2 Outline Simulation primer and OP clinic example Clinic flow, measures, issuesOpen accessMathematics of appointmentsInformation systemsClinic operations analysis cases
3 Simulation for Managers Many healthcare systems horribly complexDifficult to estimate impact of changes to system on performanceMuch easier and less expensive to experiment with a model instead of the real systemDiscrete even simulation allows capture of variability and complex interactions in systemsHanded out two nice introductions to computer simulation for healthcare managers a few weeks ago:Benneyan, J.C., An introduction to using computer simulation in healthcare: patient wait case studyMahachek, A.R., An introduction to patient flow simulation for health-care managersExample: An outpatientclinic simulation model
4 Simulation for Managers Basic components of a simulation study:Study real system to understand problem and need for simulationDevelop model of real system using simulation softwareConcurrently collect data on key inputs to simulation model (e.g. processing times, arrival rates) as well on on outputs (wait times) if possibleVerify and validate modelIterate through above 3 steps, with user involvement, until everyone satisfied model is reasonable representation of realityConduct controlled experiments with simulation model by running it for various combinations of input valuesStatistically analyze the output from the simulation experiments to draw conclusions, gain insights, support decision makingSoftware – MedModel (ServiceModel), ProcessModel, Arena, Extend, GPSS, see
5 Generic Flow ModeledWait for ProviderInitial wait
6 Using Simulation to Support Capacity Planning - Research Ran set of simulation experiments for range of volumes, exam times, staffing levels, rooms/doc, prep locationestimate initial wait time, wait time for provider, total time in clinic, length of clinic sessionDeveloped simple spreadsheet based model using Pivot Tables to find max volume subject to constraints on patient waiting and clinic lengthThe data is output from the simulation experiementsCurrently developing regression and neural network based prediction models from the simulation experimental outputDeveloping decision support toolsFamPractice_v5.xls, ClinicWhatIfLookup-v4-Example.xlsif interested in collaboration, please contact me
8 Interest in Clinic/Office Operations & Management IHIs initiative (started 1999) on the “Idealized Clinic Office Practice”
9 Improving Chronic Illness Care Higher level viewA Robert Wood Johnson Foundation programBodenheimer, Wagner, & Grumbach (2002) Improving primary care for patients with chronic illness, JAMA 288(14),Bodenheimer, Wagner, & Grumbach (2002) Improving primary care for patients with chronic illness: The chronic care model, Part 2, JAMA 288(15),
10 Some Operational Inputs and Outputs Performance MeasuresInput/Decision VariablesQuality of careAppointment Lead TimePatient Wait Time – initial, for provider, repeat waitsPatient Time in ClinicLength of clinic dayExam Room UtilizationSupport Space UtilizationProvider and Support Staff UtilizationPatient satisfactionStaff satisfactionProfitabilityVolume by Patient TypeProvider and Support StaffingAppointment Scheduling PoliciesExam Room Allocation PoliciesPatient Flow Patterns
11 A High Level Clinic Model Architecture balk, renegeQ21Models again help to organize our thinking about complex systems3
12 A Simple Patient Flow Model multiple waitsInterfaces 28:5 Sep-Oct 1998 (pp.56-69)
13 A myriad of questions – demand? Who is the underlying population to serve?What is the level of demand that can be satisfied by a clinic?How do you manage panels of patients for providers?what is the expected workload generated by a given panel of patients?What is the “appropriate panel size”?What are the basic types of patients served?Appointments, walk-ins, both?Demand for advance appt’s vs. same-day appointments
14 The Front Desk? How should the “front desk” be staffed? appointment schedulingpatient phone questionspatient check in/outbillingHow long do patients wait on the phone for scheduling appts, medical questions, billing questions?What about information systems to support patient records, appointment scheduling, billing?
15 How is appointment capacity organized? How much appointment vs. walk-in capacity is needed?appointment templateshow many of each “type” of appointment to offer?how to best sequence mix of appointments?how to estimate length of time block for each type of appt?leave appt slots open for same day appointments?open access concept (Murray and Tantau)how many?how many and how to schedule different specialty “sub-clinics” within an OP Clinic
16 Appointment Templates 2Template ID: Phys_Mon_AM_OB Provider Type: PhysicianDay / Time: Monday AM Clinic: OBHow does one design good templates?how many each type?slot length?sequencingTemplate managementBasis for generation of daily appointment schedulesStartSlotAppointmentPatientsTimeLengthTypePer Slot8:3030NEW19:0015Postpartum9:15Follow Up9:309:4510:0010:3010:4511:0011:1511:30
17 How is other resource capacity organized? How many exam rooms per provider?are the rooms assigned?Do patients get appointments with specific providers?How much support staff needed?Where are various clinical interventions done? Who does them?How much waiting room capacity is needed?
18 Appointment scheduling? Do you overbook? By how much?Performance measures for your overall appointment scheduling process?How do you measure how long your patients are waiting for an appointment?do you know when they want the appointment and whether their request was satisfied?How do you most effectively use appointment scheduling information systems?
19 Open AccessPremise – adjust capacity as needed to meet customer demandOne attempted response to chronic problem of delays to see primary care physicianaccommodate all appointment requests when patient wantsdeveloped by Kaiser Permanente (CA)popularized by Murray and Tantau (MT)Developed in early 1990sRecent articles in JAMAThree common modelstraditional access1st generation open access2nd generation open access
22 Traditional Access Stratify demand into urgent and non-urgent See urgent nowSee non-urgent laterDemand controlled by reservoir of supplyAppts booked to end of queue, schedules get saturated, little holding of capacity for short-term demandOften multiple appt typesEmphasis on matching demand to desired physicianUrgent demand “added on” or “worked in”May lead to long appt lead timesMT argue it artificially increases demandFocus on urgent condition only necessitates additional visitsDiverted patients (e.g. different physician) end up coming back anyway – 1 visit becomes 2 visits
23 1st Generation Open Access A “carve out” approach More “patient focused”I want to see my doc, and I want to see him/her nowPremise: demand can be forecasted with sufficient accuracy to allow better matching of capacity to demand“Carve out” capacity each day for projected SDA demandUrgent vs. Routine appt stratificationDeveloped by Dr. Marvin Smoller of Kaiser PermanenteSee Hawkins, S. “Creating Open Access to Clinic Appointments in the Henry Ford Medical Group”passed out in class
24 Some Problems with 1st Generation Open Access Mismatches between patient and PCPDefinition of “urgent” is fuzzy and changes as day goes onCreation of new appt types to meet urgent needs of patient who can’t come in todayQueues for routine tend to growgets shifted to use urgent capacityaffects phone-in capacity and SDA capacityBlack market or “second appt book” which fills “held” appts as they come available
25 2nd Generation Open Access “Create capacity” by doing all today’s work todayProviders responsible for panel, not appt slotsNo distinction between urgent and routineAppts are taken for the day the patient wants independent of capacityEvery effort to match patient with PCPargued that this reduces “unnecessary demand”Challengespredict total demandprovider flexibilitypanel management – how big?, how much work generated by a given panel?
26 2nd Generation Open Access What it is and what it is not….. It is a theory designed to improve appointment access and customer satisfaction.It is not a rigid formula(s)….each clinic will implement the theory in the manner that works best for them.Demand is not insatiable. Staff is not in the office until all hours of the day and night.How Clinic X tried to convey open access concepts to staff and mgt
27 Precursors to Open Access Prospective demand measurementtrack actual demand for appts by patients (when they want slot, not when got slot)track provider requests for follow-up demandPanel sizes must be manageable and equitableno method can deal with demand>>capacitytying panel size to workload can be challengingMust estimate current supply# of providers, # of available appointment slots taking into account time each provider is actually in clinicMust eliminate backlog of appointmentstemporary increase in capacity through extended hours, weekends, etc.Reduce # of appt typesPCP vs othershort and long (e.g. long = 2xshort)Develop contingency plansdealing with short term imbalances in supply or demandReduce and shape demandcontinuity of providermultiple issues at a visitgroup visitsnon-visit care (education, reference, self-care)Increase effective supply (especially of bottleneck resource)relieve providers of tasks that can be done by otherReview call center processes, staffing, etc. to assure telephone access
28 Myths and Rumors at Clinic X Correct ConceptMyth/RumorAppointment SchedulingAppointments are scheduled for when the patient would like to be seen.Appointment can be scheduled ahead of time (as far in advance as patient would like)Patient is driver of when to schedule appointment.Scheduled with PCP if in the officeCannot schedule return appointment until day want to be seen.PCP has to remain until patient is able to get to the office.Must add on as many patients as call to be seen that day.Insatiable DemandPatients are added on within a reasonable limit (contingency plans are developed).Providers are remaining in the clinic until all hours of the night.TeamingProviders are encouraged to form teams of 2-4 providers to care for patients.Teammates are utilized when PCP is out of the office.Patients still have PCP and see that individual as long as they are in the clinic.Must have only 2 people per team.Panel SizePanel size must be within reasonable limits. (Utilize Smoller’s demand model to help determine appropriate size).Panel is allowed to continue to grow without regard to demand.Appointment TypesThe pure theory dictates that there is no differentiation in appt types.Many clinics choose to continue with SDA (to maintain holds in the schedule).All appointments have to be 1 slot.All appointments are considered “routine” or same day.OvertimeSupport staff schedule is worked to decrease overtime and allow for provider support.People are staying late into the night with little support staff for assistance.OverallMany clinics are already doing a modified 2nd Generation Model and there are few changes.Drastic change in the way we do business.
29 Questions/Concerns about Open Access? Under what conditions would OA seem to be most applicable?When would it not be applicable and if so, are modifications possible?What is effect on care for chronic conditions? Will follow-up care slip through the cracks?Are we trading wait for an appointment for a wait at the clinic?What will day to day variation actually look like? How often will we be working until , say, 8pm?Effect on staff morale?How to actually implement?How to sustain?How pervasive and successful has it actually been?Impact on patient satisfaction?Impact on demand for visits?More...?
30 Measurements related to OA Patient satisfaction:Quarterly reports - all levels of careAnnual access satisfaction surveysProvider and staff satisfactionAvailability of appointments compared to modelLead time for future appointments and/or “defect rate”Percentage of patients seeing own PCP and % seeing team memberTelephone performance compared to standards:Average speed to answerHold timesCall abandonment ratesTalk timesPanel SizeVisits per month
31 Resource Based Relative Value Units Used as relative measure of clinical workload as well as basis for reimbursement by CMSDeveloped in late 1980’s by researchers from Harvard in conjunction with HCFA and physicians from numerous specialtiesAdopted in 1992 by HCFARBRVUs also used to measure physician productivityperformance monitoringincentive planscomparisons across departmentspanel managementresource allocationShortcomings as a productivity measuremedical care has changed since 1988 RBRVU development especially with respect to pre and post-encounter workdon’t fully account for effort for coordination of care, on-call, supervision of allied health professionals, remote communication with patientsCPT coding basis not very detailed for E+M (evaluation & management)for OP visit for new patient, for OP visit for established patientE+M codes cannot be combined to reflect multiple E+M tasks done at 1 visitLimited reflection of complexity variation in patient populations, provider experience or quality of careSee Johnson, S.E. and Newton, W.P. (2002) Resource-based Relative Value Units: A Primer for Academic Family Physicians, Family Medicine, 34(3), ppnice overviewreferences include the original research leading to RBRVU development
32 Measuring Work Effort – “Panels” How to translate a panel of patients to workload (# of visits, RVUs)?# of patients not a good measure of workdifferent patient types generate different numbers and types of visitsWhy might you want to be able to put a workload measure to a panel of patients? How would you use it?What are practical difficulties with measuring physician workload?effect of FFS and HMO patientssubstitution of specialist and/or ER care for primary carecovering for a colleagueHFMG built regression models based on patient age, sex, and Ambulatory Diagnostic Group (ADG) to predict workload for a panelKachal, S.K., Bronken, T., McCarthy, B., Schramm, W., Isken, N. – Performance measurement for primary care physicians, QQPHS 1996 Conference Proceedings (avail upon request)Have been using for the last 10 years for a variety of purposes
33 The Mathematics of Appt Scheduling tradeoffs between patient & provider wait, length of clinic day, provider utilizationappt timelast patientxxxxxidleclinicrun overend of exampatientwaitindividual appointments or blocks of patients given same appt time? (ex: 2 patients at start of day, then individual)
34 The Mathematics of Appt Scheduling Decent amount of research on various simplified versions of the appt scheduling problemsingle patient type usually consideredpunctuality often assumed (patients and providers)simple patient care path (one visit to provider)Important variablesmean exam time, coefficient of variation of exam timenumber of appts scheduled in a sessionpunctuality, no-show ratesrelative wait cost ratio between providers and patientsSome findingsneed good estimates of exam timesrelatively simple rules like scheduling 2 patients at the start of the clinic and then spacing appts out by mean exam time performed well in simulation experimentsthe “best” schedule depends on your objectives and parameter valuesimpact on practice has been limited (O’Keefe, Worthington, Vissers)
35 More about the math of appt scheduling Handout – annotated bibliography of recent research in appointment schedulingVissers, J. “Selecting a suitable appointment system in an outpatient setting”, Medical Care, XVII, No. 12, DecHo and Lau, “Minimizing total cost in scheduling outpatient appointments”, Management Science, 38, 12, Dec 1992.Vanden Bosch, P.M. and D.C. Dietz, “Scheduling and sequencing arrivals to an appointment system”,Bailey, N.T.J., “A study of queues and appointment systems in hospital outpatient departments”, J. Roy. Stat. Soc. B, 14, 185, 1952first paper published about the topic of appt systemsFetter, R.B. and J.D. Thompson, “Patients waiting time and physicians’ idle time in the outpatient setting”, Health Services Research, 1, 66, 1966.another early classic
36 Information Technology and Appointment Scheduling/Practice Management AppointmentsProOne-Call (Per-Se Technologies)Brickell Schedulere-MDsManage.md (ASP)The Medical OfficeMany more...The open source movement...Open source practice management projectsMedPlexus – open source EHR initiative with AAFPOSCARdev’d at McMaster in Canadastand alone appt scheduling vs. integrated with practice managementsingle appointments vs. series of appointmentscomprehensive resource scheduling?enterprise wide vs. departmental?integration with existing IS?remote access?capacityprice, vendor support, vendor viability
37 Case 1: A Partially Successful OR Engagement (Bennett and Worthington) Ophthalmology clinicnew and follow up patientsRoutine, Soon, UrgentThree ½ day clinic sessions per week3 docs (11New, 33FollowUp for regular clinic)Overbooked, overrun, excessive patient waitsMr. T suspected the appt systemFundamental issue of matching capacity to demand“systems thinking” viewUser involvementAwareness of fit within broader organizationInterfaces 28:5 Sep-Oct 1998 (pp.56-69)
38 Why might not the clinic be running smoothly? Patients late/earlyDoctors lateNo shows, cancellationsExcessive overbookingInappropriate appt lengthsHighly variable consultation timesLack of data about operationsWalk-insStaff absencesUnderstaffingNot enough spaceNot enough appt capacityPoor information flowMany more...
39 Vicious Circle of Insufficient Capacity and Overbooking Interfaces 28:5 Sep-Oct 1998 (pp.56-69)
40 Analysis HighlightsConsideration of both process and organizational issuesPatients were generally punctualwaited on avg 40 mins to see physician (51 mins including repeat waits)Simple model for “clinic appt build up”highlighted severity of demand>capacityIf demand>capacity in long term, no appointment scheduling magic is going to helpvacation notice deadline for providersSimple model to assess impact of lengthening time between routine visitsan attempt to decrease demandInterfaces 28:5 Sep-Oct 1998 (pp.56-69)
42 Analysis HighlightsUsed specialized queueing model to explore different appt scheduling patternsas expected, by spacing out appts further, wait to see provider decreased but at increase in provider idlenessof course, less appts will also exacerbate the difficulty in getting an apptDeveloped list of long term and shorter term operational strategiessome were implemented to various degreeshowever, not much really changed over 2½ yearsOP Clinics are messy, complex, and different constituencies have different goals and objectivesSimple models and “applied common sense”Interfaces 28:5 Sep-Oct 1998 (pp.56-69)
43 Demand Management Upstream Midstream Downstream population mgt prevention and wellnessself-caredisease mgtmanage chronic conditionsMidstreamwalk-in or call-incoordinate with ancillary providersmaximize visit efficiencymatch patient to providergroup visitsDownstreameducationtelephone follow-uplengthen visit intervalschange future point of service entry
44 Case 2: Simulation provides surprising staffing and operation improvements at family practice clinics (Allen, Ballash, and Kimball)Simulation quite useful for exploring impact of operational inputs on system performanceIntermountain Health Careintegrated health system based in Utah> 70 clinics, 840,000 enrollees, 2000 docsclinics ranged in size, configuration, operating tacticsDeveloped generic clinic simulation model to explore impact of different configurations/tactics on performanceMedModel – healthcare specific simulation development toolPaper has very nice description of a typical simulation analysis in healthcareProceedings of the 1997 HIMSS Conference – available upon request
45 A few highlights and things to note ( from Allen, Ballash, and Kimball) Started with “simple” model and added complexity as neededObtained “patient treatment profiles” from healthcare consulting firmFig 3,6 – “Low” MA utilization is “good”MA team had dramatic positive effect over assigned MAs – from 6 down to 4 MAs with only 4% ACLOS increase3 rooms/doc not better than 2 per docwait “moved” from waiting room to exam roomDedicating exam rooms to docs did not adversely impact performance – not the bottleneckPatient scheduling matters at higher workloadsOverbooking had significant negative impact on patient waitsProceedings of the 1997 HIMSS Conference – available upon request
46 A few highlights and things to note ( from Allen, Ballash, and Kimball) Used results as springboard to look at IHC clinics and how they operateAssessed feasibility of implementing insights gained from the modeling processNoted that significant changes (“reengineering”) of the patient care process will likely change the results of the analysisso, rerun it, that’s the beauty of having a model.Proceedings of the 1997 HIMSS Conference – available upon request
47 More Resources http://www.ihi.org/idealized/idcop/ American Academy of Family PracticeFamily Practice ManagementJournal of Medical Practice ManagementJournal of the American Board of Family PracticeManaged Care QuarterlyMedical Group Management Journal