for Long-Term Care Services:

Transcription

1 Hospital DRGs and the Need for Long-Term Care Services: An Empirical Analysis Mark R. Meiners and Rosanna M. Coffey The Medicare DRG-based Prospective Payment System (PPS) encourages hospitals to reduce length ofstayfor elderly patients. Thus, discharges to long-term care services are expected to increase. Maryland hospital data for 198 are used to identify those DRGs which mostfrequently represent patients discharged to nursing home and home health care services; explores the incentive to discharge earlier under PPS those patients needing long-term care versus short-term care; and describes characteristics ofpatients most likely to face increased pressure of earlier discharge to nursing homes and home health programs. Because only a limited set ofpatient characteristics are available from Maryland hospitals, data from a study of San Diego nursing homes are used to explore further the sociodemographic and health status measures associated with unusually long stays in a hospital prior to nursing home placement. This research suggests that the DR G reimbursement system gives hospitals a strong incentivefor earlier discharge ofpatients needing long-term care services. However, hospitals that target only long-term care patients for early discharge will not substantially gain under PPS because these patients represent a small portion of the cases treated in the hospital and a small percentage of unreimbursed days. An earlier version of this article was presented at the 111th Annual Meetings of the American Public Health Association, November 1983, in Dallas, Texas. The views expressed in this article are those of the authors, and no official endorsement by the National Center for Health Services Research or the Department of Health and Human Services is intended or should be inferred. Mark R. Meiners, Ph.D. is Economist and Director, Long-Term Care Program, Division of Intramural Research, at the National Center for Health Services Research. Rosanna M. Coffey, Ph.D. is Economist and Director of the Hospital Studies Program in the Division of Intramural Research, NCHSR. Address correspondence and requests for reprints to Mark R. Meiners, Ph.D., Senior Research Manager, NCHSR-DIR, STOP 3-5, PARK Building, 56 Fishers Lane, Rockville, MD 2857.

2 36 Health Services Research 2:3 (August 1985) INTRODUCTION Placement of hospitalized elderly patients in long-term care facilities is difficult in many areas of the country [ 1]. A hospital 'backup problem" is created when elderly patients occupy hospital beds while waiting for long-term care beds to become available. This problem has been attributed to a variety of causes, induding inadequate discharge planning, lack of alternatives to nursing home care, inadequate reimbursements to nursing facilities for public patients, and insufficient nursing home beds [2-7]. Until recently, hospitals have had little reason to be concerned about this problem because Medicare paid for most "administrative days" in the hospital for patients awaiting long-term care services. With the new Medicare Prospective Payment System (PPS) for hospital services, attitudes of hospital care providers toward the backup of elderly patients in their facilities are likely to change significantly. As of October 1, 1983, Medicare began paying for short-term hospital care with essentially a fixed rate per Medicare discharge, depending on the Diagnosis Related Group (DRG) in which the patient is classified.' In addition to the fixed price by DRG, hospitals are marginally reimbursed for excessively expensive patients once they reach a high-cost or length of stay threshold. These cases are commonly referred to as "outliers," and they generally have lengths of stay that are at least twice the average for their DRG. For days of stay beyond the outlier threshold, 6 percent of a calculated per diem amount for a DRG (based on the DRG average cost and average length of stay) is paid. Only 5 to 6 percent of Medicare patients per year are expected to become outliers (see Note 1). The cost of days less than the outlier for a DRG but beyond the average are borne by the hospital. For patients certified by the attending physician or by utilization review as no longer requiring hospital care, the hospital may bill other parties (i.e., private insurance, Medicaid, or the patient directly) after 48 hours of this certification, for further "administratively necessary days." The two "grace days" may or may not be covered by Medicare depending on when they occur in the hospital stay. This reimbursement system replaces for most short-term general hospitals an earlier cost-based per diem system for patient care services. The new system generally places hospitals at financial risk for patients who stay beyond the average length of stay and/or who exceed the national average cost of care for the DRG in which the patient is classified. It is unclear how many administrative days will become bad

3 DRGs and Long- Term Care Need 361 debts for the hospital. Conversely, the hospital gains from treating patients with fewer resources than the average for that DRG. Gains from low-cost patients may balance losses from high-cost patients within a DRG so that the hospital may break even for a particular DRG category, or the hospital may gain in some DRGs while losing in others. A good administrator or financial officer will identify and locate the hospital's gains and losses and attempt to change operating procedures to improve the hospital's financial position. One of the most visible elements of the cost of a patient's hospital care is the individual's length of stay. The Health Care Financing Administration (HCFA) publishes the national mean length of stay for each DRG (Note 1). Thus, providers can compare individual patients' time in the hospital against national lengths of stay and can tailor patient management strategies to these guidelines. Patients no longer requiring hospital services, such as those awaiting long-term care beds, are probably losing money for the hospital. However, administrative day payments from other parties and the 6 percent daily rate for longstaying patients may cover the marginal cost of additional days, so that once beyond the threshold, there may be less pressure to discharge these patients. Nevertheless, there is an incentive early in the stay to transfer potential backup patients to other facilities. It is generally in the hospital's best financial interest to shorten the length of stay for these and other patients. The incentive for earlier discharge is expected to increase the demand for long-term care services and to stimulate their growth. In the short run, however, the supply of long-term care beds is unlikely to increase sufficiently to handle the added pressure on these services. One exception may be extended-care facilities in short-term hospitals. Long-term care units in short-term general hospitals are exempt from PPS, for the time being, and retain cost-based payments for their services. This may motivate hospitals with serious backup problems to expand or establish their own long-term care units, provided that at a minimum the projected costs of these units are covered by third-party allowable expenses and/or are offset by avoided losses on hospitalized Medicare patients otherwise awaiting long-term care placement. In any case, hospitals may become more actively interested in the development of long-term care services to enable earlier discharge of elderly patients. The increased demand for long-term care services will place added pressure on sources of financing. Medicare and private supplemental benefits will need reevaluation in terms of their flexibility and adequacy to accommodate the increased use of long-term care services anticipated under the DRG-based PPS.

4 362 Health Services Research 2:3 (August 1985) It is important to examine the incentives facing hospitals to decrease lengths of stay of elderly Medicare patients, so that we can begin to anticipate pressures on long-term care services. We know of no empirical research on the relationship between the DRG-based PPS and hospital discharges to nursing homes and home health care. This article (1) identifies those hospital DRGs which most frequently represent patients discharged to long-term care services, (2) explores the strength of the incentive to discharge earlier under PPS those patients needing long-term care versus short-term care, and (3) describes characteristics of those patients most likely to face increased pressure of earlier discharge to nursing homes and home health programs. These analyses are exploratory and necessarily comparative in nature, i.e., between nursing home discharges and self-care discharges, since data are not yet available to describe pre- and post-pps changes in discharges to long-term care services. METHODS AND DATA Patient-level data primarily from hospital records are used in the study reported here to describe Diagnosis Related Groups (based on the ICD-9-CM [8]) by discharge destination -nursing homes (skilled and intermediate care), home health care programs, and self-care (routine discharge to one's own home). With these data, the types of DRGs that predominate in hospital discharges to nursing homes can be compared with DRGs for other types of discharges. To understand the incentive to early discharge by type of discharge destination, patient distributions are examined by length of stay categories, defined according to cutoff criteria for the Medicare Prospective Payment System. These criteria identify three groups: (1) a category of patients with lengths of stay below the geometric mean length of stay, which will likely make money for hospitals; (2) one with lengths of stay below the "outlier" cutoff but above the geometric mean, which will likely lose money for hospitals on those days beyond the geometric mean; and (3) another with lengths of stay greater than the outlier cutoff, which are reimbursed the flat rate plus 6 percent ofpro rata daily expenses for those days beyond the cutoff. The outlier cutoff is either about twice the geometric mean length of stay or 2 days beyond the geometric mean, whichever is smaller. In addition to length of stay criteria, high cost cutoffs are also used by HCFA as a basis for marginal per diem payments. These could not be applied accurately to the data available for the study. It is also noted that HCFA uses the

5 DRGs and Long- Term Care Need 363 geometric mean for length of stay estimates because the distribution of hospital lengths of stay is approximately log-normal. The geometric mean of a log-normal sample is an estimator for the median of the lognormal distribution and is directly related to the mean of the underlying normal distribution. The three length of stay categorizations allow comparisons of the types of patients by discharge destination most likely to gain or lose money for hospitals and thus less or more likely to be subjected to increased pressure for earlier discharge. These are reasonable approximations of the cutoff distinguishing gains from losses. They may differ for individual hospitals that have lower or higher expenses per case than the average hospital. Also, these estimates relate to national payments only and do not reflect hospitalspecific amounts which Medicare is paying in the interim until full national PPS becomes effective in Finally, these estimates do not include the offsetting effect of the "administrative day" rule by which hospitals can recover some of their losses by billing other payers. Two sources of patient-level data are used in this study: one to examine incentives to earlier discharge under PPS and the other to study patient characteristics in more detail. Hospital discharge abstract records for the year 198 for all short-term general, non-federal Maryland hospitals are aggregated by discharge destination to describe the most frequent DRGs and the proportions of patients comprising length of stay categories. Some patients' characteristics can also be examined with these Maryland data. Only Medicare patients age 65 and over, who were discharged to nursing homes, home health care, or self-care, were used in the analysis. Ninety percent of the 13,635 Maryland patients used in the study were self-care discharges; 8 percent were discharged to nursing homes; and nearly 2 percent were discharged to home health care. (National estimates for hospital discharges to longterm care services are below this level, although precise national estimates are not available separately for long-term and short-term discharges in 198 [9]). Patients discharged to short-term general hospitals and institutions other than nursing homes and those who left against medical advice are excluded from the analysis; they represent only 2.7 percent of the live discharges in Maryland. For the second source of data, nursing home records of patients admitted from short-stay hospitals are analyzed by hospital length of stay categories within detailed patient social and medical characteristics. The nursing home data come from 18 randomly chosen for-profit San Diego County, California skilled nursing homes, which are control facilities in an incentive payment experiment for nursing homes conducted by NCHSR [7, 1]. In San Diego County, there were about 7

6 364 Health Services Research 2:3 (August 1985) nursing homes in total and 41 proprietary nursing homes during the study period, November 198-April Only MediCal patients age 65 and over, most of whom were also eligible for Medicare, are used in the analysis. These two data sources are chosen for different but complementary reasons. In the case of Maryland, detailed discharge destination is collected, so that transfers to other short-term facilities can be separated from long-term facilities (skilled and intermediate care nursing homes) and from home health care programs. This important distinction is necessary for the following analysis and frequently is not made in discharge abstract records. (For example, a national sample of about 4 hospitals in the Hospital Cost and Utilization Project (HCUP) of NCHSR combined data for from 12 abstracting companies, but abstract forms generally did not distinguish nursing home discharges from transfers to other short-term hospitals.) Apart from containing diagnostic and basic demographic information, the Maryland abstracts are sketchy. For this reason, data from the study of San Diego nursing home patients admitted from hospitals are used to explore patient living arrangements and levels of functioning as they are associated with lengths of stay in the hospital prior to nursing home placement. The generalization of these data is limited strictly to their localities, but similar frequencies of patients by DRGs were found for both Maryland hospitals and HCUP sample hospitals across the nation. Further, two studies (one on Maryland data and one with the national HCUP data), which incorporated the three length of stay distinctions, gave similar results by type of hospital [11, 12]. Nevertheless, it is known that length of stay in the East is longer than in other regions of the country [13]; this is accounted for, to the extent possible, in the following analyses. Because of regional length of stay differences, criteria for the three length of stay categories described below are based on DRG-specific geometric means and cutoffs calculated from the 198 Maryland data. If the published national means and cutoffs were compared against the Maryland length of stay distributions, the proportions of patients estimated to lose money would be accurate for Maryland but overestimated nationally. To approximate the national perspective and reduce this bias, we compare Maryland patient length of stay to averages and cutoffs calculated from Maryland data. Maryland is unique in another respect, since in 198, a per-case payment system was in effect in 2 of 5 Maryland short-term general, non-federal hospitals. The effect of per-case payment in Maryland

7 DRGs and Long-Term Care Need 365 from 1977 to 1981 on length of stay, among other factors, was analyzed by Steinwachs and Salkever [14]. They found no statistically significant effect of Maryland's per-case system on length of stay. Thus, the length of stay experience in Maryland is probably more strongly related to regional phenomena than to Maryland's rate-setting environment Ṫhe San Diego data are from a region of the country with the shortest average length of stay. Because of this difference, we use national criteria for analyzing length of stay categories in the San Diego data. This approach is more realistic than using Maryland norms for San Diego analyses, but it underestimates the pressures on long-term care services from a national perspective. Another problem exists with the San Diego data. Hospital surgical procedures were not required data items for the San Diego study. This lack of information means that the nursing home analysis must be limited to a subset of DRGs that do not depend on procedures. These DRGs were identified by applying the DRG-grouper program to the Maryland hospital data with and without information on procedures and then comparing the two sets of patient distributions by DRG. DRGs that underwent a change of 1 percent or less in the number of patients in the DRG were assumed to be relatively unaffected by the procedure information and only these were used in the San Diego analysis. These non-procedurerelated DRGs covered 695 patients or 53 percent of the patients otherwise eligible for our San Diego analysis. Another caveat to bear in mind in interpreting the data in this article concerns the availability of long-term care in different geographic locations. Maryland has about 53 nursing home beds per thousand population over age 65, compared to the U.S. average of 57 beds.2 This suggests that the hospital backup problem is likely to be slightly more serious in Maryland than in the nation. Thus, comparisons of proportions of Maryland patients by hospital stay categories may overstate the situation nationally. However, because our study cannot address the extent of the possible hospital response under PPS to discharge patients at an earlier stage of recovery to long-term care providers, 198 Maryland data may still underestimate the total effect of PPS on pressure to discharge patients. San Diego County has about 35 nursing home beds per thousand population over age 65, which is considerably below the national average. San Diego also relies on residential care facilities to supply lower levels of care and, thus, has a more developed market for an alternative to skilled nursing institutions than other parts of the country. Since San Diego is different in this regard, it is more difficult to assess how it relates to the nation. Fur-

8 366 Health Services Research 2:3 (August 1985) thermore, the San Diego data force us to restrict our focus to characteristics of patients and prohibit us from inferring anything about the pressure to earlier discharge with San Diego data. Due to the limitations of the data available for this study, the results should be considered tentative and warranting further corroboration. At this stage in the development of knowledge about the expected effects of DRG-based fixed payments for acute hospital care on long-term care services, the use of data with such weaknesses is justifiable. This research provides a needed first look at this new phenomenon and points to issues which should be explored later on with better data. RESULTS TYPES OF DRGS AND LONG-TERM CARE SERVICES From data on Medicare patients hospitalized in Maryland in 198, it is clear that there are differences in the most common DRGs discharged to nursing homes, home health care, and self-care. Table 1 shows the 1 most frequent DRGs discharged to each of these three destinations, 18 DRGs in all. The 1 most frequent DRGs account for a significant percentage of cases in each discharge category: 4.2 percent of nursing home cases, 36.5 percent of home health care program cases, and 28.5 percent of self-care cases. Some categories such as stroke (DRG 14) and hip procedures (DRG 21) are more representative of nursing home discharges than self-care discharges, but others such as heart failure and shock (DRG 127) are commonly found in all three types of discharges. By examining the rankings in Table 1 and the distributions of patients within specific DRGs across the discharge destinations in Table 2, some general statements can be made. Discharges to nursing homes fall more frequently into diagnostic categories that require skilled rehabilitation services: stroke (DRG 14), hip and femur procedures (DRG 21), and major joint procedures (DRG 29); that reflect mental or behavioral problems: mental disorders (DRG 429); or that specifically reflect frailty of old age: pneumonia with complication (DRG 89) and kidney and urinary tract infections with complication (DRG 32). Discharges to home health care agencies more likely comprise patients in diagnostic categories that may require long-term management but do not necessarily represent debilitating conditions, at least in some stages of the illness: respiratory

9 DRGs and Long-Term Care Need cancer (DRG 82), diabetes (DRG 294), and major bowel procedures (DRG 148). PRESSURE FOR EARLIER DISCHARGE 367 By examining distributions of patients with specific discharge destinations by their length of stay categories relating to Medicare prospective payment rules, as described above, the extent to which certain types of patients are more likely to lose money for the hospital can be assessed. Table 3 examines these distributions for elderly Medicare patients discharged from Maryland hospitals in 198. About 7 percent of discharges to nursing homes and 68 percent of discharges to home health programs stay beyond the average for their DRG, compared to 48 percent of patients routinely discharged. Twenty percent of discharges to nursing homes have excessively long stays (i.e., stays at least twice the average for their DRG) compared to 11 percent of home health care patients and only 5 percent of self-care discharges. Viewed another way, 29 percent of all outlier cases are nursing home or home health program discharges even though they represent only 1 percent of all patients. Thus, hospital patients ultimately receiving long-term care more than proportionately comprise the PPS categories of potential underpayment to hospitals compared to cost-based payments. Table 4 presents the distribution of hospital days for the patients in Table 3. A patient staying beyond the cutoff will contribute days to all three length of stay categories. From these data and Table 3, we estimate that the average patient discharged to long-term care services potentially accounts for 7 days of unreimbursed care (i.e., for nursing care and home health care combined, the number of days in the "greater than the mean but less than the cutoff" group divided by the total number of discharges) compared to 3 days for the average patient discharged to self-care. Whether nursing home and home health program discharges should be targeted for earlier discharge by hospitals depends on whether these differential stays are due to slack in discharge planning, or unavailability of long-term care services and seriousness of the patient's medical condition. The next section looks in more detail at the characteristics of elderly Medicare patients likely to be pressured for earlier discharge. By understanding the characteristics of these patients, they can be identified early in the hospital stay as candidates for careful discharge planning.

10 o c C-4 COC -~ LO C44* C"4 C4Cq ce; PG.) 8~> -- - _ t L C6 -l ECE CC4f o #C C C~4 CW4 4 CC) * LO ~ O t co 4.~.r t f- r VX co Cl C4 P4.~ 4D 4) OU Q Cf o 4) m 4) 4 i S-U CwU X E P- rap o ou C W _UX9 n co O O f ba CM co cn e

11 q 4) O Cf)C~e LO CC4-11 / -, _ ~ v CQ, w 4) CI' e co - C4 o C- _- CC) co c to s co_- +~~~~~~~~~t CO co:c-1 LO o o en en o: C1 o: 4 co cl cl 14 es 14 _; 1 C 4 ~. _ " 6bO._ N. o; Q 1X 4) 4) A 1- L. (._ (.2 o ; O *w D~ =44 d.n *_.( X ; _) co~ ~ ~ ~ ~ ~ o~ ~ -r c r.4) 2 e,^ r o@ * ^ E 5 2 -, 22 E 2> _.._2 b r,4)q e 4) i o o 4)._2 4) 4) - (.2. O o o 4)E Co o4c & C co ) o C4 n Cl4 co en o co CI CM4 _ C4 _ c12 :

12 .4 Co I o rt- cm rl~ 6 Lr; Co o Co ~6 l L or- ~ - oo co tc - C C-4 + C4 - Lr) b bi O) O O O) O) O O O O) O CO O O O O) O _- _. -. _ I_ I cid *~ -u ce) Ito O _- m n C C)~ C14 t- U2 _ t,v bo * 4 4- CIS2 Q V. Its C4 -.2 * *sq Q~~~~~w 9-.. S q C. C. ". v u x m v!i X LI4~ v " o.b 4 O 7O C) E 4 - x ~WI~ s.4 o " CU co... -" *~CU v CU "Z lv *- 1. _ VU2 ox O O A. 2 to _ c, " o O Q.c 8 C.)q E =, S".m I4Q.: o co o - C.t C14 I_ C"4 c Cq

13 ce co +- co 1+ oo ld uo.4 n o b Q o o omo co cn Co o Q _ l Cli C-4 C-4 Clt + _ N o6cb (M CN to C4 m LO C- so _ ei Le cli C- + Y C) ce o t- + ko - ~o ~o ld4 i U- O ~o Cn o t- en LO o ^ ~~~~~~~~~~~~~~~~ v a ~~ b S+C b.

14 372 Health Services Research 2:3 (August 1985) Table 3: Percent Distribution of Patients Within Three Discharge Destinations by DRG-Based Length of Stay Categories: Maryland Medicare Patients, Age 65 and Over, 198 Leth of Stay Greater Than Less Than the Geometric or Equal Mean but Less Greater To the Than or Equal Than the Total Geometric To the Outlier Discharge -Discharges Total Mean Outlier Cutoff Cutoff Destination Number Percent Distribution Total 13, Nursing home 8, Home health care 1, Self-care 93, Source: National Center for Health Services Research, Hospital Studies Program, and the State of Maryland Health Services Cost Review Commission. Table 4: Percent Distribution of Patient-Days Within Three Discharge Destinations by Length of Stay Categories: Maryland Medicare Patients, Age 65 and Over, 198 Length of Stay Greater Than Less Than the Gemetric or Equal Mean but Less Greater Total To the Than or Equal Than the Patient- Gometric To the Outlier Discharge Days Total Mean Outlier Cutoff Cutoff Destination Number Percent Distribution Total 1,31, Nursing home 19, Home health care 3, Self-care 1,79, Source: National Center for Health Services Research, Hospital Studies Program, and the State of Maryland Health Services Cost Review Commission.

15 DRGs and Long-Term Care Need 373 PATIENT CHARACTERISTICS Table 5 shows the distribution of patients by characteristics available from the Maryland data across the length of stay groups for each discharge destination. Advanced age, though strongly related to the likelihood of discharge to a nursing home, does not necessarily mean that the patient will have an unusually long hospital stay. Only 16.3 percent of discharges to nursing homes in the oldest age group stayed beyond the outlier cutoff for their DRG compared to 2.5 percent of those aged and 25.8 percent of those aged For patients discharged to a nursing home, those aged 85 and over are much more likely to be discharged on or before the average length of stay for their DRGs than the two younger patient groups. Compared to nursing home discharges, age group differences across the length of stay distributions are insignificant for home health care, but are in the opposite direction for self-care. This suggests that some of the nursing home discharges, particularly of the oldest age groups, mnay have had short stays because they were originally admitted to the hospital from nursing homes and essentially have ready access to long-term care beds and skilled nursing care. These patients could be discharged at an earlier stage of their recovery. This speculation is explored directly with the San Diego data below. Continuing with the Maryland data, the discharge patterns for female patients are similar to those for older patients, probably because older patients are more likely to be female. Patients with more diagnoses and receiving more hospital procedures, such as surgeries and invasive diagnostic tests, are more likely to be outlier cases, regardless of the discharge destination. Nursing home patients are also more likely to have more medical problems: 6 percent of nursing home discharges versus 4 percent of self-care discharges have four or more diagnoses. This estimate supports the earlier guess that a considerable number of outliers among nursing home discharges are probably due to their greater severity of illness not reflected in DRGs. It is notable, however, that nursing home patients, who we know have more diagnoses, do not receive more procedures than self-care patients: only 33 percent of nursing home discharges received three or more procedures compared to 38 percent for self-care discharges. This agrees with findings of treatment patterns of elderly patients studied by Garnick and Short [15] for five major disease categories. For many frail older patients, procedures are not performed as frequently, probably because of the patient's poor prognosis for recovery. Additional characteristics of nursing home patients are described

16 -. -. '.. * * s bn V) Oc st Q...&.4 C -.a 94 Ot 4 'Ob rlic; 4O II C CNC LO I U, _ oq Dlu U, u CL C - U w _ P-4, t It Vs A ;.s t ztz..2 (5 J4 s. (Z) ;3 d S: ".2 t, -.; 1.4 t- r. CC4 " q V ".2 C -.l %4 t- gl. q C14 ce) - C-4 C-4 -Co t- C-4 Cf) ".4 "I Lif) if') if') t- C4 oo t- Cee) C14 Co Ct) C-4 ~c 44 -_ Ce' -f') -* C14O c C, ; C-4 CNC t- oco co LO -C O * LO lq t o_d LO C) C4*e C Co CD CC) tcs _e C4 14c CW rb iqo. 4- Q Q ~o U X o - m'-) if) - Co co Co LO o) ' o o ~o co if') o CSf,L Z"'i 4. Co ) ') F-' os,.

17 CN (. ce cl4 co r t- Ce C14 -o Co 11 m C.4 co C4 Co 6 ) co -C l:tq c u cn - o o lct 6 (s C; 6 cle C14 lq C%i O) co Co Co) C)o e) 4-l11 Uq LC) 1-4 C' LO - c C.".Cf lo' Co~V~ Co C4 ceo n. rs ceo o: om - - ) Co6 C~ _; C.4 u1 on I t-~ U- LO U- o oco -OCo 114 co C- cl ce en n C1.) " C- U Cd 4rJ U " Ubi) t.. LO- -4 4C) LO = oo co LrO OC co LO Q6 4C ~o C 1- LO Co ro ~o ce) ~o Co C) Co 1- C'". c Co CeLrs` LO1I4 co) - C C'4 C- 4) 6 r ) c.4 ".o C 114 ce) C1 ce) ce) ce ~o o C4 U C C. UfC LO t. 4o.4 O o 4) U.4- U rj2.4- cd ce -1 so Co~ ~ ~ I _ C v.. 1:4

18 ..Z!.. 6. C14 tf% 4i4 4f.. C~4 '*o-o ; C14 C,4 cf)- V. C-4 Om e c4i O en C14 H C4 Co 44 "' C4 m bid -a.4;.. qqd o. a '.l A,q -., 4z, -A 9: 8: a 4 " 4 :. 4 %.. 45 C4) C14 t 6C C~JC Its - ~ c Cr; CO' v C _ 4 S:, -4, se F.- N 13'. -. p v 1.. Cb 1.4 s C14 4 N4 C14 C4 "C14C oo o o6.4 D ~~(OC.....C.C; C2 p *4 QE bo co I*l.. c 4 oio _C4 44 _ C en co V) 4n.1, Ia g v ~.,.Ir4 C-4 S -

19 On C4 C c N) 'Nit Ui) Uir -4 Cei o C L"I NW C,4 C- C'. -4 * t. C14 _ en) I4 C14 CW4 % C14 LO U) LO it") O _ N- ~l._ :s, C._ Cen C6-4 Lo LF O O LO U-) Lir U:)._ " r- C;4 _1 LO.4 toc. el CN rl cf - 14 co O O O _ Ji ~ bo C) br 6 6. * C) C> ( O C) O O (D O PA 1. CO 4CU V C.) LO (a) C) _ - CQ " bi C.) _ co) _ CL) C4 C _ C14 Ci" _ " _ C) o o; CU o C.) -4 C)b.. U e CUY SO A, CU 2 "~~~~~~~V -c

20 378 Health Services Research 2:3 (August 1985) next using the San Diego data. Of the 695 patients available for the analysis, 217 were institutionalized in a nursing home prior to their hospital admission and subsequently discharged to a nursing home. As expected, 36 percent of these patients are in the shortest length of stay group and only 12 percent are in the outlier group compared to 25 and 22 percent, respectively, for the patients not previously institutionalized. Patients in nursing homes who are transferred to a hospital for an acute illness episode often have their nursing home bed reserved, minimizing the hospital stay. These cases are removed from the San Diego data presented in Table 6, so that effects of PPS incentives on patients without ready access to nursing home beds can be assessed. Almost three-fourths of the patients discharged to San Diego nursing homes have a hospital stay greater than the average for their DRG; 22 percent have a very long stay beyond the outlier cutoff. These are similar to Maryland statistics. The distribution of San Diego patients across length of stay categories by age shows no overall significant differences, but the direction of differences is the same as for Maryland. In fact, if statistical significance is tested across age groups for outliers only, the oldest age group has a significantly smaller proportion of patients in the outlier category compared to the younger age groups. Thus, our earlier speculation, based on Maryland data, that prior institutionalization explains this surprising finding may be part but not all of the story. The fact that fewer procedures are performed on the very old in conjunction with their shorter stay and greater number of medical problems suggests that providers aim to stabilize the very old quickly and place them in a nursing setting with less medical care in the hospital. If this hypothesis is true, then fixed prospective per case payments will reinforce these decisions. It is often argued that having no spouse and/or living alone are important determinants of the demand for nursing home care. Here we examine these factors in relation to hospital length of stay prior to nursing home placement. Although most patients discharged to nursing homes (81 percent) are not married, overall their distribution across the three length of stay categories is not significantly different from that of married patients. However, if proportions of patients beyond the average DRG length of stay are examined, statistically significant differences occur at the.5 level: 8 percent of unmarried patients - and only 7 percent of married patients - stay beyond the average. Solitary living arrangements do not significantly influence the patients' hospital length of stay in the San Diego data. Several health-related factors distinguish levels of care for longterm care patients: need for assistance in activities of daily living

21 DRGs and Long-Term Care Need (ADL) (bathing, dressing, going to the bathroom, transfer, continence, and eating [16, 17]); presence of mental illness; existence of behavioral problems; and anticipated length of nursing home stay. These factors make patients more difficult to place in nursing homes and thus are recognized as potential contributors to the hospital backup problem [18, 2, 19]. Results in Table 6 generally agree with this assessment. Patients dependent in five to six activities of daily living are less likely to be discharged from the hospital on or before the average length of stay for their DRG and more likely to have an outlier length of stay. Patients with long expected nursing home stays are significantly more likely to have a hospital stay greater than the average for their DRG. Patients with behavioral problems also have longer hospital stays. However, patients with mental disorders, defined as ICD-9-CM codes for principal or secondary hospital diagnoses, are more likely to be discharged earlier from the hospital than those without mental disorders. Thus, the San Diego data suggest that it is the manifestation of behavioral problems, not necessarily the presence of mental disorders, that makes certain patients difficult to place in nursing homes. Patients with psychobehavioral problems have been found elsewhere to be an important part of the hospital backup problem [18]. It is clear from both the Maryland and San Diego data that the DRG classification system does not capture a number of patient characteristics that are associated with differences in hospital length of stay. These simple associations should not necessarily be interpreted as needed DRG modifications. Other potentially confounding factors, most notably the success of the hospital in billing other payers and the availability of nursing home and home health services, may be important [2]. The market for home health programs in Maryland, for example, was quite underdeveloped in the early 198s, although it is now experiencing dramatic growth [21]. SUMMARY AND DISCUSSION 379 There is evidence that elderly patients requiring extended care in nursing homes or from home health agencies are backing up in hospitals [19, 2]. The new Medicare DRG-based Prospective Payment System (PPS) is expected to encourage hospitals to take a more active interest in solving this problem. This article has studied the types of hospital DRGs which comprise nursing home, home health, and self-care discharges. In addition, it has examined characteristics of patients likely to be discharged earlier as evidenced by their tendency to stay in the

22 38 Health Servzces Research 2:3 (August 1985) hospital longer than the average for their DRG. The findings should be considered exploratory because they are based only on data from Maryland and San Diego County, and because they do not capture all of the nuances of the new Medicare payment system. However, the data offer some useful insights about the effect of the new Medicare PPS on discharges to long-term care services, and they provide a basis for comparison and further analyses as other data become available. Data for the year 198 on Maryland hospital patients aged 65 years and over indicate that there are differences in the most common DRGs discharged to nursing homes, home health care, and self-care. Discharges to nursing homes fall more frequently in diagnostic categories that require skilled rehabilitation services, that reflect mental or behavioral problems, or that specifically reflect frailty of old age. Discharges to home health care agencies more likely comprise patients in diagnostic categories that may require long-term management but do not necessarily represent debilitating conditions, at least in some stages of the illness. A length of stay analysis shows that patients referred to long-term care services are more likely than other patients to lose money for the hospital by staying beyond the average number of days for their DRG; about 7 percent of Maryland discharges to nursing homes and 68 percent of those to home health care stayed beyond the average for their DRG compared to only 48 percent of patients discharged to selfcare. Elderly Medicare patients needing long-term care services account for a disproportionately large number of unreimbursed days compared to self-care discharges. These results suggest that the Medicare Prospective Payment System gives hospitals a strong incentive for earlier discharge of patients needing long-term care services. However, there are two factors which mitigate this incentive. First, because longterm care discharges are such a small portion of cases treated in hospitals, hospitals that target primarily long-term care patients for earlier discharge will not substantially gain under PPS. Eighty-seven percent of patients staying beyond the average length of stay for their DRG are self-care discharges. Eighty percent of the days that would be unreimbursed under PPS are related to self-care discharges. Thus, to profit from PPS, hospitals must reduce length of stay for all types of patients. Second, hospitals will be able to recover some of the losses on longterm care patients under PPS by billing other payers, once the hospital establishes the "administrative day" status of the patient and allows for the 2-day grace period. Further research is needed on the effect of the "administrative day" ruling to assess more accurately its effects and those of PPS generally on the long-term care system. Also, the ade-

23 DRGs and Long-Term Care Need 381 quacy of Medicare and private supplemental nursing home and home health insurance benefits needs to be reevaluated to assure that they can accommodate the expected increased use of these services. Early identification of potentially long-staying patients is important for successful discharge planning activities. The New Jersey DRG reimbursement system has witnessed the importance of an effective discharge planning system [22, 23] in helping hospitals respond to prospective payment. The data in this study provide a basis for beginning to identify the characteristics of patients, particularly long-term care patients, who have long lengths of stay. As expected, long-staying hospital patients subsequently discharged to nursing homes were patients with five to six ADL dependencies, those who had behavioral problems, and those expected to have lengthy stays in the nursing home. These factors are generally understood to make patients more difficult to place in nursing homes. However, diagnosed mental disorders are not found to be associated with longer hospital stays before nursing home institutionalization. Furthermore, although older patients are more likely to receive nursing home care, they will not necessarily have unusually long hospital stays before discharge to extended-care services. This is probably not due to prior nursing home institutionalizations but rather to poor prognoses for these patients and, thus, less intensive treatments in the hospital and earlier placement arrangements. Perhaps more critical than discharge planning activities is the need to plan and develop an appropriate supply and mix of long-term care services to accommodate the increased pressure for earlier discharge of those elderly patients most likely to be a problem for hospitals. Under the Medicare PPS, hospitals will put greater pressure on nursing homes to admit more patients. In this regard, the analysis looks only at a part of the expected pressure; it does not predict how many and which types of patients who were discharged to their homes prior to PPS reimbursement will now be discharged to a nursing home or home health agency to shorten their hospital stay. In the short run, nursing homes have little capacity with which to respond to this pressure. Hospitals and home health agencies are likely to assume a larger role in providing long-term care. Some hospitals will probably convert hospital beds to separate long-term care units to accommodate the growing elderly population [24]. This may be stimulated in the short run by interim PPS regulations which reimburse these units on a cost basis until long-term care facilities are brought under PPS regulations. Growth in the number of home health agencies will probably be accompanied by changes in their patient mix, which is

24 382 Health Services Research 2:3 (August 1985) likely to change as hospitals seek to discharge patients earlier. New Jersey apparently has experienced such case-mix changes since more patients need daily visits and visits of longer duration [22, 23]. Nursing homes also must accept a heavier-care case mix to accommodate the growing need for post-hospital extended care. Nursing homes prefer to admit private pay patients and light-care public pay patients, precisely those patients with family circumstances and care needs that make for expeditious hospital discharge to home health care settings. Nursing homes should be encouraged to concentrate on those patients whose physical and behavioral problems and family situations make "community" placement unlikely. The conflicting incentives faced by hospitals and nursing homes must be resolved if the incentives of the Medicare PPS are to work efficiently. Closer ties between hospitals and nursing homes through the provision of educational service, consultation, and medical coverage have been noted as strategies being used by hospitals to gain greater access to long-term care beds [24]. A more direct approach would be for hospitals to pay incentives to nursing homes for more difficult-to-place patients. How nursing homes respond to incentive payments to admit heavier-care patients is the subject of the NCHSR Nursing Home Incentive Reimbursement Study in San Diego. Although not designed with the Medicare PPS in mind, the San Diego study will provide insights into the effectiveness of incentive reimbursements for nursing homes and the feasibility of case mix-based payments in extended care facilities. ACKNOWLEDGMENTS The authors wish to acknowledge the computer programming assistance ofjack Bieler, Social and Scientific Systems, Inc., and Marilyn Barron and Tobin Short, National Center for Health Services Research, and the research assistance of Arlene Tave of the National Center for Health Services Research. NOTES 1. The fixed DRG rate per patient is also adjusted by specific hospital and community characteristics, by allowances early on for transition to the national Prospective Payment System (PPS), and by extremely costly cases in terms of resources used. See Federal Register 48(171), September 1, 1983, and 49(1), January 3, 1984, for the PPS regulations.

25 DRGs and Long-Term Care Need These calculations were made from population statistics in the Area Resource File compiled by the Bureau of Health Professions, U.S. Department of Health and Human Services. Nursing home beds for 198 were obtained from the National Center for Health Statistics, DHHS, unpublished data. REFERENCES 1. Feder, J., and W. Scanlon. Medicare and Medicaid Patients' Access to Skilled Nursing Facilities. Working Paper Washington, DC: The Urban Institute, November Gurenberg, L., and T. Willemain. Hospital discharge queues in Massachusetts. Medical Care 2(2):188-21, Inui, T. S., et al. Identifying hospital patients who need early discharge planning for special dispositions: A comparison of alternative techniques. Medical Care 19(9):922-29, September Scanlon, W. A theory of the nursing home market. Inquiry 17(1):25-41, April Shapiro, E., N. P. Roos, and S. Kavanagh. Long term patients in acute care beds: Is there a cure? Gerontologist 2(3):342-49, Vladek, B. C. Unloving Care: The Nursing Home Tragedy. New York: Basic Books, Weissert, W. G., et al. Encouraging appropriate care for the chronically ill: Design of an experiment in patient based nursing home incentive payments. Health Care Financing Review 5(2):41-49, December Averill, R. F., R. L. Mullin, and E. D. Elia. The Revised ICD-9-CM Diagnosis Related Groups. New Haven, CT: Health Systems International, May National Center for Health Statistics. National Hospital Discharge Survey. Unpublished data, Meiners, M. R., G. D. Heinemann, and B. J. Jones. An Evaluation of Nursing Home Payments Designed to Encourage Appropriate Care for the Chronically Ill: Some Preliminary Findings. Working Paper, presented at the American Economic Association Meeting, New York, New York, December Coffey, R. M., and M. G. Goldfarb. DRGs and Disease Staging in Reimbursing Medicare Patients. Working Paper #1, Hospital Studies Program. Rockville, MD: National Center for Health Services Research, October Short, T. L., and R. M. Coffey. Diagnosis Related Groups vs. Disease Staging: Implications for Hospital Reimbursement. Working Paper #7, Hospital Studies Program. Rockville, MD: National Center for Health Services Research, October Hornbrook, M. C. Regional Differences in Hospital Length of Stay. Hospital Studies Program, National Center for Health Services Research, Rockville, MD, March Steinwachs, D. M., and D. S. Salkever. Impact of "Per Case" Versus "Per Service" Hospital Payment in Maryland. Draft final report to the

26 384 Health Services Research 2:3 (August 1985) National Center for Health Services Research, Rockville, MD, May Garnick, D., and T. L. Short. Utilization of Hospital Inpatient Services by Elderly Americans. HCUP Research Note. RockVirile, MD: National Center for Health Services Research, forthcoming. 16. Kane, R. A., and R. L. Kane. Assessing the Elderly: A Practical Guide to Measurement. Lexington, MA: Lexington Books, Katz, S., and C. A. Akpom. A measure of primary sociobiological functions. InternationalJournal of Health Services 6:493-58, Eggert, G. M., and B. S. Brodows. Mental health patients, a growing health system problem. American Health Care AssociationJournal 8(6):24-3, November Feder, J., and W. Scanlon. The under-used benefit: Medicare's coverage of nursing home care. Milbank Memori Fund Qaly 4(3):64-25, Connor, M. J., and S. B. Greene. Should extended care following hospitalization be encouraged? Inquiry 2(3):258-63, Fall McCleary, M., and R. R. Driscoll. Health services cost review: The impact on home health services in Maryland. Caring 3(2):46-47, February Livengood, W., C. Smith, and S. Hallstead. The impact of DRGs on home health care. Home Healthcare Nurse September-October Livengood, W. DRG system in New Jersey impact on reimbursement. Caring 3(2):48-49, February Campion, E. W., A. Bang, and M. I. May. Why acute-care hospitals must undertake long-term care. The New England Journal of Medicine 38(2):71-75, January 1983.