Documentos de Trabajo

Transcription

1 Documentos de Tabajo Medical behavio: an application to cesaean section delivey in Uuguay Leonel Muinelo, Máximo Rossi y Paticia Tiunfo Documento No. 06/05 Noviembe, 2005

2 Medical behavio: an application to cesaean section delivey in Uuguay 1 Leonel Muinelo Depatment of Economics School of Social Sciences Univesity of the Republic Montevideo, Uuguay Phone: Fax: Paticia Tiunfo Depatment of Economics School of Social Sciences Univesity of the Republic Montevideo, Uuguay Phone: Fax: paticia@decon.edu.uy Máximo Rossi Depatment of Economics School of Social Sciences Univesity of the Republic Montevideo, Uuguay Phone: Fax: mito@decon.edu.uy 1 This wok could be caied out thanks to the data povided by the Depatment of Population Infomation of Health Public Ministy (Depatamento de Infomación Poblacional - Ministeio de Salud Pública). We specially thank its Diecto Da. Teesa Puppo, and also Mainés Figueoa and Fedeico Ramos. The authos would like to thank suggestions and comments fom D. Rafael Aguie. As usual, eos and omissions ae ou own esponsibility. 1

3 Resumen En Uuguay, al igual que en todo el mundo ha aumentado sistemáticamente la tasa de nacimientos po cesáeas, llegando a se en Montevideo en el año 2003 un 23% de los nacimientos en los hospitales públicos y un 42% en los pivados. Utilizando los nacimientos egistados en el año 2003 (23.474) po el Sistema Infomático Peinatal en Montevideo (Uuguay), se estima la pobabilidad de tene una cesáea contolando po los pincipales factoes de iesgo y po la endogeneidad en la elección del hospital donde se atienden. Los hospitales públicos montevideanos básicamente pagan po salaio fijo y los pivados po acto médico. Según pedice la teoía en los pimeos no hay ningún efecto del ingeso sobe la inducción y esta es ceo, mientas que en los segundos es positiva, a la vez que se educe paa mujees de meno iesgo médico. Po lo anteio, difeencias en las pobabilidades iían en el sentido de confima la hipótesis de inducción de demanda en este acto médico en paticula. Las estimaciones pemiten afima que una muje tiene 20% de pobabilidad de tene una cesáea en un hospital público y un 39% en uno pivado. A su vez, las difeencias ente los dos tipos de hospitales son mayoes paa mujees de meno iesgo. Po ejemplo, aquellas mujees que no pesentaon ninguno de los iesgos médicos consideados en este estudio tienen un 11% de pobabilidad de tene una cesáea en un hospital público y 25% si se atienden en un hospital pivado más del doble-. Palabas claves: demanda inducida, cesáeas, exogeneidad Abstact Consideing deliveies egisteed in 2003 (23.474) by the Penatal Infomation System (PIS) in Montevideo (Uuguay), the pobability of having a cesaean section delivey is estimated, contolled by isk factos and the endogeneity of the choice of hospital. At Montevideo pivate hospitals this pocedue has to be paid fo wheeas public hospitals have fixed budget payment systems. The empiical wok yields evidence to say that thee is 20 pe cent of pobability fo a woman of having a cesaean section delivey in a public hospital; while in pivate hospitals the pobability ises to 39 pe cent. At the same time, diffeences between the two types of hospitals get bigge fo lowe isk women. Fo example, women consideed to have no medical isk have double the pobability of a cesaean section in pivate hospitals than in public ones (25 pe cent vesus 11 pe cent). Keywods: induced demand, cesaean section delivey, exogeneity JEL-Classification: I11, I18, C35 2

4 1. Intoduction Health cae can be seen as an input in the individual health poduction function. In this sense, it is a deived demand, that many times, because of infomation asymmeties between the physician and the patient; it is oiginated by the physician, who indicates what and how to demand. Health economics liteatue has emphatically studied the impact of these infomation asymmeties on the agent s decision taking pocess. It was paticulaly studied the use of the physicians main position to obtain an economic benefit, like the incentive to change consumption demand in ode to maximize thei utility (demand induction). Evidence is found against neoclassical taditional theoy pedictions; fo example in Uuguay it is obseved a positive coelation between income and the amount of physicians duing the nineties. Moeove, when the payment to medical attention became an impotant popotion of physicians income (paticulaly among anesthetic-sugey pivate secto in Montevideo since 1993), a ise in the numbe of sevices povided was obseved. Thee was specially a ise of non-complex sugeies such as cesaean section (Fleis y Uestaazu, 2000). Due to the incidence of this pocedue and the diffeences among hospitals, in this wok we intend to show diffeent pactice styles contolling by isks chaacteistics of population attended in diffeent institutions. Infomation is taken fom Penatal Infomation System (SIP, CLAP-OPS/OMS, MSP) fo Montevideo in

5 2. Cesaean sections Bith can occu though vaginal delivey o though the abdomen, that is, though a cesaean section. This last pocedue was less fequent until mid XX centuy. Actually, in the past times the only cesaean sections made wee post motem in ode to save the child once the mothe was dead. In the XIX centuy it was used in obstucted labo with the fetus alive, and its application ose in XX centuy (CLAP, 1989). Duing the foties a cesaean section libealization was poduced, ising fom 2 to 4 pecent of total biths and stating a sustained and genealized incease until pesent. Fo example, in the United States it ose fom 5.5 pecent of total biths in 1970 to 23.0 pecent in 2000; in Sweden this popotion moe than duplicated duing the seventies, eaching in pecent of total biths. In Latin Ameica, Chile appeas as the county with highe ates, epesenting 40.0 pecent of total biths in In Uuguay, annual aveage ate between 1985 and 1989 was 16.8 pecent, while in 2003 became 27.0 pecent, obseving huge diffeences among institutions. Such as in the est of Latin Ameica, highe fequencies of cesaean sections wee obseved in pivate institutions. [TABLE1] 4

6 Reasons fo this development have been extendedly discussed. Development of new medical technology (hemotheapy, anesthesiology, antibiotics, chemotheapy and pe-bith diagnosis), existence of egulation and legislation elated to medical capacities, change in physicians pactice style and patients pesonal attitudes elated to cesaean section, can be pointed as the main causes. At the same time, thee is empiical evidence both fo and against this pocedue as a way of deceasing penatal motality ate. [TABLE2] Seveal oganizations suggest ideal cesaean section ates. Fo the Wold Health Oganization (WHO) it is between 10 and 15 pecent; fo the United Nations and its pogam Health fo All, it vaies fom 15 pecent fo nullipaous to 63 pecent fo women with pevious cesaean sections. Fo othe oganizations the ideal ate is the minimum. Anyway, thee is a consensus in specialized liteatue that cesaean ate depends on population chaacteistics. At the same time, cost-effective studies on cesaean section have detemined that a cesaean section without complications could epesent an ove cost between 66 and 200 pecent compaed to a egula delivey, depending on the county (Clak et al, 1991; Keele y Bodie, 1993; Eckelund y Gedthamn, 1996, Epstein y Nicholson, 2005). Fo that eason, it is elevant to study the main factos that lead to the decision of making a cesaean section, as they expose both, mothe and child, to unnecessay isks and incease medical costs. Accoding to Schwacz et al (1989) the ise in this kind of pocedue can be explained by many factos, such as the decease of associated isks, new indications based on bette penatal esults and a low pobability of the physician of getting pactice in

7 vaginal deliveies. Howeve, abusing of cesaean sections is concening when exta-health factos that go against mothe and child inteest pevail, such as economic factos, health team comfot, among othes. Taking all explanations into account, easons fo cesaean section can be gouped in thee: a) Obstetic indications i. Matenal: pevious cesaean sections, pevious tumos, uteine factue, sevee chonic hypetension, and failue of induction. ii. Fetal: abnomal fetus pesentation (coss, podalic, etc.), fetal macosomy, compomised living of the fetus, multiple pegnancy, and fetus suffeing. iii. Matenal-fetal: fetopelvic dispopotion, obstucted labo. iv. Ovulaian: pevious placenta, etc. b) Non-medical patient conditions: age, socio-economic situation, expeience in malpactice lawsuits, etc. c) Non-medical pofessional conditions: economic incentives, pofessional discetion, technology availability, etc. 3. Antecedents Gube and Owings (1996), using infomation on biths in the United States between 1970 and 1982, study the existence of induced demand on cesaean sections, given the income vaiation of obsteticians duing that peiod. The authos outline that in this peiod, thee was a decease in fetility ate of 13.5 pecent, making physicians adjust by quantity, thus inceasing the numbe of biths by cesaean section. In pesence of changes in medical emuneations, it was not possible to identify supply esponses, due to the fact 6

8 that, at the same time, thee existed changes in demand as a esponse to change in pices, which could be eal o induced. With a database of appoximately 256 thousand of biths obtained fom the National Hospital Dischage Suvey, the authos estimate logit models with the pupose of detemining the pobability of a bith concluding in a cesaean section. Contolling by state and yea of bith, they find that the pobability depends on demogaphic chaacteistics of the mothe, on having a pevious cesaean section, on the hospital chaacteistics, the numbe of obsteticians by bith occued in the state and the numbe of biths elated to state population. Results show that a 10 pecent incease of obstetician density inceases a 0.6 pecent the pobability of a cesaean section. At the same time, thee exists a monotonic incease with the age of the mothe. The pobability of cesaean section inceases fo maied and white women; in big hospitals, and in pivate ones. They also point out that pevious cesaean section is a good pedicto of epeating the pocedue. Finally, contol vaiables such as isks factos of pegnancy, bette explain the incidence of cesaean delivey. Eckelund and Gedthamn (1996) using data fom the medical egiste of 59 obstetic depatments in Sweden in 1991 (98 pecent of total biths in the county), cay out a eseach of coss section data in ode to explain the vaiation in the cesaean section ate among the diffeent obstetic depatments. The authos identify nea 20 deteminants of cesaean section ate, mainly: age of mothe, multiple pegnancies, size of obstetic depatment, being o not a univesity hospital, vaiables elated to medical pactice style, etc. On one side, they find that univesity hospitals have a highe cesaean section ate than othe hospitals, and on the othe side, that thee existed a high vaiation among and inside 7

9 six egions of Sweden. At the same time, cesaean section ate inceases with age of mothe, mostly fo pimipaous mothes. Fabbi and Monfadini (2001), based on a natual expeiment of eduction of physicians honoaies that took place in the Italian egion of Emilia-Romana between 1997 and 1998, estimate a model of induced demand of biths by cesaean section. They develop an extended vesion of the model of demand induction intoduced by Newhouse (1970) and Evans (1974), pointing out that the majo limitation to these models is supposing that patients ae homogeneous. They ty to mend this estiction consideing a distibution of patients in five classes accoding to diffeent isk levels. Estimating a pobit model of simultaneous equations, in which the pobability of cesaean section depends on the type of hospital chosen, they explain the pobability of cesaean section elated to two types of hospitals (public and pivate hospitals), contolling by patient chaacteistics and isk factos. The vecto of individual isks includes vaiables elated with the numbe of pevious cesaean section, pesentation of the fetus, age of the mothe and ecent admission to hospital. Among the vaiables that detemine the choice of the hospital, they include: distance to hospital and some chaacteistics of the woman esidence locality (availability of cas, height, etc). In this study, they cannot eject the exogeneity of the choice of hospital, and theefoe the patient s election of a povide is exogenously detemined with egad to the inductive behavio and the hospital teatment. As a consequence of this esult, the model is educed to a single equation, focalizing the attention on the evaluation of the impact in the pobability of cesaean section accoding to isk factos, type of hospital and changes in schedules. With egad to the induced demand, they povide obust evidence of the pesence and magnitude of this behavio. The authos find that the pobability of 8

10 cesaean section, afte a eduction of 20 pecent in honoaies, fo any type of isk, is systematically and significantly highe in pivate hospitals (financed with taiffs) than in public hospitals (with fixed budget). At the same time, esults suggest that a maginal incement in the pobability of cesaean section due to a eduction of physicians honoaies is significantly lowe among high isk women compaed to women with lowe isk. In esponse to the income shock, cesaean sections inceases 65 pecent fo women with low isk what epesents moe than 50 pecent of the sample. Howeve, if women with highe isk ae analyzed, which epesent no moe than 3 pecent of the sample, cesaean section pobability inceases only 4 pecent points. In the case of Uuguay, the only study that intends to give hints on the induced demand hypothesis fo medical acts was caied out by Fleis y Uestaazu (2000). The authos undeline the change of the medical emuneation system in 1993 in Uuguay that pioies the payment of medical and sugical act, finding an incease on the aveage eal monthly income of physicians along the nineties, especially in the pivate secto in Montevideo. At the same time, thee was a ise in the numbe of sugey pocedues, especially non-complex ones. In geneal, they find a positive coelation between emuneation and the physician supply of 0.8 fo the peiod consideed, suggesting, like Reinhadt test, the existence of an induced demand. Howeve, they cannot pove the hypothesis of induction of demand, because they fail in identifying unnecessay acts in the incease of the quantity of sevices povided. 9

11 4. Induced demand 4.1. A theoetical model In health, diffeences between patient and physician in tems of knowledge, suggest the possibility that physicians may take advantage of that asymmety fo thei own economic benefit. This theatens the economic maket paadigm, gadually destoying the nomative implications undelying economic ecommendations about maket policy. Since the seventies health economics have centeed its attention on the demand side. But if physicians ae capable of inducing demand, supestuctue constucted on consume theoy is theaten, and this is not coect fo this type of maket stuctue (Culye y Newhouse, 2003). Induced demand by physicians exists when the physician influence the patient s demand fo health sevices against what the physician itself consides is the best fo patient s well being. It is impotant to take into account two sides of the pevious definition. Fist, thee is a diffeence between agency s good use and induction. In fact, when the physician influences in ode to take the patient to the optimum, it would not be induction, but the good use of agency elation (Culye y Newhouse, 2003). The othe side to look at elates to the diffeence among use and demand. A physician could influence use level but not influence demand. In tems of economic vision of induced demand, thee ae theoetical easons to believe that it always exists at some level. If we conside a physician that gives the optimal quantity of infomation, the patient will be using the optimal quantity of health sevices. Envelope theoem agues that, aound this point, a small incease o decease of quantity 10

12 will have a small effect on consume welfae. On the contay, physician can ean moe money when inducing a patient to demand moe. No matte which theoetical model is used to ague physician motivation, tade off pesented (to ean moe affecting the patient o not to) implies that the physician, at least in some way, will end up inducing demand. Infomation asymmeties between physician and patient ae not a sufficient condition fo demand induction. Thee must exist incentives to do it. Fo example, when the supply of physicians ise, they might not be supplying the quantities they would like, and theefoe have an incentive to induce demand (Fleis y Uestaazu, 2000). Howeve, thee might exist obstacles to demand induction, which can be ethical (Hippocatic Oath), legal (medical licenses, sanctions to hospitals) o socioeconomic (eputation loss, time spent in convincing patients, etc.). Existence of induced demand on the supply side of health sevices was oiginally intoduced by Evans (1974) and Fuchs (1978). The main idea aises fom the fact that in geogaphical aeas with hospitals with highe beds supply thee was a highe use of sevices as well (Roeme, 1961). Evans (1974) suggested a model in which physicians maximize thei utility function with aguments as income o induction. Besides, it takes into account the disutility of induction that limits income geneation. On the othe side, Fuchs (1978) pesented induction as the ability of the physicians of changing the demand cuve in the maket. Howeve, thee ae altenative explanations among liteatue on how positive association between supply and demand can be sustained by itself in a competitive stuctue. Fo example, moe physicians (though density incease) decease time costs of patients, inceasing theefoe the demand. 11

13 As fo limits to induction, some authos find them as a taget income fo physicians (Newhouse, 1970; Evans, 1974); othes as a moal limit fo themselves (Sloan y Feldman, 1978; Gube y Owings, 1996); while othes find that thee might be limits imposed by patient pecaution (Danove, 1994). At the same time, in liteatue thee is a clea distinction among models that limit induction in a context of benefit maximization (Danove, 1994; Stano, 1987), and models that incopoate disutility of acting against the best inteest of the patient (Evans, 1974; Fuchs, 1978; McGuie y Pauly, 1991; Gube y Owings, 1996; Zweifel y Beye, 1997; Calsen y Giten, 1998; Fabbi y Monfadini, 2001). McGuie y Pauly (1991) fomalized the ideas of Evans and Fuchs in the context of a model that intends to explain the induction poduced by physicians that espond to changes in income, incopoating as a limit to induction the disutility of induction. Next, taking these woks as a basis, we develop the theoetical model used in this wok. It is assumed that diffeent obstetic depatments attend women with diffeent isks, which ae distibuted among a discete numbe of isk types ( ), with = 1,..., s,..., R, whee if isk types ae odeed and R > s, women in R type pesent moe isk than women in s type. At the same time, each hospital o obstetic depatment has a utility function in the way: U = U ( Y, I1,..., I R ) ' ' U Y > 0, U < 0 (1) U '' '', U YY < 0 12

14 Whee Y epesents obstetic depatment income, and I is the induction level of women with isk type. ' When assuming that U < 0, it is implied that pofessionals espect an ethical code, and physicians have disutility when exploiting thei agency elation in inducing demand, which is bigge when the patient isk is lowe: U <, with R > s. ' ' s U R salaies, f. Thee can exist two types of payments: fo pocedue o medical act, p, o fixed In the fist case and assuming sepaable additive pefeences, we obtain: U p = U Y R p p ( Y ) + U ( I ) (2) = 1 Y + p p p = YN N YCC (3) Whee N is natual biths, C is Cesaean section, and Y, Y ae income fom each N C intevention. At the same time, Y Y > 0, that is, the eimbusement pemium fo C N pacticing cesaean section is high enough to compensate the obstetic depatment fo any loss of time spent in it. If B epesents deliveies fo women of isk type, we have: C N p p = = R = 1 R = 1 φ ( i ) B (4) p p ( 1 φ ( i )) B (5) I = i B = 1,..., R (6) p p 13

15 Whee i epesents the induction by bith fo isk type, epesenting an effot without cost that the physician makes to induce cesaean section demand fo a given bith. φ i ) Is an induction function that detemines the cesaean ate fo isk type fo each level of effot in induction made by physicians. It is supposed that cesaean section ate ises with ' '' the induction effot, φ (.) > 0, that φ (.) = 0, y φ ( o) > 0 (a faction of biths ae coectly diagnosed, equiing a cesaean section). Moeove, due to the fact that cesaean sections ae moe fequently pefomed to isky women, we conclude that φ 0) > φ (0), given > s. (6). Physicians maximize ( s ( p U (equation 2) with espect to i, subject to estictions (3)- Fist ode conditions let us affim that thee is a tade off between the net disutility of inducing, and the net utility of the income ise. Taking these esults it is possible to make a compaative static execise, fo example, a decease in the numbe of biths ceteis paibus- will make induction ise, and theefoe thee will be a ise of biths by cesaean section. We can also make an analysis of the effect of contolling medical payments fo the amount of inteventions made. Fo example, Yip (1998, cited in Culye y Newhouse, 2003) applying this model to thoax sugeies in New Yok and Washington, analyze the impact of contolling pices in the public system (Medicae). The autho finds evidence that this pice contol aised the amount of sugeies in the public and pivate systems though an income effect. In paticula, the estimations of Yip find that thoax sugeies ecoveed a 70 pecent of loss income via pices, though a ise in volume (Culye y Newhouse, 2003). 14

16 Unde this payment system, economic incentives to induce demand ae highe than if the payment is made by a thid paty instead of the demande of medical sevices itself, making unpedictable the amount destined to cove the expenses. At the same time, even though the payment fo act pomotes a model based on healing instead of pevention, we must emphasize that it aises physician poductivity, ewads pesonal effot and pesonal esponsibility, and impoves the quality of the sevice (Fleis y Uestaazu, 2000). In the case of fixed payment, the physician income depends basically on the amount of woking hous, whee the hou-value will depend on the activity, gade and specialty. In the optimum, thee is no effect of income on induction, thus it is zeo. Theefoe, this payment modality geneates incentives to eliminate supefluous pactice that would only povoke a decease of physicians utility. Howeve, it may geneate incentives to an insufficient quantity of medical sevices supply. 4.2.Empiical model and infomation Following Fabbi and Monfadini (2001) in this study we estimated the pobability of a cesaean delivey in Montevideo, contolling fo diffeent patient s isks. The type of the hospital could affect this pobability, because thee ae two payment systems: fo act (pivate hospitals) and fixed salay (public hospitals). At the same time, given that the isk type of patients educes induction, it is impotant to estimate the diffeences fo women with lowe elative isk. As it was mentioned befoe, theoy pedicts that in the second goup the induced demand is zeo, but it is positive in the fist goup and deceases with the isk level of patients. 15

17 It is impotant to biefly systematize the payment system in Uuguay. Aound 60 pecent of opeative expenses in hospitals ae emuneations, and physicians wok basically in a dependence elation. In health secto thee ae two sub sectos with diffeent contacting ules. On one side, thee is the public sub secto whee payment modality is mainly thoough fixed salay; and on othe side, thee is the pivate sub secto whee it is necessay to distinguish between the capital city and the est of the county. In 1993, a collective bagain ageement established that physicians fom pivate secto in Montevideo would be emuneated by medical and sugical act, educing the weight of base salay in total income, while in the est of the county the payment modality is still based on fixed salay (Fleis y Uestaazu, 2000). Even though payment by act aleady existed, the ageement of 1993 inceased consideably in eal tems the pice of anesthetic-sugical acts. Real income of physicians showed a maked inceasing tend duing the nineties, being an impotant pat of this incease the ise in 1993 (Fleis y Uestaazu, 2000). At the same time, if we compae the aveage eal income of physicians fom public and pivate secto, we obseve that emuneation to physicians in public secto is appoximately a thid of emuneation to physicians in pivate secto. As we do not count with data fo befoe and afte the change of 1993, in this wok we decided to compae the medical behavio egading biths in two types of institutions in Montevideo. It is impotant to undeline that thee ae othe diffeences than the payment systems in both types of hospitals, such as the available capacity of sugical emegency, medical pactice styles 2 and socioeconomic level of patients. Due to the fact that this last 2 In Uuguay, public hospitals ae mostly univesity hospitals, paticulaly in Montevideo. Fo example, Cento Hospitalaio Peeia Rossell, the most impotant matenity in the county, is a pactice hospital, and in 2003 concentated 40 pecent of total deliveies in Montevideo. 16

18 chaacteistic is contolled fo in estimations, that a high popotion of physicians have both public and pivate jobs and that the sugey is non complex, is that we undestand that if thee ae diffeences, they ae because of the payment system 3. It is possible to ague that vaiables such as type of hospital ae detemined endogenously, due to the fact that the pobability of cesaean section depends on nonobsevable vaiables that ae coelated with non-obseved chaacteistics that undemine the choice of the hospital. This could lead to inconsistent estimations. In fact thee could exist selection bias, fo example hospitals with high-isk population, etc 4. The above makes necessay to find out whethe it is convenient o not to jointly estimate both phenomena. The model stuctue is based on a educed equation fo the potentially endogenous vaiable (hospital) and a second stuctual equation fo the outcome of inteest (cesaean section) * ' (7) y1 i = β 1x1i + u1 i * ' ' y2 i = β 2 x2i + u2i = δ1y1 i + δ 2 z2i + u2i Whee * y1i is the latent vaiable fo hospital choice, and * y2i is the latent vaiable fo the identification of a cesaean section delivey. Both vaiables ae obseved as binaies of the fom: 3 Double job is less pobable in the main matenity hospital in Montevideo and Uuguay: Cento Hospitalaio Peeia Rossell. It is a public, univesity hospital that woks mainly with esidents, and as they have not finished thei caee, they ae not allowed to wok in pivate hospitals. 4 In Uuguay, the chaacteistics of the health system make the woman to choose between a pivate o public hospital in he place of esidence. The choice among diffeent pivate institutions is only possible in the capital city of the county. 17

19 y 1i 1 = 0 if y if y * 1i > 0 * 1i 0 ( hospital) (8) y 2i 1 = 0 if y if y * 2i > 0 * 2i 0 ( c. s) x 1 i and z 2i ae vectos of the exogenous vaiables in both equations, including the fist vecto of instuments to be used fo contolling endogeneity; β 1,δ 2 ae the paamete vectos and δ1 is a scala. The eo tems ae assumed to be independent and identically distibuted as a bivaiated nomal with zeo mean and unitay vaiance, with ρ = co u 1, u ). The ( 2 exogeneity condition could be put in tems of ρ, which could be intepeted as the coelation between the unobseved explanatoy vaiables of both equations. The Wald test fo ρ, implies that, in case of not ejecting the null hypothesis, y i y u2i 1 ae not coelated, so y 1 i is exogenous in the second equation. In the liteatue on these topics it is possible to find many altenative exogeneity tests (Wooldidge, 2002; Fabbi et al, 2004; Baum et al, 2003). In the case of the fist specification (equation 8), the estimation can be done by maximum likelihood, using bivaiate pobit o seemingly unelated bivaiate pobit (SURE pobit), whee the equations ae elated solely by the petubations. In ode to contol fo the potential endogeneity of the hospital choice, one must include y 1 i as an endogenous egesso, making it possible to study it effect on the cesaean section delivey. At the same time, x 1 i must include the instuments to be used to contol the endogeneity. 18

20 Anothe possibility is to use instumental vaiables and estimate two stage models, which ae a paticula case of simultaneous equation models. In this sense it is possible to specify a linea pobability model (LPM-IV 2SLS), with the disadvantage of fixed patial effects but good fo values aound the poblational mean of the egesso- and heteoskedastic by constuction, this is why obust standad eos ae pesented (Wooldidge, 2002; Baum et al, 2003) 5. Finally, estimations can be done tough two stage pobit least squaes models (2SPLS), maintaining the specification pesented in equation 2 6. As exogenous vaiables individual isk factos ae included, such as pevious cesaean sections, fetus pesentation, eclampsy, peclampsy, hypetension, multiple pegnancies and fetopelvic dispopotion. The liteatue usually uses insuance pices as instumental vaiables fo the hospital choice as they ae coelated to this, but not with the fact of having o not a cesaean section delivey. Howeve, in Uuguay pice of public hospitals is basically zeo, and pice of pivate hospitals is egulated, existing low vaiation among pivate institutions. At the same time, as we ae woking with coss section data, thee is no tempoal vaiation in pices. Fo these easons, in the pesent wok we conside that thee is no pice constaint. As constaints in the hospital choice we use education and maital status of the mothe, because, as theoy pedicts, women moe educated and maied (o in stable union) ae moe likely to plan pegnancy and to invest on health, but these vaiables themselves do not affect the pobability of having a cesaean section (Culye and Newhouse, 2003). If we 5 Stata 9 (StataCop, 2005) softwae is used to estimate 2SLS with the command iveg2, which gives the exogeneity tests as well as the ones fo elevance and validity of instuments. 6 The ivpobit command is used (2SPLS). As the endogenous egesso (hospital) is binay, it is chosen to wok in two stages, whee the values pedicted in the fist stage ae used as instuments in the second one. 19

21 find that h is exogenously detemined, the model could be educed to an equation of the pobability of a cesaean section conditional to the hospital choice. We used data fom the Penatal Infomation System (SIP, CLAP, OPS/OMS) fo the yea 2003 povided by the Public Health Ministy (Ministeio de Salud Pública). Registeed biths by this system wee 39,937 all ove the county. This epesents 70 pecent of the egisteed biths by the Bon Alive Cetificate that counts with a univesal coveage of the county. In the capital city, Montevideo, whee in geneal SIP coveage is wide, 23,474 biths wee egisteed. In what follows, vaiables consideed as isk factos ae defined, without consideing actual medical pegnancy poblems diagnosed. Age of the mothe, measued by fou binay vaiables that: take each one the value of 1 in case of the mothe being younge than 16, between 17 and 19 yeas old, between 35 and 39 yeas old, and olde than 40 yeas. This specification instead of a continuous one is used in odeed to captue diffeences among specific stages of woman s life. Omitted age ange in estimations is between 20 and 34 yeas old, because this is the ideal age in social and epoductive tems. As matenal isk factos we consideed deliveies, pevious cesaean sections, high blood pessue (eclampsy, peclampsy, pevious hypetension), being in all cases binay vaiables that take value 1 if the chaacteistic is pesent 7. As matenal-fetus isk facto we consideed the fetopelvic dispopotion, measued by the vaiable dispopotion, which is binay and takes the value of 1 if it is pesent. 7 We did not consideed medical indications of having cesaean section. 20

22 Finally, as fetal isk factos we consideed abnomal pesentation of the fetus (pelvic o coss) though pesentation, binay vaiable that takes the value of 1 if it is pesent; and multiple, binay vaiable that takes the value of 1 if it is a multiple pegnancy. As non-medical conditions we consideed the type of institution whee the physician woks, using public as a binay vaiable that takes the value of 1 if the bith occued in a public institution and zeo othewise. With the pupose of contolling endogeneity in the choice of the hospital, we included in equation 8 education, continuous vaiable fom 1 to 4 (1 no education, 2 pimay education, 3 seconday education and 4 univesity); and stable union, binay vaiable that takes value 1 if the woman is maied o in a stable union. In Table 3 we pesent desciptive statistics of the vaiables. [TABLE3] 5. Results The pobability of a cesaean section delivey eithe in a pubic o pivate hospital is estimated fo Montevideo in the yea 2003, contolling by the potential endogeneity of the hospital choice. The diffeent Wald tests fo exogeneity in each estimation pocedue makes it possible to state that the hospital choice is endogenously detemined with the pobability of a cesaean section delivey 8. At the same time it is possible not to eject education and maital status as appopiate instuments. 8 Smith-Blundell test fo 2SLS; Wald test fo ρ in the SURE pobit and 2SPLS. At the same time, the test fo the instuments confims the appopiateness of thei use (2SLS: Patial R2=Shea= in the case of 2 21

23 As obseved in Table 4 in the hospital choice the main deteminants ae: age, education and maital status, and some medical isk facto peviously known by the mothe as hypetension and pevious cesaean sections. In this way, the pobability of choosing a public hospital is highe fo: least educated, single and young women. The significant and positive isk factos ae hypetension, peclampsy and pevious deliveies of the woman. The last vaiable can be seen as a poxy of income, consideing the fact that pooe women have highe fecundity ates, ising the pobability of choosing a public hospital. On the othe hand, having had a pevious cesaean section negatively impacts on the pobability of choosing a public hospital, meaning that women assume that futue pegnancies would have the same ending what takes them to invest in pivate insuance. [TABLE4] In Table 5 we pesent the second stage esults fo the diffeent estimation pocedues also with the esults of the classic pobit model (without contolling fo endogeneity) in ode to show the diffeences. [TABLE 5] Just looking at the coefficient sings, it is possible to say that the pobability of having a cesaean section ises with the age of the mothe, fo pivate hospitals and the diffeent isks, excluding, as expected, deliveies that have a negative sign. Given the non-lineaity of pobit models, in Table 6 we pesent the coesponding maginal effects, eplying at the same time the coefficients of the linea model in ode to make the compaison possible. instuments, F(2, 17618); SURE pobit: Chi2(1)). Tests, as well as complete outcomes ae available upon equest to the authos. 22

24 In this sense simila esults ae obseved. While the aveage pobability of having a cesaean section is appoximately 25 pecent, women in pivate hospitals have between 17 and 19 moe pecentage points of pobability o having a cesaean section, between 10 and 14 moe if they ae 40 o moe yeas old; if fetopelvic dispopotion is pesent; if they had pevious cesaean sections, etc. [TABLE 6] When the estimation poceeds without contolling fo endogeneity and focusing on the hospital type, the esults show an undeestimation given the maginal effect of 13%. Contolling by the isk factos fo people teated in the diffeent hospitals, esults let us affim that the pobability of having a cesaean section is always gate in a pivate institution. At the same time, this diffeence is even geate fo low-isk women. As it can be seen in Table 7, the pobability fo a women without any isk facto of having a cesaean section is: 11% in a public hospital and 25%in a pivate one moe than twice-. If we analyze multipaous women without any isk facto the diffeence between public and pivate, almost tiplicate, 7% in public hospitals and 18% in pivate ones, and double fo nullipaous without isk factos (20% and 39% espectively). 9 [TABLE 7] Diffeences found can be explained by non-obsevable factos such as medical indications and isk factos not egisteed in SIP, medical style pactice, hospital technology, patients attitudes, etc. Accoding to pevious studies, in Montevideo thee is a high popotion of physicians that combine public and pivate jobs, due to low emuneations in the public secto. Fo that eason, it is possible to affim that thee ae no 9 60 pecent of the women in this data do not have any of the isk factos consideed, while 16 pecent of them had a cesaean section delivey. 23

25 substantial diffeences in medical staff in both types of hospitals (Bucheli, 2000). On the othe hand, the act analyzed is a low complexity sugey, so no big diffeences ae expected among institutions technology. In this sense, it is consideed that diffeent emuneation systems explain much of the diffeent pobabilities estimated, leading to the hypothesis of induced demand. 6. Conclusions In Uuguay, as all ove the wold, the popotion of biths by cesaean section has inceased. Reasons fo this tend have been extensively discussed, undelining the existence of laws and ules elated to medical pactice, development of new technologies, changes in medical pactice styles, as well as attitudes of patients towads cesaean section. Basing on the deliveies occued in Montevideo (capital city of Uuguay) and egisteed in the Penatal Infomation System (SIP, CLAP-OPS/OMS, MSP) in the yea 2003, this ate was 23 pecent in public hospitals and 42 pecent in pivate ones. Although thee is no ageement on the ideal ate fo cesaean sections, thee is a concen about the pocedues made without clea medical justification, because they expose both mothe and baby to unnecessay isks, and they ise medical attention costs. In this wok we estimate the pobability of having a cesaean section contolling by main isk factos of population and by the type of hospital whee it is done. As the choice of hospital is endogenous, we estimate a two-stage pobit model by maximum likelihood. Pobability of cesaean section inceases with age of woman, pesence of eclampsy, peclampsy, pevious hypetension, pevious cesaean sections, multiple pegnancy and fetopelvic dispopotion, and deceases fo multipaous women and women attended in a 24

26 public hospital. In fact, pobability of cesaean section in a pivate institution is almost double than having one in a public hospital (20 pecent vesus 39 pecent). Focusing on women that do not pesent any of isk factos consideed, we found that they have an 11 pecent pobability of having a cesaean section in a public hospital and 25 pecent pobability in a pivate one. Results ae identical if we conside only women younge than 35, age in which is less pobable to have a cesaean section. Diffeences between the two types of institutions may be because of non-obsevable factos such as medical indications o isk factos not consideed, medical pactice styles, hospital technology, patient pessue, etc. Howeve, as we contol by the socioeconomic level of patients, as in Montevideo exists a high popotion of physicians that combine public and pivate wok due to low emuneations in public secto- and as cesaean section is low-complexity sugey, we conside that the emuneation system explains much of the diffeence found in the pobabilities, leading to the existence of induced demand. It is not the pupose of this wok to foment defensive medicine (Danson, 2000, taken in Culye and Newhouse, 2003). This is undestood as the medicine that physicians cay out in ode to pevent themselves fom law suits; that is, these pocedues do not benefit o involve the patient in isks but the physician ecommend them fo easons of pesonal secuity. It can be expected that the physician always poceeds in benefit of the patient and in that sense, the cuent emuneation systems acts as an incentive. But it does not appea as the ight one. It also encouages multiemployment. Fom the patients point of view, it can be agued that the physician-patient elationship in the pivate secto is moe pesonal, what can lead to pessues fom the patient o its family in ode to delive by cesaean section. In this sense, education and 25

27 infomation given to the patient befoe the delivey ae key factos. In the fist case, thee ae scientific findings that should be taught in delivey didactical health, duing pegnancy contols, etc. This include the ise in matenal death isk in cesaean section vesus nomal delivey, the coelation between the ise in the cesaean section ate and the decease of the motality ate, especially afte a given theshold and in the last decade, etc. On the othe hand, the medical and non-medical estaining of women and thei families, the socialization of the pocess, etc., will help to lowe the pessue and avoid the physicians exposue to the ethics of esistance (El País, 2001). 26

28 7. Bibliogaphic efeences Baum, C.F., M.E. Schaffe and S. Stillman (2004): Instumental vaiables and GMM: estimation and testing, Boston College, Woking Pape Nº 545, Becke, G. (1965): "A Theoy of the Allocation of Time", Economic Jounal 75. Bucheli, M. (2000). Remuneaciones del secto salud, Infome paa el Banco Inteameicano de Desaollo, Unidad Pepaatoia del Poyecto UR120 (Reestuctuación del Hospital de Clínicas de la Univesidad de la República), no publicado. Calsen, F. and J. Gytten (1998): Moe physician: impove availability o induce demand?, Jounal of Health Economics 7. Cento Latinoameicano de Peinatología (CLAP) (1989): El nacimiento po cesáea hoy, Boletín del CLAP, OPS/OOMS, Vol. 3, Nº 9. Clak, L., M. Mugfod and C. Pateson (1991): How does the mode of delivey affect the cost of matenity cae?, Bitish Jounal of Obstetics and Gynecology 98. Culye, A. J. and J. Newhouse (2000): Handbook of Health Economics, Volume 1A y 1B, Elsevie Noth Holland. Danove, D. and P. Wehne (1994): Physician-induced demand fo childbiths, Jounal of Health Economics 13. Eckelund, I. and U. Gedthamn (1996): Vaiation in Cesaean Section Rates in Sweden Causes and Economic Consequences, Cente fo Health Economics Stockholm School of Economics. El País (Anuaio 2001), Cesáeas a pesión de Viginia Alington, Elías, A., L. Escalante, M. Loenzelli and S. Milnitski (2000): La cisis de las IAMC: poblemas de gestión? Un Enfoque Institucional, Univesidad de la República Facultad de Ciencias Económicas y Administación Cáteda de Economía Institucional y de las Oganizaciones, mimeo. Epstein, A. and S. Nicholson (2005): The fomation and evolution of physician teatment styles: an application to cesaean sections, Woking Pape 11549, National Bueau of Economic Reseach. 27

29 Evans, R. (1974): Supplie-induced Demand: Some Empiical Evidence and Implications, in M. Peelman (Ed.) The economics of Health and Medical Cae, London: McMillan. Fabbi, D. and C. Monfadini (2001): Demand induction with a discete distibution of patients, Depatment of Economics, Univesity of Bologna. Fabbi, D., C. Monfadini and R. Radice (2004): Testing exogeneity in the bivaiate pobit model: Monte Calo evidence and an application to health economics, Depatment of Economics, Univesity of Bologna. Fleis, P., and I. Uestaazu (2000): El mecado de la salud uuguayo en la última década: cambios en el sistema de emuneación a los médicos e incentivos económicos, Tabajo Monogáfico de la Licenciatua en Economía. Fuchs, V. (1978): The supply of Sugeons and the Demand fo Opeations, The Jounal of Human Resouces, Vol. 13. Gube, J. and M. Owings (1996): Physician Financial Incentives and Cesaean Section, Rand Jounal of Economics 27. Jaeghe, K. and M. Jeges (2000): A model of physician behavio with demand inducement, Jounal of Health Economics 19, Keele E. B. and M. Bodie (1993): Economic incentives in the choice between vaginal delivey and cesaean section, The Milbank Quately 71. McGuie, T.G. and M.V. Pauly (1991): Physician esponse to fee changes with multiple payes, Jounal of Health Economics 10. Newhouse, J. (1970): A model of physician picing, Southen Economic Jounal, Vol. 37, Nº 2. Phelps, C. (1986): Induced demand. Can we eve know its extent?, Jounal of Health Economics 5. Phelps, C. (2003): Health economics, Cap. 7, Univesity of Rocheste. Rossite, L. and G. Wilensky (1987): Identification of physician-induced demand, The Jounal of Human Resouces, Vol. 19, Nº2. Rossite, L. and G. Wilensky (1987): Health economist-induced demand fo theoies of physician-induced demand, The Jounal of Human Resouces, Vol. 22, Nº4. Sloan, F. and A. Feldman (1978): Competition among physicians, In Competition in the health cae secto: Past, Pesent, Futue, ed. W. Geenbetg, Baltimoe: Aspen Systems. 28

30 Stano, M. (1987): A claification of theoies and evidence on supplie, induced demand fo physicians sevices, The Jounal of Human Resouces, Vol. 22, Nº 4. Tempoelli, K (2001).: Análisis de la demanda de asistencia sanitaia: la utilidad del médico como deteminante, Depatamento de Economía Univesidad Nacional del Su. Wooldidge, J.M. (2002): Econometic Analysis of Coss Section and Panel Data,The MIT Pess, Cambidge, England. 29

31 TABLE 1: Health of mothe and newbon indicatos County Population (in Life Total fecundity Motality ate less than 5 yeas Numbe of Beds in Deliveies by Health expenses thousands) expectancy at ate old physicians in hospital fo cesaean section pe capita bith (yeas) habitants habitants (pecent) (intenational dollas) Men Women Agentina , , , Chile , , , Costa Rica , , , Spain , , S/d United States , , , Fance , , United , , Kingdom Uuguay , , , Souce: Elaboation with data fom Wold Health Repot 2005 by Wold Health Oganization (WHO). Note: Last yea available.

32 Table 2: Cesaean sections in Uuguay 2003 Public institutions Pivate institutions Piecewok In Univesity Reseve Montevideo Inteio Montevideo Inteio Yes No Yes No Yes No Total biths egisteed in SIP Total Cesaean sections Pecentage of cesaean sections Souce: Own elaboation based on data fom Penatal Infomation System (CLAP, OPS-OMS-MSP).

33 Table 3: Desciptive statistics. Montevideo N= 23,474 Vaiable Mean Standad Minimum Maximum deviation Cesaean sections Public Education Maied Multiple Pevious Cesaean sections Pesentation Dispopotion Deliveies Pevious hypetension Peclampsy Eclampsy Less than 16 yeas old Between 17 and 19 yeas old Between 35 and 39 yeas old Moe than 40 yeas old

34 Table 4: Fist stage esults: choice of hospital Montevideo 2003 N=17633 Vaiable SURE 2SLS 2SPLS 1 pobit Less than 16 yeas old 1.347*** (0.106) 0.247*** (0.158) 1.344*** (0.106) Between 17 and 19 yeas old 1.967*** (0.054) 0.246*** (0.009) 1.197*** (0.054) Between 35 and 39 yeas old *** (0.037) *** (0.009) *** (0.037) Moe than 40 yeas old *** (0.064) *** (0.163) *** (0.064) Education *** (0.023) *** (0.004) *** (0.023) Stable union *** (0.036) *** (0.007) *** (0.036) Deliveies 0.697*** (0.027) 0.194*** (0.007) 0.697*** (0.027) Peclampsy 0.288*** (0.062) 0.083*** (0.014) 0.286*** (0.062) Eclampsy (0.415) (0.073) (0.412) Pevious hypetension 0.293*** (0.084) 0.078*** (0.018) 0.295*** (0.084) Multiple (0.079) (0.019) (0.079) Pesentation (0.056) (0.013) (0.056) Pevious Cesaean section *** (0.033) *** (0.008) *** (0.033) Dispopotion (0.145) (0.039) (0.147) Test Wald: exogeneity Note: *** Significant at the 1% level, **significant at the 5% level, *significant at the 10% level. Robust standad eos between backets. 1 The naïve pobit esults ae pesented in this case, the pedicted values wee used as instuments in the second stage. 33

35 Table 5: Second stage esults: cesaean section Montevideo 2003 N=17633 Vaiable SURE pobit 2SLS 2SPLS Naïve Pobit Less than 16 yeas old *** *** *** *** Between 17 and 19 yeas old *** *** *** *** Between 35 and 39 yeas old 0.187*** *** *** ** Moe than 40 yeas old 0.411*** *** *** *** Deliveies *** *** *** *** Peclampsy 0.921*** *** *** *** Eclampsy 1.376** ** ** ** Pevious hypetension 0.338*** *** *** *** Multiple 1.221*** *** *** *** Pesentation 1.708*** *** *** *** Pevious Cesaean section 1.430*** *** *** *** Dispopotion 2.293*** *** *** *** Public *** *** *** *** Education 0.058** Stable union Log pseudo-likelihood Note: *** Significant at the 1% level, ** Significant at the 5% level, * Significant at the 10%. Level. 34