Prescribing behaviour after the introduction of decentralized drug budgets: Is there an association with employer and type of care facility?

Abstract

Objective

To analyse whether prescribing patterns changed after introduction of drug budgets and whether there is an association between drug prescribing patterns and the type of employer and care facility.

Methods

Data analysed encompassed information on dispensed medicines, by workplaces, prescribed in the Region Västra Götaland, Sweden, for the years 2003 and 2006. Workplaces (n = 969) were categorized according to type of employer and type of care facility. Five prescribing indicators reflecting goals for cost-containing prescribing in Region Västra Götaland were assessed. Changes over time and differences between different types of employer and care facility were analysed by Mann–Whitney tests.

Results

In 2003, workplaces with a public employer had a significantly higher adherence to three of the prescribing indicators compared with private practitioners. Two of these differences remained in 2006. In 2003, none of the prescribing indicators differed between primary care and other care facilities. Three years later workplaces in primary care had a significantly higher adherence to three of the prescribing indicators than other care facilities. There was a statistically significant difference in change between 2003 and 2006 between primary care and other care facilities; there were no differences in change between workplaces with public and private employers.

Conclusions

Adherence to three of the prescribing indicators increased after the introduction of decentralized drug budgets. Workplaces with a public employer showed greater adherence to two of the prescribing indicators than private sector workplaces.

In order to increase cost consciousness in prescribing, the financial responsibility for prescribed drugs in Sweden has been transferred from governmental level to the county councils.

• Adherence to prescribing indicators differed between workplaces with a public employer and those with private practitioners and between primary care facilities and other healthcare facilities.

• Adherence to three of the prescribing indicators changed over time; all these changes were in a favourable direction.

Due to high expenditures on pharmaceuticals over recent decades, pharmaceutical policy reforms have been implemented in Sweden and in other countries [1], [2]. A Swedish reimbursement reform in 1997 transferred financial responsibility for prescribed drugs from the government to the county councils. This transfer was introduced stepwise: in 2003 it was partly decentralized and since 2005 the county councils have had full budget responsibility. County councils have been free to decide on how to handle this responsibility. In the county council of the Västra Götaland Region, budget responsibility was handled on a central level in 2003 in publicly owned healthcare whereas private care units had no financial responsibility for prescribed drugs. Another pharmaceutical benefits reform was introduced in October 2002, encompassing for example mandatory generic substitution and workplace codes on prescriptions. A questionnaire study investigating opinions on the reform among physicians working for the county council of Västra Götaland Region found that those employed by the county council had a more positive view towards generic substitution, compared with doctors employed elsewhere [3].

General practitioners (GPs) have reported that factors influencing prescribing include the price of the product for the patient, time–space, regulatory framework, and benefit for the patient [4], [5]. Some studies, although mostly cross-sectional, have investigated physicians’ attitudes in relation to medication costs and prescribing behaviour 6–8. GPs with high and low prescribing costs, respectively, displayed different attitudes regarding cost issues [6]. In two other studies, physicians agreed that it was important to consider cost when prescribing, but their knowledge of the actual prices of drugs was low [7], [8]. Prescribing indicators are commonly used to monitor prescribing behaviour [9], [10]. An Irish study found that GPs considered indicators based on prescribing of medicines of questionable efficacy/poor quality and drugs associated with good prescribing practice to be good indicators of prescribing quality [11]. These two groups of indicators were shown to be associated with relatively low and high prescribing rates in an Irish prescription database.

Some physician characteristics have been found to be related to certain prescribing behaviours. GPs have been reported to be more adherent to therapy recommendations than hospital specialists and private practitioners [12]. Furthermore, younger, residency-trained physicians reported greater confidence in generic drugs and more frequent generic prescribing [13]. In a Canadian study, male GPs and recently graduated specialists were more likely to prescribe newly registered prescription drugs [14]. Physician characteristics that have been related to greater relinquishment of inappropriate drugs are higher job satisfaction and valuing a high standard of medical knowledge [15].

It is important to analyse whether the Swedish reforms, including decentralized drug budgets, have had any impact on prescribing behaviour. Since type of employer influenced opinions on the reform, we were interested to analyse this factor in relation to prescribing behaviour. The aim of this study was therefore to analyse, if prescribing for a selection of drug groups, changed when financial responsibility for prescribed drugs was transferred to the county councils. A further aim was to investigate whether there was an association between adherence to the prescribing indicators and workplaces with different type of employer and type of care facility. The study was performed in Västra Götaland Region, Sweden's second largest county council.

Material and methods

Data on dispensed prescription drugs

Data on volumes of dispensed pharmaceuticals prescribed in Västra Götaland Region by workplace for the years 2003 and 2006 were collected from the National Corporation of Swedish Pharmacies (Apoteket AB). Volumes were expressed in defined daily doses (DDD). Workplaces (n = 969) were categorized according to employer (public vs. private) and type of care facility (primary care vs. other care facilities) in 2006. The category private employer encompasses all privately run healthcare facilities, i.e. both single-practitioner facilities and larger units in primary care as well as other care facilities. About three-quarters of the care units in the category other care facilities were hospital based. A sub-analysis was undertaken of workplaces with public employer by type of care facility, i.e. primary care versus other care facilities.

In addition, monthly data on sales of the prescribing indicator coxibes, described below, was retrieved for 2003 to 2006 to explore the impact of the withdrawal of rofecoxib.

Prescribing indicators

Five prescribing indicators were investigated reflecting the county council's goals for cost-containing prescribing (Table I). The prescribing indicators included the therapeutic groups ACE inhibitors, proton-pump inhibitors (PPIs), statins, non-steroidal anti-inflammatory agents, and antidepressants. Thus the prescribing indicators reflected several therapeutic areas where both high-cost and low-cost pharmaceuticals were available. Four of the prescribing indicators show the proportions of a first-line treatment in a therapeutic group whereas one (coxibes) shows the proportion of a group of drugs where caution has been recommended by the Swedish Medical Products Agency due to safety concerns. All prescribing indicators are based on dispensed volume in DDD and expressed as percentages of DDD. In 2003, the study encompassed 36.2 million DDDs of ACE inhibitors, the corresponding figure was for statins 27.1 million DDDs, for PPIs 13.8 million DDDs, for NSAIDs 16.8 million DDDs, and for SSRIs 22.6 million DDDs.

Table I.

Description of prescribing indicators analysed.

Indicator	Definition	Measurement (ATC-codes)	Preferred level of indicator
ACE inhibitors	The proportion of ACE–inhibitors of all agents acting on the rennin-angiotensin system	(C09A + C09B)/C09	High
PPI	The proportion of omeprazole of all PPIs	A02BC01/A02BC	High
Statins	The proportion of simvastatin of all statins	C10AA01/C10AA	High
Coxibes	The proportion of coxibes of all NSAIDs	M01AH/M01A	Low
SSRI	The proportion of citalopram of all SSRIs	N06AB04/N06AB	High

Statistical analysis

Changes between the years 2003 and 2006 and differences in prescribing indicators between different categories of workplaces, based on employer and type of care facility, were analysed using the Mann–Whitney test (due to non-normal distribution of data). Differences in change from 2003 to 2006 between workplace categories were analysed using the Mann–Whitney test. Tests where p < 0.01 were considered statistically significant. Missing cases were not included in the analyses. All statistical analyses were made using the statistical software SPSS v14.

Results

Comparisons by type of employer

Adherence to three of the indicators improved significantly between the years 2003 and 2006 for both public and private employers, respectively (Table II). Adherence to the indicators ACE inhibitors and statins was significantly higher among workplaces with a public employer than a private employer in both 2003 and 2006. This was also seen for coxibes in 2003 but the difference was not significant in 2006 when the level of the indicator was close to zero.

Table II.

Descriptive statistics and results from Mann–Whitney tests for workplaces with public and private employers.

					Change between 2003 and 2006
		Public employer (n = 726)	Private employer (n = 243)	Comparison public vs. private employer	Public employer	Private employer	Comparison of change public vs. private employer
Indicator	Year	Median (IR)	Median (IR)	p	p	p	p
ACE inhibitors	2003	71.4 (56.6–81.7)	48.5 (24.5–69.1)	<0.001	0.015	0.518	0.074
	2006	67.0 (49.9–78.8)	44.9 (26.0–62.5)	<0.001
PPI	2003	37.8 (24.7–51.6)	41.7 (26.8–65.6)	0.070	<0.001	<0.001	0.078
	2006	67.7 (50.4–77.7)	64.2 (51.1–77.3)	0.454
Statins	2003	59.1 (43.2–72.3)	46.1 (29.0–67.4)	0.002	<0.001	<0.001	0.438
	2006	76.9 (64.8–89.1)	64.1 (45.7–90.9)	0.001
Coxibes	2003	15.8 (5.2–27.2)	21.3 (7.8–31.2)	0.008	<0.001	<0.001	0.716
	2006	0.2 (0.0–3.1)	1.1 (0.0–5.9)	0.022
SSRI	2003	47.1 (39.9–61.4)	42.5 (24.1–57.8)	0.055	0.965	0.721	0.986
	2006	49.3 (24.8–63.3)	42.7 (26.9–58.6)	0.148

Notes: Median value, inter-quartile range (IR) of the prescribing indicators by type of employer for 2003 and 2006 are presented. Results of analyses of differences in the indicators between public and private employer and changes between 2003 and 2006 undertaken by Mann–Whitney test; p < 0.01 is considered statistically significant.

Comparisons by type of care facility

Between 2003 and 2006 there was a significant change in the preferred direction in all the prescribing indicators except ACE inhibitors for primary care workplaces (Table III). The corresponding analysis for other care facilities showed a statistically significant change in three of the indicators. For three of the prescribing indicators the change between 2003 and 2006 was larger in primary care compared with other care facilities. Adherence to the indicators PPI and SSRI was significantly higher among workplaces within primary care than among other care facilities in 2006. This was also seen for coxibes although the levels were on a very low level after the withdrawal of rofecoxib.

Table III.

Descriptive statistics and results from Mann–Whitney tests for workplaces with public and private employers by type of care facility.

					Change between 2003 and 2006
		Primary care (n = 478)	Other care facilities (n = 491)	Comparison primary care vs. other care facilities	Primary care	Other care facilities	Comparison of change primary care vs. other care facilities
Indicator	Year	Median (IR)	Median (IR)	p	p	p	p
ACE inhibitors	2003	66.1 (50.4–76.9)	67.8 (42.3–84.0)	0.244	0.191	0.021	0.180
	2006	62.5 (48.3–74.4)	61.8 (20.3–80.4)	0.293
PPI	2003	38.3 (31.0–50.3)	39.3 (18.9–59.1)	0.385	<0.001	<0.001	<0.001
	2006	70.8 (63.7–79.2)	56.5 (40.0–74.4)	<0.001
Statins	2003	56.3 (42.1–65.8)	57.5 (33.6–83.0)	0.310	<0.001	<0.001	<0.001
	2006	75.5 (63.4–83.5)	74.9 (50.0–100.0)	0.472
Coxibes	2003	15.5 (7.3–24.6)	18.2 (4.6–32.4)	0.042	<0.001	<0.001	0.964
	2006	1.7 (0.0–4.1)	0.0 (0.0–2.7)	<0.001
SSRI	2003	47.5 (37.9–57.9)	41.7 (21.5–66.7)	0.243	0.001	0.059	<0.001
	2006	52.5 (40.6–63.2)	35.9 (18.1–60.9)	<0.001

Notes: Median value, inter-quartile range (IR) of the prescribing indicators by type of care facility for 2003 and 2006 are presented. Results of analyses of changes between 2003 and 2006 and differences in the indicators between primary care and other care facility undertaken by Mann–Whitney test, p< 0.01 is considered statistically significant.

When restricting the analysis of differences between primary care and other care facilities to workplaces with a public employer only, the presented results above remained with an amendment of a statistically higher adherence for coxibes in the year 2003 among primary care workplaces compared with other care facilities.

In September 2004 a sharp drop in the coxibes was initiated across all workplace categories, both by employer and by type of care (data not shown). The indicator dropped from 20–25% in August 2004 to 5% or below in May 2005.

Discussion

We found that adherence increased to three of the prescribing indicators after the introduction of decentralized drug budgets. We also found differences in adherence to three of the prescribing indicators between physicians working in different settings. The results indicated that employer and type of care facility was associated with physicians’ adherence to guidelines on workplace level. Workplaces with public employer and primary care facilities showed greater adherence to two of the prescribing indicators than private sector workplaces and other care facilities. Change over the study period was larger in primary care than in other care facilities. However, there was no difference in change between workplaces with public and private employers. The changes in the prescribing indicator coxibes corresponded in time with the safety concerns and withdrawal of rofecoxib in Sweden.

Västra Götaland Region has focused on information efforts, including information on recommended therapy, costs, and cost-effectiveness, in recent years, especially focused towards primary care. This may be one possible explanation for the identified differences. Since 2003, budgets have been partly decentralized in public healthcare in Sweden [16], which might have contributed to awareness of limited budgetary resources. This was supported by a study that reported higher cost awareness among physicians who had experienced decentralized drug budgets than among those who had not [17]. This is in line with our findings that indicate higher adherence among publicly owned workplaces with budget responsibility compared with privately owned care facilities that did not have any economic responsibility for prescribed drugs.

The differences in adherence to recommendations between public and private care as well as between primary care and other care facilities corresponded well with previous results [12], [17]. The prescribing indicators used in this study reflect cost consciousness rather than quality of care. The finding that adherence of several indicators was lower for workplaces with a private employer corresponds well with previous findings that private care units prescribe more expensive pharmaceuticals [12]. Furthermore, a previous study showed that private practitioners were more reluctant to embrace several parts of the pharmaceutical benefits reform in 2002 than physicians employed by the county council [3]. In the present study, the category other care facilities mainly constituted hospital-based care facilities. Hospital physicians are often more specialized than GPs, which might make it difficult for employers to disseminate recommendations that often are oriented towards common drugs more frequently occurring in primary care than in more specialized areas. This might be an explanation for lower adherence to some of the prescribing indicators among other care facilities than among primary care workplaces. Furthermore, it is likely that the patient groups differ between primary care and more specialized care. Patients seeking hospital care might have different treatment needs that are not met by the treatment formularies. Jansson and Anell reported that cost awareness was higher among GPs than among specialists, which is in line with our finding [17]. Differences in prescribing patterns between different specialties have also been reported by others [18].

Strengths and limitations of the study

Sales data analysed in the present study encompassed all purchased prescribed pharmaceuticals in Sweden's second largest county council, Västra Götaland Region, thus providing a reliable picture of prescribing in this area.

Potential reasons for bias in this study include that workplaces were classified according to their status in 2006. Some workplaces that were classified as private practices might have been publicly owned in 2003, and vice versa. Another issue that was probably more widespread is that physicians often change workplace – thus the same physician might have prescribed in several units. However, this would dilute rather than enhance effects. When a prescription is dispensed both the name and the workplace of the prescriber is recorded. However, the name cannot be retrieved whereas the workplace codes can be analysed only with permission from the manager of the workplace.

The data used only encompassed sales of dispensed drugs, thus excluding hospital sales. This might have affected the picture for workplaces outside primary care if the sales pattern differed for pharmaceuticals ordered by the care unit and distributed directly to patients. However, it is unlikely to create a problem for the drug groups studied here as the major distribution path for these is by prescription. The indicator for coxibes was difficult to interpret due to the withdrawal of rofecoxib in September 2005. After this, coxibes were recommended only for limited use.

Conclusion

We found that adherence improved to three of the prescribing indicators after the introduction of decentralized drug budgets. Furthermore, there were differences in adherence to the studied prescribing indicators between different categories of employer and by type of care facility. For three indicators, adherence was higher among public employer and primary care workplaces. There were differences in change over the study period between the different categories of care facility but not between different employers. Altogether, this indicates that the introduction of decentralized drug budgets can affect prescribing patterns.

Ethics

Since prescribers and patients cannot be identified in the aggregated sales data, ethical approval was not needed for this study. Care has been taken in data processing to ensure anonymity of individual workplaces.