AN APPOINTMENT ORDER OUTPATIENT SCHEDULING SYSTEM THAT IMPROVES OUTPATIENT EXPERIENCE Yu-L Huang Industral Engneerng Department New Mexco State Unversty Las Cruces, New Mexco 88003, U.S.A. Abstract Patent wat tme and access to care have long been a recognzed problem n modern outpatent healthcare delvery systems. Despte all the efforts to develop appontment rules and polces, the problem of long patent wats perssts. Regardless of the reasons, the fact remans that there are few mplemented models for effectve schedulng that consder patent wat tmes, physcan dle tme, overtme, ancllary servce tme, as well as ndvdual no-show rate, and are generalzed suffcently to accommodate a varety of outpatent clnc settngs. The goal of ths research s to mprove the qualty and effcency of healthcare delvery by developng a patent schedulng system that meets the clncal polces wthout overbookng whle usng an nnovatve wat rato concept, a patent arrval schedule from the physcan schedule accountng for ancllary servces, an evdence-based predctve model of no-show probablty for ndvdual patent, and a model-supported dynamc overbookng polcy to reduce the negatve mpact of no-shows. Ths research provdes a step-bystep method for mplementaton of a schedulng model n outpatent clncs. Consequently, ths research wll mprove the outpatent experence for both patents and medcal provders, ncrease patent access to care, and ultmately enhance the qualty of care. Learnng Objectves The learnng objectves are: 1. Understand the fundamental reasons and shortcomngs of the current schedulng systems. 2. Defne an effectve patent central schedulng model that meets the clnc polces. 3. Develop and extend the proposed model to consder ancllary servces. 4. Learn and evaluate the mplementaton of the proposed model. 5. Develop an overbookng polcy to reduce the negatve mpact of no-shows. Introducton Patent wat tme and access to care have long been the recognzed problems n modern outpatent health care delvery systems. As competton has ncreased for lmted health care dollars, efforts have been made to ncrease effcency and reduce costs, yet only lmted gans have been made n terms of reducng patent wat tme and mprovng patent access to care. Ironcally, one of the man strateges to decrease cost has been to shft tradtonally npatent servces to an outpatent settng, whch tremendously ncreases the burden on outpatent facltes to effcently manage health care delvery. Due to the ncreasng outpatent clncs, patent wat and access to care have become two of most crtcal measurements for patents to choose ther physcans. However, lmted progress has been made n the systematc mprovement of patent wat and access to care, maybe because most facltes have focused more on the effcent schedulng of provder tme snce percevably they have a hgher value than the patents tme. The overbookng and shorter scheduled treatment tme nterval polces are the typcal approaches used to prevent physcans from dlng. The unrealstc estmaton of treatment tme ntervals at present s mostly of clncs that are stll n a physcan-centrc envronment. In ths study, even one mnute off from an estmaton of a treatment tme nterval wll have sgnfcant compoundng effects to patent watng. In addton, the naccurate defntons of appontment tme verses arrval tme, or nadequate defnton of what s ncluded n physcan treatment tme, evdently ncreases patent wat tme and decreases patent access to care. There s no clear delneaton between the physcan schedule and the patent arrval schedule. Cayrl and Veral (2003) conducted an extensve revew of the lterature on schedulng. They concluded the lmtatons on appontment rules are manly the lack of mplementaton and a generalzed model that reflects realty. Ths paper provdes an effectve schedulng method that consders patent wat tme, physcan dle tme, overtme, ancllary servce tme, as well as ndvdual no-show probablty to mprove healthcare delvery by reducng patent wat tme and mprovng patent access to care. Ths paper conssts of methods for defnng schedule tme nterval for the development of a physcan schedule, accountng for ancllary servces to develop a
patent arrval schedule, and ncorporatng overbookng to reduce the negatve mpact of patent no-shows. General Modelng of Patent Flow The general model for patent flow n and out of outpatent clncs has been developed by many researchers such as Ho and Lau (1992). Let T be physcan servce tme for patent, X be the scheduled tme nterval, S be the scheduled startng tme for patent, F be the fnsh tme for patent, A be the actual startng tme for patent, W be the wat tme for patent, and P be the physcan dle tme watng for patent. A graphcal example for patent flow s demonstrated n Fgure 1. Fgure 1. General patent flow Then the average patent wat tme, W, and the average physcan dle tme, P, can be calculated. Ths general model s used as a bass for our smulaton model. Physcan Schedule Development A successful appontment system should mnmze patent delays whle fully utlzng medcal resources. However, there s a tradeoff n that reducng patent wat tme may ncrease physcan dle tme and vce versa. Hence, ths paper proposes a wat rato approach that balances patent wat tme and physcan dle tme. Ths allows us to determne the treatment tme nterval of each vst type for a gven wat rato. Determnng the treatment tme ntervals across dfferent vst types by the same wat rato s to ensure every patent s treated equally n terms of watng. The best scheduled tme nterval s defned as the maxmum scheduled tme nterval for each vst type that satsfes medcal and clnc constrants. Let X d, mean µ and standard devaton σ are the parameters of treatment tme T. d s the number of standard devaton away from the mean. The Wat Rato, R, s defned as the degree to whch average patent wat tme W exceeds R physcan dle tme P ; W P R. The goal s to fnd the optmal R that satsfes the clnc constrants. Once the optmal wat rato R s reached, then the optmal tme nterval X can also be calculated for each vst type. To assure that our model does reflect actual clnc operatons, certan constrants need to be consdered n generatng the optmal tme nterval for each vst type. The proposed model does not am at forcng the clnc settng to ft the soluton, but rather ams at fully utlzng the avalable resources and capacty to acheve the best soluton. The clncal constrants are used to determne a rato to replace conventonal cost ratos, n large part because physcans tend to overestmate the cost of ther tme as opposed to patent tme, most lkely due to ther lack of crtera for accurately evaluatng tme cost for the patent to clncal management. Ths method effectvely elmnates the cost of patent tme and the bas nherent n cost ratos from the model n favor of well defned constrants. The underlyng constrant s to ensure the optmal tme nterval has a less than 50% chance of patent watng, Pr T X 0. 5. Asde from the constrant of the probablty of a patent wat n the absence of a pror wat, there are other clncal condtons, generally admnstratve constrants establshed by management, that can affect schedulng tme ntervals that need to be consdered, such as clnc or sesson fnsh tme, tme of last appontment, number of patents to be seen n a gven sesson, average patent wat tme, and average physcan dle tme. Once the optmal tme ntervals are determned, a physcan schedule can be constructed. An example from an Orthopedc Surgery Clnc, there are three major vst types used: Follow-up patent (FU), New Patent (NP), and Patent requrng X-ray before seeng physcan (XR). The current schedule s shown n Fgure 2. The physcan treatment tme data summary s shown n Table 1. Table 1. Physcan Treatment Tme Summary Three clnc constrants are addressed: Sesson should be fnshed by 11:30 am. Last patent should be scheduled by 11:00 am. There should be 25 patents scheduled n a sesson. From the smulaton optmzaton model, the R that satsfes clnc constrants s 18, whch also stratfes the underlyng constrant of Pr T X 0.57 0. 5.
Therefore, the optmal tme ntervals X for FU, NP, and XR are 7.4, 10.6, and 5.3, respectvely. Then the physcan schedule s constructed as shown n Fgure 2. Patent Arrval Schedule Development Once a physcan schedule s establshed, then the correspondng patent arrval schedule must be determned. The man concept behnd the arrval schedule s to provde suffcent tme between the patent arrval at the clnc and the actual examnaton tme for the patent to complete actvtes and requred pre-vst actvtes such as sgnng n, fllng out paperwork, havng vtals taken, havng an x-ray taken, provdng a specmen, and movng between lab or x-ray room and exam room. The tme assgned to pre-vst actvtes wll dffer from clnc to clnc and between specaltes. However, f the tme needed for these actvtes s not well defned, wat tme wll be compounded for ether physcan or patents. Ideally, the physcans should be able to mantan ther schedules wthout contrbutng sgnfcantly to patent wat tme. Ancllary servces such as x-ray, lab test, dagnoss vascular studes, and electrocardogram (EKG) are mportant to assst the physcan n makng an accurate assessment and are most lkely requred pror to a physcan vst n many outpatent clncs. The focus here s for clncs that prefer both ancllary and physcan servces to be done n one vst. Patent wat tme ncreases by both underestmatng and overestmatng the length of tme requred for ancllary servces. When ancllary servce tme s overestmated, ncreased wat tmes result from early completon before the physcan s ready for the patent. When ancllary servce tme s underestmated patents experence delays n beng seen by the physcan. The tme assgned to pre-vst actvtes wll dffer from physcan to physcan. Hence, the best scheduled ancllary servce tme nterval to be consdered n schedulng should be determned accordng to ndvdual physcan s schedule and needs. The goal s to mnmze the patent wat tme by determnng the best scheduled ancllary servce tme nterval n schedulng systems. Contnung wth the example from the Orthopedc Surgery Clnc, the man ancllary servce s x-ray. From the data, x-ray tme can be estmated by a Gamma dstrbuton wth mean of 6.3 mnutes and standard devaton of 3.7 mnutes. The clnc s currently askng patents who are requred to have an x-ray before seeng the physcan to come n 10 mnutes earler. Two observed problems are: 1. Ths 10-mnute tme s not accounted n the schedulng system so that patents tend to dsregard t and arrve at the scheduled tme. 2. There s no bass for ths 10-mnute decson. From the smulaton optmzaton model that mnmzes the average patent wat tme, the best scheduled tme nterval of the ancllary servce s found to be 5 mnutes. Therefore, on top of 10 mnutes for all pre-vst actvtes such as sgnng n, a 5-mnute early arrval s added to XR patents. Hence, f a patent s scheduled to see the physcan at 9:00 am and s requred to have x-ray taken, then ths patent should arrve at 8:45 am to accommodate 10-mnute pre-actvty and 5-mnute x-ray; see Fgure 2 for patent arrval schedule. Furthermore, 23% of NP and 21% of FU from data requred x-ray. After ntervewng wth medcal assstants and nurses, some rules were found: Patents who have had jont replacement wll need to have x-rays at post operaton, 3 month check, and 1year check scheduled under FU patent slots. After a year, f a patent calls n and complans of pan, they would be scheduled as FU and have an x-ray. Patents who have had bone dsplacement or fracture that was manpulated or operated on n the hosptal wll need an x-ray and are scheduled as FU. All new patents wll need x-rays f they have not had one done elsewhere. A specal case s new patents wth an ndcaton of arthrts n ther knee; f ther x-ray only has two vews, then they wll need to have an x-ray for two addtonal vews. Patents who are over 60 years old wll need x-rays. If those crtera are transferred to the schedulers, then the x-ray tme can be taken nto account when schedulng an appontment so that the frst consultaton tme and the patent wat tme can be reduced to mprove patent access to care and servce qualty. Fgure 2. The proposed physcan and patent schedules
Implementaton Results The approach thus far has been mplemented n three clncs. They are Orthopedc Surgery, Plastc Surgery, and Vascular Surgery clncs. The results ndcate the sgnfcant reducton on patent wat tme as much as 56%; see Table 2. As for physcan dle tme, t was found that there s not a sgnfcant dfference. Table 2. Implementaton Results for Three Clncs In mnutes Orthopedc Plastc Vascular Before 27.8 15.0 27.8 After 13.1 7.5 12.4 Reducton 53% 50% 56% Overbookng Polcy One of the major ssues that has been wldly studed by researchers s no-shows. Clnc no-shows are when a patent does not arrve to a prevously scheduled clnc appontment. Ths s problematc for multple reasons. For example, patents n need of an urgent clnc appontment cannot be seen when the schedule s full even f a last mnute openng occurs due to a no-show. It also deprves the clnc of needed revenue snce an empty vst slot results n non-bllable "down tme". One way to help avod ths problem s to overbook appontments, whch s where more than one patent s scheduled at the same tme. Ths can also create sgnfcant problems because f all patents do arrve for ther appontments the wat tmes can be very long, whch may lead to clnc overtme, and wll result n sgnfcant patent dssatsfacton. Many reasons for no-shows have been studed and reported such as patent moblty and physcan specalty. One schedulng approach developed to reduce no-shows s known as Open Access used by Murray and Tantau (1999). Open Access approach uses the phlosophy of dong today s work today. Ths approach seems to mprove patent access and no-show rate, but practcally burdens patents n attemptng to get appontments, whch may drve away patents and leads to proft losses. Due to the nature of ths approach, overtme s allowed to occur when the demands are hgh. Ths, n turn, ncreases the cost for clncs and burdens clncs management. Ths approach dd not focus on reducng patent wat tme, but exclusvely attempted to fll up the clnc day so that the resources could be fully utlzed. Ths paper proposes an approach usng the tradtonal appontment order schedulng to cautously overbook patents nto appontment slots by understandng scheduled patents no-show behavors so that the total costs (patent wat tme, physcan dle tme and overtme) can be mnmzed. There are two major components here. Frst, a statstcal predcton model wll determne the probablty of no-shows gven patent characterstcs and consdered wth other envronmental factors such as the predctve model bult by Glowacha et al (2009). Then, a smulaton optmzaton model wll fnd the optmal threshold of no-show rate obtaned from the predcton model to mnmze the total costs. Let be the predcted probablty of no-show for patent j p, j at appontment slot and p be the threshold of no-show rate for each slot. At any tme slot, overbookng can be performed as long as the jont probablty at a slot s stll greater than the threshold, p, p. Let W be the j j total patent wat tme, P be the total physcan dle tme, and O be the overtme for a clnc day. Gven c s the w cost of patent wat tme, c s the cost of physcan dle p tme, c s the cost of overtme, and o c s the total cost T of patent wat tme, physcan dle tme and overtme. Therefore, c c W c P c O. The objectve s to T w p fnd the optmal value for the threshold of no-show rate p that mnmzes c. Therefore, f the smulaton T optmzaton model determnes p s equal to 0.32. Ths means that clncs wll overbook a patent where the predcted probablty of no-show scheduled n a tme slot s greater or equal to 0.32. Ths paper presents a theoretcal concept for a dynamc approach whch accommodates overbookng nto a patent schedulng system based on the predcton of an ndvdual patent s no-show probablty. Ths approach s unlke the tradtonal bult-n overbookng approach that does not account for an ndvdual patent s condtons and tends to favor a physcans tme. Although the foundaton of ths proposed approach s based on the predcton of patent s no-show probablty, whch s more lkely case-by-case dfferent, the approach tself s generalzed enough for any patent schedule to be mplemented. Each clnc determnes the least amount of patents to schedule wthout overbookng and then uses the proposed overbookng approach to fnd where and how many to overbook based on the objectves of mnmzng the total cost ncludng the costs of patent wat tme, physcan dle tme and overtme. Implementaton Steps and Outcomes To mplement the proposed approach, steps from data collecton, model development ncludng determnng the optmal treatment tme and ancllary servce tme ntervals and the overbookng optmal threshold level, to mplementaton wth antcpated actvtes and expected outcomes s explaned; see Table 3. o
Table 3. Project Actvtes and Antcpated Outcomes servces, and no-show rates to accommodate the dfference of each clnc or physcan s practce. Concluson Summary of Research Desgn Ths approach uses an nterestng concept of wat rato nstead of tradtonal cost ratos of patent wat tme and the physcan dle tme. As many researchers have used tradtonal cost ratos, the ssue of the long patent wats n a physcan s offce stll exsts. In addton, ths research demonstrates an approach for patent schedulng developed to reduce patent wat tme, enhance patent flow, and mprove patent access to care, wthout sgnfcantly ncreasng physcan dle tme. The approach allows clnc management to quckly determne the best scheduled tme nterval for dfferent vst types and then ntegrate clncal constrants to construct two schedulng templates: physcan and patent arrval. Separatng the two schedules wll make t possble to create a template for patent arrval that accommodates any patent processng tasks or ancllary servces that need to be conducted n conjuncton wth a gven physcan servce. Ths wll reduce the unnecessary frst consultaton and mprove the tmely delvery and patent access. In addton, the common ssue of no-shows s consdered nto the schedulng system and the approach provdes a dynamc way of overbookng patents that mnmzes total costs. The conceptual and theoretcal framework of ths paper allows the changes of nput parameters such as provder treatment tme, ancllary servce tme(s), number of patents seen and no-show rates among physcans n a varety of clnc settngs. Ths approach allows clnc management to nput ts own parameters such as physcan treatment tme, schedulng sequence of vst types, types and tmes of ancllary Wat rato s a novel concept of the relatonshp between patent wat tme and physcan dle tme. Ths concept provdes an advantage over the tradtonal cost rato to prevent t from beng the physcan centrc soluton to ensure patents beng treated equally regardless ther vst types and condtons, as well as full utlzaton of clnc resource capacty. The man reason that cost ratos favor physcan wat tme s due to the defnton of cost values assgned to both patent wat tme and physcan dle tme. Realstcally, the cost of physcan dle tme s much hgher than that of patent wat tme. Hence, the hgh cost rato between physcan and patent wat tme becomes the most domnant factor when decdng patent schedulng tme ntervals. Therefore, t s much more objectve to consder wat tme tself nstead of placng a cost value to wat tme whle desgnng schedule tme ntervals. By adoptng the concept of wat rato ths allows a clnc to actually take nto account clnc constrants for determnng an appontment schedule to generate a better patent flow. Secondly, the concept of two schedulng systems, physcan and patent arrval schedules, has not been wldly consdered when schedulng. Accordng to the fndngs from ntervewng clnc management regardng what the patent scheduled tme means, the general answer has been the scheduled tme s the tme for a patent to see the physcan. Ths ndcates that the preactvtes such as x-ray are completely gnored, whch tends to compound the mpact of watng from the begnnng of a clnc sesson. Incorporatng two schedules makes t possble to account for any ancllary servces before seeng the physcan to customze patents needs and condtons. In addton, most clncs assume the same amount of tme to complete the requred ancllary servces, for example, 10 mnutes for any patents who needs x-rays. Ths assumpton does not account for the varaton of ancllary servce tme at all, whch generates watng for both patents and physcans. Therefore, consderng ndvdual needs for ancllary servces s mportant for a better schedulng system. Thrd, even though overbookng approaches have been wdely used n most schedulng systems, most of the overbookng slots are desgned nto the schedulng systems before actually schedulng patents. Ths tradtonal approach does not consder the ndvdual noshow behavor, wthout understandng patents characterstcs and preferences. Ths means the majorty of the tme when two patents actually show up at the same tme, one of them s bound to wat for servce. Ths paper provdes an overbookng concept that accounts for the ndvdual patent s probablty of no-show based on
patent characterstcs (age, gender, locaton, moblty) and preferences ncludng envronmental factors (weather and traffc condtons). In concluson, one of the major concerns from lterature s the gap between theoretcal concepts and feasblty n realty. Ths paper provdes a soluton for executon of the proposed approach, conceptually and theoretcally, from an ntal data collecton and model development to mplementaton of the approach. Ths approach wll not only close the gap between theory and practces but also provde evdence of applcablty for any physcans n any specalty. In short, the successful demonstraton of ths proposed approach wll shft the percepton of long wats n a physcan s offce, transform outpatent envronment to be much more pleasant, and ultmately mprove the outpatent experence. References Cayrl T and Veral E (2003). Outpatent Schedulng n Health Care: A Revew of Lterature. Prod Opns Mngt 12(4): 519-549. Ho C and Lau H (1992). Mnmzng Total Cost n Schedulng Outpatent Appontments. Mngt Sc 38(2): 1750-1764. Glowacka K, Henry R and May J (2009). A hybrd data mnng/smulaton approach for modellng outpatent no-shows n clnc schedulng. J Opl Res Soc 60(8): 1056-1068. Murray M and Tantau C (1999). Redefnng open access to prmary care. Managed Care Quarterly 7(3): 45-55. Bographcal Sketch Dr. Huang joned the Department of Industral Engneerng at New Mexco State Unversty as an Assstant Professor n 2009. He earned hs Ph.D., M.S.E. and B.S.E degrees from Industral & Operatons Engneerng Department at the Unversty of Mchgan, Ann Arbor, n 2008, 2007 and 2000, respectvely. Dr. Huang s research nterests focus on process mprovement usng smulaton optmzaton modelng methods n health care delvery systems ncludng outpatent schedulng system, pharmacy layout desgn, radology operaton mprovement, emergency delvery system, and operaton room schedulng system. He s also nterested n cancer preventon decson ad development. Dr. Huang has three years of experence as a project facltator for schedulng n outpatent clnc settngs supplemented by three years of experence as busness analyst n an ndustral suppler and two years of experence as an ndustral engneer n manufacturng systems.