PatientPaths

NOTE: patientpaths is currently (pre-alpha quality software). So if you've got suggestions/feedback on the design---especially if you work in healthcare and have a need for this sort of modelling---then please do get in touch, but be wary that it's not yet ready for clinical use.

patientpaths is a Python toolkit for clinical pathway modelling.

If you have

1. data about the capacity of your health system (e.g. number of ward/ICU beds); and

2. one (or more) forecasts of the number of infections you're expecting in the population (e.g. the outputs from an infection model based on local data)

then patientpaths can be modified to simulate the flow of these patients through your health system, e.g. presentation at emergency department → ward bed → ICU bed → ward bed (recovery) → discharge.

Features

• no proprietary software required (just Python v3.6+ with NumPy)

• customisable for the specific capacities & pathways within your health system (see Using patientpaths below)

• open source (GPLv3 licence)

Setup

pip install . will install the patientpaths executable in your Python environment, along with all the dependencies (numpy).

If you are using a linux or OSX computer which ships with a very old version of Python, like the past-end-of-life Python 2, you may need to use pip3 or another distribution-specific install command.

Using patientpaths

modify the code of the model_of_care.py file with the numerical parameters of your healthcare system, and adjust as needed the elements of your healthcare pathway and their interconnections via code in outcomes_for_moc.py file. execute example_run.py for a quick demonstration run of the software.

the outcomes_for_moc.py file contains the bulk of the pathway description and is executed by calling the outcomes_for_moc function. The outcomes_for_moc function takes 4 input parameters, a moc (model of care) string identifying a moc profile specified in model_of_care.py, two arrays specifying mild and severe cases occuring per cohort per day, and a 'risk' array, identifying specific cohorts 'at risk'.

Specificially run a pathways simulation for presenting cases (mild and severe) in cohorts (S) across days a number of days (D), inputs are:

• moc: a string identifying the model of care profile name specified in model_of_care.py (presently either: 'default','cohort','clinics','phone')
• di_mild: a numpy array of dimension [S, D], identifying mild cases presenting to the clinical pathway of strata (S) per day (D)
• di_sev: a nunpy array of dimension [S, D], identifying severe cases presenting to the clinical pathway per day (must be of same dimension as di_mild)
• risk: a one dimensional numpy array [S], which is a number for each cohort, which if greater than one identifies the cohort as being 'at risk' with slightly different dynamics.

Outputs from the simulation include a range of different arrays identifying different factors per day of the simulation, including:

• deaths: number of deaths from each of the cohorts across days of the simulation
• excess_{icu,ward,ed_sev,ed_mld,clinic_mld,clinic_sev,gp}: the vectors of numbers not admitted to each of the services on a given day
• admit_{icu,ward,ed_sev,ed_mld,clinic_mld,clinic_sev,gp}: the vectors of numbers admitted to each of the services on a given day
• avail_{ed,clinic,gp,icu,ward}: the numbers of unused capacity of each of the resources on a given day

The resources are:

• The ICU - intensive care unit, only available to severe cases, excess patients directly results in deaths
• The Wards and Emergency Departments, share beds between them and can accommodate severe and mild cases
• the Specialty Clinical pathway specifically for treatments
• and General Practitioners, specifically for treating mild cases

The specific pathway described by the code is a reimplementation of MATLAB code developed for modelling Australian Covid spread, originally modified from the pathway described by paper:

Robert Moss, James M. McCaw, Allen C. Cheng, Aeron C. Hurt, and Jodie McVernon. "Reducing disease burden in an influenza pandemic by targeted delivery of neuraminidase inhibitors: mathematical models in the Australian context." BMC Infectious Diseases, 16(1):552, October 2016. ISSN 1471-2334, doi:10.1186/s12879-016-1866-7.

The short code should be mostly self explanatory, between the code in outcomes_for_moc.py, the two functions in the components.py and the Presentation_Matrix.py class is the entirity of the functi

Licence

See LICENCE.