Healthcare interest group

From Santa Fe Institute Events Wiki

Contribute to project narrative here

Example of borders problem from Owen Densmore

Persons interested healthcare systems, health-related phenomena, health behaviors, etc... let's meet up and discuss project ideas over lunch Wednesday, 6/6, noon (we'll be over at SFI)!

(I know I missed a couple people, so please add your name!)
Anyone is welcome to join us!

I would be interested in joining you over lunch, Wed 6/6, noon. Saleha Habibullah

Summary of project topics for exploration:

1. Framework for agent-based modeling of EHR implementation effects on mortality in Pediatrics study (Han 2005).

2. Network model of preventive health services delivery.

3. Framework for targeting health behavioral interventions on the basis of agent-based models

4. East-west differences in health services delivery - a project that addresses this...

5. Where to place an intervention to best influence health related behavior.

Next meeting: Over lunch Friday, June 8, noon until ?

Next meeting: Over lunch, Monday, June 11, noon until ? - We meet with Owen Monday...--Mpoynton 23:23, 10 June 2007 (MDT)

CSSS 2007 Healthcare Interest Group Files (current) Files posted here:

Publications/ references:

  • Ammenwerth (2006) -VS
  • Del Baccaro (2006) -VS
  • Gilbert & Terna (1999) - RB
  • Han (2005) -MP
  • Holme slides (date unknown) -RB
  • McDonnell (date unknown) -RB


  • Mollie's EHR data - description of attributes -MP


  • Agent-based Modeling "black box diagram" V1 -RB
  • Black Box Template -RB

Note: If you want to share something that's publicly available over the internet (no copyright issues) - just add to the wiki here and link to it - no sense going to the trouble of adding it to the web site. --Mpoynton 23:23, 10 June 2007 (MDT)

Pubmed search on workflow yielded multiple relevant articles. I've posted abstracts here. --Mpoynton 23:23, 10 June 2007 (MDT)

--Vikas Shah 11:34, 10 June 2007 (MDT)

Pubmed query: (CPOE or "computerized provider order entry" or "electronic health record" or "electronic medical record" or EHR or EMR) AND (problem or model or mortality or study); 1300 records, many irrelevant (ideas for parsing/paring?)

Problems with CPOE:

  • Admission orders could not be written until en route patient was onsite (Han 2005); a structural problem that was corrected later
  • Increased time required for order entry (ten clicks vs written statement) (Han 2005)
  • Communications infrastructure bandwidth exceeded, slow network time
  • Reduced time at patient bedside
    • Dedicated personnel for order entry vs. bedside management by all available personnel (Han 2005)
    • Nursing order entry away from bedside (Han 2005)
  • CPOE required drugs to be kept in central pharmacy rather than in the unit (Han 2005)
  • Pharmacist access to CPOE led to lock out of further order entry (Han 2005)
  • Reduced face-to-face time between physicians and nurses
    • Potentially delaying the time that necessary order revisions could take place (Han 2005)
    • Reductions in spontaneous discussions related to patient care


  • Han et al (2005) “Unexpected Increased Mortality After Implementation of a Commercially Sold Computerized Physician Order Entry System”, Pediatrics, 116:1506.
  • Ammenwerth et al (2006) "Impact of CPOE on Mortality Rates – Contradictory Findings, Important Messages", Methods of Information in Medicine, 45: 586-593.
  • Del Beccaro et al (2006) "Computerized Provider Order Entry Implementation: No Association With Increased Mortality Rates in an Intensive Care Unit", Pediatrics, 118:290.

Approach to agent-based modeling CPOE in the ER setting:

  • Physical layout of the ER with patient rooms, healthcare providers with order entry permission (MD / DO / APRN), and terminals
  • Modeling patients as agents vs as patches
    • Agents can have a memory about whether they have been seen, by whom, and lengths of time required for their care at terminal and between visits, can decide to “leave against medical advice” (AMA) for extended lengths of time
    • Patches are probably a simpler initial approach to modeling
  • [Queuing theory] (thanks to Joshua Payne!)

Queuing theory – in the ED, I think we have a G/G/n queuing system. If we model an outpatient clinic, it would be more like an M/M/n queuing system – easier to model?--Mpoynton 00:49, 11 June 2007 (MDT)

    • Poisson process to model the arrival of patients
    • Exponential model for time it takes for patient to be cared for and leave
    • One initial “disease process” for all patients? Vs using the above processes to capture time for visit and workup
    • Tracking the “waiting room” queue
    • Triage and implications for the model developed e.g. “simpler” patients shunted away from our model?
    • Method for notification of new patient arrival?
      • “Nearest” provider is assigned to the patient
      • “Nearest” non-busy provider
      • Providers iterated with each time click until “not busy”
  • Multistep process for patient care: initial workup, order entry, follow up visit(s), discharge with healthcare provider moving between
  • Assume enough RN care to simplify the model (do not want to add too many variables!)
  • Key control parameters (initially):
    • number and location of terminals
    • number of providers
    • number of patients per provider (start with 1..?..)
    • number of rooms
    • resources
  • Other “modules” that may be implemented
    • Translator resources and stochastic increase in time required for random patients
    • Stochastic “network jams” and increase in time at terminal
    • Pre-hospital order / team assembly (some high-resource patients arrive with provider already available and “streamlined” terminal time)

I'm overwhelmed thinking about how we'll manage all of this in 3 weeks (!), so I'm starting some basic tables for everyone to modify as we go along, starting simple:

Agents Setup N Δ N
Physician 2 no
Registered Nurse 8 no
Clerks 2 no
Patients (3 acuity levels) 12 Birth/Death (develop queuing formula)

Patient Acuity RN Interaction time Documentation time Frequency of EHR access Distribution
High 1 1 1 10%
Medium .5 .5 .5 50%
Low .25 .25 .25 40%

Manipulate: Number and placement of terminals for EHR access on a unit layout

Outcome of interest: mean time to discharge/transfer OR mean patient time on unit

--Mpoynton 00:49, 11 June 2007 (MDT)