A Long Term Evaluation of the Japanese Medical Payment System for Cataract Surgeries: Did the Medical Policy Reduce the Long Hospital Stay in Japan?

Transcription

1 Healh, 016, 8, Published Online June 016 in SciRes. hp:// hp://dx.doi.org/10.436/healh A Long Term Evaluaion of he Japanese Medical Paymen Sysem for Caarac Surgeries: Did he Medical Policy Reduce he Long Hospial Say in Japan? Kazumisu Nawaa 1,, Koichi Kawabuchi 3 1 Graduae School of Engineering, Universiy of Tokyo, Tokyo, Japan Deparmen of Economics and Relaed Sudies, Universiy of York, York, UK 3 Graduae School of Medical and Denal Sciences, Tokyo Medical and Denal Universiy, Tokyo, Japan Received 30 May 016; acceped June 016; published 7 June 016 Copyrigh 016 by auhors and Scienific Research Publishing Inc. This work is licensed under he Creaive Commons Aribuion Inernaional License (CC BY). hp://creaivecommons.org/licenses/by/4.0/ Absrac In his paper, we conduced a long erm survey of he caarac surgeries. The sample period was abou 7 years, from July 005 o March 01. We evaluaed he effecs of hree revisions of he medical paymen sysem ha were done in 006, 008 and 010. For he analysis, he Box-Cox ransformaion model and Hausman es using Nawaa s esimaor were used for he lengh of say (LOS) in hospials, and he ordinary leas squares mehod was used for he non-inclusive (mainly paymens for surgeries) paymens. We analyzed a daase of 51,054 paiens obained from 60 hospials (Hp1-60) where more han 300 one-eye caarac surgeries were performed during he period. For he LOS, we found ha only he 008 revision had significan impac on shorening he LOS bu he oher wo did no. We also found very large differences among hospials even afer eliminaing effecs of paiens characerisics and ype of principle diseases as previous sudies. For non-inclusive paymens 006 and 008 revisions had significan impacs and he differences among hospials were much smaller han hose of he LOS. Keywords DPC/PDPS, Inclusive Paymen Sysem, Caarac, Lengh of Say (LOS), Box-Cox Transformaion Model, Hausman Tes How o cie his paper: Nawaa, K. and Kawabuchi, K. (016) A Long Term Evaluaion of he Japanese Medical Paymen Sysem for Caarac Surgeries: Did he Medical Policy Reduce he Long Hospial Say in Japan? Healh, 8, hp://dx.doi.org/10.436/healh

2 1. Inroducion A medical inclusive paymen sysem based on he Diagnosis Procedure Combinaion (DPC), called he DPC/ PDPS (DPC/per diem paymen sysem) [1], was inroduced in 003 for 8 special funcioning hospials. From April 004, general hospials could join he DPC/PDPS. Under he DPC/PDPS, he medical paymens consis of wo differen pars. One is inclusive paymens based on he DPC codes and Lengh of Say (LOS) in hospials (hereafer, inclusive paymens), and he oher is non-inclusive paymens based on he convenional fee-for-service sysem (hereafer, non-inclusive paymens). In principle, he firs one mainly covers he hospial fee ha is necessary o operae he hospial, and he second covers he docoral fee for medical reamens such as surgeries. For he inclusive paymens, hree periods, Periods I and II and he Specific Hospializaion Period, and per diem paymens are deermined by DPC codes, and per diem paymens become smaller as he LOS becomes longer. If he LOS exceeds he Specific Hospializaion Period, he inclusive paymen goes back o a convenional fee-for-service base. For deails of he DPC and DPC/PDPS, see Nawaa e al. []. The joining DPC/ PDPS has no been obligaory, and a hospial can freely choose o join he DPC/PDPS or no if i saisfies necessary condiions [3]. However, according o Cenral Social Insurance Medical Council [4], as of April 014, 1585 hospials had joined he DPC/PDPS and addiional 78 were preparing o join he sysem (hereafer, DPC hospials). The DPC hospials comprised 4.9% of he 7483 general hospials and had 511,439 beds, which accouned 57.0% of he oal (897,749) beds in all general hospials, and he likelihood of DPC paricipaion increased wih hospial size. The DPC hospials are required o compuerize heir medical informaion, i has become possible o analyze a large scale daase which includes informaion of many paiens, various diseases and reamens from many hospials. The DPC/PDPS has been revised by every second year afer 004. Therefore, evaluaions of revisions are very imporan for he efficien use of medical resources in Japan. In his paper, we evaluae he effecs of he 006, 008 and 010 revisions of he medical paymen sysem on he Lengh of Say (LOS) and non-inclusive paymens of caarac surgeries (Since acual inclusive paymens o hospials may be differen for idenical reamens, we analyzed he LOS raher han he inclusive paymens). The sample period was abou 7 years, from July 005 o March 01. In case of caarac surgeries, he porions of he inclusive and non-inclusive paymens are abou one-hird and wo-hirds, respecively. According o he survey of he Minisry of Healh, Labour and Welfare [5], 79,19 caarac surgeries were done for 59,318 cases and heir direc coss were 8.75 billion yen in June 013. This means ha nearly one million caarac surgeries and over 100 billion yen are spen annually. Alhough he effec of he single revision on he LOS was evaluaed [6], analyses of long erm evaluaion of he revisions on he LOS and he non-inclusive paymens have no been sudied. Among OECD counries, mos caarac surgeries were performed on a day-case basis, wihou saying hospial afer surgery (hereafer, day surgery), due o advances of medical echnology and beer anasheics [7]. Percenages of day surgeries of many OECD counries exceeded 95% in 011 [7]. In he caarac surgeries, a diry naural lens is removed, and hen replaced by an arificial lens, called an inraocular lens (IOL) [8]. The surgery will ypically ake less han 15 minues per eye [9]. Trivedi e al. [10] repored ha he lengh of hospial of caarac surgery could be reduced o an average of 3 hours and 37 minues. Aalla e al. [11] surveyed daa of 671 caarac paiens in an Ausralian hospial. 6 (33.4%) paiens were hospialized overnigh. Bu hey conclude ha many paiens who are hospialized overnigh could be safely reaed as day cases. Fedorowicz, Lawrence and Guie [1] also repored ha here was no significan difference in oucome or risk of posoperaive complicaions beween day surgeries and surgeries wih hospializaion. Even in an old sudy, Ingram e al. [13] repored ha only 13% paiens preferred o say hospial afer he surgeries among 501 caarac cases. On he oher hand, he long LOS for caarac surgeries in Japan was been very long. In our original daase, here were 56,364 one-eye caarac surgery (jus one eye was operaed in a single hospializaion) cases. Among one-eye surgeries, only 1.7% of paiens were day surgeries and average lengh of say (ALOS) was 3.8 days. The LOS shows a fa ailed disribuion on he righ hand side. Therefore, for he analysis of he LOS, he Box-Cox [14] ransformaion model (hereafer, BC model) is used as previous sudies. The maximum likelihood esimaor under he normaliy assumpion (hereafer, BC MLE) is used for he esimaion of he BC model. The BC MLE is generally inconsisen unless he small σ assumpion described in Bickel and Doksum [15] and he error erms are independen and idenically disribued (i.i.d.) random variables. Nawaa [16] proposed semi-parameric esimaors of he BC models. Using Nawaa s esimaor, we firs es he BC MLE o deermine wheher or no we can use he BC MLE for he esimaion of he BC 906

3 model by he Hausman [17] es.. Esimaors and Tess of he BC Model.1. BC Model Suppose ha LOS of paien is given by he BC model: z = x β + u, y 0, = 1,,, T, (1) λ y 1, if λ 0, z = λ log ( y ), if λ = 0, where y is he LOS, x and β are he k-h dimensional vecors of he explanaory variables and coefficiens, respecively, and λ is he ransformaion parameer. The likelihood funcion under he normaliy assumpions is given by log L ( θ) = log f ( θ), and ( θ) φ{ ( β) σ} σ ( λ ) log f = log z x log + 1 log y, () where θ ( λβ,, σ ) he variance of u. We can obain he BC MLE is obained by maximizing Le θ 0 ( λ0, β 0, σ0 ) =, φ is he probabiliy densiy funcion of he sandard normal assumpion and σ is = be he rue parameer value of θ. The BC MLE is generally inconsisen. However, if he error erms are i.i.d. random variables (hereafer, i.i.d. assumpion) and he small σ assumpion given by λσ 0 0 ( 1+ λ0x β0) 0 (pracically, P[ y < 0] is small enough under he normaliy ˆ assumpion), ˆ ˆ he BC MLE can be a consisen esimaor, θ = λ, β, ˆ σ are given by and small σ asympoics [15] of he BC MLE BC ( BC BC BC ) 1 T ( θ ) ( ) 1 BC θ0 N A B A ˆ 0,, (3) where A= E T θθ 1 log L θ0, and log f log f B = E. θ θ θ 0 θ0.. N-Esimaor Nawaa [16] considered he roos of he equaions, ( ) 1 log λx β + 1 z x β G ( θ) = + y z z x ( λx β ) g ( θ) ( β) λ T σ λ λ λx β z x β + log λ λx β + 1 log L log L = 0, and = 0. β σ When he firs- and hird-momens of u are zero, he esimaor given by Equaion (4) is consisen. The ˆ θ = λ, β, σ (hereafer, N-esimaor) becomes asympoic disribuion of he esimaor N ( N N N) 1 T ( θ ) ( ) 1 N θ0 N C D C ( ) θ where C = E θ 1 ξ θ β σ θ0 ( ) = x ( z x ), and ς ( θ) ˆ 0,, (5), D = E ( θ ) ( θ ), ( θ) g ( θ) ξ ( θ) ς ( θ) = 0 0 ( z xβ ) σ σ. =,,, (4) 907

4 .3. Hausman Tes for he BC Model The BC MLE is consisen under he i.i.d. and small σ assumpions. Therefore, we can es he null hypohesis consising of hese assumpions by he Hausman es using he N-esimaor. Under he null hypohesis, we ge, where ( ) 1 1,1 T ( λn λbc ) N( δ1 ) δ = elemen of ( A C ) B( A C ) ( ˆ ˆ ) ˆ N BC 1 N 0,1 T λ λ δ =. Since ( ) ˆ ˆ 0,, (6). Le ˆ δ 1 be he esimaor of δ 1 and under he null hypohesis, we can es using as he es saisic [5]. Noe ha we canno use wo or more parameers in he Hausman es in his case [18]. We can use he BC MLE if he null hypohesis is acceped. Nawaa and Kawabuchi [19] considered a furher es when he null hypohesis is acceped. Bu he null hypohesis consiss of compounded assumpions, and he furher es is no necessary if he null hypohesis is acceped. Noe ha we need oher esimaors and ess if he null hypohesis is rejeced [19]-[]. 3. Daa and Summary of he Revisions for Caarac Surgeries In his sudy, we use daa from he Secion of Healh Care Economics of Tokyo Medical and Denal Universiy. The daa were colleced from over 100 Japanese hospials from July 005 o March 01. For each paien, gender, age, daes of hospializaion, medical paymen amouns, DPC code, names of principle disease and disease ha caused hospializaion, Inernaional Classificaion of Diseases 10h revision (ICD-10) code of he WHO for he diseases, up o four surgeries, informaion of comorbidiies and complicaions, placemen afer hospializaion and oher informaion of paiens were available. Our original daase conained informaion of 9,766 caarac paiens and here were 56,364 one-eye caarac surgery cases. Alhough Naional Eye Insiue [8] recommended o a 4-8 week inerval beween surgeries when a paien needs caarac surgeries for boh eyes, boh eyes are operaed in a single hospializaion (hereafer, wo-eye surgeries) o a large number of paiens in Japan. 36.6% underwen wo-eye surgeries in our daase (For he res of he paiens, eiher DPC codes were no available or no surgeries were done). As menioned earlier, we analyzed paiens who underwen one-eye caarac operaions. The DPC codes, per diem paymens by he LOS and Specific Hospializaion Periods for hese paiens are given in Table 1. (In Japan, medical paymens are measured by poins and 10 yen per poin are paid o hospials.) In all hree revisions, hree periods deermined by he DPC/PDPS became shorer and per diem paymens became less. There was a clear inenion of he Japanese policy makers o reduce he LOS for caarac surgeries. The daase of 51,054 paiens from 60 hospials which had more han 300 caarac surgeries was used in he analysis. Table shows he average lengh of say (ALOS) by hospial. The ALOS of all paiens was 3.76 days Table 1. DPC codes and inclusive paymens for caarac paiens +. Periods DPC Code LOS and Per Diem Paymens (poins) Specific Hospializaion Period (days) x01x xx97x0x xx97x0x xx97xxx0 LOS 1s-3rd 4h-6h 7h-10h Per Diem Paymen LOS 1s-nd 3rd-4h 5h-8h Per Diem Paymen LOS 1s nd-3rd 4h-7h Per Diem Paymen LOS 1s-nd 3rd-4h 4h-6h Per Diem Paymen : The DPC/PDPS is revised in April of every second year. 908

5 Table. Disribuion of LOS by hospials. Hospial No. of paiens LOS No. of LOS Hospial ALOS S.D. paiens ALOS S.D. Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp43, Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp Hp All 51, wih sandard deviaion (S.D.) of 1.64 days. The maximum and minimum ALOS among hospials were 6.84 (Hp) and.07 (Hp0) days, respecively. The difference was 4.77 days and here were large differences among hospials. Table 3 shows he oal, inclusive and non-inclusive paymens per paien o hospials. The average oal cos was 9,491 yen wih S.D. of 45,60 yen, he average inclusive paymen was 74,886 yen wih S.D. of 36,09 yen, and he average non-inclusive paymen was 154,606 yen wih S.D. of,578 yen. The noninclusive paymen composed of abou wo-hirds and inclusive cos was jus one hird. However, S.D. of inclusive paymens is much bigger han ha of he non-inclusive paymens, and a large porion of he difference of 909

6 Table 3. Disribuion of medical paymens per paien by hospials (yen). Hospial Toal Inclusive Non-inclusive Hospial Toal Inclusive Non-inclusive HP1 4,199 83, ,451 HP31 1,33 63, ,341 HP 04,79 50, ,795 HP3,677 64, ,593 HP3 75, , ,579 HP33 3,553 6, ,045 HP4 6,73 99,871 16,851 HP34 08,510 63, ,15 HP5 19,0 67,85 151,350 HP35 1,968 60,918 15,049 HP6 9,684 69, ,198 HP36 43,604 8, ,193 HP7 6,709 70, ,19 HP37 40,610 81,51 159,089 HP8 16,97 61,8 155,015 HP38 73, , ,657 HP9 0,770 63,16 157,608 HP39 7,105 11, ,635 HP10 0,308 68,075 15,34 HP40,369 65, ,819 HP11 16,198 60, ,103 HP41 01,569 56, ,14 HP1 36,89 7, ,73 HP4 46,145 89, ,966 HP13 8,35 78,97 149,937 HP43 1,168 68,954 15,14 HP14 48,177 96, ,637 HP44 13,577 63, ,086 HP15 196,33 46,5 150,107 HP45 41,764 80,91 160,85 HP16 15,14 63, ,393 HP46 5,939 73,496 15,443 HP17 6,144 73,160 15,984 HP47 44,37 8, ,79 HP18 19,451 73, ,743 HP48 7,587 75, ,930 HP19 17,456 6, ,64 HP49 31,591 80, ,556 HP0 00,137 45, ,688 HP50 6,55 76, ,447 HP1 18,871 65, ,568 HP51 78,39 108, ,406 HP 84,58 17, ,534 HP5 0,903 64,59 156,311 HP3 37,39 81, ,70 HP53 41,561 79,98 16,63 HP4 16,553 63,89 15,74 HP54 46,518 84, ,700 HP5 9,000 78, ,0 HP55 49,04 90, ,198 HP6 31,6 76,43 155,018 HP56 87,360 19, ,805 HP7 31,6 76,43 155,018 HP57 50,477 95, ,686 HP8 0,438 6, ,750 HP58 38,497 8, ,909 HP9 0,863 68,050 15,81 HP59 194,05 50, ,77 HP30 1,134 68,856 15,78 HP60 9,15 79, ,550 All 9,491 74, ,606 he medical paymens caused by he inclusive paymens. The maximum oal, inclusive and non-inclusive paymens among hospials were 87,360 (Hp56), 19,554 (Hp56) and 164,406 (Hp51) yen, respecively. The minimum oal, inclusive and non-inclusive paymens among hospials were 194,05 (Hp59), 45,448 (Hp0) and 143,77 (Hp59) yen, respecively. The differences were 93,155, 84,106 and 5,634 yen, respecively. The differences of he inclusive paymens among hospials were much larger han hose of he non-inclusive paymens. 4. Resuls of Esimaion 4.1. Analysis of he LOS by he BC Mode Figure 1 shows he disribuion of LOS of one-eye caarac surgeries. Since he disribuion has a fa ail on he 910

7 Figure 1. Disribuion of he LOS. righ side, we use he BC model described in Secion 3. For he analysis of LOS, i is necessary o consider he characerisics of he paiens and he ypes of principal diseases. We also consider he effecs of advances and improvemens of caarac surgery echnologies. The explanaory variables used in he BC model were shown in Table 4. As he basic informaion of paiens, gender and age were considered. Under he Japanese mandaory public insurance sysem, he direc paymens of paiens was 10% for age 70 or over and 30% for younger han 70 in he sample period. So, he Over 70 dummy was added. Comorbidiy (number of comorbidiies), Complicaion (number of complicaions), and Emergency, Oupaien and Home dummies were used for represening condiions of paiens. The Summer and Winer dummies were used o evaluae seasonal effecs. To evaluae he effecs of hree revisions of he DPC/PDS, which is one he major purposes of he paper, we used hree (i.e., Afer 006, 008 and 010) dummies. For evaluaion of advances of medical echnology for caarac surgeries, Trend was added. Since we are analyzing a long erm effec (7 years), he rend migh have been changed and we added he squared erm of rend. The paymen goes back o he convenional fee-for-service base if he LOS exceeds he Specific Hospializaion Period, and per-diem paymen is no reduced afer he Specific Hospializaion Period. Over Period dummy was used o evaluae his effecs. Thireen ICD-10 dummies (base of hese dummies was H5.0), were used o represen he influences of principal diseases. We used sixy hospial dummies were used o evaluae he influences of he hospial direcly and a consan erm was no included in he model As a resul, No. of paiens LOS (days) x ijβ of Equaion (1) became x ijβ = β1female + βage + β3over 70 + β4comorbidiy + β5complicaion + β Emergency + β Oupaien + β Home + β Summer + β Winer ( ) + β11afer β1afer β13afer β14trend + β15 rend + β Over Period + β j-h ICD 10 dummy + β k Hp k dummy. j The resul of he esimaion is given in Table 5. The esimaes of he ransformaion parameers were ˆ λ BC = and ˆ λ N = The esimae of V ( ˆ λ ˆ ) BC λn = and we go = Therefore, he i.i.d. and small σ assumpions were acceped a he 5% significance level and we can use he BC MLE in his analysis. The esimae of Female dummy and Age were posiive and significan a 1% level. Alhough he esimae of Over 70 were posiive bu no significan a he 5%. The LOS became longer for females and for older paiens, bu we did no admi he effec of lower paymens for paiens age 70 or over in his sudy. The esimaes of (7) 911

8 Table 4. Definiions and summary of explanaory variables. Variable Definiion Summary of paiens Female Dummy variable; 1: female, 0: oherwise 1: 8,745 Age Age of paien mean: 73.4, S.D.: 10. Over 70 Dummy variable; 1: Age 70 or over, 0: oherwise 1: 36,93 Comorbidiy Number of comorbidiies, 0: 814, 1: 11,938, : 50, 3: 74, 4: 3068 Complicaion Number of complicaions 0: 33,993, 1: 10,99, : : 77; 4: 10 Emergency Dummy variable; 1: Emergency hospializaion, 0: oherwise 1: 901 Oupaien Dummy variable; 1: Oupaien, 0: oherwise 1: 48,67 Home Dummy variable; 1: reurn o home, 0: oherwise 1: 46,003 Summer Dummy variable; 1: summer +, 0: oherwise 1: 1,53 Winer Dummy variable; 1: winer ++, 0: oherwise 1: 10,147 Afer 006 Dummy variable; 1: Afer April : oherwise 1: 50,006 Afer 008 Dummy variable; 1: Afer April 008, 0: oherwise 1: 41,656 Afer 010 Dummy variable; 1: Afer April 010, 0: oherwise 1: 3,005 Trend No. of monhs from April 005 monhs from April 005 Over Period Dummy variable; 1: LOS exceeds he Specificaion 1: 1098 Hospializaion Period, 0: oherwise ICD 10 dummies H senile incipien caarac 4,0 paiens H51 1: senile nuclear caarac, 0 oherwise 1: 767 H5 1: senile caarac, morgagnian-ype, 0 oher wise 1: 535 H58 1: oher senile caarac, 0: oherwise 1: 395 H59 1: senile caarac, unspecified, 0: oher wise 1: 5398 H60 1: infanile and juvenile caarac, 0: oherwise 1: 88 H6 1: complicaed caarac, 0: oherwise 1: 161 H63 1: drug-induced caarac, 0: oherwise 1: 119 H64 1: Afer-caarac, 0: oherwise 1: 50 H68 1: oher specified caarac, 0: oherwise 1: 50 H69 1: unspecified caarac, 0 oherwise 1: 5938 H7 1: Oher disorders of lens (H71 & H7), 0: oherwise 1: 171 H8 1: diabeic caarac (H81 & H8), 0: oherwise 1: 733 Hospial Dummies Hp i 1: Hospial i; 0: oherwise + Summer: July and Augus, ++ Winer: December o February, +++ : base of ICD10 dummies. Comorbidiy and Complicaion were posiive and significan a he 1% and 5% levels, and comorbidiies and complicaions made LOS longer as expeced. Alhough he esimae of Emergency dummy was no significan, esimaes of Oupaien and Home dummies were negaive, posiive and significan a he 1% level. I is reasonable he LOS becomes shorer if a paien is an oupaien before hospializaion since he hospial is be able o 91

9 Table 5. Resuls of esimaion of he BC model for LOS by he BC MLE. Variable Coefficien Sd. Error -Saisic Variable Coefficien Sd. Error -Saisic lambda ** HP ** Female ** HP ** Age ** HP ** Over HP ** Comorbidiy ** HP ** Complicaion * HP ** Emergency HP ** Oupaien ** HP ** Home ** HP ** Winer HP ** Summer HP ** Afer HP ** Afer ** HP ** Afer HP ** Trend ** HP ** (Trend) ** HP ** Over Period ** HP ** H ** HP ** H ** HP ** H ** HP ** H ** HP ** H ** HP ** H ** HP ** H HP ** H * HP ** H ** HP ** H ** HP ** H ** HP ** H ** HP ** Hospial Dummies HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** 913

10 Coninued HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** HP ** ** : Significan a he 1% level, * :Significan a he 5% level, R = Definiions and summaries of variables are given in Table 4. conduc various medical checkups in advance. The resul of Home dummy implied ha LOS becomes shorer if paiens discharged o he place oher han heir home, and a problem of social hospializaion (a paien says a hospial wihou medical reamens because here is no place o go afer hospializaion, for deails, see Innami [3]) did no occur in caarac surgeries. Boh esimaes of Winer and Summer dummies were no significan a he 5% level, and he seasonal effecs were no admied. Among hree dummies which evaluaed effecs of he revisions, ha is one of he major subjecs of his sudy, only he esimae of he 008 dummy was significan a he 1% level and he oher wo were no significan. In he 008 revision, he Period I, he period hospials ge larges paymens, was shoren o jus one day. On he oher hand, he Period I was exended o wo days in 010 revision. So, his migh be an imporan facor affeced he behavior of hospials. As previously menioned, many paiens prefer shorer LOS. Therefore, shorening LOS no only reduces he medical paymens bu also becomes paiens benefis. The esimaes of Trend and (Trend) were significan a he 1% level and heir signs were posiive and negaive. This means ha was he LOS became shorer bu effecs becomes smaller as ime wen. The esimae of Over Period dummy was 1.17 and is -value was This means he effecs of he Specific Period was much larger han hose of oher explanaory variables. For he ICD-10 dummies, esimaes of H51, H59, H68 and H69 were negaive and significan a 1% level. H64 was negaive and significan a 5% level. On he oher hand, H5, H58, H6, H7 and H8 were posiive and significan a 1% level. These ypes of diseases affeced LOS. H63 was no significan and we did no admi he effec of his ype of disease. For he esimaes of he Hospial dummies, he maximum was.33 (Hp56) and he minimum was (Hp5); he difference was and was much bigger han he esimaes of mos of oher explanaory variables. This indicaed ha here remained large differences among hospials, even afer eliminaing he influences of variables including paien characerisics, disease ypes, hree revisions of he DPC/PDPS, ime rend, and Specific Hospializaion Period. 4.. Analysis of Non-Inclusive Paymens For he analysis of non-inclusive paymens per paien, n_ p, we used he ordinary leas squares mehod. The explanaory variables are he same as he previous case and he model is given by n_ p = β1female + βage + β3over 70 + β4comorbidiy + β5complicaion + β6emergency + β7oupaien + β8home + β9summer + β10winer (8) + β11afer β1afer β13afer β14trend + β15 ( rend) + β Over Period + β j-h ICD 10 dummy + β k Hp k dummy. 16 j We cloud no ge he paymen informaion for some paiens and he daase of 50,34 paiens was used in he analysis. The resuls of esimaion are given in Table 6. The esimae of Female dummy no significan a 5% 914

11 Table 6. Resuls of esimaion of he non-inclusive paymen model. Variable Coefficien Sd. Error -Saisic Variable Coefficien Sd. Error -Saisic Female HP17 17, ** Age ** HP18 16, ** Over HP19 175, ** Comorbidiy ** HP0 168, ** Complicaion ** HP1 164, ** Emergency ** HP 167, ** Oupaien ** HP3 174, ** Home ** HP4 166, ** Summer HP5 170, ** Winer HP6 17, ** Afer , ** HP7 169, ** Afer ** HP8 175, ** Afer HP9 171, ** Trend HP30 163, ** (Trend) HP31 173, ** Over Period 4, ** HP3 174, ** H HP33 178, ** H ** HP34 161, ** H ** HP35 161, ** H ** HP36 167, ** H HP37 17, ** H HP38 170, ** H HP39 175, ** H64 55, ** HP40 176, ** H HP41 164, ** H ** HP4 174, ** H7 110, ** HP43 17, ** H ** HP44 168, ** Hospial Dummies HP45 179, ** HP1 171, ** HP46 168, ** HP 17, ** HP47 177, ** HP3 174, ** HP48 165, ** HP4 175, ** HP49 170, ** HP5 169, ** HP50 167, ** HP6 179, ** HP51 183, ** 915

12 Coninued HP7 173, ** HP5 168, ** HP8 169, ** HP53 176, ** HP9 175, ** HP54 179, ** HP10 169, ** HP55 173, ** HP11 174, ** HP56 168, ** HP1 18, ** HP57 170, ** HP13 167, ** HP58 173,988 1, ** HP14 165, ** HP59 163,938 1, ** HP15 169, ** HP60 160,853 1, ** HP16 166, ** R ** : Significan a he 1% level, * :Significan a he 5% level. Definiions and summaries of variables are given in Table 4. level. Alhough he esimae of Age was negaive and significan, Over 70 was no significan a he 5%. The non-inclusive paymen became smaller for older paiens, bu we did no admi he effec of lower paymens for paiens age 70 or over in his sudy. The esimaes of Comorbidiy and Complicaion were posiive and significan a he 1% level, and comorbidiies and complicaions made non-inclusive paymens higher. Alhough he esimaes of Emergency and Oupaien dummies were no significan, esimaes of Oupaien and Home dummies were a he 1% level. Emergency hospializaion increased he non-inclusive paymens, bu a paien was an oupaien or reurning home reduced he non-inclusive paymens. Boh esimaes of Winer and Summer dummies were no significan a he 1% level, and he seasonal effecs were no admied. Among hree dummies which evaluaed effecs of he revisions, he esimaes of he 006 and 008 dummies were significan a he 1% level bu he signs were opposie. The 006 revision made he non-inclusive paymen lower and 008 revision made i higher. The esimaes of Trend and (Trend) were no significan and we could no find ime rends unlike he LOS case. The esimae of Over Period dummy was 4,34 and is -value was This means ha he LOS exceeded he Specific Period Hospializaion, no only LOS bu also non-inclusive paymens increased by a large amoun. For ICD-10 dummies, esimaes of H5, H58, H59, H69 and H7 were posiive and significan a 1% level. Especially, he esimae of H7 was 110,17 yen and very high. This was caused by he fac here were many paiens who received surgeries of vireous bodies wih caarac surgeries. On he oher hand, he esimae of H64 and H8 were negaive and significan a 1% level. The value of H64 was 55,15 yen. H64 is afer-caarac; ha is a paien ook a caarac surgery before and lens capsule becomes diry again, and he surgery jus makes he lens capsule clean. The paymens for his surgery was lower han regular caarac surgeries. For he esimaes of he Hospial dummies, he maximum was 183,41 (Hp44) yen and he minimum was 160,853 yen (Hp5); he difference was,559 yen or 14.6% of he average non-exclusive paymens. The difference was much smaller han ha of he LOS. 5. Discussion We found ha he differences of LOS among hospials were very large bu differences of he non-inclusive paymens (mainly paymens relaed o surgeries) were relaively small. Alhough he ypes of diseases were differen, we go he similar resuls for diabees [19] [1]. Anoher quesion is wheher he LOS and he non-inclusive paymens were relaed or no. The correlaion coefficien of he esimaes of hospial dummies beween he LOS and non-inclusive paymen models was 0.4 and -value for esing no correlaion is We could no say ha he LOS and he non-inclusive pay were correlaed. By he hree revisions, he periods deermined by he DPC/PDPS were shorened and per diem paymens were reduced. This means ha here was a srong poliical implicaion o reduce he LOS o conrol he medical paymens. Among hree revisions, only one revision significanly reduced he LOS. The resuls of his sudy shows ha he LOS did no decrease much. However, he LOS could be reduced by he effors of hospials. For example, Kobao e al. [4] repored ha he LOS was 916

13 reduced by inroducion of proper clinical pahways. More recenly, he Minisry of Healh, Labour and Welfare [5] released he Japanese medical paymens reached 40 rillion yen in fiscal year 014 (Japanese fiscal year is from April o March), his figure is expeced o increase as aging he populaion in he fuure. The medical paymen has become a big financial problem. As already poined ou by Nawaa and Kawabuchi [1], he bes answers for his problem is o rea paiens more efficienly and conrol paymens wihou degradaion of reamens. There were large differences in ALOS among hospials even afer eliminaing effecs of various facors. This suggesed ha i migh be possible for many hospials o reduce he LOS wihou degradaion of reamens for caarac surgeries. In he revision of he medical paymen sysem implemened April 014 [6], he caarac surgery is classified under he caegory of he Shor Say Operaion Basic Paymen 3. If a hospial and a paien saisfy he required condiions, a hospial ges he same amoun of paymens. In caarac surgeries wih lens inserion, a hospial ge 70,930 yen if he LOS is 5 days or less. I is essenially same as he DRG/PPS (Diagnosis-Relaed Group/ Prospecive Paymen Sysem) widely used in he Unied Saes and oher counries 1). The numbers of diseases, hospials and paiens under his sysem are currenly small, i will be expeced o increase in he fuure. Therefore, evaluaion of he new (DRG/PPS ype) paymen sysem is an imporan subjec for he financial susainabiliy of he Japanese medical paymen sysem. 6. Conclusions In his paper, we conduced a long erm survey of he caarac surgeries. The sample period was abou 7 years, from July 005 o March 01. We evaluaed he effecs of hree revisions of he medical paymen sysem ha were done in 006, 008 and 010. Abou one million surgeries are performed in Japan annually. For he analysis, he Box-Cox ransformaion model and Hausman es using Nawaa s esimaor are used for he LOS, and he ordinary leas squares mehod is used for he non-inclusive paymens. To evaluae hese changes, we analyzed a daase of 51,054 paiens obained from 60 hospials (Hp1-60) where surgeries more han 300 one-eye caarac surgeries were performed during he period. The ime rend reduced he LOS bu degrees of reducion became smaller. For he analysis of he LOS, gender, age, numbers of comorbidiies and complicaions, oupaien before hospializaion, and place o go back afer hospializaion were significan variables. The ime rend and squared of rend were significan and he LOS became shorer bu effecs became smaller as ime wen. For ICD-10 dummies, H64 and H8 were negaive and significan. H5, H58, H6 and H7 were posiive and significan. The Specific Hospializaion Period also srongly affeced he non-inclusive paymens. Among hree dummies which evaluaed effecs of he revisions, only he esimae of he 008 dummy was significan and he 008 revision reduced he LOS bu no he oher wo. There were large differences among esimaes of he hospial dummies, indicaing ha here remained large differences among hospials, even afer eliminaing he influences of various facors. For he analysis of he non-exclusive paymens, gender, numbers of comorbidiies and complicaions, being an oupaien before hospializaion, place o go back afer hospializaion, emergency were significan variables. The effecs of ime rend was no admied in his case. For ICD-10 dummies, H5, H58, H59, H64, H69, H7 and H8 were significan variables. As he LOS case, he Specific Hospializaion Period also srongly affeced he non-inclusive paymen. For he effecs of he revisions, 006 and 008 dummies were significan bu he signs were opposie. The 006 revision made he non-inclusive paymen lower bu 008 revision made i higher. For he esimaes of he Hospial dummies, he larges difference was,559 yen or 14.6% of he average non-exclusive paymens. The difference was much smaller han ha of he LOS. In his sudy, we analyzed he LOS and medical paymens for caarac surgeries. For financial susainabiliy of he Japanese medical sysem, i is necessary o evaluae oher diseases. Evaluaion of he Shor Say Operaion Basic Paymen sysem expanded in 014 is also necessary. In his paper, characerisics of hospials were no analyzed. Barbieri e al. [7] repored ha he LOS of privae hospials was shorer han ha of public ones. These are subjecs o be sudied in he fuure. Noe: 1) Alhough 10 hospials (8 naional and social insurance hospials) had adoped he DRG/PPS on a rial basis from 1998 o 004, i was no acceped by he Japanese medical sociey. Acknowledgemens The Insiuional Review Boards of he Universiy of Tokyo (number: KE1-7) and Tokyo Medical and Denal 917

14 Universiy (number: 839) approved he use of his daase. An earlier version of he paper was presened a he Inernaional Symposium on Recen Developmens in Economeric Theory wih Applicaions in Honor of Professor Takeshi Amemiya (June 0-1, 015) a Xiamen Universiy in China. We would like o hank an anonymous referee and he paricipans of he symposium for heir helpful commens. We also hank various hospials for heir sincere cooperaion. References [1] DPC Evaluaion Division, Cenral Social Insurance Medical Council (010) Heisei 4 nendo kaie ni mukea DPC seido (DPC/PDPS) no aiou ni suie (Concerning he Procedures for he 01 Revision of he DPC Sysem (DPC/PDPS)). (In Japanese) [] Nawaa, K., Ii, M., Toyama, H. and Takahashi, T. (009) Evaluaion of he Inclusive Paymen Sysem Based on he Diagnosis Procedure Combinaion wih Respec o Caarac Operaions in Japan. Healh, 1, hp://dx.doi.org/10.436/healh [3] Nawaa, K. and Kawabuchi, K. (015) Evaluaion of Lengh of Hospial Say Joining Educaional Programs for Type Diabees Mellius Paiens: Can We Conrol Medical Coss in Japan? Healh, 7, hp://dx.doi.org/10.436/healh [4] Cenral Social Insurance Medical Council (014) DPC aishou hyouin Junbi hyouin no kibo (heisei 6 nen 4gasu 1 nichi) mikomi (Esimaed Sizes of DPC Hospials and Preparing Hospials as of April 1s, 014). (In Japanese) [5] Minisry of Healh, Labour and Welfare (014) Shakai iryou shinryu koui besu chosa hesiei 5 nendo (Social medical Surveys by Treamen Pracices in Fiscal Year 013). (In Japanese) [6] Nawaa, K. and Kawabuchi, K. (015) An Economeric Analysis of Hospial Lengh of Say for Diabees Operaions in Japan by he Box-Cox Transformaion Model and Hausman Tess: Evaluaion of he 010 Revision of he Medical Paymen Sysem. Open Journal of Applied Science, 5, hp://dx.doi.org/10.436/ojapps [7] OECD (013) Healh a a Glance 013. hp:// [8] Naional Eye Insiue (009) Facs abou Caarac. hps://nei.nih.gov/healh/caarac/caarac_facs [9] Jacobs, J. and Koch, P. (009) Your Caarac Surgery Procedure & Operaion. Medical Managemen Services Group. hp:// [10] Trivedi, N.S., Cheah, E., Heyman, C. and Qiu, C. (00) Reducing he Lengh of Hospial Say of Caarac Surgery: One Year Experience in a HMO Seing. Aneshesiology, 96, A1137. hp://dx.doi.org/ / [11] Aalla, M.L., Wells, K.K., Peucker, K., e al. (000) Caarac Exracion in a Major Ophhalmic Hospial: Day-Case or Overnigh Say? Clinical and Experimenal Ophhalmology, 8, hp://dx.doi.org/ /j x [1] Fedorowicz, Z., Lawrence, D.J. and Guie, P. (006) A Cochrane Sysemaic Review Finds No Significan Difference in Oucome or Risk of Posoperaive Complicaions beween Day Care and In-paien Caarac Surgery. Saudi Medical Journal, 7, [13] Ingram, R.M., Banerjee, D., Traynar, M.J. and Thompson, R.K. (1983) Day-Case Caarac Surgeries. Briish Journal of Ophhalmology, 67, hp://dx.doi.org/ /bjo [14] Box, G.E.P. and Cox, D.R. (1964) An Analysis of Transformaions. Journal Royal Saisical Sociey Series B, 6, [15] Bickel, P.J. and Doksum, K.A. (1981) An Analysis of Transformaions Revisied. Journal of American Saisical Associaion, 76, hp://dx.doi.org/ / [16] Nawaa, K. (013) A New Esimaor of he Box-Cox Transformaion Model Using Momen Condiions. Economics Bullein, 33, [17] Hausman, J. (1978) Specificaion Tes in Economerics. Economerica, 46, hp://dx.doi.org/10.307/ [18] Nawaa, K. and McAleer, M. (014) The Maximum Number of Parameers for he Hausman Tes When he Esimaors Are from Differen Ses of Equaions. Economics leers, 13, hp://dx.doi.org/ /j.econle [19] Nawaa, K. and Kawabuchi, K. (015) Financial Susainabiliy of he Japanese Medical Paymen Sysem: Analysis of he Japanese Medical Expendiure for Educaional Hospializaion of Paiens wih Type Diabees. Healh, 7, hp://dx.doi.org/10.436/healh [0] Nawaa, K. (015) Robus Esimaion Based on he Third-Momen Resricion of he Error Terms for he Box-Cox Transformaion Model: An Esimaor Consisen under Heeroscedasiciy. Economics Bullein, 3,

15 [1] Nawaa, K. and Kawabuchi, K. (016) Comparison of he Lengh of Say and Medical Expendiures among Japanese Hospials for Type Diabees Treamens: The Box-Cox Transformaion Model under Heeroscedasiciy. Healh, 8, hp://dx.doi.org/10.436/healh [] Nawaa, K. and Kawabuchi, K. (016) Did he Revision of he Japanese Medical Paymen Sysem Work Properly? An Analysis of Averages and Variances of Lengh of Hospial Say for Type Diabees Paiens by Individual Hospial. Healh, 8, hp://dx.doi.org/10.436/healh [3] Innami, I. (009) Sakaieki nyuuin no kenkyu (Sudies of Social Hospializaion). Toyo Keizai Shinpo Sha, Tokyo. (In Japanese) [4] Kobao, T., Nakamura, A., Wakiya-Masumoo, J., e al. (006) Effecs of Implemening Clinical Pahways for Care of Paiens Undergoing Ophhalmic Surgery for Caarac, Glaucoma, and Vireoreinal Disorder. Nippon Ganka Gakkai Zasshi (Japanese Journal of Ophhalmology), 110, (In Japanese wih English Absrac) [5] Minisry of Healh, Labour and Welfare (015) Heisei 6 nendo iryou no doukou Gaisan iryouhi no nendo shuukei kekka )The Trend of Medical Paymens in Fiscal Year 014 Resuls of Esimaed Medical Paymens of he Fiscal Year). (In Japanese) [6] Cenral Social Insurance Medical Council (014) Toushin-sho (Heisei 6 nendo shinnryou houshuu kaiei ni suie) (The Repor for he Revision of Medical Paymens in Fiscal Year 014). (In Japanese) [7] Barbieri, V., Schmid, E., Ulmer, H. and Pfeiffer, K.P. (007) Healh Care Supply for Caarac in Ausrian Public and Privae Hospials. European Journal of Ophhalmology, 17, Submi your manuscrip a: hp://papersubmission.scirp.org/ 919