- Email: [email protected]
Cost management in community-acquired lower respiratory tract infection
Cost Management in Community-Acquired Lower Respiratory Tract Infection PETERG. DAVEY, M.D., Dundee,UnrtedKingdom
Pharmacoeconomic analysis must determine a credible cost model, including healthcarerelated, non-healthcare-related, and intangible costs, and must identify important treatment outcomes. Outcomes can be ranked in terms of their sensitivity to different treatments or weighted in terms of cost. Two recent studies of community-acquired lower respiratory tract infection have shown that inadequate treatment is a frequent cause of repeat visits to the physician’s office, involving extra costs that could presumably be avoided by more effective first-line treatment. When economic data are to be incorporated into clinical trials, it is important to collect data outside as well as within the trial center and to concentrate on data likely to be of significance to the decision makers the trial is designed to influence. The best way of achieving internal validity in such trials is through randomization, which minimizes bias, chance imbalance, and confounding factors. The three major determinants of the costeffectiveness of treatment are efficacy, the value attached to preventing treatment failure, and the accuracy of diagnosis. The latter is important because a drug can only benefit patients if they have the disease for which it is an effective treatment.
From the Pharmacoeconomlcs Research Centre, Department of Clinical Pharmacology, Unlversrty of Dundee, United Kingdom. Requests for reprlnts should be addressed to Peter G. Davey, M.D., Department of Clinical Pharmacology, Ninewells Hospital and Medical School, Dundee, Taysrde. DDl 9SY, United Krngdom.
6B-20s
December 29, 1995
The American Journal of Medicine
ower respiratory tract infections (LRTI) rank high on the list of reasons patients visit their family physicians; they are also among the most frequent indications for which antibiotic treatment is prescribed. Not surprisingly, the cost-effectiveness of antibiotic therapy for community-acquired (LRTI) has become the focus of intense scrutiny. In performing a comprehensive and balanced economic analysis of the use of antibiotics for the primary care treatment of LRTI, it is important to address four key questions: 1. What is a credible cost model? 2. What are the important outcomes for antibiotic treatment? 3. How should economic analysis be incorporated into randomized clinical trials? 4. What is the relevance of diagnostic techniques to the economic analysis of antibiotic treatment? This article will discuss the significance of these issues and explore some of the ways in which they can be integrated into analyses of cost management in community-acquired LRTI. Identification of the most important considerations for pharmacoeconomic analysis can not only provide answers for decisions we have to make today but can also help pinpoint questions for future research.
L
COST MODELS The methods used in a credible cost model must be statistically valid, and the various cost estimates must be reasonably accurate. The first steps in developing such a model are to identify the potential costs of failed treatment and to separate the components of these costs into healthcare financial costs, nonhealthcare financial costs, and intangible costs, which include those costs that are difficult to value financially. (In this review the terms healthcare and nonhealthcare will be used in preference to the more ambiguous adjectives direct and indirect to distinguish between costs that fall inside and outside the healthcare system, respectively.) When we consider chronic obstructive -airways disease or LRTI, healthcare costs encompass the costs of drug treatment, visits to the physician’s office, outpatient hospital visits, and hospital admissions. Nonhealthcare costs might include any copayments that the patient must make for prescrip-
Volume 99 (suppl 66)
SYMPOSIUM ON RESPIRATORYTRACT INFECTIONS/ DAVEY
TABLEI Rankingof Outcomesin Lower-RespiratoryTract Infection’ se;za; to
Interest to Purchaser
Cost Rank
US
UK
us
UK
US
UK
AntIbiotIccost Officewsds
1 2
I
4
4
4
1
2
3
3
3
3
Hospital admwons
3
3
2
4
4
1 2
1
Time off work
1 2
2
4
‘Ranked accordtng to sensMy
to disease, cost, and interest to purchaser, r the Unded States and the Unried Kmgdom.
tions or other charges, transportation costs, and such social costs as loss of earnings or productivity. Intangible costs refer to patient suffering, anxiety, and loss of confidence in the event of treatment failure. It should also be noted that, for example, the decision to switch the patient from parenteral to oral antibiotic therapy may cause anxiety for nurses as well as patients. When contemplating a change in healthcare practice, it is important to look beyond the financial ramifications and also consider the perspectives of the various persons involved. This means satisfying not only policy makers and administrators, but also fellow practitioners, patients, and, increasingly, public opinion as represented by the mass media. It should be kept in mind that an attempt to reduce healthcare costs may arouse public anger if it is perceived as not being in the best interests of patients. Moreover, a change that seems advantageous from the hospital’s point of view may seem less attractive from the perspective of the community practitioner, and these different standpoints must be taken into account.
OUTCOMES Until just a few years ago, information about the risks associated with unsuccessful treatment of community-acquired LRTI was largely unavailable.’ In a recent study, Macfarlane and coworkers reported the outcome in 206 patients who presented to a suburban general practice with LRTI over the course of 14 months.’ These investigators found that community-acquired LRTI was associated with considerable morbidity, although with no motirtlity. In this population, 36% of patients with LRTI spent 2-4 days at home in bed, 25% of those who were employed lost up to 4 days from work, and four patients were admitted to the hospital. Overall, it took a median of 7 days before patients were able to resume their normal activities, with more than a third of patients taking more than a week to return to normal activities. These observaDecember
tions highlight the substantial benefits that could be gained by more efficacious treatment of community-acquired LRTI. To estimate the advantages potentially afforded by improvements in treatment, it is useful to consider the sensitivity of the various outcomes, such as antibiotic costs, physician visits, hospital admissions, and time lost from work, to different management strategies (Table I). Since the majority of patients do consult their physicians and are treated with antibiotics, antibiotic costs and office visits probably head the list of outcomes most sensitive to alternative costs. Hospital admission ranks third and time lost from work comes at the bottom of the list, since these outcomes are applicable to only a minority of patients with community-acquired LRTI. In contrast, when these outcomes are weighed in terms of cost, the rank order is inverted. Hospital admission is undoubtedly the most expensive outcome, followed by time lost from work. Using cost as a criterion, office visits are only the third most important item and antibiotics trail in last place. Interestingly, the emphasis placed by decision makers on each of these outcome measures may vary from country to country. In the United States, the highest priority would be to avoid the expense of hospital admission. Somewhat illogically, however, in the linited Kingdom drug cost containment is used as one of the leading indicators of health service management performance, even though drugs represent only a fraction of the healthcare budget. In examining outcomes in community-acquired LRTI, it is important to look at the frequency of repeat office visits and the reasons that additional visits are required (Table II). In Macfarlane and colleagues’” study, 26% of patients requested another visit to the general practitioner, mostly for cont.inuing symptoms of infection requiring further antibiotic prescriptions. Specifically, 11% reported worsening chest problems, lY,% experienced no improvement, 34% had only7 limited improvement,
29, 1995
The Amencan Journal of Medune
Volume 99 (suppl
68)
66-21s
SYM~SIUM ON RESPIRATORYTRACT INFECTIONS / DAVEY
parenteral to oral therapy could reduce the length of hospital stay, valuable resources would be saved.
TABLE II Reasons for Repeat Office Visits *
MacFarlane etal. 53/206(26%l
Numbel
Davey etal. 16/62 (26%1 96%
Worse/no improvement
64%
Ahbiotic side effects
10%
4%
Testresults
13%
0
rimong patients wrth lower resprratory tract Infections in two general practces. dapted from MacFarlane et al. I21 and Davey et al. 131, with permrssron.
10% developed side effects from antibiotic treatment, and 13% returned for their test results. Thus, inadequate or inappropriate treatment was apparently responsible for return visits in nearly two thirds of patients, Among the patients who came for repeat visits, two thirds had previously been in good health. We likewise documented exactly the same proportion of repeat visits for LRTI in our 6-month retrospective review of all consultations made in a general practice in Scotland.” We noted that the persistence or recurrence of symptoms was the most frequent reason for additional consultations, accounting for 96% of return visits. In contrast to the Macfarlane study, in our study only 4% of repeat visits were attributable to antibiotic side effects and no extra visits were necessary for retrieval of test results. Thus, the challenge for researchers is to identify the consequences of treatment failure and to determine the value of the resources that are saved by treatment success. We can use the data from the Macfarlane study and our own study as the basis for a simple cost calculation. For example, if we assume that half of the 26% of repeat consultations in both studies were, in fact, due to failure of antibiotic therapy, then we would have an antibiotic failure rate of 13% in community-acquired LRTI. If an alternative treatment reduced the failure rate to 5%, the difference in efficacy between the two treatments would be 8%. If the extra cost of the alternative treatment were $4.00 per patient treated, then the cost of preventing one treatment failure would be $4.00 divided by 0.08, or $50. Thus, if the costs saved by preventing repeat consultations, secondline prescriptions, and time lost from work exceeded $50, then the alternative treatment could be considered cost-effective. Although the cost benefit of preventing repeat office visits may vary depending on the location of the practice and although it is difficult to quantify nonhealthcare costs, such as time lost from work, it is clear that cost savings might be achieved by the use of more effective treatments.* Similarly, if switching a patient from
65228
INCORPORATINGECONOMICANALYSISINTO CLINICAL TRIALS In most developed countries, amassing data about such outcome measures as office visits, drug costs, and hospital admissions generally presents no major problem. On the other hand, collection of data about time lost from work may be more complicated and may discourage physicians from participating in clinical trials. A booklet published by the U.K. Department of Health offers helpful guidelines about incorporating economic analysis into randomized trials.” This work points out that some types of data, such as those pertaining to physical quantities of resource (e.g., number of treatment days and number of drugs prescribed) and descriptions of patients’ quality of life may be collected largely within the trial. However, it may be advisable to look outside the clinical trial center to collect other sorts of data, such as unit costs, costs of resources, and health state preference values, which indicate how strongly people feel about being in a particular state of health. 4 useful schematic view of these recommendations is presented in Figure 1. When collecting economic data in the context of a clinical trial, it is important to give particular emphasis to those data that are of relevance to the people who are responsible for making the decisions that the trial is designed to influence, whether they be physicians or administrators. In addition, the trial data should be arranged according to a hierarchy of costs. For example, when length of hospital stay is the major consideration, the number of syringes used becomes insignificant. Clinical trial designers should also be realistic in assessing whether the collection of economic data is feasible within the trial protocol or whether it might render the trial virtually impossible to conduct. Finally, it is necessary to consider the impact of cost variation on statistical power. A brief review of the basic concepts of validity may be helpful. Statistical validity means that the trial is able to detect covariants; for example, if there is a link between antibiotic treatment and duration of hospital stay, the trial will identify it. Internal validity means that covariance is due to cause and effect or, in other words, that the antibiotic treatment was responsible for the difference in length of stay. External validity means that the results of the trial can be generalized to other centers, which underscores the importance of collecting data outside as well as within the trial.
December 29, 1995 The Amencan Journal of Medicine Volume 99
(suppl
6~)
SYMPOSIUM ONRESPIRATORY TRACT INFECTIONS / DAVEY
Physical quantities of resources
I
Descriptive quality-of-life data
I
Boxes mfssfon
rndrcate
economfc dfstribubon,
from Drummond,
data need rather
1
to be gathered
than actual
I
I
Health state preference (utilities) 1. Which
I
I
Unit costs or prices of resources
Figure
pens to patients who do not improve with treatment. In addition, large databases offer a wealth of information that can be used to construct decision trees and identify potential targets for clinical trials. Moreover, these databases have potential applicability to randomized trials. In the United States, health maintenance organizations are now being used as sources of data, with randomization being performed within the database.
Major Data Collection Effort
Data element
I
wrthrn controlled
percentages.
(Adapted
trfals? with per-
et al.‘)
Diagnostic accuracy Figure
2. Importance
of accurate
diagnoses
RELEVANCEOF DIAGNOSISTO ECONOMIC ANALYSIS Economic analysis sometimes overemphasizes the importance of choosing the most cost-effective drugs, while downplaying the significance of selecting which patients warrant treatment. To take a theoretical example: drug A costs $30 more per patient than drug BY but drug A is 80% effective while drug R has a success rate of only 50%. If treatment failure results in an additional $150 in healthcare costs, which treatment is more expensive? The answer will hinge on the accuracy of the diagnosis. If only 10% of patients receiving treatment with either drug actually have the disease, then the difference in efficacy can benefit only 10% of the patients treat,ed. As shown in Figure 2, in order for the use of the more effective treatment, to reduce healthcare costs, it is necessary for the diagnosis t,o be correct in about one half of cases.
In cost-effechveness
CONCLUSIONS The best way of achieving internal validity is randomization. The advantages of randomization are that it protects against selection bias and provides a sound mathematical basis for statistical analysis of cause and effect. In addition, even when bias occurs by chance, randomization facilitates correction by analysis of covariance. Randomization is also the only way of ensuring that the treatment allocation and trial results are independent of unknown or unmeasured confounding factors that may affect outcome.” In contrast, we must be more wary of inferring cause and effect relations from observational studies, which are associated with a higher risk of being affected by selection bias, bias in the information collected from study participants, chance imbalance, and unrecognized confounding i-actors.
Large databases can be extremely useful because they help us to measure resourc’e costs, especially the costs of treatment failure. They also aid in identifying the treatment pathways that people follow in real life, providing information about what hap-
December
When a new treatment is more expensive than the available alternatives, it must be more effective if it is to be cost-effective. Cost-effectiveness, then, depends on the value attached to preventing treatment failure, which always includes a subjective element. Finally, both effectiveness and cost-effectiveness depend on the appropriateness with which t,reatment is applied, that is to say, on the accuracy of diagnosis.
REFERENCES 1. Davey P, McDonald
T. Postmarketing surverllance of qufnolones, 1990 to 1992. Drugs 1993; 45 tsuppl 3). 46-53. 2. Macfarlane JT. Colv~lle A, Gufon A, Macfarlane RM. Rose DH. Prospective study of aetfology and outcome of adult lower-resprratory-tract rnfections fn the community, Lancet 1993; 341: 511-4. 3. Davey P, Rutherford 0, Graham B, Lynch B. Malek M. Repeat consultatfons after antfbfatfc prescrrbing for respfratory Infectron: a study In one general practice. Br J Gen Praci 1994: 44: 509-13. 4. MacDonald TM, Collrns D, McGrlchnst MM, Stevens J, McKendrtck AD, McDevftt DG, Davey PG. Antfmrcrobial practrce. The uttlrsation and economic evaluatfon of antrbrotfcs prescrrbed In prfmary care. J Antimrcrob Chemother 1995; 35: 191-204. 5. Drummohd M, Smith GT, Wells N. Economrc evaluahon rn the development of medrctnes. London: Office of Health Economrcs, 1988 (ISBN 04738837). 6. Brennan P. Croft P. Interpreting the results of observatfonal research: chance is not such a ffne thing. Br Med J 1994; 309: 727-30.
29, 1995
The American Journal of Medicine
Volume 99 (suppl
6Bl
66-238
Pharmacoeconomic analysis must determine a credible cost model, including healthcarerelated, non-healthcare-related, and intangible costs, and must identify important treatment outcomes. Outcomes can be ranked in terms of their sensitivity to different treatments or weighted in terms of cost. Two recent studies of community-acquired lower respiratory tract infection have shown that inadequate treatment is a frequent cause of repeat visits to the physician’s office, involving extra costs that could presumably be avoided by more effective first-line treatment. When economic data are to be incorporated into clinical trials, it is important to collect data outside as well as within the trial center and to concentrate on data likely to be of significance to the decision makers the trial is designed to influence. The best way of achieving internal validity in such trials is through randomization, which minimizes bias, chance imbalance, and confounding factors. The three major determinants of the costeffectiveness of treatment are efficacy, the value attached to preventing treatment failure, and the accuracy of diagnosis. The latter is important because a drug can only benefit patients if they have the disease for which it is an effective treatment.
From the Pharmacoeconomlcs Research Centre, Department of Clinical Pharmacology, Unlversrty of Dundee, United Kingdom. Requests for reprlnts should be addressed to Peter G. Davey, M.D., Department of Clinical Pharmacology, Ninewells Hospital and Medical School, Dundee, Taysrde. DDl 9SY, United Krngdom.
6B-20s
December 29, 1995
The American Journal of Medicine
ower respiratory tract infections (LRTI) rank high on the list of reasons patients visit their family physicians; they are also among the most frequent indications for which antibiotic treatment is prescribed. Not surprisingly, the cost-effectiveness of antibiotic therapy for community-acquired (LRTI) has become the focus of intense scrutiny. In performing a comprehensive and balanced economic analysis of the use of antibiotics for the primary care treatment of LRTI, it is important to address four key questions: 1. What is a credible cost model? 2. What are the important outcomes for antibiotic treatment? 3. How should economic analysis be incorporated into randomized clinical trials? 4. What is the relevance of diagnostic techniques to the economic analysis of antibiotic treatment? This article will discuss the significance of these issues and explore some of the ways in which they can be integrated into analyses of cost management in community-acquired LRTI. Identification of the most important considerations for pharmacoeconomic analysis can not only provide answers for decisions we have to make today but can also help pinpoint questions for future research.
L
COST MODELS The methods used in a credible cost model must be statistically valid, and the various cost estimates must be reasonably accurate. The first steps in developing such a model are to identify the potential costs of failed treatment and to separate the components of these costs into healthcare financial costs, nonhealthcare financial costs, and intangible costs, which include those costs that are difficult to value financially. (In this review the terms healthcare and nonhealthcare will be used in preference to the more ambiguous adjectives direct and indirect to distinguish between costs that fall inside and outside the healthcare system, respectively.) When we consider chronic obstructive -airways disease or LRTI, healthcare costs encompass the costs of drug treatment, visits to the physician’s office, outpatient hospital visits, and hospital admissions. Nonhealthcare costs might include any copayments that the patient must make for prescrip-
Volume 99 (suppl 66)
SYMPOSIUM ON RESPIRATORYTRACT INFECTIONS/ DAVEY
TABLEI Rankingof Outcomesin Lower-RespiratoryTract Infection’ se;za; to
Interest to Purchaser
Cost Rank
US
UK
us
UK
US
UK
AntIbiotIccost Officewsds
1 2
I
4
4
4
1
2
3
3
3
3
Hospital admwons
3
3
2
4
4
1 2
1
Time off work
1 2
2
4
‘Ranked accordtng to sensMy
to disease, cost, and interest to purchaser, r the Unded States and the Unried Kmgdom.
tions or other charges, transportation costs, and such social costs as loss of earnings or productivity. Intangible costs refer to patient suffering, anxiety, and loss of confidence in the event of treatment failure. It should also be noted that, for example, the decision to switch the patient from parenteral to oral antibiotic therapy may cause anxiety for nurses as well as patients. When contemplating a change in healthcare practice, it is important to look beyond the financial ramifications and also consider the perspectives of the various persons involved. This means satisfying not only policy makers and administrators, but also fellow practitioners, patients, and, increasingly, public opinion as represented by the mass media. It should be kept in mind that an attempt to reduce healthcare costs may arouse public anger if it is perceived as not being in the best interests of patients. Moreover, a change that seems advantageous from the hospital’s point of view may seem less attractive from the perspective of the community practitioner, and these different standpoints must be taken into account.
OUTCOMES Until just a few years ago, information about the risks associated with unsuccessful treatment of community-acquired LRTI was largely unavailable.’ In a recent study, Macfarlane and coworkers reported the outcome in 206 patients who presented to a suburban general practice with LRTI over the course of 14 months.’ These investigators found that community-acquired LRTI was associated with considerable morbidity, although with no motirtlity. In this population, 36% of patients with LRTI spent 2-4 days at home in bed, 25% of those who were employed lost up to 4 days from work, and four patients were admitted to the hospital. Overall, it took a median of 7 days before patients were able to resume their normal activities, with more than a third of patients taking more than a week to return to normal activities. These observaDecember
tions highlight the substantial benefits that could be gained by more efficacious treatment of community-acquired LRTI. To estimate the advantages potentially afforded by improvements in treatment, it is useful to consider the sensitivity of the various outcomes, such as antibiotic costs, physician visits, hospital admissions, and time lost from work, to different management strategies (Table I). Since the majority of patients do consult their physicians and are treated with antibiotics, antibiotic costs and office visits probably head the list of outcomes most sensitive to alternative costs. Hospital admission ranks third and time lost from work comes at the bottom of the list, since these outcomes are applicable to only a minority of patients with community-acquired LRTI. In contrast, when these outcomes are weighed in terms of cost, the rank order is inverted. Hospital admission is undoubtedly the most expensive outcome, followed by time lost from work. Using cost as a criterion, office visits are only the third most important item and antibiotics trail in last place. Interestingly, the emphasis placed by decision makers on each of these outcome measures may vary from country to country. In the United States, the highest priority would be to avoid the expense of hospital admission. Somewhat illogically, however, in the linited Kingdom drug cost containment is used as one of the leading indicators of health service management performance, even though drugs represent only a fraction of the healthcare budget. In examining outcomes in community-acquired LRTI, it is important to look at the frequency of repeat office visits and the reasons that additional visits are required (Table II). In Macfarlane and colleagues’” study, 26% of patients requested another visit to the general practitioner, mostly for cont.inuing symptoms of infection requiring further antibiotic prescriptions. Specifically, 11% reported worsening chest problems, lY,% experienced no improvement, 34% had only7 limited improvement,
29, 1995
The Amencan Journal of Medune
Volume 99 (suppl
68)
66-21s
SYM~SIUM ON RESPIRATORYTRACT INFECTIONS / DAVEY
parenteral to oral therapy could reduce the length of hospital stay, valuable resources would be saved.
TABLE II Reasons for Repeat Office Visits *
MacFarlane etal. 53/206(26%l
Numbel
Davey etal. 16/62 (26%1 96%
Worse/no improvement
64%
Ahbiotic side effects
10%
4%
Testresults
13%
0
rimong patients wrth lower resprratory tract Infections in two general practces. dapted from MacFarlane et al. I21 and Davey et al. 131, with permrssron.
10% developed side effects from antibiotic treatment, and 13% returned for their test results. Thus, inadequate or inappropriate treatment was apparently responsible for return visits in nearly two thirds of patients, Among the patients who came for repeat visits, two thirds had previously been in good health. We likewise documented exactly the same proportion of repeat visits for LRTI in our 6-month retrospective review of all consultations made in a general practice in Scotland.” We noted that the persistence or recurrence of symptoms was the most frequent reason for additional consultations, accounting for 96% of return visits. In contrast to the Macfarlane study, in our study only 4% of repeat visits were attributable to antibiotic side effects and no extra visits were necessary for retrieval of test results. Thus, the challenge for researchers is to identify the consequences of treatment failure and to determine the value of the resources that are saved by treatment success. We can use the data from the Macfarlane study and our own study as the basis for a simple cost calculation. For example, if we assume that half of the 26% of repeat consultations in both studies were, in fact, due to failure of antibiotic therapy, then we would have an antibiotic failure rate of 13% in community-acquired LRTI. If an alternative treatment reduced the failure rate to 5%, the difference in efficacy between the two treatments would be 8%. If the extra cost of the alternative treatment were $4.00 per patient treated, then the cost of preventing one treatment failure would be $4.00 divided by 0.08, or $50. Thus, if the costs saved by preventing repeat consultations, secondline prescriptions, and time lost from work exceeded $50, then the alternative treatment could be considered cost-effective. Although the cost benefit of preventing repeat office visits may vary depending on the location of the practice and although it is difficult to quantify nonhealthcare costs, such as time lost from work, it is clear that cost savings might be achieved by the use of more effective treatments.* Similarly, if switching a patient from
65228
INCORPORATINGECONOMICANALYSISINTO CLINICAL TRIALS In most developed countries, amassing data about such outcome measures as office visits, drug costs, and hospital admissions generally presents no major problem. On the other hand, collection of data about time lost from work may be more complicated and may discourage physicians from participating in clinical trials. A booklet published by the U.K. Department of Health offers helpful guidelines about incorporating economic analysis into randomized trials.” This work points out that some types of data, such as those pertaining to physical quantities of resource (e.g., number of treatment days and number of drugs prescribed) and descriptions of patients’ quality of life may be collected largely within the trial. However, it may be advisable to look outside the clinical trial center to collect other sorts of data, such as unit costs, costs of resources, and health state preference values, which indicate how strongly people feel about being in a particular state of health. 4 useful schematic view of these recommendations is presented in Figure 1. When collecting economic data in the context of a clinical trial, it is important to give particular emphasis to those data that are of relevance to the people who are responsible for making the decisions that the trial is designed to influence, whether they be physicians or administrators. In addition, the trial data should be arranged according to a hierarchy of costs. For example, when length of hospital stay is the major consideration, the number of syringes used becomes insignificant. Clinical trial designers should also be realistic in assessing whether the collection of economic data is feasible within the trial protocol or whether it might render the trial virtually impossible to conduct. Finally, it is necessary to consider the impact of cost variation on statistical power. A brief review of the basic concepts of validity may be helpful. Statistical validity means that the trial is able to detect covariants; for example, if there is a link between antibiotic treatment and duration of hospital stay, the trial will identify it. Internal validity means that covariance is due to cause and effect or, in other words, that the antibiotic treatment was responsible for the difference in length of stay. External validity means that the results of the trial can be generalized to other centers, which underscores the importance of collecting data outside as well as within the trial.
December 29, 1995 The Amencan Journal of Medicine Volume 99
(suppl
6~)
SYMPOSIUM ONRESPIRATORY TRACT INFECTIONS / DAVEY
Physical quantities of resources
I
Descriptive quality-of-life data
I
Boxes mfssfon
rndrcate
economfc dfstribubon,
from Drummond,
data need rather
1
to be gathered
than actual
I
I
Health state preference (utilities) 1. Which
I
I
Unit costs or prices of resources
Figure
pens to patients who do not improve with treatment. In addition, large databases offer a wealth of information that can be used to construct decision trees and identify potential targets for clinical trials. Moreover, these databases have potential applicability to randomized trials. In the United States, health maintenance organizations are now being used as sources of data, with randomization being performed within the database.
Major Data Collection Effort
Data element
I
wrthrn controlled
percentages.
(Adapted
trfals? with per-
et al.‘)
Diagnostic accuracy Figure
2. Importance
of accurate
diagnoses
RELEVANCEOF DIAGNOSISTO ECONOMIC ANALYSIS Economic analysis sometimes overemphasizes the importance of choosing the most cost-effective drugs, while downplaying the significance of selecting which patients warrant treatment. To take a theoretical example: drug A costs $30 more per patient than drug BY but drug A is 80% effective while drug R has a success rate of only 50%. If treatment failure results in an additional $150 in healthcare costs, which treatment is more expensive? The answer will hinge on the accuracy of the diagnosis. If only 10% of patients receiving treatment with either drug actually have the disease, then the difference in efficacy can benefit only 10% of the patients treat,ed. As shown in Figure 2, in order for the use of the more effective treatment, to reduce healthcare costs, it is necessary for the diagnosis t,o be correct in about one half of cases.
In cost-effechveness
CONCLUSIONS The best way of achieving internal validity is randomization. The advantages of randomization are that it protects against selection bias and provides a sound mathematical basis for statistical analysis of cause and effect. In addition, even when bias occurs by chance, randomization facilitates correction by analysis of covariance. Randomization is also the only way of ensuring that the treatment allocation and trial results are independent of unknown or unmeasured confounding factors that may affect outcome.” In contrast, we must be more wary of inferring cause and effect relations from observational studies, which are associated with a higher risk of being affected by selection bias, bias in the information collected from study participants, chance imbalance, and unrecognized confounding i-actors.
Large databases can be extremely useful because they help us to measure resourc’e costs, especially the costs of treatment failure. They also aid in identifying the treatment pathways that people follow in real life, providing information about what hap-
December
When a new treatment is more expensive than the available alternatives, it must be more effective if it is to be cost-effective. Cost-effectiveness, then, depends on the value attached to preventing treatment failure, which always includes a subjective element. Finally, both effectiveness and cost-effectiveness depend on the appropriateness with which t,reatment is applied, that is to say, on the accuracy of diagnosis.
REFERENCES 1. Davey P, McDonald
T. Postmarketing surverllance of qufnolones, 1990 to 1992. Drugs 1993; 45 tsuppl 3). 46-53. 2. Macfarlane JT. Colv~lle A, Gufon A, Macfarlane RM. Rose DH. Prospective study of aetfology and outcome of adult lower-resprratory-tract rnfections fn the community, Lancet 1993; 341: 511-4. 3. Davey P, Rutherford 0, Graham B, Lynch B. Malek M. Repeat consultatfons after antfbfatfc prescrrbing for respfratory Infectron: a study In one general practice. Br J Gen Praci 1994: 44: 509-13. 4. MacDonald TM, Collrns D, McGrlchnst MM, Stevens J, McKendrtck AD, McDevftt DG, Davey PG. Antfmrcrobial practrce. The uttlrsation and economic evaluatfon of antrbrotfcs prescrrbed In prfmary care. J Antimrcrob Chemother 1995; 35: 191-204. 5. Drummohd M, Smith GT, Wells N. Economrc evaluahon rn the development of medrctnes. London: Office of Health Economrcs, 1988 (ISBN 04738837). 6. Brennan P. Croft P. Interpreting the results of observatfonal research: chance is not such a ffne thing. Br Med J 1994; 309: 727-30.
29, 1995
The American Journal of Medicine
Volume 99 (suppl
6Bl
66-238