Special Article
Do the Poor Cost More? A Multihospital Study of Patients' Socioeconomic Status and Use of Hospital Resources
N Engl J Med 1990; 322:1122-1128April 19, 1990
- Abstract
- Abstract
There is controversy about whether hospitalized poor patients use more resources and whether hospitals that provide care for the poor therefore merit supplementary payment under per-case prospective payment systems. We previously reported that patients of low socioeconomic status with connective-tissue disease had longer hospital stays and higher costs than patients of higher socioeconomic status at a single hospital. To examine the generalizability of this phenomenon, we interviewed 16,908 (83 percent) of 20,278 consecutive adult patients (excluding obstetrical and psychiatric patients) admitted in 1987 to five Massachusetts hospitals, to obtain information on three direct measures of socioeconomic status (income, occupation, and education). We divided each measure into three strata. Thus, there were 15 comparisons — three measures of socioeconomic status applied to each of five hospitals.
After excluding outliers and adjusting for diagnosis-related group (DRG), we found that the patients of the lowest socioeconomic status had hospital stays 3 to 30 percent longer than those of patients of higher status, the differences varying with the hospital and the indicator of socioeconomic status (P≤0.05 for 11 of the 15 comparisons). Hospital charges were 1 to 18 percent higher for the patients of lowest socioeconomic status than for those of higher status (P≤0.05 for 9 of 15 comparisons). When we adjusted for age, severity of illness, and DRG, the patients of lowest socioeconomic status had longer stays than those of higher status in 14 of 15 comparisons (P<0.05 for 7 of the 15) and higher charges in 13 of 15 comparisons (P<0.05 for 6 of the 15). The differences between patients of high and low status ranged up to 21 percent for length of stay and 13 percent for charges.
Our findings suggest that hospitalized patients of lower socioeconomic status have longer stays and probably require more resources. Supplementary payments to hospitals for the treatment of poor patients merit further consideration. (N Engl J Med 1990; 322:1122–8.)
Media in This Article
Table 1
Response Rate and Sociodemographic Characteristics of Patients at Five Hospitals.Table 2
Percent Differences in Length of Stay and Total Charges at Five Hospitals among Patients of Lower as Compared with Higher Socioeconomic Status, in Relation to Primary Indicators of Status, with Adjustment for DRG.*Article Activity
- Article
THERE is evidence that the adoption of a prospective payment system based on diagnosis-related groups (DRGs) has enabled Medicare to restrain the growth in its payments to hospitals. In addition to the federal government, third-party payers in nearly a dozen states have adopted prospective per-case payment systems patterned after the DRG system. A number of other states are considering similar policies.1
Despite the apparent success of per-case payment in controlling hospital expenditures, there has been persistent concern about the ability and continued willingness of hospitals to provide care for patients who require more resources than the average, especially those who are exceptionally ill or of lower socioeconomic status.2 3 4 If the costs of care are higher for these patients than for others within the same DRG, hospitals with higher proportions of these patients will receive inadequate payment, and there will be an incentive to restrict access to care.
Currently, Medicare provides supplementary payment to hospitals that care for a disproportionate share of the poor, but this policy is controversial.2 Other per-case prospective payment systems in individual states have generally not provided supplementary payment.1 The advocates of supplementary payment believe that patients of lower socioeconomic status require longer stays and use more resources than patients of higher status in the same DRG. Factors such as greater severity of illness on admission, poor nutritional status that retards healing, and less social support after discharge may contribute to this increased need for resources. Alternatively, one could argue that poor patients have lower costs. Difficulty with ensuring adequate ambulatory outpatient follow-up for patients of lower socioeconomic status may encourage doctors to lower the threshold for hospital admission. If so, the average severity of illness of hospitalized poor patients might be less than that of wealthier persons in the same DRG. At the same time, persons from higher socioeconomic groups who benefit from the support services of hospitals as they recuperate may be more able to persuade their doctors to prolong their hospitalizations.
Thus far, a number of studies have tried to determine whether patients of low socioeconomic status stay in the hospital longer and incur higher costs than other patients. Virtually all these studies used retrospective data obtained from hospital discharge systems rather than direct measures of socioeconomic status.5 6 7 8 Their results have been inconsistent.5 6 7 8 9
In 1985, we performed a retrospective study in which we interviewed 402 patients with connective-tissue disease at a single hospital to obtain information on direct measures of their socioeconomic status.9 After controlling for DRG, we found that patients with lower incomes, less prestigious occupations, and less education stayed in the hospital approximately 10 to 20 percent longer than members of other subgroups. We were unable, however, to assess the overall importance of this phenomenon for other clinical conditions or its generalizability to other hospitals.
In the present study, we prospectively interviewed 16,908 consecutive adult patients admitted to five hospitals in 1987, excluding those admitted to the obstetrical or psychiatric services. Our results indicate that at these hospitals, care for patients of lower socioeconomic status was generally associated with longer stays and higher hospital charges than care for those of higher socioeconomic status.
Methods
Patients and Sources of Data
We studied five hospitals in eastern Massachusetts that varied in size, setting (location), and the presence of teaching programs. These included two major urban teaching hospitals in Boston (Beth Israel Hospital and Brigham and Women's Hospital), two suburban nonteaching hospitals near Boston that share financial management but have largely separate groups of admitting physicians and different patient populations (Choate Hospital and Symmes Hospital), and one community hospital (Cape Cod Hospital) that serves a small city, Hyannis, and that has minimal involvement with teaching provided by house staff based primarily at an affiliated teaching center.
At each hospital we collected data on all patients 18 years of age or older, excluding obstetrical and psychiatric patients, who were admitted during the first six months of 1987. The starting and ending dates for patient enrollment varied slightly at the different institutions. Because of resource limitations, we were unable to enroll all the patients admitted to one of the hospitals (Beth Israel Hospital) and instead enrolled a proportional sample chosen on the basis of sequential dates of admission. We used structured interviews to obtain information on the patients' socioeconomic characteristics. Information on age, sex, primary and secondary diagnoses, DRG, expected payer, length of stay, and total charges was obtained from the information systems of the hospital.
Interview and Survey Procedures
The items in our questionnaire included the most recent occupation and the type of work performed, the highest grade or year completed in school, the annual household income in 1986 (less than $5,000, $5,000 to $9,999, $10,000 to $14,999, $15,000 to $24,999, $25,000 to $49,999, and $50,000 or more), and home-ownership status (whether the patients or a member of their households owned or rented their living quarters).
The interviews were performed by regular personnel in the admitting office and project staff who were trained to administer the questionnaire. When a patient's medical condition or another logistical problem made it impossible to conduct the interview at the time of admission, the patient was interviewed on the hospital ward after admission or by telephone after discharge. When feasible, we interviewed proxies (close friends or relatives) for the patients who did not speak English, who were judged by the nursing staff to be too ill, or who died before they were contacted (1867 patients, 11 percent of the population interviewed). If a patient was admitted more than once during the course of the study, each admission was counted as a separate observation. If we were unable to interview the patient on readmission, we used the socioeconomic data obtained at the first admission to assess his or her occupation, education, and 1986 income.
Socioeconomic Status
In our primary analyses, we examined the relation of the length of hospital stay and hospital charges to three characteristics that we thought most directly indicated a patient's socioeconomic status: occupational prestige, education, and household income. We focused on these characteristics individually because there is no aggregate measure of socioeconomic status that is universally accepted. To compare the use of resources by different subgroups of patients, we divided the continuous variables (such as years of education) into commonly accepted categories of economic opportunity and social standing (high, intermediate, and low). These categories were comparable to the measures used previously in our work and in other studies.9 10 11 12
We defined three subgroups on the basis of the patients' occupational prestige: skilled white-collar, professional, and managerial workers; craftspersons and clerical workers; and manual and service workers. The patients' educational levels were divided into three categories on the basis of the highest grade the patient had completed: 12th grade or higher, 8th through 11th grade, and 7th grade or below. Income was categorized into three groups: $25,000 and over, $10,000 to $24,999, and less than $10,000.
In addition to these primary indicators of socioeconomic status, we examined other characteristics of the patients that might be related to their socioeconomic status or affect the patterns of medical care they received, including home ownership and health insurance status.
Severity of Illness
To gauge the severity of disease, we used DRGSCALE (SysteMetrics),13 , 14 a computerized system developed recently as a modification of disease staging15 for use with abstract data obtained at discharge. The developers of the system used multiple-regression techniques to assess the relation between patients' coexisting conditions and their use of resources within DRGs, using average hospital costs for the United States. In the creation of DRGSCALE, empirical analyses were conducted of a data base containing information on nearly 7 million hospital discharges from different regions of the United States. The scale assigns scores for relative severity of illness to patients with different constellations of coexisting conditions, standardized within individual DRGs. A score of 100, for example, represents the national average cost for the DRG; a score of 150 indicates a patient whose predicted cost exceeds the national average by half.
DRGSCALE has been validated by its developers in analyses of data from 50,000 randomly selected patients in the state of Washington.16 They found that using this system with DRGs increased the amount of the variation in cost between patients that could be explained from 19 percent with DRGs alone to 28 percent and that groupings based on the scale were monotonie with respect to cost.
Data Analysis
The basic policy question we addressed was: Do patients with lower socioeconomic status require more resources than patients with higher socioeconomic status, after adjustment for DRG? We defined resources in terms of length of stay and total charges. Because the distributions of length of stay and total charges were skewed, we used a natural log transformation. To assess the patterns observed in the different hospital settings better, we examined each setting individually. Although the use of multiple separate analyses may have given our assessment of statistical significance a conservative bias, it allowed us to ensure that the findings at the larger hospitals did not dominate our results. We used charges as a proxy for cost. Because we conducted our analyses separately at each hospital and adjusted for DRG, the bias in the results from the use of charges rather than costs is likely to be limited.
In our initial analyses, we used multiple linear regression to adjust for individual DRGs in the comparisons of lengths of stay and hospital charges within a single hospital for groups of patients of differing socioeconomic status.17 Log transformations of the lengths of stay or the total charges were the dependent variables in these analyses. The independent variables included DRG and socioeconomic status. We used dummy variables to represent each DRG and two dummy variables to represent subgroups of patients of low and intermediate socioeconomic status. After appropriate calculation, this statistical technique allowed us to use coefficient estimates from the linear regression to estimate the percentage difference between the socioeconomic subgroups with respect to the dependent variable of interest (such as length of stay), with the group of patients of high socioeconomic status as a reference. The statistical significance of the overall differences between groups was assessed with an F test. We examined each index of socioeconomic status separately at each hospital.
The existing DRG system used nationally, as well as a number of the per-case payment systems used by the states, have separate payment mechanisms for outliers — that is, patients whose lengths of stay or charges fall at the extremes of the distribution. In accordance with this approach, patients with very long hospital stays were excluded from our initial analyses. We defined outliers according to the Medicare standards for length of stay used in 1987.18 This allowed us to focus on the health policy implications of our findings.
To examine whether the differences in length of stay and charges that were associated with patients' socioeconomic status could be explained by severity of illness, we performed additional analyses in which we controlled for age and DRGSCALE score as well as for individual DRGs. We also repeated our analyses of two populations: the entire group of patients interviewed, including those who were outliers with respect to length of stay, and the subgroup of patients insured under Medicare. Because of the relatively small number of Medicare patients at some of the hospitals, we combined the Medicare population of all five hospitals for this analysis and examined only length of stay.
Finally, we assessed the generalizability of our findings by examining whether the pattern of differences was present in many individual DRGs or just a few. At each hospital we identified all DRGs with 10 or more patients (excluding outliers) and at least 2 patients from the lowest socioeconomic strata. For each of these DRGs separately, we examined whether the mean stay for the patients of lower socioeconomic status was longer than for the patients of higher status. We then used a chi-square test to determine whether this effect occurred in more than half the DRGs in each hospital. Finally, to identify the particular clinical conditions for which the effects of socioeconomic status were most marked, we examined individually all DRGs with at least 100 patients, using household income as our measure of socioeconomic status.
Results
Response Rate and Patient Population
At the five hospitals, we obtained interview data from 16,908 of 20,278 eligible patients or their proxies (83 percent) (Table 1Table 1
Response Rate and Sociodemographic Characteristics of Patients at Five Hospitals. ). Of the patients interviewed, 98 percent (16,610) responded to the question about education, 96 percent (16,228) to the questions about occupation, and 81 percent (13,625) to the question about income. Twelve percent (1955 patients) were excluded from the analyses of occupation — 1942 (mostly homemakers and students) who reported never having had a regular occupation outside the home and 13 whose job titles and activities we could not classify reliably. Of the 3370 patients not interviewed, 765 (23 percent) refused to be interviewed, 402 (12 percent) died before an interview could be conducted, 272 (8 percent) could not be interviewed because they did not speak English, 903 (27 percent) had a diminution of mental status that made the interview unreliable, and 1028 (30 percent) were not contacted.Table 1 shows basic demographic and socioeconomic information about this population. The distribution of the patients' age, sex, and socioeconomic status in terms of occupation, education, and income was significantly different at the different hospitals (P≤0.05 for all comparisons). For example, the patients at Cape Cod Hospital and Symmes Hospital were much older. Choate Hospital and Symmes Hospital had a higher percentage of patients with lower educational attainment and lower occupational prestige; Cape Cod Hospital had a higher percentage of patients with incomes under $25,000.
Overall Use of Services
Among the patients interviewed, the mean hospital stay (adjusted for DRG) at the different institutions varied by less than 9 percent — from 7.3 days at Cape Cod Hospital to 8.0 days at Symmes Hospital. Of the total group, 569 patients (3.4 percent) were excluded from the primary analysis because their length of stay exceeded the Medicare cutoff for outliers. When they were excluded, the mean stay at the different institutions decreased by about 20 percent. Except as otherwise indicated, the outliers have been excluded from the analyses presented here.
Association of Status with Length of Stay and Total Charges
Differences in length of stay and total charges among the patients categorized according to the three primary indicators of socioeconomic status are shown in Table 2Table 2
Percent Differences in Length of Stay and Total Charges at Five Hospitals among Patients of Lower as Compared with Higher Socioeconomic Status, in Relation to Primary Indicators of Status, with Adjustment for DRG.*, after adjustment for DRG. Although the magnitude of the differences in length of stay among the socioeconomic subgroups varied according to socioeconomic indicator and hospital, the overall pattern of the data suggests that the patients of lower socioeconomic status had longer stays. At Brigham and Women's Hospital and Beth Israel Hospital, the differences in length of stay between the highest and lowest socioeconomic subgroups ranged from 9 to 30 percent and were consistently statistically significant. At Cape Cod Hospital and Symmes Hospital, patients of lower socioeconomic status also had significantly longer stays when grouped according to two of our three indexes, income and education. The size of the differences ranged from 4 to 24 percent. At Choate Hospital, length of stay was significantly longer for persons of lower socioeconomic status only when it was assessed according to income. The pattern of total charges also suggested a higher use of resources for the care of persons of lower socioeconomic status. Overall, the differences among the socioeconomic subgroups (involving comparisons of three indicators at five hospitals) were statistically significant (P≤0.05) for 11 of 15 comparisons for length of stay and 9 of 15 comparisons for charges. At the three largest hospitals, seven of nine comparisons showed significant (P≤0.05) differences with respect to both length of stay and charges.Differences in length of stay and total charges among the subgroups of patients as defined according to other socioeconomic characteristics and DRGSCALE scores are presented in Table 3Table 3
Percent Differences in Length of Stay and Total Charges at Five Hospitals in Relation to Indicators of Socioeconomic Status and Disease Severity, with Adjustment for DRG.*, after adjustment for DRG. Persons who rented their living quarters tended to have longer hospital stays than those who owned them, but the results were statistically significant only at Beth Israel Hospital and Symmes Hospital. The Medicaid patients had longer stays than those who had other insurance; the results were statistically significant at three of the five hospitals.In examining the effect of DRGSCALE scores, our proxy for disease severity, we formed three groups on the basis of cutoff points 1 SD above and below the mean. As compared with patients with lower severity of illness as indicated by DRGSCALE, the most severely ill persons had lengths of stay and total charges that were generally more than 50 percent higher (Table 3). Greater severity of illness (as indicated by higher DRGSCALE scores) also correlated with lower socioeconomic status (occupation, r = −0.04; education, r = −0.06; income, r = −0.11; P<0.01 by Pearson's correlation coefficient for all comparisons).
Potential Confounders
To control for potential confounders, we examined length of stay and total charges in relation to the patients' socioeconomic status, after adjustment for age and DRGSCALE score as well as DRG. The same trends toward longer stays and higher charges among those of lower socioeconomic status persisted at the three larger hospitals. The magnitude of the differences among subgroups, however, generally decreased by about one third (Table 4Table 4
Percent Differences in Length of Stay and Total Charges at Five Hospitals in Relation to Primary Indicators of Socioeconomic Status, with Adjustment for Age, DRGSCALE Score (Severity), and DRG.*). The differences remained statistically significant at Beth Israel Hospital for all three indicators of socioeconomic status, at Brigham and Women's Hospital for occupation and income, and at Cape Cod Hospital for income.Analysis Based on Individual DRGs
To determine how often lower socioeconomic status was associated with longer hospital stays within individual DRGs, we examined patterns of care separately for the patients in each DRG that included 10 or more persons (Table 5Table 5
Proportion of DRGs in Which Patients from the Lowest Socioeconomic Subgroups Had Longer Hospital Stays Than Other Patients.*). Overall, the patients in the lowest socioeconomic group stayed in the hospital longer than the other patients for approximately two thirds of the DRGs. For example, at Beth Israel Hospital the low-income patients stayed longer than the other patients in 48 of the 62 DRGs (77 percent) in which there were at least 10 patients.To identify the clinical conditions for which the effects of socioeconomic status were most pronounced, we examined individually each of the 35 DRGs in which there were at least 100 patients, using income as our measure of socioeconomic status. Poor patients stayed longer than those with higher incomes in 31 of these DRGs (in 5 of the 31 the difference was statistically significant). The stays of poor patients were at least 25 percent longer than those of other patients in 9 of these 31 DRGs. These nine DRGs included a diverse mix of medical conditions and surgical procedures: DRG 112, vascular procedures except major reconstruction without pump; 125, circulatory disorders excluding acute myocardial infarction with cardiac catheterization without complex diagnosis; 138, cardiac arrhythmia and conduction disorders, with age ≥70, complications or comorbidity, or both; 140, angina pectoris; 174, gastrointestinal hemorrhage, with age ≥70, complications or comorbidity, or both; 197, total cholecystectomy without common duct exploration, with age ≥70, complications or comorbidity, or both; 215, back and neck procedures, age <70, without complications or comorbidity; 243, medical back problems; and 468, unrelated procedures in the operating room. Low-income patients had shorter stays in four DRGs (DRG 162, inguinal and femoral hernia procedures, age 18 to 69, without complications or comorbidity; 209, major procedures involving joint and limb reattachment; 277, cellulitis with age ≥70, complications or comorbidity, or both; and 395, red-cell disorders, age ≥18). For these four DRGs, the differences between groups were under 10 percent, except in DRG 395 (18 percent).
Analysis of Medicare Patients
When we repeated our analyses but restricted them to the subgroup insured by Medicare, we again found trends suggesting longer stays for those of lower socioeconomic status. After adjustment for DRG, the differences in length of stay between the high and low socioeconomic subgroups were 16 percent when socioeconomic status was assessed according to income, 11 percent when assessed according to education (P<0.001 for both comparisons), and 3 percent when assessed according to occupation (P >0.05).
Analysis Including Outliers
Overall, 3.4 percent of the patients interviewed were outliers. We found that patients in each of the lowest socioeconomic strata were more likely to be outliers than those in the highest strata (income, 5.0 vs. 2.4 percent; occupation, 4.2 vs. 2.8 percent; and education, 5.2 vs. 3.0 percent; P<0.001 in all cases). When we included these patients in the analysis and adjusted for DRG, differences in the use of resources between patients in the high and low socioeconomic subgroups became even larger. Depending on the index examined, the differences in length of stay at Beth Israel Hospital ranged up to 34 percent, at Brigham and Women's Hospital 19 percent, at Cape Cod Hospital 26 percent, at Choate Hospital 32 percent, and at Symmes Hospital 27 percent.
Nonrespondents
The 17 percent (3370) of the population that we were unable to interview tended to be older (mean [±SD]age, 61.7±21.2vs. 56.3±19.2 years; P<0.001) and, as assessed according to DRGSCALE score, sicker (mean score [±SD], 115.0±51.6 vs. 106.9±44.9; P<0.001). Overall, a larger percentage of these patients also had Medicaid insurance (16 vs. 8 percent), Medicare (45 vs. 36 percent), or were uninsured (8 vs. 4 percent); a smaller percentage had coverage with Blue Cross (13 vs. 23 percent), HMOs (8 vs. 13 percent), or other commercial insurers (7 vs. 12 percent) (P<0.01 in all cases). When the outliers were excluded, the mean stay (adjusted for DRG) for the nonrespondents was longer at Beth Israel Hospital by 17 percent, at Symmes Hospital by 8 percent, and at Brigham and Women's Hospital by 8 percent; it was shorter at Choate Hospital by 14 percent, and of equal length at Cape Cod Hospital. Overall, these results do not suggest any consistent bias introduced by nonresponse.
Discussion
We interviewed nearly 17,000 patients in five eastern Massachusetts hospitals to determine their socioeconomic characteristics and found that those with the lowest levels of income, occupational prestige, and education had hospital stays 5 to 25 percent longer than those with the highest levels. The findings were similar and generally statistically significant at four of the five hospitals but somewhat less consistent at the fifth institution (Choate Hospital). With regard to total charges, the magnitude of the differences was smaller, but the trends were similar to those for length of stay. When we analyzed the patterns of care within individual DRGs, we found that in about two thirds of the DRGs patients in the lowest socioeconomic group had longer hospital stays than other patients. Taken together, these findings suggest that generally, care for patients of lower socioeconomic status entails longer hospital stays and probably requires more resources than care for patients of higher socioeconomic status.
From a public policy perspective, the extent to which greater severity of illness on admission among patients with lower socioeconomic status can explain the differences we observed may be important. If the relation between socioeconomic status and the use of resources is due solely to greater severity of illness, then adopting an adjustment for severity to existing DRGs or other systems of per-case payment might correct any inequities in payment to hospitals that care for the poor. DRGSCALE, a recently developed measure of disease severity, was especially appropriate for our study, since it was designed to assess severity in terms of need for resources (i.e., hospital costs), rather than outcomes such as mortality. It could be assessed from abstract data obtained at discharge without an additional review of charts and could be used in conjunction with DRGs. Assessed in this way, severity of illness was significantly related to length of stay and the cost of care. In our study population, the patients of lower socioeconomic status did tend to be more severely ill. When we adjusted for age and DRGSCALE score as well as DRG, the magnitude of the observed differences among socioeconomic strata decreased substantially. However, statistically significant differences persisted at three of the hospitals. Although it is possible that adjustment using a better measure of severity would have eliminated all differences, we do not know of a better alternative to DRGSCALE for the assessment of clinical data based on discharge abstracts.19 , 20
One interesting issue is that of the interplay between insurance coverage, socioeconomic status, and hospital costs. Patients who are uninsured tend to have lower socioeconomic status.21 , 22 Yet in a previous study, two of us showed that uninsured hospitalized patients also tend to have shorter hospital stays.23 Both trends were evident in our data. When we controlled for insurance status (insured vs. uninsured) as well as DRG, we found that the differences in length of stay between high and low socioeconomic groups became slightly larger.
The controversy over whether to adjust the rates of hospital payment to reflect the care provided to the poor dates back to the adoption of the Medicare system. The original federal legislation contained no provision for such supplementary payment. However, the secretary of health and human services was given authority to adjust payments, as appropriate, to hospitals that cared for a disproportionate share of the poor.8 Initially, the secretary determined that no special adjustments were warranted, but in 1986 Congress enacted legislation to provide supplementary payment. Payments have continued since then, although there is still substantial debate over their appropriateness and their magnitude. The current formula for payment has a threshold; supplementary payment is provided only to hospitals that care for large numbers of the poor. Many hospitals that care for a smaller but still substantial number of poor patients receive no extra payment. For example, of the hospitals in our study, only Brigham and Women's Hospital and Beth Israel Hospital received supplementary payment during the year of the study. Our data suggest that in order to avoid disincentives for caring for the poor, it may be more appropriate to provide supplementary payment to all hospitals, based at least in part on the proportion of patients served who are of lower socioeconomic status.
One aim of this study was to evaluate the generalizability of the relation between socioeconomic status and use of resources to a variety of hospitals. Although we cannot be certain that the patterns we observed would be evident in states other than Massachusetts, we chose the hospitals in our study to represent a range of institutions with respect to teaching status and location. In our primary analysis, when we adjusted only for DRG, the trends we found were fairly consistent across institutions. When we adjusted for age and severity as well as DRG, the trends persisted clearly at three of the five hospitals. Differences in the patterns of resource use may reflect differences in the populations of patients served or the patterns of care provided at different hospitals. For example, low-income patients cared for at an inner-city teaching hospital may be different from those cared for at a suburban nonteaching hospital. Patterns of length of stay and charges are also likely to vary among hospitals in relation to common clinical practice and discharge policies, the availability of extended care facilities and home support for poor patients, and the hospital's efficiency in arranging for social services and continuing care for such persons after discharge.
In conclusion, our findings indicate that for many institutions, caring for patients of lower socioeconomic status probably entails longer hospital stays and greater use of resources. The magnitude of the differences, generally on the order of 5 to 25 percent, has far-reaching and obvious implications for hospitals that care for substantial numbers of poor persons. At the managerial and clinical levels, our findings suggest the need to identify the conditions for which the relation between socioeconomic status and increased use of resources is strongest, so that strategies can be developed for the more efficient provision of care for these patients. This is clearly a long-term approach, since it is likely to involve interventions, such as increasing the availability of services after discharge, that are beyond the control of an individual hospital. On a policy level, our findings have more immediate implications for the states that use per-case prospective payment systems, and nationally for Medicare. At least until better measures of severity are developed, they support the provision and possibly the refinement of supplementary payment for patients of lower socioeconomic status, the hospitals treating them, or both.
Supported by a grant from the Massachusetts Fund for Cooperative Innovation. Dr. Epstein is a Henry J. Kaiser Family Foundation Faculty Scholar in General Internal Medicine.
We are indebted to H. Richard Nesson, M.D., and James L. Wiczai, Jr., of Brigham and Women's Hospital; Mal Weiner and Michael H. Littman of Beth Israel Hospital; Dorcas McGurrin and Robert S. Murray of Cape Cod Hospital; and Paul L. Downey and Sandra Saleeba of Choate and Symmes Hospitals, for their support and efforts; to Amy Cohen and Stephen Fielding, Ph.D., for help with the computer analysis; to David Blumenthal, M.D., Stephen Jencks, M.D., and Joseph Newhouse, Ph.D., for helpful comments on an earlier draft of this manuscript; and especially to Christine Collins for her help in developing this study.
Source Information
From the Department of Medicine, Division of General Medicine, Section in Health Services and Policy Research, Brigham and Women's Hospital (A.M.E.); the Department of Dermatology, Beth Israel Hospital, and Harvard Medical School (R.S.S.); the Department of Health Care Policy, Harvard Medical School, and the Department of Health Policy and Management, Harvard School of Public Health (A.M.E.,J.S.W.); all in Boston. Address reprint requests to Dr. Epstein at the Department of Health Care Policy, Parcel B, 1st Floor, 25 Shattuck St., Boston, MA 02115.
- References
References
1
Hellinger
Web of Science | Medline2
Pear 3
Horn
Full Text | Web of Science | Medline4
Stern
Full Text | Web of Science | Medline5
Department of Health and Human Services, Health Care Financing Administration 6
Department of Health and Human Services, Health Care Financing Administration 7
Jencks
CrossRef | Web of Science | Medline8
Prospective Payment Assessment Commission. Technical appendixes to the report and recommendations to the Secretary, U.S. Department of Health and Human Services, April 1, 1985. Washington, D.C.: Government Printing Office, 1985.
9
Epstein
Full Text | Web of Science | Medline10
Epstein
CrossRef | Web of Science | Medline11
Davis
CrossRef | Web of Science12
Rushing
CrossRef | Web of Science | Medline13
Conklin JE, Lieberman JV, Hernberg JD. DRG refinement: a feasibility assessment using disease staging, age, and unrelated co-morbidity. New York: SysteMetrics, 1984. (HCFA grant no. 100–82–0038.)
14
Conklin JE, Louis DZ, Houchens RL. DRG refinement using measures of disease severity: final report. New York: SysteMetrics, 1987. (HCFA grant no. 18-C-98761/9–01S1.)
15
Gonnella
CrossRef | Web of Science | Medline16
Christoffersson
Medline17
Kleinbaum DG, Kupper LL. Applied regression analysis and other multivariable methods. North Scituate, Mass.: Duxbury Press, 1978:209–21.
18
Department of Health and Human Services, Health Care Financing Administration 19
Iezonni LI. Measuring the severity of illness and case-mix. In: Goldfield N, Nash DB, eds. Providing quality care: the challenge to clinicians. Philadelphia: American College of Physicians, 1989:70–105.
20
Jencks SF. DRG refinement: outliers, severity of illness, and intensity of care. Baltimore: Department of Health and Human Services, Health Care Financing Administration, Office of Research and Demonstrations, 1987: 15. (HCFA publication no. 03254.)
21
Wilensky G. Underwriting the uninsured: targeting providers or individuals. In: Sloan FA, Blumstein JF, Perrin JM, eds. Uncompensated hospital care: rights and responsibilities. Baltimore: Johns Hopkins University Press, 1986:148–66.
22
Davis
CrossRef23
Weissman
CrossRef | Web of Science | Medline
- Citing Articles (34)
Citing Articles
1
Dorte Gyrd-Hansen, Kim R. Olsen, Torben H. Sørensen. (2011) Socio-demographic patient profiles and hospital efficiency: Does patient mix affect a hospital's ability to perform?. Health Policy
CrossRef2
Richard Cookson, Mauro Laudicella. (2011) Do the poor cost much more? The relationship between small area income deprivation and length of stay for elective hip replacement in the English NHS from 2001 to 2008. Social Science & Medicine 72:2, 173-184
CrossRef3
Mauro Laudicella, Kim Rose Olsen, Andrew Street. (2010) Examining cost variation across hospital departments–a two-stage multi-level approach using patient-level data. Social Science & Medicine 71:10, 1872-1881
CrossRef4
Bryan Lipmann. (2009) Elderly Homeless Men and Women: Aged Care's Forgotten People. Australian Social Work 62:2, 272-286
CrossRef5
H. Jasti, E. M. Mortensen, D. S. Obrosky, W. N. Kapoor, M. J. Fine. (2008) Causes and Risk Factors for Rehospitalization of Patients Hospitalized with Community-Acquired Pneumonia. Clinical Infectious Diseases 46:4, 550-556
CrossRef6
Julian Perelman, Amir Shmueli, Marie-Christine Closon. (2008) Deriving a risk-adjustment formula for hospital financing: Integrating the impact of socio-economic status on length of stay. Social Science & Medicine 66:1, 88-98
CrossRef7
Susan M. Frayne, Victoria A. Parker, Cindy L. Christiansen, Susan Loveland, Margaret R. Seaver, Lewis E. Kazis, Katherine M. Skinner. (2006) Health Status Among 28,000 Women Veterans. The VA Women's Health Program Evaluation Project. Journal of General Internal Medicine 21:S3, S40-S46
CrossRef8
Magali Pirson, Michèle Dramaix, Pol Leclercq, Terri Jackson. (2006) Analysis of cost outliers within APR-DRGs in a Belgian general hospital: Two complementary approaches. Health Policy 76:1, 13-25
CrossRef9
Magali Pirson, Dimitri Martins, Terri Jackson, Michèle Dramaix, Pol Leclercq. (2006) Prospective casemix-based funding, analysis and financial impact of cost outliers in all-patient refined diagnosis related groups in three Belgian general hospitals. The European Journal of Health Economics 7:1, 55-65
CrossRef10
Adrienne Green, Jonathan Showstack, Deborah Rennie, Lee Goldman. (2005) The Relationship of Insurance Status, Hospital Ownership, and Teaching Status with Interhospital Transfers in California in 2000. Academic Medicine 80:8, 774-779
CrossRef11
S. J. Moug, E. Robertson, W. J. Angerson, P. G. Horgan. (2005) Socioeconomic deprivation has an adverse effect on outcome after ileostomy closure. British Journal of Surgery 92:3, 376-377
CrossRef12
A. L. Diamant, R. D. Hays, L. S. Morales, W. Ford, D. Calmes, S. Asch, N. Duan, E. Fielder, S. Kim, J. Fielding, G. Sumner, M. F. Shapiro, D. Hayes-Bautista, L. Gelberg. (2004) Delays and Unmet Need for Health Care Among Adult Primary Care Patients in a Restructured Urban Public Health System. American Journal of Public Health 94:5, 783-789
CrossRef13
Marie-Christine Closon, Julian Perelman. (2003) L'impact des facteurs sociaux sur les dépenses de santé : comment éviter que la recherche de l'efficience ne se fasse au détriment de l'équité et de l'accessibilité aux soins ?. Reflets et perspectives de la vie économique XLII:1, 47
CrossRef14
HE Frech, LR Mobley. (2000) Efficiency, growth, and concentration: an empirical analysis of hospital markets. Economic Inquiry 38:3, 369-384
CrossRef15
Diane Nykamp, Dale Ruggles. (2000) Impact of an Indigent Care Program on Use of Resources: Experiences at One Hospital. Pharmacotherapy 20:2, 217-220
CrossRef16
Lavoisier Cardozo, Susan Aherns. (1999) Assessing the Efficacy of a Clinical Pathway in the Management of Older Patients Hospitalized with Congestive Heart Failure. Journal for Healthcare Quality 21:3, 12-17
CrossRef17
C. J. Roe, N. Dodich, E. Kulinskaya, W. R. Adam. (1998) Comorbidities and prediction of length of hospital stay. Australian and New Zealand Journal of Medicine 28:6, 811-815
CrossRef18
David G Whiteis. (1998) Third World medicine in First World cities: capital accumulation, uneven development and public health. Social Science & Medicine 47:6, 795-808
CrossRef19
Laura B. Shepardson, Amy C. Justice, Dwain L. Harper, Gary E. Rosenthal. (1998) Associations Between the Use of Do-Not-Resuscitate Orders and Length of Stay in Patients With Stroke. Medical Care 36:SUPPLEMENT, AS57-AS67
CrossRef20
D M Carlisle, B D Leake. (1998) Differences in the effect of patients' socioeconomic status on the use of invasive cardiovascular procedures across health insurance categories.. American Journal of Public Health 88:7, 1089-1092
CrossRef21
Salit, Sharon A., Kuhn, Evelyn M., Hartz, Arthur J., Vu, Jade M., Mosso, Andrew L., . (1998) Hospitalization Costs Associated with Homelessness in New York City. New England Journal of Medicine 338:24, 1734-1740
Full Text22
Michael D. Murray, Paul Stang, William M. Tierney. (1997) Health care use by inner-city patients with asthma. Journal of Clinical Epidemiology 50:2, 167-174
CrossRef23
David G. Whiteis. (1997) Unhealthy Cities: Corporate Medicine, Community Economic Underdevelopment, and Public Health. International Journal of Health Services 27:2, 227-242
CrossRef24
J S Haas, S Berman, A B Goldberg, L W Lee, E F Cook. (1996) Prenatal hospitalization and compliance with guidelines for prenatal care.. American Journal of Public Health 86:6, 815-819
CrossRef25
S Martin. (1996) Explaining variations in inpatient length of stay in the National Health Service. Journal of Health Economics 15:3, 279-304
CrossRef26
HOWARD S. GORDON, GARY E. ROSENTHAL. (1996) Impact of Interhospital Transfers on Outcomes in an Academic Medical Center. Medical Care 34:4, 295-309
CrossRef27
Robyn McDermott. (1995) IMPROVING EQUITY AND EFFICIENCY IN THE BUSH: A NEEDS-BASED METHOD FOR HEALTHCARE RESOURCE ALLOCATION IN REMOTE COMMUNITIES. Australian Journal of Rural Health 3:2, 72-79
CrossRef28
Edward Guadagnoli, Paul D. Cleary, Barbara J. Mcneil. (1995) The influence of socioeconomic status on change in health status after hospitalization. Social Science & Medicine 40:10, 1399-1406
CrossRef29
Jennifer S. Haas, Paul D. Cleary, Edward Guadagnoli, Christopher Fanta, Arnold M. Epstein. (1994) The impact of socioeconomic status on the intensity of ambulatory treatment and health outcomes after hospital discharge for adults with asthma. Journal of General Internal Medicine 9:3, 121-126
CrossRef30
Narendra Shah, Giovanni Vatieri. (1992) The Implications of Urban Social and Economic Realities on Strategic Management of Hospitals. Healthcare Management Forum 5:4, 4-11
CrossRef31
Narendra Shah, Giovanni Vatieri. (1992) Conséquences des réalités socio-économiques urbaines sur la gestion stratégique des hôpitaux. Healthcare Management Forum 5:4, 12-20
CrossRef32
I. Steven Udvarhelyi, Lee Goldman, Anthony L. Komaroff, Thomas H. Lee. (1992) Determinants of resource utilization for patients admitted for evaluation of acute chest pain. Journal of General Internal Medicine 7:1, 1-10
CrossRef33
(1991) Do the Poor Cost More?. New England Journal of Medicine 324:1, 63-64
Full Text34
Mary Ann Adler Cohen, Asher Aladjem, Amy Horton, Jose Lima, Antonio Palacios, Ivan Hernandez, Jay Lefer, Pankaj Mehta. (1991) How Can We Combat Excess Mortality in Harlem: A One Day Survey of Substance Abuse in Adult General Care. The International Journal of Psychiatry in Medicine 21:4, 369-378
CrossRef
- Letters
