Hospital pharmacy decisions, cost containment, and the use of cost-effectiveness analysis

Pergamon

Sot'. Sci. Med. Vol. 45, No. 4, pp. 523 533, 1997

PII: S0277-9536(96)00393-0

(; 1997 ElsevierScience Ltd All rights reserved. Printed in Great Britain 0277-9536/97 $17.00 + 0.00

HOSPITAL PHARMACY DECISIONS, COST CONTAINMENT, A N D THE USE OF COSTEFFECTIVENESS ANALYSIS F R A N K A. S L O A N , '2 K A T H R Y N W H E T T E N - G O L D S T E I N ~ and A L I C I A W I L S O N ' ~Center for Health Policy Research and Education, Box 90253, Duke University, Durham, NC 27708, U.S.A. and ~'Department of Economics, Duke University, Durham, NC 27708, U.S.A. Abstract--The key hypothesis of the study was that hospital pharmacies under the pressure of managed care would be more likely to adopt process innovations to assure less costly and more cost-effective provision of care. We conducted a survey of 103 hospitals and analyzed secondary data on cost and staffing. Compared to the size of the reduction in length of stay, changes in the way that a day of care is delivered appear to be minor, even in areas with substantial managed care share. The vast majority of hospitals surveyed had implemented some form of therapeutic interchange and generic substitution. Most hospitals used some drug utilization guidelines, but as of mid 1995 these were not yet important management tools for hospital pharmacies. To our knowledge, ours was the first survey to investigate the link between hospital formularies and use of cost-effectiveness analysis. At most cost-effectiveness was a minor tool in pharmaceutical decision making in hospitals at present. We could determine no differences in use of such analyses by managed care market share in the hospital's market share. One impediment to the use of cost-effectiveness studies was the lack of timeliness of studies, Other stated reasons for not using cost-effectiveness analysis more often were: lack of information on hospitalized patients and hence on the potential cost offsets accruing to the hospital; lack of independent sponsorship, and inadequate expertise in economic evaluation. ~~ 1997 Elsevier Science Ltd Key words--cost-effectiveness, hospital pharmacies, cost containment

Because of the growth of managed care and cutbacks in public programs, hospitals in the U.S. are under increased external pressure to become more efficient. Reflecting these exogenous changes, many U.S. hospitals have been downsizing. Although pharmacy constitutes only 4 - 8 % of hospital total expense, there are forces causing this share to grow, including increases in the real price of pharmaceutical products, increasing acuity of hospitalized patients, and increasing availability and use of biotechnology-derived products. Although the pharmacy is of interest in its own right, focusing on a single department also permits a more detailed look inside the "black box" of hospital responses to managed care growth. There is a burgeoning literature on effects of managed care on cost, access, and quality of care, but there is still a paucity of information on the decision making process itself. The term "managed care" is used widely, and precise definitions differ. Iglehart (Iglehart, 1992) defined managed care as: a system that integrates the financing and delivery of appropriate medical care by means of the following features: contracts with selected physicians and hospitals that furnish a comprehensive set of health care services to enrolled members, usually for a predetermined monthly premium: utilization and quality controls that contracting providers agree to accept; financial incentives for patients to use the providers and facilities associated with the plan; and the assumption of some financial risk by doctors, thus fundamentally altering their role from serving as agent for

the patient's welfare to balancing the patient's needs against the need for cost control (p. 742). Health maintenance organizations (HMOs) that place individual providers at some financial risk fit this description. Preferred provider organizations (PPOs) do not completely fit this description since providers do not assume financial risk, although they are often included as "managed." Up to now, very few hospitals in the U.S. provide care on an "at risk" basis. Rather, they sell care to managed care at a discount, such as an all-inclusive p e r diem rate. The key hypothesis of this article is that the extent of managed care penetration has systematic effects on hospital behavior. In particular, hospitals under the pressure of managed care will be more likely to adopt process innovations to assure less costly and more cost-effective provision of care. In this context, cost-effectiveness analysis may be a useful methodology for achieving these objectives. In this article, we address the following issues. First, in general, how do hospital costs and hospital input use differ by the extent of managed care penetration in the market? Having examined the general costs and input use, we focus on hospital pharmacies. We explore the extent to which hospital pharmacies are using formularies, generic substitution, therapeutic interchange, and drug utilization guidelines. The latter tools are often used to contain cost and/or to assure that cost-effective medicines 523

524

Frank A. Sloan et al.

are used. Gauging the effectiveness of formularies, generic substitution, therapeutic interchange, and drug utilization is beyond the scope of this study. Such tools are widely believed to be effective in improving care and reducing costs, and further, are subject to ongoing refinement. Second, does the extent of use of these tools vary with the managed care share in the hospital's market area? Third, how has the spread of managed care affected hospital adoption of new drugs? It has been hypothesized that growth of managed care will retard the diffusion of product innovations in the hospital sector and in health care more generally (Weisbrod, 1992; Baumgardner, 1991). Traditionally, retrospective cost or charge-based reimbursement for hospital care provided an incentive for product innovation (Baumgardner, 1991). Fourth, to what extent is cost-effectiveness analysis used in the hospital pharmaceutical decision making process and is it more likely to be used where competitive pressures from managed care are the greatest? Such analysis potentially provides both a framework and empirical evidence on cost versus benefits of specific pharmaceutical products. When such analysis was not being used, our study also inquired into the reasons for nonuse and specifically how cost-effectiveness analysis could be improved to be of more practical use to hospital decision makers. Up-to-date data to address these issues are unavailable from any source. Thus, we conducted a survey of 103 hospitals for purposes of this study during Summer 1995. Section 1 of this article describes the study's methodology, including our survey methods. In Section 2, we present empirical results on adoption of cost containment strategies by hospital pharmacy

departments, adoption of new drugs, and use of scientific information, including cost-effectiveness analysis in decision making. Section 3 discusses implications of our findings, suggestions for improving the usefulness of cost-effectiveness analysis to hospitals, and future research on the influence of managed care on hospital decision making.

2. M E T H O D S

OF PROCEDURE

Classifying areas according to the influence of managed care

Markets have been classified into five stages by APM, Inc. and the University Hospital Consortium according to the degree of influence of managed care (Hospitals and Health Networks, 1995; Murray, 1995). In Stage 1, "Unstructured," HMO penetration is 10% or less. Hospitals and physicians function independently (Fig. 1). Utilization of hospitals is high. In Stage 2, "Loose Framework," physicians begin to organize in groups for purposes of contracting with or forming health plans. Hospital consolidation begins and excess bed supply supports deep discounts. The HMO market share rises to 11-30%. In Stage 3, "Consolidation," hospital systems form and aggressively recruit primarycare group practices. Plans begin dropping physicians; overcapacity in the hospital sector begins to fall. Providers develop continuums of care. HMO penetration rises to 30-50%. In Stage 4, "Managed Competition," capitation plans dominate with the HMO market share exceeding 50%. Risk shifts to primary-care medical groups. Use of specialists and their fees are driven down dramatically. There is even more pressure to eliminate hospital beds. In Stage 5, "Endgame," which seems to be a strength-

Managed competition may be last stage of market evolution lit

Unstructured • Independent hospitals, MDs, employers, HMOs • 0-10% HMO

penetration

Loose Framework • Lead HMOs, PPOs, or medical groups emerge • Loose provider networks and weak hospital affiliations form • Excess inpatient capacity develops • 11-30% HMO

penetration

Consolidation ] • Lead HMOs, PPOs and medical groups achieve critical mass, begin to consolidate • Hospital systems form/aggressively recruit/compete for primary care group practices • Beds close, hospital profits increase • 31-50% HMO penetration

Fig. 1. Market evolution.

J

Managed Competition ]

• Purchasers contract with integrated medical groups and hospital/ physician systems to provide comprehensive services to beneficiaries HMO penetration

• >50%

Hospital pharmacy decisions ening and continuation of Stage 4, integrated systems manage patient populations. No market has yet reached Stage 5. For purposes of our study, we randomly selected cities in each of the first four stages and randomly selected nonfederal, short-term general hospitals within each of these cities: Stage 1--Little Rock, Arkansas, Birmingham, Alabama, Newark, New Jersey, and Raleigh-Durham-Chapel Hill, North Carolina; Stage 2--Dallas Fort Worth, Texas, Louisville, Kentucky, Orlando, Florida, and Philadelphia, Pennsylvania; Stage 3--Boston, Massachusetts, Chicago, Illinois, Orange County, California, and Seattle, Washington, and Stage 4 - Los Angeles, California, Minneapolis-St Paul, Minnesota, San Diego, California, and Worchester, Massachusetts.

Questionnaire content We designed a two-part survey to be administered by telephone to directors of hospital pharmacies and pharmacists most responsible for preparing reports for the pharmacy and therapeutics (P&T) committee. The survey collected information on hospital characteristics, including provision of services on a capitated or a per diem basis, details about the hospital's formulary, generic and therapeutic interchange, drug utilization review, whether specific recently-introduced drugs had been reviewed and adopted, knowledge of cost-effectiveness studies, use of such analyses in decision making, and how such studies could be made more useful to the hospital. The second part involved detailed questions about product adoption and use of cost-effectiveness analysis whereas the first part focused on cost containment strategies. The two parts were often asked of two different people in the pharmacy department of the hospital. Usually the director completed the first part of the survey and the pharmacist completed the second part. Many of the questions were drawn from a 1992 survey developed by the American Society of Hospital Pharmacists (1992). The two-part survey took about 25 minutes to complete on average.

Response rate A total of 156 hospitals were contacted, about equally divided among hospitals in the four managed care stages (Table 1). From these, we were able to conduct 103 Part One interviews (66%). Of the 53 nonrespondents, there were 31 refusals. Of the remaining 22, two were dropped because the hospital had closed, one merged with another hospital in our sample, four had the same director of pharmacy as another hospital in the sample, and seven interviews could not be scheduled before the inter*The adjustment converts outpatient visits into day or admission equivalents.

525

Table I. Responserates by hospitalcharacteristic Characteristic

Responding (103)

Stage of managedcare Stage 1 27% Stage 2 24% Stage 3 23% Stage 4 26% Total 100% Mean number of beds 356 Public For profit Voluntary Total

Ownership type 14% 18% 68% 100%

Council of Teaching Hospitals Member 34%

Nonresponding (53) 2t% 30% 26% 23% 100% 360 6% 21% 73% 100% 30%

view closing date. Reasons given for refusing were primarily "too busy" and "we have a policy of not doing interviews.'" The refusal rate was almost evenly distributed among the four stages of managed care penetration: Stage l-7; Stage 2-9: Stage 3 8; and Stage 4 7. Respondents and nonrespondents were similar with respect to stage of competition in the market, bedsize, and teaching status (Table 1). Public hospitals were somewhat relatively overrepresented among respondents.

Annual hospital surveys We used hospital-specific information from Annual Hospital Surveys conducted by the American Hospital Association for hospital-wide data on hospital cost, outputs, input use, and general information on hospital characteristics. We used the most recently available survey, which was for 1993, to analyze cost and staffing. We compared the 1988 and 1993 surveys to gauge trends in cost and staffing for hospitals in areas with different managed care penetration.

Equation specOqcation Dependent variables. We developed dependent variables to measure hospital adoption of various cost containment-quality assurance strategies. To assess the effect of managed care penetration on hospital-wide indicators ~?[ cost and input use, we defined variables for expense per adjusted patient day and expense per adjusted admission, and numbers of total full-time equivalent (FTE) personnel and pharmacists per adjusted patient day and per adjusted admission*. We deflated expense per adjusted patient day and per adjusted admission by an area price index based on the Health Care Financing Administration's wage index. To construct the index, we weighted the wage index by the national ratio of hospital wage/salary plus fringe benefit expense to total expense. For nonlabor inputs, we assumed no differences in prices. We

526

Frank A. Sloan et al.

specified the dependent variable in natural logarithmic form. To measure restrictiveness of drug formularies, we examined a variable for self-evaluated restrictiveness of formulary. Respondents were asked to rate the hospital's formulary on a scale of one to five with one being closed and five being open. We used ordered probit analysis to assess determinants of self-evaluated formulary restrictiveness. In a separate analysis we examined specific indicators of formulary restrictiveness, including: nonformulary drugs not routinely stocked; less than 5% of total hospital expenditures on pharmaceuticals spent on nonformulary drugs; and prescribers monitored for use of nonformulary drugs. For the analysis of product innovation, we defined three dependent variables: (1) number of new drugs reviewed; (2) number of new drugs adopted; and (3) number of drugs adopted that ,,ere approved by the U.S. Food and Drug Administration (FDA) that represent a therapeutic advance because either no effective drugs were available for the condition or the new drug was shown to be more effective than the drug currently used for the condition or the new drug had important advantages such as convenience and reduced side effects. Our survey requested information on review and adoption of 9 prescription drugs introduced in the U.S. from 1992 to 1994. These were selected from 76 drugs introduced during this period. The new drugs were: cefpodoxime (Vantin); cefprozil (Cefzil); loracarbef (Lorabid); piperacillin/tazobactam (Zosyn); nabumetone (Relafen); oxaprozin (Daypro); enoxaparin (Lovenox): granisitron (Kytril); and salmeterol (Serevent). Two of these drugs, Lovenox and Serevent, were considered by the FDA to provide significant therapeutic gain. To measure generic and therapeutic interchange, we defined variables for (1) whether or not generic substitution was permitted and (2) number o f drug categories for which therapeutic interchange was permitted (up to eight). We defined generic substitution as substitution of a chemically-equivalent drug by a pharmacist which does not require consent of a prescriber. Therapeutic interchange involves substitution of a drug by a pharmacist that is therapeutically but not chemically-equivalent. The drug categories we inquired about were: anti-infectives; anti-ulcer; cardiovascular; nonsteroidal antiinflammatory drugs; thrombolytic agents; antiemetics; central nervous system drugs; and analge*These variables were: binary variables for whether the hospital sold services on a (1) capitated basis or on a (2) fixed per diem basis; Medicare days as a fraction of total inpatient days; Medicaid days as a fraction of total inpatient days; and affiliation with a multihospital system. The capitation and fixed per diem variables w e r e defined by hospital for 1993 from the American Hospital Association's Annual Survey of Hospitals.

sics. To study use of drug utilization review, we defined a binary variable for whether or not the hospital had drug utilization guidelines. To gauge the extent of use of information on cost-effectiveness analysis in pharmacy decision making, we specified a dependent variable on frequency with which information on cost-effectiveness of a new drug was provided to the pharmaceutical and therapeutics committee. We assessed responses to a question whether cost-effectiveness information was provided to the P & T committee "'often," "seldom," or "never." The survey requested information about why cost-effectiveness analysis was not used more often in pharmacy decision making as well as how such studies could be improved to be of greater practical use. Explanatoo, variables. The key explanatory variables pertain to managed care and competition among hospitals. We defined variables for each of the four managed care stages, with Stage 1, the omitted reference group. We also controlled for hospital characteristics. hospital bedsize; teaching status (Council of Teaching Hospital member); and ownership (profit, private nonprofit, the omitted reference group, and public). In preliminary analysis, we included a larger number of explanatory variables. Including the larger number of variables did not appreciably affect our findings, but the associated coefficients were not statistically significant. Especially because of the small hospital sample, these variables were dropped*.

3. RESULTS Hospital cost and input use There were substantial differences in hospital cost per patient day but not per adjusted admission (Table 2). The discrepancy between the cost per day and admission results is fully explained by shorter lengths of stay in areas with higher managed care penetration. Relative to Stage 1, Stage 2 hospitals cut length of stay, but apparently not inputs per admission (service intensity). This is apparent from the lack in change in relative cost per admission but with a dramatic increase in cost per adjusted patient day (from a 7% to a 29% differential in cost per adjusted patient day relative to Stage 1). By contrast, in Stages 3 and 4, relative service intensity fell between 1988 and 1993. In 1988 Stage 3 hospitals had a cost per adjusted patient day which was 35% higher than Stage l's. Five years later the differential had fallen to 19%, A similar pattern was observed for Stage 4 hospitals. For Stage 4 the differential fell from 49% in 1988 to 10% in 1993. Presumably ¢asemix complexity rose as well even though there are no measures of this for the hospital population as a whole. The drop in service intensity in areas of high managed care penetration is not evident for full-

H o s p i t a l p h a r m a c y decisions

527

Table 2. Hospital cost regressions Dependent variables Cost per adjusted Patient day Explanatoryvariables Intercept Stage 2

Stage 3

Stage 4

Teaching Public For profit Beds (log) Re R2(C) F*

1993

Cost per adjusted Admission

1988

5.82 (0.46) b 0.25 (0.11 )~' [0.29] 0.18 (0.11) c [0.19] 0.26 (0.10) a [0.30] 0.19 (0.10) c 0.062 (0.12) 0.042 (0.10) 0.15 (0.081 )~ 0.22 0.16 3.74

5.89 (0.33) 2 0.070 (0,074) [0.073] 0.30 (0.074) ~ [0.351 0.40 (0.072) :' [0.49] 0.18 (0.069) h 0.11 (0.082) 0.12 C0.072)b 0.067 (0.057) 0.37 0.33 7.79

1993 8.35 (0.41)" -0.024 (0.093) [-0.024] -0.047 (0.095) [-0.046] 0.072 (0.091) [0.075] 0.31 (0.087)" 0.055 (0. I 0) -0.047 (0.092) 0.070 (0.71) 0.25 0.20 4.61

1988 8.69 (0.43) ~ -0.045 (0.096) [-0.044] -0.060 (0.10) [-0.058] 0.082 (0.094) [0.085] 0.37 (0.090) ~' 0.063 (0.11 ) -0.070 (0.094) 0.009 (0.075) 0.26 0.20 4.58

"Two-tailed t-test significant at 1%. ~Two-tailed t-test significant at 5%. CTwo-tailed t-test significant at 10%. *The degrees of freedom for the 1993 analysis are (7,95). For the 1987 analysis they are (7,91).

time personnel employed by hospitals (Table 3). The differential per FTE per adjusted patient day rose appreciably for Stage 2 and 3 and did not fall for Stage 4. On average hospitals in Stage 1 added 176 FTEs between 1988 and 1993 and hospitals in Stage 2 added 433 FTEs. By contrast, Stages 3 and 4 decreased personnel: 60 FTEs for Stage 3 and 12 FTEs for Stage 4. Thus for at least Stages 3 and 4,

the increase in service intensity was mainly due to a substantial decrease in length of stay. Hospitals in Stages 1, 2, and 4 increased the number of FTE pharmacists between 1988 and 1993 (2.9, Stage 1; 3.9, Stage 2; 1.2, Stage 4). Although Stage 3 hospitals decreased the number of FTE pharmacists by 1.2 on average, they increased the number of pharmacist technicians by 2.4 FTEs.

Table 3. Hospital full-time personnel regressions Dependent variables FTEs per adjustedPatient day Explanatoryvariables Intercept Stage 2

Stage 3

Stage 4

Teaching Public For profit Beds (log) Rz R2(C) F*

1993 -5.70 (0.38) '~ 0.29 (0.088)" ]0.34] 0.30 (0.089)" [0.35] 0.19 (0.085)" [0.21] 0.15 (0.082) c 0.11 (0.096) -0.007 (0.086) 0.19 (0.067)" 0.34 0.30 7.10

t 988 -5.17 (0.32) ~ 0.13 (0.072)~ [0.14] 0.18 (0.073) b [0.20] 0.18 (0.071)h [0.20] 0.17 (0,068) ~' 0.27 (0.081 )" -0,042 (0.071) 0.095 (0.056)~ 0.37 0.32 7.61

FTEs per adjustedAdmission 1993 -3.34 (0.30) ~' 0.050 (0.068) [0.051] 0.1)58 (0.070) [0.060] -0.089 (0,067) [-0.O85] 0.25 (0.064)" 0.12 (0.075) -0. I 0 (0.067) 0.14 10.052)" 0.50 0.46 13.31

~Two-tailed t-test significant at I%. bTwo-tailed t-test significant at 5%. CTwo-tailed t-test significant at 10%. *The degrees of freedom for the 1993 analysis are (7,95). For the 1987 analysis they are (7,91).

1988 -3.19 (0.37) '~ -0.027 (0.083) [-0.026] 0.12 (0.084) [0.131 0.058 (0.082) [0.O60] 0.23 (0.078)'~ 0.18 (0.093)c -0.18 (0.081 )h 0.11 (0.065)~ 0.38 0.33 7.89

Frank A. Sloan et al.

528

Table 4. FTE pharmacists regressions Dependent variables Pharmacists peradjusted patient day Explanatoryvariables

1993

Intercept Stage 2

Stage 3 Stage 4

Teaching Public For profit Beds (log) R -+ R2(C) F*

Pharmacists peradjusted admission

1988

1993

1988

- 10.21 (0.66) ~ 0.50

-9.57 (0.56)" 0.25

-7.79 (0,63)" 0.27

-7.50 (0.54)~ 0.09

(0.15) ~

(0.13) b

(0.14) ~

(0.12)

[0.66] 0.41 (0.15)~ [0.511

[0.28] 0.26 (0.13)h [0.301

[0.31] 0.17 (0.15) [o.19l

[0.09] 0.20 (0.12) [0.22]

0.40 (0.15)" [0.49] 0.24 (0.14) b 0.18 (0.17) 0.19 (0.15) 0.14 (0.12) 0.22 0.16 3.69

0.22 (0.12) b [0.25] 0.065 (0.12) 0.23 (0.14) ~ -0.013 (0.12) 0.069 (0.098) 0.12 0.05 1.70

0.15 (0.14) [0.161 0.35 (0.13)" 0.19 (0.16) 0.13 (0.15) 0.081 (0. I I ) 0.19 0.13 3.19

0.12 (0.12) [0.13] 0.14 (0.11) 0.15 (0.13) -0.12 (0.12) 0.067 (0.094) 0.13 0.07 1.96

~Two-tailed t-test significant at 1%+ hTwo-tailed t-test significant at 5%. CTwo-tailed t-test significant at 10%. *The degrees of freedom for the 1993 analysis are (7,95). For the 1987 analysis they are (7,91)+

The number of pharmacists per adjusted patient day rose dramatically relative to Stage 1 between 1988 and 1993 in Stages 2 through 4, particularly in Stage 2 (38% difference, Stage 2; 21% difference, Stage 3; and 24% difference, Stage 4) (Table 4). The increases in pharmacists in Stage 4 hospitals stand in contrast to the decrease in overall employment, Increases in pharmacists may be seen as an attempt by hospitals to control pharmacy budgets in response to competitive pressures.

Drug formulary policies Formulary restrictiveness. The majority of hospitals regarded their formulary policy as quite restrictive (Fig. 2). Using ordered probit analysis and several alternative measures of restrictiveness, we tested for differences in self-rated restrictiveness by managed care stage. There were no differences by managed care stage. 45

J

40

g,

35 30

N N r,,

25 20

15 10 5 0

\

\ \ \ \

v Most I

1 3

Least 5

Fig. 2. Self-rated restrictions of hospital formulary.

Specific measures of formulary activity were consistent with the above findings that most hospitals had restrictive formularies. Almost 90% of hospitals did not routinely stock nonformulary drugs (Table 5). About three-fifths spent less than 5% of total pharmaceutical expenditures on nonformulary products. About half monitored prescribers for excessive use of nonformulary drugs. Formulao, management. The vast majority of hospitals evaluated their formularies for ineffective and obsolete drugs, reviewed therapeutic categories with high risk, high volume, or expensive drugs, and considered the cost impact of drugs being considered for addition to the formulary (Table 5). As with restrictiveness, there were no statistically significant differences in policies by managed care stage.

Addition of recently introduced drugs to the formulary. There were no differences by managed care stage in the number of new drugs in our list that were reviewed (Table 6). However, Stage 3 hospitals adopted fewer drugs in total and fewer of the drugs considered by the F D A to provide significant therapeutic gain. This does not continue for Stage 4, possibly reflecting a relaxation of the reluctance to add drugs in the more mature stages. The difference between Stage 3 and 1 was 1.3 drugs (out of the nine drugs). Cost was the second most frequently mentioned reason for adding a drug to the formulary and ranked third as a reason for deciding not to add a drug (Table 7). In 71% of cases, the drug had not been reviewed at all.

529

H o s p i t a l p h a r m a c y decisions Table 5. Specific indicators of formulary policy Indicator

Percent responding "yes"

Part A: Restrictiveness Nonformulary drugs not routinely stocked Expenditures on nonformulary drugs less than 5% Prescribers monitored for excessive use of nonformulary drugs

88 61 52

Part B: Management Formulary evaluated periodically for ineffective and obsolete drugs Therapeutic categories with high risk or volume or expensive drugs regularly reviewed Cost impact of formulary addition calculated always or often Well-documented reviews for all/most drugs considered for addition Pertinent deletions of similar drugs suggested always/often Formulary contains information on cost of drugs to hospital

95 92 88 81 74 32

The four most common sources of information used in formulary decision making were Micromedix, Medline, Medical Letter, and the New England Journal of Medicine (Table 8). PharmacoEconomics, a journal devoted specifically to economic evaluation of pharmaceuticals was used by 6% of hospitals.

Generic" substitution and therapeutic' interchange Generic substitution and therapeutic interchange were both widely practiced in the hospitals surveyed. Generic substitution was permitted without consent of the prescriber in 94% of hospitals. Therapeutic interchange was permitted by 79%. Therapeutic interchange was most commonly permitted for antiinfectives and antiulcer drugs (Fig. 3). In part because of the high percentages adopting generic substitution and therapeutic interchange, there were no differences among hospitals by extent of managed care penetration.

Drug utili~-ation review Most hospitals had some drug utilization guidelines (59%). Examples by therapeutic category are guidelines for antiinfectives and cardiovascular drugs. Generally however, utilization guidelines

were defined for specific disease entities rather than for particular drug therapies.

Use of cost-effectiveness analysis in decision making According to our respondents, 72% of hospital pharmacy departments used some type of cost effectiveness in decision making. However, only 37% said that information on cost-effectiveness of new drugs was often presented to the hospitals pharmacy and therapeutics committee when a drug was being considered for addition to the formulary. Information on cost effectiveness of particular drugs was most frequently found in peer-reviewed journals. Overall, 68% mentioned peer-reviewed journals as a source of cost-effectiveness information; by contrast, practitioner-oriented journals and promotional materials were mentioned by 23% and 13% of hospitals, respectively. Another 14% of hospitals used other sources of information on costeffectiveness of drugs, such as in-house studies and Micromedix. Drug categories for which respondents had seen the most studies were granistron, kytril, H2 blockers, Lovenox, and antibiotics (Table 9). Respondents gave several reasons why cost-effectiveness is not used most often in pharmacy decision making (Table 10). In terms of frequency,

Table 6. Number of new drugs reviewed and reviewed and added Dependent variables (Poisson) Explanatory variables Intercept Stage 2 Stage 3 Stage 4 Teaching Public

Reviewed 1.2ff' (0.13) 0.10 (0.13) [0.42] -0.18 (0.15) [-0.73] 0.17 (0.13) [0.70] -0.046 (0.12) 0.13

(0.14) For profit Beds

-0.35 b (0.16) 0.0007" (0.0003)

~Two-tailed t-test significant at 1%.

Reviewed and added 0.97" (0.16) -0.016 (0.16) [-0.43] -0.4T (0.19) [-1.27] -0.14 (0.16) [-0.37] --0. I 0 (0.15) 0.30 ~

(0.16) -0.22 (0.18) 0.0006" (0.0003)

Reviewed and addedlarge gain 0.14 (0.26) 0.018 (0.25) [0.018] -0, 50~ (0.30) [-0.51] -0.12 (0.26) [-0.13] -0.22 (0.25) 0.035

(0.28) -0.53 ~ (0.32) 0.0006 (0.0005)

530

Frank A. Sloan

Table 7. Reasons to add/not add drugs to hospitals formulary Reason

Percent mentioning

Part A: Reason to add Clinically superior Cost MD request Easier dosing ~Better than existing drugs" Other (each item less than 4%) Part B: Reason to not add Drug not reviewed Duplicate therapy or "me too" drug Clinically inferior High cost Other (each item less than 4%)

44 26 13 10 4 17 71 15 8 8 4

et aL

50 45 40 35 30 25 2O 15 10 5 0 Ant iinf

Ant iulf

Car dio v

NS AID

An alg esti

Ant iem

CN S

Thr om o

Fig, 3. Therapeutic interchange permitted. the top four reasons for lack of use were: "no studies on drugs of interest; .... studies do not apply to hospitals; . . . . don't have the knowledge or time or staff to evaluate these studies;" and "studies are biased due to industry sponsorship." Several ways were suggested to make cost-effectiveness studies more useful to hospital pharmacy decision making (Table 11). The top three suggestions were to use data that permits generalization to hospitals, to make the results available sooner, and to have the studies conducted by independent organizations.

4. DISCUSSION

Managed care has caused shortening in the length of hospital stays and to some extent a decrease in service intensity (Miller and Luft, 1994). Compared to the size of the reduction in length of stay, changes in the way that a day of care is delivered appear to have been minor, even in areas with substantial managed care share. As managed care reaches Stage 5, where managed care penetration is complete, there may be meaningful changes in service intensity. We conducted the more general analysis of hospital changes to provide a context for our more detailed analysis of pharmacy decision making within hospitals. It is noteworthy that the sample Table 8. Source of outside information commonly used in formulary decision making Name of source Micromedix Medline Medical Letter New England Journal of Medicine Formal Service The Formulary Fix P and T American Journal of Hospital Pharmacy Journal of the American Medical Association Annals of Internal Medicine Hospital Formulary PharmacoEconomics Annals Pharmacotherapy ASHP

Percent mentioning 23% 22% 16% 16% 15% 12% 10% 9% 8% 7% 6% 6% 5% 4%

hospitals in areas with substantial managed care have added other types of skilled personnel, such as registered nurses. There is some evidence to suggest that increasing the number of pharmacists can decrease the drug budget of a hospital through increased ability to educate physicians and monitor policies (Bastean, 1995). Barnes Jewish Christian Hospital in St Louis, Missouri provides its pharmacy with a global budget which can be used by the pharmacy for personnel or drugs. The pharmacy department decided to hire more clinical pharmacists and by doing so claim to have reduced their pharmacy costs and to be consistently under budget. Another hospital was about to decrease the number of pharmacists when a software package was brought in designed to track savings per pharmacist. After examining the data the hospital hired 12 new pharmacists (Bastean, 1995). What is not known is if the pharmacists save a hospital money over a prolonged period of time or lead to total savings in hospital costs. Are individuals who are given less expensive forms of drugs, different drugs, or no drugs more likely to have longer stays and repeat hospitalizations? An alternative interpretation is that the pharmacy has escaped the pressures of cost containment, at least for the moment. It is noteworthy that the sample hospitals in areas with substantial managed care have added other types of skilled personnel, such as registered nurses. Additional nursing positions may be justified on grounds that RNs are better able diagnose and react to unforeseen clinical events than less trained personnel and may substitute for both more and less skilled personnel (Wunderlich e t a l . , 1996). It has been hypothesized that the rate of product innovation is lower under capitation than under retrospective cost or charge reimbursement. At the very most, our analysis supports this view only in part. In addition to going against the trend to increase pharmacists with increasing managed care, Stage 3 also had a lower rate of innovation of new drugs. In Stage 3 more hospitals adopted neither of the two drugs providing significant therapeutic gain

Hospital pharmacy decisions

531

Table 9. Drug categories for which pharmacists have seen cost-effectiveness analyses Drug category

Number seeing studies

GI drugs (Zantac, Pepcid, Axid, Tagamet. Kytril, Zofran) Antibiotics (Cephalaprins, Quinolones) Anti-thrombolytics (Lovenox, tPa, Streptokinase Heparin) Cardiovascular drugs (Calcium Channel Blockers, ACE Inhibitors. Beta Blockers) Adjunct of coronary angioplasty (Riapro, Axsmibab) Antiasthmatic NSA1Ds (Toradol) Other

than in both Stage 1 and 2 hospitals combined. Stage 3 may be a transition stage, in which hospitals decide where and how to cut costs without jeopardizing effectiveness. We have focused our study on hospital pharmacies. Based on a comprehensive case study of five hospitals, four hospital systems and other organizations, Luce and Brown (1995) found that, if anything, hospital pharmacy decision makers appear to be more active in their use of technology assessment than those in the other parts of the hospital. In contrast, hospital technology assessment for nondrug technology tended to be a prudent purchase decision based on a financial assessment with little or no formal evaluation of changes in patient outcomes or practice patterns. Formulary committees in hospitals were rather sophisticated and took into account patient outcomes, total drug, and sometimes total hospital costs in their assessments. Compared with other committees, formulary committees were more familiar with technology assessment concepts. They were more discriminating about the source of clinical information, were more sensitive to study design, sample size, sources of support for the study and the institutions involved. They were more likely to examine peer-reviewed journal articles for information on effectiveness and report skepticism of company claims in the absence of evaluation by an independent group. Our study confirms that there is a substantial amount of scrutiny of the pharmaceutical literature within pharmacy departments, at least in hospitals of the size range we studied. The overall lack of differences in hospital formulary restrictiveness and management by stage of managed care penetration could mean that there are truly no differences or that our survey did not capture them. A telephone survey may be unable to capture subtle differences in the oversight and imTable 10. Why is cost-effectiveness analysis not used more often? Reason not used more often No studies on drugs of interest

Studies do not apply to hospitals Don't have knowledge/time/staff to evaluate studies Studies published too late Studies are biased Cannot locate studies SSM 454~B

Percent mentioned 34 28 15 10 7 4

38 34 25 8 8 6 4 16

plementation of policies. Many studies of drug formularies have found that restrictive formularies save money (Pearce and Begg, 1994; Segal et al., 1994; Barriere, 1986). These studies, however, have certain limitations. First, many studies examined only a few drugs, primarily antibiotics. Savings from a few drugs may not generalize. Second, they have focused on pharmacy costs rather than on total treatment expense. Saving in pharmacy expense may be offset by added expenses of other types such as increased length of stay. Third, the studies have been based on data from only a single hospital or at most a handful of hospitals. Some empirical evidence based on a more detailed measure of hospital cost and a much larger sample of hospitals suggests that restrictive hospital formularies do not save money (Sloan et al., 1993). These findings are consistent with the results of other research on the effect of formulary restrictions (Kozma et al., 1990: Moore and Newman, 1993). Other work on Medicaid drug payment limits also suggest that limitations may be counterproductive in terms of cost containment (Soumerai et al., 1991). In this sense hospitals, especially those under competitive pressures, may be making rational decisions not to adopt a restrictive formulary. To our knowledge, ours is the first survey to investigate the link between hospital formularies and use of cost-effectiveness analysis. In a 1989 survey, Rascati interviewed 130 randomly selected hospital pharmacy directors (Rascati, 1992). Virtually all of the hospitals had a formulary of some type. The main purpose of the formulary systems was said to decrease cost followed by to ensure appropriate therapy. However, at a substantial minority of hospitals, the formulary was simply a drug list with no supporting information. Most of the hospitals had formal procedures for adding drugs to the formulary, but the survey did not provide details about the nature of these procedures. Use of cost-effectiveness analysis or related techniques was not mentioned. In commenting on the Rascati article, Crane distinguished between three types of formulary systems: inventory management approach; cost accounting; and clinical decision analysis (Crane, 1993). The third approach uses cost-effectiveness concepts even if not formal cost-effectiveness analysis.

532

Frank A. Sloan et al. Table 11. How can cost-effectiveness analysis be made more useful to hospitals? Suggestion

Percent mentioning

Make generalizable to hospital settings Make available sooner Sponsor independent research More transparent explanation/presentation of methods Publish studies/bibliography in pharmacy literature Use data representative of populations hospitals treat Educate hospital staff in evaluating studies Make "head-to-head" comparisons among drugs

The vast majority of hospitals in our survey had implemented some form of therapeutic interchange and generic substitution even though evidence on the effectiveness of these techniques is also mixed. However, we did not obtain information on how frequently the techniques were used. A number of studies present findings to support the effectiveness of therapeutic interchange as a cost containment mechanism (Achusim (1992)--antimicrobial agents, H2-receptor antagonists, thrombolytic agents, plasma expanders; Brown and Clarke (1992)--cefazolin with metronidazole for cefoxitin; Lawrenz et al. (1991)--ampicillin-sulbactam for cefoxitin; Oh and Franko (1990)--intravenous famotidine for cimetidine and rantidine; Rich (1989)--H2 antagonists; Richardson (1993)--ACE inhibitors; Wright (1991)--antimicrobial agents). These studies, however, suffer from the same deficiencies noted above for studies of hospital formularies. In fact, a recent multivariate analysis, using more comprehensive measures of cost finds no effect (Sloan et al., 1993). Again, as with formularies, there may be some doubts in the hospital community about the effectiveness of therapeutic interchange. Most hospitals use some drug utilization guidelines, but as of mid 1995 these were not yet important management tools for hospital pharmacies. In most cases the guidelines cut across several fields and are not restricted to the pharmacy. Progress awaits further acceptance of further disease management strategies. At most cost-effectiveness is a minor tool in pharmaceutical decision making in hospitals at present. We could determine no differences by managed care market share. One impediment to its use is the lack of timeliness of studies. We conducted a Medline search to identify cost-effectiveness studies relating to the nine newly introduced drugs used in our survey. We were able to find only two cost-effectiveness studies, both pertaining to the same drug, enoxaparin. These studies appeared more than a year after the drug was placed on the market (Drummond et al., 1994; O'Brien et al., 1994). A follow-up literature search was conducted using the International Pharmaceutical Abstracts (IPA) system. The IPA includes references to Pharmacoeconornics, which Medline excludes. However, no further articles relating to cost-effectiveness of new drugs were discovered. We con-

29 28 22 17 17 15 14 12

ducted a third search on a new database created by the International Federation of Pharmaceutical Manufacturers' Association and the Office of Health Economics in London. This search produced several articles. Four of the seven articles simply referenced the cost work of others or included only direct drug costs. Of the remaining three, one was on enoxaparin and appeared in 1994, one was on cefpodoxime and appeared in 1994, and the last was on piperacillin and appeared in 1993. One reason for the delay in study in results is that cost-effectiveness is not part of the clinical trial process. Whether this should be a requirement however involves a weighing of timeliness of results versus the additional cost of adding more regulatory requirements. Other stated reasons for not using cost-effectiveness analysis more often pertained to the lack of information on hospitalized patients and hence on the potential cost offsets accruing to the hospital. The message here is clear. To be relevant to hospitals, it is necessary to conduct cost-effectiveness analysis on their patients. Inferences from other populations will not be fully accepted. The issue of cost offsets is an issue in own right. The hospital in Stage 1 would be primarily interested in savings realized during the stay, if interested at all. By contrast, hospitals in Stages 4 or 5 are likely to be part of a capitated network, thus downstream savings should be of much greater interest. The lack of independent sponsorship is also an issue to hospitals. One hospital pharmacist remarked that he had never seen a study that did not prove that the drug was cost-effective. Costeffectiveness studies are viewed by some as another marketing tool. Lack of financing for cost-effectiveness studies is clearly a problem. It is possible to provide financial support for a study without simultaneously producing it and/or controlling dissemination of findings. As Luce and Brown suggested, the federal government or a public/private entity could fund a technology assessment clearing house (Luce and Brown, 1995). A barrier identified by some respondents was inadequate expertise in economic evaluation. Several methods may be used to increase expertise, including but not limited to formal training. Overall, the lack of use of cost-effectiveness analysis, and more generally the other techniques

Hospital pharmacy decisions p r o p o r t e d to save m o n e y may reflect a lack of incentive to contain per diem cost at the present time. Possibly after savings from reduced admission rates a n d lengths of stay have been fully realized, hospitals will turn to specific measures aimed at reducing per diem costs. T h a t this would occur must remain s o m e w h a t speculative. Acknowledgements--We would like to thank Elizabeth Kulas and James Caverlee for their help on the empirical research.

REFERENCES Achusim, L. E. (1992) Therapeutic interchange as a costcontainment mechanism. PharmacoEconomics 2, 347. American Society of Hospital Pharmacists (1992) ASHP technical assistance bulletin on assessing cost containment strategies for pharmacies in organized health care settings. Ameriean Journal of Hospital Pharmacy 49, 155. Barriere, S. L. (1986) Controversies in antimicrobial therapy: formulary decisions on third-generation cephalosporins. American Journal of Hospital Pharmacy 43, 625-629. Bastean, B. (1995) Coordinator of PharmacoKenetics, Missouri Baptist Hospital. Telephone conversation 11/ 20/95. Baumgardner, J. R. (1991) The interaction between forms of insurance contract and types of technical change in medical care. Rand Journal of Economics 22, 36. Brown, G. R. and Clarke, A. M. (1992) Therapeutic interchange of cefazolin with metronidazole for cefoxitin. American Journal of Hospital Pharmacy 49, 1946. Crane, V. S. (1993) How to use structured decision making in developing therapeutic, cost-effective formulary systems. Hospital Formulary 28, 859. Drummond, M., Aristides, M., Davies, L. and Forbes, C. (1994) Economic evaluation of standard heparin and enoxaparin for prophylaxis against deep vein thrombosis in elective hip surgery. British Journal o[ Surgeo' 81, 1742. Hospitals and Health Networks (1995) How markets evolve. Hospitals and Health Networks March 5. Iglehart, J. K. (1993) Health policy report: the American health care system-managed care. New England Journal of Medicine 327, 742. Kozma, C. M., Reeder, C. E. and Lingle, E. W. (1990) Expanding Medicaid drug formulary coverage: effects on utilization of related services. Medical Care 28, 963. Lawrenz, C. A., Cole, P., Theodorou, A., Cook, R. L. and Bermann, L. (1991) Therapeutic interchange of ampicillin-subactam for cefoxitin. American Journal of Hospital Pharmacy 48, 2150.

533

Luce, B.R. and Brown, R.E. (1995) The use of technology assessment by hospitals, health maintenance organizations, and third-party payers in the United States. International Journal of Technology Assessment in Health Care 11, 79. Miller, R. H. and Luft, H. S, (1994) Managed care plan performance since 1980: a literature analysis. Journal of the American Medical Association 271(19), 1512. Moore, W. J. and Newman, R. J. (1993) Drug formulary restrictions as a cost containment policy in medical programs. Journal of Law and Economics 36, 71. Murray, D. (1995) The four market stages, and where you fit in. Medical Economics March 13, 44. O'Brien, B. J., Anderson, D. R. and Coeree, R. (1994) Cost-effectiveness of enovaparin versus warfarin prophylaxis against deep-vein thrombosis after total hip replacement. Canadian Medical Association Journal 150, 1083. Oh, T. and Franko, T. G. (1990) Implementing therapeutic interchange of intravenous famotidine for cimetidine and rantidine. American Journal ~?[ Hospital Pharmacy 47, 1547. Pearce, M. J. and Begg, E. J. (1994) Hospital drug therapy cost containment through a preferred medicines list and drug utilization review system. New Zealand Medical Journal 107, 101. Rascati, K. L. (1992) Survey of formulary system policies and procedures. American Journal of Hospital Pharma~3' 49, 100. Rich, D. S. (1989) Experience with a two-tiered therapeutic interchange policy. American Journal o[" Hospital Pharmacy 46, 1792. Richardson, C. E. (1993) The ACE inhibitors. P & T January. 40. Segal, R., Oh, T., Ben-Joseph, R. and Russell, W. L. (1994) A pharmacoeconomic analysis of IV H2-receptor antagonist use in 40 hospitals. Hospital Formulary 29, 379. Sloan, F. A., Gordon, G. X. and Cocks, D. L. (1993) Hospital drug formularies and use of hospital services. Medical Care 31,851. Soumerai, S. B., Ross-Degnan, E., Avorn, J., McLauglin, T. J. and Choodnovskiy, I. (1991) Effects of Medicaid drug-payment limits on admission to hospitals and nursing homes. The New England Journal of Medicine 325, 1072. Weisbrod, B. A. (1991) The health care quadrilemma: an essay on technological change, insurance, quality of care. and cost containment. Journal o1" Economic Literature 29, 523. Wright, D. B. (1991) Antimicrobial formulary management: a case study in a teaching hospital. Pharmacotherapy Suppl, to 11, 27S. Wunderlich, G. S., Sloan, F. A. and Davis, C. K. (eds) (1996). Nursing Staff in Hospitals and Nursing Homes. Is It Adequate? Institute of Medicine, National Academy Press, Washington, DC.