Trends in Homitall. Use by the Aged

d, JULIAN H. PETTENGILL*

data on hospital utilization by the aged since This article compares data from a variety 1963. Several measures of utilization including of sources on the hospital utilization experi- admissions or discharges per 1,000 population, ence of persons aged 65 and over during the days of care per 1,000 population, and average period 1963-71. These data show that, in the length of stay are employed. first 3 years fo7lowing the inception of Medi- In order to assess the impact of the HI pro- care, use of hospital services by the aged- gram, two questions are of particular interest : measured by days of care per 1~000 persons- increased at an average annual rate of 6-13 1 How much increased use of hospital services was generated in the first year of the program compared percent. Since 1969, however, hospital utili- with earlier periods? zation by this age group has declining been 2 What has been the continuing trend of hospital l-3 percent annually, mostly as a result of a use by the aped in more recent years? decline in average length of stay. Compari- son with similar data for persons under age Although neither question can be answered 65 reveals that a significant redistribution of with absolute precision, it is hoped that a compar- hospital care between this age group and ison of the data from several sources will provide those aged 65 and over has occurred since a range of values approximating the actual trend 1965. It is likely that aged persons increased csperirncc. their use of hospital services to some extent at the expense of persons under age 65. SOURCES AND LIMITATIONS OF DATA The data used in this paper are derived from ~MEDICARE’s two coordinated programs of seven different sources.2 A brief summary of the health insurance for the aged-a basic hospital principal features of each source is presented in insurance program (HI) and a voluntary supple- table 1. along with a discussion of some inherent mentary medical insurance program (SMI) - differences between these surveys that tend to went into effect on , 1966. The primary goal complicate the comparison of their results. of the Social Security Administration’s Medicare Bs table 1 shows, none of the surveys are com- program is to provide insurance protection to pletely alike. Generally speaking, the dissimilari- help persons aged 65 and over pay a major por- ties are of two types: differences in approach and tion of their large expenses for hospital and med- design and diHerences of definition, in the broad ical care. The success of the Medicare program as sense of the term. a whole has been amply documented in a previous If classified according to their ultimate source artic1e.l of information, three distinct kinds of surveys are Financial protection was the main focus of the included here. Data were collected in both the program ; however, it was apparent that the aged Health Interview Survey (HIS) and the Colum- would probably increase their hospital utilization bia University Survey (GUS) from interviews of under the hospital insurance part of t.he program. individuals in a probability sample of households. Predictions about the probable size of the increase Data reported by both American Hospital ASSO- varied from 5-10 percent to 40 percent or more. riation (AHA) sources (Guide Issue and Hospi- This article measures part of the impact of the tal Indicators) wese collected in mail surveys of HI program by examining the existing sources of hospitals registered by the Association. The re- maining three surveys used individual case rec- *Division of Health Insurance Studies, Office of Re- ords: discharges in the Hospital Discharge Sur- search and Statistics ’ Howard West, “Five Pears of Medicare-A Statistical ‘For a more detailed description of sources, see the Review,” Socaal Securzty Bulletzn, December 1071. Technical Sate, page 11.

BULLETIN, JULY 1972 3 TABLE l.-Pnncipal characteristics of the seven survey data sources

Oeneral Reference Reference Institutional Nonre- description time period population coverage I sponae rate I-- National Center Annual general Multistage proh- 42,ooOhouseholds Independent Civilian non- All general or 5 percent for Health Sta- health data ability sample of containing 134,COLlweekly samples lnstitutlonal other special tistics, Health households within persons In which rcspond- populntlon living hospitals or hos- Interview Survey. *sample of geo- ents were queried at time of inter- pital unit of an graphic segments about experience view. institution except of previous 6 chronic, tuber- months Data culosh, or were aggregated psychiatric to annual figures’ fiscal years 1963- 67, calendar years 1967-69

National Center Annual short-stay Multistage prob- Approximately Annual data ac- Clvihan non- percent for Health Sta- hospltal use data ability sample of 210,MKIdischarges cumulated on a institutional tistics, Hospital discharge records from about 409 monthly basis population. Discharge Survey. within a sample hospitals calendar years of hospitals 1965-70

Columbia Uni- Measurement of Multistage prob- 6,604persons.-..- Two independent Old-age and sur- All hospitals and 12 percent verslty Survey. the early impact ability sample of samples inter- vivors insurance extended-care of Me&care. individuals viewed in April- beneficiaries. facilities ccrtifled I within a sample May 1966and for participation of geographic November- In Medicare and sampling units. December 1967In nursing homes which respond- not participating ents were queried in the program about experience of previous 12 months I American Hospital Monthly current Single&age 813hospitals..... Monthly survey Civilian resident Short-term gen- Not Assoclatlon, Hos- data on com- stratified sample Annual figures art population era1 and other available pital Indicators. munity hospital of hospitals aggregated for special hospitals utilization and both calendar and (community). Ilnances gg7Jjea%

American Hospital Annual statistics Mail survey of all 7,123hospitals-.-. Annual survey Civilian resident All hospitals, ln- 7 percent. AsTciation, Guide of hospital utiliza- hospitals regis- Data generally population cludlng subgroups tion and finances tered with the refer to the 12 such as eom- AHA in the U 8 months ending munity hospitals y;p7pr 3%

Bodal 6ecurity Monthly, quar- Single-stage Approximately Records were All persons en- All hospitals 3 percent Administration, terly, and annual stratifled sample 39,ooOhospital sampled monthly rolled in the HI certified for par- Current Medicare statistics on dis- of all hospital stays per year. Annual figures program ticipation in the Survey charges covered admission notices arc aggregated for HI program by the HI pro- both calendar and gram, with days asc;lps, of care and charges I-- Social Security Annual statistica Census of all 100percent ______Claims data are All persons en- All hospitals ----m-_---*-m Administration, of claims for claims records accumulated rolled in the HI centificd for g;z;.F Control hospital care continuously. program partleipation in covered under HI Annual flgnrcs the HI program program. are aggregated for both calendar and gg7;esrs, vey (HDS), paid claims in the Medicare Control As a result the levels of hospital use estimated Records (MCR) , and admission notices in the from different sources are not strictly comparable. Current Medicare Survey (CMS) data. ’ Household survey data, for example, exclude the Definitional differences among the surveys in- experience of persons who died in the hospital or volve such questions as who is included in the who died after discharge but before the interview survey, what is considered hospital utilization, date. These data also contain substantial under- and when the reported hospital use occurred. reporting (5-10 percent 3) in the amount of hos- These surveys do not measure hospital utilization pital use by respondents. Thus the level of use for the same population, nor do they have a single common hospital universe or even an identical ‘iWHS Vital and Health ki’tatistics, Series 2, n’os. 6 time frame. and 7.

4 SOCIAL SECURITY estimat,ed from household data will not corre- has been published for the HDS, HIS, CMS,‘&d spond with the results from other types of sur- CUS data. Although the relative errors range veys. from 2.5 percent to about 6 percent, there is no For the purpose of comparing rates of change conclusive evidence of a radical difference among of hospital utilization over time, the situation is surveys. somewhat better. Since each individual survey has Neither the MCR nor the AHA Guide Issue remained reasonably consistent during the period data are subject to sampling variability since in terms of its design, techniques, and definitions, these are derived from a complete enumeration of annual rates of change estimated from different the relevant universe rather than a sample. These sources should be approximately the same. data have limitations, however. MCR data for Rates of change cannot be expected to match recent years are incomplete because of lags in the exactly, however. In any given year, each survey filing and processing of claims; AHA Guide Issue measures the hospital utilization experience of a data do not provide utilization estimates for dif- different group of people, in a different set of ferent age groups. hospitals, during a different 12-month period. The seven data sources do not provide equally Such definitional differences will cause rates of precise estimates of hospital use. A ranking of change reported by different surveys to vary surveys in descending order of precision would somewhat, although the size of the differences probably begin with those based on complete enu- should be small. meration (MCR and AHA Guide Issue data) and A larger proportion of the disparities among end with those based on household interviews the rates of change from various surveys is proba- (HIS and CUS data). bly due to the inherent variability of survey data Variation in rates of change reported by the that results from errors of sampling and measure- different sources makes comparison more difficult. ment. Naturally, the likelihood that a set of data The general trend and direction of change, how- will show distorted rates of change from year to ever, can still be evaluated, although the degree of year increases with the degree of variability. Un- change resulting from the Medicare program fortunately, very little information is available cannot be precisely estimated. regarding the relative variability of each source’s data. There are, however, reasons to believe that some of the survey data presented here tend to- ward greater precision than others. MEDICARE’S INITIAL IMPACT Part of the variability of any survey estimate is As noted previously, a large increase in hospi- due t,o nonresponse, reporting inaccuracies, and tal utilization by the aged was expected to occur errors in processing. Although the amount of var- in the early years of the program. When esti- iability resulting from nonresponse and process- mated levels of use during Medicare’s first year ing errors is probably similar for each type of are compared with corresponding estimates for survey, the amount due to reporting error is the last year before the program went into effect, likely to be much greater for household interview it is evident that, a substantial increase in utiliza- survey data than for surveys using either hospital tion did occur. or claims records. Response accuracy in a house- Table 2 compares CUS, HDS, and HIS data on hold int,erview depends upon the ability and will- the number of stays or discharges and days of ingness of the respondent to recall past events. A care per 1,000 population and average length of survey of records, however, makes use of informa- stay before and under Medicare. As expected, the tion compiled at the time of the event. estimated utilization rates for both periods vary The remaining and most important source of from survey to survey as a result of definitional variability is sampling error. Sampling variabil- and design differences already Giscussed. In addi- ity results from the fact that repeated samples tion the reference period for each survey is differ- using identical techniques will provide different ent. The “before” period covers April- estimates. The standard error or relative error, to April- for the CUS data ; it is calen- which describes the dispersion of these sample es- dar year 1965 for the HDS data and fiscal year timates around the true value for the population, 1966 for the HIS data. “Under Medicare” refers

BULLETIN, JULY 1972 5 TABLE S.-Hospital utilization experience of persons aged hospital utilization focuses on two effects of the 65 and over in short-stay hospitals, before and under Medi- care, three sources Medicare program. First, more than 97 percent of all persons aged 65 or over were eligible for hos- pital benefits at the bepinning of the program. A substantial number of these aged persons had no previous hospital insurance protection at all. Columbia University Survey *. stays per l,ooo persons_-__---_ 247 11 3 Second, some of the increase may have come from Days of care per 1,M)opersons.. 3,:: 3.924 Averagelength of stay (days).. 142 16 9 :; It admissions that were deferred until after the Health Interview Survey effective date of the program. DaysDischarges of care perper 1,ooO 1,000persons...persons-- 2,253177 2,656 197 1711 38 1: :: Average length of stay (days)- 12 7 13 6 03 41 In 1962, gross enrollment under private health Hospital Discharge Survey: Dischargesper 1,CWJpersons ___ 264 289 46 insurance for hospital benefits was estimated at Days of care per l.OCQpersons- 3.444 Average length of stay (days)-. 13 0 4iz E approximately 12 million aged persons.4 When duplication resulting from multiple enrollment by * Based on compound annual rates of interest 1Data from Regina Lowenstein,“Early Effects of Medicare on Health some individuals was eliminated, the net number Care of the Aged,” Social Smwfl~ Bulklln, ,table 4. of different insured persons in this age group fell to slightly more than 9 million persons, or 54 to November- to November- percent of the population aged 65 and over. Al- ‘for the CIX data and calendar though the percentage enrolled may have in- year 1967 for both the HDS and HIS data. Al- creased slightly between 1962 and 1966, it is clear though HIS data for fiscal year 1967 are availa- that nearly half of those who became eligible for ble. they are not appropriate for comparative Medicare benefits on July 1,1966, were previously purposes. Since the data collect,ed in interviews uninsured. It is likely that many of these individ- are based on a 6-month recall period, figures re- uals came into the Medicare program with signif- ported for fiscal year 1967 include hospitaliza- icant mlmet medical needs. This factor alone tions that actually occurred before the beginning could explain a substantial portion of the total of the Medicare program. increase of hospital utilization rates. Because the length of time between reference The second factor-postponement of hospitali- periods is different in each case, gross percentage zation-affected not only those with no private increase figures derived from these surveys are hospital insurance but also those who had some not comparable. In order to eliminate this prob- insurance. Clearly, any person who expected to be lem, annual rates of increase have been computed. eligible for Medicare benefits would have had an The corrected annual rates derived from each incentive to postpone hospital care where possible survey indicate that days of care per 1,000 aged if his current benefits under private health insur- persons increased between 9 percent and 16 per- ance were less complete than those provided cent in the first year of the Medicare program. under Medicare. The extent to which hospitaliza- This growth reflects the combined effects of a rise tion was actually postponed is unknown. That it in the stay or discharge rate of 5-7 percent and did occur is suggested by the fact that the dis- an increase of 4-8 percent in average length of charge rate among aged patients for the treat- stay. ment of cataracts-generally considered elective Several factors could have accounted for such care-rose 52 percent between 1965 and 1967.5 increases in utilization rates. Aged persons could A major share of the increase in hospital utili- have experienced a greater number of illnesses zation by the aged seems to be directly due to the and more serious illnesses than usual in the first impact of the Medicare program. If increased use year of the program. Fluctuations in hospital use had resulted solely from a temporary change in by the aged are related to some extent to out- the incidence of illness, utilization rates could be breaks of upper respiratory diseases in winter and to prolonged heat spells in summer. Available 4 hlarjorie Smith Mueller, “Private Health Insurance evidence, however, does not indicate that the 18- in 1970: Population Coverage, Enrollment, and Financial month period from July 1966 to December 1967 Experience,” Social Security Bulletin, February 1972, table 8. was unusual in either respect. “Nonthly Vital i3tatistics Report, vol. 20, No. 7, Oc- A more plausible explanation for the increased tober 8, 1971.

6 SOCIAL SECURITY expected to level off or decline in later, more nor- TABLE 4.-Annual percentage change in hospital utilization rates of persons aged 65 and over in all hospitals, three mal years. As described below, however, hospital sources, fiscal years 1967-71 ’ use by the aged population continued to increase - for several years following the initial period of Percentage change from %Elr program operation. previous year PFh?;zege Survey and utilization measure - - 1969 1971 1968 1969 1970 I 1971 from 1967 ‘%i ------MEDICARE’S CONTINUING IMPACT AHA Hospital Indicators 1, Admissions per 1,COOpopulation. Days of care per 1,600po ulstion. E !f 2; 1; f :9” ri p Because of the varying nature of the time peri- Average length of stay (a ays) _-_. 47 -8 -3 0 20 Current Medicare Survey 1 ods covered by the different sources, the continu- Discharges per 1,000population-. 69 56 (9 (3 5 8 (9 Covered days of care per l,CnNl ing trends of hospital utilization are evaluated by population ______.81 30 (9 Cl 55 19 Average length of covered stay means of two separate comparisons. Table 3 pres- (days) ______20 -2 3 0 (9 - 7 (“1 Me&care Control Records, ents rates of change in hospital utilization availa- Admissions per 1,ooOpopnlation.. 117 34 -.3 10 74 .3 Covered days of care per 1,ooO ble from four sources for calendar years 1967-70. population ______.199 59 -3 8 .2 9 12 6 -3 4 Average length of covered stay Similar trend data are given in table 4 for three (days) ______74 23 -3 8 .40 48 -3 a surveys that provide estimates for fiscal years - - 1 Commumty hospitals 1967-U. 3 Based on covered utilization under HI in participating hospitals. : Not available Table 3 shows that hospital utilization rates continued to rise between 1967 and 1968, but that the HIS data is three or more times larger than by the end of 1969 the period of rapid growth the comparable rate from any other source, it had come to an end. Total use, as measured by seems unlikely that it describes the real trend days of care per 1,000 persons, increased between accurately; and (2) although the HIS data may 5 percent and 34 percent from 1967 to 1968 as a overstate the rate of change, the CMS data proba- result of a rise of 4-14 percent in the discharge bly understate it. In the CMS data length of stay rate and a change ranging from -2 to 17 percent is based only on days of care covered by the Medi- in average length of stay. care program and does not include the effect of a Rates of change derived from the HIS and 1967 amendment (effective January 1, 1968) pro- CMS data for days of care per 1,000 persons and viding a lifetime reserve of 60 days of covered average length of stay must be qualified, however, hospital care for each enrollee. Before this by the following considerations: (1) Since the change, an enrollee would have exhausted his ben- percentage change in eit,her utilization measure in efits under the program on the 90th day of hospi- tal care in a benefit period. After this change, TABLE 3 -Annual rates of change m hospital utilization rates of the population aged 65 and over in short-stay hos- however, such a person could obtain additional pitals, four sources, 1967-70 covered days of care by drawing on his lifetime reserve. Since the C&IS data exclude days of care Percentage change from previous year beyond 90 during a benefit period, actual covered Year and measure 1 days of care and average length of covered stay are understated in 1968 and subsequent years but _- are comparable with 1967 data. If days of care Discharges per 1,000~pulation beyond 90 had been included, the rate of increase 1988--.-.-.-..--..------14 1 lsBS----.-.------in days of care per 1,000 persons would have been 1970_____-_-----_ ------ii! Days of care per 1,000population somewhat higher, and average length of stay 1988.-...----..-.------33 6 1969.--_ _ _ _---__--- ______----- might have shown no change between 1967 and 1970__-______------14 i Average length of stay (days). 1968. 1968.______------_------17 0 1989...---.------5 7 Despite these qualifications, however, total hos- 19m....---.----.--.------12 0 - pital utilization by aged persons increased sub- 1 Bawd on rate of admissions. stantially between 1967 and 1968. This increase 1 Rates may be too low because of a change in the base used in estimating population data In 1970 probably resulted from a combination of factors 8 Not available 4 Covered days of care under HI including an outbreak of upper respiratory dis- 5 Average length of covered stay per discharge.

BULLETIN, JULY 1972 7 ease (that occurred during the winter of under the program include two distinct periods. 196%69), some residual catching-up with pre- During the first 3 years, utilization rates in- viously unmet needs, or increasing awareness creased substantially but at a declining rate of among aged persons of the benefits available to increase each year. From 1969 to 1971 the trend them. In addition, the lifetime reserve provision reversed and total hospital use by the aged began may have had some effect upon average length of to decline absolutely. Thus, during the second pe- stay and days of care. This change may also have riod, days of care per 1,000 persons declined at an produced a higher rate of multiple admissions of average rate of 2-3 percent per year. To be sure, the same individual during 1968 and in subse- the discharge rate continued to increase at a rate quent years. of about 1 percent per year, but at the same time Between 1968 and 1969, hospital use by aged average length of stay was decreasing about 3-4 persons leveled off: the discharge rate increased percent per year. more slowly and, more important, average length These trends probably resulted from the simul- of stay declined. By the end of 1970, days of care taneous interaction of several factors. The decline per 1,000 persons had decreased as a result of the in average length of stay, for example, may re- continuing and substantial decline in average flect intensified utilization review activities. These length of stay. activities may have had some impact on physician The extent of the decrease in days of care per awareness of the high cost of hospital care and, 1,000 persons between 1969 and 1970 reported in ultimately, on their decisions with regard to the HDS data is probably overstated. Utilization length of stay. Other factors have undoubtedly rates per 1,000 persons in 1969 were based on contributed to the decline in average length of population estimates projected from the 1960 stay since 1968. Case-mix, for example, may have Census. Similar rates for 1970 were based on pop- changed to include a higher proportion of less ulation estimates derived from the 1970 Census. serious cases as the backlog of previously unmet When these figures are compared the estimated needs was gradually eliminated. percentage increase in the population aged 65 and The causes of the continuing increase in the over between 1969 and 1970 is 2.9 percent-almost discharge rate are less certain. There does not twice as much as in any earlier year. As a result seem to be any evidence that the medical needs of the 1970 utilization rates are low in relation to the average aged person have increased in the last similar rates for 1969. few years of the program. Instead, the large in- The rates presented in table 4 exhibit substan- creases in the discharge rate between 1966 and tially the same pattern for the period 1967-69 as 1969 may have represented a one-time step to a those shown previously. Differences are mainly new equilibrium level of use that will prevail in attributable to the fact that the rates in table 4 the absence of changes in medical technology or are based on fiscal-year data covering a somewhat benefit provisions of the program. The l-percent earlier set of months than the calendar-year rates annual rise from 1969 to 19’71 may represent the in table 3. The change in Medicare that caused beginning of a stable, long-term trend of slowly the CMS data to understate the increase of hospi- rising discharge rates resulting from changes in tal utilization between 1967 and 1968-the inclu- medical techniques that tend to reduce the risk sion of the lifetime reserve-had the opposite associated with the treatment of specific disorders effect on the MCR data. The MCR data are based or that allow treatment of diseases that could not on days of care covered under the program, in- have been effectively dealt with before. Of course, cluding the lifetime reserve. Since days beyond slowly increasing discharge rates need not repre- 90, which were covered and therefore counted in sent the beginning of a new trend ; they , were not covered under the program in instead represent a return to a stable trend that 1967, the rates of increase in covered days of care may have existed before the initial years of the and in average length of covered stay are over- Medicare program. stated in relation to total days of care and to Slo~vly rising discharge rates might also result average length of stay actually used. as the hospital gradually replaces the physician’s The average annual rates in table 4 indicate office as the center of the health care delivery that the first 5 years of utilization experience system-not only for aged persons, but for the

8 SOCIAL SECURITY general population as well. Other changes-such TABLE B.-Percent of total discharges and days of care for patients under age 65, in short-stay hospitals, three sources, as the development of health maintenance organi- 1965-70 zations or the implementation of some form of national health insurance-can also be expected to have an impact on hospital utilization rates of all age groups. Percent of all discharges in- 1965--_-__-__-_----__------Similarly, the impact of the HI portion of the 1966------__--_ __--__ -_--_ _- 1967------_--_--______Medicare program has not been limited to 1968------_-______1969-.-.-...-.-..-.------changes in hospital utilization by persons aged 65 1970______Percent of all days of care in- and over. 1965______1966______1967-e.______1968.---_-______1969.- ______1970-m.______PROGRAM SPILLOVERS: THE POPULATION * Based on data for fiscal years UNDER AGE 65 ’ Not available * Based on sdmlsstons During the first 3 years of Medicare trends in hospital use by the population under age 65 seem care that, in the absence of the Medicare program, to have followed a pattern largely the reverse of would have been used by persons under age 65. that of persons covered under the program. Table Partial evidence that some displacement may 5 presents rates of change in hospital utilization have occurred is given in table 6, which shows the per 1,000 persons under age 65 for the period relative share of total admissions or discharges 1965-70. Days of care and discharges per 1,000 and days of care received by the nonaged popula- persons both declined steadily from 1965 to 1969, tion for the period 1963-71. The percentage of rising again at the end of the period. The trend both discharges and days of care received by per- in average length of stay, however, was roughly sons mlder age 65 declined substantially during similar over the entire period for both the Medi- this period to approximately four-fifths of all care and the non-Medicare age groups. discharges and two-thirds of total days of care. Since hospital utilization rates for persons As before, the causes of the redistribution of under age 65 were declining at the same time that hospital care from nonaged persons to aged per- utilization rates of aged persons were increasing, sons cannot be identified with certainty. It was the latter group may have received some hospital noted earlier that from its beginning in 1966 the hospital insurance program has generated a sub- TABLE B.-Annual percentage change in hospital utilization stantial increase in demand for hospital care by rates of persons under age 65 in short-stay hospitals, three sources, 1967-70 aged persons. This rise in demand more than offset the concurrent decline in demand by per- Percentage change from previous year sons under age 65. As a result, between 1965 and total Year and measure Health Hospital AHA 1970 demand as measured by per capita days Interview Hospital of care increased about 11 percent. During this Survey “SfYe Indicators time, the capacity of the hospital system to render Discharges per 1,CCOpopulation. inpatient care, as measured by available per 1966______’ -0 8 -2 0 1967______------’ -2 6 -4 4 q. capita bed supply, increased only 7 percent.6 1968------_---_------3 4 -2 3 1969______------'0 Thus, the demand for hospital care increased 1970______------ii : '66 Days of careper 1,ooOpopulation- more rapidly than the supply of beds, with the 1968.-_-- _-_ _-______- - ’ -5 2 -1 0 9 1967.------_------‘11 -2 9 I 9 1968______-6 -2 6 result that hospitals were used more intensively 1969____-__-__-_--_-______-0s 1970.----.---.------di -2 i 37 than before. This is evident in that the national Average length of stay (days) 29 average occupancy rate (percentage of beds occu- -i Y Iyo 26 -: t pied during an average day during the year) in ‘-16 -‘: 3 -: 3 ‘-1 li community hospitals increased from 76.0 percent

1 Based on data for f&al pears. e Percentage increases were estimated from AHA Guide ’ Not available. 1 Based on admissions Issue data for community hospitals.

BULLETIN, JULY 1972 9 in 1965 to an all-time high of 78.8 percent in 1969. this phenomenon alone cannot explain the decline Thus Medicare had the effect of increasing the in hospital utilization by persons under age 65. pressure on the capacity of the hospital system. Alternatively, increasing average length of stay AS a result, physicians may have become more and declining discharge rates could have resulted selective in the choice of patients to be admitied from the fact that the price of inpatient hospital to the hospital. If doctors did in fact alter the care has increased much more rapidly since 1965 case-mix of admissions to give priority to patients than per capita income. Between 1965 and 19’70 most in need of hospitalization, then average the “hospital daily service charge” component of length of stay might well be expected to increase the Consumer Price Index of the Bureau of even though the discharge rate might decline. Labor Statistics increased approximately 88 Cases that would have been admitted if beds were percent.8 During the same period, per capita dis- more abundant-less serious cases or those nor- posable money income increased only 37 percent.0 mally admitted for diagnostic tests-may have As prices have risen, more individuals may have been treated in some other setting. The rationing decided to avoid hospitalization unless it became hypothesis would explain the fact that average absolutely necessary. The result would be fewer length of stay for patients under age 65 increased admissions and longer stays because illnesses between 1965 and 1967 while their discharge rate would tend to be more serious when the patient declined. finally did seek treatment. Conversely, as the case-mix returned to normal Those persons in the population under age 65 after the initial period of the program and as the who have good health insurance coverage were bed supply began to catch up with the demand affected only indirectly-through higher premium for hospital care, the need for rationing may have costs-by rising prices for hospital care. A sizable decreased. This phenomenon might explain part proportion (lo-20 percent in 1969) of the non- of the decline over the last 2 years in average aged population, however, is completely unpro- length of stay for all age groups, since the less tected except through programs such as Medicaid seriously ill could again be admitted to hospitals. (the federally aided State medical assistance Other hypotheses could also explain the pattern program).10 Use of hospital services by these in- of change observed for the non-Medicare popula- dividuals may be particularly sensitive to rapidly tion. The decline in the birth rate between 1965 increasing prices. and 1968 may have contributed to the decline in The above hypothesis cannot explain the rever- the discharge rate and the increase in average sal in the trend of the discharge rate for persons length of stay for the population under 65. The under age 65 that occurred in 1969 and 1970. It is largest decline in births during this period possible that during the first few years the combi- occurred between 1965 and 1966 when approxi- nation of rationing and rising prices produced mately 150,000 fewer births were recorded than in the decline in utilization, since each would tend to the previous year.’ If the decline had not have a negative effect. Once the need for ration- occurred and each birth had been counted as an ing ended, however, the effect of rising prices admission, however, they would represent only 0.6 could have been offset by the release of excess percent of all admissions under age 65. Looked at demand previously held in check by rationing. another way, if the birth rate had not declined, The increase of the discharge rate, in other the admission rate in 1966 would have been less words, may represent the return of the demand than 1.0 per 1,000 higher. Had these admissions for hospital care by persons under age 65\to pre- occurred, they would have added approximately Medicare levels. 600,000 patient days-or only 0.4 percent of total days for persons under age 65-not enough to a This component measures the change in hospital change the average length of stay statistic. Since charges for room, board, and routine care in urban areas the birth rate declines in 196’7 and 1968 were of the United States. much smaller (85,000 and 20,000, respectively), ‘Economic Report of the President, 1971, table C-16, page 215. lo Marjorie Smith Mueller, “Private Health Insurance “‘Annual Summary for the United states,” fifonth@ in 1970: Ponulation Coverage, Enrollment, and Financial vital Btatistka Report, Pl’ational Center for Health Sta- Experience,” .S’ociaZ See&& Bulletin, , tistics, 1965-70. table 8.

10 SOCIAL SECUh SUMMARY AND CONCLUSIONS sons who either died in the hospital or were de- ceased after discharge but preceding the survey The data show that the hospital insurance pro- interview. The degree of understatement is fur- gram has had considerable impact on the hospital ther increased because of proxy response, response utilization rates of the aged population. After the error, and related interview survey problems. initial large increases betlveen 1965 and 1969, hos- pital utilization measured in days of care per The Hospital Discharge Survey.-The National 1,000 persons declined slightly, reflecting a sub- Center for Health Statistics also conducts the stantial decrease in average length of stay. This HDS, which provides statistics on the utilization slight decline in hospital use may represent a experience of the civilian noninstitutional popu- movement toward a new ecplilibrium trend. lation in short-stay hospitals. The HI program also appears to have had a Since 1965 data have been collected on demo- substantial indirect effect on the hospitalization graphic characteristics of patients, their diag- rates of nonaged persons. All available data indi- noses, discharges, total days of care, and average cate that a redistribution of hospital care between length of stay. Only noninstitutional general and those under age 65 and those aged 65 or older special hospitals with six or more beds and with occurred between 1965 and 1970. During this pe- average stays of less than 30 days are included in riod the proportion of total days of care received the survey. Military and Veterans Administration by aged persons increased from about one-fourth hospitals and hospital departments of long-term to roughly one-third, and the share received by and custodial institutions are excluded. Since the nonaged population declined similarly. 1967, all federally owned hospitals have been eliminated from the sample. HDS data are derived from a systematic sam- TECHNICAL NOTE ple of discharge records within about 400 hospi- tals selected from a stratified sample of the 6,965 hospitals in the master facility inventory. In any Sources of the Data given year the sample includes about 210,000 dis- The Health Interview Survey.-Conducted by charge abstracts drawn from hospitals of all sizes the National Center for Health Statistics, the in all regions. HIS series-“Current Estimates”-has reported data annually since 1963 by age and sex of re- The Co2zcmbiu l7niversity Survey.“-This spondent on the number of short-stay hospital dis- two-part survey of the initial effects of Medicare charges and average length of stay, and, since was conducted in early 1966 and late 196’7 by the 1966, on total days of care. School of Public Health and Administrative These data are derived from household inter- Medicine of Columbia University and by the Na- views of the civilian, noninstitutional population tional Opinion Research Center of the University of the United States living at the time of the of under contracts with the Social Secu- interview. The multistage probability sample is rity Administration. The CUS was concerned designed so that those interviewed each week in with the hospital utilization experience of aged approximately 800 households constitute a repre- persons in the 12 months preceding the week of sentative sample of the population. In any year the interviews. In both years data were collected the sample includes about 134,000 persons from by means of household interviews in an independ- 42,000 households. Those interviewed report on ent, multistage, stratified sample of 6,600 old-age their hospital utilization experience in any hospi- and survivors insurance beneficiaries. The data tal except those providing primarily chronic, tu- included the number of stays, days of care, and ’ berculosis, or psychiatric care during the g-month average length of stay by age, sex, race, and period before the week of the interview. income of respondent for all hospitals, extended- Since HIS data are based on information re- care facilities, and nursing homes. In the survey, ceived in household interviews, they represent a substantial understatement of the actual level of u Regina Lowenstein, “Early Effects of Medicare on Health Care of the Aged,” Social 8ecurity Bulletin, April hospital use. They exclude the experience of per- 1971.

BULLETIN, JULY 1972 11 TABLE A.-Health Interview Survey (Nationrl Center for Health Statistics): Hospital utilization experience and rates per 1,ooO persons, by age, in short-stay hospitals, 1963-70

Discharges Days of care Average;py$p of stay (in thousands) I (m thousands) I 8 Yl?W

Number

Fiscal year: 1963------__----_____------_- 1964~~~~~-~_-_--~-~______196L. ----_------______-______- 1964______-______1967------______-__-_-______-______Calendar year. 1967------_------__------_-_- _ 1963----_--_--_--_-______1969------____-_--_____------~---- 1970* ___---_____---______

Rates per 1,000persons ’ - Fiscal year: 1963_------_____-_-_____------124 170 11,043 1 912 1964------______-__---- ______-_-__ __ _ 123 :z ’ 910 1965------_____--______1;: :s 11,03711,085 1966_-----______:z 2 1’67___---______-______126 % 197177 :*::i, 911 Calendar year 1967.--..-.-..---...------123 115 197 1,044 878 2,655 -______.______1968------_ ------_____--__ -_____-___-_ 122 111 ifi: 1,1671,124 3,547 --_-_-_ - -- -_ - ___ -- -__-_-----_- 1969.______129 118 it:: 3.678 I --___-_---_---l----l ------__---- -_____--____ 1970’..-.--.-.-..-..------133 122 234 1,139 936 3,075 --_____-_-__-__----_-__- -______--__ - - 1Based on unpublished data r Based on noninstituttonal opulation ’ Unpublished and preliminary data Source National Center for R e&h Statistics, Series 10, various numbers. short-stay hospitals referred to all those that were mate actual hospital use, because they exclude the certified for participation in the Medicare pro- experience of decedents and are subject to re- gram and that had an average length of stay of sponse error, and related problems associated with less than 30 days. interview surveys. As with other household interview surveys, the The Hospital Indicators Xurvey.-The Ameri- data reported in the CUS substantially underesti- can Hospital Association in its journal, Hospitals,

TABLE B.-Hospital Discharge Survey (National Center for Health Statistics) : Hospital utilization experience and rates per 1,000 persons, by age, in short-stay hospitals, 1965-70

Number

1966’....--.---.--.------19651967’r __-__------_------__---__------___-_------%$I f%& gg 1.1 :@ $J ii :‘iB 11 1968’..----.-...--..-.-.------1969’ __----____------_---______1970r ______.______,

Rate per 1,000persons 0

1965______152 140 1,156 957 3,444 ______--._____-- ______--____ 1966______149 136 1,203 3.712 -_------___------__----- _------.--- 1967______145 E ;tJ 1,215 iii 4,086 ______1963__-----______144 301 1,214 897 4,272 ______1969______145 E iii 1,211 891 4,276 ______1970-_____-____-___-___------146 1,173 870 4,015 -__------___ --__-___-_-- -__------_--

1 Data from NCHS Vatal and Health SYat?st~c.v,Series 13, No. 6 4 Estimated from published data * Includes discharges with age not stated. 8 Unpubhshed revisional data a Data from Monlhl~ Vttal bfattstm Report, Nos 4 and 7, 19;1, and No 2, 8 Based on non Pnstitutional population. 1972

12 SOCIAL SECURITY TABLE C.-Hospital Indicators (Amencan Hospital Association): Hospital utilieatlon experience and rates per 1,000 persons, by age, in community hospitals, 1963-71

Number

21,840 5,208 214,454 148,536 21,960 6,505 221,891 148,798 22,123 6.904 227,633 149,585 23,110 6,137 231,643 153,070 23,966 6,346 234,413 155,475 196.991 I:,’ 201,266 $1 203.474 208,589 H 865 217,418 18 870 22:057 225,582 149:245 22,295 228,668 149,789 23,Qr) 156,876 23,833 154,110

Fiscal Year: 1967-.____-_-__- ______--__--______139 125 1968______124 1969______-_--____------2 1970______146 :i 1971______149 131 Calendar Year 1963______-______------139 1964______1966______1: 1966______1967______139 124 ri61Q 1968______------141 124 3:912 1969______124 1970______-______:tl? 132 1971______------148 130

1 Unpublished data ’ Based on estimated civilian resident population. s Estimated Bourcc Hospifals, Journal of the American Hospital ASSoCi8tlon,mid- * Not avaIlable month Issues

reports monthly utilization and financial statistics The AHA Guide Issue survey data are derived of its registered hospitals in a section titled “Hos- from a mail survey of all hospitals registered by pital Indica tars.” These data are derived from a the Association as of February of each year. Since stratified probability sample of over 800 non-Fed- it is a survey of the aggregate records of hospi- eral, short-term, general, and other special (com- tals, the data contain no demographic detail with munity) hospitals drawu from the total of more respect to patients. than 5,000 such hospitals registered by the Asso- ciation. Since 1963 the survey has included the TABLE D.-Guide Issue Survey (American Hospital Associ- number of admissions and average length of stay. ation): Hospital utilization experience and rate per 1,000 Beginning with July 1966, “Hospital Indicators” persons, by age, in community hospitals, 1963-70 has also been reporting admissions and average Admissions Days of care length of stay for patients aged 65 and over. Fiscal Year Ai,‘gY (ending September 30) T;‘“gp.$ R;t;r R;t&;er of stay (days) skds) perhons 1 parkons’ The Guide Issue Survey.-The American Hos- I I .I 1963_-______------193,600 pital Association has been reporting hospital util- 1964.....-...-.--..---- 2% 201,300 1965______------26:463 ization and financial statistics in the Guide Issue 1966______- ;yg 1967______---- edition of its journal, Hospitab, on August 1 each 196%______- 27: ?7! 1969______-- year beginning with 1947. The number of admis- 1970___*-*-_-_------3;;;. ’ 241,600 sions and average length of stay for patients of - 1 Based on civilian resident population of the United Gtates all ages by type, size, and regional location of the ’ Estimated. 8ourca Hospitala, Journal of the American Hospital Association, Guide hospital are included among the published data. Issue, August 1, various Years

13 BULLETIN, JULY 1972 TABLE E -Medicare Control Records (SocialSecurityAdmm- TABLE F .-Current Medlcare Survey (Social Secunty Ad- i&ration) : Hospital utlhzatlon expenence and rates per 1,000 rnmistratlon) . Hoqxtal utilization experience and rates per HI enrollees, in all hospitals, 1967-71 1,000 HI enrollees, in all hospitals, 1967-69

Admissmns I Covered davs of care Averaee Discbarges Fiscal year Year

1967_-_-______5,079 266 61,682 3,232 12 2 All hospitals 1968______5,771 297 75.400 3,374 I3 I _ 4,101 13 4 Fiscal year , 19691970______*-~--~~------XE 306307 %E 3,944 12 9 1967______. 1971s______61300 309 78:000 3,829 12 4 1968-____--__---___-. I I 19fi9__.______1 Eased on total HI enrollment as of January 1 each year * Based on bills, not discharges, and may be shghtly understated Short-stay hospitals a Adjusted for lags in clams processmg Source Howard West, “Five Years of MedIcare-A Statistical Review,” Calendar year So&Z Securtty Bulletma, ,table 7 1967______1968______------___-_ 1969v.______The Medicare Control Becords Data.-The hos- pital insurance program’s payment operations generate a variety of data on the utilization expe- CMS data are derived from a O.5-percent sys- rience of enrolled aged persons. Records of each tematic sample of admission notices received by hospital bill for which coverage is claimed and the Social Security Administration each month. admission notices are maintained on a centralized To obtain data on discharges, covered days of basis by the Social Security Administration. Data care, and average length of covered stay, these on hospital admissions, discharges, days of care, admissions are linked at a later date with corre- and average length of stay covered under the pro- sponding claims. If the corresponding claim has gram can be derived from these control records not been filed, information is obtained directly for each year since the beginning of the program. from the hospital. A supplementary sample of These data represent the hospital utilization of long-t’erm stays-a 0.5 percent systematic sample the enrolled population for which benefits were of all admission notices with 31 days elapsed be- paid. Thus total hospital use by aged persons is tween the date of admission and the date of selec- slightly understated because some aged persons tion or recorded discharge-is also drawn to are not eligible for hospital insurance benefits and reduce the samplin g variability in estimates of some of those who are eligible have exhausted covered days of care and average length of cov- their benefits for hospital care for the year, but ered stay. As a result the CMS sample includes are nonetheless hospitalized. In addition, data for about 42,000 admissions each year. recent years are incomplete because of lags in As mentioned before, data on covered days of filing and reporting of claims. Data for 1970 and care and average covered stay for 1968 and 1969 1971 have been estimated based on an adjustment are slightly understated because the definition of for these lags. the survey excludes the effect of the 60-day life- time reserve that became operative in January The Current Nedlcare Suwey.-The hospital 1968. Before this change enrollees had 90 days of insurance sample of the CMS conducted by the hospital care covered under the program in a ben- Office of Research and Statistics of the Social efit period. Any days used beyond 90 were not Security Administration has produced annual covered and therefore were not counted in the data on hospital utilization covered under Medi- sample. After , however, an enrollee care’s hospital insurance program for the period who used more than 90 days could receive as 1966-69. The number of admissions and dis- many as an additional 60 days of covered care charges, their covered days of care and average before exhausting his benefits for that period. In length of covered stay by age, race, sex, and resi- the CMS sample no covered days of care beyond dence are included among the data. Utilization 90 were counted. Potentially, the CMS data could data for all hospitals and short-stay hospitals are underestimate total covered days of care by as available on either a fiscal-year or calendar-year much as one million days and average covered basis. stay by as much as 0.2 days.

14 SOCIAL SECURITY TABLE G.-Population estimates used in calculation of utlhzatlon rates per 1,000 persons for each survey in this report, including fiscal and calendar years, by age, 1963-71 [in thousands]

National Center for Health Statistics American Hospital Association Social Security I I Administration Culde Year Health Interview Survey r Hospital Discharge Survey f Hospital Indicators Survey 8 Issue Medicare Current Survey 1 Control Medica?

B”,‘e$ 1 II&d;; 1 Ax%;,“, v v -&- a!?%% a!i+?v%

Fiscal year

183,146 166,280 16,866 (7) (7) (3 (9 (7 (9 (;, 0 ‘;’ 185,797188,430 168,775171.138 17,02217,292 I:; $1 3 [:I I:] [:I # I:; i:\ 1oo.710192,359 173,132174;4”4 17,57817,865 11 $1 8 lQ?275 17s]367 lk)QG7 ii’, 10940 10,071 $1 id 1:; $1 11 196:198.204272 176:178,704QGO 19:19,651340 19:19,587310 19,73319,423 2OG.391 180,275 19,458 ii] (7 203,046 182,305 29,386 0 y3; $1 - Calendar year

1963______(‘1 1964------.-.------1965______Fl 1966------_----__-_----______1967______1$)374 1968--______177:051 1969______178,764 1970______180,846 1971--______(3

1 Civilian noninstitutional population of the United States at midyear number of persons in this age group, enrollment was not used to calculate from NCHS VttaZ and Health StaWrca, Serie? 10, vmous issues population under 65 because it was believed that understatement of this * Civdian noninstitutional population of the United States at July 1 from group would result NCHS Vzlal nnd Health Stafialrcs, Series 13,various issues Population data 4 Cwihan resident population at January 1 and July 1 from Bureau of the for the period 1965-60were projected from the 1960Census, data for 1970 Census, Series P-25, No 478 based on the 1970Census Age distribution on fiscal-year basis not available 6 Persons enrolled under the III portion of Medicare as of January 1 from a Estimates for subgroups do not add to total Population aged 65 and over Social Security Administration records is taken from data for enrollees under HI and/or SMI portions of Medicare, 6 Average number of persons enrolled under the HI portion of the Medi- as of January 1 and July 1 Population under age 65 is calculated from popu- care program durmg each ll-month period from Social Security Adminis- lation of all ages, less unpubhshed estimates of population aged 65 and over tration records as of January 1 and July I Medicare enrollment was used for population r Not available aged 65 and over because Census estimates are beheved to understate the 8 Unpubhshed estimates of civilian noninstitutional population as of July 1.

BULLETIN, JULY 1972 15