As specified in most guidelines for pharmaco-economic analysis, units of service and the cost per unit of service should be collected separately for completeness and transparency [1,2]. Bundling of units of service and cost per unit makes it difficult to understand the underlying process by which costs are being accumulated and particularly difficult to estimate the marginal cost of care. In this editorial, I consider the costing of physician services for pharmaco-economic analyses. Costs of physician services are examined through the use of data from an international trial of medications for a chronic condition. Information on the use of physician services was collected within the trial and augmented by highly detailed information on services provided for selected, common diagnoses requiring treatment. One important observation from the trial was that the use and cost of physician services costs were related to the underlying payment system.
Collecting units of service, whether it be alongside trials or from administrative or clinical databases, requires specification of the details to be collected. In the case of hospital services, which are clearly the highest cost source of services, there have been at least a couple of analyses providing guidance on the details to be collected in international trials. Glick and colleagues found that mean length of stay from the Medicare diagnosis-related group payment system was a good predictor of unit costs in four countries in Europe [3]. They found that only 3% of variations in hospital costs was explained by the country of -hospitalization. In contrast, Schulman and colleagues found that 11% of variation was explained by the country of hospitalization, which still seems like a low source of variation [4].
While these studies do not suggest that diagnosis and length of stay are the only data elements that need to be collected, these two data elements appear to capture much of the available information. Could the same hold true for physician services? If one captured the numbers of visits and clinical reasons for these visits, would one have captured the majority of information -necessary for costing physician services?
Collecting use of service data
For a controlled clinical trial involving use of medications for a chronic condition, 20 European centers enrolled patients (seven in Belgium, six in France and seven in Germany). The method for determining costs was the collection of units of services provided and collection of standard costs per unit for these services. Counts of units of service were collected for physician visits, other provider visits, laboratory and diagnostic tests, medications and any facility services (including hospital days). Counts of services provided at the study sites were recorded by the investigators or their staff. Counts of services provided at other sites were solicited from patients at each study visit.
Most physician visits and related services were associated with protocol-related events, either prespecified visits (every 6–12 weeks) or medication-dosing visits that varied depending upon patients’ responses to study medications. Any use of services apart from the study was considered an adverse event. Investigators indicated whether an adverse event was related to treatment. Only the medical services stemming from attributable adverse events were analyzed. Adverse events indicated as ‘definitely’, ‘possibly’, ‘probably’ and ‘uncertain’ were considered to be attributable to treatment. Adverse events indicated as ‘unlikely’ and ‘definitely not’ were considered not to be attributable to treatment. During the study, 280 occurrences of nonprotocol-related medical services were reported. Investigators indicated that 32 of these occurrences were related to the use of study medication. Related medical services consisted solely of physician services and related diagnostic procedures and diagnostic laboratory tests. Separate from the trial, we developed a questionnaire on the use of services for the five adverse events associated with the study condition: abdominal pain, constipation, myalgia, nausea and nervousness. The questionnaire asked physicians at the study centers for estimates of use of services based on their personal experiences concerning diagnosis and treatment of the five adverse events. The questions related to abdominal pain are presented below. Questions related to the other -conditions were quite similar.
Abdominal pain was a complaint offered by some study participants. Physicians were first asked about the frequency of consultations per event: how often do patients with abdominal pain associated with the study medication visit your office? How often do these patients consult you exclusively by phone or additionally by phone? Physicians were also asked which diagnostic procedures they perform and for what percentage of patients; physical examination, abdominal ultrasound, gastro-scopy, colonoscopy, blood counts, chemical laboratory tests and other tests.
Collecting unit cost data
Costs per unit of service (prices) were collected from payment databases for each country included in the study and from individual sites. Payments by service can vary by payer within a country with multiple payers, as well as among countries. Therefore, presenting the resource use and costs separately could enable analysts to alter payment schedules for their particular case. For each country there was more than one reimbursement schedule that was applicable to this study. However, variation was low within each country, particularly in terms of which services were associated with additional payments for physicians. Physicians in Belgium reported getting paid for the visit, almost all inclusive. Physicians in Germany and France reported billing for more services. This analysis includes only the average level of payment within each country for demonstration purposes.
Observed physician services costs
Multiplying use of services by unit costs yields total physician services costs. For the five common types of adverse events in this trial, total physician costs are presented in Table 1. Physician reports of Belgian costs were universally the lowest, followed by German costs and then French costs. Both French and German costs showed far more variation among conditions than Belgian costs. For abdominal pain and constipation, Belgian costs were a fraction of the German costs, while French costs were half higher again than German costs. For myalgia, the average German costs were three-times Belgian costs and French costs were three-times German costs. German and Belgian costs were approximately equal for diagnoses of nausea and nervousness, while French costs were twice as high for -nervousness and three-times as high for nausea.
Some country-specific patterns were also observed. Belgian costs consisted largely of a single charge for an office consultation. There were no additional charges for telephone consultation or physical examination. In addition, the Belgian physicians did not use or did not charge separately for abdominal ultrasound, gastroscopy, coloscopy, electromyogram or x-rays. The difference between the office consultation fee and total fee in Belgium consisted of blood counts, chemical tests, other tests and therapy. These additional services were not reported -uniformly by physicians.
German costs included a charge for office visit, telephone consultation, physical examination and blood counts for all physicians. These costs totaled very close to the Belgian values. However, the German physicians also reported that they would perform abdominal ultrasound, gastroscopy and colonoscopy on both abdominal pain patients and constipation patients, and that they would obtain electromyogram and x-rays on myalgia patients. German physicians did not report costs for other tests and only three of the German physicians would perform clinical lab tests. The higher costs of German care are associated with a higher intensity of high technology diagnostic services.
French costs were broken down into an office visit, which included telephone consultation and physical examination. With the addition of a blood count reported by physicians for all adverse events, the cost of the universal French visit was higher than either the German or Belgian visits. The French were also similar to the Germans in performing abdominal ultrasound, gastroscopy and coloscopy for all adverse events. In addition, costs included laboratory tests reported by all physicians and other tests reported by most physicians. It appears that French costs included both the high technology diagnostic costs seen in the German costs and more intensive, more expensive blood and chemical testing.
Table 1. Physician costs by service by country.
Suggestions for costing physician services
Several suggestions for costing physician services stem from this brief review of one trial and numerous other analyses not reported here. First, there are numerous reasons for visits to physicians, only a subset of which may be related to a particular regimen of treatment. In some trials, particularly where death is an outcome, so-called all-cause mortality may be an important measure. All-cause use of medical services is typically not the aim of pharmacoeconomic analyses. With only a tenth of physician visits being associated with potential adverse events, careful consideration should be given as to which visits to include in an analysis.
Second, the variation of physician costs was substantially greater between countries than within countries. The variation in physician costs was largely related to treatment intensity, which appears to have been related to the underlying payment system. Systems that use more comprehensive fees were associated with fewer additional tests and costs. Reports of physician pressures based on payment systems were common for all countries [5–7]. To get more information on physician costs it may be helpful to obtain estimates from more countries, rather than from more physicians within a country.
Third, the variation of physician costs was substantial between diagnoses, even within countries. In this sense, the results of studies of international comparisons of hospital costs are not fully supported by this view of physician costs. Given the variation in costs for each diagnosis, it would not be anticipated that total spending for each diagnosis, or for the underlying chronic condition being treated, would be similar among countries, as has been reported for some conditions [8]. It is not clear that it would be better to obtain estimates for fewer diagnoses from more countries than the reverse [3]. A more definitive statement in this regard would require an analysis of physician services costs that considers substantially more than five diagnoses and 20 centers in three countries.
Fourth, the observed variation in costs was directly related only to resource use. In addition to the observed variation, there could also be variation related to coding patterns of services. In the USA, we observe variation in costs associated with assignment of evaluation and management common procedural terminology codes [9]. For many other countries, office visits of all durations and complexity have the same payment – a visit is a visit. To the extent that other payment systems permit differences in costs through coding of services, there may be variation in resource use that would not be captured if the detail that underlies coding is not also captured.
Fifth, the difference between costs and prices, and therefore one aspect of the perspective of the trial presentation, is not as meaningful for physician services as it is for hospital services. In the case of physician services, costs to payers are the income to the physician. There is not a way to think about the underlying costs, average or marginal, in a way other than the prices paid to physicians.
Physician services costs, while only a fraction of hospital costs for serious adverse events, are an important component of costs for chronic illnesses. Accurately capturing physician services costs in pharmacoeconomic analyses requires careful planning. Analysts should understand the clinical condition being considered and the underlying payment system to anticipate the sources of variation in resources to be used. Even if information on the numbers of visits and clinical reasons for these visits is collected, the majority of information necessary for costing physician services still may not have been captured.
Acknowledgement
This work was funded by Pfizer, Inc. (NY USA). James Albright (Pfizer, Inc.) and Karin Berger (Medical Economics Research Group, Munich, Germany) were instrumental in the development of the methods upon which this work was based.
Table 1. Physician costs by service by country.
References
- Gold MR, Siegel JE, Russel LB, Weinstein MC. Cost–Effectiveness in Health and Medicine. Oxford University Press, NY, USA (1996).
- Drummond MF, Jefferson TO, on behalf of the BMJ Economic Evaluation Working Party. Guidelines for authors and peer reviewers of economic submissions to the BMJ. Br. Med. J. 313, 2758–2753 (1996).
- Glick HA, Orzol SM, Tooley JF, Polsky D, Mauskopf JO. Design and analysis of unit cost estimation studies: how many hospital diagnoses? How many countries? Health Econ. 12(7), 5172–5177 (2003).
- Schulman K, Burke J, Drummond M et al. Resource costing for multinational neurological clinical trials: methodology and results. Health Econ. 7(8), 6293–6298 (1998).
- Huttin C. Andral J. How the reimbursement system may influence physicians' decisions results from focus groups interviews in France. Health Pol. 54(2), 678–676 (2000).
- Kinder K. Assessing the impact of payment method and practice setting on German physicians' practice patterns. J. Am. Care Manag. 24(2), 118 (2001).
- Morrison I. The future of physicians’ time: past, present and future. Ann. Int. Med. 132(1), 808–804 (2000).
- Evers SM. Struijs JN. Ament AJ et al. International comparison of stroke cost studies. Stroke 35(5), 12091–12095 (2004).
- Griffin TC, Hutter JJ, Johnson KK et al. A survey of clinical productivity and current procedural terminology (CPT) coding patterns of pediatric hematologist/oncologists. Pediatric Blood Cancer 43(2), 1407 (2004).