Geo Data Institute, University of Southampton

1


Detailed Practical Models

Data Analysis and Modelling With an emphasis on
Care of People with Particular Diseases and
Conditions and for Planning and Managing Health
Services
Geo Data Institute, University of Southampton
2


Main Messages
Collaborative work by
Universities, Industry, and Service
Organisations for Developing, testing and
validating practical detailed models is mutually
beneficial

Geo Data Institute, University of Southampton
3


Main Messages
Databases Data
analysis including classifications Detailed
models at the level of individual
entities Powerful tools for making good
Planning and managing decisions

Geo Data Institute, University of Southampton
4


Personal Information

• Arjan Shahani
• 1956 1963 Statistician with Philips, London
• 1964 1967 Bristol College of Technology/Bath
  University
• Since 1967 University of
  Southampton.
• Developed Operational Research and collaborative
  work
• Evolved Health work since about 1980.
• Set up Institute of Modelling for Healthcare in
  1987 in
• Mathematics Department.
• Retired from full-time formal university work in
  1998.
• Set up Health Data Analysis and Modelling Group
  in the
• GeoData Institute, University of Southampton
• Health work continues

5


Collaborative Work from Southampton

• Collaborative work through MSc projects, PhD
  projects, and Commissioned research projects with
  many organisations
• Examples
• Designing a new sea port for container traffic
• Forecasting milk production
• Passenger flows through an airport
• Interactions between research and production for
  new products
• Flows of vegetables through an Indian Market
• Planning for batch production
• Care of people with particular diseases
• Planning and managing capacities for health
  services

Geo Data Institute, University of Southampton
6


Health Work At and Through Southampton

• Since 1980
• Courses on Health work in Southampton MSc OR.
• About 25 PhDs. About 100 MSc Projects.
• About 500K obtained for funding Research
  Assistants for work on Health Care and Health
  Services.
• Research work of the Operational Research Group
  awarded Grade 5 in a national UK evaluation
  exercise in 1996. Health work made a highly
  significant impact in this evaluation. OR Group
  was the only one in Maths Dept. to be awarded
  this top Grade.

7


Collaborative Health Work

• Collaborative work with many Health
  professionals and
• organisations in UK, Austria, Brazil, India,
  and Sweden.

Geo Data Institute, University of Southampton
8



Nature of the Southampton Models

• Based on individual patients grouped in
  appropriate classes.
• Uncertainty and variability are taken into
  account PROPERLY.
• Models can be easily used by health
  professionals
• Examples
• Care of people with diabetes
• Critical Care capacities
• Whole hospital capacities

9


A Program for Classification and Data Analysis

• Rapid extraction of information from databases.
• Patient flows
• Patient outcomes
• Use of resources
• Auditing and monitoring
• Links with detailed health care and health
  services
• models
• Demonstration of PORT

Geo Data Institute, University of Southampton
10


Health Care Modelling

• Understand and represent the PROCESSES involved
  in disease, care, patient flows etc.
• Description of community or patient groups.
• e.g. age, sex, risk groups
• Disease history or patient progress
• Interventions e.g. screening, vaccination,
  treatment, socio-
• economic actions
• Resources needed or planned
• Costs of resources

11


Example of Health Care Modelling


Data Analysis and Modelling for Early Detection
and Treatment of Breast Cancer
Geo Data Institute, University of Southampton
12



Breast Cancer

Breast cancer is a major form of cancer in many
countries.

• In the UK breast cancer results in
• Annually 27,000 new registrations
• and 15,000 deaths in England and Wales.
• 20 of all female cancer deaths.
• 5 of total female deaths.
• High mortality in women aged
• 35-54 years.
• In UK women between the ages
• of 50 and 65 are invited for
• screening at three-yearly intervals.

13



Early Detection of Breast cancer

• Delay in detection can be
• reduced by screening.
• Apparently healthy women
• can be screened through
• Self-examination.
• Clinical examination.
• X-rays (mammography).
• Evidence from many countries
• suggests that early detection reduces mortality.
• Screening using mammography is available in a
  number of countries.

Geo Data Institute, University of Southampton
14



Treatment of Breast Cancer

• Many are treatments available.
• Treatment depends on the severity of cancer.
• Stage I Small moveable tumour in breast only.
• Stage IITumour not advanced but spread to
  lymph nodes.
• Stage IIILocally advanced tumour. May be
  attached to chest muscles.
• Stage IVDistant metastases.
• Mortality rates are the usual measures of
  goodness of treatment.
• Mortality rates vary between hospitals and
  between countries.
• UK mortality rates are high in comparison with
  Western Europe and
• USA.

Geo Data Institute, University of Southampton
15



Modelling

Evaluation without formal cancer models
Intervention
Evaluation
Mortality Statistics
Screening or Treatments
eg Randomised Controlled Trial (RCT) for annual
mammography

• Noisy data.
• Expensive field trials
• Information about annual screens
• Other options need new RCT

• Cancer models data from various sources
  including any
• necessary RCT is a powerful combination

16



Modelling of Screening for Breast Cancer

• Three PhD theses V de Senna, Y Ouinten, S
  Mandurah
• Models given to UK Department of Health
• Current UK screening policy was one of the
  options evaluated
• by the Southampton Models

Geo Data Institute, University of Southampton
17



Data and Information for Screening Model

Cancer Register Life Tables Population Tumour Growth rate Age of Onset of Cancer Screening Processes Tumour Size Error probabilities
Geo Data Institute, University of Southampton
18


Screening Model


Inputs to Model Outputs from Model
Particular data sets Tumour size
Expert knowledge Aggressiveness
Screening options Age at detection
eg Starting age Stage of cancer
Intervals detected by screens
Type of test Resources costs
Etc
Geo Data Institute, University of Southampton
19



Illustrative Results From Screening Model

Stage at registration
Geo Data Institute, University of Southampton
20



Illustrative Results From Screening Model

Tumour doubling time (measure of cancer
aggressiveness)
Geo Data Institute, University of Southampton
21



Treatment Model

• Very general model that can be tuned for
  different users.

• Users define treatment for different stages of
  cancer,
• patient progress and treatment outcomes.

22



Illustrative Results From Treatment Model

Survival by stage at diagnosis.
23


Illustrative Results From Treatment Model

Survival by age at diagnosis.

24


Some Southampton Health Care Models




Particular Diseases Asthma, Breast Cancer,
Cataracts, Cervical Cancer, Chlamydial
Infection, Colorectal Cancer, Depressive
Illness, Diabetes, HIV/AIDS, Trachoma
Geo Data Institute, University of Southampton
25
Health Services Models Capture Patient Flows and
Use of Resources

Arrival of Individual patient. Patient type.
Care Unit needed
Admission rules for Care Units
Required capacities available?

No
Yes
Send elsewhere. Evaluate increased capacities?
Admit
Treat
Discharge

26


Example Planning a New Respiratory Unit


No. of beds in use from a detailed model
No. of beds
Proposed capacity of 30 beds

Days in a year
27



ExamplePlanning for a Group of Hospitals

• Outpatients, Inpatients, Day cases at four
  hospitals
• Detailed models at the level of individual
  patients grouped within about 20 specialities
• Models provided detailed quantitative
  information about
• The number of beds for each speciality
• The number of operating theatres
• Organisation of outpatient clinics taking the
  travelling times of the patients into account

Geo Data Institute, University of Southampton
28



Critical Care Capacities

• Demonstration of a model for planning
• Critical Care Capacities
• The computer solution of the model is called
  CCU-SIM
• CCU-SIM can be tuned to represent a particular
• Critical Care Unit

Geo Data Institute, University of Southampton
29


Comparing Data with Model Predictions







Data Model Mean Model 95 CL for Mean
Emergency Admissions 822 821 814, 828
Elective Admissions 190 191 187, 195
Total Admissions 1012 1012 1004, 1020
Number of Deaths 184 185 182, 188
Bed Occupancy 80.0 80.3 79.3, 81.3
Deferral rate 20 21.5 20.5, 22.5
Transfer Rate 1 1.2 1.0, 1.4

30



Distribution of Beds Occupied

• Note evidence of pressure due to heavy use of
  capacity

Geo Data Institute, University of Southampton
31



Planning Critical Care Capacities







Evaluation of Scenarios

• Many scenarios involving changes in
• Capacities
• Case-mix and numbers of patients
• Organisation of Critical Care Units
• were evaluated
• Some examples will illustrate the use of
• Southampton models

Geo Data Institute, University of Southampton
32


Increased Demand and Increased Beds


(Base case has 13 beds) Base Case 14 Beds 16 Beds
Emergency Admissions 822 819 1047
Elective Admissions 190 190 234
Total Admissions 1012 1009 1281
Number of Deaths 184 186 236
Bed Occupancy 80.0 74.2 81.8
Deferral rate 20 12.6 19.7
Transfer Rate 1 0.6 1.2

33


No Neuro Surgery Patients

Data 13 Beds 12 Beds
Emergency Admissions 770 768 771
Elective Admissions 189 194 190
Total Admissions 959 962 961
Number Died 176 175 179
Bed Occupancy 74.4 80.8
Deferral Rate 15.5 25.3
Transfer Rate 0.7 1.3


Geo Data Institute, University of Southampton
34

20 Extra Planned General Surgery Patients

Data 13 Beds 14 Beds
Emergency Admissions 822 822 819
Elective Admissions 210 209 212
Total Admissions 1032 1031 1031
Number Died 186 185
Bed Occupancy 81.8 76.2
Deferral Rate 23.0 15.1
Transfer Rate 1.3 0.6



Geo Data Institute, University of Southampton
35


Hospital Capacities


Demonstration of a model for Planning Hospital
capacities
Geo Data Institute, University of Southampton
36


Work with Hospitals


Brazil Through Rio de Janeiro University India
G.T., Mumbai. KEM, Pune Sweden Critical Care
(Prof Walther, Dr Steins) UK Basingstoke,
Bournemouth, Chichester, Dumfries, Heatherwood,
Isle of Wight, Norwich, Poole, Princess Anne,
Queen Alexandra, Royal Berkshire, Salisbury,
Southampton General, St Marys, Tunbridge,
Winchester, 7 others for Critical Care
capacities

37


Some Southampton Health Services Models




Hospital Capacities Critical Care Units. AE
MAU. Hospital Care units. Hospital (existing or
new) as a whole. Outpatient Clinics
Orthopaedics, Depressive illness, ENT, Eye,
Skin. Waiting Lists Inpatients and
Outpatients. Regional Capacities Cleft lip and
Palate, Coronary, Dental. Service Provision
Maternity Care. Whole System Primary Care,
Acute Hospital, Post-Acute Care. Forecasts of
daily emergency admissions for hospitals in
England. Met Office project
Geo Data Institute, University of Southampton
38


Collaborative Work

• MSc projects
• Southampton (England) Students June
  to September
• Linkoping (Sweden) Students September
  to December
• PhD / MPhil projects one to three years
• Research grants
• Commissioned work
• Plan for work with Indian Institutes

Geo Data Institute, University of Southampton
39
Collaborative Operational Research in Health A
Meeting in India

• Indian Health Professionals and Operational
  Researchers
• Interested people from other countries
• Meeting in January 2007
• Visiting Mumbai, Delhi, Varanasi, Pune,
  Coimbatore
• in December 2005 for planning the meeting

Geo Data Institute, University of Southampton
40
Collaborative Operational Research in Health A
Meeting in India

• Areas to be covered
• Infectious diseases
• Chronic conditions
• Planning and managing health capacities
• Indian System of Medicine (Ayurveda)
• Combining Indian and Western Systems of
• Medicine

Geo Data Institute, University of Southampton
41


Contact Details



Dr Arjan Shahani, Director, Health Data Analysis
and Modelling Group, GeoData Institute,
University of Southampton, Southampton SO15 7PJ
akshahani_at_hotmail.com aks_at_geodata.soton.ac.uk A.K
.Shahani_at_maths.soton.ac.uk Email to hotmail
address with copies to geodata and maths
addresses please

Geo Data Institute, University of Southampton