Is the NHS subject to hidden systems’ effects?

Is the NHS subject to hidden systems’ effects? : Is the NHS subject to hidden systems’ effects? Steve Gallivan Director Clinical Operational Research Unit University College London s.gallivan@ucl.ac.uk www.ucl.ac.uk/operational-research

Various collaborators : Various collaborators Martin Utley Peter Sozou Tom Treasure Ken Binmore Paul Bate Glenn Robert John Gabbay Cathy Pope . . . etc

Slide3 : Some slides carry a health warning

Outline of talk : Outline of talk Is there a link between health and transport systems? An overview of NHS reorganisations Cautionary tales

CORU research areas : CORU research areas

Slide8 : System delays and throughput Throughput System delays 100% Capacity

Parallels between traffic control and NHS operation : Parallels between traffic control and NHS operation Delays and waiting are centrally important Capacity has a major influence Both suffer from ‘congestion’ Variability and unpredictability common

Differences between traffic control and NHS operation : Differences between traffic control and NHS operation Traffic jams are just an inconvenience, patient care is far more important Experiments are much easier with traffic If a new traffic control method doesn’t work, you find out very quickly

Overview of the NHS : Overview of the NHS

Slide12 : The National Health Service was founded on 5th July 1948 To provide healthcare on the basis of need, not on ability to pay – “free at the point of service”

Slide13 : Doctors' pay -  The Royal Commission on doctor's pay; Management - Hospital Activity Analysis introduced;. Porritt Report - criticised the separation of NHS into three parts; Hospital plan - development of district general hospitals; Salmon Report - reorganisation of senior staff structure; GP's charter - group practices, health centres; New hospitals – Local reorganisation; Resources planning – distribution of resources more fairly; Royal Commission - strategic planning, better resource reallocation; Performance indicators - better health service information;                  Audit - closer examination of outcomes; Community health - moving care from a hospital to community setting; Internal market - 'purchasers' (health authorities) and 'providers' (hospitals) competition; NHS trusts - Hospitals become “NHS Trusts” – self-contained businesses;                   GP fund holders - GPs become fund holders and ‘buy’ services; Internal Market ‘abandoned’ National Service Frameworks, Information for Health, Modernisation, Primary Care Trusts, Strategic Health Authorities, booking, patient choice, electronic patient records, Treatment Centres, payment by results, electronic booking. Re-organisations of NHS

Slide14 : Today the NHS has a workforce of over one million. A massively complex public system The NHS in the 21st Century

Slide15 : Devil’s advocacy and health policy what can go wrong? Cautionary tales

Health initiatives : Health initiatives The internal market (now abandoned?) Payment by results Booked admissions Patient choice

Slide17 : Devolution of decision making can degrade system performance Policy change may accidentally introduce system instability Cautionary Tale 1

Slide18 : Modelling an internal market in the NHS

Slide19 : Modelling the evolution of a health service A dynamical queueing system , (~ road traffic?)

Slide20 : Computerised demonstrator prototype

Slide21 : Example of the evolution of the system Time Annual Trust profits Trust 2 closes

Slide22 : Trust profits Cost to community Prices Allocation of patients

Slide23 : Example where two Trusts close Trust profits Time

Slide24 : Trust profits Pathological example where all three Trusts close

Slide25 : Trust profits An example of unstable, possibly chaotic, evolution Time

Slide26 : But payment by Results policy re-introduces competition “The potential for gaming has already been acknowledged by DH”

Slide27 : Devolution of decision making can degrade system performance Booked admissions may require additional system capacity Cautionary Tale 2

Slide28 : VARIABILITY

Slide29 : A simple extension to the conveyor belt model How many beds are needed to honour commitments ?

Slide30 : pi - probability that a patient is still resident i days after admission; Probability generating functions give simple analysis

Slide31 : Variable length of stay causes variation in bed demand

Slide32 : MORAL: Coping with unpredictable variability adds to system needs

Slide33 : Devolution of decision making can degrade system performance To be able to offer choice, the system needs more capacity Cautionary Tale 3

Slide34 : Patient choice analogy - aircraft catering 14 first class passengers will be offered meat or fish dinners. Both are equally popular. How many of each should be stocked? 7 meat + 7 fish 10 meat + 10 fish 14 meat + 14 fish Stock Probability of being understocked 80% 5% 0%

Slide35 : Trust A Trust B Patients’ preferences for two equally attractive options

Slide36 : Extra capacity supplied above that expected (%) Total number of patients in programme Percentage chance of Unit being oversubscribed 1% 5% 10% 50% KEY Capacity requirements for a Unit catering for patients choosing between two equally attractive sites

Slide37 : MORAL:Choice requires substantial unused capacity (which has to be funded)

Cautionary Tale 4 : Cautionary Tale 4 Devolution of decision making can degrade system performance With apologies to Braesse Devolution of decision making may degrade system performance

Slide39 : Care pathways for two hypothetical hospitals indicating processing time formula depending on flow rate x 25+x 25+x 5x+1 5x+1 Pre- Assessment Surgery Referral Join waiting list Referral Referral E B D A C F

Slide40 : Referral E B D A C F Referral 16 units 28 units 28 units 16 units Each patient has 44 units of processing time Configuration of patient flows without patient choice

Slide41 : Referral Equilibrium configuration for patient choice showing delays associated with each link in the care pathway E B D A C F Referral 26 units 26 units 26 units 26 units Each patient has 52 units processing time NO PATIENT CAN OPT FOR A PATHWAY OF SHORTER DURATION

Devolving decisions to system users may degrade system performance for all. Why? : Devolving decisions to system users may degrade system performance for all. Why? Changing an optimisation problem with N sets of variables into N optimisation problems with 1 set of variables does not ensure efficiency GUESS

Slide43 : MORAL: Offering choice may have complex hidden costs Authoritarian agent System Optimum Multiple decision makers Degrades system performance

Slide44 : Patients Providers 1 year’s flow of cases Arthritis By pass Cataract Diabetes Epilepsy Fractures Geriatrics Private TC Also patient choice adds extra unpredictable variability to gaming model ?