Transcript Document
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
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
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
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
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
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.
Collaborative Health Work
• Collaborative work with many Health professionals and
organisations in UK, Austria, Brazil, India, and Sweden.
Geo Data Institute, University of Southampton
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
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
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, socioeconomic actions
•Resources needed or planned
•Costs of resources
Example of Health Care Modelling
Data Analysis and Modelling
for
Early Detection and Treatment of Breast Cancer
Geo Data Institute, University of Southampton
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.
Early Detection of Breast cancer
•Delay in detection can be
reduced by screening.
Cancer starts
•Apparently healthy women
can be screened through:
•Self-examination.
•Clinical examination.
•X-rays (mammography).
Woman becomes aware
Treatments start
Visit to doctor
Delay in
detection
Time
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
Treatment of Breast Cancer
•Many are treatments available.
•Treatment depends on the severity of cancer.
Stage I: Small moveable tumour in breast only.
Stage II:Tumour not advanced but spread to lymph nodes.
Stage III:Locally advanced tumour. May be attached to chest muscles.
Stage IV:Distant 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
Modelling
Evaluation without formal cancer models
Intervention
Screening or Treatments
Evaluation
Mortality Statistics
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
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
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
Screening Model
Inputs to Model
Particular data sets
Expert knowledge
Screening options
eg Starting age
Intervals
Type of test
Outputs from Model
Tumour size
Aggressiveness
Age at detection
Stage of cancer
% detected by screens
Resources & costs
Etc
Geo Data Institute, University of Southampton
Illustrative Results From Screening Model
2 year Mammography
4
3
4%
7%
Non-Screened
4
1
8%
13%
2
30%
1
59%
3
19%
3 year Mammography
4
3 4%
8%
2
60%
Stage at registration
2
35%
1
53%
Geo Data Institute, University of Southampton
Illustrative Results From Screening Model
Tumour doubling time (measure of cancer aggressiveness)
Geo Data Institute, University of Southampton
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.
Stage 1
Stage 2
Treatment
Treatme
nt
Disease
Free
Local
No
Response
Distant
Response
Local and
Distant
Stage 3
Treatment
Stage 4
Treatment
Death From
Other Causes
Progressive
Disease
Death
From
Breast
Cancer
Illustrative Results From Treatment Model
1
0.9
0.8
0.6
0.5
0.4
0.3
0.2
0.1
Survival Time (Years)
Stage 1
Stage 2
Stage 3
Stage 4
Survival by stage at diagnosis.
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
0
Probability
0.7
Illustrative Results From Treatment Model
1
0.9
0.8
0.6
0.5
0.4
0.3
0.2
0.1
Survival Time (Years)
30-39
40-49
50-59
60-69
Survival by age at diagnosis.
20
19
18
17
16
15
14
13
12
11
10
9
8
7
6
5
4
3
2
1
0
0
Probability
0.7
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
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. Admit
Evaluate increased
capacities?
Treat
Discharge
Example: Planning a New Respiratory Unit
No.
of
beds
No. of beds in
use from a
detailed model
Proposed capacity of
30 beds
Days in a year
Example:Planning 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
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
Comparing Data with Model Predictions
Data
Emergency Admissions
Elective Admissions
Total Admissions
Number of Deaths
822
190
1012
184
Model
Mean
821
191
1012
185
Bed Occupancy
Deferral rate
80.0%
20%
80.3% 79.3%, 81.3%
21.5% 20.5%, 22.5%
Transfer Rate
1%
1.2%
Model 95%
CL for Mean
814, 828
187, 195
1004, 1020
182, 188
1.0%, 1.4%
Distribution of Beds Occupied
• Note evidence of pressure due to heavy use of capacity
Geo Data Institute, University of Southampton
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
Increased Demand and Increased Beds
(Base case has 13 beds)
Emergency Admissions
Elective Admissions
Total Admissions
Number of Deaths
Base
Case
822
190
1012
184
14
Beds
819
190
1009
186
16
Beds
1047
234
1281
236
Bed Occupancy
Deferral rate
80.0%
20%
74.2%
12.6%
81.8%
19.7%
1%
0.6%
1.2%
Transfer Rate
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
20 Extra Planned General Surgery Patients
Data
Emergency Admissions
Elective Admissions
Total Admissions
Number Died
Bed Occupancy
Deferral Rate
Transfer Rate
13
14
Beds Beds
822
822
819
210
209
212
1032 1031 1031
186
185
81.8% 76.2%
23.0% 15.1%
1.3% 0.6%
Geo Data Institute, University of Southampton
Hospital Capacities
Demonstration of a model
for
Planning Hospital capacities
Geo Data Institute, University of Southampton
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
Some Southampton Health Services Models
Hospital Capacities: Critical Care Units. A&E + 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
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
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
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
Contact Details
Dr Arjan Shahani,
Director,
Health Data Analysis and Modelling Group,
GeoData Institute,
University of Southampton, Southampton SO15 7PJ
[email protected]
[email protected]
[email protected]
Email to hotmail address with copies to
geodata and maths addresses please
Geo Data Institute, University of Southampton