Dispensing patterns and financial costs of drug utilisation
in Tayside, Scotland, with particular reference to Type 1 and Type 2 diabetes
(1995)
Craig J. Currie, PhD
Senior Research Fellow
Cardiff Research Consortium
Chapter 1. Gastrointestinal system
Chapter 2.
Cardiovascular system
Chapter 4.
Central nervous system
Chapter 7.
Obstetrics, gynaecology, and urinary tract disorders
Chapter 8.
Malignant disease and immunosupression
Chapter 9.
Nutrition and blood
Chapter 10.
Muskuloskeletal and joint disorders
Chapter 14.
Immunological products and vaccines
Abstract
Drug utilisation studies provide important economic intelligence, and in some circumstances they can also provide epidemiological intelligence. The purpose of this study was to investigate drug utilisation and the related financial costs for the population of Tayside, Scotland, for the year 1995. In particular, to focus on the population of people with types 1 and 2 diabetes. The study design was a retrospective, cross-sectional, descriptive study. The study combined data for all drug consumption in Tayside along with diabetes and population registers using the Scottish community health index number to link records relating to the same individual. There were 3.4 million prescriptions dispensed to a population of 406,000 people. The prevalence of identified diabetes was estimated to be 2.13%. 6.1% of prescriptions were dispensed to people with diabetes, and they had a 2-fold age and sex standardised probability of being dispensed a prescription. Excluding drugs used in treating diabetes, this value was reduced to 1.7-fold. The total cost of these drugs was estimated to be £7.3 million per 100,00 population at 1995/96 prices. People with diabetes were thus responsible for 8% of all drug expenditure. The estimated cost of drug use in those with diabetes in the UK NHS in 1995 was £225 million at 1995/96 prices. The paper provides detailed population, consumption and cost data for all drug use in nondiabetic, type 1 and type 2 diabetes by British National Formulary sub-chapter. People with diabetes have increased drug use in almost all categories. The pattern of increased drug use reflects known differences in disease epidemiology between those with type 1 diabetes, type 2 diabetes, and those people without diabetes.
Introduction
Not only do the majority of people with diabetes require lifelong drug therapy, they have increased incidence of many other diseases that can be managed by drug therapy other than insulin or oral hypoglycaemic agents. One example would be events caused by heart disease, where people with diabetes are known to respond favorably to primary and secondary prevention.[1] [2] Because of these and other factors, there is likely to be an increase in drug utilisation in this group. Increased resource use in those with diabetes is evident in other areas of the UK National Health Service (NHS). For example, in secondary care.[3] Even though it is hypothesised that there should be an increase in drug use in this group of patients, there has been no detailed drug utilisation study discussing drug use in those with diabetes in the UK.
In addition to supplying important economic intelligence, drug utilisation studies offer a mechanism to infer relationships with other diseases. They also sometimes allow for inference to be made about disease incidence and prevalence. Because these studies give a view of the total spectrum of drug use, they also offer a mechanism for hypotheses generation. Drug use can be a good epidemiological indicator. For example, suspicion about the advent of a new disease striking young homosexual men in the US (AIDS) was discovered via the monitoring of the use of pentamidine by the Center for Disease Control, Atlanta.[4] There are thought to be three conditions necessary for a drug to offer reliable epidemiological intelligence: specificity of treatment, quality data/information, and the facility to study data at regular intervals.[5]
The Prescription Pricing Service in Newcastle, the Prescription Pricing and Information Service in Cardiff, and the Pharmacy Practice Division in Edinburgh collectively process all prescriptions in Great Britain. The information that they record from the prescription is used for the remuneration of the agencies that dispensed the prescription. They do not include the name and address of the individual person for whom the prescription is intended in their electronic record: there is no incentive. This does however limit the ability of others to link these data to other sources of health data that describe the health experiences within the UK population.
The Medicines Monitoring Unit (MEMO), Dundee University, is a research organisation that is allowed secondary access to prescriptions for the defined geographical area of Tayside, Scotland. Data from these prescriptions are then computerised. These data include a unique identification number, the Scottish Community Health Index Number (CHINo). This allows reliable record linkage of drug dispensing data. Other sources of data would typically include hospital activity data, biochemistry results, and ad hoc disease registers. A disease register existed that identified the population in Tayside with identified diabetes. This has been given the acronym DARTS (the Diabetes Audit and Research in Tayside Scotland study). DARTS uses the same CHINo and allowed dispensing data to be linked to the diabetes register. This offered a unique facility to describe drug utilisation in this group, with the added bonus of being able accurately differentiate between people with types 1 and 2 diabetes. Both MEMO and DARTS have had their validity scrutinised by the publication of previous scientific work.[6] [7]
A number of drug utilisation studies for those with diabetes have been described. These studies have been detailed by Papoz.[8] Papoz referred to only one paper by Robinson and colleagues that discussed trends in utilisation of insulin and oral anti-diabetic agents in the UK.[9] This source is now dated, and drug utilisation is only a small component of the study.
The objectives of the present study were to determine the pattern of drug utilisation for all dispensed drugs in the population of people in Tayside with and without diabetes. To then infer and compare patterns of disease experienced in the two populations by examining and comparing the classification of the drugs used. Finally, to estimate the relative costs of treatment in the two groups.
Methods
The study design was a retrospective, cross-sectional, descriptive study.
The DARTS study is described in detail elsewhere.[10] In brief, eight independent data sources (diabetes clinics, biochemistry results, a mobile eye screening facility, and various hospital activity databases) have been used to identify all people in the region with a diagnosis of diabetes. The sensitivity of this method was estimated to be 0.96, and the positive predictive value 0.96.
Subjects
The study population was the population of Tayside Health Board, Scotland as defined by inclusion on the health service register at the end of 1995. This population excludes a small number of outward migrants over the study period. The total population was 406,526 people. The population of Tayside is typical of the UK as a whole, as it was largely white (98.8%[11] vs. 94.5%[12]). Those in the population with diabetes were identified via the DARTS diabetes register. Type 1 diabetes was defined as people less than 35 years of age at diagnosis and a requirement for treatment with insulin. A small number of type 1 patients may have only been recently diagnosed or diagnosis may have occurred shortly after the study period but they may have been included because of the retrospective method. It is also accepted that some patients may be non-compliant, even with essential medicines such as insulin.[13]
Prescribing Data
Drug utilisation data were all dispensed prescriptions for the geographical area of Tayside, Scotland for the year 1995. The drug coding system used for the purposes of this study was the British National Formulary (BNF).[14] The BNF provided a useful taxonomy generally based on the physiological site of drug action.
These data were analysed to the second level of disaggregation in the BNF taxonomy. For example, “drugs used in diabetes” (BNF-6.1), a sub-category of the “endocrine system” (BNF-chapter 6).
A technical problem existed at the time of investigation in the MEMO database that resulted in the exclusion of a small number of newly marketed drugs (1.8% of dispensed prescriptions).
Financial data
Scottish Prescribing and Audit data (SPA data) were supplied by the Tayside Health Board for the financial year 1995/96. These data were dissaggregated by BNF sub-chapter. Estimates of cost for the population of people with diabetes were made by calculating the product of the estimated unit cost per prescription in each BNF sub-category, and the number of prescriptions dispensed to each population sub-group. Cost values are given at 1995/96 financial year prices. Detailed unit costs for insulin and oral anti-diabetic agents were taken from a published source from the same study period.[15] These unit costs were allocated by type of diabetes rather than a specific reference to the nature of the diabetic drug used. A small proportion of drug expenditure was not included on the BNF taxonomy. Exclusions included dressings, appliances, and prescriptions for incontinence and stoma care (total annual expenditure £1.5 million).
Statistical considerations
Diabetes is recognised to remain undiagnosed in a notable proportion of the population.[16] Perhaps as many as 50% of people who would come under this diagnostic classification if any area had a stringent screening program to identify those with diabetes. The result of this is uncertainty about the denominator in many calculations, particularly rate and risk calculations. Placing confidence intervals around risk estimates is spurious given this uncertainty.[17] Ascertainment in an area such as Tayside is likely to decrease the proportion of undiagnosed diabetes because increased awareness about the disease and greater effort to identify those with diabetes. This is a lesser problem in type 1 diabetes than type 2 diabetes. For this reason, standardised risk estimates lower than 1.3 (as an educated estimate) should therefore be treated with caution as epidemiological indicators. Data were standardised by applying rates in the non-diabetic population in each specific population sub-group to the number of diagnosed people with diabetes in the same group. The ratio between the sum of the observed and expected values was then calculated.
Results
Population analysis
There were 406,526 people registered with a Tayside general practitioner at the end of 1995 (table 1). The prevalence of diabetes was 2.13% (0.2% (990) type, 1 and 1.9% (7664) type 2). 292,811 (72%) of the total population had had a drug prescription dispensed. This proportion was increased in diabetes where 94% of people with diabetes dispensed at least one prescription. Overall, 80% of females dispensed at least one prescription over the study period by comparison with 62% of males. The difference between males and females is striking (figure 1 and table 2).
Drug utilisation analysis
Over 3.41 million prescriptions were dispensed over the year (table 3). Excluding drugs used for the treatment of diabetes, 6.1% of prescriptions were dispensed to people with diabetes (5.5% of prescriptions to males, and 7.2% to females). Including diabetes drugs, these values increased to 7.8%, 9.4%, and 6.8% for all prescriptions, prescriptions to males, and prescriptions to females respectively.
The distribution of the proportion of prescriptions dispensed by age, sex, and type of diabetes is shown in figure 2. The comparator in this illustration was all prescriptions rather than prescriptions to the nondiabetic population.
Table 2
Population data: the number of people in the community who were dispensed a prescription in each BNF category (categories with a total number of prescriptions less than 2000 over the year are not shown).
Types 1 and 2 diabetes
produce differing patterns of prescribing that closely resembled their
differing age-related prevalence. The
age and sex distribution showed a direct relationship between age and the
proportion of prescriptions dispensed.
The proportion of prescriptions dispensed to males older than 75 years
was lower than the 65-75 year-old group.
This reflected reduced life expectancy in men.
The proportion of prescriptions dispensed to those with diabetes varied widely by BNF category, for example, 10.5% of cardiovascular drugs (BNF chapter 2.), and 2.2% of prescriptions for obstetrics and gynaecology (Chapter 7).
Figure 3 illustrates the annual mean rate of prescribing by age and sex in the type 1, type 2
groups and for the whole population. This showed the consistent increase in prescribing to females
over males in all groups. It also shows
the difference in prescribing rates between those with type 1 diabetes, type 2
diabetes, and the population as a whole (this is a good proxy for the nondiabetic
population). Representing these data
in another way, standardising for age and sex, people with diabetes had a
2-fold increased probability of being dispensed at least one prescription
(table 4). The pattern of relative
probability of being dispensed a prescription showed wide variation by drug
category, type of diabetes and sex.
Excluding drugs used for treating diabetes, this value decreased to
1.7-fold. In the type 2 diabetes group
the summary value of 1.7-fold was the same for both males and females. However, in the type 1 group, males and
females showed differing patterns of dispensing, whereby males had a 1.9-fold
increase and females had a 2.4-fold increase.
Table 3
Frequency of prescribing: the number of prescriptions dispensed in each BNF category (categories with a total number of prescriptions less than 2000 over the year are not shown).
Table 4 lists these data in detail for each drug category. The probability of being diabetic and receiving a prescription was increased in almost every category. Although exceptions were few, to give one example, type 2 diabetic males have the same probability as nondiabetic males of receiving prescriptions that treated respiratory problems (BNF chapter 3).
Table 4
Relative rate of prescribing in each BNF category (categories with a total number of prescriptions less than 2000 over the year are not shown.
(As a rule of thumb given the
differences in the size of the two populations, any BNF category with a
standardised relative risk >1.3 and/or with >2,000
dispensed scripts per year exhibits a difference in relative risk that is
unlikely to be due to chance (5% level)).
Financial Costs
The study population was estimated to consume £29.7 million over the study period (table 5). This equated to £7.3 million per 100,000 people per year (nondiabetic people £6.8 million/100,000). Of this total, people with diabetes were estimated to consume £2.3 million (7.8% (1.4% type 1, 6.4% type 2)). This expenditure equated to £31.2 million per 100,000 in the diabetic population, a 4.6-fold crude relative increase in cost between the two groups. More specifically, £41.8 million/100,000 in type 1 patients, and £24.8 million/100,000 in type 2 patients. By comparison with the nondiabetic population, the crude relative cost was therefore increased six-fold in type 1 patents and a 3.6-fold in type 2 patients.
Insulin and oral anti-diabetic drugs accounted for 1.7% of expenditure in the total drug budget (BNF chapter 6.01). Only one-third of expenditure in this category was for the treatment of patients with type 2 diabetes. Anti-diabetic agents (insulin and OHAs) accounted for 77% of drug expenditure in the type 1 group, whereas they accounted for only 8.8% of drug expenditure in the type 2 group. Drugs used for cardiovascular disease accounted for the greatest proportion of expenditure in the diabetic group (29% of expenditure). This compared to 24% in the nondiabetic group. Ulcer-healing drugs were responsible for almost twice as much expenditure as were anti-diabetic agents in those with type 2 diabetes (16.6% vs. 8.8%). Cardiovascular and gastrointestinal drugs collectively accounted for more than half of all drug expenditure in type 2 diabetes (54% vs. 41% in nondiabetes).
Table 5
Financial cost of prescribing in each BNF category.
Discussion
Studies of drug utilisation offer inference about disease relationships and epidemiology, and important intelligence about resource use.
One important factor in interpreting these results is that people with diabetes receive free prescribed medicines in the UK. There was therefore no financial disincentive to collecting prescriptions in those with diabetes. Another factor that may influence increased prescribing is the increased access to their general practitioner as a result of their chronic disease state. There is also reason to hypothesise that doctors may have a lower threshold for prescribing some drugs to people with diabetes.
The data presented above are detailed. In order to give a structure to the discussion with respect to the findings, each BNF chapter will be discussed in turn.
Chapter 1. Gastrointestinal system
There was
an average 1.3-fold increase in prescribing in type 2 diabetes and a 2.0-fold
increase in type 1 diabetes. The increased use of drugs for intestinal
secretions – although a small number of prescriptions related to a small number
of people – may be related to diabetes secondary to chronic pancreatitis who
take Pancreatin and other related drugs.
Anti-diarrhoeal drugs showed the greatest increase in prescribing in
this group. It may be that there is an
IgA mediated immune paresis, especially if there is poor glycaemic control predisposing
to secondary infection and infective diarrhoea, as a consequence of an
immunological reaction. There may also
be a suggestion that these patients have increased risk of colitis. Alternatively it may simply represent
autonomic neuropathy.
Chapter
2. Cardiovascular system
There was an expectation that use of this class of drug would be raised because of the increased risk of heart disease in those with diabetes. Nevertheless, some of the detail may be of more interest, for example, anti-hypertensive agents. The general belief is that type 1 patients do not have an increased risk of hypertension unless they develop nephropathy, whilst there is increased risk of hypertension in type 2 patients. It was unlikely that large numbers of patients with nephropathy skewed these results (risk in type 1 was 9.8 vs. 3.3. in type 2 diabetes). It was likely that large numbers of patients with microalbuminurea and normal blood pressure were being treated prophylacticly with ACE-inhibitors in the type 1 group. The rate of taking therapy for hyperlipidaemia was again greater in type 1 diabetes. This may reflect the possibility that type 1 patients were screened more effectively and treated more aggressively than were type 2 patients. It was perhaps not surprising that diuretic prescribing was increased in types 1 and 2 diabetes. These drugs are indicated for hypertension, renal impairment, cardiac failure, and they are also used symptomatically for the relief of ankle edema. That ischaemic heart disease was increased in diabetic patients was demonstrated by the increased use of nitrates and calcium channel blocking drugs. Support for this hypothesis was also evident with increase prescribing for anti-platelet drugs.
It could be argued that it is surprising that the rate of prescribing in this category was not higher.
Chapter
3. Respiratory system
There was
no evidence of a relationship between diabetes and respiratory illness. In this respect this category should not
have had a notably higher risk, even taking into consideration factors such as
free prescriptions. The apparent risk
values of 1.2 in type 2 diabetes and 1.1 in type 1 diabetes seem to reflect
this hypothesis.
Chapter
4. Central nervous system
There was
some evidence that diabetes was associated with depression and possible
suicide/parasuicide (rather like other chronic illness).[18] [19] This may be reflected in the increased risk
of being prescribed drugs in this category.
Anti-depressant drugs are also used widely in the treatment of neuropathy. This may demonstrate that these findings
were robust in supporting previous evidence for these relationships. There was an increased rate of prescribing
for analgesia – more marked in the type1 than in type 2 diabetes. Although we believe that this observation
has never been described before, it may be possible that this simply reflects
free prescriptions drugs. It may also
be that they have more pain or they may simply visit the GP more frequently. This observation requires further investigation.
Chapter
5. Infections
There were
increased prescribing in anti-bacterial, anti-fungal and anti-protazoal drugs
in type 1 and 2 diabetes. Diabetic
people are known to be more prone to bacterial infections[20], in
addition, doctors may have a lower threshold for prescribing antibiotics to
these patients. The increased use of
anti-fungal agents was expected, with an increased association with fungal
infections – especially skin infections - in people with diabetes.[21] Ulceration, gangrene, and other foot
problems also lead to increased use of these agents. Increased use of the anaerobic agents (metronidazole) may reflect
an increase in anaerobic bacterial foot ulcers.[22]
Chapter
6. Endocrine system
The
increased risk of thyroid drugs is well recognised as patients with type 1
diabetes have a much higher risk of hypo and hyperthyroidism.[23]
Chapter
7. Obstetrics, gynaecology, and urinary
tract disorders
The most notable observation in this group was that females with diabetes had a much increased use of drugs for vaginal and vulval conditions. Vaginal candidiasis is strongly associated with diabetes.[24] [25]
Chapter
8. Malignant disease and
immunosupression
There has
been previous inference about a relationship between diabetes and
haematological diseases. There is a possibility that patients treated with
immunosuppressants are also treated with high dose steroids, which may expose
undiagnosed diabetes. The other
possibility is that diabetes, an autoimmune condition, may be associated with
rheumatoid arthritis and other diseases where immune suppression is
indicated.
Chapter
9. Nutrition and blood
This
possible relationship between diabetes and blood disorders may be evident here,
particularly in type 1 diabetes, and may be due to use of high dose steroids
that can cause diabetes.[26]
[27] [28] [29] [30] [31] [32] [33] [34]
Chapter
10. Muskuloskeletal and joint disorders
There was
an increased risk in this group for both type 1 and type 2 diabetes (1.5 in
both groups). This may result from
rheumatoid disease once again being an immunity disorder. Peripheral neuropathy may also cause soft
tissue damage. This may also describe
the same effect that is evident in the analgesics. People with diabetes suffer renal impairment that may in turn
cause an increase in the incidence of gout.
In addition, this observation may be associated with an increase in
diuretic use in this sub-population.
Chapter
11. Eye
The
interesting observation in this class of drugs is the increased risk in
preparations for glaucoma.
Chapter
13. Skin
An increase
in this group would be predicted because of the effects of peripheral
neuropathy and its related problems such as infection. This was reflected in the high-risk
(1.9). Topical corticosteriods (BNF
6.04) are the largest sub-group in this class, and have risk of 1.7 in type 2
diabetes and 1.6 in type 1 diabetes.
Chapter
14. Immunological products and vaccines
Vaccines
and antisera are used more frequently in type 1 diabetes (risk was 2.5). This may be due to increased awareness and
prophylactic treatment such as influenza vaccination.
Cost data
People were diabetes are found to consume almost 8% of the prescribing budget. This equated to expenditure for the NHS in England of £173 million in 1995/6 (7.8% of £2.22 billion).[35] The value of £2.22 billion is likely to have accounted for around 80% of UK NHS drug expenditure.[36] A best estimate for general practice prescribing for diabetes in 1995 would therefore be £225 million at 1995/96 prices.
Summary
Many of these drug utilisation observations were unsurprising, and support the validity of these data. There are however a few exceptions, notably, the high use of analgesics which may warrant further investigation. Overall, type 1 diabetics had a two-fold, and people with type 2 almost a 2-fold increase in use of all drugs compared to the nondiabetic population (excluding drugs used for the management of diabetes). Drugs used in the care of those with diabetes account for a large proportion of drug expenditure, the proportion being very similar to the proportion of cost for acute care.3
Acknowledgments
I should like to thanks investigators in Tayside for allowing me access to these data. In particular, Professor Tom MacDonald and Drs. Andrew Morris, Graham Leese, and John Peters (Cardiff) for helping to clinically interpret these data.
I thank Dr Mark Mark McGilchrist, Mr. Gary White and Doug Boyle for their technical expertise in developing the MEMO/DARTS information systems. Additionally, Richie McAlpine, diabetes research nurse for his efforts in validating these data.
Craig Currie was partly funded by a project grant from the SmithKline Beecham pharmaceutical company. I thank particular Dr. Eva Lydick, Worldwide Director of Epidemiology, Smithkline Beecham Philadelphia, USA.