case study

Effects of Health Care Cost-Containment Programs on Patterns of Care and Readmissions Among Children and Adolescents

Thomas M. Wickizer, PhD, MPH, Daniel Lessler, MD, MHA, and Jodie Boyd-Wickizer; BA

Managed care plans and health insur- ance companies have adopted aggressive cost-containment strategies in response to competitive pressures and to the demand by health care payers for improved cost control. Little is known about the effects of these strategies on the delivery of medical care among adult patients, and almost nothing is known about their effects among children and adolescents.

Utilization management (UM) is one of the most widely used approaches to health care cost containment. Current esti- mates suggest that health care for more than 90% of adults enrolled in group insur- ance plans, including health maintenance organizations and preferred provider orga- nizations,'-3 is subject to UM procedures, as is care for dependent children covered under these health plans. UM programs provide external review and authorization for inpa- tient care and for selected outpatient proce- dures. Common UM program activities include preadmission authorization for hospi- talization and concurrent review of the need for continued hospitalization.

The aim of UM is to ensure that treat- ment provided to patients is clinically appro- priate and medically necessary.4'5 Studies documenting high rates of unnecessary and inappropriate inpatient care,6-8 including inpatient pediatric care,9 provided the impetus for the establishment ofUM programs during the 1980s. Studies show that UM reduces utilization and health care costs,10-l5 but understanding of its effects on health care delivery remains limited.

This study analyzed data on a case series of privately insured pediatric patients, ranging in age from birth to 18 years, whose care was subject to UM review and approval. The study had 2 objectives: to examine the effects ofUM on patterns of pediatric care and to determine whether restrictions imposed on length of stay (LOS) by UM review affected the quality of pediatric care, as measured by early readmis-

sion. By analyzing a case series of patients whose care was subject to UM, we obtained information on denials and restrictions result- ing from UM-mandated preadmission autho- rization and concurrent review and were able to examine the effects of these restrictions on readmission. Because we did not have popula- tion-based comparison data on groups sub- jected to UM review and those not so sub- jected, however, we were unable to examine the sentinel effect ofUM on admissions.

Methods

Utilization Management Program

Utilization management was conducted as part of a managed fee-for-service health care plan offered by a large commercial insur- ance carrier from 1989 onward. More than 500 groups, located in 47 states, adopted the insurance carrier's program with UM to pro- mote cost containment. Health care for dependent children covered under the groups' respective policies became subject to the review procedures of the UM program. The groups to which the UM program applied had

Thomas M. Wickizer is with, and at the time of the study Jodie Boyd-Wickizer was with, the Depart- ment of Health Services, School of Public Health and Community Medicine, University of Washing- ton, Seattle. Daniel Lessler is with the Section of General Internal Medicine, Department of Medi- cine, Harborview Medical Center and the Univer- sity ofWashington, Seattle.

Correspondence and requests for reprints should be sent to Thomas M. Wickizer, PhD, MPH, Department of Health Services, Box 357660, Uni- versity of Washington, Seattle, WA 98195-7660 (e-mail:tomwick(u.washington.edu).

This paper was accepted March 4, 1999. Note. The opinions and conclusions expressed

are those of the authors and do not necessarily represent the views of the Robert Wood Johnson Foundation.

American Journal of Public Health 1353

Wickizer et al.

the same basic insurance benefit plan, which covered 80% of allowed charges to some designated (stop loss) level, usually $2500, and 100% of charges thereafter. Most of these groups were subject to an individual deductible of $150 or $200 and a family deductible of $450 or $600. Mental health inpatient coverage was more variable but was usually limited to 21 to 30 days of care for each instance of inpatient treatment. Never- theless, differences in benefit coverage had no direct effect on the action taken as a result of UM review, because the latter was undertaken before and independently ofphysician or hos- pital reimbursement, which was done accord- ing to a claims-examination process.

Utilization management review was compulsory for all patients. Two primary pro- cedures were used to review and approve care: (1) preadmission authorization, which included an outpatient review for selected diagnostic tests and surgical procedures, authorization for admission, and approval of a specified number ofdays for a patient's initial hospital stay, and (2) concurrent review, which examined requests for the patient's continued hospital stay beyond the initial treatment cycle. All requests for admission or continued hospital stay, regardless of LOS, were subject to UM review and approval. Information needed to conduct the reviews was provided by telephone and written com- munication. The UM reviews were con- ducted by a large, well-known UM firm and were done by trained nurse reviewers and physician advisers. Diagnosis-based criteria were used by the UM program to evaluate the appropriateness ofcare; regional LOS profile data were used to assign the number of days ofcare approved.

Data and Measures

From 1989 through 1993, 8568 UM reviews were conducted on patients ranging in age from birth to 18 years. We defined the UM review as the unit of observation for our study and selected all 8568 reviews for analysis. Information routinely gathered by the UM program provided the data for our analysis. This included information on out- comes of preadmission review, which were categorized as follows: (1) request for care denied, (2) outpatient care approved as a sub- stitute for requested inpatient care, or (3) inpatient care authorized. We used data on the number of days requested and approved by concurrent review for continued hospital stay (beyond the initial stay approved at the time of admission) to construct a measure of restriction of LOS, defined as the difference between the number of continued-stay days requested and approved. The total number

of days approved by the UM program and the actual LOS were identical in almost all cases (>99.5%). Among the 6300 cases reviewed for inpatient care, only 3 patients stayed in the hospital for fewer days than approved, while 14 stayed for more days than approved.

The UM data used in our study included information on the patient's sex, age, geo- graphic region, and primary admitting diag- nosis. We used the diagnostic information for a particular patient to classify cases for descriptive analyses and, along with the other variables, to provide covariates for multivariate analyses. In addition to our UM data, we obtained data from the National Hospital Discharge Surveys of 1991 to 1993 to provide comparative LOS benchmarks for our analysis.

To examine the effect on quality of care of limiting utilization through UM review, we defined a patient's first hospital admis- sion as the index admission and then con- structed a biary variable that measured the occurrence of readmission within 60 days after discharge ofthe patient from this initial hospitalization.

Statistical Procedures

We used descriptive statistics to docu- ment denials of admission and LOS restric- tions imposed by the UM program. We relied on multiple regression analysis to determine whether UM became more restrictive both in approval ofthe initial treatment cycle through preadmission review and in authorization of continued hospitalization through concurrent review. We did this by first selecting the most common diagnoses subjected to preadmis- sion and concurrent review and then deter- mining for each diagnosis the change in the number of days approved by the UM pro- gram that occurred between 1989-1991 (T1) and 1992-1993 (T2), while controlling for age, sex, and geographic region.

We used logistic regression analysis to determine whether restrictions on LOS imposed by concurrent review affected the risk of readmission. Although the relation- ship between readmission and quality of care is not fully understood, early readmission is considered an important indicator of quality of care.1619 We limited our analysis to med- ical and mental health admissions because these admission categories had sufficient numbers of readmissions and adequate vari- ability in LOS restrictions to permit analysis. Covariates entered in the logistic regression equations included sex, age, geographic region, total days requested for the treatment episode, total number of reviews performed over the study period, and diagnosis. Diag-

noses were grouped and coded in the form of dummy variables. For medical cases, the diagnostic categories used were dehydra- tion, respiratory tract infections, pneumo- nia, asthma, gastroenteritis or colitis, low birthweight, respiratory distress, and other diagnoses not included in the previous cate- gories. For the mental health analysis, the following broad diagnostic categories were used: depression, alcohol or drug dependence, and other mental health diagnoses.

Length-of-stay restriction, defined as the difference between number of days requested and approved for continued hospital stay, was measured as a continuous variable. We report the adjusted odds ratios and 95% confidence intervals associated with LOS restriction for medical (n = 949) and mental health (n = 510) cases in which there was a request for continued hospital stay through concurrent review.

Results

Descriptive information on the 8568 UM cases subjected to UM review and exam- ined in our study is presented in Table 1. The cases were well distributed among the 3 age groups shown. Medical and surgical admis- sions accounted for the great majority of the cases. Of the 8568 reviews performed, 6300 (73.5%) represented requests for inpatient treatment and 2268 (26.5%) were for outpa- tient treatment. Although not shown, the most common diagnosis for infants was low birthweight, which accounted for 10.1% of all cases, followed by insertion of ear tubes (8.4%), jaundice (7.1%), and pneumonia (6.0%). Among children aged 2 to 12 years, the most common diagnoses were tonsillec- tomy or adenoidectomy (25.3%), insertion of ear tubes (6.5%), and asthma (5.4%). The most common conditions among adolescents were depression (10.4%), tonsillectomy or adenoidectomy (6.1%), and routine delivery (5.1%).

Preadmission Review

Nearly all (99.1%) ofthe 6300 patients requesting authorization for inpatient treat- ment were approved for admission. Of the 6300 cases reviewed, authorization for admis- sion was denied in 5 cases and outpatient treatment was required to replace inpatient treatment in 51 cases. The UM program was somewhat more aggressive in denying outpa- tient treatment than in denying hospitaliza- tion. Authorization for outpatient care was denied in 85 (3.7%) of the 2268 outpatient cases. Ofthese 85 cases, 78% were tonsillec- tomies and 6% were ear tube insertions.

September 1999, Vol. 89, No. 91354 American Journal of Public Health

Health Care Cost Containment

Concurrent Review

Approximately 37% (2356/6300) ofthe patients reviewed for inpatient treatment required additional care beyond the initial treatment cycle and had 1 or more concur- rent reviews performed for continued stay. In 8% ofcases such additional care was denied, and in 38% of cases it was approved for fewer days than were requested. Restrictions imposed on LOS by the concurrent review process resulted in a reduction of requested continued inpatient treatment by 7636 days, or 3.2 days per patient. This average figure (3.2 days) was heavily influenced, however, by substantial restrictions imposed on mental health admissions (Table 2). Mental health admissions, which represented 27% of all admissions, accounted for 80% (6118/7636) of the aggregate reduction in requested days; medical admissions (n = 1400) accounted for 18% (1383/7636) of the reduction. The aver- age per-patient reductions in requested inpa- tient care were 1.0, 0.4, and 9.8 days, respec- tively, for medical, surgical, and mental health admissions.

Although not shown in Table 2, restric- tions imposed by the UM program affected different age groups differently. The greatest reduction in requested hospital days among medical admitted patients (646/1383 days) was for low-birthweight infants (n = 180). Adolescent patients with depression (Inter-

national Classification ofDiseases, Ninth Revision, codes 296.2 or 296.3) accounted for 2301 ofthe 6118 days ofrestricted mental health care shown in Table 2; adolescent patients with alcohol or drug dependence diagnoses accounted for an additional 1037 days ofrestricted treatment.

Changes in the Restrictiveness of Utilization Management Over Time

Table 3 shows changes in the number of days approved by the UM program through preadmission and concurrent review for selected common diagnoses. For most of the diagnoses analyzed, the number of days approved by the UM program declined sig- nificantly. For example, the average patient with asthma had 2.65 days of inpatient treat- ment approved through preadmission review during the baseline period (1989-1991). Dur- ing the comparison period (1992-1993), this same patient had 0.33 (12.5%) fewer days approved (P = .01). The relative magnitude of decline in the number of days approved for continued stay through concurrent review was greater for mental health than for med- ical patients, with the decline for depression averaging 6.7 days from a baseline of 15.6 days (P=.001). Exogenous changes in LOS within the general pediatric patient population do not explain the changes shown in Table 3. National Hospital Discharge Survey data

indicate that the average LOS for male patients under 15 years ofage and discharged between 1990 and 1994 increased from 4.8 to 4.9 days; the LOS for females decreased slightly, from 4.9 to 4.7 days.20

Effects of Utilization Management on Readmissions

To investigate the effects of UM on readmissions, we analyzed medical and mental health cases in which there was a request for continued hospital stay through the concurrent review process. Of 3151 cases representing index (initial) medical admissions, continued hospitalization was requested through concurrent review in 959 (30%). In 79 (8.2%) of these 959 cases of medical admission, the patient was readmitted within 60 days after discharge. Of the 561 cases representing index mental health admissions, continued hospitalization was requested in 510 (9 1%), and the patients in 30 (5.9%) of these were readmitted within 60 days after discharge.

Pediatric patients with a medical or mental health diagnosis for whom concurrent review reduced LOS were more likely to be readmitted within 60 days. For each day by which concurrent review restricted LOS, the odds of readmission among medical cases increased by 5.8% (odds ratio [OR] = 1.058; 95% confidence interval [CI] = 1.006, 1.113) and among mental health cases increased by 6.9% (OR= 1.069; 95% CI= 1.027, 1.114). Because concurrent review imposed only a small reduction in LOS for the average patient with a medical diagnosis (Table 2), the increased relative risk of readmission for such a patient would translate into only a small increase in absolute risk of readmis- sion. The reduction in LOS for the average mental health patient was larger, at just under 10 days (Table 2), which implies an increased relative risk of readmission of approximately 63%. The baseline readmission rate for the mental health cases analyzed, however, was only 6%, and thus the increase in absolute risk even for these patients would be modest.

Readmissions began occurring shortly after discharge. Approximately half of all readmissions among medical patients whose stay was restricted occurred within 14 days after discharge. Among mental health patients, 45% ofreadmissions occurred within 21 days after discharge. The great majority of the patients were readmitted for the same diagnosis that led to their initial admission. Among 12 patients with depression who were readmitted, 9 were readmitted with the same diagnosis, while the remaining 3 were radmitted with a diagnosis ofdrug or alcohol dependency. Among the 18 patients admit-

American Journal of Public Health 1355

TABLE 1-Characteristics of Pediatric Patients for Whom Authorization for Medical Care Was Requested Through a Utilization Management Program (n = 8568)

Characteristic No. of Review Cases (%)

Age Infants (birth-2 years) 2448 (28.6) Children (3-12 years) 3214 (37.5) Adolescents (1 3-18 years) 2906 (33.9)

Sex Female 3956 (46.2) Male 4612 (53.8)

Diagnostic category Obstetric 171 (2.0) Medical 4246 (49.6) Surgical 3445 (40.2) Mental health 706 (8.2)

Requested treatment setting Inpatient 6300 (73.5) Outpatient 2268 (26.5)

Region Northeast 825 (9.6) South 2773 (32.4) West 1262 (14.7) Midwest 3708 (43.3)

Year of review 1989 188 (2.2) 1990 963 (11.2) 1991 2119 (24.7) 1992 2682 (31.3) 1993 2616 (30.5)

September 1999, Vol. 89, No. 9

Wickizer et al.

ted with an initial diagnosis ofpneumonia or asthma, 14 (78%) were readmitted with the same diagnosis, 2 were readmitted for surgi- cal procedures involving the middle or inner ear, and the remaining 2 were readmitted for other medical reasons.

Discussion

Health care provided to children and adolescents, like that for adults, has come under intense cost-containment pressure. Although UM has become a key cost-con- tainment strategy of health plans and insur- ance carriers, there is little understanding of its effects on patterns of care. The UM program we analyzed restricted LOS mod- estly in relation to requests by physicians for continued inpatient treatment for med- ical and surgical admissions, by 1.0 days and

0.4 days, respectively. The average restriction in LOS was much greater for mental health admissions, at 9.8 days. On the basis of the reduction in requested hospital days of care, there is little doubt that the UM program pro- moted cost containment, especially in the area of mental health. In restricting LOS, however, the UM program increased the risk ofreadmission for patients receiving medical or mental health care. Nevertheless, data relating to the relationship among LOS, read- mission, and quality of care are contradic- tory. 6,l1-2 Some studies have reported an increased risk of readmission in association with shorter hospital stays among mental health patients,23-25 whereas others have reported no effect.2S28 Further, as noted by Epstein et al.,29 even if a shorter LOS does increase the risk of readmission, earlier dis- charge may reflect greater efficiency rather than lower quality if outcomes such as sur-

vival, functional status, and return to normal activity are unchanged.

Two important questions arise about the validity of our results: To what extent do patients comply with the UM review deci- sion, and do pediatricians routinely inflate their requested LOS? If physicians do rou- tinely inflate their requests for inpatient care, expecting them to be automatically reduced by the UM process, then our data would have less clinical and policy relevance. Similarly, if patients routinely ignore UM program decisions by staying in the hospital for more days than the program approves, our analysis would be less meaningful. As noted earlier, we found that patients almost always (>99%) adhered to the UM program decision about LOS.

To address the question ofwhether pedi- atricians routinely inflate their requested LOS, we examined the median total number

1356 American Joumal of Public Health

TABLE 2-Number of Days Requested and Approved Through Concurrent Review for Pediatric Patients for Continued Inpatient Stay"

Percentage of Days Days Requested Days Sum of Reduced

Admission Typeb Requestedc Approvedc Approvedc Days Reducedc Hospital Days

Medical (n =1400) 8.1 (14.2) 7.1 (13.2) 89.2 (25.1) 1.0 (3.8) 1383

Surgical (n =319) 3.8 (4.9) 3.4 (5.0) 88.2 (27.2) 0.4 (0.99) 135

Mental health (n=627) 25.1 (22.1) 15.3(17.5) 58.6 (34.1) 9.8 (12-5) 6118

All admissions (n = 2346) 12.1 (17.8) 8.8 (14.4) 80.1 (31.2) 3.3 (8.2) 7636

aData shown are means, unless otherwise noted; data in parentheses are standard deviations. bTable omits 10 obstetric cases that were subject to concurrent review. All requested continued-stay days were approved for these 10 cases. cDifferences among the 3 admission types were statistically significant (P< .001).

TABLE 3-Changes in the Number of Inpatient Days Approved for Pediatric Patients by Utilization Management for Selected Diagnoses

Inpatient Days Change in Inpatient Type of Review Approved, Baseline Days Approved, and Diagnosis Period, 1989-1991 1992-1993 95% Cl P

Preadmission review: Initial treatment cycle Appendectomy (n = 142) 2.91 -0.23 -0.52, 0.57 .12 Dehydration (n = 182) 1.83 -0.46 -0.71,-0.22 <.001 Asthma (n = 221) 2.65 -0.33 -0.58, -0.74 .01 Gastroenteritis (n = 173) 2.00 0.10 -0.14, 0.35 .40

Concurrent review: continued stay, days Depression (n = 290) 15.61 -6.71 -10.68,-2.73 .001 Other mental health 21.84 -7.13 -10.65, -3.62 <.001

diagnoses (n = 234)a Pneumonia (n = 1 10) 4.76 -2.00 -4.51, 0.14 .07 Low birthweight (n = 185) 17.90 3.27 -4.32,10.86 .39

Note. Cl = confidence interval. alncludes admissions for adjustment reaction (n = 24), manic depression (n = 23), emotional disturbance specific to childhood (n = 20), and bipolar affective disorder (n = 1 1).

September 1999, Vol. 89, No. 9

Health Care Cost Containment

ofdays ofhospitalization requested by physi- cians for selected diagnoses and procedures in relation to the LOS of privately insured pediatric patients as reported in the National Hospital Discharge Surveys for 1991 to 1993. These diagnoses and procedures included pneumonia, dehydration, asthma, tonsillec- tomy with adenoidectomy, appendectomy, routine delivery, depression, and alcohol or drug dependence. To enhance the compara- bility of the data, we limited selection of cases subjected to UM review and discharge records to groups in the Midwest and South, the two regions in which most of our study population was concentrated.

The median days requested for the diag- noses and procedures examined were reason- ably close to the 50th percentile ofLOS distri- bution calculated from the discharge-survey data. The median days requested for pneu- monia, dehydration, and asthma were 4, 2, and 3 days, respectively. These figures repre- sent the 65th, 45th, and 56th percentiles for each of the respective LOS distributions. The corresponding values for tonsillectomy, appendectomy, and routine delivery were 1, 3, and 2 days, representing the 50th, 48th, and 50th percentiles, respectively. For patients with recurrent depression or with alcohol or drug dependence, the median days requested represented the 64th and 66th percentiles, respectively. These LOS benchmark compar- isons suggest that the pediatricians who treated patients in our study population did not routinely inflate their requests for inpa- tient care to any significant extent.

The limitations of our study should be noted. First, we had limited information with which to control for differences in severity of illness in our logistic regression analyses. Fail- ure to control for severity of illness would, however, introduce bias into the results only if such severity were correlated with both read- mission and LOS restriction. Although sever- ity of illness may be correlated with readmis- sion, one would not expect it to be correlated with limitations on LOS imposed by UM. Key informant interviews conducted by one of us (T.M.W) among hospital nursing staff mem- bers in the Seattle area support this view. The nurses interviewed indicated that it is generally easier to justify the need for continued hospi- talization for patients with more severe condi- tions. Thus, we do not believe that our limited ability to control for differences in severity of illness poses serious problems for our analysis.

Second, our study focused on a single UM program, which may limit the general applicability of our results. A recent study found considerable variability in utiliza- tion review procedures used by UM orga- nizations.30 However, our study population was reasonably large (>8500) and included

patients from all 4 census regions of the United States. Indeed, our focus on a single UM firm may actually have increased the internal validity of our study by reducing confounding effects that might have occurred had we studied multiple UM programs with different review procedures. In all of the cases we studied, the patient's care was sub- jected to the same review procedures, based on a single set of clinical protocols. Never- theless, caution should be used in generaliz- ing the results ofour analysis.

Controlling medical care utilization will remain an important cost-containment objec- tive for both public and private health care payers as well as for health plans and health insurance carriers. Cost-containment pro- grams, such as the UM program we studied, may reduce unnecessary care, thus promot- ing more efficient resource consumption. Such programs may, however, also have unin- tended effects on the delivery of health care. Further analyses will be needed to determine the impact ofUM and other cost-contain- ment programs on the quality and outcomes of care. Only with a more complete under- standing of this effect can judgments be made about the long-term value of current approaches to cost containment. ]

Contributors T. M. Wickizer planned the study, analyzed the data, and wrote the paper. D. Lessler assisted in designing the study and analyzing the data. J. Boyd-Wickizer conducted the literature review, developed the diag- nostic code categories, and assisted in planning the data analysis. Both D. Lessler and J. Boyd-Wickizer contributed to the writing ofthe paper.

Acknowledgments This work was supported by grant 19977 from the Robert Wood Johnson Foundation.

References 1. Gold MR, Hurley R, Lake T, Ensor T, Berenson

R. A national survey of the arrangements man- aged-care plans make with physicians. NEnglJ Med. 1995;333:1678-1683.

2. Sullivan CB, Miller M, Feldman R, Dowd B. Employer-sponsored health insurance in 1991. HealthAff: 1992;11:172-185.

3. Wickizer TM. The effect of utilization review on hospital use and expenditures: a review of the literature and an update on recent findings. Med Care Rev. 1990;47:327-363.

4. Payne SM. Identifying and managing inappro- priate hospital utilization: a policy synthesis. Health Serv Res. 1987;22:709-769.

5. Institute of Medicine. Controlling Costs and Changing Patient Care? The Role of Utilization Management. Washington, DC: National Acad- emyPress; 1989.

6. Restucia JD, Gertman P. A comparative analysis of appropriateness of hospital care. Health Aff 1984;3:130-138.

7. Restucia JD, Kemper BE, Payne SM, Gert- man P, Dayno SJ, Lenhart GM. Factors affect- ing appropriateness of hospital care in Mas- sachusetts. Health Care Financ Rev. 1986;8: 47-56.

8. Siu AL, Sonnenberg A, Manning WG, et al. Inappropriate use of hospitals in a randomized trial of health insurance plans. NEngl J Med. 1986;315:1259-1266.

9. Kemper KJ. Medically inappropriate hospital use in a pediatric population. N Engl J Med. 1988;318: 1033-1037.

10. Wickizer TM, Wheeler JR, Feldstein PJ. Does utilization review reduce unnecessary care and contain costs? Med Care. 1989;27: 632-646.

11. Feldstein PJ, Wickizer TM, Wheeler JR. Private cost containment: the effects of uti- lization review programs on health care use and expenditures. N Engl J Med. 1988;318: 1310-1315.

12. Wickizer TM. The effects of utilization review on hospital use and expenditures: a covariance analysis. Health Serv Res. 1992; 27:103-121.

13. Khandker RM, Manning WG, Ahmed T. Uti- lization review savings at the micro level. Med Care. 1992;30:1043-1052.

14. Scheffler RM, Sullivan SD, Ko TH. The impact of Blue Cross and Blue Shield plan utilization management. Inquiry. 1991;28:263-275.

15. Wickizer TM. Controlling outpatient medical equipment costs through utilization manage- ment. Med Care. 1995;33:383-391.

16. Thomas JW, Holloway JW. Investigating early readmission as an indicator for quality of care studies. Med Care. 1991;29:377-394.

17. Brook RH, Lohr KN. Monitoring quality of care in the Medicare program. JAMA. 1987; 258:3138-3141.

18. DesHarnais SL, McMahon F, Wroblewski R. Measuring outcomes of hospital care using multiple risk-adjusted indexes. Health Serv Res. 1991;26:425-445.

19. Ashton CM, Del Junco D, Souchek J, Wray NP, Mansyur CL. The association between the quality of inpatient care and early readmission: a meta-analysis of the evidence. Med Care. 1997;35:1044-1057.

20. National Center for Health Statistics. Health, United States, 1998. Washington, DC: US Dept of Health and Human Services; 1998:308-309.

21. Caton CL, Gralnick A. A review of issues surrounding length of psychiatric hospitaliza- tion. Hosp Community Psychiatry. 1987;38: 858-863.

22. Ashton CM, Kuykendall DH, Johnson ML, et al. The association between the quality of inpa- tient care and early readmission. Ann Intern Med. 1995;122:415-421.

23. Appleby L, Desai PN, Luchins DJ, et al. Length of stay and recidivism in schizophrenia: a study of public psychiatric hospital patients. Am J Psychiaty. 1993;150:72-76.

24. Altman H, Sletten IW, Nebel ME. Length-of- stay and readmission rates in Missouri state hospitals. Hosp Community Psychiatry. 1973; 24:773-776.

September 1999, Vol. 89, No. 9 American Journal of Public Health 1357

Wickizer et al.

25. De Francisco D, Anderson D, Pantano R, Kline F. The relationship between length of hospital stay and rapid-readmission rates. Hosp Community Psychiatry. 1980;31:196-197.

26. Edward-Chandran T, Malcolm DE, Bowen RC. Reduction of length of stay in an acute care psychiatric unit. Can J Psychiatry. 1996;41: 49-51.

27. Holloway JJ, Medendorp SV, Bromberg J. Risk factors for early readmission among veterans. Health Serv Res. 1990;25:214-237.

28. Thomas MR, Rosenberg SA, Giese AA, Fryer GE, Dubovsky SL, Shore JH. Shortening length of stay without increasing recidivism on a university-affiliated inpatient unit. Psychiatr Serv. 1996;47:996-998.

29. Epstein AM, Bogen J, Dreyer P, Thorpe KE. Trends in length of stay and rates of readmis- sion in Massachusetts: implications for moni- toring quality of care. Inquiry. 1991;28:19-28.

30. Schlesinger MJ, Gray BH, Perreira KM. Med- ical professionalism under managed care: the pros and cons of utilization review. Health Aff 1997;16:106-123.

Edited by Daniel B. Herman and Ezra S. Susser

This volume contains some of the most influential research that has been conducted to date on the problem of homeless- ness and specific conditions that are associated with homelessness. The articles share the methodological rigor, and the rel- evance to public health policy and practice that have long been hallmarks of the American Journal of Public Health.

The collection is organized into three sections: * "The Problem of Homelessness: Description, Scope, and Underlying Causes," includes several editorials that address underlying causes and issue calls to action * "Individual-Level Risk Factors for Homelessness," includes several of the best designed studies conducted to date that illuminate the role played by individual-level factors in vulnerability to homelessness * "Health Conditions among Homeless People," addresses both a broad range of health conditions that dispropor- tionately affect homeless people as well as issues of utilization and access to health care services

1998 * 150 pp * Stock No. 808/HOAD98 $30 Non-APHA Members * $25 APHA Members

U American Public Health Association * Publications Sales P.O. Box 753, Waldorf, MD 20604-0753 Tel: 301/893-1894; Fax: 301/843-0159; Web: www.APHA.org; E-mail: [email protected]

1358 American Journal of Public Health September 1999, Vol. 89, No. 9