An empirical analysis of continuing medical care utilization in a health maintenance organization

An empirical analysis of continuing medical care utilization in a health maintenance organization

care de&ions) make it likciy that diEbrent sets d ~~rm~amts or at least di~~t same d~~in~ts WiR were s~nifi~tly rcla to the likc%ood of utiliainp pre...

1MB Sizes 10 Downloads 61 Views

care de&ions) make it likciy that diEbrent sets d ~~rm~amts or at least di~~t same d~~in~ts WiR

were s~nifi~tly rcla to the likc%ood of utiliainp preventive and na~avGntive doctor visits. For copies ~u~t~a and ittcr.uq had saint positive effects on the ~ike~i~ of using ~~~atjve ser-

nology and the resulting dominant role of the provider, we expect that the e&t of patient sociai and economic factors on continuing mcdiii care utiiiaa~ tion will be mitigated. Normatively, this would be 4 pixitive res.tuirii society pr&rs to diitribute medicai services. once contact is initiated. on the basis of provider evaluation rather than on the basis of [email protected] power and other tiaf chara~t~~t~~

Ecotmmic ~~~j~s @xnne, prices. time) on the patients’ ~~h~~ power are ~t~nt~~iy important datcrminaats of the drmand for continuing medical care. just as they arc important in the demand fat

_*\I--- ___A_ _-A a_-.x__, XI..-.__.- -L..-l_:--_ ..A1 OulGr tJolJosunu IICMOCS.numvcr, pnysmGiansWIH have a major influenoe over ~n~mption d~~io~

and may scrcca out some t#tcts of these types af var~bl~ on deal. On the other hand he may con161

162

DAVID

R.

LAIRSON. JOSFH

sider them carefully in his recommendation for one more return visit or admission to the hospital, etc. In addition, so&o-demographic factors such as education, age. sex, and health status are potentially important determinants of the demand for continuing care. Education is an environmental factor which may alter an individual’s ability to produce his own health through the use of formal medical care and by other means. Age, sex, and health status have implications for the type and amount of medical care which is appropriate in alternative health situations. Relatively old persons and persons in general poor health status may require more intensive care for the same health condition because of body deterioration and resulting difficulties of recovery. Women often require services of female medical care specialists such as obstetricians and gynecologists. A general model of the demand for continuing medical care can be written as follows:

Knwov and

JAMES R. MARSDEN

The data available to us from the Kaiser system suffer because we do not have data on physician training and work experience. key variables in our conceptual model. However, we do have data to represent ail other parts of the conceptual model (see Table 1 for a description of the data). Economic

variables

As in classical econoniic theory, it is hypothesized that demand increases with income and that quantity demanded is inversely related to price. Medical care and prepayment in particular present special problems for determining the effect of price and income on demand. Prepayment is an insurance mechanism whereby one pays in advance for the future right to use service. This arrangement does not mean, however, that there are no costs to the patient at the time service is demanded. Prepaid plans include some nominal charges at the time of service. e.g. visit fees and partial payment for diagnostic services. and the CON, - F( Econ,, Sot Demt, Diag,, Phys,) ( I) time and transportation costs are also present. InThe model states that the demand for continuing come is therefore a potentially important determinant medical care of type j is some function of patient of demand. economic characteristics i so&demographic characAnnual family income (FAMINC) represents a teristics k, the diagnosis of morbidity 1, and physician resource constraint on the consumption 6pportunities training or experience proxy m. of the family. Consumption opportunities are parThis general model will be operationalized using tially determined by family income, family size, and the prices of goods and sewias. The higher the consumer behavior observations from a large prepaid medical care program. Availability of the desired inincome the more goods uld services one can puiformation on a large population made this setting chase; consequently it is expected that family income will be directly related to the demand for continuing useful, though the specificity of the program rgtricts the generality of the results. medical m given prices and family size. Employment status (EMPST) is included because individuals not employed in market work generally have less severe time constraints on their use of mediDATA AND DEFINITION OF VARIABLES cal care. In addition, persons not employed have no Our estimates of the demand for continuing medi- loss of money wages when time is used to consume medical care. cal care are based upon economii and sociodemoThe price of a doctor oiiicc visit (PRiCj represents graphic data obtained from a househoid interview the out-o&pocket charge levied at the time of service. survey conducted during 1969-70 and corresponding Depending on the specifications of their group policy, medical care utilization and morbidity data obtained members either paid zero, one, or two dollars for an from medical records of a 5% sample of the members of the Kaiser Foundation Medical Care Program in office visit. Although everyone is a member of the same prepaid group practice. and has the same basic Portland, Oregon. A partnership of approximately 200 physicians and a complementary medical staff insurance coverage and the same facilities and services available to them, there are some charges at provide prepaid comprehensive services to approxithe point of service for some members. mately 200,000 persons in the metropolitan Portland I. ?- expewca -..-__~-~ tnar -L_* prw -_.__ ana __ quanrlry _..__~.~.. aemanaea _--- _ IC IS area wtihm a highiy organned hospitai-amoularory care medical system [S]. of return visits will be inversely related. This relationThe present analysis is based on the behavior of ship is expected to prevail in each office visit equation, but there is no reason to expect the magnitude subscribers and spouses who were in the household interview survey and who made at least one initial of the price coeillcients to be the same in each equacontact with the medical care system in 1970. In- tion. The coinsurance rate for diagnostic tests (CRDT) cluded are 797 males and 1,021 females [6]. Several measures of continuing medical care are used as is the proportion of the charge for outpatient lab test and X-rays required of Kaiser members for use of dependent variables. The measures are not intended to be exhaustive. but to represent a range of different those services. Kaiser bases the charge on prevailing types of continuing medical care. The different types community rates. Zero and one-half were the only coinsurance proportions in effect at the time of this of care imply more or less control by the physician study [8]. The coinsurance factor will be included more or less discretion on the part of the patient. and more or less time and money costs to the patient. in the return visit equations, as well as the diagnostic Measures of coniinuing care are return visits to inter- test equation because of a possible complementary relationship between office visits, and lab and X-ray nists (RI). return visits to other specialists (RS), return services. Some members may avoid visits to the physvisits to obstetrics-gynecology (RG), use of outpatient ician because of the possibility of incurring lab and/or X-ray and lab diagnostic tests (DT). and hospital X-ray charges. admissions (HA) [7].

0 1

Dummy

Employment status

Employed Not employed

E lL500 17,500 25,000

1250 3000 4250 5750


w&7499 75009999 lO,OaI-14,999 15.oaL19.9999 > 20,aIo

Midpoints

InCOllU

SO I 2

Prii

Actual

Actual

Coinsurance

0 0.5

Measurement

Variabk

423 5%

:; 64 172 336 151 75

:; 44 151 276 143 59

631 166

62 41

;:

76

27 21

5:: 163

739 282

Distribution Female Mak

601 1%

. -

18.0

110

4p

58

65

161

149

14.0 16.0

35 75 174 453

3:

685 198 95

192 206 I95 183 147 98

832 189

44 85 110 255

538 153 75 6 25

111 179 157 153 119 78

651 146

Distribution Mak Female

3.5 8.0 IO5 12.0

Midpoints

Education O-7 Years 8years 9-11 years 12-High school grad. 13-15 &une colkge 16-Cdlcge graduate >17 years

7.5 22.5 37.5 52.5 67.5

Midpoints

Actual

0 1

Dummy

Measurement

< I5 Minutes l6-29. 30-45 46-59 >60

Time to Clink

Age

Excelknt/Good Fair/Poor

Health Status

Variabk

Family size

g I5 Minutes 1629 30-45 46-59 w

Time to Hospital

Variable

Table 1. Specifkations oi explanatory variabks and popuiation distributions

-__

Actual

7.5 22.5 37.5 52.5 67.5

Midpoints

Measurement

I

1:: 95 54 24 7 4

109 274

205 309 217 27 39

z 146 152 109 58 24 9 5 1

277 374 280 30 60

Distribution Male Female

164

thVl0

R. LAUHON,~YEPH

KRWOV

Uncertainty prevails for patients who pay for lab and X-ray servkea. Persons seeking preventive care. having chronic illness, or wha have an acute condition such as a sprained a&e, may expect to receive specifk lab or X-ray services while others have a vaguer pnrcqtion of the services to be tetzommended. fn either casa. the totaI cost of lab and X-ray servkxs is unlikely to be known b&ore the visit. The eoinsuranee variable is considered to be another price laetor and is expeetcd to be inversely related to return visits and diignostic tests. Travel time to the c&tic (TRAVC) and to the hospital (TRAVH) refer to the number of minutes it takes on the averaga to travel from o&s place of msidence to the nearest Kaiser Clinic and to the Kaiser hospital, respectively. Tii is thus viewed as a general resource constraint and time to the clinic and to the hospital represent part of the price of health sew& consumption. Similar to money prioe, time is expectad to be inversely relatad to sarvice demand. Family siza (FAMSIZ) is the number of persons in the individual’s fhmily, For a given fpmily income, mote family members rcduee per capita income which may reduce the demand for madii care for a given individual in the family. Also, larger families may ptoduce care at home rather than take time from household aetivitics to utilixe formal medical care. Family size is tWefote expacted to be inversely related to the demand for continuing medical care.

aad

JAMESR.

MARSDEN

cai services are needed to minimixe the deterioration. The sample popuiation was limited to nondependents; the youngest being 20 years old. Age-squared (AGE**21 is inchrded to measure the expected non&tear effects of age on co~inuing care. That ia younger persons in the s8mpk especially women, are initially expected to have deelining rates of utR~tion as they age followed by rising rates atter they pass a certain stage of the lik cycle. Health status (HS) represents the patient’s perceived general state of health. Respondents were asked whether they [email protected] t&dr health was excellent, good, fair or poor. It is expected that those persons who perceive themselves to be in relatively poor health status will demand more medical care than persons who perceive themselves to be in relatively good health status.

Diagnnsric variables Determination of the patient’s problem is necessary for the physician to determine appropriate alternative treatments. Additionally, morbidity may in&nce to what degree one’s economic and social constraints arc binding [lo]. For example, lab fbes may influence the number of times preventive visits are made during a period, but have no influence on return visits to monitor a perceived satious chronic diseam. It is therefore impottarit to attempt to control for morbidity while examining the effect of other factors on demand. Since we arc intefcstcd in continuing medical care ~~~~~~~~~~~~ decisions which am heavily inBuaneed by the phys~at~~~~~~~~~~~~ ~~~~~~p~n~~~~i~oo utiliition in seve& studies [9]. A~~~ their the initial visit. The preacntit~gdiagno& was initiaRy ~~on~~~~~~~ coded by the International Cla&&ation of Disease ~~~~~~~~~ Adapted (ICDA) c&us. For the purpo# of simplicity on return visits have not baan [email protected] the codes were aggregated into eight groups: PrevenEducation (EDUC) mpreaents the number of years tive Care, Pregaancg, Trauma and five Diacase Cate.,.c or iormai sehooii attained by the individuai. “i%uoa- gories. Tii ia a rot&t& version of the ten-way *behavtion may produce several eftbcts on the demand for ioral clam&adon system dmd~peci by Hurtado and continuing medical care. Once morbidity is perceived Greenliok spac&aily for the study of medieai care mote educated persons may be more likely to comut%xation Cll]. The baxic structure of that system is bine formnl medical cart with other measures (diet preacnted in Pigurc 1. Preventive Care (PREV) represents the number of and rest) to regain their health. Ia addition, more highly educated persons may be better quippad to initial contacts for preventive medical care made by deal with a complex madical care systam in order a member during the study year. Tha major prevento obtain tha continuing medical care desired. Educative services indude general medical examinations, r:** qJe&_~__^._rl ia? k &mi:ji -A-r-A a.._ sramulurI”n;r, -..~-:-~r:~---A :--..,L*r:,.. --.L-Ll”Kl *L-IL”lllCi,Cil”Ili is I1I11,U &^ 1” GJ’5 al,” ImWIuBU*LLl”,” su “ILiSJ.I. 1c :1s _^. LL”l the demand for continuing medical cate, given mot- clear what effect the use of preventive medical care will have on health and the demand for continuing bidity and other patient characteristics. Sex (SEX) is entered indirectly as a variable in this care. Immunization services may reduce the risk of illness and indirectly reduce the demand for curative study by estimating separate equations for males and females. Biological and social difkrences between men continuing c&e. Use of preventive care may also ret&t a person’s general concern about health matters and women rasuit in differences in their respective patterns of medical care utilization. This fact will be and be an indicator of persons who arc healthier refkcted by including an quation for return visits hceause they work at producing their own health. In both cases we would expect an inverse relationship to obstetrics-gynazology by females and by including an explanatory vatiabk to teiket whether or not between PREV and the demand for return visits and hospital admissions, with a Positive relationship there has hcen a diagnosis of pregnancy or compliesbetween PREV and diagnostic tents because of their tion of pregnancy. cd3mplementary nature. Conversely, use of preventive Age {AGE) is measuredin years for each individual. Age is generally assumad to at’iect %eed” for medical care may uncover problems that require treat~nt continuing medical care. In that case, care. At very early stages in the life cycle the body with agate we would .expect a positive relationship between requires protection from the env~on~t. Child-bearing is Possibk only during a certain spao of the fife PREV and the demand for continuing medical care Disease (DiS), Pregnancy (PREG) and Trauma cyck. One’s body dateriorates in later years and medi-

t

f

I

DAVID R. LAIRSON. JOSEPHKRISLOV and JAMESR. MARSDEN

166

Table 2. Demand functions. male non-dependents

kS

RI

Variable

PRIC CRDT TRAVC

Equations (N = 797) DT Coeff. Coeff.

HA Coeff.

COeff.

Coeff.

X lo3

X 105

x 105

x 10s

X 10’

x 105

X 105

t

t

t

t

t

t

t

t

12,398 0.70 - 1897 0.05 - 286 0.63

17.743 1.03 - 10.371 6.26 -385 0.86

- 9693 0.48

- 2654 0.05 -306 0.58

-8340 0.41 - 6269 0.13 -352 0.67

56,071 0.66 -403 0.31

EMPST EDUC FAMSIZ AGE AGE-2

HS PREV TRAUM DIS

-3 2.27 57,202 2.37 2700 1.29 -6379 1.55 5961 2.03 -47 1.54 78,552 4.59 -196 0.02 7940 0.74 41,086 7.34

- 121.234 1.66 1.03

19,117 2.13 50,786 4.75 33,027 2.91 142,334 5.35 171.368 3.58 -97.531 1.36 1.03

0.18

0.22

ACUTE CHRON SYMPT EMOT HOSP Constant Dep Vrble Mean R2

-2 2.23 58,656 2.49 2123 1.04 -6050 1.51 5158 1.80 -41 1.38 75,339 4.50 708 0.07 7799 0.74

-3 2.03 23.684 0.85 1937 0.80 - 5317 1.12 5542

1.63 -47 1.32 71,902 3.63 5772 0.52 42,488 3.43 24.906 3.84

-2 1.91 24,459 0.88 1778 0.74 - 5335 1.12 5200 1.53

. ;; 70,019 3.54 6967 0.62 42,516 3.44

-6 1.80 171,437 2.51 5746 0.98 - 11,747 1.01 16.504 1.99 -144 1.64 66.459 1.37 275,571 10.08 27,798 0.92 101.989 6.42

-6 1.71 170.948 2.51 4844 0.82 - i 1.280 0.97 15,442 1.86 -134 1.52 68.387 1.41 274,066 9.97 24.739 0.82

- 59,445 0.70 1.01

10,160 0.96 49.400 3.91 9170 0.68 58,547 1.87 86,345 2.38 - 49,278 0.58 1.01

- 240.501 1.19 4.72

77.725 2.99 117,537 3.80 82.090 2.50 320.230 4.17 125.441 1.41 -206.174 1.02 4.72

0.09

0.10

0.19

0.20

obtain continuing care. A national study by Phelps [12] of users of office visits suggests that coinsurance has a significant negative effect on the demand for office visits, but it is not possible .to tell the effect of coinsurance on decisions to initiate care relative to decisions to continue care. One possibility is that the income effect is negative because lower income males bring more serious ailments to the system initially and are therefore called back more often for return visits. Whatever the reason, prepayment appears to reverse the traditional effect of income on the demand for medical care for males. This is consistent with the Phelps [I33 finding of a slight negative relationship between income and visit demand. Acton’s [14] study of users of New York City outpatient departments indicated that travel time. represented by distance, may be an important negative :‘ctcrminant of the demand for medical care when

Coefl.

60,746 0.72 -679 0.52

TRAVH FAMINC

CoeK

COeff. x 10’

-164 0.67 -2 2.49 la.542 1.28 680 0.55 -3151 1.27 2118 1.20 -17 0.92 33.732 3.27 - 10,755 I.85 - 5002 0.78 10,104 2.99

- 147 0.62 -2 2.63 20,227 1.46 452 0.38 - 3071 1.29 1659 0.98 -14 0.82 29.148 2.95 - 7854 1.40 - 3394 0.55 -1460 0.28 16.832 2.67 1652 0.25 31,078

1.99 - 17,948 0.42 0.23

157.795 8.72 - 2383 0.05 0.23 0.17

money prices are low. Our analyses support this hypothesis for decisions to initiate care but not for decisions regarding return visits. The Phelps [lS] study showed a positive but insignificant travel time effect on office visit demand. EDUC has the expected positive signs and FAMSIZ has the expected negative signs. however both coefficients were quite small relative to their standard errors. This is again in marked contrast to our earlier findings that education had a very significant positive association with the probability of using preventive care. and family size was inversely related to the probability of acute visits. Grossman [16] has suggested that education is an environment factor which increases an individual’s ability to produce his own health and therefore will reduce his demand for medical care (if the price elasticity of demand for health is less than one). His empirical results did not support

Continuing medicai care utilization in a health maintenanceorganitation that hypothesis, nor did those of Phelps [lTJ. Our results do not support the hypothesis of a negative education effect. Age is directly related to return visits to internists by males. with a slight negative influence after the age of 63, as implied by the negative AC3W.2 coefhcient. The age co&cknts in the fti equations were not significant. HS appears to be a aignifkant and positive determinant of continuing care demand. The positive HS coefficients imply that those reporting to be in generally poor health use more return visits than those reporting to be in generaRy good health. DiS (Di) was an important diagnostic variable determinant of return visits to internists. For both males and femaks, the results indicated that for every

167

initial contact for disease, one could expect over 0.4 return visits to internists. Although each separate disease variable in the disaggregated RI equation had a signifkant positive coefficient, HOSP had the greatest effect, with each initial visit for disaase generally requiring hospitalization generating 1.7 return visits by males and 2.6 return visits by femaks. Initial contacts for prevention, pregnancy and trauma bad insignitkant effects on return visits to internists. Continuing aisits to speci&sts Only 10% of the variance in return visits to specialists was explained by the regression equations. Some reduction in explained variance was expected because of the heterogeneity of physicians in the group as well as the patient’s diverse health conditions [18]. The

Tabk 3. Demand functions, kmaie naMependants RS

RI Variable

coeff. x 10s t

PRIG

CRDT TRAVC

- 34,612 0.75 327 0.66

Coeff. xlos t

ckeff. x 10” t .

Equations (n = 1021) M RG coeff. coaff. cceff. coeff. coeff. x IO’ x 105 x lo* x 10’ x 105 t r t t t

1914 - t 3,743 -15,560 ‘. .n n*” S.-o it.11 u.aa U.IU 28,470 27,559 .3Q407 0.67 0.63 0.66 467 1.00 05 0%

-884 A”1

v.

14

- 7967 0.27 241 0.77

EMPST 987

EDUC FAMSIZ 0.27

AGE

2506

AGE’*2

0.87 -9.1 0.01

HS PREV TRAUM PREG DIS

110,586 5.87 -5186 0.55 1279 0.10 - 5024 0.22 52539 9.70

ACUTE

CHRON SYMPT EMOT HOSP Constant Eznlk R2

0.37 -716 0.14 1671 0.56

0.2: 101,469 5.39 -3448 0.36

1 0.98 -3640 A-.4 “.8W no9 0.67

mos 1.47 3144 1.11

-13 0.46 52707 2.95 10,638 1.18

1

1.02 -4571 @.3rJ 1027 0.40 7338 2z 0.93 0:; 47,639 2.65 12643 1.40

287 0.17 -4W 1.51 - 4745 2.50 30

kii 0.31

-3971

- 3971 0.31 -7874 0.35

k!! 58ii 56*5 4.57 0.15 -2851 - 3626 331,927

39,948 4.28 54,547 4.94 50,356

25.688

0.13 30.607 j.95

0.17

0.19

23.08 3713 1.08

i.88’

43,468 4.12 21352 i.92 11,311 86$z 0.44 3.25 160,146 261,484 3.58 -1isJB5 -82g - 136.621 -113.075 i.80 1.48 1.44 1.M 0.96 0.96 1.30 1.30 0.17

X.42 3466 034

0.09

0.10

rrn.

V.W

-8143 0.28

80,078 0.50 0.::

1

1.37 3523 035

0.4; 164,096 2.G? io7 11,318 0.84 0.24 1371 -4858 1.52 0.05 32.266 -4667 2.16 2.45 -305 30 1.95 1.51 4474 400,767 4.25 0.37 138,567 -4359 2.91 0.71 2078 143,841 0.25 2.22 332166 572,186 23.00 5.05 153,363 5.64 3884

0.65 3124 0.44 4575 0.61

72081 0.45 -6 0.01

0.4; 164354 2.02 8292 0.61 2585 0.10 29,979 -2&y 380$ 4.01 145,468 3.04 129,130 1.98 564,637 4.98 124.023 i.63 194.750

0.44

-230 1.14

- 192 O.%

0:: 7633 -:6?5

:: 6281 086

0.09

0.10

-ii68

0.55 -913 0.39 -2489 1.80

0.95 - 536 0.23 - 2874 2.09

2.z 19.045

2.; 15,440

i.18 - 5886 _!g

0.08 60,216 5.74 10,167 4.03

1.77 -4458 1.02 ‘-992 0.17 59,m7 5.75 4058

68,445 1.88 0.29

27:; 5.36 1274 0.24 -6232 0.50 85,716 3.94 85,852 2.37 0.29

0.08

0.10

i.49

i3&641 2.33 56,925 10.738 0.63 0.42 764,756 - 25,876 3.23 0.86 176,335 17z608 - 794,541 - 692997 2.00 1.74 3.36 3.47 7.02 7.02 0.66 0.66 0.44

CoeK x 105 t

-534

TRAVH FAMING

HA c&f. x 10’ t

168

DAVID R. LAIRSON, JOSEPHKRISLOV and JAMESR. MARSDEN

priceand travel time variables appeared to be rcla-

the coe&ients

tively unimportant, while income was inversely related to the dependent variabk for males and positive but insignilicant for females. Employment status did not appear to be significant in either equation. EDUC (Education) posses& positive but small coetRcients relative to their standard errors. Family size was negative for males but positive for females. There is no apparent reason for the opposite signs. Age, AGE**2 and HS exhibited the same patterns as in the internists return visit equations. TRAUM (Trauma) is an important diagnostic determinant of return visits to specialists with every initial contact for trauma expected to result in over 0.4 return visits to specialists by both males and females. This is not surprising since most trauma cases involve lacerations or broken bones which often require the services of surgeons. The DIS coeglclents were also positive and quite significant. The disease breakdown again indicated the relative importance of the HOSP variable with each initial visit for disease generally requiring hospitalization generating 0.86 and 1.6 return visits to specialists by males and females respectively. Initial contacts for prevention and pregnancy did not appear to significantly alTect return visits to specialists.

CW0l-S.

Continuing visits to obstetrictans-gynecologists Forty-four per cent of the variance in return visits to obstetrics-gynecology by females was explained with the regression model. The only variables that appeared to be important were age and initial contacts for pregnancy and complications of pregnancy. The AGE and AGE**2 co&kients imply that utilixation first declines with age and then increases after the age of 78. As might be expected, the most sign& cant variable by far was PREG. That coelRcient indicated an average 3.3 return visits to obstctrics-gynecology would be made for every initial contact for pregnancy or complications of pregnancy. The relatively high explained variance and the unimportance of factors other than those associated with the occurrence of health conditions treated by obstetrics-gynecology suggest that Kaiser members receive quite similar continuing care for those conditions regardless of their economic and social charac_._?.~1__ teristics. Diagnostic tests Diagnostic variables are by far the most important correlates of diagnostic test utilization. EMOT had the highest coefRcient (B = 3.20) in the male equation while HOSP had the highest coel5cient (f? = 7.65) in the female quation These results merely support the fact that return visits and diagnostic tests are used together in the diagnosis and treatment of episodes of illness brought to the attention of the medical care system. Family income was inversely related to diagnostic test utilization by maka but unrelated fop tamales. Employment status bad a positive co&c&t in both group% suggesting that persons not employed use more diagnostic suvias tllan persons employed. As expected, education was directly related to the use of diagnostic teats for both males and females, but

were low relative to their standard

Hospital admissions The rearession analysis explained 17 and loo/, of the vari&on in male and female hospital admissions, reJpoctiwly. Income was sign&ant and inversely related to the depe&em variable in the male equation. This finding again suggests the hypotbmis that lower income males are more ill when they contact the system and before tend to use it more than higher income males. A strict economic interpretation would be that higher income persons have a higher opportunity cost of time which would induce less reet=o&;ime intensive in-patient c8n, everything Our finding differs from those found in some earlier studies on the demand for hospital services. Gerald Rosenthal 1193 concluded, in a national study of hospital utilization that low income appeared to constrain the use of hospital services while relatively high income did not appear to influence hospital utilization. Similarly, Newhouse and Phelps [20] estimated positive income elasticities for both hospital length of stay and admissions from a data set drawn from a 1963 national sample. In contrast, Phelps [21] estimated a negative income elasticity for length of stay from national data collected in 1970. Data used in the former two studies were collected before the development of public insurance programs in the United States while the latter was conducted several years after the enactment of Medicare and Medicaid. Thus, the usual effect of ‘income may be reversed when hospital care is extensively covered by insurance. Acton [22] found that both earned and nonearned income were inversely related to hospital admissions in his study of users of New York City’s pubiic outpatient department, and he concluded therefore that hospitalization is an inferior good. However, negative income coefficients do not by themselves imply inferiority without controlling for morbidity and the possible use of other non-public hospitals, or the substitution of less time intensive outpatient treatment by higher income individuals. Other factors which appear to be important are r-n . PV? PIlO --A nulz. #Vm’L*,r__ lcula,cs L___,__ onlyh _-,..\ l-iXC”, nnr., r-l=, nur: -L (LOT PREG, and several disease variables. As expected, the health status measure was directly related to hospital admissions.:Interestingly, the male HS coefficient is about W/, greater than the female HS coefRcient. For females, hospital admissions decline with age up to 43 years old and then increase with age. Although not highly significant, initial contacts for preventive care were inversely related to hospital admissions for both males and females. One interpretation would be that preventive medical care reduces hospital utilization by protecting health. Another interpretation would be that those persons who USCpreventive care also probably do other things (such as exercise, watch their diets, and refrain from excessive smoking and drinking, etc.) which improve their health and reduce the risk of the need for hospital care. This question merits further research to determine the efficacy of preventive medical care and the

Continuing medical care utilization in a health maintenance organization

link between prevention and therapeutic medical care utilization. DIS and PREG were both directly related to hospital admissions as expected. The disease coefficient was 0.10 for males and females while the pregnancy and complications of pregnancy coefficient for females was 0.60. When DIS is disaggregated, the largest coeficients for males are EMOT (B = 0.31) and HOSP (B = 1.6) compared to females where the largest disease coefficients are CHRON (B = 0.27) and HOSP (B = 0.86). Thus, when pregnancy is controlled, the diagnosis of a disease generally requiring hospitalization is much more likely to result in hospitalization for males than for femaies. This is consistent with the hypothesis that males let their health deteriorate more than females before contacting the medica! care system. SUMMARY

AND CONCLUSIONS

A different set of economic determinants was obtained for continuing care demand than was obtained in our earlier study of the demand to initiate preventive and nonpreventive medical care. In that study the coinsurance rate for diagnostic tests and travel time to the nearest clinic both appeared to be inversely related to the likelihood of initiating medical care, while the income and employment status coeflicients were posifive. In contrast, coinsurance and travel time appear to be unimportant while family income is an important determinant of adult male demand for continuing visits to internists and hospital admissions, with family income having a -negative effect on demand. Employment status was directly related to return visits to internists by males and the use of diagnostic tests by both males and females. Two explanations of the income result are plausible; one is economic while the other is medical. One would expect an inverse relationship between income and the demand for continuing care on an opportunity cost basis. If the consumption of care resulted in a loss of income because of work loss. higher income individuals lose more per unit of time, everything else equa!. The fact that the income relationship held only for males who are more likely to be the main wege earner, and the fact that the family income coefficient was most significant in the relatively time intensive hospital admissions equation, supports the opportunity cost hypothesis. Another possible explanation is that low income males let their health deteriorate further than higher income males and therefore require more intensive medical treatment once they finally initiate service. The relative impor-.. tance of the relationship between income and hospital use (the most intensive mode of treatment) also supports this hypothesis. In addition, the low income household may be viewed as a less suitable environment for the care of a sick person and therefore differential treatment would be advised for low and high income individuals with the same condition. For whatever reason, it does not appear that the purchasing power effect of income operates on continuing care in this prepaid group practice. This conclusion is tempered somewhat by the fact that utilization was measured by quantities of visits and admissions rather than value of service.

169

In most equations education and family size had the expected signs, however. they did not appear to be highly significant determinants of continuing care demand. Health status possessed the expected positive signs and appeared to be significani in most equations. The diagnostic variable results were as expected with highly significant positive coefficients in most equations. An interesting result was the negative coefficient on the PREV (initial contacts for preventive service) variable. obtained in the hospital admissions quation for both males and females. This result should be investigated further to determine whether preventive care actual!y promotes good health or that persons who use preventive care tend to be healthier because of their genera! behavior. The amount of variance of the utilization variables explained by the, regression models ranged from 0.08 in the female hospital admissions equation to 0.44 in the female obstetrics-gynecology visit equation. In general, this does not represent a large improvement over the power of other models that do not include diagnostic variables. Possibly, the categories used here are too heterogeneous or the model is misspecified because of the absence of direct physician variables. Future attempts to study the demand for continuing care should control for diagnosis to the degree possible, include variables that may influence physician behavior, including the method by which they are reimbursed for service (fee for service or prepaid by salary or capitation) length of time in practice, place of training, and the capacity at which the practice is operating. Given the range of possible treatments, these factors may substantially contribute to explaining the variation in utilization by patients. authors would like to thank M. R..Greenlick. C. Pope, P. Lairson and other investigators and staff at the Kaiser Health Services Research Center in Portland. Oregon. for their assistance during the earlier stages of this research. Ackno~ledyement-The

REFERENCES

1. Lairson D. R. and Swint J. M. Estimates of preventive ‘versus nonpreventive medical care demand in an HMO. Hlth SW. Res. Spring, 1, 1979. 2. See Grossman M. The D&for Health: A Theoretico/ and Empiriccrl Inrusrigation. National Bureau of Economic Research, New York, 1972; for a develop ment and test of the human capital approach to the demand for medical care. 3. See Feldstein P. Research on the demand for health services. Milbank meml Fund q. 44, 143, 1966; for an expanded discussion of a similar view of medical care demand. .4. For elaboration on this point see Arrow K. J. Uncertainty and the welfare economics of medical care. Am. l&n. Rev. 53,941, 1%3. 5. .T!Cs study setting and sample population have been described in Greenlick M. R. et nl. Determinants of medica! care utilization. Hlth Sew. Res. 3, 2%. 1968. 6. The entire sample poputation included 3918 persons of whom 2390 were subscribers or thci spouses. During 1970. approximately 76% of that adult population made at least one contact with the medical care sys-

tem.

170

I.&VW R. L.A~

JOSZPHKtttstaov and MSES R. M*msom