Tackling Patient Wait-Times at LVPEI with Systems Thinking. Ali Kamil, SDM 12, HKS 14 Dmitriy Lyan, SDM 11, Sr. Product Manager, Amazon

Transcription

1 Tackling Patient Wait-Times at LVPEI with Systems Thinking Ali Kamil, SDM 12, HKS 14 Dmitriy Lyan, SDM 11, Sr. Product Manager, Amazon

2 Agenda About LVPEI Opportunity, motivation, and challenges Approach General observations Analysis and recommendations Next steps Appendix

3 LVPEI is a non-profit organization focused on the delivery of eye care to patients at all levels of the economic pyramid. Hyderabad Campus Services offered: Comprehensive patient care Clinical research Sight enhancement and rehabilitation Community eye health Education Product development Centre of Excellence: Provides outpatient services to 200,000 people Performs 25,000 surgeries Trains 250 professionals at all levels of eye care Provides low vision services to 3,000 people LVPEI Eye Health Pyramid

4 Our team was engaged to identify bottlenecks and causes of high patient service time in the LVPEI outpatient department (OPD). Challenges Unpredictable demand patterns in a given day No established uniform process within the hospital. High and varying patient service times* High provider fatigue due to high patient volume and extended hour of service High turnover of doctors and staff Motivation and Opportunities Identify bottlenecks in the system. Causes of unpredictable wait times Propose a policy to reduce service times and achieve uniformity Must not compromise LVPEI s high standard of quality care and cannot turn away patients Can we increase capacity to handle more patients? How can we effectively handle walk-in patients? *service time = total time spent by patient in the hospital

5 Approach: Team conducted the project in three phases over 5 months Pre-trip (January to March) Engaged LVPEI Head of Operations to discuss current challenges Interviewed key personnel at hospitals in the Boston area, including operations leaders at Massachusetts General Hospital (MGH) and practitioners at Massachusetts Eye and Ear Infirmary (MEEI) and Mount Auburn Hospital Interviewed key stakeholders at the LVPEI hospital On-site (March 1 20) Conducted time and motion studies in two cornea and two retina OPD clinics, collecting timestamps on the flow of patient folders, and noting management practices Interviewed faculty ophthalmologists and optometrists, and OPD scheduling administrator Conducted patient surveys at the walk-in counter Post-trip (April to May) Ran statistical analyses to quantitatively identify relationships between different variables and patient service times Compared findings to our interviews and observations of management practices Derived recommendations for addressing systemic causes of increases in patient service times in the OPD

6 GENERAL OBSERVATIONS

7 Cornea Retina Patient pathways varied significantly depending on the clinic, and on the type of patient and appointment. Check-In Work-up Consultation Investigation Check-out Start Dilation Work-up 30 minutes Retina Consultation Yes Yes Retina Investigation Investigation required? No Investigations units are subject to their own process flow management practices. End Dilation required? No Cornea Consultation Cornea Investigation Yes Dilation 30 minutes

8 Additionally, several combinations of factors impact patient service time. Hospital-Specific Factors Commitment to training medical staff Patient volume vs. hospital capacity Scheduling-Specific Factors Doctor-specified appointment and walk-in templates Administrator's adherence to doctor-specified appointment templates Real-time prioritization of patients Clinic-Specific Factors Management of patient folders and staff # of Fellows, Optometrists, and Facilitators Skill levels of staff Size and layout of clinics Anticipated vs. actual patient volume Types and variety of patients that can be seen Need for diagnostics Patient-Specific Factors Lack of awareness of appointment-based system Bias for early morning arrival High volume of late arrivals and no shows

9 Service Time (h:mm) There is little discrepancy in patient service time between nonpaying (NP) and general (G) patients, but high variability ranging from two to four hours. Priority Level Patient Count Clinic 1 Clinic 2 Clinic 3 Clinic 4 Average Service Time G 182 2:44 4:08 2:17 3:26 3:12 NP 56 2:55 3:35 2:51 4:01 3:29 Average patient service times for general and nonpaying patients differed by only 17 minutes. 9:36 8:24 Mean (μ) 3h 15m SD (σ) 1h 37m Service Time Variability (All Days) 7:12 6:00 4:48 3:36 2:24 1:12 σ μ -σ 0:00

10 Walk-in patients have higher variability in service times compared to patients with appointments. Clinic 1 Clinic 2 Clinic 3 Clinic 4 6:21 5:01 4:38 5: Walk-in patient service time is higher than aptbased patients Average Service Time Total Patient Count 5: % of walk-ins arrive before 12pm Walk-in Patient Arrival Time vs. Service Time Walk-in Patient Service Time Variability (all days) 10:36 9:24 Mean (μ) 5h 4m SD (σ) 2h 3m 8:12 7:00 5:48 4:36 3:24 2:12 σ μ -σ 1:

11 Service Pressure vs. Time per Patient Time per Patient 10am 1pm Time of Day 4pm ANALYSIS & RECOMMENDATIONS

12 Require doctors to adhere to appointment based system and encourage on-time arrivals. Key Observations Only 28% of all apt. based patients arrived on time Clinics adhering to apt. based system achieved shorter service times Clinics 1 and 3 adhered to apt. based system and penalized patients for early (<30m) or late arrivals (>30m) Clinics 2 and 4 did not actively adhere to apt. system and had significantly higher service times for on-time patients Service Times for On-time Patients Clinic 1 Clinic 2 Clinic 3 Clinic 4 Avg. Service Time 2h 22 m 3h 58m 2h 4m 3h 38m St. Dev 1h 17 m 1h 33m 1h 22m 1h 23m Recommendations Require doctors to adhere to apt. based system Prioritize patients based on their appointment times and not check-in times Educate patients about apt. based system and encourage adherence :06 2:50 3:35 3:31 3:38 4:10 3:10 2: >4 hours early 3: :00 3:47 3 hours 2 hours early early 1 hour early 3:09 Arrival Time 2:22 On-time 30min window 1 hour late 3:21 2:27 Service Time 0:00 2 hours 3 hours late late Clinic 1 Clinic 4 4> hours late 2:06 2:23 4:19 3:50 3:21 2:52 2:24 1:55 1:26 0:57 0:28 0:00 4:48 4:19 3:50 3:21 2:52 2:24 1:55 1:26 0:57 0:28 0:00

13 Encourage the use of appointment system, while simultaneously employing strategies to better manage walk-in patients. Walk-in Survey Results Summary 40 patients surveyed in total 41% of patients had tried unsuccessfully to make an appointment; 50% of these were because the requested appointment time was unavailable 80% of patients who did not make appointments were unaware of the option Survey Findings In general, awareness of the appointment option is low Patients choose the walk-in option because the next available appointment is too far away The majority of walk-in patients are new to LVPEI Interview Findings Doctor scheduling for walk-in patients by time and type is often not adhered to due to over demand and incorrect triage Unexpected walk-ins are disruptive to the patient flow, but doctors have no choice but to accommodate Incorrect triage results in re-routing patients to other clinics and increased service time Walk-in patients often have primary care concerns that do not require specialized attention, or ask to see a specific doctor unnecessarily Ideas to Consider Better promote appointment system, especially among new patients Designate general doctors for walk-in clinic to reduce specialist time on general cases Require referral letters for new patients asking to see a specific doctor Enforce ophthalmologist-set guidelines for appointment booking at the walk-in counter

14 # of Patients Identify factors contributing to decreasing service times in the late afternoon : :52 Apt. Based Patients Arrivals & Service Times (all clinics) 3:57 3:36 3:23 3:29 3:28 3:01 2:35 2: :06 1 4:19 3:50 3:21 2:52 2:24 1:55 1:34 1:26 0:57 0:28 0:00 Average Service Time (8am 1pm) 3h 30m Key Observations Average service time decreases with time of day Appointment-based patients arrival time has normal distribution Average Service Time (1pm 7pm) 2h 30m Potential Factors to Consider Providers work more efficiently towards the end of the day Patients that do not require diagnostics are stacked later in the day Reduced number of walk-ins in the latter half of the day Ideas to Consider Closely observe the behavioral patterns of providers during the later half of the day. If positive behavior is identified, this practice should be replicated during the rest of the day.

15 Monitor practitioner fatigue in latter half of workday, as high pressure to serve customers can lead to increased errors and reduced service quality. Interview Findings Error rate of providers rises throughout the day for both optometrists and ophthalmologists After 4:00PM, doctors begin to observe fatigue in their teams After 4:00PM, doctors begin to observe work being completed in a hurry Key Observations Patients who arrive later in the day and patients who arrive significantly late for their appointments tend to experience lower service times. Average service time decreases with time of day With time of day, providers and staff tend to get fatigue and are prone to mistakes/errors 4:22 3:53 3:24 2:55 2:26 1:58 1:29 1:00 Apt. Based Patients Arrivals & Service Times (all clinics) Service Pressure vs. Time per Patient Time per Patient Ideas to Consider Closely observe the error rate that is created at any given time Closely observe the frequency of re-work over a given time period Determine the cause of the decline in service times during latter half of the day 10am 1pm Time of Day 4pm

16 Monitor practitioner fatigue in latter half of workday, as high pressure to serve customers can lead to increased errors and reduced service quality. Insights Workday is scheduled for 8am 5:30pm. Providers observed working until 7/8pm to service all patients High patient backlog increases pressure on LVPEI providers to service all patients in a given day Latter part of the day has been observed (via interviews) to increased fatigue and errors in service High pressure situation coupled with long workdays will lead to high turnover of staff Apt. Based Patients Target Wait Time Per Patient Time Remaining in Workday Current Wait Time Per Patient based on Backlog Walkin Patients + + Patient Arrival Rate + - Required Service Time - - Service Pressure + B Pressure Buildup Standard Workday LVPEI Patient Backlog + Workday + Total LVPEI Staff - B Reduced Time Per Task Time Per Patient Patient Check-out Rate - - Actual Service Time R Errors Created Fatigue Buildup time + Error Fraction + - <Workday> Rate of Change in Perception about Wait Times - Burnout Rate Perception Buildup Time Impact of Fatigue on Error Fraction Fatigue/Burnout Perception of Long Wait Times at LVPEI + Avg. Workday for Past Month Ideas to Consider Adherence to apt. based system and reducing number of walk-in patients Consider provider/staff rotation between highpressure clinics and regular clinics Identify rework and errors created by time of day Modeling Next Steps Consider long term impact to quality of service and reputation due to high service times and errors/rework Identify impacts to staffing and turnover due to high pressure environment Consider competitor/alternate emergence scenario

17 Identify and encourage best cross-consultation management practices Relevant Clinic Observations Cross-consultation cases comprise a non-negligible percentage in each clinic: 10 to 15% 3 out of 4 clinics employed practices to manage and integrate cross-consultation cases into existing patient flow Management of cross-consultation cases differed across clinics Passive cross-consultation management was disruptive to regular patient flow Sample Cross-Consultation Management Practices Fixed time allocation: 15 minutes every 2 hours for cross-consultations and short follow-ups Real-time prioritization: integration and prioritization of cross-consultations with existing patients Prioritization by check-in: prioritization of cross-consultation patients according to check-in time Ideas to Consider Conscious management of cross-consultation patients in each clinic Identification of good cross-consultation management practices Closer observation of the decision-making process behind the need for cross-consultation Guidelines for providers on the necessity for cross-consultation

18 Remove annual post-surgery follow-up requirement and divert patients to comprehensive clinic for ongoing long follow-ups Observations 60-70% of doctor s appointment templates are dedicated to seeing new patients 20% of all patients seen across the four days of study are new patients. Providers perform over 500 surgeries a year All patients are requested to come back for follow-ups at least once a year regardless of the need. Total Addressable Market + Adoption Rate + New Patient Arrival Rate New Patient Appointment Slots LVPEI Non-Surgical Patients LVPEI OPD Patient Backlog NS Patient Dep Rate B Hotel California Patient Service Rt. Follow Up Patient Arrival Rt Surgical Case Fraction + LVPEI Post Surgical Surgery Rate Patients + Follow Up Appointments + SG Patient Dep Rate Insights Continuing with the policy of requiring patients to come back for simple follow ups exhausts LVPEI doctors capacity to serve new patients. Dedicating more of providers time to follow up patients reduces opportunities to learn from diverse and complex cases. Ongoing reduction in time available to see new patients limits LVPEI s ability to realize its vision to reach all those in need. Ideas to Consider Removing the requirement for all patients to come in for yearly follow-ups post-surgery Transitioning fully recovered patients to comprehensive clinic for ongoing long follow-ups -

19 SYSTEM DYNAMICS BASED MODELING OF LVPEI

20 Used System Dynamics based modeling tools to simulate LVPEI operations

21 Hour*Optometrist/Patient Hour*Doctor/Patient Patient/Hour Used System Dynamics based modeling tools to simulate LVPEI operations Check-in/Checkout Selected Variables Rate Simulated the operations for the ideal scenario observed during time and motion studies of the Current system accumulates as more and more patients show up Workup time and Investigation times per patient drop to the minimum established requirement Time (Hour) Arrival Rate : Appointment Based "Check-out rate" : Appointment Based 0.7 Workup Time 0.2 Investigation Time Per Patient Time (Hour) Workup Time : Default Time (Hour) Investigation Time Per Patient : Default

22 Patient Used System Dynamics based modeling tools to simulate LVPEI operations Standard workday (9 hours) quickly devolves into workdays lasting hours. Complaints of burnout and fatigue during interviews confirmed the simulation results Increase in patient volumes at the investigation rooms tends to correlate with increase in crossconsultations or referrals Selected Variables Cross-Consultations/Investigations Implies ophthalmologists tend to increase referrals and cross consultation in patient buildup Time (Hour) Patients In Diagnostics : Default Patients In Investigation : Default

23 Used System Dynamics based modeling tools to simulate LVPEI operations System Dynamics model of LVPEI provided quantitative evidence to observations made during the time and motion studies Enables LVPEI to identify potential policies proven through the model to improve their processes Adherence to apt-based system Building a reputation of rewarding on-time arrivals Increase in headcount to avoid burnout and fatigue Additional work required to take into account exogenous factors such as provider s reputation, experience level of staff, bulk arrival vs. trickling arrival rate

24 NEXT STEPS

25 Additional studies and modeling exercises will build a comprehensive understanding of the factors contributing to patient service time in the OPD. Time and motion studies that include cross consultation patients Time and motion studies on cornea diagnostics Patient flow of walk-in patients Triage process at walk-in counter Check-in process Future Studies Patients returning to LVPEI due to incorrect diagnosis Effectiveness of short-term recommendations Simulation Models LVPEI s patient flow system for cornea and retina clinics

26 Acknowledgements Anjali Sastry Senior Lecturer MIT Sloan School of Management Janet Wilkinson - Senior Lecturer MIT Sloan School of Management Elli Suzuki COO, Global Minimum Inc. Nicole Yap Jacaranda Health, Kenya

27 Thank you

28 QUESTIONS?

29 APPENDIX

30 Service Time Overall Patient Average Service Time 6:00 4:48 3:36 2:24 1:12 0:00 Appointments Walk Ins New Follow Ups 22:48 21:36 Clinic 1 Clinic 2 Clinic 3 Clinic 4

31 Service Time Overall Patient Check-in to Dilation Average Service Time 4:48 4:19 3:50 3:21 2:52 2:24 1:55 1:26 Appointments Walk Ins New Follow Ups 0:57 0:28 0:00 Clinic 1 Clinic 2 Clinic 3 Clinic 4

32 Overall Patient Arrival Rates Clinic 1 Clinic 2 Clinic 3 Clinic :00 9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00 19:00

33 Service Time Number of Patients Appointment-based Patient Patient Type (G,NP, S, SS) & Service Time 3:36 3:28 3: :29 3: :14 3:07 3:00 2:52 2:45 3: : G NP S SS Patient Types Service Times

34 Appointment-based Patient Distribution Early/Late/On-Time Arrivals Arrivals (Total) NA 0% On Time 28% Early Arrivals 46% Late Arrivals 26% Clinic 1 NA 0% Clinic 2 NA 0% Clinic 3 NA 0% Clinic 4 NA 0% Late Arrival s 23% On Time 19% Early Arrival s 58% Late Arrivals 15% On Time 30% Early Arrivals 55% On Time 35% Late Arrivals 25% Early Arrivals 40% On Time 25% Late Arrivals 36% Early Arrivals 39%

35 Patient # Service Time Appointment-based Patient Arrival & Service Time (Clinic 1) : :52 3:47 3: : :21 3:21 2: :27 2:24 8 2:10 2:06 1: :26 4 0: :28 0 >4 hours early 0 0:00 3 hours early2 hours early 1 hour early On-time 30min window 0 0:00 1 hour late 2 hours late 3 hours late 4> hours late 0:00

36 Patient # Service Time Appointment-based Patient Arrival & Service Time (Clinic 2) 25 6: : :48 3:58 4:05 3:58 4: : :31 3: : :50 5 1:12 0 >4 hours early 3 hours early 2 hours early 1 hour early On-time 30min window 2 1 hour late 2 hours late 0 0:00 3 hours late 1 4> hours late 0:00

37 Patient # Service Time Appointment-based Patient Arrival & Service Time (Clinic 3) 25 6: :00 4: : :35 2:50 2:24 2:04 8 1: :31 5 1:17 1: :00 0:00 >4 hours early 3 hours early 2 hours early 1 hour early On-time 30min window 1 hour late 2 hours late 3 hours late 4> hours late 1 2 0:00

38 Patients # Service Time Appointment-based Patient Arrival & Service Time (Clinic 4) : :10 4: :35 3:31 3: :10 3:50 3: :50 2:44 2:52 2:23 2: : : : :57 0:28 0 >4 hours early 3 hours early 2 hours early 1 hour early On-time 30min window 1 hour late 2 hours late 3 hours late 4> hours late 0:00

39 Service Time Patient # Walk-in Patient Arrival & Service Time 10:48 9:36 9: :24 7: :12 6:17 6: : :48 4:10 4:22 4:05 8 3:36 2:24 2:37 2: :12 2 0: : : : Walkin Arrival Time vs Srvc Time Service Time

40 Difference b/w Check-in and Apt Times Appointment-based Patient Arrival vs Appointment Time Variability Early (0.10) (0.20) (0.30) Late (0.40) Day 1 Day 2 Day 3 Day 4

41 Service Time (h:mm) Appointment-based Patient Service Time Variability for On-Time (Clinic 1) 6:00 4:48 3:36 2:24 1:12 0:00 9:38 9:54 10:47 11:29 11:29 12:01 12:28 13:11 13:11 14:43

42 Service Time (h:mm) Appointment-based Patient Service Time Variability for On-Time (Clinic 2) 9:36 8:24 7:12 6:00 4:48 3:36 2:24 1:12 0:00 7:40 8:04 8:10 8:54 9:18 9:05 9:48 10:12 10:21 10:38 11:10 11:20 12:41 13:15 13:14 13:12 13:52 14:05 14:35 16:02

43 Service Time (h:mm) Appointment-based Patient Service Time Variability for On-Time (Clinic 3) 6:00 4:48 3:36 2:24 1:12 0:00 9:22 10:07 10:42 10:48 10:49 10:49 11:10 11:27 11:10 11:38 11:36 12:00 12:17 13:25 14:10 14:05 14:18 15:01 15:21 15:44 16:00 16:07

44 Service Time (h:mm) Appointment-based Patient Service Time Variability for On-Time (Clinic 4) 6:00 4:48 3:36 2:24 1:12 0:00

45 Team Profiles Ali Kamil a graduate student at the MIT System Design and Management program and an M.P.A. candidate at the Harvard Kennedy School of Government. His research interests lie in employing big data, social computing, and system dynamics based simulation tools to identify patterns in human behavior, connectivity, and activities in low-resource settings specifically in developing and emerging markets. He is a member of the MIT Media Lab's Human Dynamics group directed by Professor Alex "Sandy" Pentland. He holds a bachelor's degree in computer science and economics from the Georgia Institute of Technology. Dmitriy Lyan has professional experience in both software development and investment management. While at SDM, his research focused on identifying critical performance factors and developing simulation models to tackle management challenges faced by organizations in healthcare and education. He is currently a senior product manager at Amazon Web Services. In addition to holding a master's degree in engineering and management from MIT, he has an M.S. in financial engineering from the Peter F. Drucker and Masatoshi Ito School of Management at Claremont Graduate University and a B.S. in computer engineering from the University of California, San Diego. Elli Suzuki is a COO of Global Minimum Inc., a non-profit organization that is focused on inspiring and enabling youth in developing markets to create and implement their own solutions to most critical issues they face. Prior to MIT, Elli spent 5 years in the financial service industry, marketing multi-asset investment solutions to institutional clients Elli holds an M.S. in Management Studies from MIT Sloan School of Management. After graduation, Elli employs her management skills to disseminate innovative and affordable interventions designed to empower marginalized individuals in sub-saharan Africa. Nicole Yap holds an M.S. in Management Studies from MIT Sloan Scool of Management. She has two years of consulting experience, advising large private and public sector clients on their Customer Relationship Management (CRM) strategies. Her research focuses on the development of market-based policies and approaches that organizations can apply to sustainably reach developing markets. Nicole plans to apply her management consulting background to the development of sustainable global health strategy upon graduation in 2013.