the satisfaction survey under the census sampling methodology, and 945 were eligible under the random sampling methodology. We contacted 9490 of the total 15,817 mothers eligible for interview, for a contact rate of 60%. The contact rate did not significantly differ under the two survey methodologies (census versus random sampling). Common reasons for noncontact included no answer, incorrect telephone number, or disconnected telephone. Of 9490 women contacted, 8395 completed the satisfaction survey for a response rate of 88.4%.
We assessed whether mothers or infants were readmitted for inpatient hospitalizations (within the entire Intermountain Health Care system) within 7 and 30 days after the index hospitalization. We also assessed whether mothers or infants utilized emergency services within the same time frames after their index hospitalizations. These time frames were selected on the basis of previous research that found that between 7 and 30 days after index hospitalization is the optimal period to study the effect of changes in length of stay on maternal and neonatal outcomes.9,11,12,15,16
Total maternal and neonatal costs (adjusted to 1997 US dollars) were calculated using the standard cost manager microcomputer software system. Total costs included fixed and variable costs and have been reported previously.29 Maternal costs were measured from the time of admission resulting in birth until hospital discharge. Neonatal costs were measured from the time of birth until hospital discharge.
The major independent variable in the study was legislation implementation status, treated as a binary measure. The preimplementation period includes 18 months (July 1, 1996 to December 31, 1997) immediately before enactment of the legislation. The postimplementation period includes 18 months (January 1, 1998 to June 30, 1999) immediately after enactment of the legislation.
The electronic enterprise data warehouse also included sociodemographic, health systems, and clinical measures. The data set included one sociodemographic measure—maternal chronological age (continuous) at the time of hospital admission—and two health system characteristics, hospital type (teaching hospital or non-teaching hospital) and maternal insurance status (private-indemnity insurance or managed care, Medicaid or Medicare, managed Medicaid, and self-pay). Managed Medicaid included patients enrolled in managed care plans financed by the Medicaid program on a capitated basis.
The following six clinical characteristics were included in the study: parity (multiparous or primiparous), gestational age (continuous), maternal severity of illness score (moderate or minor), neonatal severity of illness score (moderate or minor), neonatal birth weight (continuous), and 5-minute neonatal Apgar scores (continuous).
We performed all analyses using the Statistical Package for the Social Sciences (SPSS 8; SPSS Inc, Chicago, IL). Three outcome measures, satisfaction with length of stay (“about right” versus “somewhat too long/much too long” or “somewhat too short/much too short”), hospital readmission (≥1 hospitalization versus none), and emergency room use (≥1 visits versus none) were treated as binary measures. Length of stay was analyzed both as a continuous (measured in hours) and binary (≥ 48 hours versus < 48 hours) measure. Total hospitalization cost was analyzed as a continuous measure.
We calculated univariate statistics for control and outcome variables for the entire period (prelegislation and postlegislation period). Next, control and outcome variables were compared in the postlegislation versus prelegislation periods, using the χ2 test to assess the significance of the association between legislation implementation status and control or outcome variables for binary variables and one-way analysis of variance for continuous variables. Percentile ranks were also calculated for maternal and neonatal length of stay as well as total delivery hospitalization costs.
Multiple logistic regression modeling adjusted for age, insurance status, hospital type, parity, maternal and neonatal severity of illness, gestational age, birth weight, and 5-minute Apgar score. Adjusted odds ratios (OR) (postlegislation versus prelegislation period) were calculated for maternal and neonatal length of stay (≥48 hours), maternal satisfaction rated as “about right,” maternal and neonatal hospital readmission (within 7 and 30 days after index hospitalization), and maternal and neonatal emergency room use (within 7 and 30 days after index hospitalization). All control variables were forced into logistic models. Collinearity diagnostics were examined for final models using linear regression techniques to ensure that the effects of legislation implementation status were not obscured through extraneous collinear terms in final models.
Analysis of covariance33 was used to evaluate total mean hospitalization costs adjusted for maternal age, insurance status, hospital type, maternal severity of illness, gestational age, parity, neonatal birth weight, and neonatal severity of illness. These adjusted mean hospitalization costs were compared before and after implementation of the of the Newborns' and Mothers' Health Protection Act legislation.
Overall (including prelegislation and postlegislation periods), mothers were relatively young (mean 27.4 ± 5.4 years), had private health insurance (81.3%), were multiparous (70.6%), and delivered at a non-teaching hospital (55.0%) (Table 1). As might be expected, most mothers (69.1%) and newborns (95.9%) had low severity of illness scores. Similarly, mean gestational age (39.4 ± 1.02 weeks), mean neonatal birth weight (3423 ± 404.3 g), and 5-minute Apgar scores (8.95 ± .043) were indicative of a healthy maternal and neonatal population.
With the exception of insurance status and maternal severity of illness, distributions of univariate statistics did not change significantly from the prelegislation to postlegislation period. Mothers in the postlegislation period were slightly more likely to be privately insured (82.6% versus 79.8%, P < .001) and present with moderate severity of illness at index hospitalization (31.9% versus 29.8%, P < .001) compared with mothers in the prelegislation period.
Impact of the Newborns' and Mothers' Health Protection Act Legislation on Clinical and Financial Outcomes
Length of Stay.
Average maternal and neonatal length of stay was significantly longer after implementation of the Newborns' and Mothers' Health Protection Act legislation (Table 2). The increase in length of stay is unmistakable after implementation of the Newborns' and Mothers' Health Protection Act, with no prior ramp-up in the prelegislation period (Figure 1). In addition, the proportion of mothers and newborns with stays of at least 48 hours increased substantially (Table 2).
Satisfaction With Length of Stay.
The percentage of mothers who reported their satisfaction with length of stay as “about right” increased dramatically, from 67.5% in the prelegislation period to 90.9% in the postlegislation period. This increase was statistically significant by both univariate and multivariable analyses (Tables 2 and 3). Compared with respondents, nonrespondents and those not contacted were less likely (P < .05) to have private insurance and more likely to have public, managed Medicaid, or self-pay insurance. Nonrespondents and those not contacted did not differ significantly from respondents with respect to any other sociodemographic, health delivery, or clinical system characteristics (results not shown).
Hospital Readmission and Emergency Room Utilization.
Maternal hospital readmission rates and rates of emergency room use did not change significantly in the postlegislation period compared with the prelegislation period, at either 7 or 30 days after discharge. This was found using both univariate and multivariable analysis (Tables 2 and 3).
Neonatal hospitalization readmission rates did not change significantly in the postlegislation period compared with the prelegislation period, at either 7 or 30 days after discharge in univariate analysis, or at 30 days after discharge in multivariable analysis (Tables 2 and 3). However, neonatal readmission rates were significantly lower at 7 days after index hospitalization in the postlegislation period compared with the prelegislation period when multiple logistic regression was used (Table 3). Neonatal emergency room utilization did not change significantly in the postlegislation period compared with the prelegislation period, at either 7 or 30 days after discharge, in both univariate and multivariable analysis.
Further logistic analysis found that parity was the strongest predictor of neonatal hospital readmission 7 days after index hospitalization. We calculated additional ORs stratified by parity, adjusting for maternal age, insurance status, hospital type, maternal severity of illness, gestational age, neonatal birth weight, and neonatal severity of illness. Newborns of primiparous mothers in the postlegislation period were less likely to have a hospital readmission (OR 0.48, 95% confidence interval [CI] 0.24, 0.93) within 7 days of index hospitalization, compared with newborns of primiparous mothers in the prelegislation period (Table 3). However, newborns of multiparous mothers in the postlegislation period were neither more nor less likely to have a hospital readmission within 7 days of index hospitalization, compared with newborns of multiparous mothers in the prelegislation period.
Adjusted mean hospitalization costs (mother and infant combined) were about US$116 higher in the postlegislation period compared with the prelegislation period, an increase of nearly 6% (Table 4).
In our study population, we found that length of stay (maternal and neonatal) and maternal satisfaction with length of stay increased substantially after implementation of the Newborns' and Mothers' Health Protection Act legislation. This increase in maternal satisfaction came at a cost increase of about US$116 per delivery in the postlegislation period, compared with the prelegislation period, an increase of over $1 million in hospitalization costs, or 7% of total hospitalization costs for the entire 18-month postlegislation time period.
Despite an adjusted increase in cost of just under 6% per delivery as a result of the Newborns' and Mothers' Health Protection Act legislation, the only clinical benefit that we identified was a significant (OR 0.61, 95% CI 0.40, 0.95) decrease in neonatal hospital readmission rates within 7 days of index hospitalization. This difference was largely driven by parity status, with infants of primiparous mothers (approximately 30% of our population) showing the greatest benefit. Such a difference is noteworthy, given the small proportion of newborns readmitted within 7 days. This finding appears to be consistent with that of Liu and colleagues,23 who found that newborns of primiparous mothers with stays shorter than 30 hours were more likely to be rehospitalized within 7 days. Thus, it follows that a longer initial hospital stay might benefit newborns of primiparous mothers through a reduced risk of rehospitalization.
The neonatal finding may be due to changes in post-partum care after implementation of the Newborns' and Mothers' Health Protection Act legislation. Readmission criteria for newborns might have become more stringent after implementation of the Newborns' and Mothers' Health Protection Act legislation, resulting in the lower readmission rate for newborns at 7 days. In addition, given the benefit of home health care visits for newborns,34–36 it is plausible that increased home health care utilization after implementation of the Newborns' and Mothers' Health Protection Act legislation might have resulted in lower neonatal hospital readmission rates. Unfortunately, no data were available in this study to properly assess the impact of home health care utilization on the neonatal clinical outcomes of interest.
Although cost-effectiveness analysis was beyond the scope of the study, an argument could be made that decreased neonatal readmissions produced cost savings that offset the higher hospitalization costs due to mandated length of stay increases. However, we do not believe such a cost offset occurred, for the following reasons. First, because neonatal rehospitalization was a relatively infrequent occurrence, decreases in rehospitalization, though statistically significant, did not occur frequently enough to offset increased index hospitalization costs. Second, clinical reasons for readmission, a large determinant of hospital costs, remained consistent across the prelegislation and postlegislation periods (Table 5). Third, although newborns might have utilized less hospital care in the postlegislation period, other types of service utilization (not all of which are measurable) might have increased, further reducing potential cost savings gained through reduced hospitalization.
Our study was limited by the fact that we did not have sufficient statistical power to detect differences in maternal hospital readmissions. Based on the small proportion of maternal hospital readmissions within 7 and 30 days, we would need to study over 100,000 deliveries to detect a 25% change (assuming 80% power) in readmission rates. Even more cases would be needed to detect smaller differences. Similarly, although we did have power to detect a 25% difference in 30-day maternal emergency room utilization rates, we did not have sufficient power to detect smaller differences.
Although we had fairly extensive data on women and their infants, our study has several additional limitations. First, because mothers with missing satisfaction data (either nonrespondent or no contact) were more likely to have self-pay, managed Medicaid, or public insurance, we might have overestimated patient satisfaction with length of stay, because these groups are often less satisfied with care than women with private insurance.37,38 However, additional analyses found that maternal satisfaction with length of stay increased significantly among each insurance category in the postlegislation period compared with the prelegislation period. Moreover, because non-privately insured mothers with missing satisfaction data did not differ from non-privately insured respondents with respect to any other characteristics, it is likely that we did not overestimate patient satisfaction with length of stay. Second, because our patient population is predominantly white, multiparous, and healthy, results cannot necessarily be generalized to different patient populations. Third, because our data were primarily administrative, we were unable to collect data on patient characteristics, such as maternal concerns, maternal competence, and other quality-of-life measures that might have influenced outcome measures of interest.
Notwithstanding these limitations, the study provides important information on the effects of the Newborns' and Mothers' Health Protection Act legislation on several outcomes for a population of healthy mothers and infants. The study suggests that mandated length of stay increases provided a more enjoyable experience for our population of white, healthy, multiparous women, and a significant benefit for newborns of first-time mothers, while also increasing hospitalization costs significantly.
In this not-for-profit health system, increased health care costs from postpartum length of stay increases might have had an adverse effect on the provision of other services with proven medical benefits (such as prenatal care for indigent populations). The reduction of any services because of increased costs associated with mandated length of stay increases raises serious concerns, particularly in the absence of any pre-Newborns' and Mothers' Health Protection Act data suggesting such legislation would improve outcomes. Moreover, the study raises concerns as to whether optimal postpartum length of stay should be based on legislation rather than empirical research.
More research is needed to fully understand the effects of the Newborns' and Mothers' Health Protection Act legislation on maternal and neonatal populations. Such research should be prospective in design, measure an extensive level of maternal, neonatal, and cost-effectiveness outcomes, and include several delivery systems and demographic groups. Such research will better inform policy makers to improve outcomes of care for maternal and neonatal populations.
1. Trends in length of stay for hospital deliveries—United States 1970–1992. MMWR Morb Mortal Wkly Rep 1995;44:335–7.
2. Lee K, Perlman M, Ballantyne M, Elliot I, To T. Association between duration of hospital stay and readmission rate. J Pediatr 1995;127:758–66.
3. MacDonald MG. Hidden risks: Early discharge and bilirubin toxicity due to glucose 6-phosphate dehydrogenase deficiency. Pediatrics 1995;96:734–8.
4. Maisels MJ, Newman TB. Kernicterus in otherwise healthy, breast-fed term newborns. Pediatrics 1995;96:730–3.
5. Seidman DS, Stevenson DK, Ergaz Z, Gale R. Hospital readmission due to neonatal hyperbilirubinemia. Pediatrics 1995;96:727–29.
6. Coody D, Yetman RJ, Montgomery D, van Eys J. Early hospital discharge and the timing of newborn metabolic screening. Clin Pediatr 1993;32:463–6.
7. Sinai LN, Kim SC, Casey R, Pinto-Martin JA. Phenylketonuria screening: Effect of early newborn discharge. Pediatrics 1995;96:605–8.
8. Newborns' and Mothers' Health Protection Act of 1996, 42 USC 1185 (Pub L No. 104–204).
9. Mandl KD, Brennan TA, Wise PH, Tronick EZ, Homer CJ. Maternal and infant health: Effects of moderate reductions in postpartum length of stay. Arch Pediatr Adolesc Med 1997;151:915–21.
10. Carty EM, Bradley CF. A randomized, controlled evaluation of early postpartum hospital discharge. Birth 1990;17:199–204.
11. Brooten D, Roncoli M, Finkler S, Cohen A, Mennuti M. A randomized trial of early hospital discharge and home follow-up of women having cesarean birth. Obstet Gynecol 1994;84:832–8.
12. Lemmer CM. Early discharge: Outcomes of primiparas and their infants. Neonatal Nurs 1987;16:230–6.
13. James ML, Hudson CN, Gebski, VJ, Browne LH, Andrews GR, Crisp S, et al. An evaluation of planned early postnatal transfer home with nursing support. Med J Aust 1987;147:434–8.
14. Norr KF, Nacion KW, Abramson R. Early discharge with home follow-up: Impact on low-income mothers and infants. J Obstet Gynecol Neontal Nurs 1989;18:133–41.
15. Pittard WB, Geddes KM. Newborn hospitalization: A closer look. J Pediatr 1988;112:257–61.
16. Fox MH, Kanarek N. The effects of newborn early discharge on hospital readmissions. Am J Med Qual 1995;10:206–12.
17. Beebe SA, Britton JR, Britton HL, Fan P, Jepson B. Neonatal mortality and length of newborn hospital stay. Pediatrics 1996; 98:231–5.
18. Grupp-Phelan J, Taylor JA, Liu LL, Davis RL. Early newborn hospital discharge and readmission for mild and severe jaundice. Arch Pediatr Adolesc Med 1999;53:1283–8.
19. Thompson JF, Roberts CL, Currie MH, Ellwood DA. Early discharge and postnatal depression: A prospective cohort. Med J Aust 2000:172:532–6.
20. Kotagul UR, Atherton HD, Eshett R, Schoettker PJ, Perlstein PH. Safety of early discharge for Medicaid newborns. JAMA 1999;282:1150–6.
21. Malkin JD, Garber S, Broder MS, Keeler E. Infant mortality and early postpartum discharge. Obstet Gynecol 2000; 96:183–8.
22. Lock M, Ray JG. Higher neonatal morbidity after routine early discharge: Are we sending newborns home too early. CMAJ 1999;161:249–53.
23. Liu LL, Clemens CJ, Shay DK, Davis RL, Novack AH. The safety of newborn early discharge. The Washington State experience. JAMA 1997;278:293–8.
24. Grullon KE, Grimes DA. The safety of early postpartum discharge: A review and critique. Obstet Gynecol 1997;90:860–5.
25. Gagnon AJ, Edgar L, Kramer MS, Papageorgiou A, Waghorn K, Klein MC. A randomized trial of a program of early postpartum discharge with nurse visitation. Am J Obstet Gynecol 1997;176:205–11.
26. Patterson PK. A comparison of postpartum early and traditional discharge groups. Q Rev Bull 1987;13:365–71.
27. Gjerdingen DK, Titus SL, Tuggle MN. Patients' responses to short obstetrical hospital stays during a nurses' strike. Fam Prac Res J 1988;7:227–35.
28. Classen DC, Burke JP, Pestotnik SL, Lloyd JF. Clinical and financial impact of intravenous erythromycin therapy in hospitalized patients. Ann Pharmacother 1999;33:1371–2.
29. Classen DC, Pestotnik SL, Evans RS, Lloyd JF, Burke JP. Adverse drug events in hospitalized patients. Excess length of stay, extra costs, and attributable mortality. JAMA 1997;277:301–6.
30. Burke JP, Pestotnik SL, Classen DC, Lloyd JF. Evaluation of the financial impact of ketorolac tromethamine therapy in hospitalized patients. Clin Ther 1996;18:197–211.
31. Klingner JM, Solberg LI, Knudson-Schumacher S, Carlson RR. How satisfied are mothers with 1-day hospital stays for routine delivery? Eff Clin Pract 1999;2:253–7.
32. Woodbury D, Tracy D, McKnight E. Does considering faction data? J Healthcare Qual 1998;20:33–40.
33. Andrews FM, Morgan JN, Sonquist JA, Klem LL. Multiple classification analysis: A report on a computer program for multiple regression using categorical predictors. 2nd ed. Ann Arbor: Institute for Social Research, The University of Michigan, 1973.
34. Lieu TA, Bravemen PA, Escobar GJ, Fischer AF, Jensvold NG, Capra AM. A randomized comparison of home and clinic follow-up visits after early postpartum hospital discharge. Pediatrics 2000;105:1058–65.
35. Lieu TA, Wikler C, Capra AM, Martin KE, Escobar GJ, Bravemen PA. Clinical outcomes and maternal perceptions of an updated model of perinatal care. Pediatrics 1998;102:1437–44.
36. Johnson TS, Brennan RA, Flynn-Tymkow CD. A home visit program for breastfeeding education and support. J Obstet Gynecol Neonatal Nurs 1999;28:480–5.
37. Mainous AG, Griffith CH, Love MM. Patient Satisfaction with care in programs for low income individuals. J Community Health 1999;24:381–91.
© 2002 by The American College of Obstetricians and Gynecologists. Published by Wolters Kluwer Health, Inc. All rights reserved.
38. Salganicoff A, Wyn R. Access to care for low-income women: The impact of Medicaid. J Health Care Poor Underserved 1999;10:453–67.