Development of an Economic Model for the Management of Upper Gastrointestinal Disease in Primary...

Development of an Economic Modelfor the Management of UpperGastrointestinal Disease in Primary CarePreliminary Findings

Alan Haycox, Mike Butterworth, Tom Walley and Stuart Barton

Prescribing Research Group, Department of Pharmacology and Therapeutics, University ofLiverpool, Liverpool, England

Abstract Health economic models for identifying therapeutic options that maximisehealth benefits from limited healthcare resources are being developed in a numberof therapeutic areas. The development of such a model for upper gastrointestinal(UGI) symptoms to support decision-making by primary care clinicians is ofparticular importance, given the prevalence of this symptomatology. This eco-nomic model was based upon the clinical guidelines aimed at improving themanagement of UGI disorders at the primary care level that were developed bythe International Gastro Primary Care Group. This paper discusses the derivation,methodology and results of the economic model developed to assess the resourceimplications arising from these clinical guidelines.

In order to construct the economic model, it was necessary to identify thefollowing:• every therapeutic pathway followed by patients• resource use along each pathway• the probabilities of following alternative pathways.

One crucial factor underlying the interpretation of results obtained from anyeconomic model is the time period covered by the model. The model presentedhere analysed the initial 12-month treatment period of ‘new’ patients presentingwith UGI symptoms. In order to test the implications of a longer term perspective,the model is currently being developed to analyse resource use over a 24-monthperiod.

The model demonstrates that utilising the predominant symptom approach tothe diagnosis and treatment of patients with UGI disorders appears to providesignificant benefits in terms of patient management and effective resource use.This factor, together with the more intensive use of Helicobacter pylori eradica-tion therapy, provides the potential to reduce the cost of drugs for the treatmentof UGI disorders by approximately 15% in the UK.

A major strength of the model is its adaptability to a wide range of clinical andcost scenarios. Such adaptability enables the model to effectively reflect the po-tential resource implications in countries exhibiting significantly different levels

REVIEW ARTICLE Pharmacoeconomics 1998; 14 Suppl. 2: 11-231170-7690/98/0002-0011/$06.50/0

© Adis International Limited. All rights reserved.

of cost and patient management. In this manner, the model provides one valuablemethod by which clinicians can be supported in optimising the management ofUGI disorders within current resource constraints.

This paper describes the methods by which theInternational Gastro Primary Care Group (IGPCG)guidelines were incorporated into an economicmodel and the preliminary results obtained fromthis model. The aim of the model was to evaluatethe comparative resource demands placed on gen-eral practitioners (GPs) by patients following themany different management pathways that areavailable for upper gastrointestinal (UGI) disease.The economic importance of this therapeutic areais illustrated by the fact that over 5% of GP con-sultations relate to UGI symptoms and, therefore,the role of the GP is crucial in achieving efficientpatient management in this therapeutic area.[1] TheGP also faces a bewildering array of diagnostic andtreatment options in attempting to identify the op-timal therapeutic pathway to be followed by eachindividual patient.[2] Evidence concerning thecomparative costs and effectiveness of differenttherapeutic strategies (e.g. early as compared withlate recourse to endoscopy) is still being gatheredand is frequently found to be contradictory.[3,4] Insuch circumstances, economic guidance is requiredto supplement the clinical guidance being promul-gated in this area.[5]

Economic analyses are normally undertakenfrom a ‘societal’ perspective, which comprehen-sively evaluates the implication of the burden ofdisease on all parts of society.[6] While acceptingthe theoretical superiority of such a focus, this analy-sis restricts itself to evaluating the resource impli-cations that would be imposed on the GP by thedifferent therapeutic strategies. The complexityof the analysis would be greatly increased by in-corporating costs imposed upon other parts of thepublic purse and privately borne costs. Therefore,at this stage it was felt that it was justifiable tolimit the resource perspective of the analysis tothat of the key decision-maker – the primary carephysician.[7]

The evaluation is in the form of a cost-effective-ness analysis in which the effectiveness of treat-ment is evaluated in terms of the anticipated timewithout symptoms or toxicity that would arise fromeach of the treatment options. More accurately, theoutcome is related to time without symptoms se-vere enough for the individual to seek medicalevaluation in the form of a GP visit. UGI symptomsare a very common complaint, with 40% of adultsexperiencing such symptoms during any 6-monthperiod,[8] and it is only when the frequency or se-verity of symptoms passes a certain ‘threshold’ thata medical opinion will be sought. The comparativesuccess of the treatment options is measured interms of the ability of the different therapeuticstrategies to maintain symptomatology below thissymptom trigger at which formal medical interven-tion is sought.[9]

1. Shifting the Focus from Efficacyto Effectiveness

Clinical trials evaluate the efficacy of therapy inthe hands of specialists in an environment of clin-ical certainty. In contrast, economic models focuson evaluating the effectiveness of therapy in a realworld characterised by clinical uncertainty, and

Table I. Measuring therapeutic effectiveness in the ‘real world’,which is characterised by clinical uncertainty, and where therapy isprescribed and consumed under less than perfect conditions

Outcome in upper gastric disease may differ from trials becauseof the following:a

Patients are less rigorously selected

Patients are less rigorously monitored

Drug regimen, duration and dose become dependent on patientcompliance

Therapy is provided by nonspecialists

There is likely to be less specialist diagnostic equipment availableto support clinical decision-making

a It should be noted that identifying the implications of such variations represents the bedrock of outcomes research.

12 Haycox et al.

© Adis International Limited. All rights reserved. Pharmacoeconomics 1998; 14 Suppl. 2

where therapy is prescribed and consumed underless than perfect conditions.[10] The economic fo-cus on effectiveness, therefore, implicitly incorpo-rates the impact of real world problems such asnoncompliance and suboptimal drug utilisation.[11]

Table I outlines a range of confounding factors thatmay cause real world costs and benefits to differfrom those identified in trials, while the principaldata sources of information on therapeutic effec-tiveness are outlined in table II.

In order to focus on effectiveness, economicmodels may need to adjust results obtained in clin-ical trials to reflect mainstream clinical practice.[12]

This adjustment can be achieved through the useof medical databases or the development of an ex-pert panel to identify costs and benefits that arerelevant to clinical practice rather than driven bythe research protocol.[13] For example, a researchprotocol may require all patients with dyspepsia toundergo endoscopy, significantly distorting thestructure of resource use away from that found inclinical practice. An economic model would adjustthe costs to reflect the proportion of patients actu-ally provided with endoscopy to more accuratelyreflect the costs incurred in practice.

Despite such theoretical advantages, it is impor-tant to recognise the methodological difficultiesthat are introduced into the analysis as a conse-quence of focusing evaluations on effectivenessrather than efficacy.[14] In particular, drug effec-tiveness will be affected by a large number of con-founding variables (such as compliance), which in-terfere with the ‘pure’ relationship betweenprovision of therapy and patient response. The in-terpretational difficulties encountered by suchstudies are inevitably greater than those encoun-tered in clinical trials and, therefore, the resultsobtained from studies of effectiveness must be in-terpreted with caution.[15] However, despite suchreservations, analyses of this nature represent a vi-tal component of informed decision-making in anytherapeutic area.[16]

2. The Information Needs Underlyingthe Economic Model

The data used to operationalise the economicmodel were obtained from a number of sources. Acomprehensive review of the existing literaturewas undertaken to ensure that, wherever possible,the results of individual effectiveness studies havebeen incorporated into the model both in terms ofcosts and treatment pathways.

In cases where different sources of evidenceconcerning comparative costs and effectiveness atcertain stages of the model were available, an adhoc synthesis that attempted to reflect the ‘weight’of the available evidence was made. In all cases,the aim was to ensure that the evidence built intothe economic model reflected, as far as possible,the comparative costs and benefits to GPs thatwould be likely to arise in practice in the provisionof care to patients with UGI symptoms. Both thesearch strategy and the criteria for inclusion of ev-idence were as wide as possible, in order to ensurethat all potential sources of evidence could be in-corporated into the model.

Equally, however, such comprehensiveness alsoimplies that the result of analyses of significantlydifferent quality (from large scale randomised clin-ical trials to small scale observational studies)will be included in the dataset underpinning themodel.[17]

In cases where the research evidence appearedto be absent or contradictory, the comparative costsand clinical effectiveness of different therapeutic

Table II. Sources of information concerning effectiveness: theprincipal data sources and analytical techniques used to spanefficacy and effectiveness, and to extrapolate to long term outcomefrom short term trials are provided in this tablea

Analyses attached to randomised clinical trials

Medical databases

Modelling techniques (particularly decision analysis and marker)

Observational studies

Systematic reviews of the literature

Expert clinical opinion

a Note that a comprehensive outcome analysis frequentlysynthesises information from multiple sources.

Development of an Economic Model 13


strategies were based on ad hoc synthesis of datafrom several sources, including expert opinion.Such an approach is unavoidable in the absence ofclinical effectiveness data.[18]

Clinical guidelines attempt to promote effectivemedical practice by identifying the most appropri-ate method of treatment provision for any particu-lar condition in order to reduce ‘inappropriate’treatment and lead to an improvement in the ‘qual-ity’ of patient care.[19] However, to be of practicalvalue such guidelines should have their estimatedresource implications evaluated, in order to ensurethat they are achievable from within available re-sources. Evaluations of clinical guidelines thatconcentrate entirely on either costs or benefits donot enable clinicians to make the balanced judge-ment required to provide the best possible carefrom within limited resources. The development ofan economic model provides one method of sup-porting clinical decision-making through the use oftimely and relevant information. Such a model alsoallows clinicians to evaluate the impact of custom-ising the model to make it more appropriate to theirlocal environment.

Attempting to develop an economic model isprobably the quickest method of becoming disen-chanted with the current emphasis that is beingplaced on ‘evidence-based medicine’. It becomesobvious that in many areas appropriate ‘evidence’has not yet been generated, while in other areas theavailable ‘evidence’ has been generated in a man-ner that makes its relevance to real-life clinicaldecision-making strictly limited. Evidence-basedmedicine should be utilised in areas where evi-dence is available, but the breadth of informationrequired to develop an economic model will inev-itably require assumptions to be made in certainareas. It is crucial that the assumptions are trans-parent, objective and based on the best availableevidence or opinion.[20] Such assumptions mustalso be readily modifiable in the light of new evi-dence that could improve the real world accuracyof the model. In order to quantify an economicmodel, it is necessary to identify the following:

• every potential therapeutic pathway that couldbe followed by patients

• the nature and level of resources consumed bypatients at each stage of each pathway

• the probabilities of patients following alterna-tive pathways.Every branch of the model represents a treat-

ment pathway with an associated level of resourceuse and probability of use by patients.[21] The qual-ity of the underlying clinical algorithm is crucial,given that the construction of elaborate economicmodels based on poor quality clinical algorithmswould be akin to constructing an elaborate eco-nomic castle on foundations of sand. The complex-ity and nature of the model is entirely dependenton the complexity and nature of the underlyingclinical algorithm. In this sense, the economicmodel has no independent ‘life’ but is entirely de-pendent for its existence on the form and nature ofthe clinical model.

3. Applying Economic Modelling to theIGPCG Guidelines

In developing an economic model, the key re-quirement is to ensure transparency in the underly-ing assumptions and methodology. This enablesthe quality and appropriateness of the assumptionsand methodology to be challenged and, where nec-essary, adapted to local circumstances. Crucially,such transparency also enables the model to be im-proved as higher quality and more comprehensiveempirical data become available.

The use of economic modelling enables theanalysis to be extended in a number of ways.Firstly, the results obtained in studies of each indi-vidual part of the treatment algorithm can be incor-porated into an evaluation of the comparative costeffectiveness of the overall system of care. Sec-ondly, the analysis can be extended beyond the lim-ited time-scales assessed in empirical investiga-tions to evaluate the long term cost implicationsarising from different structures of treatment pro-vision. Modelling also enables data from a wide va-riety of sources to be incorporated into the analysis.Inevitably, this means that the data underpinning

14 Haycox et al.


different elements of the model will vary signifi-cantly in quality, and that the analysis will have torely on ‘expert opinion’ for part of the systemwhere objective empirical data have yet to be gen-erated. Modelling also enables the undertaking of‘what if’ analyses, which test the responsiveness ofthe overall system to changes in the most crucial‘cost drivers’ underpinning care provision. Suchanalyses provide an invaluable guide to providersof care as regards the individual elements withinthe overall pattern of care provision that are mostcrucial in determining the cost effectiveness of thesystem as a whole.

Economics is, by its nature, a comparative dis-cipline. The assessment of cost effectiveness in-trinsically depends on the bench mark againstwhich any therapeutic strategy is being evaluated.For the purposes of this analysis, the chosen com-parator is the cost of treatment currently being pro-vided to patients with dyspepsia. The perspectiveunderlying the economic model reflected the goalof improving the diagnosis and treatment of UGIdisease at the primary care level. The model there-fore analysed the patient flows and treatment strat-egies that would result at the primary care level ifthe strategy utilised by GPs to manage patientswith UGI disorders conformed with IGPCG guide-lines. The perspective taken is that of the healthservice, with the analysis incorporating the costsof consultations, prescriptions and hospital ser-vices but specifically excluding both privatelyborne costs and costs borne by other parts of thepublic sector. The choice of appropriate methodol-ogy depends on the form, nature and context of theanalysis being undertaken. For the purposes of ouranalysis, it was decided that the most appropriatestructure would be a cost-effectiveness analysisincorporating variations in both resource use andeffectiveness. The effectiveness of therapy wasmeasured in terms of each patient’s ‘time withoutsymptoms or toxicity’, utilising the implicit as-sumption that treatment was assumed to be effec-tive when further contacts with the primary carephysician were unnecessary.

The development and quantification of the eco-nomic model required each stage of clinical deci-sion-making to be clearly and uniquely distin-guished. The costs associated with each stage ofthis process, together with the treatment pathwayslinking each stage, were then combined in order toestimate the costs of each individual branch of thealgorithm. The total cost of the overall structure formanaging patients with UGI disorders could thenbe calculated by weighting the cost of each indi-vidual treatment path by the proportion of patientstravelling along each path. The analysis assessedthe diagnostic and treatment needs of all patientswith UGI disorders being managed at the primarycare level. Thus, patients who were identified oninitial examination as requiring specialist care (e.g.patients with gastric cancer) were excluded fromthe analysis. While recognising that the actual costof managing individual patients will vary in rela-tion to their age and co-morbidities, the analysisfocused on identifying an ‘average’ cost for pa-tients following each of the individual treatmentpaths indicated by the clinical algorithm.

4. Costing of the Economic Model

Evaluating the cost of resources consumed inthe provision of care to patients with UGI disor-ders at the primary care level requires a 3-stageanalysis:1. Identify all resources consumed during the pro-cess of care provision.2. Measure the amount of each type of resourceconsumed.3. Place a value on such resource consumption.

The nature of the resource and the unit costplaced on the resources available to GPs in man-aging patients with UGI symptoms are providedin table I on page 12 of this issue.[22] The unit costsare derived from a range of sources, but each at-tempts to reflect the average cost (or price) of suchresources within the National Health Service(NHS). The amount of each type of resource con-sumed will depend on the pathway followed by eachpatient. Each stage of the patient pathway is asso-ciated with a defined level of resource consumption



Sus

tain

ed r

elie

f

Rel

apse

; PP

I [56

] (75

%);

H2R

A a

ndpr

okin

etic

[56]

(25

%)

Sus

tain

ed r

elie

f

Rel

apse

; PP

I [56

] (75

%);

H2R

A a

ndpr

okin

etic

[56]

(25

%)

Sus

tain

ed r

elie

f

Rel

apse

; H2R

A [5

6] (

50%

); P

PI [

56]

(40%

); H

2RA

and

pro

kine

tic [5

6] (

10%

)

Sus

tain

ed r

elie

f

Rel

apse

; PP

I [56

] (75

%);

H2R

A a

ndpr

okin

etic

[56]

(25

%)

64%

36%

64%

36%

64%

36%

64%

36%

Long

term

relie

f

Chr

onic

patie

nt

Long

term

relie

f

Chr

onic

patie

nt

90%

10%

90%

10%

Long

term

relie

f

Chr

onic

patie

nt

Long

term

relie

f

Chr

onic

patie

nt

90%

10%

90%

10%

40%

60%Ref

er

Sus

tain

edre

lief

36%Rel

apse

;H

2RA

[56]

PP

I [56

] (75

%);

H2R

A a

ndpr

okin

etic

[56]

(25

%)

85%

15%Ref

er;

PP

I [56

] (75

%);

H2R

A a

ndpr

okin

etic

[56]

(25%

)

Sus

tain

edre

lief

76% Rel

apse

;P

roki

netic

[56]

H2R

A [5

6] (

50%

); P

PI

[56]

(40

%);

H2R

A a

ndpr

okin

etic

[56]

(10

%)

24%

Tem

pora

ry o

rlit

tle r

elie

f

25% Lo

ng te

rmre

lief

75% T

empo

rary

or

little

rel

ief

34% Lo

ng te

rmre

lief

66%

Hea

rtbu

rn a

nddy

spep

ticsy

mpt

oms;

Pro

kine

tic [5

6]

15%H

eart

burn

;H

2RA

[56]

85%

GO

RD

-typ

e sy

mpt

oms

Fig

. 1. G

astr

o-oe

soph

agea

l ref

lux

dise

ase

(GO

RD

)-ty

pe s

ympt

oms.

H2R

A =

H2

rece

ptor

ant

agon

ist;

PP

I = p

roto

n pu

mp

inhi

bito

r. N

umbe

rs in

squ

are

brac

kets

are

the

num

bers

of d

ays

for

whi

ch th

e dr

ug is

pre

scrib

ed.

16 Haycox et al.


Sustained relief

Chronic patient90%

10% Sustained relief

Chronic patient90%


Chronic patient90%

10%Sustained relief

Chronic patient90%


Chronic patient90%


Chronic patient90%

10%

Sustained relief

Chronic patient90%


Chronic patient90%


Chronic patient90%


Chronic patient90%


Chronic patient90%


Chronic patient90%

10%Sustained relief

Chronic patient90%


Chronic patient90%


Chronic patient90%

10%

PPI [42]

25%

H2RA [42]

70%Prokinetic[42]

5%

PPI [42]

25%

H2RA [42]

70%Prokinetic[42]

5%

Refer

80%

PPI [42]

25%

H2RA [42]

70%Prokinetic[42]

5%

PPI [42]

25%

H2RA [42]

70%

Prokinetic[42]

5%

PPI [42]

25%

H2RA [42]

70%Prokinetic[42]

5%

Refer

10%

Reassess

90%

Sustainedrelief

75%

Temporaryor littlerelief

25%

Prokinetic[42 (80%);56 (16%);168 (4%)]

20%

Relapse

20%

Long termrelief

75%

Sustainedrelief

80%

Reassess

90%

Refer

10%

Temporaryor littlerelief

25%

Prokinetic[42]

Fig. 2. Dysmotility-type symptoms. H2RA = H2 receptor antagonist; PPI = proton pump inhibitor. Numbers in square brackets are thenumbers of days for which the drug is prescribed.



and a defined set of probabilities of progressing tofurther stages of the overall pathway. The amountsof different resources consumed will therefore de-pend on the therapeutic pathway followed by eachindividual patient.

The estimate of resource use at each stage of thepatient pathway is based on analysis of medicalrecords, supported, where necessary, by expertopinion. Again the aim was to evaluate averageresource use for real patients in normal clinicalpractice rather than in the ‘rarefied’ atmosphere ofa clinical trial.

Average unit costs (obtained from local or na-tional sources) were used to place a price on thequantity of physical resources consumed. Whileagain accepting the theoretical superiority of mar-ginal social costs as a basis for the costing ofresource, it was not felt that the costing basisutilised in this study would significantly bias thecost analysis undertaken. All costs were obtainedin pounds sterling for the 1995/1996 financial yearand then adjusted to a 1996/1997 price base.

The economic model was developed in the formof a decision-tree analysis using software devel-oped by ‘TreeAge’, which provided a valuablegraphic representation of the structure underlyingthe model. The range of potential treatment pathsgenerated by the management algorithm, togetherwith the associated cost, were estimated throughcomprehensive literature reviews, supported, wherenecessary, by expert opinion.

The quantification of the economic model,based on the clinical algorithm developed by theIGPCG, required a detailed description of thefollowing:• the individual pathways taken by each patient• the level of resources consumed by patients at

each stage of the diagnostic pathway• the transitional probabilities that reflect the like-

lihood of patients following each individualpathway.The drug costs used in the resource analysis

were obtained from the British National Formu-lary. The costs of outpatient referrals plus inpatientstay (gastroenterology) and diagnostic tests (en-

doscopy, barium meal, biopsy, pH measurement,Helicobacter pylori test and blood tests) were ob-tained from the NHS executive in the UK. A pri-mary care consultation was assumed to cost £10(normal) or £15 when extended to enable the GP toprovide lifestyle advice. Finally, a patient who de-velops a chronic UGI disease is assumed to con-sume health service resources valued at £236 perannum, consisting of a mixture of pharmacologicaland nonpharmacological support. Full details ofthe cost structure employed are available from theauthors. In order to add the necessary time dimen-sion to the model, the resource implications wereanalysed over a defined period of 12 months. Thistime limitation is of particular importance in thecase of patients with chronic disease who will im-pose resource demands on the health services farbeyond this period.

The basic economic model consisted of 226 dis-tinct clinical pathways, which reflect the widerange of clinical options available to the primarycare physician in diagnosing and treating patientspresenting with UGI symptoms. All patients pre-senting with UGI symptoms have their full historyand presentation assessed by the GP. It is estimatedthat 13% of patients will present with ‘alarm’symptoms requiring referral, 5% will have a prob-lem related to use of NSAIDs and 5% will have atime-limited symptomatology that is adequatelytreated by reassurance and lifestyle advice. Underthis scenario, 77% of patients exhibit symptomsindicative of either gastro-oesophageal reflux dis-ease (GORD; 35%), ulcer (32.5%) or dysmotility(32.5%). The therapeutic strategy adopted for pa-tients presenting with GORD or dysmotility-typesymptoms are presented in full in figures 1 and 2.Unfortunately, the range of strategies required forpatients presenting with ulcer-type symptoms istoo complex to present in a single diagram and anoverview is provided in figure 3, with full detailsbeing available from the authors on request.

5. Results

This research was initiated to assess the extent towhich an economic model that would be of practical

18 Haycox et al.


value in evaluating the comparative cost effective-ness of management options for patients with UGIsymptoms at the primary care level could be devel-oped. In reporting the results from the model, themain aim is for transparency and clarity in provid-

ing the overall structure of results obtained, whiletrying to avoid being buried beneath more detailedresults that would be applicable to only a smallnumber of patients. Obviously, the results obtainedmust be interpreted with caution, given that the

Temporary orlittle relief

30%

65%

Sustained relief

35%

Maintenance H2RA [140]

Therapeuticstrategy

1

Temporary orlittle relief

35%

73%

Sustained relief

27%

Maintenance prokinetic [140]

90%

Sustained relief

10%

Chronic patient

90%

Sustained relief

10%

Chronic patient

90%

Sustained relief

10%

Chronic patient

Therapeuticstrategy

2

Therapeuticstrategy

3

Ulcer-typesymptoms

No ulcerhistory

60%

Hp testingnotundertaken

50%

H2RA[28]

80%

Prokinetic[28]

5%

Refer

15%

Hp testundertaken

50%

Hp test −ve

47% Endoscopy −ve

10%

Previouslydocumented ulcer

40%

Long termrelief

65%

H2RA [ 28]

37%

PPI [28]

10%

Hp eradication

53%

Endoscopy +ve

90%

Hp test +ve

53%

Long termrelief

70%

Therapeuticstrategy

1

Fig. 3. Ulcer-type symptoms: an overview. Hp = Helicobacter pylori; H2RA = H2 receptor antagonist; PPI = proton pump inhibitor;−ve = negative; +ve = positive. Numbers in square brackets are the numbers of days for which the drug is prescribed. Because ofspace limitations in the context of this publication, the therapeutic sub-strategies in the boxes on this figure have been omitted. Thesecan be obtained from the author on request.



model remains in the preliminary stage of develop-ment. Obtaining high quality feedback representsa crucial stage in the future development of the modeland, in this regard, full details of the assumptionsconcerning resource use and treatment pathways areavailable from the authors on request.

The results were analysed in the following 3 ways:1. The comparative cost of different patient path-ways was analysed to identify potential methods bywhich ‘high cost’ pathways could be accurately fo-cused on patients in greatest need of such intensivesupport.2. The overall resource use and cost of implement-ing the IGPCG algorithm (the ‘basic’ model) wascompared with current resource use and cost in-curred in supporting patients with UGI symptomsat the primary care level.3. The cost implications of alterations in the under-lying structure of patient care (e.g. the resource im-plications of moving to a system of open accessendoscopy or implementing a system of blind erad-ication of H. pylori) were compared with the costof the ‘basic’ model.

5.1 Comparative Cost of Different Pathways

The average cost incurred during the first yearof management of patients with UGI symptoms byprimary care physicians is estimated to be £165.This estimate includes all of the diagnostic andtreatment resources utilised at the primary carelevel in supporting these patients. However, thisaverage hides a wide range of costs that are im-posed by patients presenting with widely divergentproblems. The range of costs exhibited by differentindividual patient pathways varied from £10 to£989 over the 12-month period covered by theanalysis.

The 3 main symptom classifications used by theIGPCG algorithm are ‘ulcer-type symptoms’,‘GORD-type symptoms’ and ‘dysmotility-typesymptoms’. The average cost of treating patientsexhibiting ulcer-type symptoms and patients ex-hibiting GORD-type symptoms was estimated tobe £193 and £140, respectively – again with indi-vidual patients varying widely around this average.

The average cost of patients exhibiting dysmotility-type symptoms was £128, with less marked varia-tion between individual patient pathways.

While such variations between such individualpatient types are of interest, the overriding messagederived from the analysis is the crucial importancein any cost-effective system of healthcare of ensur-ing that patients are accurately channelled downappropriate pathways. Thus, more accurate and in-tensive initial patient assessment to identify thefundamental cause of UGI symptoms will go a longway towards preventing the significant waste ofresources that currently occurs as a consequence ofinappropriate treatment.

5.2 Comparison of Cost with Current Treatment

Accurate identification of the current cost oftreatment provision within the NHS represents awidely recognised area of uncertainty, particularlyat the primary care level. A number of medical in-formation companies have developed databasesanalysing the prescribing and referral activity oflarge numbers of GPs. One of the leading compa-nies in this area is Intercontinental Medical Statis-tics (IMS). To obtain the best available estimate,the IMS MediPlus® database was analysed for pa-tients presenting with UGI symptoms. While thediagnoses as recorded by GPs on the IMS databaseare not necessarily proven diagnoses, this methodat least provided a ‘benchmark’ against which theestimated cost of implementing the IGPCG algo-rithm could be compared. The IMS database indi-cated that an average of 4.4 prescriptions per an-num are currently provided to patients presentingto primary care physicians with UGI symptoms.Implementing the IGPCG algorithm is estimated toreduce the number of prescriptions to an averageof 1.9 per patient, with approximately 70% of pa-tients treated in accordance with the algorithm re-ceiving long term (defined as >6 months) symp-tomatic relief from their first course of therapy. Itis estimated that implementing the algorithm couldsave approximately £70 million out of a currenttotal drug cost of £488 million. Such cost reductions

20 Haycox et al.


arise mainly from improving the diagnosis andsymptomatic management of patients and the moreintensive eradication of H. pylori infection.

5.3 ‘What If’ Analyses

The development of an economic model pro-vides the opportunity to interact with and influ-ence, as well as inform, clinical practice. Therehave been many suggestions for changes in policyor practice that are held to significantly improvethe management of patients with UGI symptoms.The availability of an economic model facilitatesthe calculation of the estimated changes in patientflow and resource use that would result from theimplementation of such suggestions. In this man-ner, the implications of changes in the status quocan be estimated prior to actually implementingsuch changes in patient care. The structure of theeconomic model presented here was adapted to un-dertake 2 main ‘what if’ analyses:1. The impact of ‘open access’ endoscopy (i.e. theability of GPs to refer patients for endoscopy with-out requiring a prior outpatient referral to a gastro-enterologist) on patient flows and resource use.2. The impact of ‘blind eradication’ of H. pylori(i.e. the earlier and more intensive use of drug ther-apy to eradicate H. pylori in patients exhibitingulcer-type symptoms) on patient flows and resourceuse.

Open access endoscopy was estimated to reducethe average cost of patient management by 5%.This overall saving occurred because the reductionin the cost of outpatient referrals was estimated tooutweigh the additional demand for endoscopyarising from the GP’s direct access to this service.[23]

The use of blind eradication in patients exhibit-ing ulcer-type symptoms was estimated to reducethe average cost of patient management by 7%.This remains an area of controversy, with the ma-jority of studies from the US strongly supportingthe more intensive use of eradication therapy,[24]

while a more critical response to the intensive useof eradication therapy is evident in most UK stud-ies.[25] While many additional factors need to betaken into account, the model indicates that signif-

icant potential savings could result from more ex-tensive use of blind eradication therapy.

The ability to undertake such ‘what if’ analysesrepresents one of the major advantages of incorpo-rating economic modelling into the process of de-veloping clinical guidelines. It is important to re-emphasise that the aim of the analysis remains thedesire to optimise the clinical management of pa-tients with UGI symptoms within the resourcesavailable to primary care physicians. The eco-nomic model greatly facilitates this aim by ensur-ing that clinical decisions can be taken with know-ledge of their resource implications on GPs’limited healthcare budgets.[26]

6. Sensitivity Analysis

Uncertainty arises from a number of sources ineconomic modelling. It can arise in relation to thevariable quality of the underlying data, or in rela-tion to the quality and appropriateness of the as-sumptions and methodology used to inform themodelling process. The reliability that should beplaced on the model depends upon its robustnessin response to ‘reasonable’ variations in the crucialcost drivers underlying the model. Assumptionsconcerning pathways that affect 1 patient in 10 000can afford a level of uncertainty that could not bejustified in an assumption that affects the pathwaysof 1 patient in 10.

The incorporation of a comprehensive sensitiv-ity analysis is a crucial component of any eco-nomic model.[27] Although the model has beenbased on the best available evidence, the amountof information required to quantify the modelmakes it inevitable that sources of evidence of dif-ferent quality will be combined. Some sources ofevidence are very powerful (e.g. a meta-analysis ofrandomised controlled trials), whereas othersources are less robust (e.g. expert but unprovenclinical opinion). The sensitivity analysis under-taken aims to test the sensitivity of the results toassumptions that are based on less powerful evi-dence and which are likely to significantly affectthe overall results. Such ‘cost drivers’ tend to occurat an early stage in the economic model, where



patients are being broadly divided into differenttreatment paths. As such, the sensitivity analysisconcentrates on evaluating the impact of variationsin assumptions at this early stage, given that evena significant cost variation is unlikely to affect theoverall structure of results if it affects only a smallproportion of patients.

The sensitivity analysis undertaken takes 2forms. Firstly, the crucial cost drivers underlyingthe model have been identified and the robustnessof the results are being evaluated in relation to ‘rea-sonable’ variations in these crucial elements of themodel. Secondly, Monte Carlo simulation is beingused to test the impact of simultaneous variationsin the crucial assumptions underlying the model.This approach allows a large number of individual‘patients’ to be hypothetically ‘treated’ in accord-ance with a range of potential scenarios that reflectdifferent sets of assumptions concerning the under-lying parameters of the model. The most importantconclusion from the analysis to date is that the re-sults obtained from the model appear to be robustin relation to the sensitivity analyses currently un-dertaken. However, further work concerning thisaspect of the analysis is ongoing.

7. Conclusion

At this stage, a definitive conclusion would beinappropriate, since refinement of the model is on-going. However, a number of general conclusionsarise from the analysis undertaken.

Firstly, and perhaps most importantly, the analy-sis demonstrates the feasibility and value of devel-oping an economic model in support of a clinicalalgorithm aimed at defining good clinical practicein the management of patients with UGI symp-toms. The form and nature of this relationship arecrucial. The analysis is not being ‘driven’ by eco-nomic considerations; rather, the economic modelis being developed to inform and support theachievement of good clinical practice. Prior to im-plementing a change in clinical management, cli-nicians need to be fully informed of both the costsand benefits of such a change. Only then will theybe able to assess the extent to which implementing

the proposed algorithm improves their ability toprovide healthcare to their entire patient popula-tion. Not surprisingly, many clinicians are unwill-ing to ‘sign a blank cheque’ by implementing algo-rithms whose resource implications have not beenfully identified. In the case study presented here,the clinician is in a ‘win-win’ situation in whichproposed improvements in patient care also lead toa potential saving of resources. The persuasivepower of any clinical algorithm will be signifi-cantly enhanced in cases where its potentialresource implications have been analysed prior toits implementation.

Secondly, the process involved in developingsuch an economic model inherently requires enor-mous detail concerning the structure of servicesprovided to patients at each stage of the clinicalalgorithm. Such a process, in itself, provides a use-ful check on the practicality, consistency and com-pleteness of the clinical algorithm. The develop-ment of an economic model thus forces the clinicalalgorithm to specify the nature of service deliveryat each stage of each potential treatment path.

Thirdly, the economic model has identified awide variation in costs imposed on the health ser-vice as a consequence of the treatment of UGI dis-ease. The primary factor underlying such varia-tions will be the nature of the clinical problemspresented by each individual patient, as this definesthe necessary clinical response (and hence resourceuse) from the primary care physician.

Within the general pattern of costs, it is impor-tant to recognise that patients with chronic diseaserepresent a significant and continuous drain on theresources available to the primary care physician.By using a rational approach to diagnosis and treat-ment, primary care physicians can minimise theproportion of such high cost patients. The conceptof cost effectiveness that underlies any economicanalysis inevitably depends on the time-scale un-der consideration. For example, the ‘investment’ ofdiagnostic resources in endoscopy, while expen-sive in the short term, is likely to generate informa-tion of enormous value in identifying a strategythat is most cost effective over the long term. In

22 Haycox et al.


addition, the investment of resources into H. pylorieradication therapy may keep the patient symptomfree (and hence cost free) far beyond the 12-monthperiod analysed in this model. Thus, it is importantthat the results obtained from the model are evalu-ated in the context of the time-scale under consid-eration.

The final point that should be emphasised is thatany economic model developed must be suffi-ciently flexible to incorporate local variations inresource costs and clinical practice. However, ther-apeutic strategies, i.e. for UGI disease, exhibitcommon strands that delineate ‘good practice’ ir-respective of the context. However, other strandsmust be allowed to reflect variations in local costlevels and clinical practice. While both the costsand structure of services examined in this currentanalysis are derived from UK experience, each el-ement of the model is sufficiently flexible to beadaptable to the patient management and coststructures that are operating in different countries.

References1. Jones RH. Clinical economics review: gastrointestinal disease

in primary care. Aliment Pharmacol Ther 1996; 10: 233-92. Johnsen R, Bernerson B, Straume B, et al. Prevalence of endo-

scopic and histological findings in suspects with and withoutdyspepsia. BMJ 1991; 302: 749-52

3. Saunders JHB, Oliver RJ, Higson DL. Dyspepsia: incidence ofnon-ulcer disease in a controlled trial of ranitidine in generalpractice. BMJ 1986; 292: 665-8

4. Brown C, Rees WDW. Dyspepsia in general practice. BMJ1990; 300: 829-30

5. Berwick DM. Harvesting knowledge from improvement.JAMA 1996; 275 (11): 877-8

6. Department of Health. Guidelines on good practice in the con-duct of economic evaluation of medicines. London: Depart-ment of Health, 1994

7. Eve R, Golton I, Hodgkin P, et al. Beyond guidelines: promot-ing clinical change in the real world. J Management Med1996; 10 (1): 16-25

8. Jones RH, Lydeard SE, Hobbs FDR, et al. Dyspepsia in Englandand Scotland. Gut 1990; 31: 401-5

9. Calman K. The profession of medicine. BMJ 1994; 309: 1140-310. Clancy CM, Kamerow DB. Evidence-based medicine meets

cost effectiveness analysis. JAMA 1996; 276 (4): 329-30

11. Russell LB, Gold MR, Siegal JE, et al. The role of cost-effec-tiveness analysis in health and medicine. JAMA 1996; 276(14): 1172-7

12. Evans RG. Manufacturing consensus, marketing truth: guide-lines for economic evaluations. Ann Intern Med 1995; 123(1): 59-60

13. Sheldon T. Problems of using modelling in the economic eval-uation of healthcare. Health Econ 1996; 5: 1-11

14. Freemantle N, Maynard A. Something rotten in the state ofclinical and economic evaluations? Health Econ 1994; 3: 63-7

15. Hillman A, Eisenberg JM, Pauly MV, et al. Avoiding bias in theconduct and reporting of cost-effectiveness research spon-sored by pharmaceutical companies. N Engl J Med 1991; 324(19): 1362-5

16. Freemantle N, Henry D, Maynard A, et al. Promoting cost ef-fective prescribing. BMJ 1995; 310: 955-6

17. Beek JR, Pauker SG. The Markov process in medical prognosis.Med Decis Making 1983; 3 (4): 419-58

18. Weinstein MC, Siegal JE, Gold MR, et al. Recommendationsof the panel on cost-effectiveness in health and medicine.JAMA 1996; 276 (15): 1253-8

19. Grimshaw JM, Russell IT. Effect of clinical guidelines on med-ical practice: a systematic review of rigorous evaluations.Lancet 1993; 342: 1317-22

20. Task Force on Principles for Economic Analysis of HealthcareTechnology. Economic analysis of healthcare technology – areport on principles. Ann Intern Med 1995; 122: 61-70

21. Sonnenberg FA, Beck JR. Markov models in medical decisionmaking: a practical guide. Med Decis Making 1993; 13 (4):322-38

22. Haycox A, Butterworth M, Walley T, Barton S. Developmentof an economic model for the management of upper gastro-intestinal disease in primary care: preliminary findings. Phar-macoeconomics 1998; 14 Suppl. 2: 11-23

23. Holdstock G, Wiseman M, Loehry CA. Open-access endoscopyservice for general practitioners. BMJ 1979; 1: 457-9

24. Vakil N, Fennerty MB. Cost-effectiveness of treatment re-gimens for the eradication of Helicobacter pylori in duodenalulcer. Am J Gastroenterol 1996; 91 (2): 239-45

25. Davenport PM, Morgan AG, Darnborough A, et al. Can prelim-inary screening of dyspeptic patients allow more effective useof investigational techniques? BMJ 1985; 290: 217-20

26. Ubel PA, Dekay ML, Baron J, et al. Cost-effectiveness in asetting of budget constraints. N Engl J Med 1996; 334: 1174-7

27. Clemens K, Townsend R, Luscombe F, et al. Methodologicaland conduct principles for pharmacoeconomic research.Pharmacoeconomics 1995; 8 (2): 169-74

Correspondence and reprints: Dr Alan Haycox, Departmentof Pharmacology and Therapeutics, University of Liver-pool, 70 Pembroke Place, Liverpool L69 3GF, England.



Development of an Economic Model for the Management of Upper Gastrointestinal Disease in Primary...

Documents

Transcript of Development of an Economic Model for the Management of Upper Gastrointestinal Disease in Primary...