Journal of Public Health Advance Access published January 18, 2015 Journal of Public Health | pp. 1– 9 | doi:10.1093/pubmed/fdu116
Variation in the pharmaceutical costs of New Zealand general practices: a national database linkage study Andrew M. Tomlin1, Todd D. Gillies1, Murray W. Tilyard1,2, Susan M. Dovey2 1
BPACNZ, Dunedin 9016, New Zealand Department of General Practice and Rural Health, Dunedin School of Medicine, Dunedin 9013, New Zealand Address correspondence to Susan M. Dovey, E-mail: [email protected]
2
Background Variation in prescription costs between general practices and within practices over time is poorly understood. Methods From New Zealand’s national health data collections, we extracted dispensed medicines data for 1045 general practices in 2011 and 917 practices continuously existing 2008–11. Using indirect standardization to account for patient demographics and morbidity, a standardized prescribing cost ratio (SPR: the ratio of actual : expected prescription costs) was calculated for each practice in each year. Case studies of three outlier clinics explored reasons for their status. Results SPRs ranged from 0.53 to 2.28 (median ¼ 0.98). Of 469 practices with higher than expected costs (SPR . 1.0) in 2011, 204 (43.5%) had a single medicine or therapeutic drug class accounting for .15% of total costs. Case studies contrasted practices with overall pharmaceutical expenditure influenced strongly by a few patients needing high-cost medicines, more patients using medicines in one high-cost therapeutic drug class (antiretrovirals), and high medicine use across all therapeutic drug classes. Conclusions Routine data collections can measure inter-practice variation in prescription costs, adjusted for differences in the demography and morbidity profile of each practice’s patients. Small groups of patients using high-cost medicines influence general practices’ expenditure on pharmaceuticals. Keywords data collection, general practice, health expenditures, New Zealand
Introduction Variation in healthcare treatments between general practices may be a marker for inequality and reduced care quality in at least some practices, yet the patient-centred value of general practice1 suggests that variation may also be an appropriate consequence of delivering the right care to the right patient at the right time. In developed countries, primary care is where most health care is provided to most people most of the time,2 and in New Zealand 80% of publicly funded pharmaceutical expenditure is generated outside hospitals.3 Internationally, health system costs of prescription medicines are increasing.4 – 6 This is often attributed to aging populations, increasing prevalence of long-term conditions and efforts to satisfy patient demands.7 – 9 Better understanding of variation between general practices in the medicines use of their patients may provide insights into ways to reduce
variation between practices and increase care quality as well as predict and control medicines-related expenditure. Quality prescribing is not necessarily low-cost prescribing: appropriate prescription of medicines may help patients avoid other more expensive treatments, including hospital care.10 No association has been found between prescribing costs and achievement of Quality and Outcome measures in the UK, suggesting that lower prescribing rates may not compromise healthcare quality.11 However, in discrete areas of care such as cholesterol management, there is some evidence of a
Andrew M. Tomlin, Research Consultant Todd D. Gillies, Senior Data Analyst Murray W. Tilyard, Professor,Chief Executive Susan M. Dovey, Professor
# The Author 2015. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: [email protected].
1
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
A B S T R AC T
2
J O U R NA L O F P U B L IC H E A LT H
Methods Data sources
We analysed data from two national data collections administered by the New Zealand Ministry of Health: the Pharmaceutical Collection, containing records of all publicly subsidized medicines dispensed in community pharmacies, and the Primary Health Organisation (PHO) Enrolment Collection containing records of patients registered in New Zealand general practices. Patient registration affects practices capitation funding so although registration is voluntary, most New Zealanders (4.19 million (95.3%) of an estimated total population of 4.39 million22) are part of a PHO and registered in a general practice.23 Every patient has a unique National Health Index (NHI) code recorded in the national healthcare data sets. This code, in an encrypted form, was used to link practice register and medicines data. Data elements
Patient demographic data from practice registers included date of birth, sex, ethnic group (European, Maori, Pacific People, Asian and Other) and residence deprivation quintile derived from the NZDep2006 census-based index of deprivation.24 Dispensing data included the NHI code, date of dispensing, generic name of chemical dispensed, formulation, quantity prescribed, dosage information and the cost of the dispensed drug, exclusive of taxes.
Medicines were categorized according to the New Zealand Pharmaceutical Schedule, within therapeutic groups of the 12 major therapeutic drug groups (alimentary tract and metabolism, blood and blood-forming organs, cardiovascular system, dermatologicals, genito-urinary system, hormone preparations, infections, musculo-skeletal system, nervous system, oncology agents and immunosuppressants, respiratory system and allergies and sensory organs).25 These groups provided proxy indicators of patient morbidity. We excluded medicines not included in the 12 major therapeutic groups, including extemporaneously compounded preparations, galenicals and special foods, as total expenditure on these products was small (2.3% of total costs) and they are prescribed for a wide range of patient conditions. We collated dispensing data for the 2011 calendar year for patients registered to 1045 of 1074 general practices on the PHO enrolment registers. The 29 excluded practices (1.0% of all registered patients) were atypical, because they had ,200 registered patients (12 practices), an almost exclusively male population (1 practice), a nursing home focus (3 practices), or were university health centres with mainly young registered patients (13 practices). To examine changes in prescribing costs within practices over time, we analysed dispensing data for each calendar year from 2008 to 2011 for the 917 practices (85.4% of practices existing in 2011), providing healthcare continuously throughout the 4-year period. The 128 practices excluded from this analysis had closed or changed their name during this time.
Statistical analysis
To assess variation in costs, we first calculated the mean cost and number of dispensed medicines per patient by age group, sex, ethnic group and deprivation quintile and in each of the 12 therapeutic drug groups. We also calculated the proportion of patients in each practice within each of these groups. To compare practices’ prescribing costs, we employed indirect standardization26 with the total New Zealand registered patient population as the standard population. Actual costs were derived from medicine type, formulation strength and quantity dispensed. Practice prescribing costs were calculated by summing costs for every dispensing attributed to each practice and then adjusted to account for differences in the age, sex, ethnicity, social deprivation and morbidity profile (using the medicines therapeutic group proxy) of practices’ registered patient populations. The result was a standardized prescribing ratio (SPR) for each practice of the actual cost of dispensed medicines to the expected cost based on cost rates in the standard population. A ratio of 1.0 indicated that the total expenditure for a practice was equal to the national
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
relationship between prescribing costs and care quality.12 Most prescribers consider both total cost and cost to patients when prescribing, secondary only to clinical factors, but their knowledge of medicine prices is poor.13 – 18 Previous studies have found that as well as variation in patients’ needs, doctor factors such as age, previous experience, time constraints and reliance on drug company information may lead to different prescribing choices.15,19 – 21 We aimed in this study to investigate variation, and reasons for variation, in medicine costs across all general practices in New Zealand. With a view to transparency and to enable replication, we used routinely collected data to test whether useful predictive models of pharmaceutical expenditure could be made with existing accessible data sources. As well as measuring variation between general practices and within practices over time, we explore outlier practices generating higher and lower medicines costs to better understand whether changing the medicines expenditure of outlier general practices is likely to be practically possible or desirable.
VA RI AT IO N I N T H E P H A R MACE U TI CA L CO STS O F N E W ZE A L A ND G E NE RA L P RAC TI C ES
Results A total of 4 145 415 patients were registered at 1045 practices in 2011 and were included in this study (94.5% of the New Zealand population). In 2011, 61 625 737 prescription items were dispensed to 3 055 125 patients (73.7%) at a total cost of $815 958 478. Variation in prescribing rates and practice populations by age, sex, ethnicity, deprivation and therapeutic group
There were considerable differences in prescribing costs per patient across age, ethnicity, deprivation and therapeutic groups, but not by sex (Table 1). The mean cost per patient was $429 for patients aged .55 and $120 for patients aged ,55. The average cost for patients of European ethnicity was 40% higher than for patients in the other major ethnic groups, and patient costs in the three highest deprivation quintiles were 10% higher than for patients in the least deprived areas. Only 1.2% of patients used oncology agents or immunosuppressants, yet this therapeutic group accounted for 14.3% of total pharmaceuticals expenditure. Although overall costs were similar for males and females, there were large differences in costs and use of medicines within therapeutic groups. For example, 17.3% of females and 1.5% of males used genito-urinary medicines, but mean cost per patient for these items was $27 for females and $64 for males. There was substantial variation among practices in the demographic profiles of their patients. The proportion of registered patients aged .65 ranged from 5 to 25%. In 10% of practices, .35% of patients were Ma¯ori and 93 practices had .50% of their patients living in areas categorized as being the most deprived. Variation between practices in the medicines use within therapeutic groups was also notable with the proportion of patients treated for infections and nervous system disorders ranging from ,20 to .60%. Medicines in these two therapeutic groups were also the most
used by New Zealanders. Figure 1 shows practice variation in proportion of patients using these and medicines in other therapeutic groups in 2011.
Practice variation in expected prescribing costs
The funnel plot in Fig. 2 shows the distribution of SPRs for 1045 practices in relation to total pharmaceutical expenditure expected by the demographic and morbidity profile of each practice. Fifteen practices were definite outliers (SPR outside the 3 SD control limit), and another 33 practices were possible outliers (SPR outside the 2 SD control limit). Of the 1045 practices, 469 (44.9%) had higher costs than expected. As the expected expenditure of each practice was related to both its patient list size and the demography and morbidity of its patients, the range of SPRs in smaller practices was greater than in larger practices. In small practices, a few patients with high costs influenced total practice expenditure more than in larger practices with greater total expenditure. Table 2 lists high-cost medicines and therapeutic groups and their contribution to the higher than expected costs of practices with an SPR . 1.0 in 2011. Practices with either a single medicine or medicine class accounting for .15% of total practice prescribing costs are also shown. A single medicine accounted for .15% of total practice expenditure at 74 practices despite the median practice cost for these drugs being 0%. Prominent among these medicines were chemotherapeutic agents and immunosuppressants.
Variation in expected prescribing costs within practices 2008 –11
The number of patients registered at the 917 general practices providing services throughout 2008–11 increased from 3 530 327 in 2008 to 3 725 850 in 2011 (89.9% of patients at all study practices in 2011). Over this time, pharmaceutical expenditure for these patients increased from $652 368 287 to $729 947 319 (89.4% of total prescribing costs at all study practices in 2011). Figure 3 shows the distribution of SPRs for individual practices over the 4-year period. For ease of presentation, data are illustrated for 118 practices within four DHB districts. Higher cost practices tended to have higher costs in all years, with lower cost practices remaining low cost relative to other practices. Some practices displayed considerable variation in costs with time. For example, the SPR of one practice fell from 1.45 in 2008 to 0.98 in 2009 due mainly to a single patient decreasing imiglucerase use. Figure 3 also indicates that some DHBs have relatively more practices with high prescription costs.
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
average practice based on differential prescribing costs within patient demographic and drug therapeutic groups. We calculated SPRs for 1045 practices in 2011 and, to examine variation in practice prescribing costs practices over time, 917 practices in each year from 2008 to 2011. Funnel plots of practice ratios and expected expenditure identified outlier practices, using control limits set at 2 and 3 SD from the mean expenditure of all practices, to adjust for bias due to small practice size and expenditure.27,28 We then used case studies to explore reasons for medicine costs in exemplar high and low outlier practices.
3
4
J O U R NA L O F P U B L IC H E A LT H
Table 1 Pharmaceutical use and expenditure by age, sex, ethnicity, deprivation and drug therapeutic group in 2011 Registered
Rx items (n)
Patients, n (%)
NZ total
Rx cost ($) Per patient
‘000 000
NZ total
Practice variation Per patient
‘000 000
Median % of patients (1st to 9th decile)
Age 302 023 (7.3)
2.072
6.86
9.255
30.64
6.8 (4.3 –9.8)
5– 14
582 616 (14.1)
2.001
3.44
20.325
34.89
13.5 (8.9 –18.5)
15– 24
550 154 (13.3)
2.616
4.76
37.370
67.93
12.9 (9.7 –17.2)
25– 34
499 576 (12.1)
3.535
7.08
55.720
111.53
10.9 (7.8 –16.6)
35– 44
589 025 (14.2)
5.912
10.04
103.199
175.20
13.6 (11.1 –17.0)
45– 54
593 552 (14.3)
8.379
14.12
148.676
250.48
14.4 (11.7 –16.9)
55– 64
472 424 (11.4)
10.064
21.30
169.412
358.60
11.8 (7.7 –15.8)
65– 74
307 584 (7.4)
10.705
34.80
149.213
485.11
7.7 (3.9 –11.8)
75– 84
180 378 (4.4)
10.756
59.63
93.805
520.05
4.4 (1.6 –7.9)
68 083 (1.6)
5.585
82.03
28.983
425.70
1.5 (0.3 –3.5)
85 and over Sex Female
2 147 493 (51.8)
35.363
16.47
429.452
199.98
51.0 (46.5 –56.7)
Male
1 997 922 (48.2)
26.263
13.14
386.506
193.45
49.0 (43.3 –53.5)
Ethnicity Asian European
349 439 (8.4)
3.137
8.98
43.748
125.19
3.1 (0.7 –19.7)
2 791 041 (67.3)
46.160
16.54
622.080
222.88
78.1 (18.2 –90.8)
Maori
571 439 (13.8)
7.477
13.09
90.387
158.17
8.4 (2.5 –35.6)
Pacific
296 573 (7.2)
3.564
12.02
39.113
131.88
1.6 (0.4 –15.4)
Other
10 727 (2.6)
0.928
8.61
15.937
147.94
1.3 (0.3 –5.7)
Deprivation quintile 1
834 639 (19.9)
9.327
11.17
151.638
181.68
16.4 (2.0 –37.0)
2
774 403 (18.5)
10.038
12.96
146.977
189.79
18.0 (5.4 –29.0)
3
748 726 (17.9)
11.152
14.89
151.578
202.45
18.0 (7.6 –25.7)
4
748 623 (17.9)
12.769
17.06
157.157
209.93
18.2 (5.8 –28.0)
5
757 487 (18.1)
13.567
17.91
151.235
199.65
12.1 (1.3 –46.6)
Therapeutic groupa Alimentary tract and metabolism
1 070 809 (25.8)
9.27
8.65
90.19
84.23
25.7 (20.0 –33.4)
Blood and blood-forming organs
764 398 (18.4)
7.21
9.44
96.53
126.29
18.6 (13.0 –25.5)
Cardiovascular
732 162 (17.7)
10.34
14.12
77.67
106.09
18.2 (12.0 –25.6)
1 023 346 (24.7)
2.74
2.68
18.82
18.39
24.4 (19.1 –29.8)
Genito-urinary
401 789 (9.7)
1.02
2.53
11.85
29.50
9.5 (6.4 –12.1)
Hormone preparations
487 378 (11.8)
1.89
3.89
21.08
43.25
11.3 (8.1 –16.0)
Dermatologicals
Infections Musculo-skeletal Nervous system
1 795 097 (43.3)
4.24
2.36
53.61
29.87
43.0 (34.3 –51.7)
969 126 (23.4)
2.91
3.00
71.36
73.63
23.0 (18.5 –27.6)
1 746 912 (42.1)
41.8 (33.8 –49.1)
15.79
9.04
167.04
95.62
49 011 (1.2)
0.44
9.01
116.98
2386.75
Respiratory system and allergies
971 816 (23.4)
4.69
4.83
80.36
82.70
22.6 (17.6 –28.8)
Sensory organs
423 875 (10.2)
1.09
2.57
10.46
24.67
10.1 (7.5 –12.8)
Oncology agents and immunosuppressants
Case study practices
At Practice A (SPR ¼ 1.31), a single patient was treated with one medicine that accounted for 20.4% of the practice’s total prescribing costs. To examine the influence of this single drug
1.2 (0.7 –1.7)
on the SPR, we repeated the standardization of prescribing costs excluding it. Following this, Practice A’s adjusted SPR was 1.05, but four more patients accounted for 20.5% of total prescription costs: one taking another medicine (11.8% of
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
0– 4
VA RI AT IO N I N T H E P H A R MACE U TI CA L CO STS O F N E W ZE A L A ND G E NE RA L P RAC TI C ES
total cost) and three patients using a third drug (8.7%). The patients taking the two most costly drugs accounted for 33.2% of the practice’s total prescription costs. At Practice B (SPR ¼ 2.28), 91 patients were treated with antiretrovirals costing $13 287 per patient. The median number of patients using antiretrovirals at all 1045 practices 70
50 40 30 20 10 0 ATM BLD CVD DRM GNU HRM INF MSK NRV OAI RSP SNS
Fig. 1 Practice variation in proportion of patients by therapeutic group in 2011. Box, inter-quartile range. 95% of practices lie within the whiskers.
was 0. Antiretrovirals contributed 45% of Practice B’s total medicines expenditure. Practice C (SPR ¼ 1.44) was one of 265 practices with an SPR . 1.0 but no single medicine or medicines class accounted for .15% of its total prescribing costs. Its cost per patient was higher than the national average in 9 out of 12 major therapeutic groups. Overall, its cost per patient was 69.9% higher than the national average. Practice D (SPR ¼ 0.53) had the lowest SPR of all practices. Prescribing costs per patient were 50.8% lower than the national average and the number of medicines per patient 20.9% lower. Costs were lower than expected in 11 of the 12 therapeutic groups. The total number of medicines dispensed was 14.9% lower than expected. At Practice E, the total number of medicines dispensed was 7.1% higher than expected (SPR ¼ 0.56), but prescription costs were 60.4% of expected. Prescription costs per patient ($21.24) were 71.9% lower than the national average ($75.55) and costs per item were 56.0% lower. Practices with lower than expected costs were characterized by low prescription costs per patient across most therapeutic groups and usually a low prescription rate per patient.
ATM, alimentary tract and metabolism; BLD, blood and blood-forming organs; CVD, cardiovascular system; DRM, dermatologicals; GNU,
Discussion
genito-urinary system; HRM, hormone preparations; INF, infections; MSK, musculo-skeletal system; NRV, nervous system; OAI, oncology agents and immunosuppressants; RSP, respiratory system and allergies; SNS, sensory organs.
Main finding of this study
Our findings indicate that there is substantial variation in pharmaceutical expenditure per patient between New Zealand general
2.5
Standardized prescribing ratio
2.0
1.5
1.0
0.5
0.0 0
1 000 000
2 000 000
3 000 000
4 000 000
5 000 000
Expected cost of prescribing ($) Fig. 2 Standardized prescribing ratios for 1045 New Zealand general practices by expected cost of prescribing 2011. Dots, practices. Dashed curves: 2 SD, solid curves; 3 SD, bold dots are ‘definite’ (.3 SD) or ‘possible’ outliers (2 –3 SD).
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
% of registered patients
60
5
6
J O U R NA L O F P U B L IC H E A LT H
Table 2 Influence of expensive medications on the total cost of prescribing in 2011 in practices with a standardized prescribing ratio (SPR) . 1.0 All New Zealand general practices
Practices with SPR . 1 and 15% of total cost in one TG2 or medicine
n ¼ 1045 $ per patient
n ¼ 204 No. of patients
Median (% of
No. of practices
total practice $) Antibacterials Colistin sulphomethate
Mean SPR
total practice $)
9.07
1 702 824
1.78
1
22.9
1.31
10 324.42
43
0.00
1
20.4
1.31
515.91
78 772
3.97
28
15.2 –41.8
1.24
1 479.57
10 798
1.60
3
15.2 –19.6
1.18
Risperidone
721.15
17 529
0.88
5
15.8 –32.9
1.25
Antiretrovirals
11 188.08
1305
0.00
15
16.3 –44.8
1.63
9018.50
5309
4.70
34
15.2 –34.3
1.15
Adalimumab
17 214.52
2264
3.62
20
15.1 –30.5
1.12
Etanercept
12 525.68
513
0.00
1
33.6
1.32
238.25
52 928
0.75
9
15.0 –36.1
1.17
6088.67
244
0.00
2
22.0 –24.2
1.06
Olanzapine
Antirheumatoid agents
Antivirals Adefovir dipivoxil Tenofovir disoproxil fumarate
5167.52
1286
0.00
1
1854.14
28 499
4.49
51
Bortezomib
18294.00
255
0.00
Dasatinib
46 186.64
59
Erlotinib hydrochloride
13 009.09
Imatinib mesylate Sunitinib Temozolomide
Chemotherapeutic agents
Endocrine therapy Octreotide Fluids and electrolytes Total parenteral nutrition Immune modulators Pegylated interferon alpha-2a Immunosuppressants
17.1
1.04
15.1 –46.3
1.17
1
15.7
1.15
0.00
2
15.1 –21.6
1.12
186
0.00
1
22.6
1.20
46 124.41
381
0.00
16
28 880.80
121
0.00
1
12 871.00
215
0.00
716.03
12 766
0.27
21 891.14
330
15.8 –35.8
1.20
21.9
1.21
1
16.7
1.32
5
15.9 –25.5
1.26
0.00
5
15.5 –25.4
1.26
38.98
104 228
0.00
3
16.2 –56.3
1.49
48 508.32
46
0.00
3
16.1 –56.2
1.49
10 093.84
751
0.00
3
17.0 –19.4
1.14
10 786.81
677
0.00
3
17.0 –19.4
1.14
5036.19
10 920
5.10
51
15.1 –55.6
1.19
Rituximab
16 970.49
844
0.00
2
21.1 –40.0
1.05
Trastuzumab
36 817.78
695
0.00
29
15.1 –47.7
1.22
134.14
480 549
7.06
21
15.2 –32.6
1.10
317.39
165 147
5.60
10
15.1 –27.6
1.10
81 405.00
16
0.00
5
18.3 –45.7
1.32
81 405.00
16
0.00
5
18.3 –45.7
1.32
6420.12
509
0.00
1
19.1
1.32
40 182.79
14
0.00
1
15.2
1.23
Lipid-modifying agents Atorvastatina Metabolic disorder agents Imiglucerase Vasodilators Iloprost a
Atorvastatin became available generically in mid-2011.
TG2, therapeutic group level 2; SPR, standardized prescribing ratio.
practices. This is true for both actual and standardized expenditure resulting from the adjustment of costs to account for practices’ differences in patient demographics and morbidity. Our results also indicate that practices’ prescribing costs may vary considerably over time, an important consideration when deciding which practices may benefit from interventions to alter prescribing.
What is already known
Among the factors associated with varying medicines use in general practice, patients’ age has most frequently been factored into analyses of variation,29 but age alone does not account for enough variation to set robust budgets for medicines expenditure.30 Models including age and sex account
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
Antipsychotics
Range (% of
VA RI AT IO N I N T H E P H A R MACE U TI CA L CO STS O F N E W ZE A L A ND G E NE RA L P RAC TI C ES
DHB A (24 practices)
DHB B (42)
DHB C (24)
7
DHB D (28)
1.5 1.4
1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5 0.4
Fig. 3 Variation in actual to expected pharmaceutical costs 2008 –11 for practices in four District Health Board areas. Markers indicate the practice’s mean SPR 2008 –11; lines indicate the SPR range.
for around 25% of variation,7,31,32 additional demographic factors up to 51%33 – 35 and models that include morbidity account for more variability (up to 80%).36 – 38 The use of medicines within therapeutic groups provides an indicator of patient morbidity and treatment needs within a practice and has been included in previous models of prescribing activity in general practice.30,39 Variation in practices’ mortality rates, but not prescribing costs, has been investigated by standardizing practices to account for their patients’ morbidity.40 Doctor characteristics and morbidity profiles in regional databases have been used to explore drivers of variation in general practice prescribing,41 and a network of computerized general practice prescribing has shown longitudinal changes in prescribing behaviour.36 What this study adds
We could find no previous investigations of variation in medicines costs per practice involving national data sets. This study builds on previous research showing the importance of including patient morbidity in models of practice variation, adding to previous models by including patients’ ethnicity and deprivation indicator, both associated with health care inequalities in New Zealand. We used all these variables, comprehensively recorded in New Zealand’s national health care data sets, to identify the standardized distribution of New Zealand general practices’ pharmaceuticals expenditure. We showed that analysis of routine national data collections provides a starting point for identifying practices with
potentially modifiable prescribing costs through interventions such as academic detailing.42 This extends previous work by exploring the practical implications of identifying outlier general practices. We confirmed that high-cost outlier practices are not necessarily generating modifiable pharmaceutical expenditure. Practices with more morbid patients, especially in chronic conditions, are more likely to have higher prescription costs. High costs per capita in smaller practices were often due to individual patients using very high cost but necessary medicines. Unusually large numbers of patients requiring expensive treatments also contribute to higher costs than expected. Practices with consistently higher than expected costs across most patient demographic and therapeutic drug groups may be more likely candidates for interventions aimed at reducing pharmaceutical expenditure. Limitations of this study
Our analysis included all pharmaceuticals dispensed to all patients registered to each practice as recorded in the national Pharmaceutical Collection. It is a limitation of this study that many of these medicines may not have been prescribed by doctors at practices registering patients. Some practices, particularly those in popular holiday areas, provide care for many casual patients whose prescription costs are attributed to practices they are enrolled in rather than the prescribing practices. Therefore, this analysis does not precisely quantify the total prescription costs generated by each practice.
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
Standardized prescribing ratio
1.3
8
J O U R NA L O F P U B L IC H E A LT H
We were able to analyze variation in prescription costs from 2008 to 2011 for only 917 of the 1045 practices. This limitation reflects the changing nature of general practices as businesses that may close, merge, transfer patients or move location. We included merged practices that had kept the same name throughout the study period. In many cases, mergers would have resulted in changes in the demographic profile of such practices, influencing the relative prescription ratios across the 4-year period.
This study indicates that there is considerable variation in prescription costs per patient among New Zealand general practices after adjusting for their patients’ demographic and morbidity profiles. The methodology identified outlier practices with higher expenditure than expected that may be targeted in interventions to reduce prescribing costs. While indirect standardization should not be used in creating practice league tables as bias is introduced when patient denominators differ substantially, it may be used to indicate how far practices are from the expected cost of prescribing.43 In our analysis, the expected cost was determined by the national average prescribing cost rates observed within patient demographic and major therapeutic drug groups. Each practice was thus compared with the reference population representing all registered general practice patients in New Zealand. General practices may have higher than expected prescription costs, even after adjusting for demographic and morbidity factors, for a variety of reasons. Patients with special needs may require expensive pharmaceutical treatment, medicines with little or no therapeutic benefit may be inappropriately prescribed, or general practitioners may view their role squarely in terms of service provision, with prescribing as part of that service. Low-cost prescribers, by contrast, are more comfortable with a wider view of their role in society that may require limiting patients’ pharmaceuticals use.44 Also it is possible that lower cost practices may be providing poorer quality care, particularly in the identification and management of patients with chronic conditions. Interventions aimed at controlling prescribing costs need to be informed by the reasons that practices have outlier costs as well as by robust evidence of the expenditures they generate.
This research was conducted as a component of the usual employment of the authors by Bpacnz and the University of Otago. Bpacnz is a registered charity operating under contracts with PHARMAC and District Health Boards Shared Services (DHBSS). Bpacnz is co-owned by the University of Otago, ProCare Health Limited, South Link Health, General Practice New Zealand and Pegasus Health.
References 1 Toop L. Primary care core values: patient centred primary care. Br Med J 1998;316:1882 – 3. 2 Green LA, Fryer GE, Yawn BP et al. The ecology of medical care revisited. N Eng J Med 2001;344:2021 – 5. 3 Tordoff J, Norris P, Reith D. “Price Management” and its impact on hospital pharmaceutical expenditure and the availability of medicines in New Zealand hospitals. Value Health 2008;11(7):1214 – 26. 4 Moon J, Bogle R. How to lose a billion pounds. Int J Clin Pract 2007;61(1):2 – 3. 5 Minhas R. Waste not, want not: free money, moral hazard and valuebased prescribing. Int J Clin Pract 2011;65(3):240 – 2. 6 Sermet C, Andrieu V, Godman B et al. Ongoing pharmaceutical reforms in France: implications for key stakeholder groups. Appl Health Econ Health Policy 2010;8(1):7 – 24. 7 Favato G, Mariani P, Mills R et al. ASSET (Age/Sex Standardised Estimates of Treatment): a research model to improve the governance of prescribing funds in Italy. PLoS ONE 2007;2(7):e592. 8 Hassell K, Atella V, Schafeheutle E et al. Cost to the patient or cost to the healthcare system? Which one matters the most for GP prescribing decisions? A UK-Italy comparison. Eur J Public Health 2003;13:18–23. 9 Carthy P, Harvey I, Brawn R et al. A study of factors associated with cost and variation in prescribing among GPs. Fam Pract 2000;17(1):36–41. 10 Zhang Y, Soumerai S. Do newer prescription drugs pay for themselves? A reassessment of the evidence. Health Aff 2007;26(3):880– 6. 11 Fleetcroft R, Cookson R, Steel N et al. Correlation between prescribing quality and pharmaceutical costs in English primary care: national cross-sectional analysis. Br J Gen Pract 2011;61(590):e556– 64. 12 Fleetcroft R, Schofield P, Duerden M et al. Achievement of cholesterol targets and prescribing of higher-cost statins: a cross-sectional study in general practice. Br J Gen Pract 2012;62(605):815– 20. 13 Allan G, Lexchin J, Wiebe N. Physician awareness of drug cost: a systematic review. PLoS Med 2007;4(9):1486 – 96. 14 Andersson K, Carlsten A, Hedenrud T. Prescribing behaviour after the introduction of decentralized drug budgets: is there an association with employer and type of care facility? Scand J Prim Health Care 2009;27:117–22.
Acknowledgements
15 Buusman A, Morten A, Merrild C et al. Factors influencing GPs’ choice between drugs in a therapeutic drug group. A Qualitative Study. Scand J Prim Health Care 2007;25(4):208 – 13.
We thank PHARMAC for providing permission to use the Pharmaceutical Collection.
16 Mason A, Drummond M, Towse A et al. Influencing prescribing in English primary care: the views of primary care organisations. J Health Serv Res Policy 2004;9(3):153 – 8.
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
Conclusions
Funding
VA RI AT IO N I N T H E P H A R MACE U TI CA L CO STS O F N E W ZE A L A ND G E NE RA L P RAC TI C ES
9
31 Majeed A, Head S. Controversies in primary care. Setting prescribing budgets in general practice. Capitation based prescribing budgets will not work. Br Med 1998;316(7133):748– 50.
18 Reichert S, Simon T, Halm E. Physicians’ attitudes about prescribing and knowledge of the costs of common medications. Arch Intern Med 2000;160(18):2799– 803.
32 Greenhalgh T. Controversies in primary care. Setting prescribing budgets in general practice. Effective prescribing at practice level should be identified and rewarded. Br Med 1998;316(7133):750– 3.
19 Weekes L, Mackson J, Fitzgerald M et al. National prescribing service: creating an implementation arm for national medicines policy. Br J Clin Pharmacol 2005;59(1):112– 6.
33 Whynes D, Baines D, Tolley K. Explaining variations in general practice prescribing costs per ASTRO-PU (age,sex, and temporary resident originated prescribing unit). Br Med 1996;312:488 – 9.
20 Denig P, Haaijer-Ruskamp F, Wesseling H et al. Towards understanding treatment preferences of hospital physicians. Soc Sci Med 1993; 36(7):915 – 24.
34 Forster D, Frost C. Use of regression analysis to explain the variation in prescribing rates and costs between family practitioner committees. Br J Gen Pract 1991;41(343):67– 71.
21 Watkins C, Harvey I, Carthy P et al. Attitudes and behaviour of general practitioners and their prescribing costs: a national cross sectional survey. Qual Saf Health Care 2003;12:29– 34.
35 Haynes K, Bilker WB, TenHave TR et al. Temporal and within practice variability in the health improvement network. Pharmacoepidem Drug Safe 2011;20:948 – 55.
22 Statistics New Zealand. National Population Estimates: December 2010 Quarter. 2012. http://www.stats.govt.nz/browse_for_stats/population/ estimates_and_projections/ NationalPopulationEstimates_HOTPDec10qtr.aspx (28 January 2013, date last accessed).
36 Omar R, O’Sullivan C, Petersen I et al. A model based on age, sex, and morbidity to explain variation in UK general practice prescribing: cohort study. Br Med 2008;337:a238.
23 Ministry of Health. Visiting a Doctor. 2013. http://www.health.govt.nz/ yourhealth-topics/health-care-services/visiting-doctor (28 January 2013, date last accessed). 24 Salmond C, Crampton P, Atkinson J. NZDep2006 Index of Deprivation. 2007. 25 Pharmaceutical Management Agency (PHARMAC). New Zealand Pharmaceutical Schedule: December 2012. PHARMAC, 2012. www. pharmac.govt.nz/2012/11/27/Sched.pdf (28 January 2013, date last accessed).
37 Rice N, Dixon P, Lloyd D et al. Derivation of a needs based capitation formula for allocating prescribing budgets to health authorities and primary care groups in England: regression analysis. Br Med 2000;320 (7230):284 – 8. 38 Morton-Jones T, Pringle M. Explaining variations in prescribing costs across England. BMJ 1993;306(1731):4. 39 Lloyd D, Harris C, Roberts D. Specific therapeutic group age-sex related prescribing units (STAR-PUs): weightings for analysing general practices’ prescribing in England. BMJ 1995;311(7011): 991 – 4.
26 Kirkwood B, Sterne J. Essential Medical Statistics, 2nd edn. Oxford: Blackwell Science, 2003.
40 Carey IM, Shah SM, Harris T et al. A new simple primary care morbidity score predicted mortality and better explains between practice variations than the Charlson index. J Clin Epidemiol 2013;66(4):436–44.
27 Morton A, Mengersen K, Rajmokan M et al. Funnel plots and risk-adjusted count data adverse events. A limitation of indirect standardisation. J Hosp Infect 2011;78(4):260– 3.
41 Orzella L, Chini F, Rossi PG et al. Physician and patient characteristics associated with prescriptions and costs of drugs in the Lazio region of Italy. Health Policy 2010;95(2 – 3):236 – 44.
28 Spiegelhalter D. Funnel plots for institutional comparison. Qual Saf Health Care 2002;11(4):391– 1.
42 Rognstad S, Brekke M, Fetveit A et al. Prescription peer academic detailing to reduce inappropriate prescribing for older patients: a cluster randomised controlled trial. Br J Gen Pract 2013;63(613): E554 – 62.
29 Purves I, Edwards C. Comparison of prescribing unit with index including both age and sex in assessing general practice prescribing costs. Br Med J 1993;306:496 – 8.
43 Rixon A. Performance league tables. BMJ 2002;325:177– 8.
30 Sleator D. Towards accurate prescribing analysis in general practice: accounting for the effects of practice demography. Br J Gen Pract 1993;43(368):102– 6.
44 Jaye C, Tilyard M. A qualitative comparative investigation of variation in general practitioners’ prescribing patterns. Br J Gen Pract 2002;52: 381 – 6.
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
17 Prosser H, Walley T. A qualitative study of GPs’ and PCO stakeholders’ views on the importance and influence of cost on prescribing. Soc Sci Med 2005;60(6):1335 – 46.
Variation in the pharmaceutical costs of New Zealand general practices: a national database linkage study Andrew M. Tomlin1, Todd D. Gillies1, Murray W. Tilyard1,2, Susan M. Dovey2 1
BPACNZ, Dunedin 9016, New Zealand Department of General Practice and Rural Health, Dunedin School of Medicine, Dunedin 9013, New Zealand Address correspondence to Susan M. Dovey, E-mail: [email protected]
2
Background Variation in prescription costs between general practices and within practices over time is poorly understood. Methods From New Zealand’s national health data collections, we extracted dispensed medicines data for 1045 general practices in 2011 and 917 practices continuously existing 2008–11. Using indirect standardization to account for patient demographics and morbidity, a standardized prescribing cost ratio (SPR: the ratio of actual : expected prescription costs) was calculated for each practice in each year. Case studies of three outlier clinics explored reasons for their status. Results SPRs ranged from 0.53 to 2.28 (median ¼ 0.98). Of 469 practices with higher than expected costs (SPR . 1.0) in 2011, 204 (43.5%) had a single medicine or therapeutic drug class accounting for .15% of total costs. Case studies contrasted practices with overall pharmaceutical expenditure influenced strongly by a few patients needing high-cost medicines, more patients using medicines in one high-cost therapeutic drug class (antiretrovirals), and high medicine use across all therapeutic drug classes. Conclusions Routine data collections can measure inter-practice variation in prescription costs, adjusted for differences in the demography and morbidity profile of each practice’s patients. Small groups of patients using high-cost medicines influence general practices’ expenditure on pharmaceuticals. Keywords data collection, general practice, health expenditures, New Zealand
Introduction Variation in healthcare treatments between general practices may be a marker for inequality and reduced care quality in at least some practices, yet the patient-centred value of general practice1 suggests that variation may also be an appropriate consequence of delivering the right care to the right patient at the right time. In developed countries, primary care is where most health care is provided to most people most of the time,2 and in New Zealand 80% of publicly funded pharmaceutical expenditure is generated outside hospitals.3 Internationally, health system costs of prescription medicines are increasing.4 – 6 This is often attributed to aging populations, increasing prevalence of long-term conditions and efforts to satisfy patient demands.7 – 9 Better understanding of variation between general practices in the medicines use of their patients may provide insights into ways to reduce
variation between practices and increase care quality as well as predict and control medicines-related expenditure. Quality prescribing is not necessarily low-cost prescribing: appropriate prescription of medicines may help patients avoid other more expensive treatments, including hospital care.10 No association has been found between prescribing costs and achievement of Quality and Outcome measures in the UK, suggesting that lower prescribing rates may not compromise healthcare quality.11 However, in discrete areas of care such as cholesterol management, there is some evidence of a
Andrew M. Tomlin, Research Consultant Todd D. Gillies, Senior Data Analyst Murray W. Tilyard, Professor,Chief Executive Susan M. Dovey, Professor
# The Author 2015. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: [email protected].
1
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
A B S T R AC T
2
J O U R NA L O F P U B L IC H E A LT H
Methods Data sources
We analysed data from two national data collections administered by the New Zealand Ministry of Health: the Pharmaceutical Collection, containing records of all publicly subsidized medicines dispensed in community pharmacies, and the Primary Health Organisation (PHO) Enrolment Collection containing records of patients registered in New Zealand general practices. Patient registration affects practices capitation funding so although registration is voluntary, most New Zealanders (4.19 million (95.3%) of an estimated total population of 4.39 million22) are part of a PHO and registered in a general practice.23 Every patient has a unique National Health Index (NHI) code recorded in the national healthcare data sets. This code, in an encrypted form, was used to link practice register and medicines data. Data elements
Patient demographic data from practice registers included date of birth, sex, ethnic group (European, Maori, Pacific People, Asian and Other) and residence deprivation quintile derived from the NZDep2006 census-based index of deprivation.24 Dispensing data included the NHI code, date of dispensing, generic name of chemical dispensed, formulation, quantity prescribed, dosage information and the cost of the dispensed drug, exclusive of taxes.
Medicines were categorized according to the New Zealand Pharmaceutical Schedule, within therapeutic groups of the 12 major therapeutic drug groups (alimentary tract and metabolism, blood and blood-forming organs, cardiovascular system, dermatologicals, genito-urinary system, hormone preparations, infections, musculo-skeletal system, nervous system, oncology agents and immunosuppressants, respiratory system and allergies and sensory organs).25 These groups provided proxy indicators of patient morbidity. We excluded medicines not included in the 12 major therapeutic groups, including extemporaneously compounded preparations, galenicals and special foods, as total expenditure on these products was small (2.3% of total costs) and they are prescribed for a wide range of patient conditions. We collated dispensing data for the 2011 calendar year for patients registered to 1045 of 1074 general practices on the PHO enrolment registers. The 29 excluded practices (1.0% of all registered patients) were atypical, because they had ,200 registered patients (12 practices), an almost exclusively male population (1 practice), a nursing home focus (3 practices), or were university health centres with mainly young registered patients (13 practices). To examine changes in prescribing costs within practices over time, we analysed dispensing data for each calendar year from 2008 to 2011 for the 917 practices (85.4% of practices existing in 2011), providing healthcare continuously throughout the 4-year period. The 128 practices excluded from this analysis had closed or changed their name during this time.
Statistical analysis
To assess variation in costs, we first calculated the mean cost and number of dispensed medicines per patient by age group, sex, ethnic group and deprivation quintile and in each of the 12 therapeutic drug groups. We also calculated the proportion of patients in each practice within each of these groups. To compare practices’ prescribing costs, we employed indirect standardization26 with the total New Zealand registered patient population as the standard population. Actual costs were derived from medicine type, formulation strength and quantity dispensed. Practice prescribing costs were calculated by summing costs for every dispensing attributed to each practice and then adjusted to account for differences in the age, sex, ethnicity, social deprivation and morbidity profile (using the medicines therapeutic group proxy) of practices’ registered patient populations. The result was a standardized prescribing ratio (SPR) for each practice of the actual cost of dispensed medicines to the expected cost based on cost rates in the standard population. A ratio of 1.0 indicated that the total expenditure for a practice was equal to the national
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
relationship between prescribing costs and care quality.12 Most prescribers consider both total cost and cost to patients when prescribing, secondary only to clinical factors, but their knowledge of medicine prices is poor.13 – 18 Previous studies have found that as well as variation in patients’ needs, doctor factors such as age, previous experience, time constraints and reliance on drug company information may lead to different prescribing choices.15,19 – 21 We aimed in this study to investigate variation, and reasons for variation, in medicine costs across all general practices in New Zealand. With a view to transparency and to enable replication, we used routinely collected data to test whether useful predictive models of pharmaceutical expenditure could be made with existing accessible data sources. As well as measuring variation between general practices and within practices over time, we explore outlier practices generating higher and lower medicines costs to better understand whether changing the medicines expenditure of outlier general practices is likely to be practically possible or desirable.
VA RI AT IO N I N T H E P H A R MACE U TI CA L CO STS O F N E W ZE A L A ND G E NE RA L P RAC TI C ES
Results A total of 4 145 415 patients were registered at 1045 practices in 2011 and were included in this study (94.5% of the New Zealand population). In 2011, 61 625 737 prescription items were dispensed to 3 055 125 patients (73.7%) at a total cost of $815 958 478. Variation in prescribing rates and practice populations by age, sex, ethnicity, deprivation and therapeutic group
There were considerable differences in prescribing costs per patient across age, ethnicity, deprivation and therapeutic groups, but not by sex (Table 1). The mean cost per patient was $429 for patients aged .55 and $120 for patients aged ,55. The average cost for patients of European ethnicity was 40% higher than for patients in the other major ethnic groups, and patient costs in the three highest deprivation quintiles were 10% higher than for patients in the least deprived areas. Only 1.2% of patients used oncology agents or immunosuppressants, yet this therapeutic group accounted for 14.3% of total pharmaceuticals expenditure. Although overall costs were similar for males and females, there were large differences in costs and use of medicines within therapeutic groups. For example, 17.3% of females and 1.5% of males used genito-urinary medicines, but mean cost per patient for these items was $27 for females and $64 for males. There was substantial variation among practices in the demographic profiles of their patients. The proportion of registered patients aged .65 ranged from 5 to 25%. In 10% of practices, .35% of patients were Ma¯ori and 93 practices had .50% of their patients living in areas categorized as being the most deprived. Variation between practices in the medicines use within therapeutic groups was also notable with the proportion of patients treated for infections and nervous system disorders ranging from ,20 to .60%. Medicines in these two therapeutic groups were also the most
used by New Zealanders. Figure 1 shows practice variation in proportion of patients using these and medicines in other therapeutic groups in 2011.
Practice variation in expected prescribing costs
The funnel plot in Fig. 2 shows the distribution of SPRs for 1045 practices in relation to total pharmaceutical expenditure expected by the demographic and morbidity profile of each practice. Fifteen practices were definite outliers (SPR outside the 3 SD control limit), and another 33 practices were possible outliers (SPR outside the 2 SD control limit). Of the 1045 practices, 469 (44.9%) had higher costs than expected. As the expected expenditure of each practice was related to both its patient list size and the demography and morbidity of its patients, the range of SPRs in smaller practices was greater than in larger practices. In small practices, a few patients with high costs influenced total practice expenditure more than in larger practices with greater total expenditure. Table 2 lists high-cost medicines and therapeutic groups and their contribution to the higher than expected costs of practices with an SPR . 1.0 in 2011. Practices with either a single medicine or medicine class accounting for .15% of total practice prescribing costs are also shown. A single medicine accounted for .15% of total practice expenditure at 74 practices despite the median practice cost for these drugs being 0%. Prominent among these medicines were chemotherapeutic agents and immunosuppressants.
Variation in expected prescribing costs within practices 2008 –11
The number of patients registered at the 917 general practices providing services throughout 2008–11 increased from 3 530 327 in 2008 to 3 725 850 in 2011 (89.9% of patients at all study practices in 2011). Over this time, pharmaceutical expenditure for these patients increased from $652 368 287 to $729 947 319 (89.4% of total prescribing costs at all study practices in 2011). Figure 3 shows the distribution of SPRs for individual practices over the 4-year period. For ease of presentation, data are illustrated for 118 practices within four DHB districts. Higher cost practices tended to have higher costs in all years, with lower cost practices remaining low cost relative to other practices. Some practices displayed considerable variation in costs with time. For example, the SPR of one practice fell from 1.45 in 2008 to 0.98 in 2009 due mainly to a single patient decreasing imiglucerase use. Figure 3 also indicates that some DHBs have relatively more practices with high prescription costs.
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
average practice based on differential prescribing costs within patient demographic and drug therapeutic groups. We calculated SPRs for 1045 practices in 2011 and, to examine variation in practice prescribing costs practices over time, 917 practices in each year from 2008 to 2011. Funnel plots of practice ratios and expected expenditure identified outlier practices, using control limits set at 2 and 3 SD from the mean expenditure of all practices, to adjust for bias due to small practice size and expenditure.27,28 We then used case studies to explore reasons for medicine costs in exemplar high and low outlier practices.
3
4
J O U R NA L O F P U B L IC H E A LT H
Table 1 Pharmaceutical use and expenditure by age, sex, ethnicity, deprivation and drug therapeutic group in 2011 Registered
Rx items (n)
Patients, n (%)
NZ total
Rx cost ($) Per patient
‘000 000
NZ total
Practice variation Per patient
‘000 000
Median % of patients (1st to 9th decile)
Age 302 023 (7.3)
2.072
6.86
9.255
30.64
6.8 (4.3 –9.8)
5– 14
582 616 (14.1)
2.001
3.44
20.325
34.89
13.5 (8.9 –18.5)
15– 24
550 154 (13.3)
2.616
4.76
37.370
67.93
12.9 (9.7 –17.2)
25– 34
499 576 (12.1)
3.535
7.08
55.720
111.53
10.9 (7.8 –16.6)
35– 44
589 025 (14.2)
5.912
10.04
103.199
175.20
13.6 (11.1 –17.0)
45– 54
593 552 (14.3)
8.379
14.12
148.676
250.48
14.4 (11.7 –16.9)
55– 64
472 424 (11.4)
10.064
21.30
169.412
358.60
11.8 (7.7 –15.8)
65– 74
307 584 (7.4)
10.705
34.80
149.213
485.11
7.7 (3.9 –11.8)
75– 84
180 378 (4.4)
10.756
59.63
93.805
520.05
4.4 (1.6 –7.9)
68 083 (1.6)
5.585
82.03
28.983
425.70
1.5 (0.3 –3.5)
85 and over Sex Female
2 147 493 (51.8)
35.363
16.47
429.452
199.98
51.0 (46.5 –56.7)
Male
1 997 922 (48.2)
26.263
13.14
386.506
193.45
49.0 (43.3 –53.5)
Ethnicity Asian European
349 439 (8.4)
3.137
8.98
43.748
125.19
3.1 (0.7 –19.7)
2 791 041 (67.3)
46.160
16.54
622.080
222.88
78.1 (18.2 –90.8)
Maori
571 439 (13.8)
7.477
13.09
90.387
158.17
8.4 (2.5 –35.6)
Pacific
296 573 (7.2)
3.564
12.02
39.113
131.88
1.6 (0.4 –15.4)
Other
10 727 (2.6)
0.928
8.61
15.937
147.94
1.3 (0.3 –5.7)
Deprivation quintile 1
834 639 (19.9)
9.327
11.17
151.638
181.68
16.4 (2.0 –37.0)
2
774 403 (18.5)
10.038
12.96
146.977
189.79
18.0 (5.4 –29.0)
3
748 726 (17.9)
11.152
14.89
151.578
202.45
18.0 (7.6 –25.7)
4
748 623 (17.9)
12.769
17.06
157.157
209.93
18.2 (5.8 –28.0)
5
757 487 (18.1)
13.567
17.91
151.235
199.65
12.1 (1.3 –46.6)
Therapeutic groupa Alimentary tract and metabolism
1 070 809 (25.8)
9.27
8.65
90.19
84.23
25.7 (20.0 –33.4)
Blood and blood-forming organs
764 398 (18.4)
7.21
9.44
96.53
126.29
18.6 (13.0 –25.5)
Cardiovascular
732 162 (17.7)
10.34
14.12
77.67
106.09
18.2 (12.0 –25.6)
1 023 346 (24.7)
2.74
2.68
18.82
18.39
24.4 (19.1 –29.8)
Genito-urinary
401 789 (9.7)
1.02
2.53
11.85
29.50
9.5 (6.4 –12.1)
Hormone preparations
487 378 (11.8)
1.89
3.89
21.08
43.25
11.3 (8.1 –16.0)
Dermatologicals
Infections Musculo-skeletal Nervous system
1 795 097 (43.3)
4.24
2.36
53.61
29.87
43.0 (34.3 –51.7)
969 126 (23.4)
2.91
3.00
71.36
73.63
23.0 (18.5 –27.6)
1 746 912 (42.1)
41.8 (33.8 –49.1)
15.79
9.04
167.04
95.62
49 011 (1.2)
0.44
9.01
116.98
2386.75
Respiratory system and allergies
971 816 (23.4)
4.69
4.83
80.36
82.70
22.6 (17.6 –28.8)
Sensory organs
423 875 (10.2)
1.09
2.57
10.46
24.67
10.1 (7.5 –12.8)
Oncology agents and immunosuppressants
Case study practices
At Practice A (SPR ¼ 1.31), a single patient was treated with one medicine that accounted for 20.4% of the practice’s total prescribing costs. To examine the influence of this single drug
1.2 (0.7 –1.7)
on the SPR, we repeated the standardization of prescribing costs excluding it. Following this, Practice A’s adjusted SPR was 1.05, but four more patients accounted for 20.5% of total prescription costs: one taking another medicine (11.8% of
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
0– 4
VA RI AT IO N I N T H E P H A R MACE U TI CA L CO STS O F N E W ZE A L A ND G E NE RA L P RAC TI C ES
total cost) and three patients using a third drug (8.7%). The patients taking the two most costly drugs accounted for 33.2% of the practice’s total prescription costs. At Practice B (SPR ¼ 2.28), 91 patients were treated with antiretrovirals costing $13 287 per patient. The median number of patients using antiretrovirals at all 1045 practices 70
50 40 30 20 10 0 ATM BLD CVD DRM GNU HRM INF MSK NRV OAI RSP SNS
Fig. 1 Practice variation in proportion of patients by therapeutic group in 2011. Box, inter-quartile range. 95% of practices lie within the whiskers.
was 0. Antiretrovirals contributed 45% of Practice B’s total medicines expenditure. Practice C (SPR ¼ 1.44) was one of 265 practices with an SPR . 1.0 but no single medicine or medicines class accounted for .15% of its total prescribing costs. Its cost per patient was higher than the national average in 9 out of 12 major therapeutic groups. Overall, its cost per patient was 69.9% higher than the national average. Practice D (SPR ¼ 0.53) had the lowest SPR of all practices. Prescribing costs per patient were 50.8% lower than the national average and the number of medicines per patient 20.9% lower. Costs were lower than expected in 11 of the 12 therapeutic groups. The total number of medicines dispensed was 14.9% lower than expected. At Practice E, the total number of medicines dispensed was 7.1% higher than expected (SPR ¼ 0.56), but prescription costs were 60.4% of expected. Prescription costs per patient ($21.24) were 71.9% lower than the national average ($75.55) and costs per item were 56.0% lower. Practices with lower than expected costs were characterized by low prescription costs per patient across most therapeutic groups and usually a low prescription rate per patient.
ATM, alimentary tract and metabolism; BLD, blood and blood-forming organs; CVD, cardiovascular system; DRM, dermatologicals; GNU,
Discussion
genito-urinary system; HRM, hormone preparations; INF, infections; MSK, musculo-skeletal system; NRV, nervous system; OAI, oncology agents and immunosuppressants; RSP, respiratory system and allergies; SNS, sensory organs.
Main finding of this study
Our findings indicate that there is substantial variation in pharmaceutical expenditure per patient between New Zealand general
2.5
Standardized prescribing ratio
2.0
1.5
1.0
0.5
0.0 0
1 000 000
2 000 000
3 000 000
4 000 000
5 000 000
Expected cost of prescribing ($) Fig. 2 Standardized prescribing ratios for 1045 New Zealand general practices by expected cost of prescribing 2011. Dots, practices. Dashed curves: 2 SD, solid curves; 3 SD, bold dots are ‘definite’ (.3 SD) or ‘possible’ outliers (2 –3 SD).
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
% of registered patients
60
5
6
J O U R NA L O F P U B L IC H E A LT H
Table 2 Influence of expensive medications on the total cost of prescribing in 2011 in practices with a standardized prescribing ratio (SPR) . 1.0 All New Zealand general practices
Practices with SPR . 1 and 15% of total cost in one TG2 or medicine
n ¼ 1045 $ per patient
n ¼ 204 No. of patients
Median (% of
No. of practices
total practice $) Antibacterials Colistin sulphomethate
Mean SPR
total practice $)
9.07
1 702 824
1.78
1
22.9
1.31
10 324.42
43
0.00
1
20.4
1.31
515.91
78 772
3.97
28
15.2 –41.8
1.24
1 479.57
10 798
1.60
3
15.2 –19.6
1.18
Risperidone
721.15
17 529
0.88
5
15.8 –32.9
1.25
Antiretrovirals
11 188.08
1305
0.00
15
16.3 –44.8
1.63
9018.50
5309
4.70
34
15.2 –34.3
1.15
Adalimumab
17 214.52
2264
3.62
20
15.1 –30.5
1.12
Etanercept
12 525.68
513
0.00
1
33.6
1.32
238.25
52 928
0.75
9
15.0 –36.1
1.17
6088.67
244
0.00
2
22.0 –24.2
1.06
Olanzapine
Antirheumatoid agents
Antivirals Adefovir dipivoxil Tenofovir disoproxil fumarate
5167.52
1286
0.00
1
1854.14
28 499
4.49
51
Bortezomib
18294.00
255
0.00
Dasatinib
46 186.64
59
Erlotinib hydrochloride
13 009.09
Imatinib mesylate Sunitinib Temozolomide
Chemotherapeutic agents
Endocrine therapy Octreotide Fluids and electrolytes Total parenteral nutrition Immune modulators Pegylated interferon alpha-2a Immunosuppressants
17.1
1.04
15.1 –46.3
1.17
1
15.7
1.15
0.00
2
15.1 –21.6
1.12
186
0.00
1
22.6
1.20
46 124.41
381
0.00
16
28 880.80
121
0.00
1
12 871.00
215
0.00
716.03
12 766
0.27
21 891.14
330
15.8 –35.8
1.20
21.9
1.21
1
16.7
1.32
5
15.9 –25.5
1.26
0.00
5
15.5 –25.4
1.26
38.98
104 228
0.00
3
16.2 –56.3
1.49
48 508.32
46
0.00
3
16.1 –56.2
1.49
10 093.84
751
0.00
3
17.0 –19.4
1.14
10 786.81
677
0.00
3
17.0 –19.4
1.14
5036.19
10 920
5.10
51
15.1 –55.6
1.19
Rituximab
16 970.49
844
0.00
2
21.1 –40.0
1.05
Trastuzumab
36 817.78
695
0.00
29
15.1 –47.7
1.22
134.14
480 549
7.06
21
15.2 –32.6
1.10
317.39
165 147
5.60
10
15.1 –27.6
1.10
81 405.00
16
0.00
5
18.3 –45.7
1.32
81 405.00
16
0.00
5
18.3 –45.7
1.32
6420.12
509
0.00
1
19.1
1.32
40 182.79
14
0.00
1
15.2
1.23
Lipid-modifying agents Atorvastatina Metabolic disorder agents Imiglucerase Vasodilators Iloprost a
Atorvastatin became available generically in mid-2011.
TG2, therapeutic group level 2; SPR, standardized prescribing ratio.
practices. This is true for both actual and standardized expenditure resulting from the adjustment of costs to account for practices’ differences in patient demographics and morbidity. Our results also indicate that practices’ prescribing costs may vary considerably over time, an important consideration when deciding which practices may benefit from interventions to alter prescribing.
What is already known
Among the factors associated with varying medicines use in general practice, patients’ age has most frequently been factored into analyses of variation,29 but age alone does not account for enough variation to set robust budgets for medicines expenditure.30 Models including age and sex account
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
Antipsychotics
Range (% of
VA RI AT IO N I N T H E P H A R MACE U TI CA L CO STS O F N E W ZE A L A ND G E NE RA L P RAC TI C ES
DHB A (24 practices)
DHB B (42)
DHB C (24)
7
DHB D (28)
1.5 1.4
1.2 1.1 1.0 0.9 0.8 0.7 0.6 0.5 0.4
Fig. 3 Variation in actual to expected pharmaceutical costs 2008 –11 for practices in four District Health Board areas. Markers indicate the practice’s mean SPR 2008 –11; lines indicate the SPR range.
for around 25% of variation,7,31,32 additional demographic factors up to 51%33 – 35 and models that include morbidity account for more variability (up to 80%).36 – 38 The use of medicines within therapeutic groups provides an indicator of patient morbidity and treatment needs within a practice and has been included in previous models of prescribing activity in general practice.30,39 Variation in practices’ mortality rates, but not prescribing costs, has been investigated by standardizing practices to account for their patients’ morbidity.40 Doctor characteristics and morbidity profiles in regional databases have been used to explore drivers of variation in general practice prescribing,41 and a network of computerized general practice prescribing has shown longitudinal changes in prescribing behaviour.36 What this study adds
We could find no previous investigations of variation in medicines costs per practice involving national data sets. This study builds on previous research showing the importance of including patient morbidity in models of practice variation, adding to previous models by including patients’ ethnicity and deprivation indicator, both associated with health care inequalities in New Zealand. We used all these variables, comprehensively recorded in New Zealand’s national health care data sets, to identify the standardized distribution of New Zealand general practices’ pharmaceuticals expenditure. We showed that analysis of routine national data collections provides a starting point for identifying practices with
potentially modifiable prescribing costs through interventions such as academic detailing.42 This extends previous work by exploring the practical implications of identifying outlier general practices. We confirmed that high-cost outlier practices are not necessarily generating modifiable pharmaceutical expenditure. Practices with more morbid patients, especially in chronic conditions, are more likely to have higher prescription costs. High costs per capita in smaller practices were often due to individual patients using very high cost but necessary medicines. Unusually large numbers of patients requiring expensive treatments also contribute to higher costs than expected. Practices with consistently higher than expected costs across most patient demographic and therapeutic drug groups may be more likely candidates for interventions aimed at reducing pharmaceutical expenditure. Limitations of this study
Our analysis included all pharmaceuticals dispensed to all patients registered to each practice as recorded in the national Pharmaceutical Collection. It is a limitation of this study that many of these medicines may not have been prescribed by doctors at practices registering patients. Some practices, particularly those in popular holiday areas, provide care for many casual patients whose prescription costs are attributed to practices they are enrolled in rather than the prescribing practices. Therefore, this analysis does not precisely quantify the total prescription costs generated by each practice.
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
Standardized prescribing ratio
1.3
8
J O U R NA L O F P U B L IC H E A LT H
We were able to analyze variation in prescription costs from 2008 to 2011 for only 917 of the 1045 practices. This limitation reflects the changing nature of general practices as businesses that may close, merge, transfer patients or move location. We included merged practices that had kept the same name throughout the study period. In many cases, mergers would have resulted in changes in the demographic profile of such practices, influencing the relative prescription ratios across the 4-year period.
This study indicates that there is considerable variation in prescription costs per patient among New Zealand general practices after adjusting for their patients’ demographic and morbidity profiles. The methodology identified outlier practices with higher expenditure than expected that may be targeted in interventions to reduce prescribing costs. While indirect standardization should not be used in creating practice league tables as bias is introduced when patient denominators differ substantially, it may be used to indicate how far practices are from the expected cost of prescribing.43 In our analysis, the expected cost was determined by the national average prescribing cost rates observed within patient demographic and major therapeutic drug groups. Each practice was thus compared with the reference population representing all registered general practice patients in New Zealand. General practices may have higher than expected prescription costs, even after adjusting for demographic and morbidity factors, for a variety of reasons. Patients with special needs may require expensive pharmaceutical treatment, medicines with little or no therapeutic benefit may be inappropriately prescribed, or general practitioners may view their role squarely in terms of service provision, with prescribing as part of that service. Low-cost prescribers, by contrast, are more comfortable with a wider view of their role in society that may require limiting patients’ pharmaceuticals use.44 Also it is possible that lower cost practices may be providing poorer quality care, particularly in the identification and management of patients with chronic conditions. Interventions aimed at controlling prescribing costs need to be informed by the reasons that practices have outlier costs as well as by robust evidence of the expenditures they generate.
This research was conducted as a component of the usual employment of the authors by Bpacnz and the University of Otago. Bpacnz is a registered charity operating under contracts with PHARMAC and District Health Boards Shared Services (DHBSS). Bpacnz is co-owned by the University of Otago, ProCare Health Limited, South Link Health, General Practice New Zealand and Pegasus Health.
References 1 Toop L. Primary care core values: patient centred primary care. Br Med J 1998;316:1882 – 3. 2 Green LA, Fryer GE, Yawn BP et al. The ecology of medical care revisited. N Eng J Med 2001;344:2021 – 5. 3 Tordoff J, Norris P, Reith D. “Price Management” and its impact on hospital pharmaceutical expenditure and the availability of medicines in New Zealand hospitals. Value Health 2008;11(7):1214 – 26. 4 Moon J, Bogle R. How to lose a billion pounds. Int J Clin Pract 2007;61(1):2 – 3. 5 Minhas R. Waste not, want not: free money, moral hazard and valuebased prescribing. Int J Clin Pract 2011;65(3):240 – 2. 6 Sermet C, Andrieu V, Godman B et al. Ongoing pharmaceutical reforms in France: implications for key stakeholder groups. Appl Health Econ Health Policy 2010;8(1):7 – 24. 7 Favato G, Mariani P, Mills R et al. ASSET (Age/Sex Standardised Estimates of Treatment): a research model to improve the governance of prescribing funds in Italy. PLoS ONE 2007;2(7):e592. 8 Hassell K, Atella V, Schafeheutle E et al. Cost to the patient or cost to the healthcare system? Which one matters the most for GP prescribing decisions? A UK-Italy comparison. Eur J Public Health 2003;13:18–23. 9 Carthy P, Harvey I, Brawn R et al. A study of factors associated with cost and variation in prescribing among GPs. Fam Pract 2000;17(1):36–41. 10 Zhang Y, Soumerai S. Do newer prescription drugs pay for themselves? A reassessment of the evidence. Health Aff 2007;26(3):880– 6. 11 Fleetcroft R, Cookson R, Steel N et al. Correlation between prescribing quality and pharmaceutical costs in English primary care: national cross-sectional analysis. Br J Gen Pract 2011;61(590):e556– 64. 12 Fleetcroft R, Schofield P, Duerden M et al. Achievement of cholesterol targets and prescribing of higher-cost statins: a cross-sectional study in general practice. Br J Gen Pract 2012;62(605):815– 20. 13 Allan G, Lexchin J, Wiebe N. Physician awareness of drug cost: a systematic review. PLoS Med 2007;4(9):1486 – 96. 14 Andersson K, Carlsten A, Hedenrud T. Prescribing behaviour after the introduction of decentralized drug budgets: is there an association with employer and type of care facility? Scand J Prim Health Care 2009;27:117–22.
Acknowledgements
15 Buusman A, Morten A, Merrild C et al. Factors influencing GPs’ choice between drugs in a therapeutic drug group. A Qualitative Study. Scand J Prim Health Care 2007;25(4):208 – 13.
We thank PHARMAC for providing permission to use the Pharmaceutical Collection.
16 Mason A, Drummond M, Towse A et al. Influencing prescribing in English primary care: the views of primary care organisations. J Health Serv Res Policy 2004;9(3):153 – 8.
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
Conclusions
Funding
VA RI AT IO N I N T H E P H A R MACE U TI CA L CO STS O F N E W ZE A L A ND G E NE RA L P RAC TI C ES
9
31 Majeed A, Head S. Controversies in primary care. Setting prescribing budgets in general practice. Capitation based prescribing budgets will not work. Br Med 1998;316(7133):748– 50.
18 Reichert S, Simon T, Halm E. Physicians’ attitudes about prescribing and knowledge of the costs of common medications. Arch Intern Med 2000;160(18):2799– 803.
32 Greenhalgh T. Controversies in primary care. Setting prescribing budgets in general practice. Effective prescribing at practice level should be identified and rewarded. Br Med 1998;316(7133):750– 3.
19 Weekes L, Mackson J, Fitzgerald M et al. National prescribing service: creating an implementation arm for national medicines policy. Br J Clin Pharmacol 2005;59(1):112– 6.
33 Whynes D, Baines D, Tolley K. Explaining variations in general practice prescribing costs per ASTRO-PU (age,sex, and temporary resident originated prescribing unit). Br Med 1996;312:488 – 9.
20 Denig P, Haaijer-Ruskamp F, Wesseling H et al. Towards understanding treatment preferences of hospital physicians. Soc Sci Med 1993; 36(7):915 – 24.
34 Forster D, Frost C. Use of regression analysis to explain the variation in prescribing rates and costs between family practitioner committees. Br J Gen Pract 1991;41(343):67– 71.
21 Watkins C, Harvey I, Carthy P et al. Attitudes and behaviour of general practitioners and their prescribing costs: a national cross sectional survey. Qual Saf Health Care 2003;12:29– 34.
35 Haynes K, Bilker WB, TenHave TR et al. Temporal and within practice variability in the health improvement network. Pharmacoepidem Drug Safe 2011;20:948 – 55.
22 Statistics New Zealand. National Population Estimates: December 2010 Quarter. 2012. http://www.stats.govt.nz/browse_for_stats/population/ estimates_and_projections/ NationalPopulationEstimates_HOTPDec10qtr.aspx (28 January 2013, date last accessed).
36 Omar R, O’Sullivan C, Petersen I et al. A model based on age, sex, and morbidity to explain variation in UK general practice prescribing: cohort study. Br Med 2008;337:a238.
23 Ministry of Health. Visiting a Doctor. 2013. http://www.health.govt.nz/ yourhealth-topics/health-care-services/visiting-doctor (28 January 2013, date last accessed). 24 Salmond C, Crampton P, Atkinson J. NZDep2006 Index of Deprivation. 2007. 25 Pharmaceutical Management Agency (PHARMAC). New Zealand Pharmaceutical Schedule: December 2012. PHARMAC, 2012. www. pharmac.govt.nz/2012/11/27/Sched.pdf (28 January 2013, date last accessed).
37 Rice N, Dixon P, Lloyd D et al. Derivation of a needs based capitation formula for allocating prescribing budgets to health authorities and primary care groups in England: regression analysis. Br Med 2000;320 (7230):284 – 8. 38 Morton-Jones T, Pringle M. Explaining variations in prescribing costs across England. BMJ 1993;306(1731):4. 39 Lloyd D, Harris C, Roberts D. Specific therapeutic group age-sex related prescribing units (STAR-PUs): weightings for analysing general practices’ prescribing in England. BMJ 1995;311(7011): 991 – 4.
26 Kirkwood B, Sterne J. Essential Medical Statistics, 2nd edn. Oxford: Blackwell Science, 2003.
40 Carey IM, Shah SM, Harris T et al. A new simple primary care morbidity score predicted mortality and better explains between practice variations than the Charlson index. J Clin Epidemiol 2013;66(4):436–44.
27 Morton A, Mengersen K, Rajmokan M et al. Funnel plots and risk-adjusted count data adverse events. A limitation of indirect standardisation. J Hosp Infect 2011;78(4):260– 3.
41 Orzella L, Chini F, Rossi PG et al. Physician and patient characteristics associated with prescriptions and costs of drugs in the Lazio region of Italy. Health Policy 2010;95(2 – 3):236 – 44.
28 Spiegelhalter D. Funnel plots for institutional comparison. Qual Saf Health Care 2002;11(4):391– 1.
42 Rognstad S, Brekke M, Fetveit A et al. Prescription peer academic detailing to reduce inappropriate prescribing for older patients: a cluster randomised controlled trial. Br J Gen Pract 2013;63(613): E554 – 62.
29 Purves I, Edwards C. Comparison of prescribing unit with index including both age and sex in assessing general practice prescribing costs. Br Med J 1993;306:496 – 8.
43 Rixon A. Performance league tables. BMJ 2002;325:177– 8.
30 Sleator D. Towards accurate prescribing analysis in general practice: accounting for the effects of practice demography. Br J Gen Pract 1993;43(368):102– 6.
44 Jaye C, Tilyard M. A qualitative comparative investigation of variation in general practitioners’ prescribing patterns. Br J Gen Pract 2002;52: 381 – 6.
Downloaded from http://jpubhealth.oxfordjournals.org/ at University of Wisconsin-Oshkosh on November 16, 2015
17 Prosser H, Walley T. A qualitative study of GPs’ and PCO stakeholders’ views on the importance and influence of cost on prescribing. Soc Sci Med 2005;60(6):1335 – 46.
