Embed
Email

Bed Presentation: Re-thinking bed management

Document Sample
Bed Presentation: Re-thinking bed management
Description

Looks at the link between queuing theory, the erlang equation & Poisson statistics in explaining bed occupancy and the adverse consequences of having too few beds and hence too high an average occupancy.

Shared by: Dr Rod Jones
Stats
views:
251
posted:
9/1/2009
language:
English
pages:
25
Re-thinking Bed management

Dr Rod Jones (ACMA) Statistical Advisor Healthcare Analysis & Forecasting Camberley, UK +(0)1276 21061 hcaf_rod@yahoo.co.uk



Re-thinking Bed Management

• • • • • • • • Randomness is very important Queuing theory is not complicated Why current bed planning is flawed Bed numbers, occupancy & turn-away Economies of scale High throughput has consequences Elective vs emergency flows Can you ‘plan’ medical bed occupancy?

Dr Rod Jones (2003) Tel: 0118 3773511



Randomness in healthcare

• Poisson randomness describes arrival events • Widely used in telecommunications, business and industry • Is the basis of queuing theory • Is the forgotten but controlling factor in most healthcare demand

Dr Rod Jones (2003) Tel: 0118 3773511



Poisson randomness

• Standard deviation equals square root of the average • Maximum variation is three times the standard deviation • But is a skewed distribution • Skew increases as size decreases



Dr Rod Jones (2003) Tel: 0118 3773511



Poisson randomness

P is o ra d me sfo a a e g a a ra o 1p rd y o s n n o n s r n v ra e rriv l te f e a



• You are expecting 1 per day but must be able to cope with 6 or 7 actual arrivals • On 37% of days your resources stand idle as there are no arrivals



4% 0 3% 5 3% 0 2% 5



re un y F qe c



2% 0 1% 5 1% 0 5 % 0 % F q ec re u n y



0 3% 7



1 3% 7



2 1% 8



3 6 %



4 2 %



5 0 %



6 0 %



Atul a rriv lsea hd y ca a a c a



Dr Rod Jones (2003) Tel: 0118 3773511



Poisson randomness

• At 5 per day need to be able to cope with 15 yet on 1% of occasions no arrivals • All outpatient referrals to consultants less than 10 per work day • Guaranteed 2 week cancer wait – almost impossible!

Randomness for an average of 5 arrivals per day

18%



16%



14%

Frequency of actual arriv als



12%



10%



8%



6%



4%



2%



0%



0



1 3%



2 8%



3



4



5



6



7



8 7%



9 4%



10 2%



11 1%



12 0%



13 0%



14 0%



Frequency 1%



14% 18%



18% 15% 10%



Dr Rod Jones (2003) Tel: 0118 3773511



Special & Common Combine

Standard deviation associated with healthcare demand

100,000 Emergency admissions Elective demand GP Referral A&E Poisson (Average^0.5)



Apparent standard deviation



10,000



1,000



100



10



1 1 10 100 1,000 10,000 100,000 1,000,000



Demand (admissions, FCE, GP referrals, A&E attendances)



Dr Rod Jones (2003) Tel: 0118 3773511



Elective demand

Total Elective Demand (ON + DC) in Surgical Specialties

22,500 22,000



Annual Demand



21,500



21,000



20,500 Poisson statistics suggests that one standard deviation should be 150, however, actual is higher than twice this value. The likelihood of deviation from the expected 'average' is high.



20,000



19,500



93



94



95



96



97



98



99



00



01



92 /



93 /



94 /



95 /



96 /



97 /



98 /



99 /



00 /



19



19



19



19



19



19



19



19



20



Dr Rod Jones (2003) Tel: 0118 3773511



20



01 /



02



Implications

• Size for financial stability - very much larger than any PCT • HRG’s - 95% have fewer than 1,000 p.a. thus unable to forecast prices • Why so much contract negotiation? • Size of A&E, bed pools, etc, etc • Not able to guarantee performance targets except with excess resources • Booked admissions initiative needs statistical support

Dr Rod Jones (2003) Tel: 0118 3773511



Queuing theory

• A.K. Erlang - line not available to next caller • Now widely used in industry

– supermarket, bank, petrol station queues, etc



• Healthcare applications

– A&E resources & waiting time – ICU beds – Turn-away experienced by any bed pool

Dr Rod Jones (2003) Tel: 0118 3773511



Turned-away or join the queue

• When arrivals exceed resources you either go elsewhere or join a queue • Hence - trolley waits, cancelled operations, borrowed beds, hidden queues • Best illustrated by plotting % occupancy vs bed pool size



Dr Rod Jones (2003) Tel: 0118 3773511



Throughput (per bed)



Dr Rod Jones (2003) Tel: 0118 3773511



Benchmarks - size



Dr Rod Jones (2003) Tel: 0118 3773511



Benchmarks – why?

Region Average Number of Acute Beds per NHS Trust 425 440 390 330 260 350 380 370 Average weighted Occupancy 80% 80% 82% 85% 87% 87% 85% 88% Average weighted Turn-away 0.8% 1.4% 2.0% 4.4% 4.7% 4.9% 5.3% 6.5%



Trent Northern South & West North Thames Anglia & Oxford West Midlands North Western South Thames



Dr Rod Jones (2003) Tel: 0118 3773511



Let’s sweat those assets



Dr Rod Jones (2003) Tel: 0118 3773511



Medical bed planning



Dr Rod Jones (2003) Tel: 0118 3773511



Medical bed planning



Dr Rod Jones (2003) Tel: 0118 3773511



Bed days

• Is a fundamental time-related unit of healthcare demand • Can be diminished by shifts to other healthcare settings and new methods • Can be converted to beds by adding the appropriate occupancy



Dr Rod Jones (2003) Tel: 0118 3773511



Is 75% day case achievable?

Specialty General Surgery

ON EM ON EM ON EM



LOS (days) 0 1

7% 8% 4% 16% 8% 9% 25% 20% 15% 18% 24% 22%



2

25% 16% 26% 13% 17% 12%



Urology



T&O



Dr Rod Jones (2003) Tel: 0118 3773511



Hidden Gain

• 0 LOS patients increase daytime occupancy leading to that part of A&E trolley waiting due to unavailable beds • 1 day LOS can potentially be treated as day case – the hidden consequence of insufficient day case resources • 2 day LOS are potential day case candidates if intensive input is available • Short stay emergency imply need for streaming of patients • The above do not save overnight beds but reduce daytime occupancy to the point that the ‘system’ (including A&E) starts to work again



Dr Rod Jones (2003) Tel: 0118 3773511



HRG-based LOS variation

HRG F82

700 600



500



This is probably as good as it gets in terms of iso-resource. Annual average LOS still varies between 2.33 to 2.75 days via variation in total beddays



Frequency



400

Average LOS = 2.5 days



300



200



100



0 0 1 2 3 4 5 6 7 8 9 10 >10



LOS (days)



Dr Rod Jones (2003) Tel: 0118 3773511



Controlling LOS

• Are the clinicians responsible?

– Studies in USA show that clinicians take on the LOS of the hospital where they work – By implication it is the ‘system’ rather than the clinician per se that is responsible – Change the system, provide the resources and the clinicians will deliver on LOS & day case rates



• All within the context of variation in LOS

– Concentrate on the system and be very clear about how the LOS distribution will change – Never measure success by average LOS always use the LOS distribution – An increase in day case rate should increase average LOS!

Dr Rod Jones (2003) Tel: 0118 3773511



Optimum efficiency

• Gain benefits of scale

– L&D hospital only has 2 bed pools – ‘surgical’ and ‘medical’



• Analyse daily occupancy by HRG to create specialty pools within the larger pool • Remove all 0 LOS patients to other settings • Make the shift to 75% DC sooner rather than later

Dr Rod Jones (2003) Tel: 0118 3773511



Hot & Cold Sites?

• Forfeits economy of scale • Elective demand is just as variable as emergency demand • Implies adequate bed provision on both sites • Ignores realities of medical bed demand • Same effect if an elective factory opens nearby

©Dr Rod Jones (2009) hcaf_rod@yahoo.co.uk



Conclusions

• Understanding randomness is important • A little bit of queuing theory goes a long way to explaining a lot of things • Some things are mathematically impossible - unfortunately they are part of your performance targets! • If planning was that easy we would all have been doing it years ago

Dr Rod Jones (2003) Tel: 0118 3773511




Related docs
Other docs by Dr Rod Jones